Folder Library now uses the standard data-fn/data-v3-play markup so the song_preview plugin handles hover-previews, the same way it does for grid and list views. Removes ~200 lines of preview-specific code: dedicated audio element, indicator rendering, toolbar toggle, and backend manifest checking. The preview experience is now unified across all library surfaces.
Signed-off-by: Kyle <kyle.j.t@live.co.uk>
Expose a has_preview boolean from pack manifests and use it to gate hover previews. backend: plugins/folder_library/routes.py now (optionally) reads sloppak.load_manifest to set m['has_preview'] (guarded so plugin still loads without sloppak). frontend: plugins/folder_library/screen.js skips preview requests when song.has_preview is false and removes the HEAD-probe + previewMissing cache. docs: plugins/folder_library/CLAUDE.md updated to document has_preview and the preview behavior. This prevents unnecessary HEAD/audio 404s and console noise.
Signed-off-by: Kyle <kyle.j.t@live.co.uk>
Remove audio member resolution from folder_library and delegate hover-preview audio to the canonical song_preview plugin endpoint. This separates concerns: folder_library now only builds the preview URL with the filename, while song_preview handles manifest resolution (preview: key, stem fallback) and Range support.
Changes:
- Remove _audio_member() resolver and audio_member field from pack metadata
- Update _previewUrl() to point to /api/plugins/song_preview/audio?file=<filename>
- Add _previewMissing cache to avoid re-requesting packs with no preview
- Add HEAD probe to check preview availability before playing
- Add test coverage for _previewUrl with special character encoding
Signed-off-by: Kyle <kyle.j.t@live.co.uk>
Make the hover-preview feature opt-out instead of opt-in (defaults to on). Increase dwell delay from 500ms to 800ms to prevent accidental triggers while clicking/dragging. Replace 4-bar equalizer indicator with 9-bar waveform with staggered animation and fade-in entrance. Add guards to prevent preview during drag operations. Update button styling and all user-facing docs to reflect new defaults.
Signed-off-by: Kyle <kyle.j.t@live.co.uk>
Move audio member selection logic from frontend to backend. The frontend was probing multiple candidate audio files with error/retry logic, often resulting in 404s. Now the backend resolves the best available audio file (preview.ogg, stems/full.ogg, or any stem) during metadata extraction and includes it in song metadata as `audio_member`. The frontend makes a single, guaranteed request instead of multiple probes. Simplifies the preview flow and improves reliability.
Signed-off-by: Kyle <kyle.j.t@live.co.uk>
Implement optional audio preview when hovering over songs in the Folders view. Features include:
- Toolbar toggle button (off by default) to enable/disable hover preview
- ~0.5s dwell delay to avoid accidental audio playback while scrolling
- Dedicated <audio> element (never touches main player)
- Equalizer indicator animation overlay on song art while playing
- Per-surface localStorage persistence
- Automatic fallback through preview.ogg → stems/full.ogg → stems/audio.mp3
Bump Folder Library from 1.8.0 to 1.9.0. Remove 'auto play on hover' from roadmap as it's now implemented.
Signed-off-by: Kyle <kyle.j.t@live.co.uk>
Rebased onto merged main (was stacked on #1023/#1024, whose venue work is
now in main) so it no longer carries a stale content_packs.py that would
revert 1023's build_pack fixes.
- build_vst_pack: slice a fat .vst3 tree to one platform (keep its binary
dir + shared bundle files, drop the two foreign platform dirs and src/
build trees). Pins create_system=3 like build_pack — without it the same
tree hashes differently on a Windows runner (native .vst3 are built there),
breaking the precomputable-hash guarantee exactly where it matters.
- Publish wiring: 'python tools/content_packs.py <vst-root> --vst --version N
--publish' builds+uploads vst-<plat>-vN releases for mac/win/linux and emits
a platform-keyed {url,sha256,bytes} manifest — the shape rig_builder's
data/vst_packs.json consumes. publish() refactored onto a shared
_publish_release helper (venue behaviour unchanged).
- Tests: slice keeps target+shared/drops foreign, per-platform binary,
reproducibility, unknown-platform reject, and a simulated-win32 guard that
fails if the create_system pin is dropped. selfcheck covers the VST path.
Original build_vst_pack by Matthew Harris Glover; reworked for the create_system
fix, publish wiring, and rebase.
Signed-off-by: byrongamatos <xasiklas@gmail.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(career): host higher venues as opt-in content packs
Move the club and arena venue packs (~678 MB of crowd MP4s) out of the
bundle and download them on demand, keeping the bar starter bundled so
career still works offline. Leans on career's existing pack pipeline
(_download_pack: stream -> sha256 -> extract -> validate -> swap), which
already degrades gracefully when a pack is absent.
- venues.json: club/arena gain `pack` URLs pointing at per-pack, versioned,
immutable releases (venue-<id>-v<N>, matching the existing venue-arena-v1).
Arena's sha256/bytes are the real published asset (verified end-to-end);
club is a placeholder until its release is published.
- tools/content_packs.py: reusable, reproducible pack build/publish/manifest
tool. Byte-identical output for identical media (fixed order/mtime/perms,
STORED) so a pack's hash can be known before upload. --local (file://) for
offline tests, --publish for the per-pack release. Has a --selfcheck.
- .github/workflows/content-packs.yml: workflow_dispatch automation that
builds/publishes packs and opens the venues.json manifest-bump PR, so
publishing is never a manual checklist.
- test: round-trips a tool-built pack through career's real _download_pack.
Part of the nightly-slimming effort (feedBack-desktop#122). The desktop
bundle change (stop shipping club/arena) is a companion PR.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* fix(career): don't offer a venue pack until its release is published
A committed venues.json entry carries a 0-byte placeholder (and all-zero sha)
until its release exists. Previously has_pack was true as soon as a `pack`
object was present, so the UI showed a "Download" button that could only fail
(the placeholder URL 404s). Gate on a real, publish-stamped size via
_pack_published(): the card shows "coming soon" and the download endpoint 404s
until the pack is actually published. Caught by a real bundle+runtime smoke.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* fix(content-packs): address CodeRabbit review on #1023
- workflow: stop interpolating dispatch inputs into Bash (template
injection flagged by zizmor). Pass venues/version via env, validate
formats, use an argument array.
- content_packs: reject top-level files the career downloader would
refuse (PACK_FILENAME_RE) before publishing — a stray .DS_Store would
otherwise ship and fail _validate_pack_dir for every client. + test.
- content_packs: pin ZipInfo.create_system=3 so packs hash identically
across Windows/Unix runners (was the documented reproducibility caveat).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* docs(changelog): note opt-in career venue packs (#122)
Signed-off-by: Matthew Harris Glover <matthew@harrisglover.com>
* docs(content_packs): correct --publish usage in module docstring
--publish is a flag (no tag arg) and publish() deliberately omits
--clobber; the docstring said otherwise.
Signed-off-by: byrongamatos <xasiklas@gmail.com>
---------
Signed-off-by: Matthew Harris Glover <matthew@harrisglover.com>
Signed-off-by: byrongamatos <xasiklas@gmail.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* feat(server): add session-sync relay WebSocket /ws/sync/{session_id}
Cross-device followers (splitscreen's upcoming LAN pop-out mode,
feedBack-plugin-splitscreen#21) need a machine-crossing replacement for
BroadcastChannel — the one link in the follower architecture that cannot
leave the host browser. Chart data already streams per-client over
/ws/highway, so all that's missing is a dumb live-state channel.
Add a fan-out room endpoint: a JSON text frame from one client is relayed
verbatim to every other client on the same session id. No schema, no
history, no persistence — rooms are created on first join and GC'd when
the last socket leaves. The statelessness is deliberate: an idle room is
indistinguishable from a nonexistent one, and a host that crashes and
rejoins the same id resumes publishing to reconnecting subscribers with
no server-side coordination.
Caps for a LAN-exposable port: 16 KB frames (1009), 16 sockets/room and
32 rooms (1013), 120 msg/s sustained / 240 burst per socket (1008),
text-only (1003), session id validated against [A-Za-z0-9_-]{4,64}. An
over-limit socket is closed individually; a peer that dies mid-fan-out
is dropped without wedging delivery to the rest.
Closes#1030
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Kris Anderson <topkoa@gmail.com>
* fix(ws_sync): bound stalled peer sends; cap ws frames at the transport
Review feedback (CodeRabbit on #1032):
- A peer that stops draining its socket left send_text() pending forever;
since publishers await the fan-out gather, one stalled peer stalled
every publisher's receive loop behind it. Fan-out sends are now bounded
by SEND_TIMEOUT_SECONDS (5 s) so a stall becomes an eviction through
the existing failed-send drop path.
- uvicorn buffers inbound WS frames up to its 16 MB default before the
handler's 16 KB check ever runs, so the DoS bound wasn't enforced at
the transport. main.py now passes ws_max_size=64 KB (no client sends
large frames: the highway WS receives only small control messages, and
the relay keeps its tighter application cap as the primary limit).
Regression tests for both; the desktop's own uvicorn spawn gets the
matching --ws-max-size flag with the feedBack-desktop follow-up work.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Kris Anderson <topkoa@gmail.com>
---------
Signed-off-by: Kris Anderson <topkoa@gmail.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(player): drum-part picker for multiple drum charts (feedpak 1.17.0)
The last mile of the multiple-drum-parts feature: let a player CHOOSE which
drum chart plays. #1020 taught the loader + highway WS to carry several drum
parts (song_info.drum_parts + ?drum_part=<id> + a part_id echo on drum_tab);
this adds the host-chrome selector that drives it.
A "Drum part" <select> sits beside the arrangement switcher in the advanced
settings popover, shown only when a song has 2+ drum charts (drum_parts is
always present — empty for non-drum songs — so single-drum / no-drum songs
hide the row and nothing changes for them). Selecting a part re-streams that
part's tab over the highway WS, exactly like an arrangement switch.
- static/highway.js:
- reconnect() gains a third `drumPart` arg → sets `?drum_part=<id>` on the WS
URL (mirrors the existing `arrangement` param one line up). Empty/undefined
→ the primary part, i.e. byte-identical to today for any pack untouched.
- song_info handler populates #drum-part-select from msg.drum_parts and
shows/hides #v3-drum-part-row on `length > 1` (parallel to the #arr-select
block right above it).
- drum_tab handler carries msg.part_id onto hwState.drumTab (plugins can read
bundle.drumTab.part_id) and reflects it as the picker's selected value, so
the dropdown stays honest even when the server resolves an unknown/absent
selection to the primary.
- static/app.js:
- changeArrangement() gains an optional `drumPart`; at reconnect it forwards
the explicit part, else preserves the current picker selection — so an
ARRANGEMENT switch keeps the chosen drum part (parts are song-level).
- new changeDrumPart(id) delegates to changeArrangement with the current
arrangement held + the new part applied (a part switch is the same
re-stream, so it reuses all the transition ceremony). Exported on window.
- static/v3/index.html: the #drum-part-select row (hidden by default).
No plugin change: the drum renderers just draw whatever drum_tab streams.
RUNTIME-VERIFIED (Playwright, the core player, a 2-drum pack + a no-drum pack):
10/10 — the picker populates with both parts and shows for the multi-drum song;
song_info.drum_parts reaches getSongInfo(); the primary is pre-selected;
selecting the 2nd part drives highway.reconnect with the id and the WS URL
carries `?drum_part=drums-2`; the picker then reflects the server's part_id
echo; a no-drum song hides the row; no page errors. ESLint 0 errors (the two
max-lines warnings are pre-existing on these files). No pytest touched (JS-only).
Stacked on #1020.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017xGPjDBF8NTwTK7VQvizix
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
* Update reconnect source contract test
---------
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Editor now authors an arrangement's instrument as first-class data (a manifest
'type' field). Core dropped it: the sloppak loader never read 'type', and
'is this a bass?' was defined three different ways across call sites (name-only
in note_pitch_midi and the highway scale-degree path; path_bass+name in bass
selection; name-only in arrangement_string_count). So an authored type=bass
chart not named 'bass' got 6-string lane counts and guitar open-string MIDI.
- Add optional Arrangement.type; sloppak load_song lifts the manifest type onto it
- Add arrangement_is_bass(arr) = type=='bass' OR path_bass OR 'bass' in name
(None/whitespace safe), and route string count, note_pitch_midi, the highway
scale-degree base, and bass-player selection through it
- Back-compat: no bass signal -> unchanged 6-string / guitar behavior
Companion to editor #335 (first-class instrument type). Scale degrees are
display-only and never feed a grader.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(sloppak): load multiple drum parts (feedpak 1.17.0 drums-as-arrangements)
A song can now ship SEVERAL drum charts (a second drummer, an aux-percussion
layer). The Arrangement Editor already writes them per the feedpak 1.17.0 FEP
(feedpak-spec#63): the primary stays the song-level `drum_tab:` key (what this
app has always played), and each part rides the manifest as a `type: drums`
arrangement entry carrying a per-arrangement `drum_tab` file pointer and NO
note `file` — an entry this loader's file/notation gate already skips, which
is exactly why old builds are unaffected by such packs.
lib/sloppak.py:
- The arrangements loop collects drum-part pointer entries instead of merely
skipping them — but still NEVER turns one into a fretted Arrangement. That
skip is the grading invariant (an empty drum chart must not reach the
fretted pipeline / note-detection grading) and is now pinned by test.
- New `LoadedSloppak.drum_parts`: [{id, name, drum_tab}], primary FIRST. The
entry aliasing the song-level file contributes its id/name but is never
loaded twice (the primary's payload IS `loaded.drum_tab`, same object).
Legacy single-drum packs read as a one-part list; a pointer-only pack (a
writer omitted the alias) promotes its first part so has_drum_tab, the
default stream, and the drum-only placeholder keep working.
- The song-level drum_tab loading block is extracted verbatim into
`_load_drum_tab_file()` and shared by both paths, so every part gets the
same permissive posture: missing file → that part silently absent;
traversal / parse / validation failure → that part skipped with a warning,
never an aborted load. (The 9 pinned drumtab-load tests pass unchanged.)
lib/routers/ws_highway.py:
- `song_info` gains `drum_parts` (names only; always a list, empty without
drums) so a part picker can bind unconditionally.
- `?drum_part=<id>` on the WS URL selects which part's tab streams as the
`drum_tab`/`drum_hits` messages; the default and any unknown id fall back
to the primary — byte-identical legacy behavior. The `drum_tab` message
carries `part_id` only when a parts list exists, keeping the legacy frame
unchanged.
Tests: tests/test_sloppak_drum_parts.py (9) — the grading invariant +
parallel-ids pin, primary-first resolution with alias identity, legacy
one-part list, pointer-only promotion, per-part failure isolation (bad JSON,
path traversal, duplicate rels), and the drum-only placeholder with pointer
entries. Full suite: the only failures are 9 machine-environmental tests
(installed desktop plugins under LOCALAPPDATA, CRLF/path-shape assertions)
that fail identically on an untouched origin/main checkout on this box.
tools/check_spec_conformance.py passes against the spec's current HEAD
(`drum_tab` and `type` are declared keys); the semantics of the
per-arrangement placement land in feedpak-spec#63 — this PR should merge
after it.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017xGPjDBF8NTwTK7VQvizix
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
* Fix drum-part review findings
* Normalize drum part pointer identities
* fix(sloppak): enforce drums grading invariant + green the suite
- Gate the drum-pointer skip on type FIRST: a type:drums/drum entry never becomes
a fretted Arrangement even if it carries a note file/notation (with drum_tab it
is collected as a drum part, without it dropped+warned). Closes the spec
§5.2/§7.5 MUST-NOT hole (a malformed drums+file entry was being fretted-graded).
- Make test_drum_pointer_with_wrong_type_logs_warning robust (attach handler to the
feedBack logger + set WARNING, restore in finally) and fix the root-cause level
leak in test_tuning_provider_isolation.py (finally restored the handler but not
the level, leaking ERROR onto the feedBack tree and turning the suite red under
full ordering).
- Restore the chart-transform CHANGELOG bullet (#952) the drum entry had truncated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* Fix 3D Highway background controls under Venue override
When the Venue visualization override is active, the entire Background control group (style dropdown and intensity/reactive knobs) is now disabled since the venue scene owns rendering. The dropdown, intensity, and reactive controls are greyed out with a tooltip explaining the state.
Also fixes a cold-load bug where the screen hook subscription could fail to bind if the event bus wasn't ready during renderer init, by re-attempting binding on each retry tick.
Includes test coverage for both the Venue override greyout behavior and the cold-load screen hook binding fix.
* Add accessibility features and explicit global reads to background control
Refactor the player chrome background control to use an explicit `_bgReadGlobal()` function instead of relying on the implicit behavior of `_bgReadSetting(null, ...)`. The control is a single shared instance across splitscreen panels and must always read/write the global slot.
Add accessibility improvements:
- aria-pressed on toggle buttons to expose state to screen readers
- aria-label on select and intensity controls
- aria-describedby pointing disabled controls to a visually-hidden reason span
- The reason span carries dynamic explanatory text for why a control is greyed out
Add comprehensive tests verifying the new `_bgReadGlobal` helper ignores per-panel overrides and that all accessibility attributes are set and updated correctly.
* Gate player control slot on v3 UI version
Add explicit check for `window.feedBack.uiVersion === 'v3'` in _pcSlot() per docs/plugin-v3-ui.md. This prevents the plugin from attempting to mount player controls on non-v3 hosts (e.g., legacy v2 shell). Complements the existing `playerControlSlot` typeof check and improves compatibility robustness.
Updated test mocks to include `uiVersion: 'v3'` and added test case verifying that mounting is skipped when uiVersion is not v3, including a guard to ensure the retry loop terminates properly.
* Clarify 3D highway style control behavior
Document that the style controls group also greyes out when the Venue scene override is active, since the controls don't apply in that mode.
* Restore style dropdown tooltip when Venue override exits
The style dropdown's tooltip was cleared whenever the Venue override was inactive, permanently discarding the "Background style" hint set at mount time. Since the sync runs on every settings change, the tooltip was lost on the first sync and never returned.
This brings the dropdown in line with the intensity slider and reactive toggle, which already restore their base tooltip when they're re-enabled.
Includes a test asserting the tooltip returns after the Venue override exits.
* fix(highway_3d): skip player-control retry loop on non-v3 shells
_pcAcquire only runs once the renderer is viable inside the v3 player
chrome, and player-chrome.js sets uiVersion synchronously as it builds
that chrome — so a missing 'v3' at acquire means v2, not a not-yet-ready
v3. Bail before scheduling the retry loop instead of spinning it out to
the ~3s budget for a slot that will never appear.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: byrongamatos <xasiklas@gmail.com>
---------
Signed-off-by: byrongamatos <xasiklas@gmail.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
feedpak 1.16.0 (spec §5.3) added two OPTIONAL presentational fields to a
stems[] entry: `name` (display label, Readers fall back to the id) and
`description` (free text). The server dropped both while normalizing
manifest stems, so no client could ever display them.
Pass them through at the one place stem descriptors are built
(sloppak.load_song) and let both payload builders — the WS `ready` stems
list and the REST `/api/song/{f}?stems=1` preload list, which are pinned
against each other by test — carry them forward. Omit-when-absent, so a
stem without the fields does not grow null keys; non-string or blank
values are dropped rather than surfaced.
No behaviour change for existing packs or clients: the fields are
additive and every consumer that reads {id,url,default} keeps working
unchanged. The stems plugin / stem mixer display work lands separately.
Signed-off-by: topkoa <topkoa@gmail.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* fix(settings): enable Linux nightly AppImage auto-update in System settings
Fixes the Settings → System "App updates" panel so it actually works on
Linux, and adds Nightly as a selectable channel — previously missing
entirely, so Linux self-update couldn't be reached from this UI at all.
- The channel dropdown no longer gets permanently disabled the moment
the desktop bridge reports 'unsupported', which is the normal state
whenever the channel isn't Nightly on Linux. It stays enabled so the
user can switch to Nightly, the only way out of that state.
- Shows live download progress ("Downloading update… N%") and an
explicit button state machine (Check → grayed out while busy →
Restart now once staged), instead of a frozen "Checking…" during the
~1.5GB background download.
- Renders every status update from the triggering action's own return
value (checkNow()/setChannel()'s result) rather than a separate
follow-up getStatus() call, which can race against other state
changes and show a stale result even after a real success.
- setupAppUpdates() no longer re-syncs the channel to the backend on
every Settings-panel re-render — only once per page load — so a
redundant sync can no longer stomp an in-flight download's state.
- Routes update-flow events into the existing diagnostics.js
console-capture + contribute() snapshot API, so the user's existing
"Export Diagnostics" button now captures the full update decision
trace end to end — no new UI or log file. This diagnostic tracing is
what actually root-caused the bugs above, from real on-device
captures rather than guesswork.
Companion PR in feedBack-desktop (the underlying update engine).
Verified end-to-end on a Steam Deck: channel switch → check → live
download progress → restart button → relaunch onto the new build,
confirmed via a real Export Diagnostics capture showing a clean,
fully-accounted-for trace.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* refactor(settings): extract + unit-test the app-update status view; dedupe diag log
- Extract the status→UI state machine from renderFrom into a pure, exported
_appUpdateStatusView() (DOM-free) and cover it with tests/js — settings.js's
large module graph made importing it for a full harness impractical, so the
pure function is the testable seam. Behavior-preserving; renderFrom applies
the returned shape to the DOM exactly as before.
- Dedupe the [update-diag] renderFrom console line so the ~1.5s download poll
no longer floods the diagnostics ring buffer with byte-identical entries;
every real state/percent change still logs, and the structured contribute()
snapshot stays unconditional.
Left the 'audio_engine' diagnostics key as-is: the server export filters
client contributions to loaded plugin ids (diagnostics_bundle.py path-traversal
guard), so a dedicated key would be silently dropped from the bundle.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Byron Gamatos <xasiklas@gmail.com>
---------
Signed-off-by: Byron Gamatos <xasiklas@gmail.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Byron Gamatos <xasiklas@gmail.com>
* feat(input): add gamepad/controller support
Adds full gamepad/controller navigation and playback control, driven
by requests from players who use fee[dB]ack on a TV/console setup and
from wheelchair users for whom a controller is far more convenient
than a keyboard + mouse. Confirmed working end-to-end on a Steam Deck
across several rounds of on-device testing.
- static/v3/gamepad.js: polls navigator.getGamepads() and dispatches
synthetic keydown events (Arrow/Enter/Space/Escape) on the focused
element (falling back to document), reusing the app's existing
keyboard pipeline (static/js/shortcuts.js's scope-aware dispatcher,
player shortcuts, text-field/modal guards) instead of a parallel
action-mapping table. Only acts on gamepads reporting the W3C
"standard" mapping — which is what Steam Input presents for the
Deck's built-in controls, both in Gaming Mode and in Desktop Mode
via a non-Steam shortcut — so button order is guaranteed correct
and a non-standard/raw device safely no-ops instead of misfiring.
Handles Steam Input's virtual-pad duplicates (a real controller
plus 1-2 mirrored XInput slots) without spamming connect toasts or
losing input when the live pad isn't at index 0. Xbox-style face
button mapping: bottom face = Space (play/pause, and activates the
focused control), right face = Escape (back), top face reveals the
player screen's tool rail (focuses it into visibility via the
existing CSS :focus-within rule). D-pad/stick repeat while held,
mirroring OS keyboard auto-repeat.
- static/v3/gamepad-nav.js: fills the one real gap in that reuse
strategy — no screen but the song library grid had any arrow-key
navigation, and Chromium doesn't run native Enter/Space button
activation for untrusted synthetic events even when dispatched at
the focused element. Gated entirely on `!e.isTrusted`, so it only
ever reacts to gamepad-originated events and never touches real
keyboard/mouse users: emulates Tab-order (the sidebar + active
screen's real, already-focusable buttons/links) for Arrow keys,
explicitly .click()s the focused element for Enter/Space, and gives
Escape a consistent "go back" behavior — an existing in-screen back
button if one's visible (reusing each screen's own drill-down logic
for free), else the main menu. Every branch defers via
`e.defaultPrevented` to any screen that already handles the key
itself (the song grid, the player, settings), so nothing here
overrides existing behavior.
- static/v3/songs.js: adds real 2D d-pad/arrow-key navigation to the
song library's virtualized grid (only a slice of the library is
ever in the DOM), including fetching/scrolling off-screen rows into
view and correcting for the sticky filter toolbar's occlusion.
- static/v3/index.html: wires up the two new scripts.
* chore: regenerate stale tailwind.min.css
Rebuilt in a fresh clone (not the local working copy). Several plugin
directories (audio_engine, plugin_manager, community_charts, etc.) are
gitignored locally but present on disk from checking out plugin repos
for local dev/testing — Tailwind's content scan picks them up
regardless, so a rebuild against the contaminated local working copy
bakes in extra utility classes that don't belong in the real,
git-tracked build. A clean checkout reproduces CI's expected output
exactly.
* fix(gamepad): check all matching back buttons, not just the first
document.querySelector on the combined [data-ap-back], [data-albums-back],
#v3-pl-back selector only ever inspects the first match in DOM order —
since screens stay in the DOM (hidden, not removed) when you navigate
away, a hidden back button from an unrelated screen could sort before
the one that's actually visible, incorrectly falling through to
showScreen('v3-home') instead of clicking it. Uses querySelectorAll +
find(visible) instead.
* fix(gamepad): address CodeRabbit findings on connect/disconnect and grid nav
- gamepad.js: anyLiveConnectedPad -> anyLiveStandardPad, filtering by
mapping === 'standard' like firstLiveStandardPad already does, and
applied at the top of the gamepadconnected handler too. A still-
connected non-standard raw mirror could otherwise mask the real
pad's disconnect (toast never fires, polling never stops).
- songs.js _gpMove: an unset cursor now always seeds at index 0
before the first press, instead of applying that press's delta
immediately (ArrowDown/Right previously skipped straight past row
0; Left/Up only looked right by accident of clamping). Matches the
existing convention in shortcuts.js's legacy _handleLibArrowNav.
- songs.js _gpBlockedTarget: form-control/button blocking now
requires the element to be visible (offsetParent !== null), not
just present. Screens stay in the DOM hidden (not removed) when you
navigate away, so a real button focused on some other now-hidden
screen could leave document.activeElement pointing at it and block
all grid navigation indefinitely. (An el.closest('#v3-songs') scope
was tried first and reverted — it fixed that case but broke
blocking for the topbar search input, which lives outside
#v3-songs's DOM subtree even while v3-songs is active; visibility
is the distinction that actually matters, not DOM nesting.)
Skipped two CodeRabbit suggestions, verified against current code:
gating songs.js's grid keydown listener to synthetic-only events
would regress the real keyboard accessibility this PR intentionally
added (v3-songs' grid had none before); renaming the _gp* helpers to
drop their underscore prefix would break from this codebase's own
established module-private naming convention.
Verified in-browser: first arrow press lands on index 0, stale hidden
focus no longer blocks grid nav, the topbar search input still
correctly blocks it, and normal nav resumes after blur.
* test(gamepad): unit-cover the controller + nav state machines
- gamepad.test.js (10): standard-mapping filter, Steam Input duplicate-slot
dedup, disconnect masking, button edge-detection, d-pad/stick repeat timing,
analog deadzone — driven via a fake navigator + manual rAF queue.
- gamepad_nav.test.js (10): !isTrusted/defaultPrevented gating, arrow focus
traversal + clamping, hidden-element skipping, Enter/Space click activation
(not into text fields/body), Escape visible-back-button vs home fallback.
songs.js grid nav is left to on-device coverage (async + windowed-DOM heavy).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Byron Gamatos <xasiklas@gmail.com>
---------
Signed-off-by: Byron Gamatos <xasiklas@gmail.com>
Co-authored-by: Byron Gamatos <xasiklas@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* Add mid-song background picker to player chrome
Mount a background style/intensity control in the player's plugin popover so users can switch backgrounds mid-song without leaving for Settings. Uses ref-counting to manage the shared control across multiple renderer instances. The control syncs bidirectionally with settings.html and the settings bus, so changes from either UI stay agreed. Moved _pcAcquire() to after _isReady to avoid acquiring for non-viable (e.g. WebGL2-missing) renderers.
* Grey out background controls that current style ignores
Add _PC_USES table to track which settings (intensity, reactive) each background style actually consumes. Disable and grey out controls when the active style doesn't use them, preventing user confusion. Updates _pcPaint() to support disabled state with tooltip explanations, and guards click/change handlers against disabled controls.
* Add background control tests and changelog entry
Document the new background controls feature in the 3D Highway plugin that allows changing the highway background mid-song from the player's Plugin Controls popover. Add a comprehensive test suite for the background control system covering refcounting, settings sync, greying out unsupported controls, and teardown behavior.
* Generalize background control refcounting language
Update CHANGELOG and test comments to reflect that the 3D highway background control refcounting applies to any multiple renderer instances, not exclusively splitscreen. Change test name and clarify that multi-instance behavior is exercised with stubbed instances, not real splitscreen sessions (whose visualizer does not currently work).
* Reorder 3D Highway changelog entry, bump version
Moved the 'Background controls in the player' entry to a different position in the Unreleased changelog section. Updated 3D Highway plugin version from 3.32.0 to 3.33.0.
* fix: store screen.js and CHANGELOG.md with CRLF to match main
The merge of main was run with merge.renormalize=true (needed — this repo has CRLF committed while core.autocrlf=true, so a plain merge sees all 16k lines as changed). That rewrote screen.js and CHANGELOG.md to LF, which autocrlf then stored. main has both as CRLF, so every line differed and GitHub reported 16,428/16,112 for screen.js and refused to render it.
Restaged with the CRLF blobs written directly so they are what get stored. No content change; the diff drops to 316/0 and highway_3d_render_order.test.js leaves the diff entirely.
Signed-off-by: Kyle <kyle.j.t@live.co.uk>
* Unbind screen:changed hook on last release
Ensure the highway_3d control removes its screen:changed listener when the last reference is released to avoid listener/closure leaks across plugin reloads. Added a best-effort off() call and clears _pcScreenHook so future acquires re-bind correctly. Tests updated: mock feedBack on/off implemented, helpers added (screenHooks, fireScreenChanged), and a new test verifies the subscription is removed on final _pcRelease and re-subscribed on re-acquire.
* fix(highway_3d): show greyed-out reason on hover for disabled bg controls
A native-disabled <button>/<input> receives no pointer events, so its
`title` tooltip never appears — the "greyed out, says why on hover"
affordance was dead in the browser while the tests passed on the
swallowed control title. Move the reason onto a non-disabled wrapper and
set pointer-events:none on the disabled control so the hover reaches it.
Also add aria-disabled so screen readers get the state.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Byron Gamatos <xasiklas@gmail.com>
---------
Signed-off-by: Kyle <kyle.j.t@live.co.uk>
Signed-off-by: Byron Gamatos <xasiklas@gmail.com>
Co-authored-by: Byron Gamatos <xasiklas@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
`<EmbeddedFilePath>` is matched on filename stem so a format variant of the
same recording can win — an `.ogg` beside the declared `.mp3` is copied out
losslessly rather than transcoded. But the search spanned every directory in
the archive, so an unrelated file that merely shared the stem could stand in
for the declared asset: exactly the substitution the registry lookup added in
#1007 exists to prevent.
Candidates are now confined to the registry path's own directory. A genuinely
absent asset still falls through to the legacy stem match and then the first
audio asset, as documented.
Found by an adversarial pass over #1007 rather than a report — no known file
triggers it, since GP8 writes embedded audio to Content/Assets/ and that is
the only directory scanned. It needs a hand-edited archive to reach.
Both tests fail on main and pass here; their ZIP ordering is deliberate, so
the fall-through target differs from the decoy (otherwise fixed and unfixed
code return the same file and the tests prove nothing).
Claude-Session: https://claude.ai/code/session_01SFDokqh2H6mEjk1Kgbi6JW
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(gp8): AssetId is a key into <Assets>, not a filename stem
GPIF declares the backing track's audio as:
<BackingTrack><AssetId>0</AssetId>
<Assets><Asset id="0">
<EmbeddedFilePath>Content/Assets/<hash>.mp3</EmbeddedFilePath>
so AssetId indexes the <Assets> registry, which names the exact path in
the ZIP. `_resolve_audio_asset` instead compared it against each audio
file's FILENAME STEM. GP8 names embedded files by hash while ids are
small integers, so that match essentially never hit: every such file
logged "declared AssetId not found" and fell through to "first audio
asset". Silently correct while a file carries exactly ONE audio asset —
but with two, a backing track declaring id 1 resolved to asset 0, i.e.
the wrong recording, for both extract_sync and extract_audio.
Found while verifying embedded-audio extraction for a reported GP8
import; that file logged the warning on the normal path.
- `_asset_path_from_registry()` reads <Asset id=N><EmbeddedFilePath>,
normalising separators (a writer may emit backslashes). It never
decides a path exists — the caller verifies membership in the archive,
since the value comes out of the file and a stale entry must fall
through rather than resolve to nothing.
- Resolution is now a ladder: registry → legacy stem match → first audio
asset. Steps 2 and 3 are the previous behaviour, kept so existing
files and odd shapes are unaffected. Same-stem OGG preference is
preserved on the registry path too, so quality behaviour is unchanged.
Tests: registry resolution on the real-world shape (integer id, hashed
filename), the second asset finally being reachable (the actual bug), a
registry entry pointing at a missing file falling through, backslash
normalisation, OGG preference among same-stem duplicates, malformed and
absent registries degrading, and the legacy stem match still working.
Suite 1725 passed vs 1720 on main, same 99 pre-existing env failures.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01929LgKdJMyPGLf8N1WpEVW
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
* docs(changelog): record the GP8 AssetId resolution fix
Every other change in this release notes itself; this one shipped without
an entry, and the GP import path has had three fixes in two days — the
history is worth being able to read later.
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
---------
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Finding un-split songs took five taps (cycle each stem pill to its
"lacks" state) — and was quietly wrong even then: the drawer offered
five of the canonical six stems, so a piano-only song lacked all five
listed and matched a hand-built "not split" filter despite being split.
The stems section gains a "Not split" toggle that sets stem_lacks to
every instrument stem in one tap (the same lacks-ALL query Stem
Splitter's missing-stems view runs, backend semantics unchanged), and
piano joins the pill row (already in the backend's allowed set).
(Rebuilt on current main after #1003/#810/#92e78be rewrote the drawer
region — the original branch conflicted whole-file.)
Signed-off-by: topkoa <topkoa@gmail.com>
The keys LH/RH hand arc, the lift slice. split_hands was purely
heuristic (mean pitch vs middle C / largest-gap), so ANY edit to a keys
arrangement re-derived hand splits that could contradict the score's
authored grand-staff assignment — the documented "produces wrong hand
splits" failure, now that authored hands actually reach the wire
(editor #299 emits per-note `hand`; core #990 round-trips it).
- decode_wire_notes carries `hand` through ('lh'/'rh' strict enum;
junk → None so a hand-edited pack can't steer the split).
- split_hands: an authored hand always wins, and explicit notes are
REMOVED from their simultaneous group BEFORE the heuristic math runs
— one authored assignment must never skew its chordmates' guesses
(e.g. an authored LH melody note above middle C dragging the group
mean down and flipping the rest). All-explicit groups skip the
heuristic entirely; unassigned notes behave exactly as before.
Design per the piano-pedagogy review of the arc: binary lh/rh + absent
= unassigned; per-note explicit > heuristic precedence; crossing-hands
textures are exactly why the override is load-bearing.
Tests: five new in test_notation_lift.py (authored wins incl. a
crossing-hands case, group-removal-before-math with exact mean
arithmetic, all-explicit group, junk enum, decode carry-through); the
decode shape pin updated for the new key. Suite: 1723 passed (+5 vs
main; the pre-existing env failures reproduce identically on pristine
main).
Claude-Session: https://claude.ai/code/session_01EBQCHCNA81E9tHmSDHSe2Q
Signed-off-by: ChrisBeWithYou <chris@rifflarr.local>
Co-authored-by: ChrisBeWithYou <chris@rifflarr.local>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
The editor's keys LH/RH hand arc needs a per-note hand assignment
('lh'/'rh', from a MusicXML grand staff import today, hand-editable
later) to survive a sloppak save → reload: the editor already emits it
on the wire, but note_from_wire dropped unknown keys, so the field died
on every reopen.
- Note gains `hand: str | None = None` (None = unassigned; the
heuristic hand split keeps owning unassigned notes). Distinct from
`right_hand` (the bass plucking finger) — hence the spelled-out
`hand` wire key, since `rh` is taken.
- note_to_wire emits it default-omitted and validates on emit; older
readers ignore it (feedpak: unknown note keys are permitted).
- note_from_wire decodes it as a strict enum — anything but 'lh'/'rh'
(junk, wrong case, bools) falls back to unassigned rather than
poisoning downstream hand-split / hands-separate practice logic.
Groundwork consumers land separately: notation_lift.split_hands
respecting per-note overrides, and the editor's hand surface.
Editor counterpart: feedBack-plugin-editor #299.
Tests: three new wire round-trip tests in tests/test_song.py (literal
key, default-omitted, junk rejection both directions), matching the
teaching-marks test style.
Claude-Session: https://claude.ai/code/session_01EBQCHCNA81E9tHmSDHSe2Q
Signed-off-by: ChrisBeWithYou <chris@rifflarr.local>
Co-authored-by: ChrisBeWithYou <chris@rifflarr.local>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* feat(playlists): flag songs that are not in your current tuning
Making the library's tuning filter instrument-aware does not repair playlists
already built under the old guitar-first behaviour. Those keep their
wrong-tuning songs, so a player still hits a surprise retune mid-practice and
reasonably concludes nothing was fixed.
Adds a per-playlist check: each row is marked against the player's current
tuning, with a summary ("3 of 24 songs are not in your tuning"), a filter to
show only those, and an explicit removal that lists every affected song by
title and states they stay in the library. Flagging is the feature -- nothing
is ever removed without being asked for, and removal reuses the existing
per-song DELETE rather than adding a bulk destructive endpoint.
Reuses the tuner capability's coverage report and `window.feedBack
.workingTuning`, the same pair the library cards already score against,
rather than introducing another source of truth.
Two deliberate departures:
- A coverage report reads "not covered" both for a real mismatch and for a
bail-out it could not evaluate. Only a report carrying an actual reason
counts as a mismatch; the rest render as unknown. This differs from the
library grid, which paints every not-covered song amber -- acceptable on a
grid, not on a hand-curated playlist where a false warning costs trust.
- With no tuning perspective available it makes no claim at all, rather than
defaulting to guitar and reproducing the original bug in a new place.
Playlist rows carry `tuning_offsets` and `bass_only`; a tuning *name* cannot
be scored, since two "Custom Tuning" rows are different tunings.
Fully correct once the instrument-aware tuning filter lands. That dependency
is confined to `rowTuningForCheck()` in static/v3/playlists.js, marked SEAM.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SFDokqh2H6mEjk1Kgbi6JW
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
* build(tailwind): regenerate for the playlist tuning-check classes
CI's tailwind-fresh gate rebuilds static/tailwind.min.css and hard-fails if
the committed file differs. The new chip/summary/filter markup introduces
classes the previous build never saw.
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
* fix(playlists): stay within the shipped Tailwind class set
Reverts the regenerated static/tailwind.min.css and reworks the tuning-check
markup to use only classes already in the committed sheet.
Regenerating that file is not reproducible off CI: nothing pins tailwindcss,
autoprefixer or caniuse-lite, so a local `npx -y tailwindcss@3.4.19` resolves
different browser data and rewrites unrelated bytes -- a clean checkout of
main rebuilds with the -webkit-backdrop-filter prefixes dropped. Committing
that output fails the tailwind-fresh gate no matter how many times it is
regenerated.
Six utilities were new: bg-fb-good/10, border-fb-accent/50,
hover:bg-fb-accent/10, list-disc, list-inside, max-h-48, plus gap-x-3/gap-y-2.
Substituted bg-fb-good/30, the amber border already used by the mismatch
state, hover:bg-fb-card, a literal bullet in a div, max-h-32 and gap-3. Visual
intent is unchanged.
The removal-confirm test pinned the <li> markup; it now accepts either
wrapper, since what it guards is that every song is named and escaped ahead
of any DELETE, not which element wraps it.
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
* Use instrument tuning in playlist checks
---------
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(library): tuning filter answers for your instrument, not always guitar
The library indexed exactly one tuning per song, chosen guitar-first (lead >
rhythm > combo, bass only as a last resort), and nothing consulted the
player's instrument. A bassist filtering by tuning was shown the guitar
chart's tuning, so playlists built by tuning contained songs needing a
retune. Reported by a tester building bass practice sets; Covet "Shibuya" is
the clean case, with a custom guitar tuning over a standard bass chart.
Indexes each arrangement role's own tuning and makes the facet, filter, sort
and labels answer for one perspective. `guitar-lead` reads the original
unprefixed columns and adds no payload keys, so the default response is
unchanged. The same defect existed inside guitar -- lead and rhythm charts
can disagree -- so perspective is three-valued (guitar-lead, guitar-rhythm,
bass) driven by one PERSPECTIVES table rather than parallel column families.
Songs with no chart for the perspective fall back to the song-level tuning
rather than vanishing (18 of 59 packs in the test library have no bass
chart), but the fallback is marked inferred in the facet counts and on the
row instead of being silently coalesced. "Only real charts" reuses the
existing `arrangements_has` filter rather than adding one.
Bass-specific handling, from measured content:
- Bass tuning arrays are padded to six entries; charts never reference
string index 4 or 5. Truncated to four before naming and grouping.
- Grouping uses a canonical open-pitch key, so [-2,0,0,0] and
[-2,0,0,0,0,0] are one facet row instead of two.
- Offsets above +1 semitone are refused a name. Bassists tune down, near
never up; one pack ships [5,5,5,5,4,4] (A-D-G-C, unplayable, and its own
notes sit in the song's real key under standard tuning). Naming that
would send a player to retune to a tuning that does not exist.
Rhythm deliberately does not truncate -- padding is a bass finding, and
cutting a seven-string array would invent a tuning the chart lacks.
Adds an opt-in `tuning_match=playable` mode alongside exact match: a chart is
offered when your lowest open pitch is at or below its lowest open pitch, so
a five-string bass covers four-string standard and drop-D with no retune.
Open strings only -- note range is not indexed and the scan stays
manifest-only -- so it fails conservative: unknown low pitch is excluded, and
the upper bound is unchecked and documented rather than guessed.
Existing installs would otherwise never populate: the tree-signature fast
path reports "unchanged" forever on a settled library. Rows with NULL marker
columns re-extract, and the fast path is disabled until that backfill
converges (writes use '' rather than NULL, so it self-clears).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SFDokqh2H6mEjk1Kgbi6JW
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
* test(v3): accept the tuning-perspective indirection in the badge guard
The album-art badge now reads shownTuningName(), so the source-pattern guard
no longer matched the inline `tuning_name || tuning` form and CI went red.
Accept the helper, and pin the helper's own fallback in a companion test so
the guard still fails if a guitar player's tuning label is ever dropped.
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
---------
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Playlists could only ever be listed alphabetically (system playlists first).
Users who group playlists by purpose had no way to put the ones they reach
for daily at the front.
Adds a nullable `position` column and orders by
`(system_key IS NULL), (position IS NULL), position, name COLLATE NOCASE`,
so manually-ordered playlists lead, unpositioned ones keep sorting
alphabetically behind them, and system playlists stay pinned first.
Drag-reorder mirrors the existing within-playlist song reorder, adapted for
grid tiles (insert side decided on the horizontal midpoint since tiles flow
left-to-right then wrap). System playlists are neither drag sources nor drop
targets. `POST /api/playlists/reorder` requires an exact permutation of the
current non-system ids, so a duplicate, omission, extra, unknown id, or a
system id is rejected rather than silently producing duplicate positions;
booleans are rejected explicitly because `sorted([True, 2]) == sorted([1, 2])`
would otherwise slip through the permutation check.
`POST /api/playlists/sort-alpha` clears the manual order again.
Claude-Session: https://claude.ai/code/session_01SFDokqh2H6mEjk1Kgbi6JW
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(highway_3d): fret wires flash on a confirmed hit
The fret wires were static scenery: gold inside the anchor lane, grey
outside, and nothing tied them to what the player was actually doing.
Give them a job. Widen the lane/neck contrast so the wires around the
active lane read as a focus cue, and flash the wires bracketing a note
when a scorer confirms it. A fretted note lights the wire behind it and
the wire it is pressed against; a chord lights only the outermost wires
of its shape, so it reads as one bracketed block rather than a picket
fence; an open string has no fret of its own and its gem is drawn as a
slab spanning the lane, so it lights the lane's edge wires instead.
Gated on the provider verdict, never the proximity heuristic -- the
latter only means "near the strike line", so it would flash on every
passing note whether or not it was played. With no scorer attached the
neck behaves exactly as before.
Emissive (and emissiveIntensity) carry the flash, not albedo: these are
MeshStandard materials in a scene with no envMap, so raising albedo
alone barely brightens them.
Every value is a named constant -- see FRET_WIRE_* -- because the look
is a taste call that wants tuning by eye, not a derivation.
Signed-off-by: Kris Anderson <topkoa@gmail.com>
* feat(highway_3d): cap the fret-wire flash at one outer pair
Fast passages overlap their decay tails: consecutive notes on nearby
frets left three, four, five wires glowing at once — the picket fence
the chord rule was written to avoid, arriving through time instead of
through a shape.
The apply pass now decays every wire's glow state as before, but flashes
only the outermost pair of the lit span (or the single wire when only
one is above threshold). Interior wires keep decaying invisibly — the
base tier loop re-seeds their materials each frame — so the bracket
tightens naturally as the outer tails expire, and a hit inside the
current span widens nothing.
Net effect: at most two wires are ever lit, and everything currently
glowing reads as one bracket, exactly like a chord.
Signed-off-by: topkoa <topkoa@gmail.com>
* feat(highway_3d): chord flash frames the lane, not the shape
The lit lane strip spans the anchor's width (minimum ~4 frets), which
can run a fret past the chord's outermost fret. The chord flash
bracketed the shape (wire behind its lowest fret, wire at its highest),
so on those anchors the bracket sat one wire INSIDE the lit lane —
reading as misaligned rather than as a frame around what's lit.
Chord hits now light the anchor lane's edge wires: the exact wires the
lane strip itself spans, and the same pair open strings already use, so
every hit shape inside a lane produces the same bracket. The shape's
own outer pair survives only as the fallback for charts with no
anchors. Fretted and open intensities merge into one entry (they light
the same two wires now), and an all-open chord on an anchor-less chart
still degrades to no flash rather than a bad index.
Signed-off-by: topkoa <topkoa@gmail.com>
* feat(highway_3d): gem rims flash string-coloured, wire-fashion
On a confirmed hit the gem's outline now flashes in the STRING'S OWN
colour with the same intensity treatment as the fret wires — the
FRET_WIRE_HIT_INTENSITY emissive ramp, faded by the provider's alpha —
instead of the fixed spring-green mHitBright rim. Just the rims: the
lateral face fill keeps its existing green, and the sustain trail is
untouched.
Mechanics mirror the wires' pattern. mRimFlash[s] is one material per
string (created with the other per-string materials, palette-retint
aware, fog-exempt, disposed in teardown); drawNote() assigns it as the
outline on a good verdict and records the verdict alpha into a
per-frame per-string max (_rimFlashIn); the flash pass applies the
intensity ramp once per string. Shared-per-string is the same
compromise mGlow already makes — two same-string gems flashing in
different phases share the brighter alpha.
No decay tail of our own, deliberately: the material is only assigned
while the provider confirms the note, and the provider's alpha already
fades. When it goes silent the outline reverts, so idle intensity never
shows.
Signed-off-by: topkoa <topkoa@gmail.com>
* feat(highway_3d): wire flash is a lightning strike, not a lingering glow
The flash was instant-on with a 0.32 s exponential tail, and a held
sustain kept re-feeding it — wires stayed lit for the whole note. The
requested feel is a shock: light hits the frets, they jolt, it's over.
The flash is now a one-shot pulse triggered on the input's rising edge:
a near-instant crack up (RISE 25 ms), a fast fall (FALL 160 ms) shaped
(1-u)^2 so it drops hard then eases out, with a 26 Hz flicker biting
into the fall (the electric shudder — the crack itself stays clean),
then hard zero. A held 'active' verdict keeps the input high
continuously, which by construction triggers nothing new: one strike
per hit, and the wires go dark while the note rings on. A re-strike
after the provider goes silent re-triggers cleanly.
Seeking backward or a long stall clears all pulse state, and a pulse
whose strike time lands ahead of the playhead after a seek is
discarded. The outer-pair bracket rule is unchanged — it now selects
across pulses instead of decay tails.
Knobs: FRET_WIRE_HIT_RISE / _FALL / _FLICKER_HZ / _FLICKER_DEPTH
(replacing FRET_WIRE_HIT_DECAY).
Signed-off-by: topkoa <topkoa@gmail.com>
* fix(highway_3d): one wire strike per judged hit, not per wire edge
The strike trigger was a rising edge on each WIRE's input, which merged
distinct hits: two consecutive correct notes on the same fret kept that
wire's input continuously high, so the second note produced no strike at
all. The wires must respond to what the player did — one strike per
judged hit-zone event.
The trigger is now per event identity, using the same seen-map pattern
as _sparkSeen: the first frame a note gets a good verdict its key
(string|fret|time — or the chord key for a strum, which strikes once as
a unit) lands in _fwStruck and requests a strike on its wires; the
event never fires again however long its verdict stays live. Because
every producer is gated, any nonzero input in the apply pass IS a fresh
strike, so it restarts a pulse already in flight — a rapid re-hit on
the same wire re-cracks instead of being swallowed.
Seeks clear the map (replayed notes strike again); it is size-bounded
like _sparkSeen. Envelope, flicker, and the outer-pair rule unchanged.
Signed-off-by: topkoa <topkoa@gmail.com>
* Revert the lightning-strike experiment — back to the decaying glow
Reverts d003532 and e05d90e. The wire flash returns to its original behaviour: instant-on at the provider's alpha with a smooth exponential fade (FRET_WIRE_HIT_DECAY 0.32 s), held sustains keep their wires lit while the note rings, and no flicker. The outer-pair bracket, lane-framed chords, and string-coloured gem rims are untouched.
Signed-off-by: topkoa <topkoa@gmail.com>
* test: wire-tier assertions follow the named constants
The two render-order tests pinned the old literal hexes (idle 0x666688). The tiers moved to named constants with a retuned idle (FRET_WIRE_IDLE_HEX 0x4A4A60); the tests now assert the code uses the constants AND pin the constants' values, so a future retune is a deliberate two-line change here rather than a silent one.
Signed-off-by: topkoa <topkoa@gmail.com>
* fix(highway_3d): clamp provider alpha in the rim-flash path (review)
The wire-flash path clamps the note-state provider's alpha to 0..1; the rim-flash accumulation used it raw, so a provider returning >1 would over-drive emissiveIntensity. Clamped to match.
Signed-off-by: topkoa <topkoa@gmail.com>
* test: add fret inlay dots (renderOrder 3) to the hierarchy header (review)
Signed-off-by: topkoa <topkoa@gmail.com>
* test: accurate depth-flag claims, anchored depth assertions (review)
Two review findings on the render-order test file, both correct:
The header claimed ALL 3D-highway materials use depthTest:false, making
renderOrder "the only" draw-order control — but the accent halo
materials set depthTest:true. Now says "nearly all", names the
exception, and calls renderOrder the primary control. A header someone
trusts mid-debug must not overclaim.
The fret-wire depthTest/depthWrite assertions matched anywhere in
screen.js, which is full of other depthTest:false materials — the test
would keep passing if the wire material dropped the flags. Both are now
anchored to the wire material literal via FRET_WIRE_IDLE_HEX (unique to
it), as two separate anchored matches so property order inside the
literal still isn't pinned.
Signed-off-by: topkoa <topkoa@gmail.com>
---------
Signed-off-by: Kris Anderson <topkoa@gmail.com>
Signed-off-by: topkoa <topkoa@gmail.com>
The lane maps chart time to z exactly as notes do, over the window
[now - BEHIND, now + AHEAD]. That puts its near edge at +TS*BEHIND — BEHIND
seconds PAST the hit line, toward the player. Nothing is ever drawn there:
drawNote and the chord frames both clamp to Math.min(0, dZ(dt)), so notes stop
dead at z = 0. The overhang was therefore lane surface with nothing on it.
Clamp the floor geometry's near edge to the hit line. The far edge is
deliberately untouched — it still lands at -AHEAD*TS, aligned with the note
horizon, which is why the span stays AHEAD+BEHIND in the sliced path and the
clamp is applied per slice (a slice entirely past the line collapses to zero
length and is skipped before the arpeggio probe, so it costs nothing).
All four floor sites move together — the sliced lane (which also feeds both
divider loops), the fallback lane, its dividers, and the fret boundary
extension lines. They shared the identical `+ TS * BEHIND` shift; fixing only
some would leave fret lines poking past a lane that now stops.
Closes#991
Signed-off-by: Kris Anderson <topkoa@gmail.com>
app.js exported `esc` as an implicit global back when it was a classic
script. a9fce29 made it an ES module and 14b4058 carved `esc` into
js/dom.js; the window re-export list was rebuilt without it.
Out-of-tree plugins load screen.js as a classic script and call `esc()`
bare, so nothing in-tree catches the break: no-undef, a call-graph scan
and a grep all pass while the plugin throws in the field. The MIDI
plugin builds its device list with esc() inside the same try block that
catches requestMIDIAccess() failures, so the ReferenceError surfaced to
testers as "MIDI Access denied esc is not defined" — access had actually
been granted.
Pin the whole plugin-facing global surface by name, mirroring
tests/test_plugin_context_contract.py. Verified both ways against a
running app: without the fix the spec fails with "missing or not
functions: esc".
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* perf(scan): skip full library re-stat when the tree is unchanged
Startup scans globbed the whole DLC tree twice (*.feedpak, *.wem) and
stat()'d every file to detect changes — ~100k filesystem round trips on
a 50k-song library, and painful on a slow NTFS-3G FUSE mount (the "big
drive churns on every launch" report).
Adds/removes/renames of songs all bump the mtime of the containing
directory (verified on the target mount), so after a full pass we persist
{reldir: mtime_ns} for every library dir (scan_dir_signature.json, keyed
by DLC path). The next scan re-stats only those dirs — a handful vs 100k
ops — and skips the entire listing/stat pass when none changed.
Blind spot: a pack rewritten in place under the same name bumps the file
mtime but not its dir's. Rare for a song library, and the manual Refresh
(/api/rescan + /api/rescan/full) now passes force=True to always do the
full pass. force threads through kick_scan -> _scan_runner and coalesces
like the rescan-pending flag.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(scan): track directory-form songs' own dir in the signature
CodeRabbit: _library_dirs recorded only each song's parent. For a
directory-form song (loose-song folder or directory sloppak bundle),
adding/removing/replacing a file INSIDE the folder bumps that folder's
own mtime, not its parent's — so the fast path would skip a rescan it
should run. Record the song's own dir when f.is_dir(). File-form
sloppaks (a single .feedpak zip) aren't dirs, so the flat file library
is unaffected.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Tester, mid-gig: "the second song in the gig started when the first one ended.
But it showed the flyover intro again."
The flyover is arriving at the venue, and you arrive once. #968 stopped it
replaying on an arrangement SWITCH (same filename), but a gig's song 2 is a
genuinely different file, so it took the full-teardown path and played the
arrival flyover again — the camera flew in from the back of the room before
every track of the set.
The play queue now answers isContinuation(): false for the first song of a set
(or a standalone play — an arrival), true for song 2..N. onSongLoaded carries
the room over to the new song's loop on a continuation, and only a real arrival
plays the intro.
Verified on the built AppImage: isContinuation goes false (song 1) -> true
(song 2) across an advance, and song 2 no longer flies in.
Also confirmed NOT a bug, same session: "didn't show the author for the second
song." The credits card shows on a queue advance whenever the song carries
authors — reproduced with a song that has them as the advanced-to track. The
tester's song 2 simply had no `authors:` metadata (most auto-converted feedpaks
don't). No code change.
Tests: isContinuation across start/advance/clear, and that onSongLoaded gates the
flyover on the continuation check. Both fail on pre-fix source. JS 1214/1214.
Tester: "Passports does not advance in the song queue."
The play queue tells playSong "don't clear the queue I'm driving" by passing
options.fromQueue. But window.playSong is wrapped by a CHAIN of plugins —
nam_tone, midi_amp, fretboard, invert_highway, tabview — and each wrapper
forwards only (filename, arrangement), silently dropping the options object. So
fromQueue never reached playSong: it cleared the queue the instant its first
song started, and a gig/album/playlist never advanced.
Reproduced on the real build via a queue.start + a hooked clear(): the queue
went inactive with 0 remaining immediately after start, and the clear stack ran
through nam_tone -> midi_amp -> invert_highway -> fretboard -> session.js.
Fixing six plugin wrappers is whack-a-mole and the next plugin re-breaks it.
Fix it at the source instead: the queue raises an out-of-band flag
(_consumeInternalPlay, one-shot) beside the wrapper chain, not through it, and
playSong's clear-guard honours it. options.fromQueue stays as the in-band path.
The flag is consumed on read so a later MANUAL play still abandons the queue.
Verified on the real build: the gig queue stays active after start and advances
on song:ended (Iron Maiden -> Blind Guardian), and a manual play still clears.
Tests drive the real clear-guard against the queue for: a dropped-options
wrapper (the bug), the one-shot manual-play-still-clears invariant, and the
in-band fromQueue path on its own. All 3 fail on the pre-fix source. JS 1211/1211.
* fix(career): a gig's song pool is the whole genre, and re-roll varies it
Two tester reports, one root: the gig song pool was built from only two sets —
songs played ON THIS PASSPORT'S INSTRUMENT, and songs never played AT ALL
(`filename NOT IN song_stats`).
A song played on a DIFFERENT instrument's arrangement is in neither: it has a
stats row (so the "unplayed" filler skipped it), and its played bucket is that
other instrument's, not this passport's. It could never be gigged.
- "Metalcore says 137 songs only shows 1 in the gig list" — a library of
metalcore all played on another instrument. Reproduced: a guitar passport with
137 bass-played metalcore songs got a 404, zero songs. The "1" the tester saw
was whatever handful happened to be on-instrument or truly unplayed.
- "Passport re-roll does not change songs" — a set drawn from that filler was the
library's first N in table ORDER, every call. Re-roll re-proposes, so it
returned the identical set. Reproduced: 3 proposals, byte-identical.
_unplayed_genre_songs -> _fill_genre_songs: the pool is now every library song of
the genre the set hasn't already picked (a stats row on some other instrument has
no bearing on whether a song can be in THIS gig), and it is shuffled so re-roll
actually re-rolls.
Both reproduced against the real propose logic before the fix and pinned as
regression tests (both fail on the pre-fix routes.py). Full career suite green.
* fix(career): load the gig's venue pack when the gig starts
Tester: "Venue doesn't load when starting song from passport. Loads standard
particles."
crowd.setManifest(venue) — the call that actually loads a venue's crowd/stage
pack — is reached ONLY through pushCrowdManifest, and pushCrowdManifest is
called ONLY from refresh(), the career tab's own reload. A gig navigates AWAY
from the career tab to the player, so refresh() never runs during it. startGig
set the venue override and nulled _appliedManifestVenue but never re-pushed, so
the venue visualization turned on (3D highway) while its pack never loaded — the
song played over the bare highway backdrop, or over whatever venue a previous
refresh() had left applied.
startGig now pushes the crowd manifest for the gig venue right after setting the
override, using the career state the booking screen already fetched.
This is a call-graph fact, not a guess (pushCrowdManifest has exactly one other
caller and startGig is not it), but it is fixed by static analysis — I could not
reproduce the user-visible symptom locally because this instance happened to have
a manifest already applied from a prior refresh. On-device confirmation on a real
passport gig is still owed.
Guard test: startGig must push the manifest after setting the override (fails on
the pre-fix source). Career suite green.
* feat(career): extract the whole setlist before the gig starts
A feedpak is a zip, and the first play of one pays for its extraction into
sloppak_cache. Inside a set that cost landed BETWEEN songs: the player finished
a number and then sat there waiting for the next one to unpack, mid-gig.
A setlist is a known list up front, so unpack it all while the poster is still on
screen. New POST /gigs/prepare walks the set through resolve_source_dir; the
poster's Play button shows "Preparing set…" while it runs.
Best-effort by design, at every level:
- a corrupt pak in the set does not sink the prepare (it is reported in
`failed`; the play itself surfaces the error exactly as it does outside a
gig — slow beats blocked)
- a host without the library resolvers degrades to a no-op rather than 500
- a failed request just falls through to the old lazy extraction
Ordering matters and is pinned: the set is unpacked BEFORE the stage is borrowed
(venue/viz overwritten) and before the queue starts, so a proposal cancelled
while unpacking leaves nothing half-applied to unwind.
Tests unpack REAL zips rather than mocking the extractor: every song of the set
lands on disk before the first note, a re-prepare does not duplicate the unpack,
one bad pak still leaves the good one prepared, and no-library / empty-setlist
degrade cleanly. 18/18.
NB the other half of the gig report — the per-song results popup interrupting
the set (and worse, claimAutoExit'ing so the queue would not advance until it was
dismissed) — is fixed in the note_detect plugin repo, which is not part of this
checkout.
* fix(career): bound the prepare request; validate the setlist (PR #971 review)
Both CodeRabbit findings were right.
1. A HUNG PREPARE COULD BLOCK THE GIG FOREVER.
`await fetch(...)` only rejects on a network ERROR. A server that accepts the
connection and then never answers hangs indefinitely — and the gig would never
start. That makes this optimisation the exact thing the PR promises it can
never be: the reason you cannot play.
The request is now bounded by an AbortController (PREPARE_TIMEOUT_MS, generous
because unpacking a setlist is real work — but a CEILING, not a wait). Past it
we start the gig and let the first play extract lazily, as it always did. The
Play button is restored in a `finally`, so a timeout cannot strand the poster
on "Preparing set…" with Play disabled — which would have been the same bug
wearing a different hat.
2. THE `songs` BODY WAS UNVALIDATED.
A str is iterable: "abc" would have prepared three one-character "songs". And
the endpoint unpacks zips, so an arbitrary caller could ask for unbounded work.
Now list-only, string entries, blanks dropped, capped at MAX_GIG_SONGS.
Tests: the fetch is abortable and the button is re-enabled on EVERY path
including the abort; non-list bodies, non-string/blank entries, and an
oversized setlist. 50 career tests, JS 5/5, eslint clean.
* fix(career): path-traversal guard on prepare; a cap test that actually tests the cap
CodeRabbit again, and the first one is a real hole I put there.
1. PATH TRAVERSAL. sloppak.resolve_source_dir() does a bare `dlc_root / filename`
with NO containment guard — so `../../x` walks straight out of the library, and
my new endpoint handed it attacker-supplied filenames. Every filename now goes
through _resolve_dlc_path first, the same check every other filename-bound
handler applies. Pinned: `..`, backslash traversal, an absolute POSIX path and
a Windows drive path are all refused, and nothing outside the library is
unpacked.
2. THE CAP TEST WAS VACUOUS. It asserted `prepared == 0` against a fixture with no
library — where the endpoint exits before extraction — so it passed whether or
not MAX_GIG_SONGS existed. It now runs against a real library and asserts the
endpoint CONSIDERED at most MAX_GIG_SONGS of the 82 it was handed. Verified to
fail when the cap is removed.
Same class of mistake as the notedetect gigBlock: a test that passes for the
wrong reason. Worth saying out loud since it is twice in one day.
3. E702 — semicolon-joined statements in the new tests, split.
51 career tests; full suite green.
Starting a gig dropped the player onto the fallback 2D highway with no venue.
startGig() calls setViz('venue'), which installs the 3D renderer — whose init is
async — and then immediately starts its play queue. playSong() re-initialises
that same renderer a tick later. A renderer mints a fresh readyPromise per
init() and rejects the previous one with "superseded"; highway.js only checked
that the RENDERER OBJECT was unchanged, which it is. So it treated a healthy,
re-initialising renderer as a failed one, tore it down, and reverted to 2D:
renderer async init failure: Error: superseded
viz picker: reverted to default renderer (async-init-failure)
The guard now also checks the PROMISE identity: a rejection from an init cycle
the renderer has already moved on from is ignored. The renderer-identity guard
stays (a rejection for a renderer since REPLACED is also not ours), and a
genuine failure of the CURRENT cycle still reverts — both init() call sites go
through _setRenderer, which re-wires the handler every time, so the new cycle is
always watched.
Reproduced and fixed against the real build:
before: vizSelection=default viz-picker=default venue=inactive viz:reverted
after: vizSelection=venue viz-picker=venue venue=ACTIVE (no revert)
Also widens the paused-frame throttle's opt-out. The throttle fires whenever the
CHART CLOCK is stalled — not only on a pause, but through a count-in and the
credits/author overlay too. Its opt-out only asked "is a crowd video rolling",
but the venue scene animates on a clock of its own with no pack at all (backdrop
breathe, parallax, haze drift, warmth pulse — Math.sin(t) in the draw loop), so
that motion was still being throttled. It now claims frames for both sources; a
plain 3D highway with no venue reads motion mode 'off' and keeps the #654 GPU
saving.
HONEST CAVEAT on that second part: I could not get the throttle to fire in a
reproduction. A control run on the shipped code showed 100 draws/sec while
paused, not the ~10/sec a firing throttle would give — so the change is
defensible on its own terms (a stalled clock is genuinely not a static picture)
but it does NOT have a demonstrated symptom behind it. The viz fix above does.
Tests: the superseded guard, and that the throttle opt-out covers both motion
sources. All fail against the pre-fix source. eslint 0 errors; JS 1207/1207.
* feat(song-info): publish the playable stem list, so stems can preload
The stems plugin could only learn its stem list from the highway's WS `ready`,
which arrives once the highway is already up. So it fetched, decoded, and then
handed every stem's PCM to its audio worklet — copying the WHOLE SONG — with the
player already on screen.
For a 4-minute 6-stem pack that is over half a GIGABYTE of memcpy, in one frame,
on the main thread. Measured on a real load: a 698 ms frame, right as the
song-credits card appeared, with the venue video visibly stopping. That is the
"the video pauses when the author appears" report.
GET /api/song/{f}?stems=1 now returns the same list — [{id, url, default}] plus
full_mix_url — so the plugin can start the whole load at `song:loading`, before
the highway (and the venue) is drawn, where a stalled frame costs nothing.
Nothing about the work changes; only WHEN.
Opt-in via the query param so the library's own metadata calls — the hot path —
pay nothing. Deliberately NOT stored in the metadata cache: that is a
fixed-column table, and widening it would mean a schema migration plus a stale
row for every song already scanned, to cache something that is a plain manifest
read on an already-unpacked pack.
The safety property: REST and the WS must publish the SAME list. If they
disagreed the plugin would preload a graph and then throw it away and rebuild —
strictly worse than not preloading. So both now resolve `default` through one
shared helper (stem_default_on, extracted from load_song), and a test rebuilds
the WS's payload from load_song and requires the REST helper to produce the
identical list, rather than pinning either against a snapshot.
Also pinned: the mixdown is lifted OUT of the stem list (spec 5.3 — `full` is
not a layer; listing it beside the instruments would play the whole song on top
of the stems) while staying reachable as full_mix_url, a single-`full` pack keeps
it as its only playable stem, and an unreadable pack yields an empty list rather
than failing the request. Full suite 2608 passed.
Consumed by feedBack-plugin-stems (preloadSong).
* fix(song-info): call load_song for the stem payload — do not reimplement it
CodeRabbit caught a real bug, and it would have hit most real libraries.
load_song() falls back to the DEPRECATED `original_audio:` key when a pack has no
reserved `full` stem — which is every pack written before feedpak 1.15.0. My
payload rebuilt the full-mix rule from extract_meta and returned None for those:
REST would say "no full mix" while the WS said there was one.
Worse than a wrong field: the plugin would preload a graph WITHOUT the pristine
mix and — because the stem signature still matched — never rebuild. Unity
playback would silently downgrade to the lossy stem recombination.
That is exactly the drift this PR claims to prevent, and my test had a hole: I
only covered packs that carry a `full` stem.
So stop reimplementing. The payload now calls load_song, whose LoadedSloppak
already carries the partitioned stems and the resolved full mix, and builds the
URLs exactly as ws_highway does. Drift is now impossible by construction rather
than by agreement. extract_meta is reverted to its original shape (it never
needed to change), and the shared stem_default_on helper stays as the one place
`default: off` is resolved.
Tests rewritten to compare against load_song — the WS's own function — for a
reserved-`full` pack, a LEGACY original_audio pack (the case that was broken), and
a single-`full` pack. Also documents the `?stems=1` contract in CHANGELOG.md.
Full suite green.
* fix(venue): don't replay the flyover on an arrangement switch; keep the venue off other screens
Two bugs from a live career session.
1. CHANGING ARRANGEMENT REPLAYED THE ARRIVAL FLYOVER.
changeArrangement() reloads the song through the normal load path, so
highway.js re-emits `song:loaded` — same filename, new arrangement. The venue
could not tell that from a fresh arrival, so it reset the machine and flew the
camera in from the back of the room again, mid-set, every time the player
switched lead -> rhythm. The player is already on stage.
onSongLoaded now compares the filename. A repeat of the song already on stage
keeps the video pipeline running and only re-syncs the mood: the performance
restarts, so the loop follows the reset machine with a quiet crossfade, never
the intro. A genuinely different song still gets the full teardown + flyover.
2. THE VENUE SHOWED UP ON THE VIRTUOSO HIGHWAY.
The venue was gated purely on `isVenueViz()` — the selected visualization,
which is a GLOBAL preference and says nothing about what is on screen.
Virtuoso borrows the same highway_3d renderer for its practice charts, so with
Venue selected it inherited the backdrop: the crowd and the stage behind a
chromatic exercise.
Selecting Venue is a preference for the PLAYER; it is not a licence to paint
the venue over whatever else happens to be using the renderer. The venue is now
gated on viz AND screen (`shouldBeActive`), and follows `screen:changed` — it
tears down on leaving the player and rebuilds on return. Nothing else changes:
stop() already unbinds the videos from the renderer, so deactivating is enough
to clear the backdrop.
Tests: both decisions exposed as pure predicates and pinned — arrangement switch
vs new song (including the first load, and a malformed payload that must not
suppress the flyover forever), and the venue's screen scope. The existing syncViz
test encoded the OLD contract (activate regardless of screen), so it now states
the new one and additionally asserts the venue does NOT activate on virtuoso.
Includes a guard test: with Venue selected AND on the player, the venue IS
active — without it, every "not active" assertion could pass vacuously.
All 8 new/updated assertions fail against the pre-fix source. eslint clean;
JS 1199/1199; pytest 2597 passed.
* fix(highway): the paused-frame throttle was throttling the whole venue
Pausing the song dropped the venue, the crowd and the stage to ~10 fps —
"everything around the highway drops fps by a lot".
draw() caps paused frames to one per _PAUSED_FRAME_INTERVAL_MS (100ms), on an
assumption stated plainly in highway-constants.js: a heavy WebGL renderer "does
a full render every frame even while paused. That is pure waste." That was true
when a paused chart was a still picture.
The venue broke the assumption. Its video backdrop keeps playing and its crowd
reacts on a clock of their own, and BOTH are drawn into the same canvas as the
notes — so a throttle aimed at static notes throttled the entire room. The
scene only got a texture upload 10 times a second while the transport sat
paused.
Renderers can now declare that their picture is not static while the chart
clock is stopped: an optional needsContinuousFrames(). The throttle is skipped
only when it returns exactly true, and the probe fails closed — a renderer that
doesn't implement it, or one that throws, keeps the throttle unchanged. So the
GPU saving that motivated #654 survives everywhere it was actually valid.
highway_3d implements it and claims continuous frames ONLY while a crowd video
is genuinely rolling (bound, unpaused, not ended, readyState >= 2). With no
venue pack — the common case — the paused scene really is static, so it keeps
the throttle and the GPU still idles.
Tests extend tests/js/highway_pause_throttle.test.js, which guards this code
path source-level (the draw loop owns the rAF + WebGL lifecycle and is
deliberately not reproduced in a vm — see the file header). The new guards pin
that the capability GATES the early return rather than merely being called near
it, that the probe fails closed on absent/non-function/throwing/truthy-but-not-
true, and that the 3D renderer keys off the real video elements and can still
return false. All 3 fail against the pre-fix source.
eslint 0 errors; JS 1202/1202; pytest 2597 passed.
* perf(folder_library): render only the songs on screen (#965)
A song list rendered EVERY song it held. On a flat 50,944-song library that is
one <div> with 50,938 children and ~1,300,000 DOM nodes — ~4.2 GB of renderer
RSS, for a screen the user may not even be looking at (it was built while the
visible screen was v3-home).
It is not just this plugin's problem. A million-node document poisons unrelated
code: any `document.querySelector` that MISSES has to walk the whole tree before
returning null. That is exactly how song_preview's per-frame menu check ended up
consuming ~50% of the renderer and dropping the app to 2.7 fps
(feedBack-plugin-song-preview#7 fixes the per-frame walk; this fixes the tree it
was walking).
So render only what is on screen. Rows are uniform height (grid cards uniform
size), so the window is pure arithmetic — no per-row observers. Off-window songs
are represented by padding ON THE LIST rather than spacer elements: a spacer div
would become a grid ITEM in grid view and shift the columns, whereas padding
behaves identically in both layouts. Lists at or below VIRTUAL_MIN (200) render
in full exactly as before, so normal folders are untouched.
Two ordering fixes this forced, both real bugs waiting to happen:
- Both expand handlers populated the list BEFORE showing it. A windowed list
measures a real row and the scroller viewport, and both are zero under
display:none. Show first, then populate.
- _render() now tears down the previous render's scroll listeners. Without it
they survive against detached nodes and leak on every re-render.
Verified in real Chromium over CDP with 50,000 rows — the DOM glue, not just the
maths:
at top rendered= 25 rows scrollHeight=2,200,000px [0..24]
scroll 500k rendered= 31 rows scrollHeight=2,200,000px [11357..11387]
scroll 1,100k rendered= 31 rows scrollHeight=2,200,000px [24994..25024]
scroll to end rendered= 25 rows scrollHeight=2,200,000px [49975..49999]
25-31 rows in the DOM instead of 50,000; scroll height exact and constant (the
scrollbar stays honest); the last row lands on song 49,999.
Tests: _visibleWindow is pure and exposed via __test — top/middle/bottom/past-
the-end windows, the grid row-packing case, the padding-plus-rendered-equals-
total invariant that keeps the list from changing height as you scroll, and the
degenerate zero-height case (a list still display:none) falling back to
render-everything rather than to an empty list. eslint clean; full JS suite
1186/1186.
* fix(folder_library): re-window on resize and on show/hide (PR #967 review)
CodeRabbit caught two real bugs in the first pass. Both are mine.
1. GRID RESIZE. perRow and rows were captured once when the list was filled, but
paint() also runs on resize — and resizing changes the grid's column count.
The window maths then sliced against the OLD column count: wrong songs on
screen, and padding sized for a row count the layout no longer had (so the
scrollbar lied). metrics() now recomputes perRow/itemH/rows together on every
paint, so the geometry can never disagree with itself.
2. STALE WINDOWS ON SHOW/HIDE. paint() only ran on scroll and resize. Expanding
or collapsing any section moves every list below it, and a windowed list's
contents are a function of its POSITION — so those lists kept the window from
their old position and showed blank padding where songs should be until the
user happened to scroll. Both toggles now call _repaintVirtualLists().
Re-opening an already-populated section had the same flaw.
Collapsed lists also kept doing layout work on every scroll tick. paint() now
bails early when the list is display:none or detached, and forgets its last
window so re-showing repaints from scratch instead of short-circuiting on a
stale memo.
Tests: grid re-window on a column-count change, the padding+rendered=rows
invariant at two different perRow values, and a test that PINS THE FAILURE MODE —
a mismatched perRow/rows pair must not silently look correct. 12/12.
Re-validated the DOM glue in real Chromium with 50k rows (25-31 rows rendered,
scroll height exact). eslint clean; JS 1189/1189; pytest 2597 passed.
CHANGELOG entry added (also flagged).
`Path.resolve()` is a filesystem call — it lstats every component of the path.
`_resolve_dlc_path` and `safe_join` both re-resolved their ROOT on every single
call, and those run once per song, per art fetch, per scanned row.
Found while profiling a 2-fps report: on a real 50,944-song library the server
was issuing ~23,500 stat/lstat calls per second, re-walking the same three
parent directories over and over, and burning ~50% of a core doing it. It is
worst exactly where big libraries live — the library was on an NTFS-3G (FUSE)
mount, where every stat is a userspace round trip through mount.ntfs-3g (itself
visible in top). The cost was the constant re-resolution, not the work.
A root is fixed for the life of the process, so resolve it once
(safepath.resolved_root, lru_cache). Measured, 5,000 lookups against a real
library path:
before: 15,264 stat syscalls (54.2 ms)
after: 277 stat syscalls ( 0.8 ms) 55x fewer
Containment is unchanged, which is the part that matters:
- safe_join still resolves the CANDIDATE on every call — following its
symlinks IS the zip-slip / traversal defence, so it is never cached. Only
the server-owned root is.
- _resolve_dlc_path keeps its lexical containment check (deliberately does not
follow symlinks, so junction-mounted libraries keep working).
Tradeoff, documented on resolved_root: if a root's symlink is re-pointed at a
NEW target while the server runs, the old target stays in effect until restart.
Fine for a library path fixed at startup; the cache is keyed on the Path, so
switching library dir is a different key.
Tests: root resolved once across 500 lookups (the regression), a different root
is a different entry, and the containment contract re-pinned — traversal,
Windows drive-absolute, backslash, NUL, empty, and a symlink escaping the root
must still be refused. Full suite green.
The Velvet Room (50 stars) now ships in every build like the bar and
arena: 4 reactive crowd loops, 2 stingers, and a balcony flyover intro
rendered from the AXA Music Stage scene (110 spectators, state-scaled
stage washes over the venue's own neon). Audio files are dive-bar
placeholders until club-scale recordings land.
The installed/delete test now asserts bundled-fallback semantics:
with every venue bundled, deleting a downloaded pack reveals the
bundled copy instead of uninstalling.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
CodeRabbit follow-up on #961: presence checks alone would pass with
swapped loop filenames; assert the full loops/stingers/sfx objects.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Feedback Arena (150 stars) now ships in every build like the bar:
4 reactive crowd loops, 2 stingers, and a flyover intro rendered
from the UE5 arena scene (200 spectators + 396-body intro fill,
state-reactive rig lighting). Served by the existing bundled-pack
fallback; venues.json unchanged. Audio files are dive-bar
placeholders until arena-scale recordings land.
Largest file is 89MB — future re-renders must stay under GitHub's
100MB hard limit.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(career): Gold tier — a family-style goldImprov artifact upgrades an earned badge
The drill-state relay's goldImprov map (virtuoso gold_improv mints,
gained-only merged like drill nodes) turns an earned badge gold when the
passport's genre — or its genre family — has a verified improv artifact.
Gold never substitutes for the badge bar: gold-without-bronze stays
in_progress.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(career): Gold tier frontend — relay, ceremony, slam, gold ink everywhere
The drill-state relay now carries virtuoso's goldImprov map; a badge that
comes back gold gets its own ceremony + notification (tier-suffixed seen
ids — the bronze moment stays seen under its legacy id, a gold slam marks
both), a gold stamp slam in the book, gold ink on the shelf-cover mini
stamp, and the real gold foil chip. The bronze page's dashed 'Gold rung
coming' preview becomes a live invitation to jam the style in Virtuoso.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(career): gold review fixes — family-space style matching, intake guards, rail counter
The review's showstopper: virtuoso mints goldImprov under raw
STYLE_PALETTES ids ('punk', 'djent', 'disco'), which are mostly NOT
family keys — the tier check now matches in family space (artifact style
and passport genre bucket through the same _genre_family keyword match),
so a 'punk' gold reaches a 'punk rock' passport. Also: non-dict
goldImprov 400s loudly instead of silently dropping; evidence-free
artifacts (no verifier) never mint; goldImprov gets the same pre-merge
size bound byNode has (junk under the cap could otherwise persist
forever and wedge every later relay at the post-merge check); the
instrument-rail badge counter counts gold (earning gold no longer made a
badge vanish from the rail); first-artifact-wins is now asserted against
the persisted snapshot instead of vacuously.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* fix: remote transcription posts to /transcribe, not /align (stem-splitter#17)
transcribe_vocals_remote() POSTed the vocal stem to /align. That endpoint is FORCED ALIGNMENT —
"here are the lyrics, tell me when each word is sung" — and its `text` field is required. We have
no lyrics; transcribing them is the entire point. So the server rejected every request with a 422
from FastAPI's validation layer, before its handler ever ran, and remote transcription has never
worked for anyone.
It now posts to /transcribe (added in feedBack-demucs-server#14), which takes only the audio.
`language` moves from the query string to the FORM BODY, where the server actually reads it
(Form("")). As a query param it was silently ignored, so an explicit hint did nothing and
Whisper's auto-detection quietly decided instead — loading the wrong wav2vec2 aligner. It
"worked", it was just wrong, which is the failure mode that hides for months.
Error bodies are no longer cut at 300 chars. The body IS the diagnosis: a 422's JSON names the
field it rejected, a 500's traceback answers on its LAST line. Both got decapitated — which is
part of why this stayed invisible for so long. The message explaining the bug was inside the part
that got cut.
Nothing caught any of this because every test of this module tested the MAPPER, fed a hand-written
dict. The mapper was always fine. The request was never exercised, and the request was the bug.
tests/test_lyrics_transcribe_remote.py now pins it: the endpoint, the form field, the multipart
upload, the bearer token, an instrumental returning no lyrics rather than an error, and a 404
saying the server is too old. Verified they FAIL against /align + params.
Signed-off-by: topkoa <topkoa@gmail.com>
* fix: make the error-body cap an actual bound; correct the docstring's endpoint
- _err_body() appended the truncation marker AFTER slicing to _MAX_ERR_BODY, so the result could
exceed the cap it exists to enforce (4014 chars for a 4000 bound). A cap that is only a
suggestion surprises exactly the callers who trust it — a log line, a job record persisted to
disk and re-read on every load. The marker now fits inside the bound.
It also stripped after measuring, so a short JSON body followed by kilobytes of trailing
whitespace got truncated: real content cut to make room for blanks. Strip first, then measure.
- The public docstring still advertised /align — the exact contract this PR exists to change, in
the one place a reader would look for it. It now says what the function does and why, and that
an older server answers 404.
Found by Copilot and CodeRabbit on #959.
Signed-off-by: topkoa <topkoa@gmail.com>
* fix: keep the exception line when truncating — the tail is the answer
_err_body() kept only the HEAD of an over-long body. On a traceback the last line is the
diagnosis, and the docstring said exactly that while the code threw it away: a 4000-char window
holding "Traceback (most recent call last)" and none of the exception is a window onto nothing.
Same mistake as the 300-char cap it replaced, one level up — cutting off precisely the part the
function exists to preserve.
Head AND tail now, both inside the bound: two thirds head (what was being attempted), one third
tail (what actually went wrong), with the marker between them. Verified the test FAILS against
head-only truncation.
Found by Copilot on #959.
Signed-off-by: topkoa <topkoa@gmail.com>
* fix: every failure out of transcribe_vocals_remote() is a RuntimeError; 404 says why
The docstring promised one failure mode — RuntimeError — and the caller (_maybe_transcribe_lyrics)
catches exactly that so one song's failed lyrics don't take down the batch around it. But a DNS
failure, a timeout, a reset connection or an unreadable stem escaped as requests.RequestException
or OSError, walked straight past that handler, and turned "this song's lyrics failed" into "the
whole batch died".
A 404 now explains itself. Bare "404" sends someone hunting for a typo in their server URL; the
real answer is that their server predates /transcribe, and we are the only ones in a position to
know that.
Found by Copilot on #959.
Signed-off-by: topkoa <topkoa@gmail.com>
---------
Signed-off-by: topkoa <topkoa@gmail.com>
* feat(career): gigs frontend — poster, runner strip, summary, encore
Career v3, WS3 (frontend half), rebuilt cleanly on merged main (v3-a/b/c
in) after git interleaved the structurally-similar canvas functions:
- Book a gig from any opened passport: /gigs/propose renders as a GIG
POSTER (venue presents GENRE NIGHT, numbered bill) with re-roll and
Save/Copy poster (natively-drawn canvas via blob-io, audible failure
paths, slash-safe filenames).
- Play the gig: venue override + Venue viz handoff, then
playQueue.start(..., {source:'gig'}) — the queue's auto-advance runs
the set; zero new playback machinery.
- Floating gig strip (body-level, pointer-events none, z 35 per the
chrome invariant) tracks set position and names what's next.
- Completion = song:ended with an empty queue → POST /gigs → summary
poster overlay with per-song accuracies; encore fires the crowd
celebrate + confetti (reduced-motion: neither). song:stop with a dead
queue = abandoned (no log); end-of-song teardown (queue still active)
must NOT abandon.
- Gigs played render as dated rows in the passport book.
vm tests: runner advance/abandon semantics via the queue-state seam.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(career): stage restore survives a setViz throw; size-register row
CodeRabbit on #956: a setViz failure nulled the restore snapshot AFTER
the overrides were written, permanently borrowing the stage — snapshot
now captured before any write, write failures keep it intact. Also
registers career screen.js in docs/size-exemptions.md 'Planned, not
exempt' (1,516 lines; max-lines WARNS non-blocking — the split plan
needs Byron's sign-off).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* docs: register career screen.js in the size register (planned, not exempt)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(career): gigs backend — propose a setlist, log the completed set
Career v3, WS3 (backend half). A gig is career's verb:
- POST /gigs/propose {instrument, genre, size}: setlist from the
passport's own stubs — qualifying songs (per the genre's badge bar,
family-aware) shuffled for a free re-roll, topped with the
highest-accuracy near-bar songs as stakes, and filled from UNPLAYED
genre songs when the passport is young (the first gig is how stubs
start). Names the highest venue the current stars can book.
- POST /gigs: logs a COMPLETED set only (abandoned sets never log — no
fail state). Per-song accuracy = MAX(last_accuracy) from song_stats,
freshly written by the set's own plays; encore = avg ≥ the data-driven
bar (passports.json gig.encore_accuracy, 0.75). Appends to the career
state file (same atomic _save_json pattern).
- Passports view: per-passport gigs (newest first, capped 20) and
per-instrument gig_count — the profile wall's gig line lights up.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(career): gig backfill offsets by qualifying taken, not picks length
CodeRabbit on #954: after the stakes loop appends near-bar songs,
qualifying[len(picks):] overshoots and skips eligible qualifying songs
— a stocked passport could still get a short set.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(career): profile passport wall, home career card, shareable PNG card
Career v3, WS2. The identity artifact leaves the plugin tab:
- Profile: #v3-profile-passports-mount (core, one div) filled by career
on v3:profile-rendered — per-instrument shelves of earned covers,
hours, gig count, open-career link. Absent-not-empty.
- Home: the plugin-count stat tile becomes #v3-dash-career-slot with the
old stat as fallback content; career replaces it with a trading-card
tile (leather + foil shine, badge count, hours, closest-stamp ask) on
the existing v3:dashboard-rendered event.
- Shareable card: static/js/blob-io.js (downloadBlob lifts the idiom
duplicated verbatim in settings-io/diagnostics-export — both
refactored; copyImageBlob wraps ClipboardItem, returns false to signal
the download fallback). Earned passports get Save/Copy card: a
natively-drawn 480×640 canvas (leather, stamp ring, stubs+hours line);
copy falls back to download with a notice when the clipboard refuses.
- Mount-point convention documented in docs/plugin-v3-ui.md.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(career): external surfaces stay absent until a passport exists
CodeRabbit on #955: a bare commitment produced a zero-passport wall and
replaced the dashboard fallback. Docs also now say mounts may hold
fallback content and plugins REPLACE, never append.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(career): practice invitations — closest stamps + bring-these-up
Career v3, WS1. The passport now points at the practice that pays:
- Stubs carry next_star_at (the same primitive _stars() uses) and each
passport exposes `nearest`: the top 3 non-qualifying songs by distance
to their next star, in the worklist order.
- "Closest stamps" strip above the shelf: the in-progress graded
passports nearest to minting, each row naming the ask — N more songs
(with the nearest title + best %) or the blocking Virtuoso drill.
Rows open the passport.
- "Bring these up" list on the stubs page of in-progress passports;
earned pages stay memorabilia (no homework on a won badge).
- Invitation-voiced throughout: no meters, no completion pressure.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* docs(test): comment says qualifying-bar ranking, matching the assertion
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
The enrichment fallback made the passport rack real (hundreds of MB
sub-genres) but only the five exact umbrella keys carried Virtuoso
drills. Genres now resolve to a family by keyword substring (MB's
vocabulary is open — 'metalcore' must hit metal without an alias),
first-match-wins in list order ('blues rock' → blues), and inherit the
family's requirement from the same genres map. Exact entries still win;
per-instrument scoping unchanged; unmatched genres stay songs-only.
Data: families for metal (incl. djent/grindcore/thrash/doom), blues,
jazz (bebop/swing/bossa), funk (disco), rock (punk/grunge/shoegaze).
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* fix(sloppak): the full mix is a stem — drop the invented `original_audio` key (#933)
Core read, served, and depended on `original_audio:` — a top-level manifest key
this repo invented in #583 that the feedpak spec never defined. The format
already had a home for the pre-separation mixdown: it is a stem. feedpak 1.15.0
(feedpak-spec#53) RESERVES the id `full` for it, so read it from there.
The key existed to work around a bug in our own reader. The packer's comment
said so plainly: "we must NOT list the full mix as a playable stem — the player
sums every entry in `stems` and does not gate playback on `default`, so a listed
full mix plays on top of the stems". Faced with a reader that would double the
song, the packer put the mixdown outside `stems` and invented a key to point at
it. The fix belongs in the reader, and that is what this is.
load_song() now partitions the stem list: `full` comes out as
LoadedSloppak.full_mix, the instruments stay in .stems. Nothing that sums stems
or draws one fader per stem can see the mixdown, so retaining it is safe — which
is what lets the packer put it where the format says it goes.
- ws_highway: `song_info` gains full_mix_url / has_full_mix. The old
original_audio_url / has_original_audio remain as deprecated aliases for one
release so an older stems plugin keeps working (#945).
- `stems` on the wire, and stem_ids / stem_count in the library index, are now
INSTRUMENT stems only — a separated pack that retains its mixdown no longer
advertises a bogus "full" chip or an inflated stem count.
- enrichment: fingerprint against the mixdown wherever it lives. This widens
coverage — _song_audio_file() previously returned None for any pack without
the invented key, so fingerprinting silently did nothing for nearly every pack.
- sloppak: `original_audio:` is still READ as a deprecated fallback, because
every pack in the wild carries it and would otherwise lose its pristine mix.
tools/migrate_full_mix_stem.py rewrites those packs into the spec shape
(original/full.ogg -> stems/full.ogg, add the `full` stem at default:off, drop
the key); the fallback and the aliases die with #945.
The spec gate keeps the debt honest: the grandfather entry now tracks #945, and
the gate fails if it goes stale.
Verified: spec gate OK (4/4, incl. ingesting the spec's new example pack that
retains `full`); 2493 python tests, 995 js tests; migrator round-tripped over
real packs from the library and the results pass the spec's reference validator.
* fix(migrate): discover directory-form packs instead of silently skipping them
iter_packs() searched only files, so a directory-form pack (`song.sloppak/`, the
authoring shape) was walked INTO and never yielded — silently missed by a run
that's meant to be exhaustive. Discover suffix-named directories too (yielded
whole, not descended into), and route packs through migrate_pack/verify_pack.
Directory packs are REPORTED as `dir-form-unsupported`, not rewritten in place:
a single-file pack is replaced atomically (a fully-built temp archive swapped in
with one os.replace), but a populated directory can't be swapped that way, so an
interrupted in-place rewrite could leave an authoring pack half-migrated. The
status is a problem status, so it counts against the run's exit code and shows
in the summary — the operator re-packs or migrates it as a `.feedpak` instead of
it vanishing from the report. Addresses a CodeRabbit review finding.
Signed-off-by: Kris Anderson <topkoa@gmail.com>
* fix(migrate): verify requires an explicit `off` on a retained full mix
verify_zip accepted any non-truthy `default` on a multi-stem `full` (missing,
empty, boolean, `false`/`no`/`0`, malformed) as "ok". But core defaults an ABSENT
`default` to True — ON (lib/sloppak.py: `s.get("default", True)`) — and treats an
empty/unrecognized string as ON too, so a migrated-shape pack whose `full` stem
has a missing or blank default beside instrument stems would actually play the
mixdown on open and double the song. verify was certifying that as safe.
Require an explicit normalized `off` beside instrument stems: `on`-ish values are
reported `full-stem-default-on` (actively plays), everything that is not a
normalized `off` is reported `full-stem-default-not-off`. The migrator already
writes the literal `off`, so its own output is unaffected; this also certifies
the pack is in the tool's canonical, most-portable shape. The len>1 gate is kept,
so a sole `full` stem (which IS the audio) is not policed.
Adds parametrized coverage for missing / empty / boolean / off-ish / malformed
defaults, and a sole-full-stem case. Addresses a CodeRabbit review finding.
Signed-off-by: Kris Anderson <topkoa@gmail.com>
---------
Signed-off-by: Kris Anderson <topkoa@gmail.com>
Co-authored-by: Kris Anderson <topkoa@gmail.com>
* fix(sloppak): bound the unpack cache, and add a way to read a song without unpacking it
A tester's sloppak_cache reached 60 GB from an 1800-song library — his entire
library, unpacked, none of it played. Stems are already-compressed audio, so an
unpacked pack is ~1.1x its zip: the cache is a second, DECOMPRESSED copy of every
song it touches. It had no size cap, no LRU, and no cleanup of any kind — not even
when the song itself was deleted.
Two halves:
1. resolve_source_dir() now evicts least-recently-used songs to stay under a cap
(FEEDBACK_SLOPPAK_CACHE_MAX_MB, default 4 GB ≈ 130 songs of recency; 0 disables).
The sweep runs on unpack — the only moment the cache grows — so it can't drift.
An evicted song is dropped from _source_cache too: get_cached_source_dir() is
the only thing media.py consults before falling back, so a stale path there
would 404 every stem for the rest of the process instead of re-unpacking.
get_cached_source_dir() now also verifies the dir still exists, which makes
"just delete sloppak_cache/ to reclaim disk" safe advice.
2. read_member_bytes() reads ONE file out of a pack without unpacking it — the
same trick read_cover_bytes() uses so the library grid doesn't explode every
pack to show a cover. Unpacking a whole song to read a few KB of JSON is ~45x
write amplification; doing it in a loop over the library is what produced the
60 GB. rig_builder's library-wide tone batch is the caller that did exactly
that (fixed separately); this gives it, and everyone else, the right primitive.
Eviction is concurrency-safe: unpacks run 2-at-a-time, so a dir being written is
marked in-flight and the sweep skips it — checked and rmtree'd under one hold of
the guard, and the marker is released even if the unpack raises (a leaked marker
would make that dir permanently un-evictable).
read_member_bytes normalizes both the requested path AND the archive's stored
member names through safe_join, taking the last match — so './arrangements/x.json',
backslash members from Windows tooling, and duplicate members that normalize to
the same path all read back exactly as unpack-then-read did. Zip-slip is rejected
before anything is opened.
Tests: tests/test_sloppak_unpack_cache.py. All bite-tested (reverted each fix,
watched it fail) — including one that was passing vacuously: a freshly-unpacked
dir is the most-recently-used, so the LRU never reaches it and the in-flight race
test proved nothing until the packs were sized to force the sweep that far.
* test: split semicolon-joined statements (E702)
CodeRabbit on #950. Style only; no behaviour change.
* feat(library): effective genre falls back to MusicBrainz enrichment
Converted packs rarely carry a genres manifest key — on Byron's real
library 1188/1190 songs had no genre, starving the genre facet and the
career passport rack (2 usable genres) while song_enrichment already
held MB genres for 636 matched songs.
The effective-genre expression now resolves: per-song override → pack
genre → json_extract(enrichment.genres, '$[0]') for MATCHED rows only
(review/failed candidates could carry the wrong recording's genres).
Same fast-path gating as before: the plain indexed column is used
unless overrides or enrichment genres actually exist; stand-in DBs
without the table degrade via the OperationalError guard. Career
passports pick this up automatically through _effective_genre_expr().
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* docs: changelog names manual rows in the enrichment fallback
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
CodeRabbit on #947: the prior lastPlayPosition POST already stamped
last_played_at, so the assertion passed even if the seconds-only path
left it unchanged — a guard that cannot fail. Assert on a fresh row.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
CodeRabbit follow-up on #942:
- add_play_seconds() now stamps last_played_at (like touch_position):
an unscored play that ran to the natural end WAS played — recent /
Continue ordering must see it. Resume position stays untouched.
- stats-recorder post() treats non-2xx as failure: a 4xx/5xx JSON error
body parsed as an object read as success, silently dropping the
accrued seconds instead of re-queuing them.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(career): passport visuals pack — tilt, emerging ink, gold foil
Career v2, WS4. All CSS + a rAF-throttled pointer handler, no deps:
- Trading-card tilt on EARNED artifacts only (shelf covers + the badge
page stamp): pointer-tracked perspective rotateX/Y with a glint sweep
following the pointer. The cover's jitter rotation moves into a CSS
var so the tilt transform composes with it; the blanket cover-hover
translate is scoped :not(.pp-tilt) so it can't fight the tilt.
Hover-capable pointers only; off under prefers-reduced-motion.
- Emerging-stamp ink: the ghost stamp fills by qualifying/required via
a conic-gradient — the stamp visibly carves in, no numbers added.
- Gold foil preview: the "coming" note gains a dashed foil chip with a
periodic shimmer — honest, never earnable-looking.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(career): tilt rAF race + freshly-slammed stamp becomes a card
CodeRabbit on #944: a queued tilt frame closed over the departed card
and re-applied vars after pointerleave; and the just-slammed stamp
never regained pp-tilt until the next open.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Career v2, WS3. Bronze in blues/rock/metal/funk/jazz now also asks for
the genre's signature Virtuoso drill, data-driven in passports.json:
blues_shuffle, rock_power_backbeat, melodic_metal_gallop,
sixteenth_pocket, vl_shells — with career-side display labels the
passport page renders instead of raw node ids.
- virtuoso_nodes becomes {instrument: [node_ids]} so a keys passport
never demands a guitar drill; a flat list keeps meaning guitar
(virtuoso's content is guitar-first).
- _node_cleared also accepts keysCleared (a top-tier clean pass in one
key — virtuoso's FIRST gained-only artifact). The depth rungs
additionally require a maxed speed tier, too high a bar for Bronze.
- Genres without a curated entry stay songs-only.
Note: pre-release behavior change — v1 passports aren't in any shipped
build, so no earned badge can demote in the wild.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Career v2, WS2. Nothing measured play time before (the achievements
plugin's final-position shortcut double-counts loops and mis-reads
seeks). Now:
- stats-recorder.js accrues WALL-CLOCK seconds across song:play/resume ↔
pause/stop/ended spans (single spans clamp at 2h against suspend
inflation) and piggybacks them as `seconds` on the POSTs it already
sends; failed POSTs restore the accumulator; a session reset flushes
first so time can't re-attribute to the next song/arrangement.
- POST /api/stats accepts optional `seconds` (finite, 0 < s ≤ 6h) on the
scored and position branches, plus a new seconds-only branch for
unscored plays that ran to the natural end — banks time WITHOUT
touching the resume position (song:ended must not overwrite Continue)
and still counts as playing today for the streak.
- song_stats gains additive idempotent `seconds_total`; record_session/
touch_position accrue, new add_play_seconds() for the seconds-only
path; the legacy-encoding stats merge sums seconds across duplicates.
- Passports surface it: "14.2 h in Blues" under the badge stamp and on
the shelf cover sub-line — a true fact that only grows, never a
target or a meter (Stage 5 post-cap, per the career design).
Tests: seconds accrual/validation/seconds-only branch (stats API),
per-instrument-and-genre summing (career), fmtHours formatting (vm).
Full suites: pytest 2480, JS 1165.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Earning a genre badge now stages the full moment (career v2, WS1):
- venue-crowd.js gains a public celebrate(): machine.force('ecstatic')
commits instantly (bypassing STABLE_MS/DWELL_MS; the stamped
lastSwitchAt makes the dwell window HOLD the forced state before the
real perf machine reasserts) + a cheer stinger via the same
_lastStingerAt=-Infinity bypass the end-of-song reaction uses. If a
stinger/intro owns the idle layer, the ecstatic loop is queued via
_pendingLoop exactly like onPerformanceState. No-op without a
manifest/active venue.
- career detectNewBadges() calls badgeCeremony(): crowd first, then a
body-appended full-screen overlay 300ms later (it cannot live in
#pp-overlay — #plugin-career is display:none during playback): dimmed
backdrop, the bronze stamp slamming in with a shine sweep, a 42-piece
canvas confetti burst, click-or-4s dismiss.
- prefers-reduced-motion: chime + fbNotify only, no overlay.
Tests: machine.force commit/dwell-hold/bogus-state, celebrate export +
no-manifest no-op, celebrate-called-once-per-badge, crowd-absent and
crowd-throwing degradation.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* chore(career): refresh bar pack to v4 + restore crowd-SFX setting
Pack v4: per-character desynced animation starts, flyover intro,
per-venue reaction sounds (sfx-up/sfx-down in manifest).
Settings: re-add the crowd sound reactions toggle that was dropped
when settings.html became the passports data panel.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(v3): crowd mood-change sound reactions (cheer up / boo down)
Port the venue-crowd SFX runtime that the settings toggle and the
pack's sfx-up/sfx-down files were built for: on a committed mood
transition, play the venue's own cheer (up) or boo (down) one-shot,
gated by the feedBack-venue-crowd-sfx setting (default off).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* chore: rebuild tailwind.min.css for settings toggle classes
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Retitle the toggle "Fullscreen" (from "Start in fullscreen") and reword
the description to "Run fee[dB]ack in fullscreen mode. On macOS, changes
take effect on the next launch."
The macOS note is honest about a native-fullscreen limitation: AppKit
drops the first programmatic fullscreen-enter on a window created
windowed, so on macOS the desktop side applies the pref at next launch
rather than live. Windows/Linux apply it live on the first toggle. The
note is self-scoping text (no platform-detection code needed).
Signed-off-by: gionnibgud <gionnibgud@gmail.com>
Adds a "Start in fullscreen" toggle to the Settings → System panel,
addressing the desktop request in feedBack-desktop#97: users want the
app to launch fullscreen without hitting the OS hotkey every time.
The block ships hidden and is gated exactly like the App-updates block:
setupWindowOptions() only unhides + wires it when the feedBack-desktop
bridge exposes window.feedBackDesktop.window.{getStartFullscreen,
setStartFullscreen}. Web/Docker builds have no such bridge, so the
section never appears there. Persistence lives desktop-side because
only the Electron main process can read the pref at window-creation
time — core just proxies through the bridge.
The desktop bridge + launch behaviour land in a follow-up
feedBack-desktop PR.
Signed-off-by: gionnibgud <gionnibgud@gmail.com>
CodeRabbit follow-up on #936 (the one Major — overlay outside the click
root — was verified false: the host mounts every screen.html root inside
#plugin-career, ✕-close confirmed working live):
- Tabs: aria-selected/aria-controls + role=tabpanel/aria-labelledby.
- Book overlay: role=dialog + aria-modal + aria-label; focus moves to
the close button on open and returns to the opener on close.
- seenBadges(): guard non-object JSON so a corrupt stored value cannot
throw on every passport refresh (covered by a new corruption test).
- Fresh-session suppression test (badge seen → no re-notification).
- Stylelint declaration-empty-line-before nit in .pp-stamp.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
The Passports tab beside Venues renders the badge journey physically:
- Per-instrument passport book: embossed CSS-leather cover, 3D page-turn
spread (badge page left, ticket stubs right), Escape/backdrop close.
- Wax-seal commitment ceremony (Stage 0) — pressing the seal commits the
instrument; opening a first passport runs the ceremony implicitly.
- Rubber-stamp badge slam: earned badges chime + notify immediately, the
slam (with ink bleed, page shake, deterministic sin-hash jitter) plays
when the passport is next opened, then the badge is marked seen.
- Ticket-stub repertoire: qualifying songs as collected stubs.
- Brochure rack: unopened genres as "Explore next" invitations — no
greyed slots, no completion meters (the anti-list as layout).
- Drill relay: on virtuoso:progress bus events the career screen posts
the full virtuoso.progress localStorage snapshot to the drill-state
intake (debounced; one-time bootstrap when the server has none).
- Four synthesized sfx (stamp/seal/page/chime, 13 KB total) as plugin
assets; prefers-reduced-motion disables the theatrics.
Pure logic (ppKey, ppJitter, badge diff/seen) is covered by a bare-vm
node --test suite via a window.__careerPassportTest seam.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
The badge-journey layer on top of career stars (Christian's career-mode
v2 design, composed with the shipped venue system). Badges are computed
on read from song_stats × the library's effective genre — never stored:
Bronze = N genre songs at min_stars (data-driven in passports.json,
default 5 songs at 2★) plus any configured virtuoso drill nodes.
New endpoints under /api/plugins/career/:
- GET /passports passport walls per instrument: badges, ticket
stubs (qualifying songs), library genres, drills
- POST /passports/commit instrument commitment (idempotent wax seal)
- POST /passports/open open a genre passport (implies commitment)
- POST /drill-state intake for the relayed virtuoso.progress
snapshot (career's frontend listens on the bus)
Instrument attribution reuses progression.instrument_for_arrangement via
the song_stats arrangement index; the genre column goes through the
host's override-aware effective-genre SQL. Non-graded instruments (bass,
drums) render shown-not-judged — repertoire, never a false badge denial.
Persisted state (commitments, opened passports, drill snapshot) lives
under CONFIG_DIR/career/ and rides the settings export bundle via
settings.server_files; a minimal settings.html documents it.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Two review nits: check_allowlist_closed() raised a traceback when
--baseline-exceptions pointed at a missing file (the error now says CI
derives it from the base branch and local runs should omit the flag), and
check_key_coverage() would traceback on a reader with a SyntaxError (now a
::error:: naming the module — belt-and-braces, since such a module can't
pass pytest either, but this job may run first).
Signed-off-by: topkoa <topkoa@gmail.com>
Design change, at the maintainer's direction: the gate now checks out
feedpak-spec at HEAD instead of a pinned SHA. .feedpak-spec-ref, its
40-char validation step, and the pin-bump machinery are gone.
Rationale: it is vital that the app conforms to the spec — the current
spec, not a snapshot. The pin bought determinism at the cost of a
maintenance loop (bump PRs, a PAT, weekly latency) and a window where the
gate verified against a stale spec. Tracking HEAD makes the dev flow fully
self-serve with zero upkeep: gated PR -> FEP -> spec merge -> re-run
checks -> green. Nothing to bump.
The trade-off is accepted with eyes open, and the docs state it: the
normal FEP is additive and can only loosen the gate, so it cannot redden
anyone's PR. Only a breaking spec change (rare, deliberate, MAJOR per the
spec's compatibility policy) turns PRs red repo-wide — which is the
correct org-wide signal that the app is out of conformance. The CI job
logs the spec SHA each run verified against, so any red run is
reproducible.
Failure messages now also say why it matters beyond the one PR (also at
the maintainer's direction): non-conformance that lands shows up as red CI
on every teammate's PR until it is resolved, and only its author can clear
it — the FEP route keeps everyone else unblocked. Tone softened throughout
(the exceptions-file header now explains rather than shouts).
Signed-off-by: topkoa <topkoa@gmail.com>
_parse_exceptions() now validates the document shape — top level must be a
mapping, 'exceptions' must be a list, each entry a mapping, and YAML parse
errors are caught — each failing with a ::error:: instead of an
AttributeError traceback. CI output must say what to fix. Parametrised
tests cover all four malformed shapes.
docs/feedpak-spec-gate.md: the Limitations section still described the
pre-flow-aware scanner (KEY_OPS, name-list-only receivers). Now states the
actual residual gaps: function-parameter manifests are recognised by name
only, and helper-mediated literal keys (song.py's
_gap_fill_manifest_absent(manifest, "album")) are unseen by both the scan
and the readers-complete guard, since they share one detector.
Signed-off-by: topkoa <topkoa@gmail.com>
Review found lib/routers/chart.py binding `m = load_manifest(p) or {}` and
reading eight manifest keys through it. `m` was not in MANIFEST_VARS and
`m.get` did not match the readers-complete regex, so the module was
invisible to BOTH halves of the gate — unlisted and unscanned. Same for
lib/routers/song.py (binds `manifest` from load_manifest for enrichment
gap-fill). Both are now in READERS.
The structural fix, not the name-list patch: keys_touched() now discovers
receivers flow-aware — any local assigned from load_manifest(...) is a
manifest dict, whatever it is called. MANIFEST_VARS remains only as the
fallback for manifests that arrive as function parameters (ws_highway).
A plain `m = {}` is not a receiver; test pins that.
readers-complete now reuses keys_touched() itself instead of a parallel
KEY_OPS regex — the two detectors diverged once already (that is exactly
how chart.py slipped through), so now there is one detector and one truth.
check_reverse() gets a 300s subprocess timeout: the validator executes at a
pinned SHA, but a pathological pack or validator bug should fail the job,
not hang the runner to the Actions-level timeout.
Tests: flow-aware receiver under an arbitrary name (read + write), and the
negative — a plain dict named `m` stays out of the scan. 17 pass.
Signed-off-by: topkoa <topkoa@gmail.com>
The CI gate catches spec drift at merge time; these two additions catch it
at write time, which is where "developer didn't read the spec first"
actually happens.
CLAUDE.md (Song Formats): a spec-is-sacrosanct paragraph next to the spec
pointer — the spec defines the format, the app implements it, any new
manifest key/file/directory lands in the spec first via the FEP process,
and the gate has no in-repo bypass. AI agents and contributors both hit
this while writing feedpak-touching code, not after CI reddens.
.github/pull_request_template.md (new — the repo had only issue templates):
a feedpak-surface section requiring either "doesn't touch pack I/O" or a
link to the landed FEP + the .feedpak-spec-ref bump, plus the standing
changelog/tests/DCO checklist.
Signed-off-by: topkoa <topkoa@gmail.com>
Self-review finding: the gate is what keeps the app from drifting off the
feedpak spec, but the gate itself had zero pytest coverage — a refactor
could quietly weaken keys_touched or the allowlist logic and nothing would
notice. The protector needs protecting.
tests/test_spec_gate.py pins the load-bearing behaviours:
- read/write classification: get() reads; subscript Store and setdefault()
write (the two forms that were blind spots in review); the
load_manifest-wrapped get; unrelated dicts and non-literal keys ignored.
- exceptions file: duplicate keys and issue-less entries rejected.
- the closed allowlist: growth fails, shrink and steady state pass,
bootstrap skips.
- live-tree checks, same as CI: READERS matches the codebase, and the only
non-spec key core touches is the grandfathered original_audio.
Also fixes stale "Layer 2/3" docstrings on check_forward/check_reverse
(they are layers 3/4 since allowlist-closed landed) — the same
docs-lag-the-code class this PR's review kept catching; now the numbering
is asserted by the printed [n/4] headers next to them.
Signed-off-by: topkoa <topkoa@gmail.com>
The allowlist-closed diff hardcoded `origin main`, but ship-ci.yml also runs
this workflow for PRs into release/** and for pushes to release/**, where a
main baseline diffs against the wrong branch and can fail changes that have
nothing to do with the allowlist. It now resolves the base:
PR -> github.event.pull_request.base.ref (the branch it merges into)
push -> github.ref_name (the branch itself; its tip already contains the
change, so the diff is a no-op — enforcement happens at PR time)
Also from review, all documentation drift introduced by my own earlier
commits:
- The layer count said "three" in the module docstring, the workflow comment,
the docs, and the changelog. There are four (allowlist-closed was added).
- The changelog listed three scanned modules; there are five.
- The docs and changelog stated the rule for keys core *reads*, omitting
writes — which are equally gated, and land in every pack we emit.
- The CI summary line labelled grandfathered keys "pending spec", implying
adoption is the only resolution. For original_audio it is not: the fix is
removal. Relabelled "grandfathered (tracked debt)".
- feedpak-spec-exceptions.yml said an entry clears when core "stops reading"
the key; the rule is "no longer reads or writes".
- Replaced a bitwise `&` over two bools with two named results and an
explicit `and` — both checks must run (a stale READERS list and an
undeclared key are separate failures; short-circuiting would hide one), and
`&` reads like a typo.
Signed-off-by: topkoa <topkoa@gmail.com>
Review found the gate was scanning less than it claimed.
READERS missed two modules that genuinely touch feedpak manifests:
lib/routers/ws_highway.py (reads `authors`) and lib/gp2notation.py (loads
manifest.yaml, stamps feedpak_version, writes the file back). Keys touched
there were going entirely unchecked.
The scan also only recognised writes done via subscript, so
`manifest.setdefault("k", v)` — exactly how gp2notation.py stamps
feedpak_version — was invisible. setdefault with a literal key now counts as
a write.
The deeper problem is that READERS is hand-maintained, and a hand-maintained
list rots; that is how both modules went unnoticed. check_readers_complete()
now re-derives the set: any module under lib/ (or server.py) that both
touches manifest keys and shows a feedpak signal must be listed, or the build
fails. It is a guard on the gate itself.
The list stays explicit rather than becoming a glob, because `manifest` is
overloaded here: lib/loosefolder.py (the loose-folder manifest.json) and
lib/diagnostics_bundle.py (the diagnostics bundle manifest) have their own
unrelated manifests, and scanning those would flag *their* keys as feedpak
drift. Both score zero on the feedpak signals, which is what keeps them out.
Now scanning 5 modules: 20 reads, 2 writes, all spec-declared.
Signed-off-by: topkoa <topkoa@gmail.com>
The gate's purpose is to make a non-conforming change *not merge*, so the
person merging must stop and decide whether to take it through the format
process. The escape hatches defeated exactly that: a developer who did not
want to write a FEP could name their key `x-whatever`, or append an entry to
feedpak-spec-exceptions.yml with any issue link, and merge. Both were
self-serve and in-repo. That is a speed bump with a signed excuse note, not a
gate.
The relief valve is the FEP process itself, not something in this repo. The
spec's governance already says so: "A change is not part of the format until
it lands here."
Removed the `x-` prefix bypass. It was invented here, not in the spec — the
spec reserves no experimental namespace. Its "unknown keys are reserved for
forward-compatibility" rule is about *tolerating* other implementations'
keys, not a licence for core to mint its own.
feedpak-spec-exceptions.yml is now a CLOSED grandfather list. A new check
(allowlist-closed) diffs it against the base branch and fails any PR that
ADDS an entry; removal stays allowed, so the list can only shrink. Deleting
an entry does not by itself pass the gate — key-coverage still fails while
core reads the key, so the entry goes when the code goes.
Every failure message now points at the FEP process and at bumping
.feedpak-spec-ref to the merged spec SHA, which is the one supported way a
new manifest key reaches core.
CI fetches the base branch to diff the allowlist; the bootstrap flag covers
the one case with no baseline — the PR introducing the gate.
Signed-off-by: topkoa <topkoa@gmail.com>
Six more findings from CodeRabbit and Copilot on #934. All valid; four were
my own docs lagging the write-checking change in d0626f5.
check_forward() now discovers example packs recursively, so a pack nested
under examples/<group>/ can't slip past the "every example pack" contract.
Taken WITHOUT the suggested is_file() filter, which would have broken it: a
feedpak is dual-form — a zip (foo.feedpak) or a directory (foo.feedpak/) —
and the spec's own examples ship as directories, so is_file() would have
matched zero packs. Suffix matching covers both forms.
load_exceptions() rejects duplicate keys instead of silently keeping the
last one, which would quietly retarget the tracking issue for a piece of
debt this file exists to track.
The sloppak import is wrapped so a missing dependency produces a CI-legible
::error:: rather than a bare traceback.
Docs caught up with the code: the exceptions file header, its stale-entry
rule, and the changelog all said "reads" when the gate checks reads AND
writes.
Signed-off-by: topkoa <topkoa@gmail.com>
Review feedback from CodeRabbit and Copilot on #934. All six findings were
valid; one is fixed the other way round from how it was suggested.
Key-coverage now checks manifest WRITES as well as reads. Copilot correctly
spotted that `ast.walk` ignored subscript context, so
`manifest["year"] = ...` (lib/songmeta.py) scored as a read — but the fix is
not to drop writes. A key core *writes* is spec surface pointed outward: it
lands in every pack we emit, so an undeclared one seeds the ecosystem with
non-spec data. Subscripts are now classified by ctx (Store = write, Load =
read) and both sets are checked, with distinct error messages. Today: 19
reads, 2 writes, all declared.
Workflow:
- persist-credentials: false on the repo checkout — the job runs repository
code and never pushes (CodeRabbit / zizmor artipacked).
- .feedpak-spec-ref must be a full 40-char SHA. actions/checkout resolves
branches and tags in `ref` too, so a non-SHA there would silently un-pin
the spec — precisely what the file exists to prevent.
- Pin jsonschema==4.26.0, for the same reason the spec SHA is pinned: an
upstream release must not redden this job on a PR that changed neither
this repo nor the spec.
Script:
- check_forward() guards a missing examples/ dir instead of raising an
unhandled FileNotFoundError.
- TemporaryDirectory() instead of mkdtemp(), so a local run doesn't leak.
Signed-off-by: topkoa <topkoa@gmail.com>
The spec already carries the pre-separation mixdown as a stem
({id: full, file: stems/full.ogg}), so the key added a second, redundant
location for audio to a format that already had one. Adopting it into the
spec would make that permanent; the resolution in #933 is to remove it.
No behaviour change — the gate is agnostic about which way a violation
resolves, and only insists that one of the two happens deliberately and in
the open before the code merges. This just stops the exception entry, the
changelog, and the docs from presupposing adoption.
Signed-off-by: topkoa <topkoa@gmail.com>
feedpak is published as an open format with its own repo, normative spec,
JSON Schemas, and reference validator. That makes the spec a contract with
everyone outside this repo: third-party packers, converters, and players
build against it, and it is meant to be the complete description of a pack.
Nothing enforced that. #583 added a manifest key (`original_audio`) that
core, lib/enrichment.py, and the stems plugin all now depend on, but which
was never added to the spec — so a spec-compliant pack stopped being a
fully-working pack, the reference validator could not warn authors about a
key it had never heard of, and third-party tooling began emitting an
`original/` directory reverse-engineered from an example in a code comment.
See #933.
We cannot mechanically prove core interprets a key the way the spec means.
We can prove three surface properties, and they cover the drift that
actually happens:
1. key-coverage — every manifest key core reads is declared in the spec's
manifest.schema.json (AST scan of lib/sloppak.py, lib/enrichment.py,
lib/songmeta.py).
2. forward — core's load_song() ingests every example pack the spec ships.
3. reverse — every pack committed here passes the spec's own
tools/validate.py (7/7 pass today).
The spec is pinned by SHA in .feedpak-spec-ref rather than tracked from its
default branch, so a change over there cannot redden an unrelated PR here;
bump it in its own PR, where a red result is precisely the signal that core
does not satisfy the new spec.
A gate with no legitimate way to say "yes, deliberately, not yet" gets
switched off the first time it blocks a release, so there are two escape
hatches: the reserved `x-` key prefix (always permitted, and it tells every
third-party packer the key is not stable surface), and
feedpak-spec-exceptions.yml, which requires a tracking issue per entry. An
exception that goes stale — the spec caught up, or core stopped reading the
key — fails the build, so the allowlist cannot become somewhere drift
quietly accumulates. `original_audio` is seeded there against #933 so the
gate lands green and starts blocking the next instance immediately, rather
than requiring #933 to be resolved first.
Dev/CI tooling only; never on the serve or Docker path (constitution
Principle I). jsonschema is installed in the CI job, not added to
requirements.txt.
Signed-off-by: topkoa <topkoa@gmail.com>
Three more rules, all learned by shipping the bug first. Every one of them
produced a symptom that pointed nowhere near its cause.
RULE 5 REWRITTEN — `isConnected` lies about a panel that is a pane, in BOTH
directions:
- true when the panel is not here (it is in a pane window)
- FALSE when the panel is perfectly fine — the host detaches the element the
moment a pop-out starts, before the new window has loaded
Code that rebuilds on that `false` builds a SECOND panel while the host still
holds the first. Docking brings both home. The one the user can see is the
original, which the module no longer points at — so its close button closes the
other, invisible panel ("the X doesn't work"), and the chip gets re-attached to
the impostor ("the pop-out icon vanished"). Two baffling symptoms, one duplicate,
nothing in the stack trace.
Ask the pane system where the element is (`panes.isOpen(id)`), not the DOM.
RULE 6 (new) — a plugin that can be re-injected must be able to remove itself.
Without a teardown the second run duplicates every observer, timer and listener —
and leaves a stale pane registration, which is worse than untidy: `element` is
resolved LAZILY at open time, so the host gets a node from a dead instance. Pop
out, and it moves a panel nobody owns. Includes the teardown people forget:
panes.unregister().
RULE 1 EXTENDED — panel-internal id lookups. document.getElementById returns null
once the panel has moved, so every update it guards silently stops happening
while the user is looking at the panel. Search from the panel instead. Elements
outside the panel never move and are fine as they are — audit which is which.
Signed-off-by: topkoa <topkoa@gmail.com>
Every rule here is something that has already gone wrong on this feature —
mostly in core's own code, twice in the two plugins that adopted it first.
They are cheap to get right up front and miserable to diagnose later,
because a broken pane almost always LOOKS perfect.
The traps, and why each one is easy to walk into:
- Your code still runs in the main window. That is exactly why moving the
element works at all — and exactly why `document.body.appendChild(tooltip)`
inside a popped-out panel lands in the window the user is NOT looking at.
- Don't hide your own panel when it pops out. Core hides it and leaves a
stub. A plugin that also hides it hides the node that just moved — which is
precisely how core's own chip shipped a blank pop-out window.
- Use `hidden` or a class, not inline `display`, for show/hide. `.fb-paned`
forces the panel visible while it is out; when it docks and that class is
removed, an inline `display:none` reasserts itself and the panel returns
invisible.
- `isConnected` does not mean "docked". A panel in a pane window IS connected,
just not to this document. The test you meant is
`el.ownerDocument === document`.
- `element` is a function so it can be resolved late: return the LIVE node, and
re-attach the chip if you rebuild your panel (Camera Director rebuilds on
every mode change).
- rAF is throttled while the main window is backgrounded — which it is, whenever
the user is looking at your pane. Event-driven panels don't care; continuously
animating ones will stutter exactly when they are the only thing on screen.
- Don't synchronise anything. One realm, one panel. Writing sync code means
you have misunderstood the model.
Also states what core guarantees back, including the one that cost the most to
learn: the element is evacuated BEFORE the pane window's document is destroyed,
so it comes home alive rather than as a photograph of a panel with every
listener in its subtree silently gone.
Signed-off-by: topkoa <topkoa@gmail.com>
Two comments pointed at `main.ts` and `pane-hosts.ts` as though they were in
this repo. They are not — they are in got-feedback/feedBack-desktop, and a
contributor reading only this codebase would go looking for files that do not
exist.
Named the repo and the paths, and said the part that actually matters: nothing
here depends on that code. In a plain browser a pane window is simply a pop-up;
the desktop side only upgrades it. And the frame-name prefix is a contract
across two repos with no build-time link between them, so the comment IS the
link — worth saying out loud.
Signed-off-by: topkoa <topkoa@gmail.com>
The pane window got a ONE-TIME snapshot of the app's theme classes and the
interface-scale custom property. The app changes both at runtime — Interface
size emits `scale:changed`, the theme emits `theme:changed` /
`v3:cosmetics-applied` — so an already-open pane went on rendering at the old
scale, in the old palette, the moment the user touched either.
"Looks identical" has to keep being true, not merely start out true.
A pane window now follows those three events for as long as it is open, and
stops on unplace(). The inline style is assigned wholesale rather than merged:
unlike the class lists (where pane.html's own `fb-pane-window` must survive),
there is nothing in the pane document's inline style to preserve — and
concatenating on every change would grow the attribute without bound as the
user dragged the scale slider.
Also: the dock's focus() always smooth-scrolled, ignoring
prefers-reduced-motion — which panes.css already honours for the card's flash
animation. A smooth scroll is motion too, and someone who asked for less of it
meant this as well.
Signed-off-by: topkoa <topkoa@gmail.com>
Two from review.
1. _whenReady's own comment said a SecurityError means the pop-out is not
reachable from this realm and "no amount of waiting will fix it" — and then
it waited the full 10s deadline anyway. Ten seconds of a detached panel and a
half-popped-out UI, for a condition we had already diagnosed as fatal.
It now gives up after a 1s grace instead. Not instantly, deliberately: a
throw *during* the navigation from about:blank to /pane would otherwise take
down a pop-out that was about to work perfectly. A second is far more than
that transition needs and far less than a user should spend staring at a
detached panel.
2. attachChip() took opts.header on trust. It's a public plugin API, and a
truthy non-Element header (a selector string, a wrapper object, a ref) is an
easy mistake — one that surfaced as a confusing DOM exception from deep
inside core instead of a TypeError naming the offending pane.
Signed-off-by: topkoa <topkoa@gmail.com>
panes.css hides an empty dock (.fb-pane-dock.is-empty { display: none }), but
the element was created without the class — so between creation and the first
card it was a visible-to-CSS, announced-to-screen-readers role="region"
landmark containing nothing.
Harmless in practice today (the dock is created lazily, on the same tick as the
card that prompted it), but the CSS contract should hold from first paint rather
than from the first _syncEmpty(), and any future caller of dock() gets the right
thing for free.
Found by CodeRabbit on #928.
Signed-off-by: topkoa <topkoa@gmail.com>
`spec.width` / `spec.height` are plugin-controlled, and the window host builds
window.open()'s feature string by concatenation:
'popup,width=' + spec.width + ',height=' + spec.height
`spec.width || 380` passed anything truthy straight through. So a width of
'300,menubar=1' would not merely be an invalid size — it would inject window
features. Less dramatically, any non-numeric value produced a malformed feature
string and a pane that failed to open for no visible reason.
They now go through _size(): Number, round, reject anything not finite and
positive, clamp to 120..4000. A hostile or careless value falls back to the
default instead of reaching window.open() at all.
Verified against the obvious inputs: '300,menubar=1' -> 380 (default), '300' ->
300, 0/-50/NaN/{}/'abc' -> 380, 1e9 -> 4000, 5 -> 120.
Signed-off-by: topkoa <topkoa@gmail.com>
A ~15-line console.info dumping computed styles, sizes, child counts and the
element's inline style on every single pop-out. It was instrumentation written
to chase the "panel comes home dead" bug, and it should have gone out with the
rest of the debugging — it survived the cleanup.
Removed rather than downgraded to console.debug: nothing here is worth keeping
even behind a flag. The failures it was built to diagnose are all handled and
commented now, and the paths that can still go wrong (window never loads, adopt
throws) already log a console.error that says what happened.
Signed-off-by: topkoa <topkoa@gmail.com>
_copyStyles appended the app's stylesheets to the pane document — which
already links panes.css — so they landed AFTER it. In the app document
panes.css loads last, after tailwind/style/v3, and its rules win ties. In the
pane window that order was silently inverted, letting core styles override the
pane chrome and the .fb-paned placement rules.
Cascade order is not a detail here. "Looks identical" has to include the order
things are said in, or the same markup with the same sheets can still render
differently.
The clones now go in BEFORE pane.html's own link, preserving their relative
order among themselves and leaving panes.css last, exactly as in the app.
Signed-off-by: topkoa <topkoa@gmail.com>
The window host's place() is asynchronous — it opens the window, waits for
/pane to load, and only then adopts the element in. But the manager emits
`panes:opened` as soon as place() returns, and the chip reacts by putting its
"popped out" stub where the element used to be.
So for that gap the user saw BOTH: the real panel still sitting in its
original spot, and a stub next to it claiming the panel had left. On a window
that never loads, that lasts the full 10s readiness timeout.
Detach the element as soon as we commit to moving it. That is not destructive:
the node keeps its owner document, its listeners and its closures — it is
simply out of the tree, waiting for a document to be adopted into. And if the
window never loads, closePane() puts it straight back at its home, which is
exactly what the failure path already does.
The dock host has no such gap; its place() moves the element synchronously.
Signed-off-by: topkoa <topkoa@gmail.com>
A pane id is plugin-controlled, and it becomes a KEY in the persisted
{ paneId: hostId } map. `__proto__` and friends are not ids, they are booby
traps:
- `map['__proto__'] = 'window'` on a plain object corrupts the map, and
can reach Object.prototype.
- `map[id]` on a polluted (or hand-edited) object can return a value straight
off the prototype chain for a pane that was never remembered at all — so a
pane could be "restored" to a host nobody ever put it in.
Three layers, because each is a one-liner:
- Reject `__proto__` / `constructor` / `prototype` as pane ids at
registration, so they never reach storage.
- Re-key whatever comes out of localStorage onto a null-prototype object, so
a corrupt or hand-edited value cannot smuggle a prototype in.
- Read with an own-property check.
Also from the same review:
- Removed `window.__fbPaneWindows`. It was exposed for pane-desktop.js back
when that file needed to reach the window handles; the rewrite dropped that
need and nothing has referenced it since. Dead global, and its comment
described a collaborator that no longer exists.
- Corrected the /pane cache comment. It claimed a stale page would leave the
window blank, which stopped being true when the readiness check gained a
`doc.body` fallback — it would still work, just without the pane window's
own layout. A comment that describes a failure mode the code no longer has
is worse than no comment.
Signed-off-by: topkoa <topkoa@gmail.com>
Six more findings from CodeRabbit on #928. Three are real bugs.
1. THE CHIP HID DOCKED PANES. `_onOpened` decided "did the pane take my
element?" from `ownerDocument !== document`. That is true for a pane in a
pop-out window — and false for a pane moved into the DOCK, which lives in
this very document. So docking a pane stamped `.fb-pane-detached`
(display:none !important) onto the panel the user was looking at, and put
the stub next to it instead of at its home.
The element cannot answer this question — `isConnected` is true in a pane
window, `ownerDocument` is this one in the dock. Both were live bugs. Ask
the manager, which knows exactly what it handed to the host:
`panes.elementOf(id)`. That holds for every host, and for reconciling after
the fact (detail == null), which is what a plugin rebuilding its panel
mid-pop-out triggers.
2. `.fb-paned` FORCED `display: block !important`. A panel that is
`display:flex` or `grid` would be silently re-laid-out while detached —
the exact opposite of "placement only", and precisely the kind of surprise
this feature exists to avoid. Removed.
Making a hidden panel visible is a separate job, and it now belongs to the
manager, which does it without touching the panel's display MODE: clear
`hidden`, and clear an inline `display:none` if that is how the panel hides.
3. VISIBILITY IS NOW RESTORED. The hosts used to set `el.hidden = false` and
never put it back, so the docs' "core only changes placement" was a lie and
a panel's hidden state was quietly lost. The manager stashes both `hidden`
and the inline `display` on open and restores them on dock: a panel that was
closed when you opened its pane from the tray goes back to being closed; one
that was open stays open.
Plus:
- The launcher rebuilt its whole list on every panes:opened/closed — including
the one fired by clicking a button in that list — destroying the button under
the user's finger and dropping focus to <body>. It now restores focus to the
toggled pane's button.
- `_copyStyles` cloned every stylesheet link, including the panes.css that
pane.html already loads. Skip sheets the pane document already has.
- Docs: the chip may route to the DOCK, not always a window (it goes through
detach() → the host router). `header` precedence was documented backwards —
an explicit `header` wins. And the visibility contract above is now written
down rather than being a surprise.
Signed-off-by: topkoa <topkoa@gmail.com>
A plugin that rebuilds its panel (Camera Director does, on every mode
change) takes the chip with it. attachChip() returns a detach(); call it
before re-attaching, and again in teardown, or you leave a stub pointing at
DOM that no longer exists.
Found by CodeRabbit on #928.
Signed-off-by: topkoa <topkoa@gmail.com>
Three real findings from CodeRabbit on #928, all in current code.
1. closePane() adopted the element out of the pane window ONLY when its
original home was still connected. If the panel never had a parent (a
plugin that builds it lazily and hands it straight over) or its container
was torn down while the pane was out (a screen change), the whole block was
skipped — leaving the element inside a window we then close, which strips
every listener in its subtree. That is exactly the "comes home dead" failure
this ordering exists to prevent; the guard just moved it from the common
path to the rare one, where it is far harder to spot.
Adopting and re-homing are two different jobs and only one of them is
allowed to fail. Adopt UNCONDITIONALLY — that is what rescues the element —
and insert only when there is somewhere to insert it. With no home the
element ends up owned by this document but not in it: detached, intact,
listeners alive, ready for the plugin to re-insert.
2. The pane window's `beforeunload` handler was registered TWICE, comment block
and all — a bad scripted edit on my part. Harmless (the handler is
idempotent via panes.isOpen) but dead duplicate code. Also fixed the stale
comment further down that still claimed there was no beforeunload listener
at all.
3. _copyStyles ASSIGNED className on the pane document's <html> and <body>
instead of merging. pane.html sets `class="fb-pane-window"` on <html>, and
panes.css hangs the pane window's own chrome off exactly that — so copying
the app's classes over it silently took the pane window's own layout with
them. Merge both class lists, and append the interface-scale inline style
rather than replacing the attribute.
Also guarded the docs' integration example behind a feedBack.panes check: the
doc says the API is optional, and then showed an example that would throw on a
host without it.
Not applicable (reviewed against e5cbea2, the branch's first commit, before the
rebuild): the prototype-pollution findings in pane-bridge.js and pane-mirror.js,
and the `panes[]` manifest validation in plugins/__init__.py. All three files are
gone — 188bdaa deleted the entire cross-realm bridge, mirrorGlobal, and manifest
layer when panes switched to moving the real DOM node.
Signed-off-by: topkoa <topkoa@gmail.com>
.fb-pane-stub is a <button>, and both of its sibling controls
(.fb-pane-chip, .fb-pane-card-btn) have an explicit :focus-visible outline.
It didn't, so keyboard focus fell back to the UA default and looked
inconsistent next to them.
Found by CodeRabbit on #928.
Signed-off-by: topkoa <topkoa@gmail.com>
Docking a popped-out panel brought it home DEAD. It rendered perfectly —
right markup, right size, right place — and every control in it was inert:
the close button, the sliders, the presets, even the pop-out chip. A
photograph of a panel.
Closing a pane window tears down its document, and the panel was still
inside it. The node itself survives (the manager holds a reference), but
every event listener in its subtree goes with the document that hosted
them. Two paths did this:
1. closePane() called the host's unplace() — which closes the window —
BEFORE adopting the element back. Order is now reversed, and the
comment says why so nobody helpfully "tidies" it back.
2. The user closing the pane window themselves was only noticed by the
`closed` poll, which by definition runs AFTER the document is gone.
The window now gets a `beforeunload` listener that brings the element
home while its document is still alive.
That listener has to be attached AFTER /pane loads: window.open() hands
back a throwaway about:blank document, and anything registered on it is
discarded when the real page replaces it. This is the same trap that made
the pane window blank in the first place — adopt into about:blank and the
panel is destroyed a moment later — and it is now handled in both places.
The `closed` poll stays, but only as a last-resort net for a CRASHED pane
window, where nothing can be saved.
Also fixed while chasing this:
- The chip stamped `.fb-pane-detached` (display:none !important) onto the
element to hide it in the main window — and that element is the one we
move, so the class travelled with it and blanked the pane window. The
chip now only hides an element the pane did NOT take, and marks the hole
with its stub otherwise. "Did not take" is an ownerDocument test, not
isConnected: a panel sitting in a pane window IS connected, just not
here, and a plugin that rebuilds its panel (Camera Director does, on
every mode change) re-runs attachChip while popped out.
- The stub was inserted "before the element", which is nowhere — the
element has left the document. The manager now hands over the element's
recorded home, and the stub goes there.
- GET /pane sent no cache headers. A stale copy is especially nasty here:
the opener waits for an element inside that page before adopting, so an
old cached version means the pane window just sits there blank.
Verified in the desktop app: pop out, use the controls in the pane window,
dock back, use them again. Panel comes home alive.
Signed-off-by: topkoa <topkoa@gmail.com>
Math.round let 431/433 (99.54%) display as 100%. Floor at every
accuracy display site (HUD, library badges, dashboard, lessons,
profile, playlists, calibration overlay); stored fractions and
mastery thresholds unchanged.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Both pop-outs opened blank.
window.open() returns immediately, and the window it hands back already has
a document — an `about:blank` one, whose readyState is 'complete'. So the
host cheerfully adopted the panel into THAT, it worked for a few
milliseconds, and then /pane finished loading, replaced the document, and
took the panel with it. Blank window, vanished element.
Waiting for 'load' is no better: it may already have fired for about:blank
before we could listen.
So don't trust readyState and don't trust 'load' — wait for the one thing
that exists only in the document we actually want: pane.html's
#fb-pane-root. Poll for it (guarding the cross-document window while it is
mid-swap), give up after 10s, and on failure bring the element home rather
than stranding it in a window that never loaded.
Also drop the popup's 'beforeunload' listener: it was registered on the
about:blank window and discarded along with it, so it never fired. The
`closed` poll is what notices a user shutting a pane window — as it must be
anyway, since a crashed renderer never says goodbye either.
Signed-off-by: topkoa <topkoa@gmail.com>
The first cut of this got the model wrong. A pane was a SECOND
implementation of the plugin's panel — its own sliders, its own styling,
driven over a cross-realm bridge (ctx, a state store, capability RPC,
mirrorGlobal, a stream sampler). Popping out gave you something that
resembled the panel you popped, and every feature it did not reimplement
(presets, tabs, EQ, language) was simply gone.
What a user wants from "pop this out" is the thing they popped out.
So: MOVE THE REAL ELEMENT. Same-origin windows can adopt each other's
nodes, and an adopted node keeps its event listeners and its closures.
The panel goes on running the plugin's own code, against the plugin's own
state, in the plugin's own realm — it is merely being DISPLAYED in another
window. Copy the app's stylesheets into that window and it looks identical
too, because it is identical.
The plugin's side collapses to two lines:
feedBack.panes.register({ id, title, element: () => panelEl });
feedBack.panes.attachChip(panelEl, id);
and everything comes along: the CSS, the listeners, the presets, the
state. Nothing to keep in step, because there is no second copy.
Deleted, all of it now pointless: pane-bridge (ctx + transports), pane-hub
(the cross-realm server), pane-runtime (the pane realm's boot), pane-streams
(the rAF sampler that existed because an AnalyserNode can't cross a window),
pane-mirror (mirrorGlobal), pane-plugins + the manifest `panes[]` key and its
server-side validation, panes.state(), and both built-in demo panes. ~1200
lines. None of it was wrong — it was all correct machinery for the wrong
problem.
Consequences worth knowing:
- The window MUST be opened by the renderer with window.open(), not by the
desktop's main process: a window we did not open gives this realm no handle
to its document, and without the handle there is nothing to adopt into.
Electron turns the same-origin window.open() into a real BrowserWindow
anyway (setWindowOpenHandler → 'allow'), so we get the OS window AND the
live DOM link. The desktop side finds it by frame name.
- `.fb-paned` neutralises PLACEMENT only (position/inset/width/z-index/shadow).
A plugin panel is nearly always a fixed overlay pinned to a corner of the
app; alone in a 380px window that positioning is nonsense. Colours, borders,
padding, fonts and the panel's own internal layout are untouched — the whole
promise is that what you popped out is what you get.
- The element is returned to its EXACT home on dock: same parent, same position
among its siblings.
- The plugin's code still runs in the main window. So a document.body
.appendChild() inside a panel (a tooltip, a popover) lands in the main
window, not the pane — anchor to the panel instead. And a continuously
animating panel may run slowly while the main window is backgrounded, since
its rAF lives there. Both documented.
Signed-off-by: topkoa <topkoa@gmail.com>
mirrorGlobal covers the case where a pane drives a plain global that some
renderer reads each frame. It does not cover the far more common one: a
plugin whose MAIN-realm code is the authority — it clamps, it persists, it
emits events, it owns the audio graph or the camera rig — and which must
therefore APPLY the pane's values itself rather than have core splat them
somewhere.
Camera Director is the case that forced this. Its brain is the sole writer
of the camera store, the sole broadcaster on splitscreen's channel, and the
only thing that clamps an axis to its legal range. A pane cannot write
window.__h3dCamCtl behind its back without desynchronising its presets, its
persistence, and the panel's own sliders — and running the brain inside the
pane realm would make it a SECOND store writer and a second broadcaster,
racing the real one.
So: `panes.state(id)` hands the main realm the open pane's store
(get/set/all/subscribe). A plugin seeds it on `panes:opened`, subscribes,
and applies what comes back through its own API. The pane stays
realm-agnostic — it only ever touches ctx.state — and the plugin stays the
single source of truth.
For that to work, the hub now broadcasts EVERY change to the store, not just
the ones a pane asked for: it subscribes to the store on connect rather than
echoing pane-originated writes by hand. A value the plugin clamps or corrects
therefore reaches the pane window immediately, and there is exactly one path
by which state arrives in a pane — so it cannot drift.
Signed-off-by: topkoa <topkoa@gmail.com>
Three things a plugin needs before it can actually use panes.
## mirrorGlobal — the camera-director problem
The 3D highways read their free camera from a plain global,
`window.__h3dCamCtl` (highway_3d/FREECAM_BRIDGE.md), once per frame in
_resolveFreeCam(). A camera panel in the main window just writes that
object and the camera moves. A panel in a POP-OUT window cannot:
window.__h3dCamCtl there is a different object in a different realm, and
writing it moves nothing.
So a pane declares one field — `mirrorGlobal: '__h3dCamCtl'` — and
pane-mirror.js (main realm, where the renderers live) copies that pane's
state onto the global whenever it changes. highway_3d, keys_highway_3d
and drum_highway_3d are NOT modified and do not know panes exist.
The rule that makes it work: MUTATE THE OBJECT, NEVER REPLACE IT. A
renderer may be holding the reference, and swapping in a new object would
leave it reading an orphan. Keys the pane doesn't set are left alone
rather than deleted — the global may carry a renderer's own bookkeeping.
Closing the pane deliberately leaves the global as-is: closing the camera
panel should not snap the camera back to a default, which is exactly what
happens today (nobody clears __h3dCamCtl).
## Manifest-declared panes
"panes": [{ "id": "camera_director", "title": "Camera Director",
"script": "panes/camera.js", "mirrorGlobal": "__h3dCamCtl" }]
Declaring a pane beats calling panes.register() from screen.js because it
becomes openable FROM THE RAIL OR THE TRAY WITHOUT THE PLUGIN'S SCREEN
EVER HAVING BEEN VISITED — core registers a stub from the manifest and
fetches the script only when the user opens it. A pane you can only reach
by first navigating to the screen it was meant to replace is not much of a
pane.
The script sets `window.feedBackPane_<id> = { mount, unmount }`, mirroring
the existing window.feedBackViz_<id> convention, and the SAME file is what
a pop-out window loads in its own realm.
`script` is validated as a relpath under the plugin's src/ and served
through the sandboxed /api/plugins/<id>/src/ route — the containment rule
`styles` already has for assets/. Traversal, absolute paths, drive letters,
backslashes and non-.js are rejected; a bad entry is dropped with a warning
rather than failing the whole plugin, because one malformed pane should not
cost the user a working plugin.
Note the projection is written TWICE — _nav_entry() and the /api/plugins
route re-project independently — so panes had to be added to both, plus the
pending branch (a pane can be opened while its plugin is still installing
deps; the script is fetched on open, not at discovery).
## docs/plugin-panes.md
The contract, and the one rule it all hangs on: mount(root, ctx) runs in a
realm that may not have the app in it. Everything comes through ctx, or the
pane works docked and silently dies popped out.
Verified: manifest validation rejects ../.., C:\, non-.js, dupes and
missing fields while passing a good entry; /api/plugins projects panes[] for
all 20 plugins. mirrorGlobal mutates the global IN PLACE — a reference held
the way _resolveFreeCam holds it sees the change, and a renderer's own field
on that object survives — both for a local write and for a write arriving
over the channel from a pop-out realm.
pytest: 2401 passed, 8 failed — all 8 reproduce on a clean main (including
the one in tests/test_plugins.py) and are unrelated.
Signed-off-by: topkoa <topkoa@gmail.com>
Registers a `desktop` pane host at priority 20, above the browser pop-up
host (10) and the dock (0), whenever the Electron bridge exposes
feedBackDesktop.panes. A popped-out pane then gets a real BrowserWindow:
it remembers where you put it, can float above everything, minimizes to
the system tray, and appears in the tray's menu.
In a plain browser — or on an older desktop build that predates the
bridge — this file registers nothing and the browser pop-up host handles
detach exactly as before. Nothing else in the pane system changes. That
is what the host registry is for.
Two things only this realm can decide, so it owns them:
- The user closed a pane window (or it crashed). Close the pane, or the
dialog its pop-out chip hid never comes back and the user is left with
no way to reach their own UI.
- The tray asked to toggle a pane it has no window for. Main cannot know
what opening one means — the pane might belong in the dock — so it asks.
Unlike a browser pop-up, this host needs no user gesture, so it sets
autoRestore: true — a pane you left popped out comes back popped out,
where you left it, on the next launch.
Pairs with got-feedback/feedBack-desktop#103.
Signed-off-by: topkoa <topkoa@gmail.com>
A pane can now leave the main window entirely. Same `mount(root, ctx)`,
same file, different JS realm — which is what the ctx-only contract in the
previous commit was for.
## A purpose-built document, not the app shell with a flag on it
`GET /pane` serves static/panes/pane.html: the bridge, the runtime, and
the pane's own script. No highway, no library, no v3 shell, no <audio>,
no Tailwind.
The splitscreen follower takes the other road — it reloads the whole app
at `/?ssFollower=1` and hides what it doesn't want — and pays for it with
an anti-flash block that must run before any script parses (index.html),
bail-outs in app.js and shell.js, and ~40 lines of CSS hiding core
elements by id. It loads the entire app to throw it away. A pane window
has nothing to throw away, so it boots in milliseconds and there is
nothing to flash.
The cost is that `window.feedBack` in a pane realm is a deliberate,
documented SUBSET. The runtime installs exactly what a pane is promised —
`panes.register`, and the no-op chip/dock calls a shared script may make
at load — so a pane reaching for something it was never given fails
loudly at authoring time instead of subtly at runtime.
## The channel
BroadcastChannel('feedback-panes'), same origin. This works only because
Electron's setWindowOpenHandler returns `action: 'allow'` for same-origin
URLs: `deny` would push the window to the system browser, a different
Chromium instance, where BroadcastChannel cannot reach it and the pane
would silently never sync. That flag is load-bearing.
hello -> snapshot resync-on-open, always. The snapshot is the only way
the pane realm learns anything.
state main is authoritative. A pane's write is a REQUEST;
main applies it and echoes to every realm, so a
losing write self-corrects instead of splitting brain.
rpc / rpc:reply ctx.call() -> the capability bus, with a 10s deadline.
Without one, a main window that died mid-call leaves
the pane's promise pending forever.
event allowlisted bus events, JSON-safe. A CustomEvent
carrying a DOM node (highway:canvas-replaced does)
would throw on postMessage and take the channel down
for everyone, so detail is round-tripped through JSON.
stream one coalesced message per pane per frame, OVERWRITING
anything not yet flushed. Queueing would build a
backlog: Chromium throttles a backgrounded window, and
the main window is exactly what's backgrounded while
the user looks at the pane.
sub / unsub refcounts the main-realm sampler.
bye both directions.
## The follower clock
The pane extrapolates between broadcasts: anchor + observedRate * elapsed,
capped at 2s. observedRate is learned from the broadcasts themselves
(dt/dwall) so it tracks the speed slider without being told about it, and
seeks/pauses are excluded from the fit — a jump is not a tempo. Capping it
means a dead main window decays into a frozen clock rather than one that
confidently runs away. This is splitscreen's hard-won trick, generalized:
panes just call ctx.playhead().
## Failure modes, all of them
- Main window closes -> `bye {main-closed}` and the pane says so plainly,
rather than showing a frozen playhead that looks live. The host also
closes its windows outright; a pane that cannot be fed should not be on
screen.
- Pane window X'd or crashed -> a `closed` poll reaps it (a crashed
renderer never sends `bye`), the pane closes, and the chip's dialog comes
back. Without this the user's dialog stays hidden with no way back.
- Popup blocked -> a toast, and we bail BEFORE the manager records
anything, so the caller's dialog stays exactly where it was.
- Nobody answers `hello` in 5s -> the window says so instead of spinning.
- A pane with no `script` is a closure in this realm and cannot honestly
cross a window boundary. The window host declines it (canHost) and the
router falls back to the dock.
- A browser blocks window.open() outside a user gesture, so a popped-out
pane cannot be auto-restored on page load — it would only ever produce a
"blocked" toast. Such a pane comes back in the DOCK, and the chip pops it
out again on the next click. (autoRestore: false. The desktop host will
set it true.)
Hosts may now declare `remote: true`, meaning the pane's mount() runs in
another realm: the manager then owns only the authoritative state store and
never calls mount() itself. That is the seam the Electron BrowserWindow +
tray host drops into next, with no change here.
Verified: popped Now Playing and Mixer into real windows. The pane realm has
no window.highway, no capability bus and no <audio>, yet the Mixer renders
its faders via ctx.call('audio-mix','list-faders') across the channel — and
dragging that fader IN THE PANE WINDOW moved the main window's song volume
to 55 and persisted it. Closing the pane window un-hid the mixer dialog,
removed the stub and restored the chip, while the other pane window stayed
open.
Signed-off-by: topkoa <topkoa@gmail.com>
The option-heavy player UIs (mixer, camera director, viz, audio routing)
all live in the rail popovers, which are exclusive: openPopFor() closes
the last one before opening the next. You cannot watch the mixer while
riding the camera, and both vanish the moment you look at the highway.
Add `window.feedBack.panes` — a core registry for live UI that is
authored once as `mount(root, ctx)` and hosted anywhere. Panes are
non-exclusive, and they survive song switches structurally: the dock is a
body child outside every .screen, so the per-song teardown never sees it.
The adoption cost for a plugin is two calls:
feedBack.panes.register({ id, title, icon, mount, unmount });
feedBack.panes.attachChip(myExistingDialogEl, id);
attachChip injects THE standard pop-out chip. Clicking it opens the pane
and hides the plugin's dialog, leaving a stub to bring it back. Core owns
the hide/restore, so every plugin's pop-out looks and behaves the same —
which is the point. It hides via a dedicated .fb-pane-detached class, not
.hidden/[hidden], because the dialogs we attach to already toggle those.
Everything a pane may touch arrives through `ctx` — never a global. That
is what will let the same mount() run inside a pop-out window, a separate
JS realm with no window.feedBack, no window.highway and no audio graph:
ctx.call(domain, cmd, payload) -> the capability bus
ctx.on(event, fn) -> the feedBack bus (allowlisted)
ctx.subscribe(stream, fn) -> playhead / meters
ctx.state.get/set -> persisted, main realm is the only writer
ctx.playhead(), ctx.song(), ctx.toast(), ctx.close()
ctx tracks every subscription it hands out and drops them on unmount, so
a pane cannot leak listeners across a dock/undock cycle.
Streams exist because an AnalyserNode cannot cross a window boundary:
levels are reduced to numbers in the realm that owns the audio graph.
One shared rAF loop, refcounted against live subscriptions, dirty-checked
before fan-out, and stopped dead when the last pane closes.
Hosts register themselves with the manager rather than being imported by
it — the dock lands at priority 0 (the floor, always available), so the
OS pane window can drop in later without this code changing.
Ships two built-in panes: Now Playing (the reference pane — reads the bus,
a stream, and levels, and touches no globals) and Mixer (the same faders
as the rail, via ctx.call('audio-mix', ...), with the chip attached to the
real #mixer-control). Plus a "Panes" rail popover to open panes that have
no dialog of their own; the system tray will mirror that list.
Note the dock sits at z-index 110, not on the docs/plugin-v3-ui.md ladder
(transport 20, rail 30, popovers 40) — those live INSIDE #player's
stacking context, and #player is itself fixed at z-index 100. A dock below
100 is invisible on the one screen panes exist for. Body-level ladder:
#player 100 < dock 110 < toasts 120 < modals 200.
Pop-out windows, the system tray, manifest-declared panes and mirrorGlobal
(the window.__h3dCamCtl proxy the camera director needs) follow.
Signed-off-by: topkoa <topkoa@gmail.com>
* feat(career): career plugin — stars from song_stats, venue tiers, pack downloads (career mode PR2)
Bundled plugin: per-song stars from best_accuracy (60/75/85% → 1/2/3★),
cumulative stars unlock bar → club → arena (data-driven venues.json).
Venue packs (UE-rendered crowd loops) download on demand to
CONFIG_DIR/plugin_uploads/career/ on a background thread with sha256 +
zip-slip validation, served via FileResponse. Career screen (promoted
sidebar entry) shows progress and pushes the active venue's manifest
into the crowd video layer (v3VenueCrowd, PR1) — degrades cleanly when
either side is absent. Pack URLs land in venues.json in PR3.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(career): keep manifest cleanup path alive on delete; badge only for installed venues
Codex preflight: nulling _appliedManifestVenue on delete skipped
pushCrowdManifest's setManifest(null) cleanup, leaving the crowd layer on
a deleted pack; and the 'playing here' badge showed for an override venue
whose pack was removed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(career): generation-guard in-flight manifest fetches
Codex preflight: a manifest fetch resolving after a newer refresh (pack
deleted, venue switched) could re-apply a stale pack over the user's
newer selection — fetches now carry a generation token and bail when
superseded.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(career): exclude orphaned song_stats from star totals
Codex preflight: scans hide rather than delete stats of removed songs, so
stars now apply the same existing-song filter other stats surfaces use
(filename IN (SELECT filename FROM songs)).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(career): 50/150 star thresholds + star collection overview
Byron's progression tuning: club at 50★, arena at 150★. /state now
returns star_detail rows (title/artist joined from the library, stars,
best accuracy, next-star threshold) sorted closest-to-next-star first,
and the career screen renders a collection panel: tier summary plus a
per-song list with a 'N% to next star' practice hint.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(career): venue select/unselect UX, intro manifest support, fullmatch guards
- 'Play here' now also defaults the visualization to Venue (remembering
the prior viz); active venues show 'Leave venue' which restores it and
sets the '__none__' override so no installed venue silently reapplies.
- Pack manifests may ship an intro block (flyover video + ambience mp3);
files validate like loops/stingers, .mp3 added to the serving whitelist.
- Codex preflight: whitelist regexes use fullmatch (trailing-newline names
could validate but 500 on serving).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(career): let pushCrowdManifest clear the manifest on Leave venue
Codex preflight: nulling _appliedManifestVenue before refresh skipped the
setManifest(null) cleanup branch, leaving the crowd playing.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(career): refresh tailwind output
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(venue): reactive crowd video layer behind the 3D highway (career mode PR1)
Two crossfading video backdrop planes in the highway_3d venue background
style, driven by a new venue-crowd.js state machine that maps
v3:live-performance-state to crowd states (bored/neutral/engaged/ecstatic)
with 3s stability + 8s dwell hysteresis, plus one-shot reaction stingers
on streak milestones and end-of-song accuracy. Inert without a venue pack
manifest (career plugin, PR2) or the feedBack-venue-crowd-dev flag — the
static bg plate behaves exactly as before.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): retry renderer binding + preserve mid-stinger transitions
Codex preflight P2s: (1) videos created before highway_3d registered its
globals never reached the backdrop planes — binding is now idempotent and
retried from start/perf-event/re-activation paths; (2) a crowd-state
switch committing while a stinger played was dropped because the machine
had already advanced — it is now deferred and played when the stinger ends.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): per-video load tokens + unbind renderer on stop
Codex preflight round 2: (1) the global load token let a stinger cancel a
committed loop load on the other layer — tokens are now per-element, and a
stinger preempting an in-flight loop on its own layer requeues that loop
for when the stinger ends; (2) setManifest(null)/deactivate left the last
crowd frame bound and visible over the static plate — stop() now unbinds
both layers from the renderer and zeroes the mix.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): flush deferred loop on stinger failure, source accuracy from perf events
Codex preflight round 3: (1) a failed/timed-out stinger left a deferred
loop switch queued forever; the failure path now flushes it. (2)
stats:recorded only carries {filename, arrangement}, so the end-of-song
reaction now uses the accuracyPct from the song's last
v3:live-performance-state event (a real percentage) instead of a field
that never existed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): requeue mid-crossfade loops preempted by stingers; hard-stop on manifest swap
Codex preflight round 4: (1) idleLayer() still points at the fading-in
layer during a crossfade, so a stinger firing mid-fade overwrote the new
loop with nothing requeued — the fading loop is now tracked and requeued
like an in-flight load; (2) swapping venue packs while active now goes
through stop() so _stopGen invalidates the old manifest's in-flight loads.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): generation-gate stinger handlers; recrop on video size change
Codex preflight round 5: (1) an ended/timeout handler orphaned by stop()
could fire into a later stinger's lifecycle on the reused element — handlers
now detach unconditionally and carry a generation token; (2) the renderer
only re-applied cover-crop on camera aspect changes, so a src swap with a
different intrinsic size kept stale repeat/offset — it now recrops when
videoWidth/Height change.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): bail loop-fade completion when a stinger preempted the layer
Codex preflight round 6: the loop crossfade's completion callback could
still run between a stinger's start and its canplaythrough, promoting the
stinger's layer to active and pausing the real loop — it now bails when
the fading loop was preempted.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): keep rear video layer opaque during crossfades
Two half-transparent layers let the static bg plate bleed through (~25%
at mid-fade) — visible as a flash of the old still image on every state
transition. The crossfade is now always the front layer fading over an
opaque rear layer.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): reset active layer with mix on stop
Codex preflight: stop() zeroed the mix but left _activeLayer at 1, so a
restart flashed layer 0's stale frame until the new loop loaded.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): reset crowd mood to neutral on song load
Codex preflight: a song ending in ecstatic/bored left the next song's
crowd stuck in that mood until the hysteresis window passed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): cancel in-flight fade when a stinger preempts it
Codex preflight: the orphaned ramp kept pushing the mix toward the layer
whose src the stinger had just replaced.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): don't let null accuracy resets wipe the end-of-song value
Codex preflight: Number(null) is 0, so idle HUD resets overwrote
_lastAccuracyPct before stats:recorded consumed it, suppressing the
end-of-song stinger.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): abort stale stinger state on song load
Codex preflight: a stinger straddling a song change could fade back into
the previous song's layer or flush its pending loop.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): always detach load listeners, gate only the callback
Codex preflight: superseded loads left canplaythrough/error listeners
attached to the persistent video elements — unbounded growth over a
session.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(venue-crowd): flyover intro with crowd-ambience ducking
On song:loaded, an optional pack intro plays once: a camera flyover video
(idle layer, one-shot) with bar-crowd ambience audio that ducks out on
song:play, near the flyover's landing, or at handoff — whichever first.
Machine commits and stingers defer during the intro; stop()/song-change
abort it. Packs without an intro behave as before.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(venue-crowd): fall back to the loop when the intro fails to load
Codex preflight: a failed/timed-out intro left the song with no crowd
loop at all.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Follow-ups to #829 (CodeRabbit's review nit + the consumer adoption the PR
body promised):
- freqs_to_midis: reject non-finite frequencies (NaN/Infinity) — a provider
handing one through would otherwise raise inside int(round(...)) and 500
GET /api/tunings. Tests cover nan/inf/-inf alongside the existing garbage
cases.
- v3 instrument badge: TUNING_NOTE now prefers the exact integer midis the
server serves (tuningMidis) over reconstructing the note from the lowest
string's frequency via log2 against a hardcoded 440 — which can land a
semitone off at non-440 reference pitches. Frequency path kept as the
fallback for older cached responses.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MS2YFb6UUSwJVV6CmEa25i
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
* fix(nav): the library sometimes showed the legacy screen — map 'home' inside showScreen
Testers: "randomly, when moving to the library from another menu option, the library shows the
old interface — never when a song ends."
━━━ WHAT WAS ACTUALLY HAPPENING ━━━
#home is the PRE-V3 library screen. The v3 shell replaced it with #v3-songs, and the mapping DID
exist — but only inside WRAPPERS on window.showScreen, and only for callers that go through
`window`. THREE independent parties monkey-patch it, each capturing whatever happens to be there
at the time:
app.js publishes the raw function
-> shell.js wraps it, adding the home -> v3-songs mapping
-> the stems plugin wraps it AGAIN (src/main.js:1029), capturing the current value
Plugins load ASYNCHRONOUSLY. The chain links up in whatever order the race settles, and any
capture taken before shell.js installs — or any re-assignment after it — silently drops the
mapping. Hence "randomly".
AND THE INTERNAL CALLERS NEVER TOUCHED window.showScreen AT ALL. closeCurrentSong and the
Esc-from-settings shortcut call the IMPORTED showScreen, which no wrapper ever sees. Reproduced
in a browser: the unwrapped function with 'home' lands on the dead legacy screen EVERY time.
"Never when a song ends" is the tell, and it is what identified the mechanism: closeCurrentSong
resolves its target through _resolvePlayerOrigin(), which ALREADY applies this mapping. That one
path was fine — which is exactly why the bug looked random rather than total.
PRE-EXISTING, not a regression from the module carve: the onclick="showScreen('home')" links and
the wrapper-only mapping both date to 2026-06-22.
━━━ THE FIX ━━━
The guard lives inside showScreen now: ONE place, in the function every caller routes through,
instead of a chain of monkey-patches that must each remember. Wrapper order stops mattering, and
the module-internal callers are covered for the first time.
Verified in a browser: the raw, unwrapped showScreen('home') — which reproduced as #home — now
lands on #v3-songs, and cannot be undone by any wrapper order.
━━━ AND A [P1] I INTRODUCED, WHICH CODEX CAUGHT ━━━
My first cut mapped BOTH 'home' and 'v3-home', copied straight from _resolvePlayerOrigin.
That is correct THERE and wrong HERE. _resolvePlayerOrigin computes where to RETURN TO after a
song, and landing on the Songs list from the dashboard is the right behaviour. But #v3-home is
the v3 DASHBOARD — a real screen that the shell's Home nav, the onboarding tour and the dashboard
re-render listener all target. Redirecting it would have made Home unreachable.
A LEGACY ALIAS IS NOT THE SAME THING AS A RETURN TARGET. Only 'home' is mapped now, and a test
pins that: re-adding 'v3-home' to the guard fails it.
4 tests, bite-tested both ways.
node 1049, pytest 2425, ESLint 0, Codex 0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(nav): nobody may monkey-patch window.showScreen — add screen:changing, make the shell listen (#924)
window.showScreen was wrapped by THREE independent parties, each capturing whatever happened to be
there at the time:
app.js publishes the raw function
-> static/v3/shell.js wrapped it (to call syncActive, and to map home -> v3-songs)
-> the stems plugin wrapped it AGAIN (to tear down on leaving the player)
Plugins load ASYNCHRONOUSLY, so the chain linked up in whatever order the race settled. A capture
taken before shell.js installed silently dropped the mapping it carried — and the library opened on
the dead legacy #home screen. Testers saw that as "randomly, the library shows the old interface"
(#923).
#923 fixed the symptom by moving the mapping inside showScreen. This removes the CAUSE: neither
wrapper ever needed to be one.
━━━ TWO EVENTS, AND THE DISTINCTION IS THE WHOLE POINT ━━━
screen:changing emitted BEFORE anything happens. "I am leaving `from`." Teardown/cancel here.
screen:changed emitted after the DOM and data settle. "I am on `id`." Now carries `from`.
screen:changing is new, and it exists because Codex caught me collapsing the two. The stems plugin
tore down its audio graph BEFORE showScreen did anything; screen:changed fires at the very END,
after core awaits library and provider loads — so moving the plugin onto it would have delayed
teardown behind a slow fetch, or skipped it entirely if that fetch threw, and stems would keep
playing on a non-player screen. A test pins the ordering: screen:changing must precede the first
await.
shell.js is a plain screen:changed listener now, like app.js, audio-mixer.js and tour-engine.js
already were. window.showScreen is an unwrapped function again, and tests/js/
no_showscreen_monkeypatch.test.js fails CI if anything in static/ ever assigns to it again — so the
hazard is structurally impossible rather than merely avoided.
━━━ AND A FALLBACK THAT COULD NEVER FIRE ━━━
My retry-if-the-bus-is-late path listened for `slopsmith:capabilities:ready`. Core dispatches
`feedBack:capabilities:ready` (capabilities.js:1536) — the slopsmith: name is the PRE-DMCA event
and nothing has emitted it since the rename. Codex caught it. A guard that cannot fire is worse
than no guard: it reads as protection and is decoration.
(The same dead-event bug turned out to be sitting in THREE of the stems plugin's fallbacks, where
it has silently disabled its lifecycle wiring whenever the bus was late. Fixed in
feedback-plugin-stems#38.)
VERIFIED. A/B against origin/main: the nav highlight and topbar title follow IDENTICALLY with
shell.js as a listener; screen:changing -> screen:changed fire in order with the right {id, from};
window.showScreen is unwrapped; and showScreen('home') still lands on v3-songs.
node 1053, pytest 2425, ESLint 0, Codex 0.
Closes#924
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Testers: "randomly, when moving to the library from another menu option, the library shows the
old interface — never when a song ends."
━━━ WHAT WAS ACTUALLY HAPPENING ━━━
#home is the PRE-V3 library screen. The v3 shell replaced it with #v3-songs, and the mapping DID
exist — but only inside WRAPPERS on window.showScreen, and only for callers that go through
`window`. THREE independent parties monkey-patch it, each capturing whatever happens to be there
at the time:
app.js publishes the raw function
-> shell.js wraps it, adding the home -> v3-songs mapping
-> the stems plugin wraps it AGAIN (src/main.js:1029), capturing the current value
Plugins load ASYNCHRONOUSLY. The chain links up in whatever order the race settles, and any
capture taken before shell.js installs — or any re-assignment after it — silently drops the
mapping. Hence "randomly".
AND THE INTERNAL CALLERS NEVER TOUCHED window.showScreen AT ALL. closeCurrentSong and the
Esc-from-settings shortcut call the IMPORTED showScreen, which no wrapper ever sees. Reproduced
in a browser: the unwrapped function with 'home' lands on the dead legacy screen EVERY time.
"Never when a song ends" is the tell, and it is what identified the mechanism: closeCurrentSong
resolves its target through _resolvePlayerOrigin(), which ALREADY applies this mapping. That one
path was fine — which is exactly why the bug looked random rather than total.
PRE-EXISTING, not a regression from the module carve: the onclick="showScreen('home')" links and
the wrapper-only mapping both date to 2026-06-22.
━━━ THE FIX ━━━
The guard lives inside showScreen now: ONE place, in the function every caller routes through,
instead of a chain of monkey-patches that must each remember. Wrapper order stops mattering, and
the module-internal callers are covered for the first time.
Verified in a browser: the raw, unwrapped showScreen('home') — which reproduced as #home — now
lands on #v3-songs, and cannot be undone by any wrapper order.
━━━ AND A [P1] I INTRODUCED, WHICH CODEX CAUGHT ━━━
My first cut mapped BOTH 'home' and 'v3-home', copied straight from _resolvePlayerOrigin.
That is correct THERE and wrong HERE. _resolvePlayerOrigin computes where to RETURN TO after a
song, and landing on the Songs list from the dashboard is the right behaviour. But #v3-home is
the v3 DASHBOARD — a real screen that the shell's Home nav, the onboarding tour and the dashboard
re-render listener all target. Redirecting it would have made Home unreachable.
A LEGACY ALIAS IS NOT THE SAME THING AS A RETURN TARGET. Only 'home' is mapped now, and a test
pins that: re-adding 'v3-home' to the guard fails it.
4 tests, bite-tested both ways.
node 1049, pytest 2425, ESLint 0, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
19 declarations + 23 TOP-LEVEL STATEMENTS. 922 lines. app.js 3,243 -> 2,325 (-28%).
The panel registry, both global keydown dispatchers, the library arrow-nav, and the whole
plugin-facing shortcut API.
━━━ MOST OF THIS SUBSYSTEM WAS NOT DECLARATIONS ━━━
A declaration-seeded dependency closure reports this cluster as 10 names, 246 lines.
It is 42 statements and 922.
window.registerShortcut, createShortcutPanel, getAllShortcuts, unregisterShortcut,
clearWindowShortcuts, the panel registry, and BOTH global keydown dispatchers are bare TOP-LEVEL
STATEMENTS at app.js's top level. A call-graph scan sees NONE of them.
That blind spot has now cost three times:
* it nearly shipped a dead library A-Z rail (#896) — 43 of library.js's exports were
referenced only from app.js's window contract;
* it threw "Assignment to constant variable" in the session carve (#921), where the autoplay
gate's top-level statements wrote state that had just become a read-only import;
* and here it under-reported the slice by 3x.
The extractor takes them by construction now — any top-level statement that TOUCHES a moved
binding comes along — and the SEED is closed to a FIXED POINT, because those statements have
their own dependencies (_modifiersMatch, _isShortcutActive, _handleLibArrowNav, _gridColumns…)
that the declaration closure never walked. Seed -> pull the statements -> the statements need
more names -> re-seed. Iterate until it stops growing.
━━━ syncLibrarySong GOES ACROSS THE SEAM, NOT THROUGH AN IMPORT ━━━
The library arrow-nav calls it on Enter. It cannot be imported: syncLibrarySong reaches
showScreen/playSong, and a module importing app.js closes a cycle. It is the ONE name here that
had to stay behind, so it comes across the host seam — which is exactly what the seam is for.
host.js throws loudly if the wiring is ever dropped, and tests/js/host_contract.test.js fails in
CI if the hook drifts.
VERIFIED. A/B against origin/main in two browsers, IDENTICAL, zero page errors — and driven for
real, not merely present: the plugin API (register / unregister / getAll / panels), THE GLOBAL
KEYDOWN DISPATCHER actually firing a registered shortcut, that same shortcut correctly SUPPRESSED
while typing in a text input, and `?` opening the help modal.
node 1045, pytest 2425, ESLint 0 (no-cycle clean), host contract 2/2, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
36 declarations + the 4 autoplay/auto-exit gate statements. 359 lines.
app.js 3,772 -> 3,242. Bodies VERBATIM.
━━━ THIS WAS "THE UNCUTTABLE HEART", AND IT IS 359 LINES ━━━
At the start of this epic, seeding a dependency closure from count-in, from loops, from
section-practice, or from the JUCE seek shim all returned the SAME 178-function, 3,360-line set.
playSong and showScreen called each other; everything called them; nothing could be cut anywhere.
The conclusion — correct at the time — was that NO closure-based carve could touch it at any
seed, and the answer was a host seam.
That was true THEN. Every slice taken out since (transport, loops, count-in, section-practice,
the library, the edit modal, settings) removed edges, and the strongly-connected component
DISSOLVED. This closure is 36 declarations with an interface width of FOUR.
The lesson is not that the seam was wrong — the seam is what MADE this possible, by letting the
carves proceed against a cyclic core instead of stalling on it. The lesson is to RE-MEASURE. An
SCC is a fact about a graph at a moment, not a property of the code.
━━━ THE BUG NO SCAN COULD SEE, AND THE A/B DID ━━━
First cut passed every gate — no-undef clean, no-cycle clean, 1045/1045, pytest green — and
THREW IN THE BROWSER: "Assignment to constant variable."
window.feedBack.holdAutoplay / holdAutoExit and their two event handlers are TOP-LEVEL
STATEMENTS, not declarations. They WRITE this cluster's state (_autoplayHeld, _autoExitTimer, …),
and an imported binding is READ-ONLY — so left behind in app.js, every one threw the instant the
module existed.
A dependency scan that walks DECLARATIONS cannot see them. Mine didn't. This is the same blind
spot that nearly shipped a dead library A-Z rail (#896): app.js keeps its public API in top-level
statements, and a call-graph is blind to every one of them.
The extractor now finds them by construction — any top-level statement that WRITES a moved
binding comes with the carve — and the gate statements live beside the machinery they drive,
which is where they belonged anyway.
━━━ ZERO OUTSIDE WRITES, BY MOVING THE BOUNDARY RATHER THAN BUILDING MACHINERY ━━━
The autoplay scalars and the wake-lock state were written from outside the cluster, which would
have forced a setter or a state container. But the writers — _releaseAutoplay, _acquireWakeLock —
plainly belong here. Pulling them in left ZERO outside writes, so every export is a plain import.
Same move as settings (#920): measure the writers before you reach for a container.
VERIFIED. A/B against origin/main in two browsers, IDENTICAL, zero page errors — including the
autoplay gate driven end to end: a plugin HOLDS autoplay, the song loads but does not start, the
RELEASE fires it, and a stale release is a no-op. That is the exact machinery that was throwing.
node 1045, pytest 2425, ESLint 0 (no-cycle clean), host contract 2/2, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
22 declarations, 446 lines. app.js 4,218 -> 3,772. Bodies VERBATIM.
Settings load/save, the AV-offset nudge, the default-arrangement pin, the instrument pathway,
and the app-update channel.
━━━ INTERFACE WIDTH 1, AND IT GOT THERE BY DRAWING THE BOUNDARY IN THE RIGHT PLACE ━━━
app.js calls loadSettings() and nothing else.
The first cut was NOT clean: _defaultArrangement was written from OUTSIDE the cluster, and an
imported binding is READ-ONLY, so that one write would have forced a setter or a state
container — as it did for the player (player-state.js) and the library (library-state.js).
But the writers were saveSettings and pinCurrentArrangementDefault, which ARE settings
functions. Widening the slice to include them left ZERO outside writes. Every export is now a
plain read-only import and no container is needed.
Worth naming, because I reached for a container twice before: the fix for "this binding is
written from outside" is sometimes a container, and sometimes it just means the boundary is in
the wrong place. Measure the writers before you build machinery.
━━━ handleSliderInput STAYS A HOST HOOK, DELIBERATELY ━━━
It lives in settings now (it is a settings control), but player-controls.js must NOT import it:
this module already imports player-controls (_applyMastery, _autoplayExitEnabled, …), so a
direct back-import would close a cycle. player-controls keeps reading it through the host seam,
and app.js — the root, which imports both — wires it. That is exactly what the seam is for, and
the contract test proves the wiring survived.
VERIFIED. A/B against origin/main in two browsers: the window contract, the settings screen
rendering, the AV-offset and default-arrangement controls present, and a real `input` event
dispatched on a slider — which is the path that goes through the host seam. IDENTICAL, zero
page errors.
node 1045, pytest 2425, ESLint 0 (no-cycle clean), host contract 2/2, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
4 functions, 234 lines. app.js 4,452 -> 4,218. Bodies VERBATIM.
INTERFACE WIDTH ZERO — nothing in app.js calls into this cluster. app.js needs only the names
on the window contract, so the markup's onclick= handlers resolve. That is what makes it the
cleanest slice left.
AND IT ONLY BECAME CLEAN BECAUSE THE LIBRARY CAME OUT FIRST (#896). Every dependency the modal
has is a module now: it reads six bindings out of ./library.js (loadLibrary, loadFavorites,
loadTreeView, _removeLibCardsForFilename, libView, _lastLibSelected) plus dom.js and the L
container. Before that carve, extracting this would have dragged the whole library with it.
Checked, and it matters: the modal never WRITES any of those six. An imported binding is
READ-ONLY, so a single write would have forced a setter or a state container. Every use is a
read, so plain imports suffice.
Acyclic: edit-modal -> { dom, library-state, library }, and library imports none of them back.
VERIFIED. A/B against origin/main in two browsers: the window contract, the modal actually
OPENING off a real library row, its title and year fields rendering, and the data-edit-save
wiring (rather than an inline onclick embedding the filename — the fix this cluster's harness
exists to guard). IDENTICAL, no new page errors.
node 1045, pytest 2425, ESLint 0 (no-cycle clean), host contract 2/2, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The runtime stylesheet moves to CONFIG_DIR. static/tailwind.min.css is never written again.
━━━ TWO DIFFERENT THINGS WERE SHARING ONE PATH ━━━
static/tailwind.min.css a BUILD ARTEFACT. Committed, image-baked, generated by scanning
the in-tree plugins only. CI's tailwind-fresh check verifies it.
the RUNTIME sheet PER-INSTALL STATE. Additionally scans whatever the user installed
into FEEDBACK_PLUGINS_DIR, so it differs machine to machine.
Writing the second over the first meant that MERELY RUNNING THE DEV SERVER from a git checkout
silently modified a tracked file. `git add -A` then swept a 100KB reshuffle of minified CSS
into the commit and ci/tailwind-fresh went red with a diff that explains nothing — on a PR
whose real change touched no Tailwind classes at all. It also wrote app state into the app
directory, which is read-only in some deploys.
A new route serves the runtime sheet when there is one and falls back to the committed one
otherwise. It is registered BEFORE the /static mount, which would otherwise swallow the path.
━━━ A PERSISTED SHEET MUST NOT OUTLIVE ITS REASON (Codex [P2] x2) ━━━
1. THE USER REMOVES THEIR PLUGINS. Startup only rebuilds when user plugins exist, so nothing
would ever overwrite the stale sheet — and it still carries classes for plugins that are
gone. With no user plugins the COMMITTED sheet is complete by definition. Guarded.
2. THE APP IS UPGRADED, and my first guard for this was WRONG. I compared mtimes. Codex: that
is not a freshness signal across install methods — archives and container images routinely
PRESERVE SOURCE MTIMES, so a just-shipped stylesheet can carry an OLDER timestamp than a
runtime sheet a user built days ago. The mtime check then calls the stale one FRESH and it
masks the new core CSS indefinitely — permanently, if no Tailwind toolchain is present to
trigger a rebuild.
Freshness is decided by CONTENT now. Each runtime build stamps a sidecar with the sha256 of
the committed sheet it was made from. Core ships new CSS -> that file changes -> the hash
changes -> the runtime sheet is correctly judged stale. Timestamps only gesture at the
question that hashing answers.
Falling back to the committed sheet is always safe: at worst it lacks a just-installed plugin's
classes for the seconds until the async rebuild lands.
VERIFIED END TO END. Ran the real dev server with 3 plugins installed: it rebuilt Tailwind over
them (123,291 bytes), wrote the sheet + sidecar to CONFIG_DIR, still served /static/
tailwind.min.css at 200 — and `git diff` on the tracked file came back CLEAN.
8 tests. Bite-tested: reverting to the shared path fails 3, dropping the staleness guards fails
2 more.
pytest 2425, pyflakes 0, Codex 0.
Closes#911
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
18 functions, 1,245 lines. highway.js 3,972 -> 2,727 (-31%). The biggest R3c slice: notes,
sustains, chords, strum groups, unison bends and lyrics — everything the default renderer
paints each frame.
━━━ MUTABILITY, NOT LOCATION, DECIDES WHERE A THING BELONGS ━━━
Three per-instance caches came out with this slice, and they are why it needed care:
_frameMismatchWarned a warn-once Set of chord ids (feedBack#88)
_chordRenderInfo a WeakMap of chord -> chain info
_lyricMeasureCache Map<fontSize, Map<text, width>>
All three are MUTATED. Left at module scope they would be SHARED ACROSS PANELS — one
highway's lyric widths and chord chains stomping another's, silently, with nothing throwing.
createHighway() is a factory (the constitution publishes window.createHighway so a plugin can
build a second highway), so they are lifted onto hwState, which is exactly what hwState is for.
The shimmer LUT went the OTHER way — to MODULE scope in highway-geometry.js. It is a
deterministic xorshift table, byte-for-byte identical for every instance, so sharing it is not
merely safe but BETTER: built once for the page rather than once per panel.
Same slice, opposite directions, decided entirely by whether the thing mutates.
━━━ MY SCRIPT WAS WRONG TWICE. THE GATES CAUGHT BOTH. ━━━
1. HAND-LISTED THE MOVE SET. I listed 10 functions and missed six that drawChords needs
(_ensureChordRenderCache, bsearchChords, getChordTemplateInfo, _computeChordBox,
_updateFretLinePreview, _drawFretLineChordPreview). The no-undef gate named every one. The
set is now DERIVED from the dependency closure — 18, not 10.
2. JUDGED PURITY TOO EARLY, and this one is subtle. I classified _computeChordBox as pure
because its ORIGINAL body never mentions hwState. Then the call-site rewriter injected
`fretX(hwState, …)` INTO it — fretX takes hwState now (#916) — leaving a function that
references an hwState it was never given. Purity has to be judged from the body AS IT WILL
BE, so the classifier iterates to a fixed point: a function needs hwState if it mentions it,
OR calls anything that now takes it. That moved _computeChordBox to the stateful side.
VERIFIED. A/B against origin/main: IDENTICAL, zero page errors. The PLUGIN BUNDLE contract is
byte-identical (b.fretX arity 3, b.getNoteState arity 2, both stable references, both correct
under the old calling convention). PERF GATE PASSES AT 1.92ms against its 12ms budget — and
this is the slice that could really have cost something: the ENTIRE per-frame drawing path is
now cross-module. It costs nothing measurable.
TESTS. highway_teaching_marks follows strumGroupBuckets to the new module. The two source-shape
harnesses now read highway.js AND every static/js/highway-*.js, rather than being re-pinned at
whichever file currently holds a function — re-pinning breaks again next time, and a shape
assertion that silently stops finding its target is indistinguishable from one that passes.
node 1045, pytest 2416, ESLint 0, no-undef 0, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
fretX, fillTextReadable, _noteState, _paintGemGlow -> static/js/highway-state-primitives.js.
50 call sites rewritten. highway.js 4,105 -> 3,965.
The first slice that changes signatures. Each of these four gains hwState as an explicit
FIRST PARAMETER.
━━━ hwState IS A PARAMETER, NOT AN IMPORT ━━━
createHighway() is a FACTORY. The constitution publishes window.createHighway so a plugin can
build a SECOND highway for its own panel, and highway.js says so itself. Import hwState as a
module singleton and two panels silently share one clock, one render scale, one string
palette — each driving the other. Nothing throws. The picture is just wrong, in a way no test
would catch.
(The exact opposite of the app.js carve, where player-state.js and library-state.js ARE
module singletons — correctly, because there is exactly one app. Same epic, same language,
opposite answer, decided entirely by whether the thing is a factory.)
━━━ THE PLUGIN BUNDLE NEARLY BROKE, SILENTLY ━━━
The renderer bundle hands two of these STRAIGHT TO PLUGINS:
b.fretX = fretX;
b.getNoteState = _noteState; // stable reference
highway_3d calls both EVERY FRAME, with the old arity. Handing out the new 3-arg versions
would have passed `note` where hwState belongs — no throw, no error, just wrong geometry and
wrong judgment state INSIDE A PLUGIN, which no core test would ever see. Green CI, broken 3D
highway.
So hwState is bound ONCE per instance, in the factory, and the bundle hands out those views.
A per-frame arrow would have fixed the arity and reintroduced exactly the per-frame allocation
the bundle's stable-reference contract (feedBack#254) exists to prevent. b.project needs none
of this — project() is pure and its arity never changed.
VERIFIED IN A BROWSER, against the real bundle, on both builds:
fretX arity 3 3 (NOT 4 — the bound view preserves it)
getNoteState arity 2 2
fretX(5,1,800) in 0..800 True True
getNoteState null w/o provider True True
getNoteState honours provider True True
fretX is a stable reference True True
IDENTICAL. Without the bound views fretX would have reported arity 4 and computed garbage.
Also caught on the way: my generated module imported STRING_BRIGHT_FALLBACK, a name
highway-constants.js does not export. ESLint does not flag that — but importing a name a
module does not export is a runtime SyntaxError that kills the WHOLE module. These four need
no constants at all; the import is gone.
TESTS. highway_note_state pins the signature AND the stable-reference contract — it caught the
bundle break. Retargeted at the module and the new arity; both contracts still asserted, and
the "no fresh arrow per frame" rule is now asserted explicitly rather than implied by
`getNoteState: _noteState`.
PERF GATE PASSES AT 1.94ms against its 12ms budget — fretX and _noteState are now CROSS-MODULE
calls, per note, per frame. It costs nothing measurable.
node 1045, pytest 2416, ESLint 0, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
6 functions, 53 lines. highway.js 4,158 -> 4,105. NOT ONE CALL SITE CHANGES.
project, roundRect, bnvNormalizedPoints, teachingFingerLabel, teachingDegreeLabel,
chordHarmonyLabels — the shared primitives every drawing function leans on.
━━━ PURITY IS THE WHOLE POINT OF THIS SLICE ━━━
Every one of these is a pure function of its arguments. None touches hwState. None closes over
the canvas context — roundRect() already took `ctx` explicitly, and the rest need nothing but
numbers. project() reads only the module-level constants from #914.
That matters because createHighway() is a FACTORY: a plugin can build a second highway for its
own panel, so anything holding per-instance state must be PASSED hwState rather than importing
it, or two panels silently share one clock and palette. These six hold no state at all, so
they move VERBATIM — the module boundary is invisible to every caller.
The asserts are mechanical and in the extractor: it REFUSES to move a function whose body
mentions hwState, or that references `ctx` without taking it as a parameter. Purity is
checked, not assumed.
━━━ WHAT IS DELIBERATELY LEFT BEHIND ━━━
The four primitives that DO need hwState — fretX, fillTextReadable, _noteState, _paintGemGlow
— stay in the factory for now. They need an explicit hwState parameter threaded through 53
call sites, which is a real behavioural change and belongs in its own commit rather than
smuggled in beside a provably-identical move. Separating the provable from the risky is the
whole discipline of this epic.
TESTS. Three harnesses brace-match these functions out of the source and run them in a
sandbox; they now read static/js/highway-geometry.js. `export function x` still contains
`function x`, so the extractor needed no change — only the path.
VERIFIED. A/B against origin/main: 15 probes IDENTICAL, zero page errors. PERF GATE PASSES AT
1.91ms against its 12ms budget — and this is the one that could plausibly have cost something:
project() runs for every visible note on every frame and is now a CROSS-MODULE call. It costs
nothing measurable. That is the answer #910 was built to give.
node 1045, pytest 2416, ESLint 0, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
29 constants, 190 lines. highway.js 4,267 -> 4,158. The first real slice, and the one that
every later one imports.
━━━ WHY ONLY THE CONSTANTS MAY LIVE AT MODULE SCOPE ━━━
createHighway() is a FACTORY, not a singleton. The constitution publishes
window.createHighway precisely so a plugin can build a SECOND highway for its own panel, and
highway.js already says so at the top of the closure:
// R3c: per-instance mutable state in one object, so extracted renderer/ws
// modules can close over it as a factory arg without cross-panel sharing.
So hwState — all 79 mutable properties — must NEVER become a module-level singleton: two
highways would silently share it, and one panel would drive the other's clock, scale and
colour tables. Extracted functions will take it as an ARGUMENT.
That is the OPPOSITE of the app.js carve, where a single state container (player-state.js,
library-state.js) was exactly right, because there is exactly one app. Same epic, same
language, opposite answer — because one is a singleton and the other is a factory.
These 29 are pure literals: numbers, strings and colour tables, never reassigned, never
mutated. Sharing them across instances is not merely safe, it is what you want — one copy of
the shimmer LUT bounds and the string palettes rather than one per panel. Anything with a
runtime dependency (document, window, performance, localStorage) stays in the factory;
checked, and none of these has one.
ESLint now knows static/highway.js is a module. It could not have known before this commit:
the flip (#913) changed the SCRIPT TAG, but the file had no import/export yet, so it still
parsed as a script and lint stayed green. The first `import` is what makes the config wrong.
TESTS. Four source-shape harnesses asserted `const _AUTO_SCALE_MIN = …` etc. lived in
highway.js. They now read highway.js AND every static/js/highway-*.js — deliberately, rather
than being re-pinned at whichever file currently holds a constant. Re-pinning just breaks
again on the next carve, and a source-shape assertion that silently stops finding its target
is indistinguishable from one that passes. Bite-tested: renaming two constants away fails
them.
VERIFIED. A/B against origin/main: 15 probes IDENTICAL, zero page errors. AND THE PERF GATE
PASSES AT 1.97ms against its 12ms budget — which is the point of having built it (#910)
first: these constants moved from closure scope to module scope, and V8 does not treat those
identically. It does here. Now I know rather than hope.
node 1045, pytest 2416, ESLint 0, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two lines. index.html: defer -> type="module". highway.js: one explicit assignment.
highway.js can now `import`, which is the whole point — the carve can begin.
━━━ THE ONE THING THE FLIP ACTUALLY BREAKS: window.createHighway ━━━
A top-level `function createHighway()` in a CLASSIC script IMPLICITLY becomes
window.createHighway. In a module it does not — module declarations are module-scoped, and
the name vanishes from the global object the instant the tag grows type="module".
The constitution names window.createHighway as PUBLIC EXTENSION CONTRACT (alongside
window.playSong / showScreen / feedBack). NOTHING IN-TREE CALLS IT. That is exactly why this
would have shipped: the only consumers are third-party plugins rendering their own highway
panel, and I cannot grep those. Green CI, green tests, and a broken plugin API.
Verified by removing the assignment and reloading:
flip WITHOUT an explicit assignment: window.createHighway === undefined <-- gone
flip WITH it: window.createHighway === function
So it is assigned explicitly now — same object, same behaviour, no longer an accident of how
the file happens to be loaded.
━━━ AND A CORRECTION TO #912 ━━━
#912 (merged) rewrote 73 bare `highway.x` -> `window.highway.x` on the stated grounds that
the flip would turn every one of them into a ReferenceError. HAVING NOW ACTUALLY FLIPPED IT,
THAT WAS WRONG. highway.js already did `window.highway = highway`, which puts the name on the
GLOBAL OBJECT — and bare-identifier resolution falls back to the global object whether or not
a lexical global binding exists. Measured on both builds: bare `highway` resolves either way.
#912 is defensible as hygiene and it does not hurt, but it was not a precondition and it fixed
no latent bug. A correction is posted on the PR so its commit message does not mislead. The
real hazard was the factory, not the instance — same class of breakage, wrong name.
ORDERING is unchanged: classic-defer and non-async type="module" share ONE post-parse
execution queue, in document order, so highway.js keeps its position at index.html:1244.
VERIFIED. A/B against origin/main: 15 probes IDENTICAL, zero page errors — window.highway,
window.createHighway, the full API surface, a real song playing, the chart clock advancing,
and the seek->setTime sync. THE PERF GATE PASSES at 1.85ms against its 12ms budget (module
evaluation costs nothing at render time), which is exactly what #910 was built to tell me.
node 1045, pytest 2416, ESLint 0, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
73 bare `highway.x` references -> `window.highway.x`, across app.js and 10 other files.
Provably a NO-OP today. It is the precondition for flipping highway.js to a module.
━━━ WHY THIS HAS TO LAND FIRST ━━━
highway.js is a CLASSIC script. Its top-level `const highway = createHighway()` therefore
creates a GLOBAL LEXICAL BINDING — visible as a bare name to every other classic script AND
to every ES module. 73 call sites quietly rely on that.
The moment highway.js becomes a module, that binding is gone. `const` in a module is
module-scoped, not global. Every one of those 73 sites becomes a ReferenceError, and the
flip is impossible until they say what they mean.
`window.highway = highway` is already set, to the same object, on the same line. So this is
an identity rewrite — verified in the browser below.
━━━ THE REWRITE BIT ME THREE TIMES. REGEX IS NOT ENOUGH FOR THIS. ━━━
1. A SHADOWED LOCAL. capabilities/note-detection.js does `const highway = window.highway`.
Its 9 bare uses are LOCAL and already correct; a blind rewrite would have emitted
`const window.highway = window.highway`. Excluded.
2. HALF-CONVERTED GUARDS — the dangerous one. Six sites read
`typeof highway !== 'undefined' && highway && typeof highway.setTime === 'function'`.
The regex converted the CONSEQUENT and left the TEST, which is WORSE than not touching
them: after the flip `typeof highway` is 'undefined', so each guard is PERMANENTLY FALSE
and the code behind it silently never runs. transport.js's was the seek->setTime sync:
the chart clock would have quietly desynced after every seek, with nothing failing.
All six now test window.highway.
3. TWO MORE BARE REFERENCES, found by Codex [P2] and confirmed by an AST scan: app.js:3114
and :3176 use `highway && typeof window.highway.getSections === 'function'`. My grep
searched for `typeof highway`, not `highway &&`. After the flip these throw, the catch
swallows it, and the editor silently falls back to a ±4s edit window and arrangement 0.
Regex missed a shadow, a half-conversion, and two bare reads. The final check is an
AST pass that resolves scopes and reports every `highway` identifier not bound locally.
It now reports ZERO.
VERIFIED. A/B against origin/main in two browsers, 15 probes, IDENTICAL, zero page errors:
window.highway is the same object as the bare global, the whole API surface resolves, a real
song plays, the chart clock advances, getPerf().drawMs > 0 — and `seek syncs chart` passes,
which is the exact guard I nearly broke in (2).
node 1045, pytest 2416, ESLint 0, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
highway.getPerf() (additive) + tests/browser/highway-perf-baseline.spec.ts.
No behaviour change. This lands BEFORE highway.js is touched, because a perf-gated refactor
without a perf gate is just a refactor.
━━━ FRAME RATE IS THE WRONG THING TO MEASURE ━━━
The highway AUTO-SCALES. When the smoothed draw cost passes _DRAW_BUDGET_HI_MS (12ms) it
LOWERS THE RENDER RESOLUTION to protect the frame rate (#654). Exactly right for players —
and it means a real perf regression does NOT show up as dropped frames. It shows up as a
BLURRIER PICTURE at a perfectly healthy 60fps.
Benchmark fps and you measure the feedback loop, not the renderer, and conclude nothing
changed while the image quietly degrades.
So the gate pins the scale (setRenderScale(1) + setMinRenderScale(1), which clamps autoScale
to [1,1]) and measures drawMs — the renderer's own cost. None of that was reachable before:
neither drawMs nor the effective scale escaped the closure. Hence getPerf().
The threshold is the app's OWN: _DRAW_BUDGET_HI_MS is the cost at which the highway itself
starts sacrificing resolution in production. Exceeding it is not an arbitrary benchmark line
— it is the renderer failing its own budget. Current cost ~2.2ms, so ~5x headroom: far more
than headless-CI variance, far less than any regression worth shipping.
━━━ I WROTE THIS GATE WRONG THREE TIMES. EACH TIME IT PASSED. ━━━
1. VACUOUS ASSERTION. First cut asserted "the auto-scaler wasn't forced to intervene", i.e.
effectiveScale == 1. I injected a 10x regression (drawMs 2.4 -> 22.4ms, nearly DOUBLE the
budget) and it PASSED. Of course it did: setMinRenderScale(1) sets the scaler's FLOOR to
1, so effectiveScale CANNOT drop below it. The very pinning that stops the scaler hiding
a regression also stops it ever reporting one. A guard that cannot fail.
2. MEASURING AN IDLE RENDERER (Codex [P2]). playSong() takes ~3-4s to actually start — it is
fetching and decoding stems. My "if not playing after 2s, togglePlay()" fired BEFORE
autoplay, started playback, and then the app's own autoplay toggled it straight back to
PAUSED. The renderer idled through the entire measurement. Now it WAITS for playback
rather than racing it, and asserts the chart clock advanced DURING the sampling window —
not merely at some point beforehand, which the first fix would have accepted.
3. UNENCODED FILENAME (Codex [P2]). playSong() decodes its argument before building the
/ws/highway path, so every real caller passes encodeURIComponent(filename)
(app.js:2879, 4137). Raw, a name containing # ? % or / yields an invalid WebSocket URL,
the song never loads — and on those libraries the gate would have silently measured an
idle renderer instead of failing.
Every one of those bugs made the gate PASS. That is the whole hazard of a perf test: it
fails safe in the wrong direction.
BITE-TESTED, and this is the only reason I trust it: a 10x regression injected into the draw
path FAILS the gate under live playback (22.0ms vs the 12ms budget) and the clean build
passes at ~2.1ms with the chart clock advancing 5.6s across the sample.
node 1045, pytest 2412, ESLint 0, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The guard in startup_events() read:
if getenv_compat("FEEDBACK_DEMO_MODE") or getenv_compat("FEEDBACK_DEMO_MODE") == "1" \
and not _DEMO_JANITOR_STARTED:
`and` binds tighter than `or`, so that is `A or (B and C)`. The not-already-started half
never ran when the env var was truthy — the only case that reaches it at all. A second
startup started a SECOND janitor thread, overwrote the handle, and shutdown then joined
only the last: the first leaked and kept firing registered hooks hourly, forever.
The guard now lives INSIDE start_janitor(). A caller cannot get operator precedence wrong
if there is nothing left for it to get wrong.
━━━ THREE WAYS TO WRITE THIS GUARD WRONG. I HIT ALL THREE. ━━━
1. NO GUARD — the original bug. Double-start, orphaned thread.
2. GUARD ON THE FLAG (`if _DEMO_JANITOR_STARTED: return`). Codex [P2]. stop_janitor()
DELIBERATELY leaves that flag True when a hook outruns its join timeout, so that a later
startup cannot spawn a janitor beside a live one. But the hook usually finishes a moment
later: the thread exits and the flag is stale. A flag-keyed guard then refuses to start a
replacement for the rest of the process — demo cleanup silently dead. (The original bug
accidentally MASKED this by always starting.)
3. GUARD ON LIVENESS ALONE (`if thread.is_alive(): return`). Codex [P2], second pass. A
timed-out stop leaves the old thread ALIVE BUT DOOMED — its stop event is set and it
exits as soon as its current hook returns. Treating that as a running janitor skips the
replacement, and we are back at (2) a second later.
So: a janitor counts as running only if its thread is alive AND it has not been told to stop.
━━━ AND EACH JANITOR NOW OWNS ITS STOP EVENT ━━━
start_janitor() used to `_DEMO_JANITOR_STOP.clear()` a single SHARED Event. Start a
replacement while a doomed thread is still finishing a hook and that clear RESURRECTS it: it
loops back to wait(), sees the flag cleared, and carries on. Two janitors — the exact bug we
started from. A fresh Event per janitor makes it impossible; the old thread waits on its own
event, which stays set, so it can only exit.
Env semantics UNCHANGED, verified across every value ("", "1", "0", "true", "false", "off"):
the old expression and demo_mode_enabled() agree on all of them. The only behavioural change
is the idempotency fix.
FOUR tests, and each of the three wrong guards fails a different subset:
no guard -> 2 fail (double start; orphaned thread)
guard on the flag -> 2 fail (never restarts after a timed-out stop)
liveness alone -> 1 fail (no replacement for a doomed janitor)
liveness + not-stopping -> all pass
pytest 2416, pyflakes 0, Codex 0.
Closes#902
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Chromium treats a getDisplayMedia audio track as a voice call by
default: echo cancellation, noise suppression, auto gain control and
mono downmix. Music through that pipeline is the tester-reported
"tin can" sound on ASIO/exclusive outputs.
Request the raw path explicitly (EC/NS/AGC off, stereo, 48 kHz) —
all constraints are best-effort so unsupported ones degrade silently
instead of failing the capture. The [asio-diag] loopback line now
dumps track.getSettings() so logs prove which processing actually
applied.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
lib/demo_mode.py (342). server.py 1,870 -> 1,649.
The read-only request guard (its 96-entry blocked-route table + the middleware) and the
hourly session janitor (registry, hook runner, thread). Bodies VERBATIM.
THE MIDDLEWARE NEEDS `app`, SO THE MODULE TAKES IT. _demo_mode_guard is an
@app.middleware("http") and cannot exist without an app object. Rather than have a module
under lib/ reach for a global, it exposes install(app) and server.py — which owns the app —
hands it over. The janitor is symmetrical: start_janitor() / stop_janitor(), called from
server.py's startup and shutdown hooks, where the process lifecycle actually lives.
register_demo_janitor_hook IS PART OF THE PLUGIN CONTRACT. It is a key in plugin_context,
so plugins hold it as a LIVE REFERENCE from setup(). server.py imports this exact object
and puts it in the dict unchanged — identity preserved, and
tests/test_plugin_context_contract.py (#898, merged) fails if that ever stops being true.
This is the first carve that guard has actually protected.
━━━ stop_janitor()'s ORDER IS LOAD-BEARING ━━━
The obvious way to write it — clear the "started" flag, then join — is WRONG, and I wrote
it that way first. server.py's original deliberately returns EARLY, leaving
_DEMO_JANITOR_STARTED True and the thread handle intact, when the thread outlives the join:
# Leave _DEMO_JANITOR_STARTED True so a new janitor is not
# spawned by a subsequent startup while the old one is alive.
Clearing the flag first quietly reintroduces exactly the double-janitor leak the flag
exists to prevent. Preserved byte-for-byte, and the reason is now written down at the
function rather than only at its single call site.
━━━ A BUG MOVED VERBATIM, ON PURPOSE ━━━
if getenv_compat("FEEDBACK_DEMO_MODE") or getenv_compat("FEEDBACK_DEMO_MODE") == "1" \
and not _DEMO_JANITOR_STARTED:
`and` binds tighter than `or`, so this is `A or (B and C)` — the not-already-started
re-entry guard is DEAD whenever the env var is truthy, which is the only case that runs.
A second startup leaks a janitor thread (the handle is overwritten, so shutdown joins only
the last). Verified. Preserved exactly and filed as issue #902: a carve whose whole value
is being provably behaviour-neutral is not the place to change behaviour.
pyflakes caught three more missing imports on the way in (uuid, warnings x2). Five carves,
ten missing imports, every one a NameError on a live path.
pytest 2399, pyflakes 0, Codex 0.
Refs #48
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
lib/scan.py (326). server.py 2,098 -> 1,870.
The background scan, its spawn ProcessPoolExecutor, and the kick/runner plumbing that
serialises passes. Bodies VERBATIM except the seam reads.
Everything shared is read LATE off appstate — the same contract every module in
lib/routers/ uses, and it is not cosmetic: tests monkeypatch CONFIG_DIR and swap meta_db,
so a value captured at import time pins the wrong one for the life of the process.
CONFIG_DIR -> appstate.config_dir
meta_db -> appstate.meta_db
_default_settings -> appstate.default_settings()
_stat_for_cache -> appstate.stat_for_cache()
━━━ THE SCAN STATUS IS REBOUND, NOT MUTATED ━━━
_background_scan does `global _scan_status; _scan_status = {**INIT, ...}` at every stage
transition. It REPLACES the dict; it never updates it in place. So nothing may hold that
dict by value — a reference captured once goes permanently stale at the first stage change
and would report "listing" forever while the scan ran to completion.
Hence `scan.status()`, a getter, and hence appstate publishes scan_status as a CALLABLE.
appstate.py already said so in a comment; this is the code that makes it true. (Same for
the plugin_context entry, which was already `lambda: dict(_scan_status)` — late-bound, so
it survives the move unchanged. The contract test from #898 covers it.)
━━━ appstate.server_root: A TRAP CLOSED PERMANENTLY ━━━
_background_scan seeds the builtin content, which needs the directory holding server.py.
`Path(__file__).resolve().parent` is correct in server.py and silently WRONG anywhere under
lib/ — it yields lib/, which holds no docs/ or data/ — and it fails by finding NOTHING
rather than by raising, so the seeds would just quietly never run.
lib/builtin_content.py (#900) closed that by taking the root as a parameter. This adds the
other half: server.py publishes it ONCE as appstate.server_root, so no module under lib/
ever has a reason to derive it. Documented at the slot.
pyflakes caught two more missing imports on the way in (loosefolder_mod, enrichment) —
each a NameError on a live scan path, and the suite would have handed them over one failure
at a time. It stays part of every server.py slice.
TESTS. The two scan fixtures (test_settings_api::scan_module,
test_feedpak_extension::scan_server) patched server._make_scan_executor to swap the spawn
pool for an in-process ThreadPool; they now patch it on lib/scan.py. Worth noting WHY that
still works: the fixtures re-import `server` per test, but `scan` stays cached in
sys.modules — and it picks up the fresh CONFIG_DIR anyway, because the appstate reads are
late-bound. The seam is doing exactly the job it was built for.
━━━ TEST ISOLATION: A REGRESSION THE CARVE ITSELF CREATED (Codex [P2]) ━━━
background_scan() deliberately NEVER sets running=False — ownership of that flag lives in
_scan_runner, so a kick_scan() racing the terminal write cannot observe a stale False and
start a second runner. Correct in production.
But the scan fixtures call background_scan() DIRECTLY, skipping the runner. That was
harmless while the state lived on `server`, which the fixtures RE-IMPORT per test. It is
NOT harmless now: `scan` stays cached in sys.modules across sys.modules.pop("server"), so
the status dict OUTLIVES the test. One direct call leaves the shared scanner marked
"running" forever, and every later scan or rescan returns "already in progress" and quietly
does nothing.
Verified: after a direct call, kick_scan() returns False and starts no scan at all.
The suite passed anyway, on ordering luck — which is exactly how this class of bug ships.
tests/conftest.py::reset_scan_state now snapshots and restores lib/scan.py's module state
around the two fixtures that drive it directly.
pytest 2398, pyflakes 0, Codex 0.
Refs #48
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
One word. server.py's TuningProviderRegistry.get_merged():
except Exception:
- logger.exception("tuning provider %r raised during get_merged()", provider_id)
+ log.exception("tuning provider %r raised during get_merged()", provider_id)
There is no `logger` in server.py — the module logger is `log`. So the handler written to
swallow-and-report a bad provider instead raised NameError from inside the except, and that
NameError propagated out of get_merged().
The effect was the exact OPPOSITE of what the handler is for: one misbehaving plugin took
the whole merged-tunings call down for every other provider, AND the traceback named the
wrong problem ("name 'logger' is not defined" rather than the provider that actually blew
up). Doubly silent: nothing was ever logged either, because the logging call was the thing
that crashed.
Found by pyflakes while carving server.py (R3b). It survived because NOTHING exercised the
failure path — no test ever had a provider raise. That is the whole reason this class of
bug is invisible: it lives only on error paths, so the suite is green and the feature is
broken exactly when it matters.
tests/test_tuning_provider_isolation.py is that path:
* a raising provider must not lose the HEALTHY providers' tunings, nor the defaults
* and the failure must actually be LOGGED — swallowing is only acceptable if it reports
Bite-tested: restoring `logger` fails both.
(The log assertion attaches caplog's handler to the feedBack logger directly. It sets
propagate=False, so pytest's root capture sees nothing from it — test_plugins.py has a
capture_logger() for this, but it is not importable here: pyproject pins pythonpath to
[".", "lib"], so `tests` is not a package. Three lines beat churning 21 call sites in an
unrelated file to convert that helper into a fixture.)
pytest 2410, pyflakes 0 undefined names in server.py, Codex 0.
Closes#899
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
lib/builtin_content.py (321 lines moved). server.py 2,418 -> 2,098.
The calibration/diagnostic sloppaks and the starter library: _copy_builtin_packs,
_write_builtin_pack, the two seed helpers, their source tables, and the seed marker.
━━━ THE ONE SIGNATURE CHANGE, AND WHY THE CARVE IS UNSAFE WITHOUT IT ━━━
server.py has:
def _feedBack_server_root() -> Path:
return Path(__file__).resolve().parent
That is correct IN server.py: the repo root in dev, resources/feedBack when bundled — the
tree that actually holds docs/ and data/.
Move that body into lib/ unchanged and it keeps working, silently, and returns lib/. There
is no docs/diagnostics under lib/, so every seed would find nothing, log "source missing"
at debug, and return. Nothing raises. Nothing fails. The starter library simply never
appears, and the calibration sloppak is never seeded — on a fresh install, in the field.
A verbatim move whose MEANING changed because __file__ did.
So this module cannot compute a root: `server_root` is a PARAMETER, and server.py — the
only place that legitimately knows where it lives — passes it in. The trap is now
structurally impossible rather than merely avoided. (_copy_builtin_packs already took the
root that way; the two seed helpers now do too.)
Everything else is byte-identical. CONFIG_DIR is read late as appstate.config_dir and the
DLC root through dlc_paths._get_dlc_dir — the same seam every router in lib/routers/ uses,
late-bound because tests monkeypatch it.
━━━ PYFLAKES FOUND THREE MISSING IMPORTS THE TESTS WOULD HAVE FOUND ONE AT A TIME ━━━
The moved code uses `secrets`, `stat` and `tempfile`; none was in my import block. Each is
a NameError on a live path. `python3 -m pyflakes` names all three in one shot — this is the
Python twin of the no-undef gate that guarded every frontend carve, and it should run on
every server.py slice from here.
It also flagged a PRE-EXISTING one I deliberately did not touch: server.py's
TuningProviderRegistry.get_merged() calls `logger.exception(...)` in an except handler and
there is no `logger` in the module (it is `log`). So a raising tuning provider takes down
the merged-tunings call for everyone, with a NameError naming the wrong problem. Filed as
issue #899 rather than smuggled into a carve whose whole value is being behaviour-neutral.
The constants lost their underscore prefix: they cross a module boundary now (the seed
tests read them), so `_BUILTIN_STARTER_SOURCES` was a lie.
pytest 2397, pyflakes 0, Codex 0. Guarded by the plugin_context contract test (#898).
Refs #48
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
tests/test_plugin_context_contract.py (3 tests). No production code changes.
server.py is about to be carved apart around startup_events(), and `plugin_context` — the
20-key dict handed to every plugin's setup() — is built inline inside it. Issue #48 flagged
this while planning the split and asked for exactly this guard:
"Plugin context[...] are passed as live references into already-loaded plugins.
Refactoring must preserve the exact callables — moving them to a new module is fine,
but renaming or wrapping them breaks third-party plugins. We'd want a
'plugin context unchanged' assertion in CI."
It never got written. Writing it FIRST, because a key silently dropped or renamed by a move
is invisible to every other test in the suite — nothing in-tree reads most of these — and
would break plugins at runtime, in the field.
This is the backend's version of the window contract, and the frontend carve just taught me
what that costs: 43 of library.js's exports were referenced ONLY from app.js's top-level
window block, invisible to any call-graph scan, and trusting the scan would have shipped a
dead A-Z rail with CI fully green. A contract only external code reads has to be pinned BY
NAME, before the move, not after.
THE SURFACE IS BIGGER THAN server.py's DICT. Shipped plugins read `log` and `load_sibling`,
and neither is in it — plugins/__init__.py layers them on per-plugin. A test pinning only
server.py's 18 keys would have missed both.
━━━ CODEX CAUGHT ME WRITING A VACUOUS ASSERTION ━━━
My first identity test built a dict locally and called setup() on it — asserting
`dict(x)['k'] is x['k']`, which is trivially true and blind to everything the loader does.
[P2], and correct. It now drives the REAL plugins.load_plugins() with a probe plugin, which
matters: the loader DOES deliberately wrap one key (register_library_provider is scoped
per-plugin so a plugin cannot forge owner attribution and impersonate another). The test
pins that single intentional exception so it cannot quietly become two.
Codex then caught [P2] number two: my hand-rolled teardown restored only PLUGINS_DIR and
LOADED_PLUGINS, while load_plugins() also mutates sys.path, sys.modules and
PENDING_PLUGINS — order- and environment-dependent. tests/test_plugins.py already had a
fixture that does this properly, so `reset_plugin_state` moved to tests/conftest.py: ONE
copy, shared, rather than a second that will drift.
BITE-TESTED IN FIVE DIRECTIONS — drop a key, rename a key, drop a per-plugin key, wrap
extract_meta in the loader (all key names intact, identity broken), and remove the
register_library_provider scoping (the impersonation guard). Each fails.
pytest 2399, Codex 0.
Refs #48
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A plugin reload silently did nothing for scriptType:"module" plugins. ES modules are
evaluated ONCE PER URL PER DOCUMENT, so re-inserting a <script type="module"> whose src
the module map has already seen fires `load` without re-running the body — and the loader
then recorded the reload as applied. A no-op that reported success.
THE ISSUE UNDERSTATES IT. #879 says "upgrades are fine — a new version yields a new URL".
That is true of screen.js and FALSE of the plugin. I drove a real browser through
install(1.0.0) -> upgrade(1.1.0) -> rollback(1.0.0), counting evaluations of src/main.js:
ONE.
Not three, not two. The upgrade re-runs the one-line screen.js shim at its new ?v= URL;
the shim does `import './src/main.js'`; a relative specifier resolves against the base URL
WITH THE QUERY DROPPED; that is the same URL as before; the module map hands back the
already-evaluated v1.0.0 module. The plugin's own code never re-ran. Busting the entry
point cannot fix this, whatever token you hang off it.
So the token goes in the PATH: /api/plugins/<id>/g/<n>/screen.js. From there
'./src/main.js' resolves to /api/plugins/<id>/g/<n>/src/main.js — every relative import
inherits it, at every depth, for free. No import-specifier rewriting (which could never
see `import(expr)` anyway). Same browser drive after the fix: THREE evaluations.
Keyed on the plugin ID, not id@version: EVERY re-load of a module plugin needs a fresh
path, not just a rollback. First load keeps the stable ?v= URL, so the ETag/304 live-edit
caching the R0 rails depend on is untouched. Classic-script plugins are not affected and
never take a /g/ path.
━━━ A PATH REWRITE, NOT TWO MIRRORED ROUTES ━━━
Codex caught this, and it was right. The token shifts the BASE URL, so EVERYTHING the
module graph resolves relatively moves with it — not only imports.
`new URL('../assets/worklet.js', import.meta.url)` from /api/plugins/x/g/1/src/main.js
resolves to /api/plugins/x/g/1/assets/worklet.js. Mirroring only screen.js and src/ would
have fixed imports and 404'd every asset, worklet and wasm file the graph reaches — and
would have broken again the next time someone added a plugin route.
So the /g/<token> segment is STRIPPED BEFORE ROUTING. Every plugin route, present and
future, works under the prefix with no extra wiring. The token is opaque and never joined
into a filesystem path, so containment still rests entirely on the same safe_join.
Codex then caught a [P3] in that: eagerly re-encoding raw_path with latin-1 raises
UnicodeEncodeError on a valid plugin file like src/工具.js, 500ing a request the plain
route serves fine. raw_path is informational and Starlette routes on scope["path"], so the
mutation is simply gone — and leaving raw_path as the client sent it is more truthful for
logs anyway.
TESTS. tests/js/plugin_module_rollback.test.js (5) + 8 in test_plugin_src_route.py:
identical bytes under the prefix, the whole graph one and two levels deep, ASSETS (the
Codex [P2]), every plugin route, non-ASCII filenames (the [P3]), an opaque token, and
containment asserted as PARITY with the un-prefixed route rather than a guessed 404 —
`../screen.js` legitimately 200s on both, because the URL normalises before routing.
All bite-tested: reverting the fix fails the rollback tests, disabling the rewrite fails
the asset tests.
Two harnesses re-anchored on `script.src = _pluginScriptUrl(` — the URL literal they keyed
on now lives in the helper, further down the file, so their slice ran off the end.
node 1045, pytest 2404, ESLint 0, Codex 0.
Closes#879
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/library.js (1,988) + static/js/library-state.js (29) — bodies VERBATIM.
app.js 6,313 -> 4,451.
THE BIGGEST SLICE OF THE CARVE: 145 declarations, ~1,900 lines, 30% of what was left.
The grid, the artist tree, the A-Z rail, filters, pagination, selection, favourites, the
scan banner, and the library-provider plumbing.
A LOW module: it imports only leaves (./dom.js, ./format.js, ./library-state.js,
./tuning-display.js — all four import nothing themselves) and needs ZERO host hooks. It
calls nothing in app.js. That is not luck; it is why this cluster was picked. Two entry
points that WOULD have dragged the playback core in were left behind in app.js:
* syncLibrarySong reaches showScreen/playSong
* _handleLibArrowNav Enter on a selected row plays the song
Both are one hop from the library, and app.js is the root, so it imports from both sides
for free. Pulling them in swallows playSong, showScreen and the whole remaining core — I
measured it: the closure jumps from 145 declarations to 189.
library-state.js holds exactly FIVE fields. An imported binding is read-only, and of the
library's outward bindings only these five are genuinely WRITTEN from outside — by
showScreen, deleteSongFromModal and syncLibrarySong, none of which can move in. The other
23 are read-only from outside, so they stay plain exports (ES live bindings mean app.js
still sees every reassignment).
━━━ THE EXPORT LIST NEARLY SHIPPED A DEAD A-Z RAIL ━━━
59 exports — and 43 of them CANNOT be found by a call-graph scan. They are referenced only
from app.js's TOP-LEVEL statements: the Object.assign(window, {...}) contract and the
scattered window.X = X lines, which live outside every function, so a closure walk over
declarations never sees them. Among them are the four handler names app.js composes AT
RUNTIME into onclick="" strings — filterTreeLetter, filterFavTreeLetter, goTreePage,
goFavTreePage — the library A-Z rail and its pagination. No static tool can see those at
all. Had I trusted the call-graph, the rail would have died silently on click with nothing
failing in CI.
━━━ AND MY OWN SCANNER LIED ━━━
The cycle-risk pass reported "(none)" for this carve. It was wrong, and it could not have
been right: a dangling `else if` bound to an inner `if` instead of the outer chain, so its
`imported` map was ALWAYS empty and the check reported clean no matter what. A guard that
cannot fail is worse than no guard. Fixed, and it then found the real edges — dom.js,
format.js, tuning-display.js, library-state.js. All four are leaves, so the carve is
genuinely acyclic; I just now know it instead of assuming it.
(The AST rewriter had its own trap: `MAP[name]` with an object literal and name ===
'constructor' hits Object.prototype.constructor — truthy — and it happily rewrote
`constructor(id)` into `L.function Object() { [native code] }(id)`. Every identifier in
the file is looked up, so the lookup must not see the prototype chain. It is a Map now.)
TESTS. legacy_shim_hits SPLIT (loadLibraryProviders + setLibraryProvider -> the module;
syncLibrarySong stayed in app.js). v3_library_refresh now reads app.js AND the module,
rather than being re-pinned to whichever file happens to hold the emit this week.
VERIFIED. A/B against origin/main in two browsers: the whole window contract, cards render,
grid/tree/sort/filter/clear round-trip — and, specifically, the A-Z rail: 28 onclick
handlers composed at runtime, identical on both, and a real .click() on a letter works.
IDENTICAL on all 33 + 7 probes, no new page errors.
pytest 2396, node 1040/1040, host contract 2/2, ESLint 0 (no-cycle clean).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/format.js (17). One function. Retires the formatTime hook: 12 -> 11.
WHY A MODULE FOR ONE FUNCTION. formatTime was a host hook — loops.js and
section-practice.js both reached back through the seam for it. It is ALSO, by pure
accident of who calls it, inside the dependency closure of the library carve that comes
next. Leaving it there would have made loops.js and section-practice.js import the
LIBRARY in order to format a timestamp — nonsense, and a cycle waiting to happen.
Same rule as the transport carve: a hook is a cycle you agreed to live with; an import is
a dependency you actually have. formatTime has a real owner. It just isn't app.js, and it
certainly isn't the library. Give it a home and both consumers import it directly.
A leaf on purpose. Anything else that turns out to be a shared pure formatter belongs
here too; nothing does yet (I checked — formatBadge, _safeImageUrl and _fetchJsonOrThrow
have no callers outside the library), so nothing else is here.
node 1040/1040, host contract 2/2, ESLint 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/transport.js (377) — bodies VERBATIM. app.js 6,643 → 6,316.
THIS IS THE FIRST CARVE THAT SUBTRACTS HOOKS INSTEAD OF ADDING THEM.
Every carve before this one added host hooks: a module pulled out of app.js still had
to call back into it. But four modules were all reaching through the seam for the SAME
handful of names — _audioSeek, _audioTime, setPlayButtonState, _songEventPayload,
jucePlayer. Those names have an owner, and it isn't app.js. Give them one, and the
consumers import them directly:
count-in.js 5 hooks -> 0 (host import deleted)
juce-audio.js 4 hooks -> 0 (host import deleted)
loops.js 6 hooks -> 4
section-practice.js 10 hooks -> 7
----------------------------------------------------------
configureHost() 20 hooks -> 12
A hook is a cycle you agreed to live with. An import is a dependency you actually have.
Prefer the import whenever the name has a real owner.
_audioSeekGen now stays PRIVATE. It has exactly one writer — _resetAudioSeekState(),
which moved with it — so readers get audioSeekGen() and nobody outside can desync it.
Strictly better than the hook it replaces, which handed out a getter and left the writer
behind in app.js.
THE SCAN HAD A HOLE, AND IT BIT. Picking the carve by dependency closure over app.js's
own top-level decls said this cluster was downward-closed. It wasn't:
_currentPlaybackSnapshot reads loopA/loopB — which live in ./js/loops.js, and loops.js
imports transport. The scan saw nothing, because loopA STOPPED BEING an app.js decl the
moment loops.js was carved out. Any dependency scan of a partly-carved monolith has to
resolve the imports too, or it will confidently hand you a cycle. Added that pass; it
found exactly one back-edge, and _currentPlaybackSnapshot stays in app.js (as does
restartCurrentSong, which calls _cancelCountIn). app.js is the root — it imports both
sides for free.
TESTS. Four harnesses retargeted (play_button_reroute_guard, song_event_payload,
song_seek -> transport.js; playback_app_adapter SPLIT, since
_installPlaybackTransportAdapter stayed behind).
The two CENSUS tests — "≥8 song:* emit sites", "every seek callsite passes a reason" —
now scan app.js AND every static/js/*.js, not one file. Pointed at a single file, their
count silently shrinks as code leaves, which reads as "someone deleted an emit" or, worse,
passes while genuinely missing sites. Both bite-tested: stripping a _songEventPayload()
from an emit and adding a reason-less _audioSeek() each fail the suite.
VERIFIED. A/B against origin/main, real song, real playback: song:play payload is exactly
{audioT, chartT, perfNow, time}; song:seek carries reason "seek-by" with finite from/to;
all five song:* events fire; seekBy advances the clock; restartCurrentSong returns to zero;
the play button's aria-pressed tracks state. IDENTICAL on all 21 probes, zero page errors.
pytest 2396, node 1040/1040, host contract 2/2, ESLint 0 (no-cycle clean), Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(app): carve resume-session out of app.js (R3a)
static/js/resume-session.js (157) — the snapshot taken when you leave a song and the
pill that offers it back. Bodies VERBATIM. app.js 7,727 → 7,601.
Fifth slice out of the strongly-connected core. ONE hook (playSong) + a
currentFilename getter.
S.pendingResume JOINS THE CONTAINER — on demand, exactly as intended. app.js WRITES
it (playSong({ resume }) arms it; the song:ready listener consumes it) while this
module reads it, so it cannot be a plain export: an imported binding is read-only.
Same reason isPlaying is there. The container grows one field per carve that needs
it, never speculatively.
THE CONTRACT TEST CAUGHT THE MISSING HOOK, again on a path nothing executes:
"playSong is read by a module but never wired by app.js — it would throw at runtime".
Second time it has caught a real wiring gap the moment it appeared.
A REAL TRAP, worth remembering: I first did the S.pendingResume rewrite by feeding
acorn's identifier RANGES from node into python, and it corrupted the file
(`_pS.pendingResume null;`). **Acorn's offsets are UTF-16 code units; Python's string
indices are code points.** static/app.js contains emoji, so every offset past one
drifts. Do an AST-driven rewrite in the SAME language that produced the offsets.
`node --check` caught it; a silent version of that bug is very easy to imagine.
VERIFIED. A/B against origin/main in two browsers, real song: the window API
(resumeLastSession / _snapshotResumeSession / _readResumeSession /
_clearResumeSession), snapshot, read-back, and clear — IDENTICAL, zero page errors.
HONEST LIMIT: my probe never got the snapshot to actually PERSIST (there is a guard
beyond the 3s minimum position that a scripted playSong does not satisfy), so that
path is verified only as identical-to-main, not as observed-working. The real
coverage is tests/browser/resume-session.spec.ts, which drives the flow properly.
Zero harnesses broke. pytest 2396, node 1040/1040, ESLint 0 (no-cycle clean),
tailwind clean, Codex 0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(app): carve the JUCE/desktop audio shims out of app.js (R3a)
static/js/juce-audio.js (994) — bodies VERBATIM. app.js 7,603 → 6,643.
THE LARGEST SINGLE SLICE of the whole carve phase: 960 lines, ~13% of what was left.
Three self-installing IIFEs:
_installJuceEngineRoutingWatcher (444) routes a song to the JUCE engine or HTML5 as
the desktop output enters/leaves exclusive/ASIO
_installRendererBusFeeder (337) feeds the highway renderer bus from whichever
transport is actually running
_installJuceAudioElementShim (156) patches audio.play/pause so the rest of the app
keeps talking to the <audio> element while JUCE
owns the transport
They EXPORT NOTHING — all three publish through `window.*` (_juceMode,
_reevaluateJuceRouting, _reevaluateRendererBus, …). So app.js needs only a
side-effect import plus the one binding it actually uses
(_resetJuceAudioShimChain, which the shim IIFE assigns).
THE ORDERING QUESTION, CHECKED RATHER THAN ASSUMED. Importing this module runs the
IIFEs EARLIER than before: imports evaluate ahead of app.js's body, and therefore
ahead of configureHost(). A hook read at IIFE-execution time would THROW. So I walked
the AST at IIFE-body depth to see what they actually touch when they run: nothing but
listener registration, and `audio.play`/`audio.pause` patching — and `audio` is itself
an imported module now. Verified in the browser: both are patched on the carved build
exactly as on main, which proves the shim installs correctly at its new, earlier point.
(Had I got this wrong, host.js throws loudly rather than silently misbehaving — which
is the whole reason it has no no-op defaults.)
VERIFIED. A/B against origin/main in two browsers: the entire window.* surface the
IIFEs publish (_juceMode, _juceOutputIsExclusive, _reevaluateJuceRouting,
_reevaluateRendererBus, _clearJuceRerouteMemo), audio.play/pause patched, a real song
loading and togglePlay driving the public mirror — IDENTICAL, zero page errors.
Harnesses: juce_engine_reroute (19 tests) + renderer_bus_feeder (13) slice the IIFEs by
signature — retargeted, and each sandbox gains a `host` object routed at its EXISTING
stubs so every assertion holds unchanged. test_plugin_runtime_idempotence is SPLIT: 3 of
its 4 source-asserts stayed in app.js, the sm.emit('song:resume') one moved.
pytest 2396, node 1040/1040, ESLint 0 (no-cycle clean), tailwind clean, Codex 0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/player-controls.js (229) — the speed + mastery sliders and the four
playback-preference reads (autoplay-exit, up-next, countdown-before-song,
confirm-exit). Bodies VERBATIM. app.js 7,914 → 7,727.
The fourth slice out of the strongly-connected core, and by far the easiest: ONE
hook (handleSliderInput) and NO shared mutable state. The three groups are the same
surface — the controls under the highway — and the preference reads are the
one-line localStorage lookups half of app.js consults before deciding whether to
auto-start, show the Up Next pill, run a count-in, or confirm on exit. They travel
with the controls that set them.
Zero missed members on the first build (the no-undef pass was clean), which is the
first time that has happened in this phase.
TWO HARNESSES ARE SPLIT, and both taught something:
* speed_reset spans BOTH files — playSong (app.js) resets the speed controls
(module). Its presence GUARDS still read `src.includes('function setSpeed')`
against app.js, so once the code moved they silently evaluated FALSE and the
helpers were quietly dropped from the sandbox. A guard that disables itself is
worse than no guard. Repointed at the file the code actually lives in.
* Its `host.handleSliderInput` stub had to route at the sandbox's EXISTING spy,
not a fresh `() => {}`. The test asserts the slider was actually refreshed
(`deepEqual(__sliderInputs, ['speed-slider'])`); a fresh stub swallows the call
and the assertion passes VACUOUSLY. Same failure mode as a no-op host default —
the thing this whole seam design exists to prevent.
* autoplay_exit is split too: _autoplayExitEnabled moved, but the auto-exit
machinery around it (_clearAutoExit, holdAutoExit, _resolvePlayerOrigin) stayed.
VERIFIED. A/B against origin/main in two browsers, real song: setSpeed(0.75) ->
playbackRate 0.75; applySpeedPreset(100) -> 1; the speed slider; setMastery;
setAutoplayExit / setCountdownBeforeSong / setShowUpNext — IDENTICAL, zero page errors.
pytest 2396, node 1040/1040, ESLint 0 (no-cycle clean), tailwind clean.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/count-in.js (389) — bodies VERBATIM. app.js 8,223 → 7,913.
The third slice out of the strongly-connected core, and the first that WRITES
shared state rather than only reading it. #889's container is what makes it possible.
imports: loops (setLoop/loopA/loopB — a count-in inside an A-B loop must begin at
A), audio-el, player-state, host
hooks : _audioSeek, setPlayButtonState, _songEventPayload, togglePlay + a
jucePlayer getter
Nothing imports count-in back — app.js and section-practice both reach it through
the seam — so the graph stays acyclic.
app.js's autoplay path used to reach IN and set this module's credits timers itself
(_creditsTimer, _creditsHideOnPlay) and read _countingIn. It cannot now, and should
not have to, so the module exports the OPERATIONS instead — armCreditsHideOnPlay(),
scheduleCreditsHide(), holdCreditsThen(start), isCountingIn() — and owns its own
timer invariants. Third time this has happened (section-practice's resetSelection,
loops' state) and each time the constraint produced better code than was there
before: the module keeps its own promises instead of trusting a caller 6,000 lines
away to zero the right fields.
THE no-undef GATE FOUND FIVE MISSED MEMBERS, one at a time: showSongCreditsOverlay
and startSongCountIn (my name regex matched startCountIn, not startSongCountIn),
then _creditLineLabel, _CREDITS_MAX_MS, and _CREDIT_ROLE_VERBS. A call-graph closure
does not see a const table; only the undefined-symbol pass does.
AND A REAL TRAP: I computed _CREDIT_ROLE_VERBS's span against the ALREADY-MODIFIED
app.js and applied it to the clean one — the line numbers had drifted, so the slice
would have cut somewhere else entirely. Recomputed every span from the clean file
with acorn. Never carry line numbers across an edit.
VERIFIED. A/B against origin/main in two browsers with a real song: playback state,
the public feedBack.isPlaying mirror, audio position, cancel-count-in — IDENTICAL,
zero page errors. Unit coverage moved with the code: loop_restart's count-in
cancellation-token test and the 5 song_credits_overlay tests now read count-in.js;
loop_restart's sandbox gains a `host` object routed at its EXISTING stubs, so every
assertion is unchanged.
HONEST LIMIT: I could not make the count-in OVERLAY actually render headlessly —
its autoplay path needs a fresh-load _pendingAutostart that a scripted playSong()
never arms. Behaviour is identical to main on every probe and the unit tests cover
the logic, but the on-screen 1-2-3-4 and the credits card want a human look.
pytest 2396, node 1040/1040, ESLint 0 (no-cycle clean), tailwind clean, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/player-state.js — one exported object, two fields. app.js's 70 reference
sites rewritten. Provably a no-op; nothing shrinks.
WHY NOW. Every slice carved out of app.js so far only ever READ the state it shared
(loopA/loopB, _audioSeekGen, currentFilename), so a read-only getter hook was enough
and no container was needed — twice I checked and twice I got away with it. That
runs out at count-in: it genuinely WRITES `isPlaying` (it starts and stops playback,
4 sites) and `lastAudioTime` (2). `import { isPlaying }` then `isPlaying = true`
THROWS — an imported binding cannot be assigned to. So the state has to live on an
object: `S.isPlaying = true` is a property write, and works from any module holding
the same S. Same shape stems, studio, and editor all converged on.
DELIBERATELY SMALL. app.js has ~104 top-level `let` scalars; lifting all of them is
a ~977-site rewrite for no benefit, because most are private to one cluster and
travel with it. Only what a carved module must WRITE goes here. Add on demand.
THE REWRITE IS AST-DRIVEN, NOT TEXTUAL — and that is not fussiness. Of 100 textual
occurrences of these two names, only 70 resolve to the module binding:
* 22 are member accesses (`someObj.isPlaying`, `window.feedBack.isPlaying`)
* 4 are the LOCAL PARAMETER of `function setPlayButtonState(isPlaying)` — a blind
replace yields `function setPlayButtonState(S.isPlaying)`
* 1 is an object key
* 2 are shorthand properties `{ isPlaying }`, which must become
`{ isPlaying: S.isPlaying }` — and acorn gives a shorthand's key and value the
SAME range, so rewriting both produced `isPlaying: S.isPlaying: S.isPlaying`
until I deduped by range
A find-and-replace corrupts all 29. The rewrite walks the AST, skips shadows, member
properties and keys, and replaces identifier RANGES.
`window.feedBack.isPlaying` — the PUBLIC mirror — is a different thing and is
untouched. Two test sandboxes stub it; those were left alone deliberately.
VERIFIED WITH REAL PLAYBACK. A/B against origin/main in two browsers, real song:
togglePlay -> the public mirror goes true -> false -> true across two toggles, the
audio element's paused state follows, seekBy works — IDENTICAL, zero page errors.
Harnesses: 8 vm-sandbox suites slice playback code out of app.js and now see
S.isPlaying — juce_engine_reroute, loop_restart, play_button_reroute_guard,
playback_app_adapter, song_restart, song_seek, speed_reset, and the python
idempotence source-assert. Each gets the same container in its sandbox; every
assertion is unchanged.
pytest 2396, node 1040/1040, ESLint 0, tailwind clean, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/loops.js (261) — bodies VERBATIM. app.js 8,421 → 8,224.
The second slice out of the strongly-connected core.
It OWNS the loop state — loopA, loopB, _loopMutationGen. Nothing outside writes
them: restartCurrentSong() looked like it did, but it declares its own local `let
loopA/loopB` shadows, so the module-level bindings only ever change in setLoop /
setLoopStart / setLoopEnd / clearLoop. All four move here. No state container.
DIRECTION IS THE WHOLE DESIGN. loops and section-practice are mutually dependent —
the SCC in miniature. clearLoop() must drop section-practice's selection, and
practiceSection() must call setLoop(). Both edges cannot be imports or no-cycle
(rightly) rejects it. So:
section-practice -> reaches loops through the HOST SEAM (host.setLoop, …)
loops -> imports section-practice DIRECTLY
section-practice is the higher-level feature — a consumer of loops, not the reverse
— so it is the one that takes the indirection. app.js hands the loop module's
exports across into the seam for it. Graph stays acyclic; no-cycle passes.
THE CONTRACT TEST EARNED ITS KEEP IMMEDIATELY. It failed on the first build with
"these hooks are wired by app.js but no module reads them: playSong". My dependency
scan had counted a mention of playSong() inside a COMMENT in loops.js as a real
call. Wired but unused is precisely the "fossil of a rename" case the test exists
for — and it caught it on a path no test executes.
VERIFIED BY DRIVING BOTH SIDES OF THE SEAM. A/B against origin/main in two browsers,
real song loaded:
* setLoop(5,12) -> true; getLoop() -> 5,12 — IDENTICAL
* clearLoop() (loops -> section-practice, a direct import) -> getLoop() ->
null,null — IDENTICAL
* onPhraseNext() (section-practice -> loops, ACROSS THE SEAM) -> ok — IDENTICAL
* loadSavedLoop / saveCurrentLoop / deleteSelectedLoop on window — IDENTICAL
* zero page errors either side. An unwired hook throws, so a live app is itself
proof the seam is wired.
Harness: loop_api extracts the loop helpers by signature — retargeted to loops.js,
`export` stripped for the vm sandbox, and the sandbox's existing _audioSeek /
_audioTime / formatTime spies are now routed through a `host` object so every
assertion holds unchanged, just through the indirection the real code uses. It is
SPLIT: one test still reads app.js for the window.feedBack API surface, which stayed.
pytest 2396, node 1040/1040, ESLint 0 (no-cycle clean), tailwind clean.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/host.js (99) + static/js/section-practice.js (1,214).
app.js 9,461 → 8,409.
THE FIRST SLICE OUT OF THE STRONGLY-CONNECTED CORE. What is left in app.js is not
a tree, it is a cycle: seeding a dependency closure from section-practice, from
loops, from count-in, or from the JUCE seek shim all return the SAME 178-function
set, and setLoop() and practiceSection() call each other directly. No closure-based
carve can cut it at any seed. So it is cut BY NAME, and the calls back into app.js
go through a host seam.
61 functions + its own 24 _sectionPractice*/_sectionParents* scalars (read nowhere
else) move out. 11 hooks come back in. 4 of those are read-only GETTERS —
loopA/loopB/_audioSeekGen/_loopMutationGen are only ever READ here, never written,
so app.js keeps owning them and NO state container is needed (a 977-site lift
avoided).
app.js used to reach IN and reset the module's state by hand (clearLoop() zeroed the
selection; changeArrangement() invalidated the parent count). It cannot now — an
imported binding is read-only — so those are exported as resetSelection() and
invalidateParentCount(). Strictly better: the module owns its own invariants instead
of trusting two callers on the far side of the file to zero the right three fields.
═══ THE SILENT-NO-OP PROBLEM, SOLVED ═══
The obvious host seam is an object of no-op defaults. That is a TRAP and we walked
into it once: the plugin loader's seam defaulted populateVizPicker to `() => {}`, so
a dropped wiring line would have left the viz picker quietly not refreshing with NO
test, boot check, or bot noticing. Two layers stop it here:
1. RUNTIME — host.js is a Proxy with NO defaults and NO stubs. Reading an unwired
hook THROWS. An unwired hook cannot degrade into a no-op because there is
nothing to degrade INTO. configureHost() also rejects a non-function at WIRE
time, and refuses to run twice.
2. STATIC — tests/js/host_contract.test.js asserts the hooks the modules USE are
exactly the hooks app.js WIRES. This is the layer that matters: a runtime throw
only fires if the broken path executes, and the whole danger of a seam is the
paths that never run in a smoke test. VERIFIED TO BITE in all three drift
directions: drop a hook from configureHost -> fails; rename host.setLoop in the
module -> fails; wire a hook nobody uses -> fails.
Writing that guard took three tries and each failure is instructive: (a) the
configureHost regex anchored `});` at column 0, ran past the indented close, and
swallowed app.js's 66-name window contract — 77 "hooks"; (b) an import-stripping
regex with `[\s\S]*?` ate 14,000 characters INCLUDING the drift the bite test was
meant to catch — a guard with a hole is worse than no guard, because you trust it;
(c) `host.js'` in the import path backtracked from `js` to a "hook" called `j`.
The bite tests are what surfaced all three.
CODEX FOUND A REAL RACE [P2]. configureHost() was inside the async boot function,
after several awaits — but the window handlers (onPhraseNext, …) go live during
app.js's SYNCHRONOUS module evaluation. A user clicking one in that window would hit
"[host] … was read before configureHost() ran". It is now a bare top-level statement
sitting immediately before the window contract, so the seam is always wired before a
handler can be reached. Verified live: invoking a handler 1.2s in — well before the
boot awaits settle — works.
VERIFIED. A/B against origin/main in two browsers with a REAL song loaded: popover
toggle, practice-mode change, phrase-next, and clearLoop (all of which cross the seam
— setLoop/clearLoop/_audioTime/loopA/loopB) — IDENTICAL, zero page errors. Since an
unwired hook throws, a live app is itself proof the seam is wired.
Harnesses: section_practice_dismiss retargeted; loop_api's clearLoop sandbox gains a
resetSelection SPY (not a stub) and ASSERTS it fires — the guarantee is still tested,
just through the seam.
pytest 2396, node 1040/1040, ESLint 0 (no-cycle clean), tailwind clean, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/audio-el.js — one exported const. app.js's diff is 5 lines.
This is a HINGE, not a carve: nothing shrinks, but almost everything left in
app.js is blocked behind it.
WHY. `audio` is `document.getElementById('audio')` with 162 references in app.js
and 173 outside any one cluster. Every remaining cluster measured — settings,
app-updates, count-in (208 fns), exit-confirm (212), library-render (220) — lists
`audio` among its inbound symbols, because they all touch playback and playback
reaches for the element directly. A module that needs it cannot import app.js to
get it (that closes a cycle and fails import-x/no-cycle), so today the only way to
carve any of them would be a host seam — the exact thing #878 had to build and
#880 had to tear out.
WHY IT'S SAFE. `audio` is a `const` and is NEVER reassigned anywhere in core, so a
read-only import binding is exactly right and no state container is needed. The
162 call sites are untouched — the binding keeps its name, it is just imported
instead of declared. (Contrast the reassigned scalars — isPlaying, _avOffsetMs —
which CANNOT be shared this way: an imported binding cannot be written to. Those
still need containers, and that is the next problem, not this one.)
TIMING. app.js is <script type="module">, so it evaluates after the HTML is parsed
and its imports evaluate just before its body — the same moment app.js used to run
this exact lookup. If the element had not been in the document, `audio` would be
null and app.js's top-level `audio.addEventListener(...)` calls would throw and
kill the module. They don't.
VERIFIED WITH REAL PLAYBACK, not a boot check. A/B against origin/main in two
browsers: app alive with zero page errors (which is itself the proof the import
resolved), #audio is an AUDIO element, togglePlay/seekBy live, and playSong() on a
real library song sets audio.src and the element reports a duration — IDENTICAL on
both sides.
pytest 2396, node 1038/1038, ESLint 0 (no-cycle clean), tailwind clean, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two leaves, one PR. app.js 9,651 → 9,457.
static/js/settings-io.js (155) — exportSettings + importSettings, the Settings
backup bundle. Imports nothing. The two-phase rationale comment (server first and
atomic; then a best-effort localStorage merge) is the contract and moved with the
code.
plugin-updates → INTO static/js/plugin-loader.js, not a module of its own.
checkPluginUpdates + updatePlugin are plugin MANAGEMENT; they belong with the code
that loads plugins. A new file for 50 lines would have been a file for its own
sake.
All four are inline handlers on the Settings screen and already in app.js's window
contract, so app.js re-exposes the imported bindings unchanged.
VERIFIED BY DRIVING BOTH FLOWS. A/B against origin/main in two browsers:
* checkPluginUpdates() -> hits the API and settles the button back to
"Check for Updates" — IDENTICAL
* exportSettings() -> POSTs /api/settings/export and writes
"Exported feedBack-settings…" to #backup-status — IDENTICAL (fetch intercepted
so the assertion is on the real call, not a stub)
* all four resolve on window — IDENTICAL
* zero console/page errors either side
Zero harnesses broke. pytest 2396, node 1038/1038, ESLint 0, tailwind clean, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/tuning-display.js (228 lines) — bodies VERBATIM. app.js 9,838 → 9,650.
A LEAF: imports nothing.
Tuning NAME resolution (Drop D / Eb Standard / raw-offset fallback), bass
detection, effective string count, and the target FREQUENCIES + note names the
tuner checks against. Pure functions over a small MIDI/note-name table; the 3
_TUNING_* tables are read nowhere else and move in.
NOT A SLICE — a node-level extract. The span 2309-2535 INTERLEAVES the functions
with the `window.*` / `window.feedBack.*` assignments that publish them, and one
of those is `window.feedBack = window.feedBack || {}` — the BUS BOOTSTRAP, not
tuning code at all. Every ExpressionStatement stays exactly where it was; only the
16 functions and 3 tables move. app.js re-exposes the imported bindings from the
same lines, so the public surface and its ordering are untouched (constitution II
names window.feedBack).
app.js -> { plugin-loader, viz, diagnostics-export, dom, highway-colors,
tuning-display }
HARNESSES — 4 broke, and 3 of them broke in the SAME informative way: they sliced
app.js from `function isBassArrangement(` UP TO the marker
`window.feedBack.parseRawTuningOffsets = parseRawTuningOffsets;` — an end-marker
that (correctly) stayed behind in app.js. The module is now nothing BUT the tuning
helpers, so there is no block to slice: they read it whole and strip `export ` so
the vm sandbox still evaluates it as a script.
tuner_auto_open is SPLIT — its autoplay-gate test still reads app.js, so it keeps
APP_JS and gains TUNING_JS. Retargeting its path wholesale (my first attempt)
silently pointed the autoplay test at the wrong file.
VERIFIED BY DRIVING THE CONTRACT. A/B against origin/main in two browsers, through
the real window surface: displayTuningName -> "E Standard" / "Drop D" /
"Eb Standard", parseRawTuningOffsets('-2,0,0,0,0,0') -> [-2,0,0,0,0,0],
isBassArrangement, effectiveStringCount, displayTuningTargets, and
window.feedBack.displayTuningName / .songTuningContext — IDENTICAL on both, zero
console/page errors either side.
pytest 2396, node 1038/1038, ESLint 0, tailwind-fresh clean, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/highway-colors.js (601 lines) — bodies VERBATIM.
app.js 10,415 → 9,837. Under 10k.
DECOMPOSED, not sliced. The "settings" blob measured 47 fns / 19 inbound and was
not carvable as-is. Seeding the closure from a FUNCTION (initHighwayColors) found
only 18 fns and left 4 HWC_* constants used outside it — i.e. the seed was wrong,
not the cluster. Re-seeding from the STATE (every function touching HWC_*/`_hwc*`)
found the true cluster: 45 top-level nodes, lines 2751-3331, CONTIGUOUS, with ZERO
foreign nodes inside the span.
INBOUND: 0. EXPORTS: 2 (initHighwayColors, hwcInitSettingsUI).
The other 43 symbols — the HWC_* tables, the 12 presets, the theme store, the
share codec, the picker handlers, the window.feedBack.highwayColors facade — are
used nowhere else in core and stay private. No inline on*= handlers here (the
Settings buttons are wired by addEventListener inside hwcInitSettingsUI), so
nothing needed re-exposing on window. The three bus listeners register inside
initHighwayColors, which app.js calls — not at module top level — so no ordering
change.
THE no-undef GATE EARNED ITS KEEP. My closure said INBOUND=0; the module actually
uses `uiPrompt` (the "name this theme" prompt). It was missed because uiPrompt is
no longer an app.js DECLARATION — it's an IMPORT BINDING (from #882's dom.js), and
I was collecting declarations only. `no-undef` with typeof:true caught it.
=> Lesson for the next carve: seed `tops` from ImportDeclaration bindings too.
=> And it VALIDATES carving dom.js early: this module just imports uiPrompt from
it. Had dom.js still been stranded in app.js, this carve would have needed a
host seam.
app.js -> { plugin-loader, viz, diagnostics-export, dom, highway-colors }
plugin-loader -> viz
highway-colors -> dom
viz, diagnostics-export, dom -> (leaves)
VERIFIED BY DRIVING THE FACADE. A/B against origin/main in two browsers:
window.feedBack.highwayColors installed, identical method surface, 12 presets,
identical default slot colours, and a share-code encode→decode round-trip
returning #112233 — IDENTICAL on both, zero console/page errors either side.
Harnesses: highway_colors_facade + highway_string_colors retargeted (both
brace-extract blocks out of the source by signature).
pytest 2396, node 1038/1038, ESLint 0, tailwind-fresh clean, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/dom.js (203 lines) — esc, _escAttr, _isElementVisible, _trapFocusInModal,
_confirmDialog, uiPrompt. Bodies VERBATIM. app.js 10,593 → 10,414.
A GATHER, not a slice — the six lived in six different places (108, 635, 659,
2617, 2623, 8892). They belong together because they are the BOTTOM of the UI
stack: `esc` alone has 25 call sites and `_escAttr` 23, and every later carve that
renders HTML will need them.
That is the actual point of doing this one now. Give them a home and the next
carve imports them; leave them in app.js and the next carve that renders HTML has
to invent a host seam to reach back into app.js — exactly the trap the
plugin-loader carve had to work around until the viz layer became a module. This
is the cheapest possible way to stop that recurring.
app.js -> { plugin-loader, viz, diagnostics-export, dom }
plugin-loader -> viz
viz, diagnostics-export, dom -> (nothing)
Zero imports. Six exports (every one is used outside the cluster).
VERIFIED BY DRIVING THE MODALS, not just booting — they are interactive, so a
green suite says little. A/B against origin/main in two browsers:
* window.uiPrompt / _confirmDialog / _trapFocusInModal all resolve
* uiPrompt() mounts its modal, accepts typed input, and resolves with the typed
value ('typed') — IDENTICAL on both
* _confirmDialog() mounts and resolves true on confirm — IDENTICAL
* zero console/page errors either side
pytest 2396, node 1038/1038, ESLint 0, tailwind-fresh clean, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/diagnostics-export.js (280 lines) — bodies VERBATIM.
app.js 10,858 → 10,592.
Chosen BY MEASUREMENT, not by eye. Ran the transitive closure over four candidate
clusters and took the one with the smallest interface:
diagnostics 7 fns 235 lines span 4110-4378 imports 1 exports 2
shortcuts-modal 5 fns 235 lines span 104-9897 imports 4 exports 5
settings+updates 47 fns 1012 lines span 1409-7636 imports 19 exports 30
library-render 220 fns 4064 lines span 20-10536 imports 126 exports 117
diagnostics is contiguous and nearly closed; its one inbound symbol
(_DIAG_FILE_LABELS) lives inside the region and is read only by _renderDiagPreview,
so it moves in and the module ends up a LEAF — imports nothing.
app.js -> { plugin-loader, viz, diagnostics-export }
plugin-loader -> viz
viz, diagnostics-export -> (nothing)
Exports exactly 2: previewDiagnostics + exportDiagnostics, both already in app.js's
window contract (they're inline handlers in the Settings screen) — so app.js keeps
re-exposing them, now as imported bindings. The preview renderer, the file-label
table, and the byte/HTML formatters are used NOWHERE else in core and stay private.
VERIFIED BY DRIVING IT, not just booting. Zero harnesses broke — because the
diagnostics export flow had NO source-level test at all, which is exactly why a
green suite proves nothing here. So the flow was exercised for real: A/B against
origin/main in two browsers, window.previewDiagnostics() invoked, the preview
container rendered identical content on both, both entry points resolve on window,
zero console/page errors either side.
pytest 2396, node 1038/1038, ESLint 0, tailwind-fresh clean, Codex 0.
NOTE for the next carve: library-render is NOT a cluster — 220 functions and 126
inbound symbols is most of app.js entangled together. It cannot be carved as a
unit; it needs decomposing from the inside first.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
static/js/viz.js (770 lines) — the viz picker, renderer selection, Auto-match,
the WebGL2 probe, the 3D-promotion nag, the notation hints. Bodies VERBATIM.
app.js 11,603 → 10,857.
THE SEAM IS GONE. #878's plugin-loader needed configurePluginLoader({
populateVizPicker }) purely because _populateVizPicker lived in app.js and
importing app.js would have closed a cycle. viz.js is a LEAF — it imports NOTHING
— so plugin-loader now imports _populateVizPicker straight from it. The _host
object, the configure function, its loud-default guard, and the wiring line in
app.js are all deleted. The second carve simplifies the first.
app.js -> { plugin-loader, viz }
plugin-loader -> viz
viz -> (nothing)
NOT A PURE MOVE — one listener block had to be SPLIT. app.js had a single
top-level `if (window.feedBack) { … }` registering four handlers, and only two
were viz. song:loaded / arrangement:changed / song:ready (the mastery slider)
stay in app.js and now call the imported _autoMatchViz / _maybeShowNotationViewHint.
The viz:reverted handler MOVES, because it REASSIGNS _cancelPendingAutoLabel and
an imported binding is read-only — `_cancelPendingAutoLabel = null` would throw if
the listener stayed behind while the state moved.
ORDER CHECKED, NOT ASSUMED: viz.js's song:ready listener now registers BEFORE
app.js's own (imports evaluate first). Safe — _pendingPromotionNag is only ever
set inside _populateVizPicker, which runs at boot/plugin-refresh, never from
inside the other song:ready handler, so the two are independent.
VERIFIED — the listeners are the risk here, so they were DRIVEN, not just booted.
A/B against origin/main in two browsers:
* viz picker: 6 options (auto|default|venue|drum_highway_3d|keys_highway_3d|
highway_3d), selected highway_3d, Auto label — IDENTICAL. This alone proves
plugin-loader's direct import of viz.js works.
* emit('viz:reverted') -> picker resets to default, localStorage resets to
default, the warning logs — IDENTICAL. The MOVED listener fires.
* emit('song:ready') -> mastery slider enables, no throw — IDENTICAL. The SPLIT
listener still does both halves.
* plugin screens, module injections, 37 capability participants — IDENTICAL.
* zero console/page errors on both.
pytest 2396, node 1038/1038, ESLint 0, tailwind-fresh clean. no-cycle re-bitten on
the 3-module graph (viz -> plugin-loader fails).
Codex preflight raised a [P2] claiming viz.js's top-level bus guards would be
false because "app.js only creates the event bus later" — FALSE POSITIVE. app.js
does not create the bus; capabilities.js does, from its own <script type="module">
at index.html:122, and module scripts execute in document order, so the bus exists
long before app.js's import graph evaluates. Instrumented the setter: by viz.js's
turn `window.feedBack.on` is already a function, and the viz:reverted listener is
provably attached (firing it resets the picker). The ordering is also enforced by
test_app_shell_loads_capability_registry_before_app_runtime.
Harnesses: 5 tests retargeted to viz.js across legacy_shim_hits, venue_scene_3d,
venue_viz (each SPLIT — their non-viz tests still read app.js).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(audio): route feedpak full-mix natively under exclusive output
Song playback runs through the renderer, which WASAPI-exclusive (and
ASIO) output silences. Route single-mix feedpaks (stem-less
original_audio packs AND single-stem packs) onto the engine's backing
transport when the output device type is exclusive-style, and migrate
back to HTML5 when it isn't. Extends /api/audio-local-path to resolve
/api/sloppak/.../file/... URLs via the same containment guards as
serve_sloppak_file. Multi-stem packs stay on the WebAudio path
(Phase 2). Includes [feedpak-route] transition-gated diagnostics
logging.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(audio): renderer-bus feeder — mix renderer song audio into engine output (Phase 2)
Under exclusive-style output the native backing transport (Phase 1, #824)
carries loose /audio/ songs and feedpak full-mixes, but not the stems
plugin's multi-stem WebAudio graph or tracks JUCE rejected. The feeder taps
the renderer-side master with an AudioWorklet, re-points the owning
AudioContext at a null sink so it keeps rendering without a device, and
pushes ~10 ms chunks over IPC into the desktop engine's renderer bus
(feedBack-desktop#90 follow-up). Inert in the Docker sphere and in shared
mode. Validated by the fix12 tester spike: null-sink rendering works,
clocks hold, no overflow.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(diag): --debug ASIO routing diagnostics in static bundle
Gated on window.feedBackDesktop.audio.debugEnabled() (desktop --debug);
inert in the Docker sphere and normal desktop runs.
- [asio-diag] getCurrentDevice= full device object on outputType change
(catches ASIO drivers reporting a non-'ASIO' type name)
- [asio-diag] renderer-bus: full feeder decision vector, change-gated
(running/exclusive/stems/juceMode/elementSong/want/mode)
- [asio-diag] setSink: every sink flip with ctx state + rate
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(audio): loopback feeder mode — all app audio under exclusive/ASIO
Tester-confirmed (2026-07-11 log): song previews and other
plugin-private audio bypass the per-surface feeder taps and leak to the
default WASAPI device under ASIO output. Also confirmed: the element
capture path poisons itself when highway_3d already owns #audio's
one-shot MediaElementSource (InvalidStateError with _elCtx assigned
pre-throw → TypeError every later tick).
- New preferred mode 'loopback': one getDisplayMedia frame-audio capture
(desktop main answers with the app's own frame) covers song, previews,
and UI sounds for the whole exclusive session — engages even with no
song loaded. Local playback silenced via suppressLocalAudioPlayback,
page-mute IPC fallback otherwise.
- Sticky fallback to the existing stems/element surface modes when
capture is unavailable (old desktop main, denied, Docker sphere).
- Element capture: assign module state only after the whole chain
succeeds; close the context on failure — collision now retries clean.
- Failed engage now disables the bus and tears down loopback (no more
bus-enabled-with-no-producer stranding).
- Tests: 12 (5 new — loopback engage/preference/mute-fallback/sticky
fallback, collision retry).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(audio): loopback feeder mode — all app audio under exclusive/ASIO
Tester-confirmed (2026-07-11 log): song previews and other
plugin-private audio bypass the per-surface feeder taps and leak to the
default WASAPI device under ASIO output. Also confirmed: the element
capture path poisons itself when highway_3d already owns #audio's
one-shot MediaElementSource (InvalidStateError with _elCtx assigned
pre-throw → TypeError every later tick).
- New preferred mode 'loopback': one getDisplayMedia frame-audio capture
(desktop main answers with the app's own frame) covers song, previews,
and UI sounds for the whole exclusive session — engages even with no
song loaded. Local playback silenced via suppressLocalAudioPlayback,
page-mute IPC fallback otherwise.
- Sticky fallback to the existing stems/element surface modes when
capture is unavailable (old desktop main, denied, Docker sphere).
- Element capture: assign module state only after the whole chain
succeeds; close the context on failure — collision now retries clean.
- Failed engage now disables the bus and tears down loopback (no more
bus-enabled-with-no-producer stranding).
- Tests: 12 (5 new — loopback engage/preference/mute-fallback/sticky
fallback, collision retry).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(audio): close loopback capture context on teardown (release tap worklet)
The loopback context was reused across engages (_lbCtx || new), but teardown
only stopped the stream + deactivated the tap — never closing the context or
detaching the worklet node. Each exclusive<->shared switch orphaned a live
tap worklet on the long-lived context. Use a fresh context per session and
close it on disengage. Adds a test asserting the context is closed on teardown.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(diag): install-time + uncaught-error diagnostics for the reroute chain
2026-07-11 tester log showed the routing watcher and renderer-bus feeder
never installed (zero [feedpak-route]/[renderer-bus] lines) plus an
uncaught SyntaxError with no source location — nothing in the log said
why. New:
- global error/unhandledrejection tap logging message + filename:line:col
(error events carry the location even for parse errors in other scripts)
- explicit install / NOT-installed lines for watcher and feeder (incl.
loopback capability probe)
- DOMException detail (name/message/stack head) in the feeder retry warn
— the console-message forward stringified it to [object DOMException]
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(static): force conditional revalidation on /static (Cache-Control: no-cache)
Without Cache-Control Chromium's heuristic freshness (10% of file age)
serves /static/app.js from disk cache for hours-to-days without
revalidating. Desktop consequence: a new build's window ran the previous
build's app.js — the 2026-07-11 ASIO investigation traced 'routing
watcher never installed' + a stems module-plugin SyntaxError to exactly
this (stale loader predating scriptType support). no-cache keeps caching
but revalidates via ETag — unchanged files still cost only a 304.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(diag): gate install-time + uncaught-error [asio-diag] lines on --debug
The error tap and install lines from the previous diag commit were
unconditional. Now: error/rejection taps check _asioDiagEnabled() at
event time; install lines log deferred once the async debugEnabled()
resolves true. The NOT-installed anomaly lines stay bridge-gated
(window.feedBackDesktop present) instead — a broken bridge can't deliver
the debug flag, they fire at most once, and only in the broken state
they exist to witness. Docker sphere: fully silent.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Byron Gamatos <xasiklas@gmail.com>
The first carve, and deliberately the riskiest: app.js IS the plugin loader (the
R0 host rails), so it goes first while the module graph is still one edge deep.
static/js/plugin-loader.js (829 lines) — bodies VERBATIM. app.js 12,217 → 11,439.
Core's first `static/js/` module, exactly as constitution II anticipates.
CLOSURE (measured with acorn, not regex — brace-matching stripped source drifted):
the block at app.js:11246-12031 is contiguous and self-contained. It needs only
TWO things from the rest of app.js, and exports only TWO:
exports: loadPlugins (the window contract), bootstrapPluginsAndUi (boot)
inbound: window.showScreen — already the public host contract (constitution II),
so it is called through `window`, not re-coupled as an import
_populateVizPicker — injected via configurePluginLoader()
WHY A SEAM, NOT AN IMPORT. plugin-loader must not import app.js: app.js imports
it, so that would close a cycle. I checked whether _populateVizPicker could just
move into the module instead (which would delete the seam entirely) — it drags 9
further symbols (_canRun3D, _autoMatchViz, _showPromotionNag, …), i.e. a whole
viz cluster. That is its own carve, so the seam stays.
THE SEAM'S DEFAULT IS LOUD, ON PURPOSE. A no-op stub is the classic silent
failure for this pattern (see the editor's setHostHooks trap, hit twice): drop the
wiring call and the loader keeps working while the viz picker quietly stops
refreshing — no test, no boot check says a word. The default now console.errors,
so the smoke harness catches it. VERIFIED BY BITE TEST: removing
configurePluginLoader() from app.js surfaces
"[plugin-loader] host seam not configured" at boot. The seam IS exercised on the
plugin-startup path, so an unwired hook cannot pass silently.
no-cycle is now LIVE on core's own graph for the first time. eslint.config.js
gains `static/app.js` + `static/js/**` to the module block — app.js now `import`s,
so parsing it as a script would be a syntax error. VERIFIED BY BITE TEST: making
plugin-loader import app.js back fails with "Dependency cycle detected".
HARNESSES (the R3a note said budget one conversion per carve — it was five):
retargeted capability_inspector_nav, plugin_hydration_wipe,
plugin_loader_script_type, plugin_style_injection, legacy_shim_hits (SPLIT — one
test needs the loader, one still needs app.js) + test_plugin_runtime_idempotence.
legacy_shim_hits was missed by a symbol-name grep because it greps for a code
STRING; only the failing run found it. test_capability_events' NEGATIVE asserts
now span app.js + the loader — carving code out of app.js would otherwise make
them vacuous instead of failing.
VERIFIED: A/B against origin/main in two browsers — mounted plugin screens, 14
loaded plugin scripts, the 3 module plugins injected as <script type="module">,
37 capability participants, 14 shims, window.loadPlugins: IDENTICAL, zero
console/page errors on both. /static/js/plugin-loader.js serves 200; R0 rails
intact (src/main.js 200, conditional GET 304, script_type passthrough).
pytest 2396, node 1032/1032, ESLint 0, Codex 0.
Codex preflight caught a REAL [P1] first pass: static/js/plugin-loader.js was
untracked, so a checkout would have served an app.js importing a nonexistent
module — a failed static import kills the whole module and every window handler
with it. Now tracked.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
One attribute. #871/#872/#874/#875 exist to make this line safe.
app.js's 385 top-level `function` declarations stop being implicit `window`
properties: 87 stay reachable via the explicit contract (#874's Object.assign
block + the 47 pre-existing `window.X = X` assignments), and 298 become
module-private. Verified NO unexposed name is read from outside app.js.
Strict mode (modules are always strict) checked ahead of the flip: app.js parses
clean as `sourceType: module` (no octal, dup params, `with`), and has no implicit
globals, no `eval`/`new Function`, no top-level `this`. `registerShortcut` is
called bare at 15 top-level sites but is assigned at `window.registerShortcut`
(app.js:10387) before its first call (10648), and a bare identifier in a module
still resolves through the global object — verified `typeof
window.registerShortcut === 'function'` in the browser.
HARD GATE — app.js IS the plugin loader:
- /api/plugins script_type passthrough: editor/stems/studio = "module"
- /api/plugins/stems/src/main.js -> 200; conditional GET -> 304 (live-edit ETag)
- deep graph: stems/src/transport.js, editor/src/state.js -> 200
- window.loadPlugins present; 5 plugin screens mount; the 3 migrated plugins
injected as <script type="module">
- 37 capability participants, 14 compatibility shims, bus + capabilities v1
Every one of the shell's 336 inline handlers resolves on window under module
scope, and the A-Z rail / pagination execute 6/6 with no ReferenceError. A/B
against origin/main: the ONLY unresolvable handler is `editorToggleStemMixer`,
which is equally broken on main (a dead handler in the editor plugin — not
defined anywhere in its source; pre-existing, flagged separately).
Codex preflight raised a [P1] claiming restartCurrentSong / requestExitSong /
editRegionInEditor / returnToEditorFromHighway would ReferenceError — FALSE
POSITIVE. It scanned only #874's new Object.assign block and missed app.js's 47
scattered `window.X = X` assignments; all four are at app.js:7086/7204/8492/8511
and all four resolve as `function` in the browser with app.js loaded as a module.
pytest 2396, node 1032/1032, ESLint 0 errors, tailwind-fresh clean.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The 12 capability <script> tags become type="module". No JS changes — the
capability scripts already self-register on the window.feedBack bus,
version-negotiate (`capabilities.version !== 1` → bail), and self-guard for
idempotency. They never import or call app.js; it is pure pub/sub.
Verified they export nothing by name: no top-level declaration in
capabilities.js or capabilities/*.js is read by any other script, so losing
global scope costs nothing.
This is the first REAL exercise of the ordering fix from #872. A module defers to
after HTML parse, so the capabilities now execute AFTER the document is parsed —
while app.js still calls `window.feedBack.on(...)` at its top level. That only
works because #872 put every classic script into the same deferred queue, where
document order IS execution order: capabilities.js (line 122) still runs before
app.js (line 1237). Had app.js stayed a plain classic script it would have run
during parse, hit a bare `{}`, and died on `.on is not a function`.
A/B against origin/main, 11 probes — capabilities.version, registered
participants (37), compatibility shims (14), the bus, workingTuning, theme,
setViz/showScreen/playSong, mounted plugin screens: IDENTICAL, zero console/page
errors on both. 12 module tags served and executed; pytest 2396, node 1032/1032,
ESLint 0, Codex 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
app.js is a classic script, so each of its 385 top-level `function foo()` decls
is implicitly a property of `window`. As an ES module it will not be — module
scope is not global scope — and every name reached from outside this file would
silently vanish. This adds the explicit `window.*` assignments BEFORE the flip.
Provably a NO-OP: all 66 are top-level function declarations, so while app.js is
still a classic script `Object.assign(window, {...})` only re-assigns what
`window` already has. That is what makes it safe to land on its own, ahead of
the flip that needs it.
The consumers are wider than the inline handlers in index.html:
- inline on*= handlers in static/v3/index.html
- on*= handlers app.js BUILDS inside template literals (goFavPage,
updatePlugin, hideScanBanner, ...) — they resolve against window at CLICK
time, but live in a JS string, so scanning the HTML alone never finds them
- static/v3/*.js (showScreen alone has 17 consumers), capabilities
- feedback-desktop and the external plugin repos — easy to miss, they live in
other repos and no core test covers them
- capabilities/visualization.js reads window.setViz behind a `typeof` guard,
so losing it DEGRADES IN SILENCE rather than throwing
Constitution II names window.playSong / window.showScreen / window.feedBack as
the public extension contract, so this is an obligation, not a convenience.
FOUR names are invisible to every static tool. app.js:2156-2157 picks the
handler NAME at runtime —
const letterFn = favoritesOnly ? 'filterFavTreeLetter' : 'filterTreeLetter';
— and interpolates it into `onclick="${letterFn}('A')"`. The names exist only
inside string literals, so ESLint, no-undef, and any grep for `onclick="fn` all
miss them. They are the library A-Z rail and its pagination: drop one and those
buttons throw at click time and nowhere else.
New tests/js/window_contract.test.js scrapes the HTML's handlers AND app.js's
template-literal handlers, and pins the 4 runtime-composed names by hand.
Verified to BITE: dropping showScreen, goTreePage, or setViz each fails it with
the right message.
On-device: 28 A-Z rail buttons render with their real onclick sources
(filterTreeLetter('A'), ...) and 8/8 execute with no ReferenceError; all 66
names resolve on window in the browser.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Puts every external `<script>` in the v3 shell into the deferred queue, and
keeps each script's boot() firing at DOMContentLoaded exactly as it does today.
Behaviourally a no-op; it is what makes the ES-module flips safe.
WHY. `type="module"` defers execution to after HTML parse. Classic-`defer` and
module scripts share ONE "execute after parsing" list and run in DOCUMENT ORDER,
but a plain classic script runs DURING parse — ahead of all of them. So the
moment capabilities.js becomes a module while app.js is still plain, app.js runs
FIRST, and its 11 top-level `window.feedBack.on(...)` calls (app.js:6245-6722)
hit a bare `{}` — `_ensureFeedBackEventBus()` (capabilities.js:33), which
attaches .on/.emit/.off, would not have run yet. TypeError, app.js dies
mid-parse. Deferring everything now keeps document order == execution order
through the rest of the migration.
THE CATCH (Codex preflight caught this — a real ordering change). 22 scripts
guard their boot with `if (document.readyState === 'loading')`. A deferred
script runs at readyState 'interactive', so that test is FALSE and the else-branch
fires boot() immediately, at the script's position in document order — instead of
at DOMContentLoaded, after every script has evaluated.
That matters far more than one call site: a scan of the shell's scripts found
**43 forward references** where a script's boot() reads a global that a LATER
script defines (shell.js -> profile.js's window.v3Onboarding, songs.js ->
settings.js's window._confirmDialog, badges.js -> songs.js's
window.displayTuningName, ...). Every one of them resolves today only because
all boots happen at DOMContentLoaded. So the guards now treat 'interactive' as
not-ready (`!== 'complete'`), restoring that exactly.
Codex's specific finding (first-run onboarding silently skipped) did NOT
reproduce — shell.js's boot() awaits /api/profile, and that yield lets the
remaining deferred scripts run first. But the race it described is real, the
guard is silent when it fails (`&& window.v3Onboarding`), and the other 42
forward refs have no such await protecting them. Fixed at the root rather than
at the one site.
VERIFIED. A/B against origin/main on a fresh profile, 13 probes (onboarding
overlay, v3Onboarding/v3Songs/v3Profile/fbNotify/v3Badges/uiPrompt/showScreen,
bus, capabilities.version, createHighway, plugin scripts, mounted screens):
IDENTICAL, zero console/page errors on both. pytest 2396, node 1028/1028,
ESLint 0 errors, Codex 0.
New guard: test_every_external_script_defers_so_document_order_is_execution_order
fails if any external tag is plain classic — verified to fail on a single
reverted tag, so it actually bites.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(ui)!: remove the classic v2 shell — v3 is the only UI (R3a)
Deletes `static/index.html`, the `/v2` route, and the `FEEDBACK_UI` v2/legacy
opt-out. `/` and `/v3` both serve `static/v3/index.html`, which has been the
default since 0.3.0.
This is step 0 of the core-frontend ES-module migration (R3a). Both shells load
the same `static/app.js`, so every later step of that migration — exposing the
window contract, the `defer` ordering fix, the `type="module"` flips — would
otherwise have to be made and verified twice. Removing the fallback now halves
that surface before any of it is touched.
Incidentally fixes a latent bug in `index()`: its guard read
`if getenv_compat("FEEDBACK_UI") or getenv_compat("FEEDBACK_UI") in ("v2", "legacy")`,
whose left operand is truthy for *any* non-empty value — so `FEEDBACK_UI=v3`
actually served the **v2** shell.
- `static/tailwind.min.css` regenerated: the content globs scanned the deleted
file, so v2-only utility classes are now purged (CI's tailwind-fresh job
rebuilds and diffs it).
- Constitution amended to 1.3.0 — Principle II's frontend file list now names
`static/v3/index.html`.
- Tests: 4 suites read the v2 shell (3 via a constructed `path.join` that a
literal grep misses). Their v2 halves are paired duplicates of v3 tests that
stay, so they are dropped; `alpha_warning_banner` and the capability-registry
script-order test retarget to `static/v3/index.html`.
BREAKING CHANGE: `FEEDBACK_UI=v2` / `=legacy` and the `/v2` route are gone.
Unset the variable and use `/`. No chart, settings, or plugin data changes, and
no plugin API changes — v3 reuses the same engine.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs: drop the stale '/ is v2' plugin-verification guidance (CodeRabbit)
The v3-only rewrite updated the intro paragraphs but left three lines that
still instructed plugin authors to verify in 'both / (v2) and /v3' — now the
same shell. Historical 'in v2 it was X' contrasts are kept: they still orient
authors whose plugins also ship to users on older cores.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The artist page + external-links payload (/api/artist/{name}/page, /links,
/links/refresh) plus their exclusive helpers (_artist_links_payload,
_artist_links_from_mb, the URL-slot table) move to lib/routers/artist.py. Bodies
verbatim except @app->@router and the seam reads (meta_db->appstate.meta_db,
CONFIG_DIR->appstate.config_dir, _default_settings->appstate.default_settings).
MB link enrichment is reached as enrichment.X; the URL-safety validator is
imported from lib/library_registry.py. No new seams.
server.py: 2,507 -> 2,413 (-94).
Verified: pyflakes clean; route set unchanged (143); full pytest 2395 passed.
eslint 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Song audio (/audio/{f}), the local-audio-path resolver (/api/audio-local-path),
and raw sloppak-member serving (/api/sloppak/{f}/file/{rel}) — plus the shared
_resolve_sloppak_local_file helper — move to lib/routers/media.py. Bodies verbatim
except @app->@router and the cache/static path seams (AUDIO_CACHE_DIR->
appstate.audio_cache_dir, STATIC_DIR->appstate.static_dir, SLOPPAK_CACHE_DIR->
appstate.sloppak_cache_dir — all already in the seam). No new slots.
The two test fixtures that redirect STATIC_DIR to a temp dir now also patch
appstate.static_dir (the moved routes read the seam, not server's global).
server.py: 2,638 -> 2,507 (-131).
Verified: pyflakes clean; route set identical (143), all unique-path (no
catch-all, no shadowing); full pytest 2395 passed (the audio-local-path +
sloppak-file-traversal cases). eslint 0. Boot smoke: /audio, /api/sloppak/{}/file
serve 404 for unknown, 0 tracebacks.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The 14 /api/enrichment/* routes — status, kick/cancel, per-song state, the
Match-Review queue (accept/reject/pick/search/rematch/refresh/states), and the
AcoustID fingerprint identify endpoints — plus the route-exclusive candidate
sanitizer move to lib/routers/enrichment.py. Bodies verbatim except @app->@router
and the seam reads (meta_db->appstate.meta_db, CONFIG_DIR->appstate.config_dir).
The enrichment engine (transport, matcher, worker, upload caps) already lives in
lib/enrichment.py from the earlier subsystem move and is reached as enrichment.X.
No new seams.
server.py: 2,925 -> 2,638 (-287).
Verified: pyflakes clean; route set identical (143); full pytest 2396 passed (the
enrichment route + Match-Review + identify cases, which fake the network on the
enrichment module). eslint 0.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The library query surface (songs/albums/artists/stats/genres/tuning-names/
practice-suggestions), the provider list/art/sync endpoints, and collection CRUD
move to lib/routers/library.py. The registry itself — LibraryProviderRegistry,
LocalLibraryProvider, SmartCollectionProvider, the collection-provider lifecycle
(_sync/_unregister_collection_provider), and the shared query/collection helpers
(_library_filter_args, _split_csv, _sanitize_collection_rules, _safe_art_redirect_url,
the filter-key sets) — moves to lib/library_registry.py.
The PLUGIN CONTRACT is untouched: server.py still constructs the singleton
(LocalLibraryProvider needs meta_db), still exposes register_library_provider /
unregister_library_provider to plugins via plugin_context (with the per-plugin
ownership scoping in plugins/__init__.py), and injects the registry + local
provider into appstate. The router reads appstate.library_providers /
appstate.local_library_provider at call time; the provider classes are duck-typed
so no plugin imports a base class. Acyclic: library_registry imports
routers.art (for LocalLibraryProvider.get_art) + appstate, never server.
server.py: 3,692 -> 2,925 (-767).
Verified: pyflakes clean; ORDERED route table identical set (library block mounts
at one site — all exact/specific-path, no catch-all, no shadowing); full pytest
2397 passed (incl the plugin register/unregister + collection-as-provider tests).
eslint 0. Boot smoke: /api/library + providers list "local"; a created collection
surfaces as a `collection:N` provider through the seam; collection CRUD clean.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(audio): loopback feeder mode — all app audio under exclusive/ASIO
Tester-confirmed (2026-07-11 log): song previews and other
plugin-private audio bypass the per-surface feeder taps and leak to the
default WASAPI device under ASIO output. Also confirmed: the element
capture path poisons itself when highway_3d already owns #audio's
one-shot MediaElementSource (InvalidStateError with _elCtx assigned
pre-throw → TypeError every later tick).
- New preferred mode 'loopback': one getDisplayMedia frame-audio capture
(desktop main answers with the app's own frame) covers song, previews,
and UI sounds for the whole exclusive session — engages even with no
song loaded. Local playback silenced via suppressLocalAudioPlayback,
page-mute IPC fallback otherwise.
- Sticky fallback to the existing stems/element surface modes when
capture is unavailable (old desktop main, denied, Docker sphere).
- Element capture: assign module state only after the whole chain
succeeds; close the context on failure — collision now retries clean.
- Failed engage now disables the bus and tears down loopback (no more
bus-enabled-with-no-producer stranding).
- Tests: 12 (5 new — loopback engage/preference/mute-fallback/sticky
fallback, collision retry).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(audio): close loopback capture context on teardown (release tap worklet)
The loopback context was reused across engages (_lbCtx || new), but teardown
only stopped the stream + deactivated the tap — never closing the context or
detaching the worklet node. Each exclusive<->shared switch orphaned a live
tap worklet on the long-lived context. Use a fresh context per session and
close it on disengage. Adds a test asserting the context is closed on teardown.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Byron Gamatos <xasiklas@gmail.com>
Upload/delete, the catalog-metadata write-back, user-meta, overrides, gap-fill,
and the per-song info payload (11 routes) move to lib/routers/song.py with their
exclusive helpers (the atomic upload commit + the song-IO lock, the upload caps,
the gap-fill proposal builders). Bodies verbatim except @app->@router and the
seam reads: meta_db->appstate.meta_db, art_override_paths->appstate.art_override_paths,
and the scan/ingest helpers that stay in server.py (the scan lifecycle owns them)
-> new appstate seam callables: kick_scan, invalidate_song_caches, stat_for_cache,
and scan_status() (a getter — the underlying dict is reassigned). The gap-fill
MBID/ISRC regexes are reached as enrichment.X; _MULTIPART_OVERHEAD_SLACK (shared
with the staying AcoustID-identify route) moves to lib/enrichment.py beside
_ACOUSTID_MAX_UPLOAD_BYTES.
ROUTE ORDER: song_router mounts AFTER art_router — get_song_info's catch-all
`/api/song/{filename:path}` would otherwise shadow `/api/song/{path}/art*`
(Starlette matches first-registered; the :path converter is greedy).
server.py: 4,478 -> 3,692 (-786).
Verified: pyflakes clean; ORDERED route table preserves the specific-before-catch-all
invariant; full pytest 2397 passed (incl the art/cover 304 + CAA-fetch tests that
caught the shadowing before it was fixed). eslint 0.
BEHAVIORAL — needs an on-device pass (upload a sloppak, edit metadata write-back,
delete a song) before merge.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GET/POST /api/settings, /api/settings/reset, and the two-phase atomic
export/import bundle (/api/settings/export|import) move to lib/routers/settings.py
with their exclusive helpers (the relpath allowlist validator, the atomic writer,
the library-DB snapshot + sqlite integrity gate, the config-type validator, the
bundle schema). Bodies verbatim except @app->@router and the seam reads:
meta_db->appstate.meta_db, CONFIG_DIR->appstate.config_dir,
_running_version->appstate.running_version(), and _default_settings->
appstate.default_settings (the canonical defaults builder stays in server.py —
the scan + artist-links code share it — and is injected as a new seam callable).
server.py: 5,539 -> 4,478 (-1,061).
Verified: pyflakes clean (bar the pre-existing File/safe_join/tuning_name/ET);
route table IDENTICAL (143); full pytest 2397 passed (154 settings cases incl the
export→import round-trip + library-DB snapshot/restore + relpath-allowlist SSRF/
traversal guards, retargeted onto the settings module). eslint 0.
BEHAVIORAL — needs an on-device settings export→import round-trip sign-off before
merge (do not merge on green CI alone).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The six song-art routes — GET /api/song/{f}/art, .../art/cover-search,
.../art/candidates, POST .../art/upload, .../art/url, DELETE /api/art/{f}/override
— plus their exclusive helpers (the ETag/304 response machinery, _save_art_override,
_url_host_is_internal, _fetch_art_url + the art size/redirect caps) move to
lib/routers/art.py. Bodies verbatim except @app->@router and the seam reads:
meta_db->appstate.meta_db, ART_CACHE_DIR->appstate.art_cache_dir, and the three
shared art helpers that stay in server.py (used by the song/delete routes too)
-> appstate.<callable> (_song_pack_art_exists, _art_override_paths — already
seam-injected for the enrichment worker — plus a new art_safe_name slot). The
CAA / release-search transport lives in lib/enrichment.py and is reached as
enrichment.X. LocalLibraryProvider.get_art now calls art_router.get_song_art.
server.py: 5,988 -> 5,540 (-448).
Verified: pyflakes clean (bar the pre-existing File/safe_join/tuning_name/ET);
route table IDENTICAL (143); full pytest 2399 passed (33 art serve/candidates/
override/url cases incl the SSRF-guard _url_host_is_internal + _fetch_art_url
size-cap tests, retargeted onto the art module); test_packaging 43; eslint 0.
Boot smoke: /art 404, /art/candidates 404, DELETE /override 200 from the router.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(server): move the metadata-enrichment subsystem into lib/enrichment.py (R3)
MusicBrainz / Cover-Art-Archive / AcoustID transport, the match-scorer glue, and
the background enrichment worker (~930 lines, 61 defs) leave server.py as one
cohesive unit. Bodies are verbatim; the only changes are seam reads:
meta_db / config_dir / sloppak_cache_dir / art_cache_dir -> appstate.<slot>
_song_pack_art_exists / _art_override_paths (stay in server.py for the art +
delete routes) -> appstate.<callable> (new seam slots, injected by reference)
_env_flag -> env_compat.env_flag_compat (the existing identical helper)
_artist_title_from_filename -> imported from metadata_db (its home)
the User-Agent VERSION lookup: Path(__file__).parent ->
Path(__file__).resolve().parents[1] (lib/enrichment.py -> app root)
server.py drives the worker through the module (import enrichment; the routes +
scan lifecycle call enrichment.X). Tests that faked the network on `server`
(_mb_http_get, _enrich_network_enabled, _caa_http_get, ...) now patch the same
names on `enrichment` — the module attribute is resolved at call time, so one
setattr reaches both the routes and the worker's internal callers. Acyclic:
enrichment imports appstate/appconfig/dlc_paths/metadata_db/mb_match/
acoustid_match/sloppak/loosefolder, never server.
server.py: 6,917 -> 5,988 (-929).
Verified: pyflakes clean (bar the pre-existing File/safe_join/tuning_name/ET);
route table IDENTICAL (143); full pytest 2400 passed (140 enrichment/art cases
incl the offline-safety + transport-error-pauses-pass contracts that fake the
network); test_packaging 44 passed (enrichment.py resolves under lib/); eslint 0.
Boot smoke: /api/enrichment/status + POST /kick serve from the new module.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test: reset enrichment worker state between tests (CodeRabbit)
lib/enrichment.py now owns the worker, and it stays imported for the whole
session while the `server` fixtures pop-and-reimport `server` — so the cancel
Event / status dict / caches would leak across tests, and a stale `_enrich_cancel`
could short-circuit a later direct `_background_enrich()`. An autouse conftest
fixture clears that process-global state before each test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test: tighten the enrichment-reset fixture (CodeRabbit)
Narrow the import guard to ImportError (not blind Exception, BLE001), and stop
clearing _caa_index_locks — it's guarded by _caa_index_locks_guard, so an
unlocked clear() would race a still-alive worker, and its per-release mutexes
carry no test state.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The merged-tuning-catalog route moves to lib/routers/tunings.py, verbatim except
@app->@router, CONFIG_DIR->appstate.config_dir, and the two seam substrates it
needed:
- lib/appconfig.py — the pure config.json reader `_load_config` (used by ~11
server sites + future config-reading routers). server.py re-imports it, so
those call sites and any `server._load_config` test reference are unchanged.
- appstate.tuning_providers — the TuningProviderRegistry instance injected by
reference (a stable object mutated in place via register()/unregister()), so
the router reads the same registry plugins populate through plugin_context.
The instance stays defined in server.py, so `server.tuning_providers` still
resolves — zero test retargets.
The tuning constants (DEFAULT_REFERENCE_PITCH, TUNING_PRESET_MIDIS,
freqs_to_midis) already live in lib/tunings.py and are imported directly.
server.py: 6,960 -> 6,917.
Verified: pyflakes clean (bar the pre-existing unused `tuning_name` import);
route table IDENTICAL (143); full pytest 2400 passed (110 tuning/config cases);
eslint 0. Boot smoke: /api/tunings serves referencePitch + tunings + tuningMidis.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The three /api/diagnostics/* routes (export, preview, hardware) plus their
exclusive payload-cap helpers and the `_diag_*` normalisers. Bodies verbatim
except @app->@router, CONFIG_DIR->appstate.config_dir, _running_version()->
appstate.running_version() (a new seam slot; the impl stays in server.py where
the settings region also calls it), and the builtin-plugins lookup in
_diag_plugins_roots: Path(__file__).parent -> Path(__file__).resolve().parents[2]
(routers -> lib -> app root; plugins/ ships at the app root in every packaging
path).
The pure caps/normalisers (_diag_cap_console/_dict/_contributions,
_diag_coerce_bool, _diag_normalize_include, _DIAG_MAX_*) are re-exported from
server.py so the existing `server._diag_*` / `server._DIAG_*` tests keep
resolving — none of them monkeypatch these, so no test retargets.
server.py: 7,216 -> 6,960.
Verified: pyflakes clean (bar the intentional re-export lines); route table
IDENTICAL (143); full pytest 2399 passed (77 diag/packaging + 122 diagnostic-
matched cases incl the cap/coerce/normalize suites); eslint 0; Codex pending.
Boot smoke: /hardware, /preview, and POST /export (200 application/zip) all
serve from the new router location.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GET /api/version + its exclusive _safe_http_url URL validator. Bodies verbatim
except @app -> @router and the VERSION-file lookup: Path(__file__).parent (the
app root when this lived at the top level) -> Path(__file__).resolve().parents[2]
(routers -> lib -> app root). VERSION ships at the app root in every packaging
path (Dockerfile COPY VERSION /app/, desktop bundle).
server.py: 7,275 -> 7,214.
Verified: pyflakes clean; route table IDENTICAL (143); pytest 2401 passed (33 in
test_version_endpoint, incl the URL-validation + env-override cases); eslint 0.
Boot smoke: /api/version returns the real version + validated source/license URLs.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The last of the progression cluster. XP award (1 route) + per-song practice stats
(record/recent/best/top/per-song, 5 routes) — meta_db-only apart from the two
seam accessors record_stats uses (get_progression_content, builtin_diagnostic_
filename, both landed by shop/progression). Bodies verbatim; @app -> @router,
meta_db -> appstate.meta_db, _as_int from metadata_db, _clean_str from reqfields.
Three scattered source blocks (xp, the stats block, and the separated
/api/stats/{filename:path} which the /api/library/practice-suggestions route
splits off) are assembled into one module and mounted once. Registration order
is preserved WHERE IT MATTERS: the /api/stats/{filename:path} catch-all is
assembled LAST inside the router, so it still can't shadow the fixed /recent
/best /top paths — verified against the live route table (recent/best/top all
precede the catch-all) and the route SET is identical to origin/main (143).
server.py: 7,478 -> 7,275.
Verified: pyflakes clean; route set identical + catch-all-last; pytest 2401
passed (113 across song_stats/profile/progression); eslint 0. Boot smoke:
/stats/recent /best /top all 200 (not shadowed), xp/award 200.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(audio): route feedpak full-mix natively under exclusive output
Song playback runs through the renderer, which WASAPI-exclusive (and
ASIO) output silences. Route single-mix feedpaks (stem-less
original_audio packs AND single-stem packs) onto the engine's backing
transport when the output device type is exclusive-style, and migrate
back to HTML5 when it isn't. Extends /api/audio-local-path to resolve
/api/sloppak/.../file/... URLs via the same containment guards as
serve_sloppak_file. Multi-stem packs stay on the WebAudio path
(Phase 2). Includes [feedpak-route] transition-gated diagnostics
logging.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(audio): renderer-bus feeder — mix renderer song audio into engine output (Phase 2)
Under exclusive-style output the native backing transport (Phase 1, #824)
carries loose /audio/ songs and feedpak full-mixes, but not the stems
plugin's multi-stem WebAudio graph or tracks JUCE rejected. The feeder taps
the renderer-side master with an AudioWorklet, re-points the owning
AudioContext at a null sink so it keeps rendering without a device, and
pushes ~10 ms chunks over IPC into the desktop engine's renderer bus
(feedBack-desktop#90 follow-up). Inert in the Docker sphere and in shared
mode. Validated by the fix12 tester spike: null-sink rendering works,
clocks hold, no overflow.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(diag): --debug ASIO routing diagnostics in static bundle
Gated on window.feedBackDesktop.audio.debugEnabled() (desktop --debug);
inert in the Docker sphere and normal desktop runs.
- [asio-diag] getCurrentDevice= full device object on outputType change
(catches ASIO drivers reporting a non-'ASIO' type name)
- [asio-diag] renderer-bus: full feeder decision vector, change-gated
(running/exclusive/stems/juceMode/elementSong/want/mode)
- [asio-diag] setSink: every sink flip with ctx state + rate
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
4 routes (overview/add-paths/onboarding/events, spec 010) + their EXCLUSIVE
helpers (_goal_ui_progress, _progression_overview 112L) + the
_PROGRESSION_EVENT_TYPES whitelist. Bodies verbatim; @app -> @router,
meta_db -> appstate.meta_db, _clean_str from reqfields.
The two SHARED server accessors read through the seam: get_progression_content
(added for shop #851) and builtin_diagnostic_filename (new slot — a trivial
const-returning fn shared with the stats router's api_record_stats). Both are
injected via the second appstate.configure() after their defs (the import-top
configure runs before them). The cache + fns stay in server.py, so
test_progression_api's server._progression_content patch is untouched — 0 retarget.
server.py: 7,798 -> 7,594.
Verified: pyflakes clean; route table IDENTICAL (143); both seam accessors wired;
pytest 2401 passed (63 in test_progression_api); eslint 0. Boot smoke: GET
/api/progression 200 (drives _progression_overview + both accessors), events
400 on bad body.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The progression-content substrate: `_get_progression_content` (a lazy,
double-checked-locking content cache) is now published into the appstate seam as
a CALLABLE. The cache global + lock + the function stay in server.py (startup
uses it, and test_progression_api patches `server._progression_content`
directly), so ZERO test retargeting — routers just call
`appstate.get_progression_content()`.
Because the accessor is defined at server.py:1152 but the import-top configure()
runs at :346, a second `appstate.configure(get_progression_content=...)` publishes
it right after the def (configure is idempotent/additive).
First consumer: routers/shop.py (3 routes: buy/equip/list). Bodies verbatim;
@app -> @router, meta_db -> appstate.meta_db, _clean_str from reqfields,
_get_progression_content() -> appstate.get_progression_content(). This unblocks
stats/progression/profile next (all share the accessor).
server.py: 7,880 -> 7,845.
Verified: pyflakes clean; route table IDENTICAL (143); pytest 2401 passed
(test_progression_api's server._progression_content patch still works via the
kept cache); packaging guard; eslint 0. Boot smoke: GET /api/shop 200 (drives
appstate.get_progression_content), buy 400 on bad body.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
work keeper-chart prefs (3) + favorites toggle + tags list + saved toggle +
session/continue — 7 routes across 5 domains, all meta_db-only (0 setattr, 0
helpers). Bodies verbatim; @app -> @router, meta_db -> appstate.meta_db,
_clean_str from reqfields.
These were scattered singletons with no natural neighbor, so they're grouped as
"small library/user-state endpoints" and mounted once. All paths are distinct
and non-overlapping, so registering them together doesn't change routing:
verified the route SET is identical to origin/main (143) AND that no moved path
shadows or is shadowed by another (order-independence check).
server.py: 7,880 -> 7,833.
Verified: pyflakes clean; route set identical + order-independent; pytest 2401
passed; packaging guard 45; eslint 0. Boot smoke: tags/session GET 200,
favorites/saved toggle (400 on missing filename), work/charts 200.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(highway): lift 79 per-instance closure vars into `hwState` (R3c H lift)
Collapses createHighway()'s 79 mutable closure `let`s into one per-instance
`hwState` object. Scope-resolved rewrite via acorn + eslint-scope: 1059 edits
(1057 references + 79 defs - the deleted `let canvas, ctx, ws`), with the four
names shadowed in inner scopes (chartTime/ctx/notes/chordTemplates) resolved
correctly so only closure-bound refs move. Enables the later module split:
extracted renderer/ws modules close over `hwState` as a factory arg, so
multi-panel plugins (highway_3d, note_detect, splitscreen) don't share one
highway's state.
Container is `hwState`, NOT `H` — `H` is already canvas height (70 uses). The
frame-time gate caught that collision instantly (0 draws, `H._drawHooks is not
iterable` in the shared draw-hook path).
PERF (the whole risk): identical to the pre-lift baseline. Draw p50 2.1-2.2 ms,
p95 2.7-3.0 ms (pre-lift 2.7-3.2), measured on the Arcturus feedpak, headless.
Each closure-slot read became a `hwState.<slot>` monomorphic property load; the
hot loop pays nothing. On-device: Byron confirmed the 2D highway plays smoothly.
Tests: the ~30 highway JS suites brace-extract functions/patterns from the
source; their state references + the monotonic-clock vm sandbox now use
`hwState.<slot>` (the const _CHART_MAX_INTERP_MS etc. stay top-level, not
lifted). node --test: 1030/1030 green. Two self-inflicted over-replacements
caught and reverted (`_lefty` is a prefix of the 3D-local `_leftyCached`;
`STRING_COLORS` a suffix of `DEFAULT_STRING_COLORS`) — substring replaces on the
brace-extract regexes need word care.
Transformer saved at ~/.local/share/feedback-editor/highway-h-lift.mjs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test(highway): pin the setNoteStateProvider assertion to hwState._noteStateProvider (CodeRabbit)
The [^}]* form matched an unqualified _noteStateProvider =, so a regression to
closure-level state could still pass. Require the hwState-qualified assignment.
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* perf(harness): add 2D-highway frame-time measurement + capture the R3c gate (H0)
scripts/perf-baseline.mjs gains a `--song` mode: it wraps requestAnimationFrame
before any page script, TAGS the frames the highway actually painted (via
highway.addDrawHook), starts playback, and reports draw-frame p50/p95/p99 over
`--frames` seconds across `--runs` runs. Tagging is the point — ~half the rAF
callbacks are other cheap loops (~0.1 ms); averaging them in would hide a
renderer regression, so only draw frames are counted.
This is the metric the plan says gates the highway.js split (R3c) but the harness
never measured (it did server latency + boot + heap only, and the R0 numbers were
against an empty library). docs/perf-baseline.md now records the pre-lift baseline
on the Arcturus feedpak: p50 ~2.2 ms, p95 spread 2.7-3.2 ms across 3 runs. The
H-container lift (next) re-runs this on the same box and must stay within noise.
Maintainer/CI-only tooling; the existing no-`--song` run is unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* perf(harness): close page on error + MD040 fence + CHANGELOG (CodeRabbit)
- frameTimeOnce now closes its page in a finally, so a failing run doesn't leak
the page until the final browser.close() (CodeRabbit).
- fenced code block gets a bash language hint (MD040).
- CHANGELOG mentions the new --song frame-time mode.
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Unblocked by the DLC-path substrate (#843): chart's only server-module deps are
now app + meta_db (both seam); _get_dlc_dir/_resolve_dlc_path come from dlc_paths,
sloppak/loose detection from the shared lib modules. 4 routes (split/unsplit/
work/fileinfo), meta_db-only otherwise. 0 setattr targets, 0 helpers to relocate.
Bodies verbatim; @app -> @router, meta_db -> appstate.meta_db. include_router at
the original site; 143-route table identical to origin/main. No test retargeting.
server.py: 8,003 -> 7,909.
Verified: pyflakes clean on the router; no new undefined/dead in server.py; route
table identical; pytest 2401 passed (74 across work_charts/context_menu/
group_filter/packaging); eslint 0. Boot smoke: chart/work 200, chart/fileinfo
resolves the real pack path through _resolve_dlc_path.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(ws_highway): swallow a mid-stream WebSocketDisconnect quietly
Pre-existing (byte-identical to origin/main), flagged by CodeRabbit on #844: the
outer try's only guard was `except Exception as e: log.exception("highway_ws
unhandled error")`. The inner `except WebSocketDisconnect` covers ONLY the
post-`ready` keep-alive loop, and the drum_tab/notation sends have their own
localized guards — but the ~13 other streamed sends (beats, sections, notes,
chords, anchors, …) fall through to the blanket handler. Since
WebSocketDisconnect is an Exception subclass, a routine tab-close mid-load was
logged as an error at whichever send was in flight.
Fix: a dedicated `except WebSocketDisconnect: return` before the blanket handler,
matching the two localized guards and the lib/ coding guideline.
tests/test_ws_highway_disconnect.py drives highway_ws with a fake websocket that
raises WebSocketDisconnect on the first streamed send (`loading`), over a real
minimal sloppak. Negative-checked: removing the guard makes it fail (the
disconnect is logged as `highway_ws unhandled error`); the fix passes. Uses a
raw handler on `feedBack.server` rather than caplog, since configure_logging()
reroutes that logger through structlog where caplog doesn't observe it.
Full suite green. Closes the review thread on #844.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test(ws_highway): assert streaming stops after disconnect (exactly one send) — CodeRabbit
The stub now raises only on the first send and the test asserts ws.sends == 1,
so a handler that caught the disconnect and kept streaming would fail. Still
negative-checked: removing the guard fails the test.
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The single largest handler in server.py — the 902-line /ws/highway/{filename}
chart streamer — plus its 3 exclusive helpers (_pick_smart_arrangement,
_sanitize_authors, _sanitized_song_offset). server.py: 9,008 -> 8,003 (-1,005,
the biggest single R3 cut).
Clean move despite the size: the handler's only server-module deps are the 4
path constants + 3 exclusive helpers + app + log. Everything else it uses is
either NESTED inside the handler (_evict_audio_cache, _fill_scale_degree,
_manifest_entries, _tone_names, _xml_rank, _send_keepalives) or imported from
the shared lib modules (song/audio/sloppak/drums/notation/dlc_paths/metadata_db).
- Path constants read through the appstate seam: added static_dir /
sloppak_cache_dir / audio_cache_dir slots (config_dir already there);
server.py configures them. Bodies otherwise verbatim (@app.websocket ->
@router.websocket, PATHS -> appstate.*, log -> module logger).
- sloppak_cache_dir IS setattr-patched, so the 3 test_highway_ws_* suites now
also `setattr(appstate, "sloppak_cache_dir", ...)` next to their existing
server patch. _sanitize_authors unit tests import it from routers.ws_highway
(it moved). No other test churn.
- Removed 23 now-dead imports from server.py (song/audio/drums/notation/
bisect/contextvars/structlog/WebSocket*/_arr_smart_sort_key) — diffed against
the origin/main unused-import baseline so only NEWLY-dead ones went.
owns_tmp (assigned, never read) moved verbatim — it's pre-existing dead on
origin/main too; left as-is to keep the move faithful.
Verified: route table identical to origin/main (143, paths/methods/order);
handler body verbatim spot-checked; pyflakes clean (server has no new
undefined/dead); pytest 2400 passed; packaging guard 53; eslint 0. Boot smoke:
the highway WS streams the full chart (song_info/beats/sections/notes/chords/
notation/anchors/drum_tab -> ready) BYTE-for-byte the same message sequence as
origin/main across arrangements 0/1/2, zero tracebacks.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The keystone for the path-dependent routers (ws_highway, song, audio-local-path,
sloppak): the "where do the song files live + safe containment" layer leaves
server.py so a router can reach it.
- `_resolve_dlc_path` (pure — args only) moves verbatim.
- `_get_dlc_dir` reads the env-derived paths through the appstate seam
(appstate.dlc_dir/dlc_dir_env/config_dir) instead of module globals, so
lib/dlc_paths.py does no import-time IO.
- appstate gains `dlc_dir`/`dlc_dir_env` slots (config_dir already there);
server.py configures them. All three are env-derived, so a setenv+reimport
fixture reconfigures them for free — ZERO setattr retargeting.
- server.py RE-EXPORTS both (`from dlc_paths import _get_dlc_dir,
_resolve_dlc_path`), so its 24+16 call sites AND the tests that reach
`server._get_dlc_dir()` / `server._resolve_dlc_path()` directly
(test_dlc_junction, test_highway_ws_*) resolve unchanged — no test edits.
server.py: 9,085 -> 9,008.
Verified: _resolve_dlc_path byte-identical to origin/main; _get_dlc_dir's only
change is the three path identifiers -> appstate.*; pyflakes clean; route table
identical (143); pytest 2407 passed (test_dlc_junction + both highway_ws suites
green via re-export); packaging guard 52; eslint 0. Boot smoke: library scans
(8 songs, _get_dlc_dir), a real song resolves + serves art (_resolve_dlc_path),
and the highway WS reaches `ready` with notes.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two pre-existing cover-upload issues CodeRabbit flagged on #841 (verbatim move,
so correctly not fixed there):
1. One `except Exception as e` wrapped BOTH the PIL decode and the img.save/
tmp.replace, returned 400 for both, and echoed `e` — so a disk/permission
failure was mislabeled a client error and could leak a filesystem path. Now:
decode/validation -> 400 "Invalid image" (generic); save/replace failure ->
logged 500 "could not save cover" (no detail).
2. A shared `{pid}.png.tmp` let two concurrent uploads clobber each other's temp
file. Now a unique `tempfile.mkstemp` in the cover dir, atomic replace to
publish. Plus an existence re-check just before publishing so an upload that
raced a playlist delete can't leave an orphan cover.
mkstemp itself is INSIDE the try (Codex catch): an unwritable dir / full disk
raises there and is the same persistence failure as save/replace, so it hits the
logged generic-500 path instead of escaping as an unhandled 500. Cleanup guards
`tmp is not None` for the mkstemp-failed case.
Did NOT add a full per-playlist critical section (CodeRabbit's "heavy lift"):
FeedBack is single-user (Principle I), so a cover upload racing a delete on the
same id can't happen — documented in the code rather than building a lock
framework for a precluded race.
tests/test_playlist_cover_errors.py pins all of it; negative-checked three ways:
the old single-except-400 shape fails the 500 + no-leak-400 tests, and moving
mkstemp back outside the try fails the temp-creation-500 test. Fix passes 5/5.
Full suite 2405 passed; boot smoke: valid cover 200, bad image -> generic
"Invalid image" 400, no .tmp litter.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The biggest router yet — 12 playlist routes + custom covers — and the first that
needs the config-path seam. server.py: 9,302 -> 9,085 (-217).
Three causally-linked pieces, all required for playlists:
- `config_dir` joins the appstate seam (the plan always put the path constants
there; deferred in S3, needed now). It's env-derived, so the ~49
pop-and-reimport fixtures reconfigure it for free — ZERO setattr retargeting.
STATIC_DIR/SLOPPAK_CACHE_DIR (patched via setattr) stay in server.py until a
router that reads them is extracted, and get retargeted then.
- `_clean_str` (pure request-field sanitizer, 14 callers) -> lib/reqfields.py;
server.py imports it back. Unblocks wanted/saved/collections/profile/... later.
- routers/playlists.py: bodies verbatim, `@app`->`@router`, `meta_db`->
`appstate.meta_db`, `CONFIG_DIR`->`appstate.config_dir`, `_clean_str` from
reqfields, `_ART_CACHE_HEADERS` as a local const (art keeps server.py's).
The two exclusive cover helpers (_playlist_cover_path/_url) move with it.
include_router at the original site; full 143-route table identical to
origin/main. One test retarget: test_playlists_api called
`server._playlist_cover_path` directly -> now imports it from routers.playlists
(reads appstate.config_dir, which the `server` fixture configures).
Verified: pyflakes clean; route table identical; pytest 2401 passed (28 in
playlists+collections+appstate); packaging guard 51 (auto-picked up reqfields);
eslint 0; boot smoke drives create/rename/add-song/cover-upload/serve/delete —
the cover writes 1.png under CONFIG_DIR THROUGH appstate.config_dir and serves
200 with an mtime cache-bust token; a wrong-typed name field still 400s via
_clean_str; demo untouched.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two pre-existing races in the loops routes, flagged by CodeRabbit on #839 (the
verbatim extraction moved them unchanged from server.py, so they correctly
weren't fixed there):
1. save_loop computed `COUNT(*)` OUTSIDE meta_db._lock, then inserted inside it.
Two simultaneous unnamed POSTs read the same count and both mint "Loop N".
Fix: one lock scope around COUNT + INSERT.
2. list_loops read the shared single connection (check_same_thread=False) with
no lock, so it could overlap a POST/DELETE commit. Fix: read under the lock,
like every writer.
Low severity in context — FeedBack is single-user (Principle I), so concurrent
unnamed-loop POSTs essentially can't happen — but each fix is one lock scope.
tests/test_loops_concurrency.py pins both with a threading.Barrier that releases
16 workers into save_loop at once. Negative-checked: reverting the COUNT back
outside the lock fails the uniqueness assertion 5/5 runs; the fix passes 3/3.
pytest 2400 passed; on-device two unnamed POSTs -> ['Loop 1','Loop 2'].
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Third router. Practice loops (saved A/B regions per song): GET/POST/DELETE
/api/loops, meta_db-only (0 setattr targets, 0 helpers to relocate per
router_scan.py). Bodies verbatim; @app -> @router, meta_db -> appstate.meta_db.
include_router at the original site; 143-route table identical to origin/main.
server.py: 9,337 -> 9,301.
No test retargeting (test_demo_mode only names the paths in middleware regexes).
Verified: pyflakes clean; route table identical; pytest 2398 passed; packaging
guard green; eslint 0; boot smoke drives POST (auto-names "Loop N") / GET / DELETE
/ missing-fields error, and demo mode 403s both writes while allowing the read.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(server): extract the artist-alias routes into routers/artist_aliases.py (R3)
Second router, picked by router_scan.py: the artist-aliases / Tidy-up (P4) group
ranks at 0 monkeypatch.setattr targets and 0 helpers to relocate. 5 routes
(list/set/merge/delete aliases + raw-artist picker), all meta_db-only.
Bodies verbatim; only @app.<m> -> @router.<m> and meta_db -> appstate.meta_db.
include_router mounts at the original site; full 143-route table identical to
origin/main (paths, methods, order). No test retargets: test_artist_alias drives
via TestClient(server.app) + server.meta_db, neither of which moved.
Verified: pyflakes clean on the router; no new undefined in server.py; JSONResponse
still used in server.py (not dead); pytest 2398 passed (18 in test_artist_alias);
packaging guard green; eslint 0; boot smoke drives all 5 routes end-to-end
(set ACDC->AC/DC, read back, 400 on missing fields, delete).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs: credit both router extractions in the size-exemptions rationale (CodeRabbit)
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The packaged desktop app died at startup:
File ".../Resources/slopsmith/server.py", line 71, in <module>
import appstate
ModuleNotFoundError: No module named 'appstate'
feedback-desktop's scripts/bundle-slopsmith.sh copies a HARDCODED list of core
files into the bundle -- server.py, VERSION, lib/, data/, static/,
plugins/__init__.py. The root-level appstate.py (#833) and routers/ (#834)
shipped fine in Docker, passed every test, and were silently dropped from the
packaged app.
Both now live under lib/, the one core directory every packaging path already
copies wholesale -- Dockerfile `COPY lib/`, docker-compose.yml, and the desktop
bundler's `cp -r lib` -- and that all three put on sys.path (on Windows via the
embeddable-Python ._pth, where PYTHONPATH is ignored: build-windows.sh writes
`../slopsmith` and `../slopsmith/lib`). No feedback-desktop change and no new
release are needed for this to take effect.
lib/ is also the CORRECT home under Principle V, and always was once the design
settled: with the injection seam, appstate.py constructs nothing and does no
import-time IO, and a route module only builds an APIRouter. The premise that
forced root placement -- "appstate opens sqlite at import" -- stopped being true
when configure() replaced ownership. The Dockerfile / .dockerignore /
docker-compose.yml entries added for the root layout are reverted; nothing else
in core changes (git mv, so --follow survives).
tests/test_packaging.py is the guard: it walks server.py's module-level imports,
keeps the ones resolving inside this repo, and fails if any lives outside a
directory the packagers copy -- with the ModuleNotFoundError spelled out. So the
next root-level core module can't ship broken. Negative-checked: restoring
appstate.py to the root fails it; the message names the file and the four
packaging files a root module would have to teach.
(It also has to skip `origin in {"built-in","frozen"}` -- on 3.14 the frozen
stdlib reports origin="frozen", and Path("frozen").resolve() lands inside the
repo, which flagged `os` and `stat` as first-party.)
Verified: the bundler's copy replicated exactly into a temp dir and booted with
PYTHONPATH=<bundle>:<bundle>/lib -- `import server`, `import appstate`,
`import routers.audio_effects` all resolve, appstate.meta_db is server.meta_db,
143 routes. The same simulation against origin/main reproduces the production
ModuleNotFoundError. pytest 2398 passed (2348 + 50 new); route table still
identical to origin/main (paths, methods, order); docker build context reaches
lib/appstate.py and lib/routers/ with no __pycache__; native uvicorn boot smoke
serves /api/version, /api/library, the moved audio-effects router, and all three
migrated plugins' src/ graphs; eslint 0 errors.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The first route module through the appstate seam (#833). Picked BY MEASUREMENT,
not by the plan's guess: a transitive dep-closure scan over every route group
ranked audio-effects at 0 monkeypatch.setattr targets and exactly one exclusive
helper. (The same scan disproved the plan's assumption that artists/aliases was
free -- api_artist_links reaches _mb_http_get and _enrich_network_enabled, both
setattr targets.)
Bodies are verbatim. The only edits are mechanical:
@app.get(...) -> @router.get(...)
audio_effect_mappings.x -> appstate.audio_effect_mappings.x
The singleton read must stay a module attribute resolved at call time, so a
re-imported server re-publishes a fresh DB into the seam and monkeypatch reaches
this module. `routers/` never imports `server`: server -> routers -> appstate.
`app.include_router(...)` sits exactly where the routes used to be defined --
FastAPI matches in registration order, so the mount site preserves it. Verified
by diffing the FULL route table against origin/main: 143 routes, identical
paths, methods AND order.
server.py: 9,445 -> 9,386 lines. `fastapi.Query` went dead with the move and was
removed (the other four unused imports are pre-existing on main).
Packaging: COPY routers/ /app/routers/ plus `!routers/` + `!routers/**` in
.dockerignore (that file opens with a blanket `*`). Verified against the real
docker daemon: routers/ reaches the build context, __pycache__ does not.
Verified: pyflakes clean on routers/; no new undefined name in server.py;
pytest 2348 passed (75 in the audio-effects + demo-mode suites); eslint 0
errors; boot smoke drives all five routes end-to-end (create -> read back ->
activate -> clear -> delete -> 404 on missing -> 400 on bad body), Query(...)
still 422s on a missing required param, and demo mode still 403s all four
moved write routes while allowing the read.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(server): add appstate.py, the router seam (R3)
Routes moving out of server.py need `meta_db` and friends, but must not
`import server` -- that goes circular the moment server imports them back.
server.py keeps CONSTRUCTING its singletons and now injects them once via
`appstate.configure(...)`; routers read them back as module attributes at call
time (`import appstate; appstate.meta_db`). The Python analogue of the frontend
refactor's `configureX({...})` seams and of the plugin `setup(app, context)`
contract: dependencies flow one way, server -> routers -> appstate.
Two properties are load-bearing, both pinned by tests/test_appstate.py:
1. `import appstate` constructs nothing and touches no disk. This is why the
~49 test fixtures that `sys.modules.pop("server")` + re-import (to rebuild
meta_db under a patched CONFIG_DIR) keep working UNTOUCHED. A singleton
owned by appstate would survive that pop and go stale -- verified.
2. Reads must be late-bound. `from appstate import meta_db` freezes the binding
and defeats both a later configure() and monkeypatch.setattr -- the same
read-only-binding trap as ES imports.
configure() raises on an unknown slot instead of silently creating a global
nothing reads, and the suite asserts server ACTUALLY calls it. Negative-checked:
dropping the configure() call fails exactly the two wiring tests while the other
five stay green -- those five are the false-green a seam test must not be.
The new suite imports server through an `isolated_server` fixture that patches
CONFIG_DIR to tmp_path and closes both DB connections on teardown. An unguarded
`import server` constructs MetadataDB + AudioEffectsMappingDB under the real
`~/.local/share/feedback` (reproduced: running the file alone created
web_library.db + audio_effects.db there). The full suite now leaves the real
config dir untouched.
Packaging: `COPY appstate.py /app/` plus a .dockerignore allowlist entry. That
file opens with a blanket `*` exclusion, so root-level Python must be re-allowed
explicitly -- without it the image build fails on the COPY. Verified against the
real docker daemon (build context reaches /app/appstate.py). docker-compose.yml
gains the dev bind-mount; docker-compose.nas.yml runs the baked image, so the
COPY covers it. `routers/` will need the same two entries when it lands.
Verified: pyflakes clean; pytest 2348 passed (2341 + 7 new); eslint 0 errors;
boot smoke serves /api/version, /api/library, /api/audio-effects/mappings, and
all three migrated plugins' src/ graphs, with `appstate.meta_db is server.meta_db`
asserted against the live import.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test(appstate): address CodeRabbit — restore slots on teardown, really re-import
Two real findings on #833, both fixed:
(1) `isolated_server` closed server's DB connections but left `appstate.meta_db`
and `appstate.audio_effect_mappings` published and pointing at the closed
handles -- a live-looking, dead singleton for any later test. Teardown now
snapshots and restores both slots.
(2) `test_reimporting_server_republishes_the_fresh_singletons` never performed a
second import: it only re-asserted what `test_server_wires_the_seam` already
covers, so it could not detect the very staleness it names. (I introduced
that regression while fixing Codex's CONFIG_DIR isolation finding.) It now
pops `server`, re-imports under a SECOND CONFIG_DIR, and asserts the seam
republishes -- `second_server.meta_db is not first_db` and
`appstate.meta_db is second_server.meta_db`.
Negative-checked both directions: simulating an appstate-OWNED singleton
(configure() only-first-wins) now fails the re-import test, and dropping
server's configure() call still fails exactly the two wiring tests.
NB CodeRabbit's committable suggestion inserted the snapshot above the
fixture docstring, which would have demoted it from __doc__; written by hand
instead.
pytest 2348 passed; the full suite leaves the real ~/.local/share/feedback
untouched.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Move-only, same shape as the MetadataDB extraction. The core-owned song/tone ->
audio-effect-provider routing index leaves server.py for a flat lib/ module.
The class body is byte-identical; server.py reconstructs exactly from
origin/main minus the cut range plus the import-back and the call site.
server.py: 9,705 -> 9,433 lines.
The only non-verbatim change is the constructor seam: `__init__` takes
`config_dir` instead of reading the module-level CONFIG_DIR, so the module does
no IO at import (Principle V). The `audio_effect_mappings` singleton stays in
server.py -- no route, no test, and none of the `monkeypatch.setattr(server, ...)`
targets move. No import went dead.
Verified: pyflakes clean on the new module; no new undefined name in server.py;
pytest 2341 passed; eslint 0 errors; boot smoke drives the extracted DB
end-to-end (POST a mapping -> GET reads it back -> audio_effects.db lands in
CONFIG_DIR, proving the config_dir seam) and all three migrated plugins still
serve their src/ module graphs.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Move-only. The library metadata cache -- the `MetadataDB` class (4,018 lines)
plus the query helpers it owns (keyset paging cursors, the tuning grouping key,
smart-arrangement naming, tag normalisation, the startup DB-restore swap) --
moves out of server.py into a flat `lib/` module. Every moved block is
byte-identical to its server.py original; server.py is exactly origin/main
minus the six cut ranges, minus the now-dead `import contextlib`, plus the
import-back block and the constructor call site.
server.py: 14,037 -> 9,705 lines.
The one non-verbatim change is the seam that lets the class leave server.py:
`MetadataDB.__init__` now takes `config_dir` explicitly instead of reading the
module-level CONFIG_DIR, so `lib/metadata_db.py` does no IO at import
(Principle V). The `meta_db` singleton stays in server.py, so `server.meta_db`
(282 refs) and `server.app` (67 refs) resolve unchanged and no route moves.
None of the 114 `monkeypatch.setattr(server, ...)` targets moved.
Logging still goes through the `feedBack.server` logger, so log filters and
caplog assertions resolve to the same logger object.
`tests/test_settings_export_library_db.py` imports `_apply_pending_db_restore`
from metadata_db (the test moves with its subject); no other test changed.
Verified: pyflakes clean on the new module (zero undefined names, zero unused
imports) and no new undefined name in server.py; pytest 2341 passed;
node --test 1030 passed; eslint 0 errors; uvicorn boot smoke serves
/api/version, /api/library, and all three migrated plugins' src/ module graphs
(stems, studio, editor -> 200).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(highway_3d): size Butterchurn output canvas buffer to fill the highway
The 3D-highway Butterchurn background set only the output canvas CSS size
and called setRendererSize(), but never sized the canvas DRAWING BUFFER
(canvas.width/height). Butterchurn does not size the output canvas itself
(renderToScreen viewports to the reported size into the default
framebuffer), so the buffer stayed at the browser default 300x150 while the
viewport was the full highway. Only the bottom-left ~300x150 of the pattern
was drawn, then CSS-stretched across the whole highway -- zoomed, soft, and
aspect-wrong, worse the larger the panel.
Add _bcApplySize(cssW, cssH): set the drawing buffer to the device-pixel
render size (round(css * min(DPR, 1.5))), confine every layer (canvas,
backdrop, scrim, tint) to the highway rect, and report the same device px
to setRendererSize so buffer == on-screen viewport. Seed the buffer at
create and switch createVisualizer to pixelRatio:1, textureRatio:1 (DPR is
now folded into the reported size, so buffer == viewport == internal
texsize, no double-counting). render() and resize() both route through it.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
* chore(highway_3d): bump to 3.31.5 (3.31.4 taken by #823 on main)
Rebased onto main; #823 already shipped 3.31.4 (per-panel camera), so this
Butterchurn buffer-sizing fix advances to 3.31.5.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Byron Gamatos <xasiklas@gmail.com>
* fix(gp2rs): write beat times at 6-decimal precision
The editor/timeline derives per-bar BPM from beat spans
(bpm = beats*60/span), which amplifies rounding: at millisecond
(3-decimal) precision a constant-tempo GP import (e.g. 140 BPM) shows a
spurious per-bar "tempo drift" of ~0.05-0.7 BPM because most bar lengths
don't land on a ms boundary (worse for fast/odd meters). gp2rs computes
these beat times exactly from the GP tempo map, so the only precision
loss is the ebeat/startBeat format string. Writing them at 6 decimals
(microseconds) makes the derived tempo match GP's authored value.
Verified on GP5 imports (Highway to Hell 116, Equivalence 140, Living
After Midnight 138): the derived per-bar BPM collapses from two drifting
values to the single authored constant.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013JgxKh99UAeQqmhzSc73tv
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
* test(gp2rs): compare ebeat times by value, not string
The 6-decimal beat-time write makes _assert_ebeats' exact-string compare fail
("0.500" vs "0.500000"). These tests only assert spacing, so parse both sides
to float — precision-agnostic, no need to rewrite every parametrized list.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Byron Gamatos <xasiklas@gmail.com>
The tunings catalog is served as frequencies scaled to the reference pitch,
so every consumer that needs note identities (the v3 instrument badge's
TUNING_NOTE, plugins converging on the host profile model) reconstructs MIDI
numbers client-side via log2 — a rounding footgun at non-440 references, and
N copies of code the host can run once.
Add `tuningMidis` to the response: the same catalog keyed instrument-count →
name → absolute open-string MIDI notes (low → high). Built-ins come straight
from TUNING_PRESET_MIDIS (no float round-trip at all); provider-contributed
entries are inverted from their frequencies at the served reference via the
new freqs_to_midis() (the inverse of open_midis_to_freqs, garbage-guarded).
Purely additive — referencePitch/tunings are unchanged.
Tests: every built-in round-trips at 440; round-trip holds at 430/432/444/450
(the exact case client-side reconstruction drifts on); garbage rejected.
Claude-Session: https://claude.ai/code/session_01MS2YFb6UUSwJVV6CmEa25i
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(audio): route feedpak full-mix natively under exclusive output
Song playback runs through the renderer, which WASAPI-exclusive (and
ASIO) output silences. Route single-mix feedpaks (stem-less
original_audio packs AND single-stem packs) onto the engine's backing
transport when the output device type is exclusive-style, and migrate
back to HTML5 when it isn't. Extends /api/audio-local-path to resolve
/api/sloppak/.../file/... URLs via the same containment guards as
serve_sloppak_file. Multi-stem packs stay on the WebAudio path
(Phase 2). Includes [feedpak-route] transition-gated diagnostics
logging.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(audio): renderer-bus feeder — mix renderer song audio into engine output (Phase 2)
Under exclusive-style output the native backing transport (Phase 1, #824)
carries loose /audio/ songs and feedpak full-mixes, but not the stems
plugin's multi-stem WebAudio graph or tracks JUCE rejected. The feeder taps
the renderer-side master with an AudioWorklet, re-points the owning
AudioContext at a null sink so it keeps rendering without a device, and
pushes ~10 ms chunks over IPC into the desktop engine's renderer bus
(feedBack-desktop#90 follow-up). Inert in the Docker sphere and in shared
mode. Validated by the fix12 tester spike: null-sink rendering works,
clocks hold, no overflow.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Opening the keys highway could silently switch the user's configured MIDI
device. Two coupled defects in the plugin's MIDI selection:
1. _midiAutoConnect only consulted the plugin's own localStorage pick
(keys3d_midi_pick); with none saved it fell straight through to "first
non-loopback device", ignoring the core midi-input domain's global
selection (Settings -> Input Setup, window.slopsmith.midiInput.getSelected()).
2. _midiConnect unconditionally persisted every connect to BOTH the local
pick and the shared domain selection (mi.select). So the first-device
guess got frozen locally and overwrote the global default that other
consumers (drums, Input Setup) rely on.
Make the domain-wide selection the source of truth: _pickMidiTarget now
resolves global -> legacy local pick (fallback + name-recovery for stale
ids) -> first device, and gates the "don't grab a random device" recovery
guard on any configured preference. Gate persistence behind an explicit
`persist` flag so only a deliberate device selection writes the local pick
and the shared global; auto-connect and programmatic (audio-input) opens
open the resolved device for the session without touching either store.
mi.select() is not needed to open (open takes the logicalSourceKey directly),
so dropping it from the auto path costs nothing.
Interim step toward instrument-scoped selection in the midi-input domain
itself (the input_setup wizard is already per-instrument, but the domain
stores a single selection); tracked as a separate core follow-up.
Pure decision logic extracted to _pickMidiTarget and covered by unit tests
in data_layer.test.js.
Signed-off-by: gionnibgud <gionnibgud@gmail.com>
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
_resolveFreeCam() (keys/drum) and _freeCamFor() (highway_3d) guarded the panel
map lookup with only `i != null`, so a non-integer / negative / string index
from panelIndexFor() could resolve an unintended or inherited property (e.g.
map['toString']) instead of cleanly falling back to the global camera. Gate the
index on `Number.isInteger(i) && i >= 0` before `map[i]`, matching the hardening
already applied in _bgPanelKey(). Extend the resolver tests with float/string
(prototype-key) cases. Behavior change only for malformed indices.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Kris Anderson <topkoa@gmail.com>
Finish the docstring pass for the CamDir bridge functions changed in this PR:
convert the two per-panel _freeCamFor delegating wrappers to JSDoc, matching the
pure _resolveFreeCam / _ssApi helpers. Comment-only.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Kris Anderson <topkoa@gmail.com>
- _bgPanelKey() treated any non-null panelIndexFor() return as a valid panel id,
so a NaN/non-finite index minted a bogus "panelNaN" localStorage key instead
of falling back to "main". Gate on Number.isInteger(idx) && idx >= 0. (The
camera path is already NaN-safe — panelsMap[NaN] misses and falls through.)
- Add a NaN/negative-index case to the resolver tests (drum 22, keys 57, pass).
- Convert the camera-bridge helpers' comments to JSDoc (_bgPanelKey, _freeCamFor,
_resolveFreeCam, _ssApi across the three plugins) to lift docstring coverage on
the changed surface. Comment/robustness only; no behavior change beyond the
NaN guard.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Kris Anderson <topkoa@gmail.com>
Follow-up to the _bgPanelKey alias fix: _freeCamFor already treats
panelIndexFor as potentially throwy and catches to keep framing stable, but
_bgPanelKey called it bare. A throwing splitscreen build would take down
background-settings resolution (and the render path) even though the camera
path falls back safely. Wrap the call in try/catch, falling back to 'main'.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Kris Anderson <topkoa@gmail.com>
Song playback runs through the renderer, which WASAPI-exclusive (and
ASIO) output silences. Route single-mix feedpaks (stem-less
original_audio packs AND single-stem packs) onto the engine's backing
transport when the output device type is exclusive-style, and migrate
back to HTML5 when it isn't. Extends /api/audio-local-path to resolve
/api/sloppak/.../file/... URLs via the same containment guards as
serve_sloppak_file. Multi-stem packs stay on the WebAudio path
(Phase 2). Includes [feedpak-route] transition-gated diagnostics
logging.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* fix(playback): throttle legacy bridge-hit recording; emit loop-set for manual A/B
window.feedBack.getLoop() is a read surface plugins legitimately poll
(note_detect HUD ticked it at ~30 Hz), but every call recorded a
playback.loop-api bridge hit: compat-shim bookkeeping, a
playback:bridge-hit event, and a diagnostics snapshot rebuild +
stringify per call — real main-thread cost and a saturated hitCount in
the capability inspector, even with no song playing.
- _recordPlaybackBridge now throttles per bridgeId|surface (5 s window).
Bridge hits are a 'surface still in use' signal, not a call counter.
- setLoopEnd() (manual A/B buttons) now emits the same loop-set
transport event as setLoop(), so event-driven consumers no longer
need to poll getLoop() to see button-armed loops.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* docs(changelog): note loop-api bridge throttle fix
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(audio-effects): accept pre-rebrand chain plan schema as alias
The rebrand renamed PLAN_SCHEMA to 'feedBack.audio_effects.chain_plan.v1'
but shipped plugin bundles (rig_builder <= 2.9.x) still send the
slopsmith-era id, so _validatePlan rejected every plan and providers fell
back to their heavyweight legacy load paths (full chain rebuild per poll
cycle — audible as continuous distortion during songs). Accept the old id
as an explicit alias.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Three review findings on the per-panel camera work:
- highway_3d: _bgPanelKey() resolved splitscreen via window.feedBackSplitscreen
only, while _freeCamFor() uses the feedBackSplitscreen||slopsmithSplitscreen
alias it claims to "mirror". If the rename lands, per-panel background settings
would silently stop being per-panel while the camera stayed per-panel. Resolve
the alias the same way in _bgPanelKey.
- drum_highway_3d: applyCamera()'s "before first positionCamera()" guard tested
`_camBaseH == null`, but _camBaseH/_camBaseD were initialized to 0, so the guard
never fired (and could apply a base-0 pose for a frame). Initialize to null.
- keys + drum: the PR claimed the Camera Director resolver was unit-checked, but
nothing exercised it. Extract the resolver into pure, exported helpers
(_resolveFreeCam + _ssApi), delegate the per-instance _freeCamFor to them, and
add tests/camera_bridge.test.js covering per-panel select, global fallback,
null-when-absent, throw-safety, and the slopsmith-alias resolution. Drum 15→21,
keys 50→56, all pass; behavior unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Kris Anderson <topkoa@gmail.com>
Address a review note on the free-camera block: the comments described the
bridge as "per-panel-aware" without naming the actual globals. Spell out that
_freeCam comes from _freeCamFor(highwayCanvas) — window.__h3dCamCtlPanels[
panelIndexFor(canvas)] when split, else the global window.__h3dCamCtl, else
null — and update the nearby comment that mentioned only __h3dCamCtl. Comment-
only; no behavior change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Kris Anderson <topkoa@gmail.com>
Make the three 3D highways read the Camera Director bridge per panel so each
splitscreen panel renders its own camera (independent orbit/height/zoom/tilt/pan),
instead of all panels sharing the focused camera.
- Add a shared `_freeCamFor(canvas)` resolver to each highway: prefer this panel's
entry in `window.__h3dCamCtlPanels[panelIndexFor(canvas)]`, fall back to the
global `window.__h3dCamCtl`, else null (100% stock). Defensive on the splitscreen
global name (feedBackSplitscreen || slopsmithSplitscreen), NaN-safe, allocation-free.
- highway_3d (guitar): source `_freeCam` from the resolver (was global-only).
- keys_highway_3d: adopt the bridge for the first time — layer dolly/height/orbit +
pan/pitch offsets onto the pan/zoom follow rig at the camera write.
- drum_highway_3d: adopt the bridge — new per-frame `applyCamera()` folds the static
base pose + kick-pulse dip + free-cam offsets.
- In a follower (popped-out) window there is one panel, so the resolver yields
whatever camera the plugin set in that window; no highway change needed for pop-out.
Camera Director absent → resolver returns null → renderers behave exactly as before.
Bump each plugin patch version. Existing plugin tests pass (drum 15, keys 30); the
keys "default look unchanged" test confirms the stock path is byte-identical.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Kris Anderson <topkoa@gmail.com>
* fix(playback): throttle legacy bridge-hit recording; emit loop-set for manual A/B
window.feedBack.getLoop() is a read surface plugins legitimately poll
(note_detect HUD ticked it at ~30 Hz), but every call recorded a
playback.loop-api bridge hit: compat-shim bookkeeping, a
playback:bridge-hit event, and a diagnostics snapshot rebuild +
stringify per call — real main-thread cost and a saturated hitCount in
the capability inspector, even with no song playing.
- _recordPlaybackBridge now throttles per bridgeId|surface (5 s window).
Bridge hits are a 'surface still in use' signal, not a call counter.
- setLoopEnd() (manual A/B buttons) now emits the same loop-set
transport event as setLoop(), so event-driven consumers no longer
need to poll getLoop() to see button-armed loops.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* docs(changelog): note loop-api bridge throttle fix
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(drums): capture velocities alongside times in unmapped-percussion reporting
Both drum converters opt-in out_unmapped capture (convert_drum_track_from_midi,
convert_drum_track_to_drumtab) gain an index-aligned `velocities` list next to
`times`, carrying each dropped note real dynamics — MIDI velocity verbatim; GP
velocity with the same 1-127 gate as mapped hits, falling back to the 100
import default. A hand-mapping UI (the editor unmapped-notes dialog) can then
restore mapped notes at their source dynamics instead of flattening to v:100
(editor-side consumer: feedBack-plugin-editor#111).
The GP path chronological sort now reorders times and velocities in LOCKSTEP
so multi-voice measures cannot silently reassign dynamics. Additive: callers
that ignore the new key are unaffected.
Tests: extended tests/test_midi_import_drums.py + tests/test_gp2rs_drums.py
(alignment, lockstep sort, out-of-range fallback) — 26 passing.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JEoFeTPSnz4NpwwCG52hnu
* docs(gp2rs): clarify velocity-default comment, mark dead-path fallback
- The mapped-GP velocity comment conflated GP's authoring default (95,
Velocities.default) with the drumtab render default (100,
DEFAULT_VELOCITY in lib/drums.py) used when `v` is omitted. Clarify
both defaults and that only the latter applies to omitted hits.
- Mark the `else: times.sort()` fallback in the unmapped-percussion
time/velocity sort as belt-and-suspenders — times and velocities are
always appended together under the same len<100 guard, so lengths
can't actually diverge.
No behavior change; comment-only maintainability nits from PR review.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: ChrisBeWithYou <chris@rifflarr.local>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* feat(library): sort and badge by personal difficulty rating
Adds sort=difficulty/difficulty-desc to the library API (correlated
subquery over song_user_meta.user_difficulty, unrated songs pushed to
the bottom either direction, same pattern as the existing mastery
sort) and surfaces the rating as a badge on library cards in both the
v2 grid/tree views and the v3 grid. The rating itself already existed
(song_user_meta) — this just makes it sortable and visible, so it's
no longer only readable in the per-song edit drawer.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* fix(library): escape difficulty badge, wire tree view, add changelog+tests
- Wrap song.user_difficulty in esc() at both badge call sites
(static/app.js ~2082 and ~2283) for XSS-consistency with the
sibling tuning badge, which already uses esc().
- server.py: query_artists (the classic tree view's data source, used
by /api/library/artists) never batch-attached user_difficulty the
way query_page does for the grid, so the tree-view difficulty badge
added in 75673c3 was unreachable dead code (song.user_difficulty was
always undefined there). Now attaches it via the existing
user_meta_map() helper, same pattern as query_page.
- Add an [Unreleased] CHANGELOG.md entry for the difficulty sort +
badge feature, matching the repo's existing entry format.
- Add tests/test_library_filters.py::test_difficulty_sort_pushes_unrated_to_bottom
asserting unrated songs sort to the bottom in both sort=difficulty
and sort=difficulty-desc directions, and
::test_tree_view_songs_carry_user_difficulty covering the
query_artists fix above.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(library): chunk user_meta_map + rebuild stale tailwind css
Address review-bot findings on the difficulty sort/badge:
- user_meta_map now chunks filenames into 400-row batches (like
overrides_map) before the IN (...) query. query_artists (tree view)
passes every song across up to 50 artists, which could push the
placeholder count past SQLite's older variable limit; query_page's
small pages are unaffected. (CodeRabbit: Stability & Availability)
- Rebuild static/tailwind.min.css: the ◆N difficulty badge introduced
bg-blue-900/30 + text-blue-300, which were never compiled into the
committed stylesheet, failing the tailwind-fresh CI gate.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* feat(keys_highway_3d): sharp-layout modes, lane-color opacity, octave lines
Add a Highway layout section to the settings with a rebuilt way to draw
sharps/flats and lanes on the 3D piano highway:
- Sharps & flats layout (keys3d_bg_sharpMode): floating (original raised
plane), flat (one plane, zero-overlap piano-shaped tiled lanes with the
naturals evened out), and realistic (one plane, bars sized like the
physical keys). All geometry lives in pure laneSpanFlat()/laneSpanReal()
helpers. Default: realistic.
- Lane color opacity (keys3d_bg_laneOpacity, 0-1): fades the pitch-class
lane tint; at 0 it is a dark floor with guide lines only at the key-block
boundaries (E-F and each octave), toward 1 full vivid colored lanes. The
strips, per-lane separators and block lines crossfade with the value.
Default: 0.
- Octave separators (keys3d_bg_octaveGaps, default on) and Octave line
contrast (keys3d_bg_octaveContrast, 0-1): the B->C octave line is a dark
layer scaled by lane opacity plus a bright layer scaled by its inverse,
so it auto-shifts dark->bright as the lanes fade.
Settings re-read on init() so they apply on the next chart build. All other
behavior (MIDI scoring, palettes, camera, themes, hit feedback) is unchanged.
Unit tests cover the new defaults, the sharp-mode setting, and the lane
geometry (tiling/evening for flat, uniform/overlap for realistic).
Signed-off-by: LegionaryLeader <legionaryleader@gmail.com>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* Update plugins/keys_highway_3d/settings.html
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
* Update plugins/keys_highway_3d/screen.js
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
* fix(keys_highway_3d): don't trim active-range boundary lane when neighbor sharp is out of range
laneSpanFlat() trimmed a white key's edge for its neighboring black
key's lane even when that neighbor midi fell outside
range.activeLow..range.activeHigh — the neighbor's lane is never
drawn (see the activeLow/activeHigh skip around the lane-strip loop),
so the trim left a dark, unfilled sliver at the active-range boundary
with no sharp lane to fill it. Gate the trim on the neighbor being
in-range; callers that don't pass a range (e.g. the raw-tiling unit
tests) keep the prior unconditional-trim behavior.
Also add the CHANGELOG entry for this PR's feature set, following the
existing keys_highway_3d wording convention (no plugin-local
CHANGELOG exists; plugin.json was already bumped 0.1.2 -> 0.2.0 by
the original commits).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Signed-off-by: LegionaryLeader <legionaryleader@gmail.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
convert_midi_track_to_keys_wire builds its own inline tempo map and
divides by the raw midi.ticks_per_beat at two sites (the tempo-table
precompute and the tick_to_seconds closure). A malformed header with
division == 0 raised ZeroDivisionError, and an SMPTE division (which
mido returns as a NEGATIVE signed short) produced negative/garbage
note times.
Guard the divisor with `ticks_per_beat if ticks_per_beat > 0 else 480`
so both the zero and negative cases fall back to the SMF default. The
`> 0` form (not `or 480`) is required because a negative value is
truthy and would slip past `or`. Positive-division behavior is
unchanged.
Follow-up to #796, which fixed the same class of bug in the newer
convert_midi_tempo_map / _build_tick_to_seconds path.
Adds two focused tests: division == 0 no longer crashes and emits a
non-negative time, and a negative/SMPTE division yields sane
non-negative times.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(midi): convert_midi_tempo_map — extract tempos, time signatures, beat grid
The keys/drums MIDI converters always built a tempo-aware tick->seconds
map internally (to bake note times) and then discarded it — and never
read time_signature meta at all — so every MIDI import landed with no
bars, no measures, and an implied 4/4 no matter what the file said.
New lib/midi_import.py helper convert_midi_tempo_map(midi_path,
track_index) extracts the grid a .mid actually carries:
- tempos: {time, bpm} per tempo event (deduped per tick, 120 default)
- time_signatures: {time, ts: [num, den]} — the song-timeline shape
- beats: one row per beat on the editor grid shape — numbered downbeats
with a den hint, measure:-1 interior beats; the beat unit follows the
active signature (6/8 = six eighth-note rows per bar)
Event scope mirrors _build_tick_to_seconds: SMF type 0/1 merge meta
from all tracks, type 2 reads ONLY the chosen track (independent
timelines — callers must never share one grid across type-2 tracks).
Mid-bar signature events (ill-formed but seen in the wild) apply at the
next bar boundary. All times compute from absolute ticks through the
cumulative tempo table and round once at emit — rounding error never
accumulates with song length. A bar-count safety valve guards malformed
files. Consumer: the editor's multitrack MIDI import (tempo-seed
dialog, feedBack-plugin-editor roadmap Phase 3).
Tests: tests/test_midi_tempo_map.py — 10 cases driving the REAL
function against real .mid files built with mido (no stubs): default
grid, tempo bends, 500-bar rounding-drift check, 4/4->3/4 and 6/8
signatures, mid-bar signature deferral, duplicate-tick last-wins,
type-2 meta isolation from a bogus sibling track, empty files, grid
coverage bounds. Full MIDI-adjacent suite green: 55 passed
(test_midi_tempo_map + test_midi_import + test_midi_import_drums +
test_gp2midi) under the project venv.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JEoFeTPSnz4NpwwCG52hnu
* fix(midi): make convert_midi_tempo_map robust — real division guard, symmetric tempo default, single-pass meta
- Push the ticks-per-beat fallback into _build_tick_to_seconds (the single
place ticks route through), guarding on `> 0` so a division==0 (malformed)
or negative SMPTE-division header no longer raises ZeroDivisionError or
walks the beat grid into negative times. Mirror the guard at the
convert_midi_tempo_map beat_ticks site. The local `or 480` was cosmetic
before — the closure still divided by the raw division.
- Seed tempos_out with a 120 BPM row at time 0 when the first set_tempo
lands after tick 0, symmetric with the (0, 4, 4) time-signature default,
so the sidecar matches the grid the head of the song actually used.
- Collapse the duplicated meta_source/note_source lists into one
source_tracks walked in a single pass (meta collection + end_tick).
- Fix a weak assert in test_mid_bar_signature_applies_at_the_next_boundary
(operator-precedence `(A and B) or C`) to assert den == 4 outright.
- Add tests: non-positive division (0 + negative SMPTE), first tempo after
start seeds 120@0, explicit SMF type-0 file, and the _TEMPO_MAP_MAX_BARS
safety valve.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: ChrisBeWithYou <chris@rifflarr.local>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
`injectPlayerButton()` anchored the injected Tuner button with
`controls.querySelector('button:last-child')`, which can match a NESTED
button that is not a direct child of `#player-controls`. `insertBefore(btn,
nestedButton)` then throws `NotFoundError` (the reference node must be a
direct child); since injection runs from the tuner's `screen:changed`
handler, the throw propagated out of the player-screen transition and
stalled its render. The v3 path was already safe (plugin-control slot);
only the classic anchor was bad.
Use `:scope > button:last-of-type` (direct child only) with a
`parentNode === controls` guard before insertBefore, falling back to
appendChild. Bump plugins/tuner 1.3.3 → 1.3.4.
Test: tests/plugins/tuner/js/inject_player_button.test.js — extracts the
real function and runs it over a faithful DOM model whose insertBefore
enforces the direct-child invariant; the nested-last-button case reproduces
the throw on the old anchor and passes on the new one (5 tests).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: byrongamatos <xasiklas@gmail.com>
applySavedPrefs() rebuilt state.filters without the `genre` key, so with
saved prefs restored from localStorage state.filters.genre was undefined.
Clicking Filters ran renderDrawer(), which indexes f.genre.includes(g)
whenever the library has >=1 genre -> TypeError, renderDrawer aborts, and
openDrawer never removes translate-x-full. The drawer stayed off-screen so
the menu appeared dead. Only triggered for users with saved prefs AND a
non-empty genre list, matching the intermittent report.
Carry genre: [] alongside the other session-only facets (mastery, match),
mirroring the default and clear-all shapes which already include it.
Also rename the visible "Stems (sloppak)" drawer label to "Stems (feedpak)"
to match the public format name used elsewhere in the UI.
Signed-off-by: topkoa <topkoa@gmail.com>
The whisperx->transcribed alias runs before the membership check, so the literal whisperx never reaches _ALLOWED_LYRICS_SOURCES (same reason sng is omitted). Remove the dead entry. Per CodeRabbit review on #799.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
The feedpak spec (§7.1) defines the lyrics_source vocabulary as {authored, transcribed, user}, but the reader only accepted the legacy {xml, notechart, whisperx, user} set and silently downgraded anything else to "xml". A spec-compliant writer (e.g. the stem_splitter plugin, which emits transcribed for WhisperX-produced lyrics) therefore lost its provenance badge.
Widen the allowed set to the union of the spec vocabulary and the legacy values so both validate, and alias the legacy whisperx engine name to the spec transcribed so existing packs normalise to the spec badge.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
The v3 Settings "Save" button posts dlc_dir together with demucs_server_url,
default_arrangement and av_offset_ms in one request. POST /api/settings
validated dlc_dir first and early-returned "DLC directory not found" before it
ever processed demucs_server_url, so on a machine whose DLC path doesn't resolve
(fresh install, unplugged/network drive, a path carried over from another
machine) setting the Demucs server address silently failed — reported in
got-feedBack/feedBack-demucs-server#3 (macOS 07-05 nightly).
- server: a non-resolving dlc_dir is now recorded as a warning and skipped
rather than aborting the whole POST, so the co-submitted keys still persist.
The bad path is surfaced via a new additive `warnings` field and folded into
`message` so the settings status line still shows it.
- client (v3): the Demucs input now autosaves on blur/enter via a single-key
persistSetting POST, like every other v3 setting, so it never depends on the
coupled Save button.
- tests: cover the decoupling and the unchanged happy path.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(keys_highway_3d): add note-colour palettes and selectable camera angles
Give the 3D keys highway player-facing view options and a tuned default
look, so the piano highway is readable out of the box and customisable
from the settings panel without touching code.
Note colours (settings -> Note colours, `keys3d_bg_palette`):
- Octaves (new default): each octave its own hue climbing the rainbow,
darker sharps, so pitch height reads at a glance on any note range.
- Rainbow: the original per-pitch table (colours unchanged).
- Vivid / Pastel: per-pitch variants.
- Emerald / Ice: single-hue two-tone (uniform naturals, darker sharps).
The pick drives the notes, key glow, lane guides and hit flames, live.
Camera (settings -> Camera angle, `keys3d_bg_camera`):
- Classic (the original low rig) / Elevated / Overhead (new default).
- Height, distance and tilt fine-tune sliders nudge the base vantage the
auto-pan/zoom follow-motion orbits; presets apply live.
The new defaults are opinionated for plug-and-play (octaves palette,
overhead camera, tilt -0.6); anyone who prefers the original look can
pick Rainbow + Classic. Settings changes are re-read on init() so they
apply on return from the settings screen, not only after a relaunch.
Scoring, hit-timing and MIDI handling are untouched -- these are purely
visual. Numeric FX keys clamp to declared ranges (FX_RANGES); the pure
colour/camera helpers are covered by unit tests (node --test).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: LegionaryLeader <legionaryleader@gmail.com>
* fix(keys_highway_3d): make Classic camera reproduce the original rig + drop per-frame camera alloc
Review follow-ups on the palettes/camera feature:
- "Classic" preset now reproduces the historical rig exactly. The tuned
plug-and-play downward aim (camTilt -0.6 x CAM_TILT_UNITS = -33) is baked
into CAM_PRESETS.overhead.lookY, and camTilt now defaults to 0 (neutral).
The default overhead look is byte-identical (effective lookY still -33),
but "pick Classic for the original look" is now actually true instead of
leaving a -33 down-tilt applied. settings.html tilt slider defaults to 0.
- _rig() writes into a hoisted reusable object instead of allocating a fresh
{y,z,lookY,lookZ} literal every frame, honoring the module's documented
"no per-frame allocations in draw()" discipline. Callers read it
synchronously and never retain it, so one shared instance is safe.
Tests updated for the neutral camTilt default; adds an invariant test that
the default overhead framing is unchanged and Classic + neutral tilt == the
historical LOOK_Y. Full JS suite green (1069 pass).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Signed-off-by: LegionaryLeader <legionaryleader@gmail.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
shadowBlur cost scales with the blurred device-pixel area. The lit-sustain
trail can span half the canvas, the canvas is DPR-scaled (4x pixels on a
2x Mac), and the blur ran on every frame exactly while a sustain is HELD
— i.e. at the moment the player most notices a hitch. Sustain-heavy songs
(e.g. fingerpicked acoustic charts) hit this constantly.
Replace the blur with three inflated low-alpha fills of the same trail
quad: reads as the same soft shimmering glow (the shimmer LUT still
drives per-frame flicker, feedBack#254 intent preserved) at a flat,
area-independent cost. Also drops the now-dead shadowBlur reset in the
crackle pass; no shadowBlur uses remain in highway.js.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Byron Gamatos <xasiklas@gmail.com>
* feat(v3): add a handedness (left-handed) choice to the instrument selector + onboarding
Left-handed players could already mirror the highway, but only via a buried
Settings toggle they had to find AFTER setup -- so a lefty went through the tour,
the tuner and calibration all right-handed first (community callout).
Add a "Handedness: Right / Left" row to the v3 instrument badge popover, alongside
Instrument / Strings / Tuning (all player-orientation choices). It writes the same
lefty preference -- highway.setLefty when a live highway exists (flips it
immediately + persists), else the 'lefty' localStorage key the highway reads on
init -- and keeps the Settings "Left-handed" checkbox in sync. The first-run
tour's "Choose your instrument" step, which runs before the tuner/audio-
calibration steps, now calls it out so lefties flip it up front.
Frontend-only, additive. Full core JS suite green (938). Tests:
tests/js/badges_handedness.test.js.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UR2Cr7GEu3yMY7SrfxH6c1
* docs: split the spliced Handedness/Colorblind CHANGELOG entries
A rebase pasted the Handedness bullet over the Colorblind preset entry's bold
lead, merging two unrelated Added entries into one run-on bullet. Restore them
as two separate bullets.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
The cover picker only offered CAA covers from a song's MATCHED release, so an
unmatched song (the city-pop pile) got nothing but Current/Pack/Upload/URL. Add
a search box: GET /api/song/{fn}/art/cover-search?q= searches MusicBrainz
release-groups and returns each album's CAA front-250 thumb; the picker renders
them as pickable tiles (same apply→/art/url path; covers with no CAA art
self-hide). Pre-filled from the song's artist + album/title (romaji fallback), so
a blank-artist pack pre-fills "Junko Yagami …". Reuses the throttled _mb_http_get.
* feat(v3 library): click a "No match" badge to fix it — batch → popup handoff
Connects the two halves: the "No match" badge (the unmatched pile) now opens the
Fix-metadata popup for that song in one click, instead of right-click → menu.
The resting badge becomes interactive (pointer-events-auto + hover), carrying a
data-meta-fix hook; wireCards opens window.__fbFixMatch(playTarget) on click and
stops propagation so it doesn't also play the card. Batch tile states stay
non-interactive. Loop becomes: Unmatched filter → see the pile → click one →
fix it. tailwind.min.css regenerated for the badge's hover classes.
* feat(v3 library): show the author's romaji, not blank/native script (English base)
Two changes so an English-speaking base never sees a blank name or native script:
- Filename romaji fallback: a blank-artist CDLC pack ("Artist_Title_v1_p") shows
nothing useful (artist blank; title = the raw filename), and a match fills it
with kanji/kana. query_page + pack_fields now surface the author's own romaji
parsed from the filename ("Junko Yagami — BAY CITY") when the pack has no
artist of its own — display-only, keyset-safe (raw title stashed for the
cursor), a real pack artist or a user override still wins.
- Smart adopt: "Use these values" now KEEPS the readable romaji name + title the
card already shows and takes only album/year/genre (+ art via the pin) from the
match, so identifying a Japanese song gives "Junko Yagami — BAY CITY — FULL MOON"
with the right cover, never native script.
Tests: romaji fallback fires for a blank-artist CDLC pack (grid + pack_fields
agree) and is left alone when the pack has a real artist.
librosa.sequence.dtw's default step sizes permit pure horizontal/vertical
moves; on songs whose chroma is self-similar for long stretches the flat
cost surface let the path collapse (minutes of score onto one audio frame),
so auto-sync produced monotonic-but-garbage sync points and the per-bar
warp imported charts badly out of sync while reporting success.
Use the standard music-sync step pattern [[1,1],[1,2],[2,1]] (local tempo
ratio bounded to 0.5x-2x), falling back to unconstrained steps if the
global length ratio makes it infeasible.
Validated on the reported song (138 BPM tab, YouTube audio): coarse points
now track 1:1, refine holds slopes 0.77-1.04, warped downbeats hit onset
peaks at 3.3x background energy.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Switching the active window / alt-tabbing away (most often on Windows) can
trigger a GPU context reset. The 3D highway's WebGL renderer had no
webglcontextlost handler, so a lost context was left to escalate into a
render-process crash -- matching the intermittent "randomly crashes when I
change windows" desktop reports.
The renderer now binds webglcontextlost/webglcontextrestored on its own WebGL
canvas (ren.domElement): the loss is preventDefault()'d so the browser keeps the
context restorable, draw() bails while the context is down so no GL work runs on
a dead context, and on restore the viewport is re-applied and rendering resumes
(Three re-uploads scene resources on the next frame). Listeners are removed in
teardown.
Root cause is a strong hypothesis -- the crash is intermittent and
unreproducible -- but the fix is low-risk and additive and closes a real gap:
there was no context-loss handling anywhere in the renderer.
plugins/highway_3d 3.31.2 -> 3.31.3. Tests:
tests/js/highway_3d_context_loss.test.js (source-contract, like the other
highway_* tests). The sibling keys_highway_3d / drum_highway_3d renderers share
the same gap -- follow-up in their repos.
Claude-Session: https://claude.ai/code/session_01UR2Cr7GEu3yMY7SrfxH6c1
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds a one-click preset to the shared "Highway String Colors" picker,
next to the existing Okabe-Ito "Colorblind-friendly" preset. Contributed
by a deuteranopic player who found the Okabe-Ito set still hard to
separate: it retunes the six main strings and keeps that set's 7/8-string
colors. Applies to both the 2D and 3D highways via the shared picker,
which writes the slot->hex map the renderers already consume.
Additive frontend-only change to HWC_PRESETS in static/app.js; the picker
UI and both highways pick it up automatically (the preset list is
generated from HWC_PRESETS and applied by id). All 20 highway
string-color JS tests pass; app.js syntax-checks clean.
Claude-Session: https://claude.ai/code/session_01UR2Cr7GEu3yMY7SrfxH6c1
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds Genre as a fifth Details field (edit / lock / revert / Yours-Pack
provenance), backed by the existing override store. To make it actually useful,
the genre FILTER and FACET now resolve the per-song override (effective genre =
override else scanned pack genre) — guarded so the common no-override case stays
on the plain indexed column — so a corrected/added genre is immediately
browsable. Genre stays a library-only overlay: it is NOT a write-to-file field
(split WRITE_FIELDS = the four file-safe fields from the five DETAIL_FIELDS), so
Write to file leaves the genre override in place and the copy says so. The
Match→Details bridge also carries a candidate's first genre.
Tests: effective-genre facet + filter, and that a value-less lock doesn't invent
an effective genre.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Connects the popup's two tabs. A match only improves the underlying canon+art;
by design it never silently re-titles the grid. This adds the explicit opt-in
path: each Match candidate (search or Identify-by-audio) gets a "Use these
values →" action that copies its title/artist/album/year into the Details tab
as pending (unsaved) inputs and lands you there for review — pinning the match
too so the art/canon follow. You then Save (overlay) or Write to file. Queue-
review candidates are unchanged (they still accept/pin on click).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(library): "Write to file" in the Fix-metadata popup's Details tab
Completes the confirmed edit model: Save keeps edits as a reversible display
overlay (files untouched); "Write to file" bakes the shown title/artist/album/
year into the pack itself via the existing POST /api/song/{fn}/meta (writes the
manifest, re-stats, coalesces a rescan). On a real file write the now-redundant
override values are cleared (locks kept) and the tab re-renders, so the fields
read from the file as "Pack". Loose-folder / unwritable packs fall back to a
DB-only update and say so (may revert on a full rescan). Secondary button next
to Save; touches only the four file-safe fields, the rest of the pack verbatim.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(library): mirror server year coercion in Write-to-file grid sync
update_song_meta coerces a non-numeric/empty year to "" before persisting,
but writeToFile optimistically set song.year to the raw typed text — so the
library card flashed e.g. "abcd" until the next natural refresh. Apply the
same integer coercion client-side so the in-memory song matches what was
written.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* feat(library): metadata override + lock store, enforced by enrichment (popup slices 1–2)
Backend foundation for the Fix-metadata popup. Not yet surfaced in the UI (the
display + 3-tab popup are the next slices); no PR until it's user-visible.
Slice 1 — the store:
- `song_field_override(filename, field, value, locked)` table: a reversible
DISPLAY overlay (never written to the pack), filename-keyed so it survives a
rescan (never purged by delete_missing) and is dropped only with the song.
- DB methods (partial upsert that drops empty+unlocked rows; batch map) +
`GET`/`PUT /api/song/{fn}/overrides` (field allowlist title/artist/album/
year/genre; clearing rides PUT since DELETE /api/song/{path} shadows sub-
routes; PUT demo-blocked).
Slice 2 — locks respected by enrichment:
- The auto-matcher composes a per-song `_compose_lock_filter` onto the global
apply-filter, so a match still applies IDENTITY (mbid/release → art) but never
re-canonicalizes a LOCKED display field.
- Gap-fill (write-to-file) skips locked album/year/genre — writing the matched
value would be exactly the clobber the lock exists to prevent.
- Review/manual picks bypass the filter (an explicit confirm overrides a lock).
Tests: store semantics + rescan-survival + API; the lock filter + reader; an
auto-match leaving a locked field un-canonicalized; gap-fill excluding locked
keys.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(library): show per-song overrides in the grid (popup slice 3)
The grid now displays the user's per-song title/artist/album/year override
in place of the pack value ("grid shows only overrides") — a matched
MusicBrainz canon never silently re-titles a card; canon stays in the
Details drawer + art. Overlaid in Python over the visible window, keyset-safe
like the P4 artist-alias re-label: the seek still runs on the raw column, and
the one overridable keyset column (title) stashes its raw value for the cursor
so paging never skips/dupes. The private stash is dropped from the payload.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(library): 3-tab Fix-metadata popup — Details / Cover art / Match (slice 4)
Turns the thin single-song fix-match modal into the Plex-style metadata
editor reached from a card's "Fix metadata…" menu:
- Details tab: type + lock the displayed title/artist/album/year. Values
ride the reversible override store (GET/PUT /api/song/{fn}/overrides); each
field sits on its pack value (Yours/Pack provenance + revert-to-pack), a lock
pins it against auto-match, and Save repaints the grid via library:changed
(slice-3 overlay). This is the real tool for the blank-artist city-pop pile
MusicBrainz can't surface — you just type the right title.
- Cover art tab: hands off to the shared image picker (its own modal); the
pick refreshes the thumbnail everywhere.
- Match tab: the existing MusicBrainz search + candidate/pick flow, refactored
into shared body/footer helpers (the queue-review flow is untouched).
Backend: GET /overrides now also returns the pack baseline so the Details tab
can pre-fill + show provenance. tailwind.min.css regenerated (build-tailwind.sh)
for the popup's new utility classes.
Identify-by-audio (AcoustID) is deferred: it lives in unmerged PR #759, off
main — the Match tab gains the button once #759 lands and this branch rebases.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(library): wire "Identify by audio" in the tabbed popup's Match tab
The AcoustID Identify button (#759) merged in referencing an out-of-scope
`panel` in the wiring — a leftover from the pre-popup fix-match modal that my
tab refactor renamed to `root`. Under strict mode that threw, so the handler
never attached and the button did nothing. Scope it to `root` (the tab body),
which is where the search-results area it renders into lives.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(library): make "Identify by audio" outcomes unmistakable
An empty AcoustID result read the same as a broken button. Each state now says
plainly which outcome it is — ✓ fingerprinted-but-no-match vs no-audio vs off vs
unavailable — and, in the popup, points at the manual fallback (Search, or set
the album in Details + cover in Cover art by hand). A ✓ marks the states that
actually ran, so "worked, found nothing" no longer looks like a failure.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(v3 library): persistent "no match" badge + Unmatched quick filter
The Refresh-Metadata batch (#764) shows a transient per-tile "no match" only
while a pass runs, so the unmatched pile goes quiet at rest. Two additions make
it visible + reachable:
- Persistent per-card "No match" badge: query_page now marks each row
`unmatched` (a cheap failed-set membership like favs/estd), and enrichBadge
paints a subtle resting marker for those cards — tracked in a `_unmatched` set
so a batch tile clearing falls back to it instead of wiping it. A live batch
tile still wins while a pass runs.
- "Unmatched" toolbar toggle (local-only): one click applies the same filter as
the drawer's Match → Unmatched (match_state='failed'), so the no-match pile is
a click away right after a batch. Re-queries + reflects active state.
Test: query_page flags a failed row + the match=unmatched filter returns it.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(v3 library): repaint persistent no-match badge after metadata tile-clear
_clearMetaTiles removed every .v3-meta-tile node — including the new
persistent 'No match' resting badge, which derives from _unmatched rather
than _metaTile. A metadata rescan's tile-clear therefore dropped the badge
until the next scroll re-rendered the card. Repaint it from _unmatched.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* feat(gp_autosync): piecewise time-warp helpers + refine_sync onset pass
auto_sync computes per-bar sync points but consumers only ever applied the
scalar bar-1 audio_offset, so any tempo drift between the recording and the
tab's authored tempo accumulated over the song. Add the librosa-free helpers
needed to apply the full piecewise mapping:
- bar_start_times(gp_path): per-bar score times sharing auto_sync's axis
(GPIF bar-resolution map, GP3/4/5 per-tick integration)
- build_warp_anchors(points, bar_starts): monotonic (score, audio) anchors
- warp_time(t, anchors): piecewise-linear map with edge-slope extrapolation
- warp_song_times(song, warp): retime a lib.song.Song in place (notes,
sustains, chords, beats, sections, anchors, handshapes, phrase levels,
tone changes, tempo overrides)
- gp_has_expandable_repeats(gp_path): detects GP3/4/5 repeat/volta/direction
markup whose playback expansion auto_sync's as-written points cannot map
Also implement refine_sync() — the editor's refine-sync endpoint has imported
it since the snapshot but it never existed in lib, so the Refine button 500'd.
It densifies the DTW points to every Nth bar and re-times each with a local
onset phase sweep (radius clamped under half a beat to avoid one-beat locks,
short scoring grid + median residual snap against the first beats). Synthetic
click-track validation: ~13ms mean / ~40ms max error from ±180ms coarse input
across 117-123 BPM recordings of a 120 BPM tab.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* review: Copilot round 1 — normalize bar_start_times GP3/4/5 parse failures to ValueError, document ImportError
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(enrichment): loose MusicBrainz search fallback (find aliased artists)
The MB text search used a strict field-phrase query
(`recording:"<title>" AND artist:"<artist>"`). A field phrase only matches
MusicBrainz's *primary* artist/title — it never searches ALIASES — so a
recording stored under a non-Latin primary name (大橋純子) whose romanized
form ("Junko Ohashi") is only an alias returns ZERO results, even though MB
has it. Whole swaths of a community library (e.g. romanized J-pop / city-pop
charts) were unsearchable.
- `build_recording_query(..., loose=True)` drops the field scoping + phrases
for plain AND-ed term groups (`(telephone number) AND (junko ohashi)`),
which searches the whole document incl. aliases.
- `_mb_search_recordings` runs the strict query first (unchanged, high
precision) and only on an EMPTY result retries once with the loose query —
so mainstream matches are untouched and the extra throttled request is spent
only on a miss. Results are re-scored by rank_candidates, so recall goes up
without lowering match quality (auto-accept still needs the per-field floors).
Verified live: "Junko Ohashi / Telephone Number" and "Anri / Windy Summer"
(both 0 under the strict query) now surface the real records; "AC/DC /
Highway to Hell" still hits strict at score 1.0 with no loose retry.
Follow-up (separate): alias-aware SCORING so these can auto-confirm, not just
appear as manual candidates.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(enrichment): alias-aware scoring (auto-confirm non-Latin-primary artists)
Builds on the loose-search fallback: that surfaces a recording stored under a
Japanese primary name (大橋純子) via its romanized alias, but the SCORER still
compared the reference ("Junko Ohashi") against the primary only → artist
similarity 0 → below the auto floor, so it could only ever be a manual
candidate, never an auto-fill.
- mb_match: `cand_artist_sim` takes the best similarity over the candidate's
primary name AND its `artist_aliases`; score_candidate + classify use it.
- server: `_mb_artist_aliases(id)` fetches an artist's aliases (one throttled
lookup, process-cached — a one-artist discography costs ONE request) and
`_alias_enrich` attaches them ONLY to promising near-misses (title agrees,
primary artist doesn't) so a normal pass spends zero extra requests. Wired
into both the auto-matcher (_enrich_one) and the manual search proxy.
Verified live: "Junko Ohashi / Telephone Number" → 大橋純子 candidate goes from
score 0.5 (loose-only) to 1.0 (auto-confirmable), ranked #1; "AC/DC / Highway
to Hell" unchanged at 1.0 with no alias lookup.
Stacks on #771 (feat/mb-loose-search-fallback).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(enrichment): keep live exclusion in the loose search fallback
The loose fallback dropped the strict path's -secondarytype:Live filter, so a
studio chart whose strict query missed could fall back to — and, since
score_candidate doesn't penalize live takes, auto-confirm — a live-only
recording. Apply the same live gate to the loose query (skipped only when the
source title is itself a live take).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* feat(enrichment): loose MusicBrainz search fallback (find aliased artists)
The MB text search used a strict field-phrase query
(`recording:"<title>" AND artist:"<artist>"`). A field phrase only matches
MusicBrainz's *primary* artist/title — it never searches ALIASES — so a
recording stored under a non-Latin primary name (大橋純子) whose romanized
form ("Junko Ohashi") is only an alias returns ZERO results, even though MB
has it. Whole swaths of a community library (e.g. romanized J-pop / city-pop
charts) were unsearchable.
- `build_recording_query(..., loose=True)` drops the field scoping + phrases
for plain AND-ed term groups (`(telephone number) AND (junko ohashi)`),
which searches the whole document incl. aliases.
- `_mb_search_recordings` runs the strict query first (unchanged, high
precision) and only on an EMPTY result retries once with the loose query —
so mainstream matches are untouched and the extra throttled request is spent
only on a miss. Results are re-scored by rank_candidates, so recall goes up
without lowering match quality (auto-accept still needs the per-field floors).
Verified live: "Junko Ohashi / Telephone Number" and "Anri / Windy Summer"
(both 0 under the strict query) now surface the real records; "AC/DC /
Highway to Hell" still hits strict at score 1.0 with no loose retry.
Follow-up (separate): alias-aware SCORING so these can auto-confirm, not just
appear as manual candidates.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(enrichment): keep live exclusion in the loose search fallback
The loose fallback dropped the strict path's -secondarytype:Live filter, so a
studio chart whose strict query missed could fall back to — and, since
score_candidate doesn't penalize live takes, auto-confirm — a live-only
recording. Apply the same live gate to the loose query (skipped only when the
source title is itself a live take).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
Every real Guitar Pro 6 (.gpx) file failed to import with
"GPX BCFS sector pointer out of range (malformed file)".
A real .gpx's BCFZ-declared decompressed size isn't 0x1000-aligned, so
its last (small) container file lands in a partial trailing sector.
_parse_bcfs raised whenever a sector read would run past the buffer
end, rejecting the whole container before score.gpif could be extracted
-- so no GP6 file could be charted in the song editor. (GP7/GP8 .gp
files take the ZIP path, not BCFS, which is why this wasn't caught
earlier.)
Clamp the final sector read to the buffer end (the per-file size field
trims the padding anyway), matching canonical GPX readers (alphaTab /
PyGuitarPro). A sector whose start is past the end still raises, so the
malformed-file guard is preserved.
Verified against two real GP6 files -- both now unpack to valid GPIF
with all tracks. Adds the previously-missing positive BCFS round-trip
coverage: partial-final-sector, multi-file, sector-aligned baseline,
and the preserved out-of-range guard.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
The virtualized v3 Songs grid rebuilt its entire visible window
(grid.innerHTML = _renderCardsRange(...) + a full wireCards pass) every
time it slid by one row. Each row-boundary crossing was therefore a heavy
synchronous frame — reparse ~60 cards, re-attach hundreds of listeners,
reflow — that stalled the main thread and buffered held-arrow key-repeats,
flushing them in a burst. Testers saw the library "go super fast for a
second then slow down," skipping "every so many scrolls," up or down, at
the same spots each time. It hitched scrolling back up over already-loaded
songs too, because the cost was DOM teardown, not fetching.
renderWindow() now reconciles the window in place: it reuses the card
nodes that stay on-screen and builds only the row that enters/leaves
(~6 nodes per slide instead of ~60). Nodes are keyed by absolute index
(data-idx) with a real-vs-skeleton + select-mode signature (data-sig) so
hole-fills after a page fetch and select-mode toggles still rebuild
exactly the nodes that changed. wireCards()'s existing data-wired guard
then wires only the freshly-built nodes, so per-slide listener churn drops
with it. Everything keyed off data-fn (favorites, ⋮ menu, right-click,
selection, accuracy badges, A–Z rail) is unaffected.
Follow-up to the stage-2 virtualized grid (#636 item 3). Frontend-only.
Tests: tests/js/v3_songs_window_recycle.test.js — window stays [start,end)
contiguous and in-window node identity is reused across a down-then-up
scroll; select-mode toggle and a rail-seek jump rebuild correctly.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
Adds a media-server-style "Refresh Metadata" control to the Songs toolbar
(beside "⟳ Refresh") — the metadata counterpart to a file scan.
- Re-matches the songs currently SHOWN (the visible grid window) against
MusicBrainz: a per-view refresh that's visible even on an already-matched
library. The button doubles as Stop while a pass runs; a batch progress bar
+ per-tile queued→working→done badges show what's happening. User-pinned
`manual` matches are never re-matched.
- Backend: POST /api/enrichment/{cancel,states,rematch}; /status gains
total/matched/current/cancelling; a cooperative cancel Event is checked
between songs in the match + art phases so Stop halts without waiting for
the whole queue. `states` is read-only (open); `cancel`/`rematch` are
demo-blocked.
- Matcher: when a pack's `artist` field is blank (common in community
charts), derive artist/title from the CDLC `Artist_Song-Title` filename
convention as a SEARCH SEED so text matching can identify it — the displayed
values still come only from the confirmed MusicBrainz match, nothing
estimated is shown as author-set. Rescues blank-artist packs that otherwise
always failed.
Tests: enrichment_states_for, the three new routes, cancel-halts-a-pass,
kick-clears-stale-cancel, filename parse, blank-artist seeding, and
present-artist-not-overridden.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(enrichment): AcoustID audio-fingerprint identification (opt-in)
Text search can only guess the version; the definitive fix is content-based —
fingerprint the actual audio with Chromaprint (fpcalc) and look it up on
AcoustID, which maps the fingerprint to the EXACT MusicBrainz recording (the
approach Lidarr uses). Sidesteps the studio-vs-live ambiguity entirely.
- lib/acoustid_match.py: pure response parsing + config gating (unit-tested);
normalizes AcoustID hits into the same candidate shape as mb_match so the
review UI + editor Match popup render fingerprint and text hits identically.
- server.py: _fpcalc (Chromaprint subprocess), _acoustid_lookup (throttled,
offline-guarded HTTP), _identify_by_fingerprint (also available to the
library-enrichment pipeline), and POST /api/enrichment/identify (upload the
master audio → candidates).
- Fully OPT-IN and graceful: absent the fpcalc binary or an ACOUSTID_API_KEY
the whole path is a no-op / 503 and the text matcher runs unchanged.
Requires (both optional): the `fpcalc` (Chromaprint) binary on PATH/$FPCALC,
and a free AcoustID application key in $ACOUSTID_API_KEY. Pure parsing/gating
is unit-tested; the fpcalc + live-lookup path needs those two to exercise.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(enrichment): make AcoustID self-serve — opt-in toggle + API key in settings
Fingerprinting was env-var only (ACOUSTID_API_KEY), so only an operator could
enable it. Add two core settings so a user can turn it on themselves:
- acoustid_enabled (bool, default OFF — opt-in)
- acoustid_api_key (string, ≤128 chars, trimmed; env var stays a fallback)
_acoustid_available()/_acoustid_lookup() now resolve (enabled, key) from
settings via _acoustid_settings(). /api/enrichment/identify distinguishes
"not set up" (412 needs_setup — the UI nudges the user to enable it) from
"set up but fpcalc/network missing" (503) so the client never fakes a match.
Verified: default off; POST round-trips + trims; 412 vs 503 gating; bad
types/over-length rejected. acoustid_match unit tests green (8/8).
* fix(enrichment): POST the AcoustID lookup instead of GET
A Chromaprint fingerprint is multi-KB (a 3.5-min track ≈ 3.5k chars), so
sending it as a GET query param overflows the request URL for longer songs and
fails spuriously. AcoustID accepts the same params form-encoded — POST them.
* fix(acoustid): space-separate the lookup `meta` (was silently dropping metadata)
The meta value was `+`-joined ("recordings+releasegroups+compress"). Sent over
the wire the literal `+` percent-encodes to %2B, which AcoustID does NOT split
into flags — so every hit came back with an empty `recordings` array and the
parser produced zero candidates (a fingerprint match that resolved to nothing).
AcoustID wants the flags space-separated. Verified against real fingerprints:
`+`-joined → 0 recordings; space-joined → 28, resolving Highway to Hell and
Living After Midnight to their canonical studio albums as the top hit.
* feat(acoustid): resolve the canonical original album + year from the fingerprint
AcoustID hits resolved the right recording but a weak album/blank year: the
album picker took the first studio-typed group (a later comp/soundtrack typed
"Album" could win) and the year took an arbitrary release (often a reissue).
Request the `releases` meta (which carries per-release dates) and use them to
(1) pick the EARLIEST original studio album among the groups and (2) fill the
year from that album's earliest release. Verified against real fingerprints:
Smoke on the Water → Machine Head (1972) not a later comp; Highway to Hell →
1979; Living After Midnight → British Steel (1980). +2 unit tests.
* feat(acoustid): per-song "Identify by audio" for the library metadata tooling
Add POST /api/enrichment/identify/{filename} — fingerprints an EXISTING library
song's own master audio (resolves the sloppak's original_audio or a loose
folder's audio), the library counterpart to the upload-based /identify used by
the editor. Wire an "Identify by audio" action into the match-review / Fix-match
modal: it renders fingerprint hits in the same candidate list and pins the pick
via the existing /review/{f}/pick. Shared _acoustid_gate() (412 needs_setup /
503) for both endpoints; 404 when a pack has no full mix. Both identify routes
added to the demo-mode block list (they spend fpcalc + the AcoustID budget) —
fixes a pre-existing miss on the upload route.
* fix(acoustid): regenerate stale tailwind CSS + cap identify upload
- static/tailwind.min.css was stale vs a fresh rebuild (ci/tailwind-fresh red);
regenerated with the pinned tailwindcss@3.4.19 (byte-stable).
- /api/enrichment/identify read the whole multipart upload into memory before
writing it; stream it to the temp file with a 256 MB cap (413 over) so an
oversized upload can't balloon RAM. fpcalc reads from the temp file anyway.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(acoustid): pre-parse upload guard + settings UI to enable it
- /api/enrichment/identify is now async: a pre-parse Content-Length check +
request.form(max_part_size=…) reject an oversized body BEFORE Starlette spools
the multipart to temp disk (mirrors the song-upload endpoint), and the blocking
fpcalc subprocess + AcoustID HTTP run off the event loop via run_in_executor.
- The v3 Metadata-matching settings card gains an 'Identify by audio' opt-in
toggle (acoustid_enabled, default OFF) + an AcoustID key input
(acoustid_api_key), wired in match-review.js — so the advertised feature is
reachable from the UI instead of only via a manual settings POST. Reuses
existing classes only; committed tailwind.min.css stays fresh.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(enrichment): rank the canonical studio take over live/comp versions
A flat MusicBrainz /recording text search ties every take of a song at the
same score, so "AC/DC — Highway to Hell" returns a wall of live bootlegs and
compilations with the 1979 studio version buried (or below the fetch limit).
- build_recording_query: drop live-ONLY recordings (`-secondarytype:Live`).
Compilations are deliberately kept — they REUSE the studio recording, so
filtering them cuts the very recording we want (verified against MB).
- _best_release / parse_recording_doc: pick the canonical studio album
(primary Album, no Live/Compilation/Remix/... secondary type) for the
displayed album/year, and expose a `studio` flag.
- rank_candidates: since the combined score caps at 1.0 (perfect text match
ties), break ties on the studio flag and — when the caller knows the audio
length — on duration proximity, so the studio take wins over live/extended
cuts. The studio distinction is intentionally NOT scored (a live take is
still the right SONG), only re-ordered.
- /api/enrichment/search: accept an optional `duration` param so a caller that
has the audio but no library row (the editor's create modal) can pass the
master-track length for the duration tiebreak.
Verified end-to-end against live MusicBrainz: AC/DC "Highway to Hell" now
returns the 1979 studio recording at #1 with the correct album + year.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(enrichment): official releases outrank unofficial studio albums
_best_release sorted (clean, status_ok, date), so an UNofficial bootleg Album
outranked an official Single/EP/comp — regressing canonical album/year and
seeding cover-art from a bootleg for single-only songs. Order status_ok before
clean: official first, then prefer a clean studio album among the official
releases (still surfaces the studio album over an official live/comp album).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(enrichment): keep live recordings for genuinely-live charts
build_recording_query unconditionally added -secondarytype:Live, but denoise()
strips a '(Live at …)' qualifier from the query — so a chart that IS a live take
had its only correct recording filtered out (both background enrichment and
manual search). Skip the live filter when the source title carries a
parenthetical live marker; a bare title word ('Live and Let Die') still filters.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(enrichment): drop the studio tiebreak when the chart is a live take
Follow-through on keeping live recordings for live charts: rank_candidates still
ranked the studio take ahead of a tied live one, so a live chart would auto-match
the studio recording. Skip the studio tiebreak when the source title has a live
marker — duration proximity + score then pick the right live version.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* settings: add host instrument profiles
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
* settings: add instrument pathway selection
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
* fix(settings): profile-aware saves/resets/switch, provider tunings, bass-5
Five regressions from the instrument-profiles rework:
1. save_settings canonicalized profiles on EVERY save -> empty/unrelated POST
froze default profiles into config.json (broke
test_empty_post_preserves_all_existing_keys). Gate on the save touching
instrument settings; GET already virtualizes profiles.
2. pathway is profile-mirrored, so the Gameplay reset (flat-key delete) was a
no-op. reset_settings now resets pathway inside the persisted profiles too.
3. Per-profile tuning validation rejected provider/custom tunings (tuner
plugin, /api/tunings). _valid_tuning_for_key now accepts a name unknown to
every built-in table while still rejecting a built-in misapplied to the
wrong key.
4. First-migration overwrote an explicit active_instrument_profile with the
legacy-inferred one, so a fresh-config switch to 'bass' was lost. Use
setdefault so an explicit request wins.
5. Pre-existing test_instrument_fields_persist used bass-5 + 'Drop D' (a
4-string tuning). Updated to the valid 'Drop A'.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(settings): partial-merge instrument_profiles; clamp tuning on string-count switch
Two partial-update follow-ups:
- save_settings normalized a POSTed instrument_profiles by FILLING every omitted
profile with defaults and replacing wholesale, so a one-profile update reset
the others. Validate each PROVIDED profile individually and merge the partial
over the persisted set inside the lock — /api/settings is partial-merge.
- the string-count picker posted only string_count, so the backend silently
reset a now-invalid tuning to Standard while the UI kept the old value
(settings/tuner desync). Clamp + post the valid tuning too, mirroring the
instrument-switch path.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Signed-off-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(library): serve art/load songs mounted through a library junction
A song library mounted through a directory JUNCTION/symlink subfolder (a
library shared across app installs; the desktop app's own mounts) had broken
album art and couldn't load: the scanner's rglob follows the junction and
indexes the songs, but _resolve_dlc_path (via safe_join's .resolve()) followed
the junction to its real target, saw it outside DLC_DIR, and rejected every
song reached through it → 403 on /art, 404 on /art/candidates, broken covers.
- _resolve_dlc_path now uses LEXICAL containment (os.path.normpath, no symlink
following) so an in-library junction is allowed, while `..` traversal and
absolute paths are still rejected (the traversal tests pin this).
- safe_join is left STRICT (.resolve()-based) — it is the zip-slip / plugin-
asset / avatar guard, where following a symlink out IS the defense — but
gains an explicit NUL guard (on Python 3.13/Windows resolve() no longer
raises on an embedded NUL, so the byte was leaking through; strictly-more-
rejection, no effect on the zip-slip contract).
Tests: test_dlc_junction (junction allowed; `..`/absolute/NUL rejected; the
safe_join-stays-strict contrast). Existing traversal/safepath/art-candidates
suites stay green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(library): reject Windows drive-letter paths in _resolve_dlc_path
The new test_absolute_path_rejected pins 'C:/Windows/system32/x' → None, but on
POSIX a drive-letter path isn't absolute, so Path(dlc)/'C:/…' becomes the
contained relative dir '<dlc>/C:/…' and slipped through the lexical containment
check (red on the Linux CI). Not an escape, but the traversal contract should
hold cross-platform (a shared library is reached from either OS). Reject a path
that is absolute or drive-qualified in either POSIX or Windows semantics before
the containment check. Legitimate relative/junction paths are unaffected.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two commits (delete beethoven-ode_to_joy, re-add The Adicts' Ode to Joy) never
updated _BUILTIN_STARTER_SOURCES: it still listed the deleted pack and omitted
the added one. The listed-but-missing file made the all-present gate never fire,
so NO starter content seeded on first run — and the on-disk-but-unlisted pack
would bundle as dead weight. Both starter-seed guard tests were red on main,
reddening ci/test on every core PR. Sync the manifest to disk.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The previously pinned BtbN autobuild release (2026-06-19) was pruned
upstream, so the release build's curl download 404'd (exit 22).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Add two more public-domain starter songs alongside Für Elise, wired into
_BUILTIN_STARTER_SOURCES so they seed into DLC_DIR/starter/ on first run:
- The Star-Spangled Banner (lead) — John Stafford Smith; cleaned the "Unknown"
artist / placeholder album, author "Fee[dB]ack".
- Ode to Joy (lead/rhythm/bass + drums) — Beethoven. Replaces the raw
"Ode to Joy (VST Cover)_The Adicts.feedpak" that was committed to main but
never added to the seed list (so it bundled 23 MB of dead weight and never
appeared). Fixed metadata (artist Beethoven, year 1824, author "Fee[dB]ack"),
and repointed the stem from the 22 MB editor WAV to the byte-identical-render
full.ogg (both exactly 85.324 s), shrinking the pack 23.8 MB -> 1.7 MB.
Add guard tests asserting every _BUILTIN_STARTER_SOURCES entry has its file
committed and that a seed run lands them all — this catches exactly the
listed-but-missing (or committed-but-unlisted) mismatch that left Ode to Joy
un-seeded.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Ship a public-domain Für Elise (keys) feedpak as starter content so a fresh
install isn't an empty library. server._seed_builtin_starter_content() copies
bundled packs into DLC_DIR/starter/ exactly once, guarded by a marker in
CONFIG_DIR — unlike the always-reseeding diagnostic seed, a user who deletes
the starter song does not get it back. `starter/` is deliberately outside the
diagnostics/tutorials library carve-out so the song surfaces as a normal
library entry.
Extract the shared symlink-safe, mtime-aware copy loop into
_copy_builtin_packs() and route both the diagnostic and starter seeds through
it (diagnostic behavior unchanged; existing tests green).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(plugins): don't treat transient absence from /api/plugins as uninstall
The backend clears its plugin registry at the start of load_plugins()
and repopulates it incrementally while HTTP stays up, so every backend
restart (desktop: Audio Quality soundfont switch, LAN toggle, update
restart) serves a window of partial — even empty — /api/plugins
responses. loadPlugins() treated absence from the current response as
an uninstall, with three destructive consequences for still-loaded
plugins:
1. Their settings-panel and screen DOM were wiped while their
_loadedPluginScripts entry survived, so the NEXT refetch failed the
DOM-existence check and re-evaluated the plugin's screen.js
mid-session. For the desktop audio_engine plugin that re-ran init()
against the surviving native audio chain and exactly duplicated
every VST/NAM/IR stage (the alpha testers' "chain duplicates after
leaving the Audio menu" / blown-out gain reports).
2. _reconcilePluginStyles dropped their stylesheet, leaving them
visible but unstyled until they reappeared.
3. The stale-contribution sweep unmounted their UI contributions and
unregistered their capability participant with no re-registration
path (plugin scripts don't re-run thanks to the loadedScripts
guard).
Absence is now a non-signal everywhere in loadPlugins: the DOM wipe and
style reconcile are scoped to plugins the response actually names, and
the absence sweep is removed. Present plugins still fully re-sync via
_registerLegacyPluginUiContributions each round; failed plugins are
present in the response and still cleaned up; nav is rebuilt from the
response so genuinely uninstalled plugins drop out of it, and their
(un-unloadable) already-evaluated scripts keep their DOM until reload.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* test: update idempotence contract to the absence-is-not-uninstall invariant
The removed-plugin sweep contract pinned the old behavior this branch
deletes; pin the new invariant instead (no absence sweep + respondedIds
scoping on the DOM/style reconcilers).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
main's JS suite has been red since the recent v3-library and player
refactors landed. 17 of 18 failures were test harnesses/regexes that
went stale behind real, intentional code changes; one was a genuine
contract violation in the code.
Code fix:
- session-resume seek passed 'resume' as its _audioSeek reason; the
documented contract (enforced by song_seek.test.js) requires
multi-word kebab-case. Renamed to 'session-resume' — no consumer
string-matches specific reasons, so this is rename-safe.
Test updates (each pins the CURRENT contract):
- highway_colors_facade: inject HWC_PRESETS + applyHighwayStringPreset
(new preset feature); lock presets/applyPreset into the surface test
- loop_api: stub _updateEditRegionBtn (new edit-region UI hook)
- song_close: sandbox gets window.feedBack.playQueue; assert a real
close abandons the queue (the new queue-aware behavior)
- v3_keep_practicing: the shelf moved from client-side /api/stats/recent
dedupe+gating to the server-side practice-suggestions recommender —
tests now pin that (fetch, arrangement-aware card click, Promise.all)
- v3_songs_tuning: card row variable renamed song → shown (grouped cards)
- live_guitar_tone_source: accept literal ’ where ’ drifted in copy
- legacy_shim_hits: normalize CRLF before fixed-width region() slicing
(Windows-only failure; char windows shrank by one char per line)
Suite: 987/987 locally (Windows), previously 968/987 (and 18 red on CI).
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Off state had a 1px border, on state none — toggling grew/shrank the
button 2px and shifted the row. On state now carries a same-color
border (border-fb-primary, already in the prebuilt CSS).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
w-4 icon (16px) vs text-sm line-height (20px) made the shuffle button
4px shorter than its neighbor at equal py-2.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The repo is 'got-feedback/feedBack' (capital B) after the rename, so ${GITHUB_REPOSITORY} produced an invalid Docker tag ('repository name must be lowercase'). Use ${GITHUB_REPOSITORY,,}. nightly/rc already hardcode lowercase 'feedback'.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Crossing-arrows toggle next to Play all / Play album on the playlist
detail page. When on, playQueue.start Fisher-Yates-shuffles the queue
once at start — on a copy, so the stored playlist order is untouched —
swapping per-slot album arrangements in lockstep so each slot keeps its
pinned arrangement (#685 contract preserved). Prev-less queue semantics
are unchanged: auto-advance simply walks the shuffled order.
Preference is global, persisted as localStorage v3PlaylistShuffle.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two pre-existing failures the segfault had been masking (the run aborted at ~25%, so they never ran until #735 let the suite complete):
1) Tuner group (~24): plugins ship a bare-named routes.py, so sys.modules['routes'] leaked between plugin test dirs (achievements ran first, tuner got its module). Each plugin conftest now pops the stale 'routes' and an autouse fixture binds sys.modules['routes'] to that plugin's module for the duration of its tests (covers runtime 'import routes' in test bodies).
2) Diagnostics group (5): _SONG_FILENAME_RE never matched the tests' .feedpak/.archive filenames — it also lacked 'feedpak' (the current primary format), a real redaction gap. Added feedpak to the regex and switched the tests off the fake .archive to the real .feedpak. Verified: full suite 2183 passed, 0 failed.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Root cause of the flaky pytest segfault (exit 139): the background scan and enrichment daemon threads (_scan_runner/_enrich_runner) use the shared MetadataDB connection, but test fixtures closed that connection in teardown without stopping them. A daemon thread mid-query on a freed SQLite conn is a native use-after-free → SIGSEGV. The app's startup kicks a scan, so almost any app-booting fixture was vulnerable. It only surfaced now because got-feedback/feedBack#728 added a push trigger, so ci/test runs on every push to main.
Fix: server.py retains the scan/enrich thread handles and adds _join_background_db_threads(); every test fixture now joins the workers before conn.close(). Verified: the full suite runs to completion (no segfault) where it previously crashed at ~25%.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replaces the fee[dB]ack text wordmark in the #v3-brand sidebar header with the exported PNG logo (static/v3/brand/feedback-logo-light.png, 664x165). Sized width:100% + height:auto so it fits the 256px sidebar's content width (~208px inside the p-6). Updated both the no-JS fallback (index.html) and the shell.js boot render. Inline style avoids introducing a new Tailwind utility (constitution P-II).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Nightly builds main directly (old release/v* discovery pinned nightlies
to shipped branches forever). ship-ci adds push triggers on main and
release/** for post-merge signal. New rc.yml builds :rc images from
release branches during stabilization.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* v3 library: artist pages — in-your-library view, similar-in-library, links-only web
The greenlit artist-pages feature. Every page renders from LOCAL data;
an optional, opt-in external-links strip is the only network surface.
Server:
- artist_enrichment table (mb_artist_id PK, url_rels JSON, genres JSON,
fetched_at) — never purged; one row per matched MusicBrainz artist.
- GET /api/artist/{name}/page — all-local: canonical name (+ raw alias
variants), song/album/mastered counts, album list, similar-in-library
(top artists by shared genre, self excluded, empty is fine), and the
artist's MB id when any matched/manual song carries one. THE DENOMINATOR
LAW: "N mastered" counts songs you OWN (best_accuracy >= 0.9 across the
artist's library songs), never a global discography — a stored score for
a song no longer in the library does not count.
- GET /api/artist/{name}/links — lazy, cached-forever: returns the cached
row, else (external links enabled + network + a known MB artist id) ONE
throttled artist lookup (inc=url-rels+genres+tags), whitelisted into
{official, tour, video, social[], wikipedia}. Every URL passes the same
http(s) scheme gate as art redirects, so a hostile javascript:/data:/file:
can never reach an href. POST .../links/refresh re-fetches. Offline /
no-mbid / links-disabled → empty, no error. Both routes demo-blocked.
- Settings keys artist_pages_enabled (default ON — local-only) and
artist_external_links (default OFF — opt-in per the dev-chat thread).
Frontend (static/v3/songs.js): an in-place sub-render mirroring openAlbum()
with a "← Song Library" back + scroll restore. 2x2 album-art mosaic header
(borrows the playlist-cover renderer), canonical name + "also shown as"
variants + a Matched·MusicBrainz pill when known; stats strip that omits
the mastered segment at zero (invitational, never "0 mastered"); Play all /
Shuffle (playQueue) + Save as smart playlist (collections rule {artist});
album rail → openAlbum; song list via the artist filter + wireCards;
"Similar in your library" chips → open that artist; and the external-links
row under an "On the web · opens your browser" divider, each link
target=_blank rel=noopener noreferrer with its domain shown — rendered only
when external links are on AND links exist. Empty modules hide.
Entry points: card ⋮ "Go to artist", the grid card artist line, and a
"View artist page" link in the Details drawer — all via
window.__fbOpenArtistPage.
Tests: tests/test_artist_page.py — page counts/albums/alias folding, the
denominator law (owned-only, best-across-arrangements), similar ranking +
empty, mb-id only from matched rows, links whitelist + scheme gate (a
javascript: and an ftp:// URL both rejected), disabled-by-default no
network, cache-no-second-fetch, refresh, demo block. 21 pass (35 with
artist_alias). node --check clean; tailwind rebuilt.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* fix(v3 artist pages): unfiltered album view + select-mode row guard on artist page
Artist-page album click no longer applies the global library filters: openAlbum
gains an ignoreFilters option that builds the /api/library request scoped only to
artist+album (no drawer/genre/tuning/search params), so the album view and
Play-album match the artist page's full-shelf counts. The normal albums-view
click path is unchanged (ignoreFilters defaults off).
Select-mode row clicks on the artist page now toggle selection instead of playing.
Extracted the grid/tree capture-phase select guard into a shared bindSelectGuard()
and attach it to the persistent artist-page host too. Each host is bound once at
shell build; innerHTML re-renders reuse the same element, so there is no
double-binding.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* v3 library: multi-candidate cover picker — select from list, media-server style
Auto-best already ships; this adds the "pick from a populated list" surface
Christian asked for, as ONE reusable component (song covers now; album/artist
art reuse the same picker later).
Server:
- GET /api/song/{filename}/art/candidates — assembles WITHOUT hoarding: the
Current image + its provenance, Pack art when present, and Cover Art Archive
candidates for the matched release (and any release ids stored on a review
row's candidates). New _caa_release_index() fetches the CAA release INDEX
json (image list + types + thumb sizes) through the existing throttle +
offline gate, cached as caa_index_{id}.json beside the covers (indexes are
stable). Capped at 12; fetched on demand. Demo-blocked (it spends the rate
budget) and offline → instant tiles only, no error.
Frontend: new static/v3/image-picker.js — window.__fbOpenImagePicker({filename,
title}), a body-appended singleton modal (match-review anatomy: overlay, focus
trap, Esc). Current image + provenance badge on the left; a tile grid on the
right whose instant tiles — Current, Pack original, Upload, Paste URL — work
immediately even offline, while CAA candidates load behind ONE /art/candidates
fetch with skeleton tiles + a "the source is rate-limited" caption. The fetch
is tied to an AbortController and cancelled when the modal closes.
Applying a pick reuses EXISTING routes so there's no new write path and the
design's key trick holds: a chosen cover POSTs to …/art/url (the override
lane — never evicted by the art-cache LRU, survives a re-match); "Pack
original" DELETEs the override; Upload POSTs …/art/upload (GIF stays
upload-only + local-only). Silent-on-success; the drawer/card art refreshes
via the existing cache-buster.
Entry points: the Details drawer art click (the old direct file dialog is now
the Upload tile) and a card ⋮ "Change cover…" action.
Tests: tests/test_art_candidates.py (matched row lists index images; review
row pulls in candidate releases; unmatched/offline → instant tiles only;
index cached, no second fetch; demo blocked) over a fake index seam.
30 pass with test_art_layer green (same seams). node --check clean; tailwind
rebuilt for the new file.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* fix(v3 cover picker): uiPrompt over window.prompt, visible-only focus trap, gate CAA to matched rows, index-cache lock, abort-on-reopen; +traversal tests
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* v3 library: first-hour polish — zero-states, match progress, provenance, alias search
Six launch-eve fixes for a brand-new user's first hour with a fresh,
being-matched library. Each is small and reuses shipped idioms.
- Invitational repertoire meter: with no practice data yet, the home meter
no longer reads "0 of N mastered" (debt framing) — it shows an empty bar
with "grows as you master songs". A count of 0 read as failure on day one.
- "Start here" starter shelf: growth_edge_suggestions() distinguishes two
empties — attempts exist but all mastered (honest empty shelf) vs nothing
attempted yet (day one) → new starter_suggestions() returns up to 8
approachable songs (90–480s, shortest first) flagged starter:true, and the
client renders a "Start here" shelf instead of a blank home.
- Library-visible match progress: while the background pass runs, a quiet
"Matching your library — X of Y" line sits by the review chip (5s poll,
single guarded interval, cleared the moment the pass stops — no leak,
no toast, silent completion).
- One-time transparency toast: the first time an install is seen matching a
real library, one fbNotify names what's contacted (MusicBrainz / Cover Art
Archive), that results are stored locally, that files aren't changed
without you, and where the switch is. localStorage-gated, wrapped so a
blocked notifier can't break the chip.
- Empty-library dead-end card: a genuinely empty local library (no songs, no
query/filter) shows "Your library is empty" + drop-files hint + Open
Settings, instead of a bare grid under dead dropdowns.
- Alias-aware search: searching a canonical name ("AC/DC") now also finds
songs whose raw tag is a merged variant ("ACDC"), via the artist_alias
table. Probe-guarded so a no-aliases library keeps the exact original
3-term query; pure predicate, keyset-safe.
- Details-drawer provenance line: matched/manual rows show "Matched:
<artist — title> (source) · Fix match" under the Identity fields — the
wrong-match escape hatch at the point of the data, wired to the same
fix-match flow the card menu uses. New read-only GET
/api/enrichment/song/{filename} backs it.
Tests: tests/test_starter_suggestions.py (starter vs normal-shelf behaviour,
length window, attempts-exist path unchanged) + alias-search cases added to
tests/test_artist_alias.py. 34 targeted pass; node --check clean; no new
Tailwind classes.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* fix(v3 library): stop match-progress poll when leaving the library screen
The 5s enrichment poll (_pollTimer) was cleared on pass completion and on
fetch error, but not when the user navigated away from the library. Leaving
v3-songs mid-pass left the interval pinging /api/enrichment/status in the
background until the pass ended. Subscribe to the existing feedBack
'screen:changed' event: clear the poll when any non-v3-songs screen shows,
and refresh (re-arming if a pass is still running) on returning to v3-songs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
Funding was cleared to come back online 2026-06-30 (offending
functionality fully removed). Restore the v3 topbar donate button
(hand-re-applied revert of cad7885 — the topbar was refactored since,
so this re-adds the Support Us! anchor alongside the new v3-search-wrap)
pointing at the feedBack-branded Patreon page
https://patreon.com/got_feedback.
Rebuild static/tailwind.min.css: the button's utilities
(bg-fb-accent, hover:bg-red-600, shadow-fb-accent/20, sm:inline-flex)
were purged when the button was removed and are needed again.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* library: scraper options — per-source + per-field auto-apply toggles, review-queue order (R1)
Grows the Settings→Library "Metadata matching" card into the full
scraper-options panel — not everybody needs the same things out of a
scraper:
- Sources: enrich_src_musicbrainz gates the background matcher (phase 2;
identity hashes still stamp, and manual Fix-match/search stays
available — same contract as the master toggle). enrich_src_caa gates
the Cover Art Archive fetch (phase 3). Rows skipped by an off toggle
stay unevaluated, so re-enabling picks them up on the next pass —
nothing is permanently forfeited.
- Auto-apply fields: enrich_apply_names/year/genres filter what an
AUTOMATIC match may canonicalize (_enrich_field_filter, applied on all
three automatic paths: cache copy, mbid/isrc exact keys, text auto).
MusicBrainz ids always stamp — they're identity, not display; the art
fetch and future re-matching need them. A match the user confirms in
the review modal applies in full. enrich_apply_art gates the art fetch
alongside the CAA source toggle (two axes, one behaviour today —
future art sources slot in without re-teaching the panel).
- Review queue order: enrich_review_order = missing_first (default,
today's behaviour) | artist | recent, read by GET /api/enrichment/review;
unknown stored values degrade to the default.
- Settings card: Sources / Auto-apply / Review-queue-order groups wired
in match-review.js; the master toggle is relabelled "Match songs
automatically" so it doesn't read the same as the new MusicBrainz
source toggle. No tailwind rebuild needed — every class was already
scanned from core source.
Tests: tests/test_scraper_options.py (9) — settings validation,
MB-source-off stamps-without-matching + re-enable, per-field stripping
on auto matches with ids preserved, review-accept full-apply despite
toggles, CAA gating on both axes, review-order modes incl. the
unknown-value fallback. Full-suite failure set A/B-identical to the
base (39 env/pre-existing).
Stacked on feat/enrichment-art (#715) — merge that first.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* library: per-field auto-apply honours "nothing forfeited" (backfill + no partial seeding)
The R1 per-FIELD auto-apply toggles settled a `matched` row with the
disabled fields stripped, but enrichment_pending() never revisits an
unchanged-hash matched row — so re-enabling a field never backfilled it,
and enrichment_cache_lookup() (which gates only on mb_recording_id) could
seed sibling charts with the stripped blanks. This broke the same
"nothing is permanently forfeited" contract the source (mb_on) and art
(art_on) toggles already keep.
Fix: persist an `apply_mask` marker (sorted blocked apply-keys) on every
AUTOMATIC match:
- migration: additive `apply_mask TEXT` column (idempotent ALTER).
- enrichment_pending(allowed_keys=...): re-queues a `matched` row whose
apply_mask names a field that is now re-enabled → backfill on re-enable,
converges (a fully-applied row is never re-queued).
- enrichment_cache_lookup: only fully-applied donors (apply_mask empty/NULL)
may seed siblings; a partial row is skipped and the sibling falls through
to its own re-filtered match.
- _enrich_apply_mask()/_enrich_blocked_apply_keys() helpers; threaded through
_enrich_one → apply_enrichment_match. Review/manual writers leave it NULL
(a confirmed pick applies in full).
Tests: re-enable-backfills-and-converges; partial row is not a cache donor
(fully-applied one is). 13 scraper-options tests pass; 170 enrichment/
settings tests green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: ChrisBeWithYou <christian.a.cowan@gmail.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* library: cover-art layer — CAA auto-fetch + user overrides, GIF local-only (R3/P9)
Third slice of the enrichment series (stacked on the matcher): covers.
- Serve chain for GET /api/song/{fn}/art: USER OVERRIDE -> PACK ART ->
COVER ART ARCHIVE cache -> 404. Behaviour change, deliberate: a
user-uploaded cover now OVERRIDES pack art (previously the upload only
filled the no-art gap, which made custom art look broken on any song
that already shipped a cover).
- GIF is allowed as an override and kept VERBATIM (animation intact) —
a local-only bonus. Everything else normalizes to RGB PNG as before.
One override per song (saving either kind removes the other), and
nothing ever writes art INTO a pack file — test-pinned: the pack's
cover.jpg is byte-identical after a GIF upload.
- Art by URL: POST /api/song/{fn}/art/url fetches server-side (http(s)
only, 10 MB cap enforced while streaming) into the same override slot.
DELETE /api/art/{fn}/override drops it — under /api/art because the
greedy DELETE /api/song/{path} catch-all shadows anything beneath it
(the same dodge the chart split/unsplit routes use).
- Cover Art Archive fetch as phase 3 of the enrichment pass: matched
songs that LACK pack art get their release's front cover, throttled +
identified + offline-guarded exactly like the MusicBrainz client
(pytest can never reach the network; a transport error pauses the
pass without burning the row). The cache is keyed by RELEASE MBID —
ten charts of one album cost one fetch — and every outcome writes an
art_state (pack/user/caa/none/error) so a row is evaluated once.
- LRU cap (200 MB) on the CAA side of the cache only; user overrides
are never evicted, and evicted rows reset so a later pass may
re-fetch. Deleting a song removes its override files (CAA files stay
— they may be shared by other charts of the release).
No frontend changes: the grid, the review modal, and the player pick
the new art up through the same route they already use. The
upload/paste-a-link surfaces in the Details drawer land with the
context-menu slice once the drawer PR merges.
13 new tests (tests/test_art_layer.py) + demo-mode routes; full-suite
failure set byte-identical with the change stashed vs applied.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* library: harden cover-art layer — SSRF guard, demo/size caps, override-delete state reset
Follow-up hardening on the R3 cover-art layer:
- remove_song_art_override: reset the enrichment row (set_enrichment_art(fn,
None, None)) when an override is deleted, so a row previously settled as
'user' re-queues and the CAA fallback resumes. Previously a removed override
stranded the row (enrichment_art_pending only re-queues art_state IS NULL),
leaving the song with no art at all.
- Base64 art upload: block it in demo mode (was open — a write/disk-fill vector,
worse now that GIFs are stored verbatim), validate the filename resolves to a
real song (mirrors the url route), and cap the decoded payload at 10 MB.
- Art-by-URL: reject hosts that resolve to loopback/private/link-local/reserved/
multicast/unspecified addresses (SSRF, e.g. cloud metadata) and stop following
redirects (allow_redirects=False) so a redirect can't smuggle the request to an
internal target. Fails closed on unresolvable/unparseable hosts.
- _caa_http_get: stream with a per-file 10 MB cap (bounds any one response
independently of the aggregate LRU); guard release_id against a conservative
token before interpolating it into a cache-file path (no separators/dots).
Tests: delete-override→CAA-fallback, upload unknown-song/oversize rejection,
SSRF internal-host guard, and a demo-mode block assertion for art/upload.
Note: art_state='error' rows are intentionally not auto-retried — there is no
per-row attempt counter on the art side, so an unbounded retry could storm CAA
for permanently-bad rows; a bounded retry would need extra state, left out here.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* library: opt-in gap-fill — write confirmed missing metadata into the pack (R4a)
The write-back contract agreed with the spec chair (alignment doc §7),
made executable, now that feedpak-spec 1.14.0 (mbid/isrc) is merged:
opt-in + user-initiated, adds ABSENT keys only, spec'd-keys allowlist,
values only from a CONFIRMED identity, atomic write + .bak. Single-song
only — batch write-back stays an open question with the chair.
- songmeta.gap_fill_sloppak: append-only manifest writer. Every added
key is absent by definition, so the new lines are APPENDED — the
author's existing bytes (key order, comments, formatting) survive
verbatim, unlike the metadata editor's full re-serialize. Directory
form gets a one-time manifest.yaml.bak + temp + atomic replace; zip
form reuses the editor's backup/temp/replace rewriter. Raises on any
already-present key: the never-clobber rule lives in the writer, not
just the callers.
- GET /api/song/{fn}/gap-fill: read-only preview — which of
album/year/genres/mbid/isrc are missing from the file (absent or
empty; year 0 = empty), with the values the enrichment match
supplies. Only a CONFIRMED identity is eligible (matched or a user
pin); review-tier rows are refused until a human confirms —
wrong-match > fast, same as everywhere else in the enrichment layer.
- POST /api/song/{fn}/gap-fill {keys}: writes the user-confirmed
subset. Proposals are RECOMPUTED under _song_io_lock, so a key that
gained an author value between preview and confirm is skipped, never
replaced. mbid/isrc written in canonical form only (validated).
DB stays scanner-consistent (album/year/genre columns + mtime/size
re-stat, cache invalidation + scan kick — the metadata editor's
contract). Demo mode blocks the write.
- Details drawer (Identity section): "Write missing info to file…" →
per-key checkbox confirm ("Only adds what's missing — nothing already
in the file is changed. A backup (.bak) is kept.") → written
confirmation; not-eligible states explain themselves. v3 only; no
new tailwind classes.
- Rides along: _manifest_exact_ids now strips ISRC display separators
(spec 1.14.0's strip rule) — a hand-authored "AU-AP0-90-00045" hits
the exact-match tier instead of silently falling back to text.
Tests: tests/test_gap_fill.py (10) — preview eligibility incl.
review-refusal + empty-as-gap, author-bytes-preserved-verbatim on dir
AND zip (with .bak content pinned), skip-not-replace on the mixed
request, the writer's ValueError guard, key validation, demo block,
DB sync; +1 hyphenated-ISRC test in test_mb_enrichment.py. 46 targeted
green; full-suite failure set A/B-identical to the main base (39
env/pre-existing).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* gap-fill: align preview with append-only writer (no cleared-value 500)
The R4a preview offered present-but-empty manifest values (album: '',
genres: [], year: 0) as gaps, but the append-only writer's never-clobber
guard raises on ANY key already present — so a user-confirmed POST for
those keys turned into a 500 "write failed" instead of filling the gap.
Appending can't fill an empty-but-present key anyway (it would duplicate
the YAML key).
Fix: _gap_fill_manifest_absent now treats only genuinely-MISSING keys as
gaps; a present-but-empty value is left to the metadata editor (which
re-serializes and can replace in place). This closes the preview→POST
mismatch — the preview never offers what the writer would refuse.
Tests: test_preview_treats_empty_values_as_gaps replaced by
test_preview_excludes_present_but_empty_keys (present-but-empty not
offered; genuinely-absent still offered) + test_write_present_but_empty_
key_is_refused_not_500 (POST → clean 409, file untouched, no .bak; a
genuinely-absent key alongside still writes). Closes the write-path blind
spot in the original empty-value test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
write_song_metadata dispatched zip-form packages on `suffix == ".sloppak"`
only, while core reads both suffixes everywhere else (sloppak.SONG_EXTS,
.feedpak being the current write extension). Editing a zip-form .feedpak's
title/artist/album/year therefore silently fell back to a DB-only update,
which looked fine until the next full library rescan re-derived metadata
from the file and reverted the edit — the exact failure this module exists
to prevent. Directory-form packages were unaffected (manifest-presence
dispatch, not suffix).
Gate on SONG_EXTS, add TestWriteSongMetadata regression coverage (both zip
suffixes, mixed-case suffix, directory form, unknown-suffix fallback), and
correct the stale scan_worker comment claiming .sloppak-suffix-only
detection (the code already accepts both via is_sloppak).
Claude-Session: https://claude.ai/code/session_01H1ZBEcZoJinde9ms5fAjwc
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* v3 library: artist sort orders titles within an artist (tree-view feel)
Tester report: "the list is set up by artist, but the cards are
alphabetical(-ish random)". Real: the tree orders artist -> album -> title,
while the grid's artist sort ordered within an artist by RAW FILENAME —
community-pack filename noise, so an artist's cards looked shuffled.
- artist / artist-desc gain a title secondary (direction baked per entry so
the legacy `dir=desc` append can't land on the title term; titles stay
A->Z under Z->A artists).
- The two-term (value, filename) keyset cursor can't seek a three-term
order, so artist sorts leave _KEYSET_SORTS and page by OFFSET — measured
trivial at real library sizes; title/recent keep their keyset. Restore
via a composite sort-key column if 50k-song libraries ever hurt.
- The tree view says "List view groups by artist — the selected sort
applies to the card grid" when a non-artist sort is active, instead of
silently ignoring the picker.
- Keyset proof-tests repinned to the title sort (same property, a sort
that still keysets); 2 new tests pin the title-within-artist order and
the OFFSET pagination's no-skip/no-dupe across pages.
Full-suite failure set identical to the same-main baseline.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* v3 library: honor legacy sort=artist&dir=desc (fold dir into effective sort)
Codex/review follow-up to the title-within-artist change: the new artist
ORDER BY bakes in `ASC` (for the title secondary), so the global `dir=desc`
append is suppressed and `sort=artist&dir=desc` silently returned A->Z
instead of Z->A — a regression on the legacy /api/library dir contract.
Fold `dir=desc` into the canonical sort key BEFORE the sort_map lookup via
the existing _effective_keyset_sort helper (same fold the cursor side already
does), so the ORDER BY is built from the effective sort. Only artist/title
fold (they have `-desc` twins); title/recent/tuning/year/mastery are
unaffected — verified by the keyset/filter suites.
New test pins that legacy `sort=artist&dir=desc` matches the explicit
`artist-desc` ordering (Z->A artists, A->Z titles within each).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
The LIBRARY group's main entry now reads "Song Library" in the sidebar
(and in the topbar page title, which mirrors nav labels). The nav key
and screen id are unchanged, so routing, saved hashes and promoted-
plugin anchors are untouched.
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
A results screen that offers "Up next: <song> — starting in 10s" needs to
know WHAT follows without reaching into queue internals. peekNext() returns
{filename, index, total} for the next track (null when nothing follows),
pure — peeking never plays or mutates.
First consumer: the note_detect results card's queue-advance strip (the
"Playlist Play All has no way to progress" tester issue).
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
The guitar/bass highway_3d renderer self-detects panel-canvas size changes
in its draw() loop and re-runs applySize() every frame, because the
splitscreen host overrides hw.resize and never calls renderer.resize().
The drum and keys highways lacked that fallback — they only re-framed when
the host explicitly called resize(w, h) — so their panels stayed framed for
the pre-fullscreen size while the guitar/bass panels adapted. Symptom: a
too-small, off-center highway in the drum/keys panels after maximizing a
split-screen session.
Port highway_3d's per-frame drift check into both draw() loops: re-apply on
backing-store change (canvas.width/height) AND on CSS-box drift
(clientWidth/clientHeight vs the last applied logical size, throttled to
every 10th frame). Track _lastHwW/_lastHwH + _appliedW/_appliedH per
instance and reset them in destroy() so a reused instance re-frames on the
next song.
plugins/drum_highway_3d -> 0.3.1, plugins/keys_highway_3d -> 0.1.1.
Tests: tests/js/drum_keys_highway_3d_resize_reframe.test.js.
Signed-off-by: Kris Anderson <topkoa@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* v3 library: context-menu unification — Fix match, Refresh metadata, Get info, multi-version remove (R2)
The ⋮ overflow and the native right-click menu already render from one
builder (openCardMenu), so every entry here lands in BOTH surfaces on
grid cards and tree rows alike — parity is structural, not maintained.
New entries (local library):
- Fix match… — opens the match modal in a single-song mode: no queue,
no Skip, the search panel open and pre-filled; a pick pins the match
exactly like the review flow (window.__fbFixMatch).
- Refresh metadata — POST /api/enrichment/refresh/{fn}: resets the
song's match to unscanned (canonical values + candidates cleared,
backoff zeroed) and kicks a pass. An EXPLICIT user action, so it may
discard a manual pin — the automation never does, but the user asking
for a re-match is the one party who owns that pin. Silent on success.
- Get info… — GET /api/chart/{fn}/fileinfo: file location + folder
(selectable/copyable under the v3 no-select default), format, size,
modified; for feedpaks the manifest summary (arrangements, stems,
cover/lyrics presence, authors, and whichever identity keys are
actually authored — mbid/isrc/genres/track/disc); plus the match
verdict ("Matched (text, 96%)" / "Pinned by you" / "Not scanned").
Under /api/chart because the GET /api/song/{path} catch-all would
swallow the suffix.
- Remove from library — with the multi-version interstitial: on a
multi-chart work, "remove the song" is ambiguous (a grouped card
stands for several files), so a modal lists EVERY version with
checkboxes (the card's own chart pre-ticked) and deletes exactly
what was picked — one file or the batch. Single-chart songs keep the
plain confirm.
Refresh + Get info are demo-mode blocked (cache mutation / path
exposure). apply_enrichment_match now zeroes `attempts` on an explicit
reset to unscanned, matching the stub upsert's identity-change rule.
5 new tests (refresh resets even a manual pin then re-matches via the
fake transport; 404s; fileinfo manifest/identity/match shapes;
traversal guard) + the demo-mode route list. Full-suite failure set
identical to the same-main baseline. tailwind.min.css regenerated.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* v3 context menu: review fixes — target displayed chart, harden Get-info
Three fixes from the PR review round:
- songs.js: Fix match / Refresh metadata / Get info now act on the DISPLAYED
chart (playTarget) rather than the group representative, matching Play. On a
grouped card where an intrinsic (tuning/arrangement) filter attached a
display_chart, these three previously fixed/refreshed/showed-info for the
wrong file. (__remove stays on `song`: it needs the group's work_key/
chart_count and already pre-ticks the shown chart.)
- server.py fileinfo: 404 ("not a chart") unless the path is a sloppak or a
loose song. The route previously stat'd ANY file under DLC_DIR, leaking its
path/size/mtime for e.g. a notes.txt the user keeps there. `format` can no
longer be "other".
- server.py fileinfo: the directory size sum skips symlinked entries so a link
inside a song folder can't pull in (or leak the size of) a file outside it.
Verified on the runtime that rglob does not descend symlinked subdirs.
+1 regression test (non-chart file -> 404).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* v3 library: Albums view — the client half of the album-condense work
Follow-up to the query_albums endpoint: the UI that consumes it, plus the
track-order plumbing the endpoint's track list needs.
- Albums view (a fourth view toggle next to grid/tree/folder): album cards
(cover / title / artist / track count) from /api/library/albums,
respecting the active filter drawer; clicking one opens the track list
with per-track play and a Play-album button that feeds the play queue
(falls back to plain playSong when the queue plugin is absent).
- Track order: the scanner now reads the feedpak `track`/`disc` fields
(spec 1.12.0) into new nullable songs columns (idempotent ALTERs), and
the album track list orders by the new `track` sort — disc, then track
number, unauthored charts to the bottom by title. Charts without
authored numbers keep working; they just sort alphabetically.
- The albums view persists like the other view choices.
3 new tests: manifest track/disc extraction (and unauthored -> None),
the disc->track->title sort order over /api/library, and the put()
round-trip. Full-suite failure set matches the known env baseline (one
tuner-config name swapped inside the suite-ordering flake family — the
file passes 25/25 in isolation on clean main).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* v3 albums: honour Genre/Match filters in album grid + detail (review fixes)
The Albums view only partially respected the filter drawer:
- /api/library/albums silently dropped the `genre` and `match` params the
client sends via queryParams(), so with a Genre or Match filter active the
album grid surfaced albums with zero matching tracks. Thread match_states/
genre through the endpoint -> query_albums -> _build_where, mirroring the
/api/library grid route. (SmartCollection/pass-through providers keep their
existing kwarg handling.)
- The album-detail track list built its own params (provider/artist/album/
sort only), so it ignored ALL active filters — the track list and the
Play-album queue could include songs the user had filtered out. Reuse
queryParams({...}, {catalog: true}) so detail honours the same filters as
the grid while pinning this album's artist/album and track order.
+1 regression test (albums endpoint honours the Genre filter).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
_patchCardFav (the Song Details drawer's like -> card heart sync) hardcoded
`classList.toggle('text-white', !fav)`, so toggling the like from the drawer
left List-View rows' `text-fb-textDim` idle class in place — the exact
dim-heart bug #654 fixed for the on-card click handler, reintroduced on the
drawer path. Read the per-heart `data-fav-idle` and swap that class instead,
mirroring wireCards. +regression assertion in v3_favorites_toggle.test.js.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
refreshRail was refactored (PR #702 work-grouping) to build a
`railParams = { sort_letters: 1 }` object (adding group when grouping is
active) before calling queryParams(), instead of the inline
queryParams({ sort_letters: 1 }). Behaviour is unchanged — it still opts
into the active-sort breakdown — but the source-assertion test lagged and
went red on main. Point the assertion at the new railParams shape.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(a11y): app-wide "Interface size" setting (Accessibility)
Adds a dedicated Accessibility -> Interface size control so users on large,
low-DPI displays can enlarge the app's menus, buttons and text (reported: eye
strain on a 32" 1440p panel with no OS scaling).
Mechanism: a host-owned scale capability (window.feedBack.scale) applies a
RELATIVE root font-size (a % of the user-agent base, never a px literal, so a
raised browser/OS base font is respected) and publishes an always-present
--fb-scale token. The rem-based v3 chrome scales together; the gameplay highway
canvas (device-pixel sized) is deliberately untouched, so playback resolution
and FPS are unchanged. Medium (100%) clears the override, so default rendering
is byte-identical to before -- zero blast radius.
- Settings -> new Accessibility tab: Small/Medium/Large/Extra-Large presets
(0.90/1.00/1.15/1.30) + a fine-tune slider (to 150%).
- Applied pre-paint from an inline <head> script (mirrors the ss-follower
pattern) so there is no flash-of-reflow on load.
- window.feedBack.scale read-API (get/set + scale:changed, fires once on load)
so canvas/WebGL surfaces that cannot inherit rem can follow the size. Shape
mirrors the working-tuning read-API; persists as a durable preference.
- Cosmetic px->rem sweep so text scales cleanly at the larger stops (2 v3.css
font-sizes + 20 text-[Npx] utilities across 8 v3 files; tailwind.min.css
rebuilt byte-stable via the pinned toolchain).
- One-time first-run nudge for the large/low-DPI display profile that deep-links
to the control (never fires once the setting is touched, or on other displays).
v3-only. Verified headless: core apply/persist/reset/reload/UI-sync + nudge
gating, with no console errors.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* fix(v3): regenerate tailwind.min.css to satisfy tailwind-fresh CI (PR #664 review)
Regenerate static/tailwind.min.css via scripts/build-tailwind.sh
(tailwindcss@3.4.19) so a fresh build matches the committed artifact and
the tailwind-fresh CI job's `git diff --quiet` passes. Two consecutive
regenerations are byte-identical.
Also (Fix 2) switch the fine-tune interface-size slider to the documented
transient-preview path: oninput now calls scale.set(v, { persist:false })
so dragging previews without writing localStorage/emitting a commit each
tick, and a new onchange commits with persistence on release.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* Fix v3 Songs A–Z rail: reliable taps, precise drag, hittable size
Follow-up to #634. Three rail bugs reported on macOS + Windows (0.3.0,
2026-06-29 — =Scr4tch=, MajorMokoto):
- Taps often did nothing ("clicked O, nothing happened"). pointerdown
calls setPointerCapture, after which the browser retargets the
follow-up click to the rail container, so the click handler's
closest('.v3-azrail-letter') resolved null and a plain tap (no
pointermove) had no other path. Drive the jump from pointerdown
itself; reduce the click handler to keyboard activation only
(e.detail === 0, Enter/Space).
- A drag landed short of the release ("where you release isn't where
you get sent"). Every letter crossed fired jumpToLetter with
behavior:'smooth'; stacked smooth-scroll animations over the
virtualized grid lagged and settled imprecisely. jumpToLetter now
takes a smooth flag and scrolls instantly ('auto') while scrubbing,
animating only discrete taps/keyboard jumps, so the grid tracks the
finger and the release lands on the let-go letter.
- The rail was too small at 1440p and didn't scale. Letters were a
fixed .62rem glued at right:2px (~13px-tall target). They now scale
with the viewport (clamp(.72rem, 1.4vh, 1.05rem)), sit off the edge
with taller/wider equal-width hit targets and a hover/active
highlight so the scrub target is visible.
Keyboard arrow-nav and present-letter gating are unchanged. Tests:
tests/js/v3_az_rail.test.js gains pointerdown-seek, keyboard-only click
guard, and instant-vs-smooth assertions (809 JS tests; the 13
pre-existing unrelated failures are unchanged).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* fix(v3): ignore non-primary buttons on A–Z rail pointerdown (PR #653 review)
Right- or middle-clicking the A–Z rail (or a secondary multi-touch
pointer) no longer triggers a seek; only the primary tap/drag scrubs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
Favoriting from the tree / "List View" flipped the glyph ♡→♥ but the
heart stayed grey until a re-search — reported macOS+Windows, open since
0.3.0 / 2026-06-25.
One shared wireCards() [data-fav] handler serves both the grid card and
the List-View row, but they render with different idle colours (grid
text-white, List View text-fb-textDim) and the handler only ever removed
the grid's text-white. So in List View text-fb-textDim lingered next to
the freshly-added text-fb-accent and won by CSS source order — the glyph
changed but the colour didn't, until a re-search re-rendered the row.
Each heart now declares its idle colour via a data-fav-idle attribute;
the handler swaps exactly that class (so only one colour class is ever
present) and writes the new state back onto the in-memory song model so a
re-render / virtualized-grid recycle agrees instead of reverting.
Tests: tests/js/v3_favorites_toggle.test.js. Full JS suite 810 tests; the
13 pre-existing unrelated failures are unchanged.
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* feat(v3): native right-click context menu on library cards
Right-clicking a song card now opens the same overflow (more) menu at the
pointer, so the actions (Play, Add to playlist, Save for later, and any plugin
card actions) are reachable without aiming for the small button. openCardMenu
gains an optional pointer position (fixed + viewport-clamped when opened via
right-click, the existing card-anchored absolute otherwise). Adds a "Save for
later" row to the menu too, closing the gap where Save was only the inline
button. Shares one menu definition + the libraryCardActions registry, so plugin
actions appear in both paths.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* fix(v3): valid JS emoji escape in card-menu toast + dismiss orphaned menu on re-render (PR #688 review)
Replace the Python-style `\U0001f3b5` escape (invalid in JS, rendered
literal text) with the 🎵 emoji character in the playlists-updated toast.
Also tear down any open card context menu at the start of render(), the
full filter/scan-refresh path: the right-click menu is appended to
document.body, so replacing root.innerHTML would otherwise leave it
floating orphaned until the next document click. Reuses _closeCardMenu.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* feat(v3): genre filter facet (reads the feedpak genres field)
Adds a Genre facet to the library Filters drawer, populated from each song's
primary genre. Server: a genre column (idempotent ALTER, indexed) written from
the sloppak manifest's genres list on scan (primary = genres[0]); a genre band in
_build_where (OR within the selected set); GET /api/library/genres for the facet's
distinct list. Client: a multi-select Genre section mirroring the tuning/mastery
facets. Follows the merged spec 1.12.0 genres field (#40).
v1 stores only the PRIMARY genre (genres[0]); secondary genres aren't filterable
yet. Threaded like the mastery filter (separate query_page kwarg, so query_artists
/query_stats are unaffected) -- genre filters the grid view. Needs a rescan to
backfill genre on existing packs (only packs whose manifest carries genres).
Verified live: a sloppak tagged genres:[Metal, Rock] -> /api/library/genres
returns [Metal]; ?genre=Metal returns it; ?genre=Rock (secondary) returns none; a
plain song stays ungenred.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* fix(v3): scope genre facet to local provider (PR #690 review)
The /api/library/genres facet always read the local meta DB, so a remote
provider showed local genres while the `genre` filter was a no-op on that
provider's grid. Make the endpoint provider-aware: return an empty facet
for remote providers (kind != "local") and keep serving genres for the
local library and its smart collections, which share the local DB. The
v3 client now passes the active provider, mirroring the tuning-names facet.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
First slice of the host theme contract (#644): give plugins a host-owned way to
read the active theme + its device affordances, so a feature renders correctly
under any theme instead of binding to whichever one the dev saw.
theme-core.js previously only APPLIED themes and emitted --fbv-* vars only while
a theme was equipped (nothing to read in the default state), with no read API.
Now, all additive + feature-detected:
- Always-present default `fb` palette as `--fbv-*` on :root (the un-themed look
is unchanged — fb-* utilities still use their compiled defaults; this only
hands plugins a stable host token to read + derive surfaces from). Adds two
keystone ROLES the palette lacked: `on-accent` (legible fg on the accent fill)
and `focus-ring`.
- window.feedBack.theme.get() -> {id, isThemed, tokens}; .capabilities() ->
{glow, gradients, motion} (the device-affordance signal; recolor-only themes
report defaults, a theme may opt out via `capabilities` in its payload, motion
is reduced-motion-gated); .prefersReducedMotion().
- Normalized `theme:changed` event from the single apply() chokepoint.
The apply side stays on window.v3Theme; the read surface is attached defensively
so it survives the feedBack bus being (re)built by capabilities.js regardless of
load order. Verified via a headless render (apply/unequip intact, defaults
present + restored, capability opt-out honored, event payload correct) +
tests/js/v3_theme_read_api.test.js. See docs/host-theme-contract.md.
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Completes the Refresh feature: the scanner now reports a delta so the toast can
say what changed instead of a generic confirmation.
Server: delete_missing returns both deltas from its one query -- rows pruned
(removed) and current files not yet in the DB (added) -- and the scan retains
added/removed on the terminal scan-status (previously wiped to 0). Client: the
completion toast shows "N songs added / M removed" (or "up to date"), and a scan
we merely attached to (background / Settings) toasts only when it actually
changed something, so a periodic no-op pass stays silent. library:changed now
carries the delta too.
Verified end to end: empty rescan -> added 0; add a song -> added 1; remove it
-> removed 1.
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Adds GET /api/library/albums + MetadataDB.query_albums (plus the local and
smart-collection provider delegations): distinct (artist, album) groups with a
track count and a representative cover song, paged by album, honoring the same
filters as /api/library (including the new mastery bands). Rows with no album
name are excluded. Album detail needs no new endpoint -- it reuses the existing
/api/library?artist=&album=.
Backend foundation for the album-condense "Albums" view + play-album; the client
view consumes it next.
Verified: 4 tracks sharing one album -> /api/library/albums returns one card
(artist, album, count=4, cover); /api/library?album=... lists its 4 tracks.
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* library: MusicBrainz text matching + Match-Review UI — P8
Replaces the enrichment plumbing's no-op matcher (P7) with the real
pipeline, per the library-metadata design: a wrong match is worse than a
slow one, so medium confidence goes to a human review queue and never
straight to canonical values.
- lib/mb_match.py (new, pure — no network/DB/server imports): denoise
(author credits, (440Hz)/(Live)/(No Lead)/(v2) parentheticals,
diacritics/punctuation, ACDC / AC DC / AC/DC folding via compacted
token equality), token-set similarity, scoring with year/duration
corroboration bonuses, tier classification (auto needs combined
>= 0.95 AND per-field floors — a perfect-title cover by the wrong
artist, or a chart with no artist, can never auto-match), Lucene
query building, MusicBrainz response normalization.
- Matcher precedence in _enrich_one: content-hash cache copy (another
chart of the same recording matches with no network) -> manifest
mbid (tier 0) / isrc (tier 1) exact keys, feature-detected and
strictly shape-validated, read-only -> text search tiers
(auto / review / failed).
- Lifecycle: review rows store their ranked candidate list (JSON) and
write NO canonical fields until a human accepts; failed rows retry on
an exponential backoff (1 h doubling, 7 d cap) via the attempts
column; user-rejected rows never auto-retry; an identity edit
re-queues anything and resets the backoff; never-overwrite-manual is
enforced inside the single writer (apply_enrichment_match) so no call
path can forget it.
- Network: _mb_http_get is the one transport seam — throttled to
<= 1 req/s through P7's _enrich_throttle, identified with a real
User-Agent from VERSION, and a 503 pauses the whole pass without
burning attempts. Offline guard: no sockets under
FEEDBACK_ENRICH_OFFLINE or FEEDBACK_SKIP_STARTUP_TASKS, so pytest can
never reach MusicBrainz; the pass still stamps identity hashes
(two-phase), which is why every P7 test passes unchanged.
- Routes: GET /api/enrichment/review, POST
/api/enrichment/review/{filename}/accept|reject|pick, GET
/api/enrichment/search (throttled manual-search proxy). All four are
demo-mode blocked.
- Match facet: match= CSV accepted by /api/library AND
/api/library/stats (the A-Z rail's letter counts stay lockstep with
the grid) — review / matched (incl. manual) / unmatched / pending,
the same EXISTS idiom as the mastery facet.
- UI: static/v3/match-review.js (new, self-contained) — an ambient
"N to review" chip beside the song count (rendered only when
non-zero; silent on success, no toasts), and a review drawer on the
filter-drawer slide idiom (Escape + focus trap; row click accepts,
"Not a match" rejects, "Search instead" is the fix-match escape
hatch). songs.js gets the chip mount, a Match filter section, and
session-only match state; also fixes the latent applySavedPrefs bug
where restored filters dropped the mastery key, which made the
filter drawer throw for anyone with saved prefs.
- static/tailwind.min.css regenerated (scripts/build-tailwind.sh) for
the new utility classes; conflicts with sibling PRs resolve by
re-running the script.
Nothing is ever written to pack files — canonical values live only in
the song_enrichment display cache. Cover art caching and acoustic
fingerprinting are follow-up slices.
22 pure unit tests + 19 server tests (fake transport injected over the
_mb_http_get seam) + demo-mode route cases; full-suite failure set
A/B-identical with the change stashed vs applied.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* library: match-review modal + configurable auto-apply confidence (P8 R0)
Follow-up to the initial P8 commit, folding in the first round of tester
feedback on the review surface and the matcher's knobs:
- Review GUI is a centred MODAL now, not a sidebar — one chart at a
time (the scraper-review model from media-server / emulation-frontend
apps): the chart's current metadata with explicit amber
"Missing: album / year / cover art" chips (art detected via the art
request failing), candidates each carrying "Adds: year - genres -
ISRC" / "Shows as: ACDC -> AC/DC" per-field chips, and Skip /
Not a match / Search instead / Use selected with prev-next + arrow-key
navigation. Chip + window API surface unchanged, so songs.js needed no
edits for the rework.
- Auto-apply confidence is a SETTING: default drops 0.95 -> 0.90
(mb_match.AUTO_MIN; classify() takes an auto_min override). The
per-field floors are untouched and threshold-independent — a
perfect-title cover by the wrong artist still can't auto-match at any
setting. New validated settings keys: enrich_enabled (bool) +
enrich_auto_threshold (0.5–1.01; >1.0 = "Always review", since a
capped score can equal exactly 1.0). Read once per pass; disabling
gates only the BACKGROUND matcher — manual search/fix stays available.
- Settings -> Library -> "Metadata matching" card: enable toggle,
confidence select (85 / 90 / 95 / Always review), a Match Now button
(new POST /api/enrichment/kick, single-flight like every other kick,
demo-mode blocked), and a live status line fed by the same fetch as
the review chip. Markup in index.html per the v3 settings pattern,
wired by match-review.js, null-guarded so v2 no-ops.
- Review queue orders missing-data charts first — confirming those has
the most to gain; complete charts only stand to be re-labelled.
Tests: threshold moves the auto/review boundary via settings; the
enable toggle gates matching but not the manual proxy; settings
validation; kick route; queue ordering; classify(auto_min=...) floors.
Full-suite failure set byte-identical to the pre-change baseline.
tailwind.min.css regenerated for the modal's utility classes.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* fix(library): lock MusicBrainz throttle across sleep + de-dup enrich queue (PR #710 review)
Hold a module-level lock across _enrich_throttle's read/sleep/write so the
background daemon and threadpooled sync search route serialize outbound MB
requests instead of bursting past the 1 req/s limit. De-dup the enrich queue
by filename so a changed-hash failed row isn't processed twice per pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* library: metadata-enrichment plumbing (cache table + worker lifecycle) — P7
The pipeline around a deliberately NO-OP matcher, so the real
MusicBrainz text matcher (next slice) replaces exactly one function and
inherits the queue, throttle, lifecycle, and safety contracts:
- song_enrichment cache table: one row per song holding the match
lifecycle (unscanned -> matched(source,score) | manual | failed) plus
the canonical values a confident match supplies. A DISPLAY cache -
canonical values are never auto-written into pack files. Never purged
on rescan (only by the explicit per-song delete); dead rows filtered
at read time; re-derivable, so a lost row just re-enriches.
- Identity hashing: sha1 of normalized artist|title|album|duration.
Filename-free, so a renamed pack keeps its enrichment; unchanged hash
makes re-enrichment a no-op (idempotent).
- Queue rules (test-pinned): no row / unscanned / identity-changed ->
re-match; matched + current hash = settled; a MANUAL row is the
user's pinned pick and is never auto-reset (state and hash both
survive metadata edits); failed waits for the matcher's backoff
policy (attempts column ready).
- Worker: _kick_enrich/_enrich_runner mirror the scan's single-flight +
coalescing pattern, kicked when a scan pass fully completes (the scan
pool is a no-network process pool by design; the 5-minute periodic
rescan is the natural retry hook). One bounded pass per kick - no
drain-loop, since the no-op matcher legitimately leaves rows
unscanned. _enrich_throttle() is the <=1 req/s seam every matcher
must call before a network request, and the never-hold-meta_db._lock-
across-a-fetch rule is documented at the seam.
- CONFIG_DIR/art_cache dir helper (the cover-art slice adds the LRU
cap) + GET /api/enrichment/status (worker flags + counts by state).
8 new tests; full-suite failure set identical to unmodified main.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* fix(library): lock enrichment reads on shared conn + skip redundant stub writes (PR #707 review)
Issue 1: wrap the SELECT+fetch in enrichment_pending, get_enrichment and
enrichment_state_counts in self._lock so request-thread reads no longer
interleave with the worker's execute+commit on the shared connection.
Issue 2: guard upsert_enrichment_stub so an already-settled row (manual
pick, or a non-manual row whose content_hash already matches) skips the
UPDATE/commit — stops the no-op matcher re-writing every song each pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* v3 library: curated album — your version of an album, one chart per slot — P6
A curated album is a hand-picked, ORDERED practice set of works with a
chosen chart per track (metadata-design 7.2) - the repeatable gameplay
loop. No new tables: a playlists row with kind='album' plus two per-slot
columns.
- Schema (additive, idempotent): playlists.kind ('album' | NULL=mix),
playlist_songs.arrangement (the pinned arrangement NAME - names
survive rescans; the index is resolved at play), playlist_songs.
work_key (stamped at ADD time = "resolved to preferred once at add,
pinned thereafter").
- Orphan-at-read self-heal: an album keeps every slot. A slot whose
pinned chart was deleted resolves to the work's CURRENT keeper at
read (marked "(auto)"; membership is never rewritten - if the file
returns, the slot resolves back to itself), and reports missing when
the whole work is gone so the set's denominator stays honest. Mixes
keep hiding dead songs byte-identically.
- Slot editor: PATCH /api/playlists/{pid}/songs/{fn} pins/clears the
arrangement and/or swaps the slot's chart - validated to the SAME
work via the stored stamp, position + pin kept, duplicate members
rejected. The per-slot pick is independent of the work's global
preferred: a rehearsed set stays the same notes even if the global
keeper is re-picked later.
- UI: "New album" on the Playlists screen (album chip + disc cover);
the album detail adds a set-scoped "Album repertoire" meter (N of M
mastered - per-track mastery, never one album score), per-track
accuracy, and a per-row slot editor listing only the work's charts.
"Play album" runs the play-queue front-to-back honoring pins (the
queue already supported per-index arrangements); per-row play uses
the resolved chart + pinned arrangement.
12 new tests; playlists/collections regressions green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* fix(v3): count all album slots (list vs detail parity) + surface slot-edit PATCH failures (PR #706 review)
_playlist_count applied the mix "dead-filter" to every playlist, so an album's
list-card count dropped orphaned/missing slots that its detail view still
renders and plays (5-track album, 2 pins deleted → card "3" vs detail 5). Count
ALL slots for kind='album' (mirroring get_playlist's is_album discriminator);
mixes/other kinds keep the dead-filter. openSlotPicker's Apply now checks the
jsend return and, on a rejected PATCH (swap-to-other-work / duplicate pin),
shows an inline error and keeps the picker open instead of closing as success.
Adds album count-parity + mix dead-filter regression tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* v3 library: multi-chart work grouping engine + work-charts API — P5a/P5b
Charts of the same song (same normalized artist+title) now GROUP under a
computed work_key, with a materialized representative filter so the grid
can collapse them without breaking keyset paging:
- work_key = normalize(artist+title) (diacritics/punct/case folded,
leading "The" folded on artist); resolves the effective artist via the
artist_alias table when present (feature-detected, no hard dep).
- Sparse, never-purged-on-rescan tables: chart_group_pref(work_key,
preferred_filename) + chart_group_split(filename, split_key); purged
only by the explicit per-song delete.
- Materialized work_display(filename, work_key, effective_work_key,
is_group_representative, group_size) read-model: lazy rebuild via a
dirty flag set on put/delete; set_chart_preferred does an incremental
re-flip (no full rebuild). Auto-pick representative = most
arrangements -> most plays -> newest -> filename; a user pref wins and
degrades to auto-pick if its file disappears.
- group=1 on query_page/query_stats = one extra representative
predicate applied identically to page + total + sort_letters, so the
keyset cursor (sort_value, filename) stays a valid total order and
counts works, not charts. Grouped rows carry chart_count + work_key.
- Charts API: GET /api/work/{work_key}/charts (members + which is the
keeper, your pick vs auto), PUT/DELETE .../preferred, and
POST /api/chart/{filename}/split + /unsplit (under /api/chart so the
DELETE /api/song catch-all can't shadow them).
Tests: 15 grouping-engine + 7 charts-API tests, including grouped
keyset pagination (no skip/dupe across pages).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* v3 library: grouped grid card + persistent "N charts" chip — P5c
Flip the v3 grid to the grouped library (group=1 on /api/library and the
rail's /api/library/stats fetch): one card per song, showing the
representative (preferred/auto-pick) chart. group rides page, total and
sort_letters identically so the A-Z rail's cumulative-seek math and the
virtualized sizer stay consistent, counting works not charts; the keyset
cursor chains with group on every page.
- New groupingActive() helper, default ON per the design; the persisted
per-view toggle (P5e) lands there. Only the local provider implements
group=; smart collections and remote providers ignore it and stay
flat, so it is safe to send unconditionally.
- chartsChipHtml(): a "flag N charts" chip rendered ONLY when
chart_count >= 2 - single-chart cards emit byte-identical markup.
First in the fixed-height chip row + shrink-0 so it never clips and
card height is unchanged.
- Chip click = feature-detected window.__fbOpenChartsDrawer (the Charts
drawer arrives in P5d); until then a no-op. Plain-click / play / the
arrangement chips are untouched and play the representative.
- The library-home repertoire meter's stats fetch deliberately stays
ungrouped: its mastered numerator counts chart filenames, so a works
denominator could exceed 100% - reconciling that is P5e's
mastery-anchor work. The tree view stays flat (query_artists has no
grouping; its opener is wired in P5d).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* v3 library: Charts drawer + openers — P5d
The single deep-management surface for a work's charts (design UX-2/3):
a body-appended slide-in drawer (filter-drawer idiom) listing every
chart of the work as a radiogroup — the checked row is the keeper the
grid card plays.
- Rows show format / tuning / arrangements / year / your accuracy (or
"not played") plus the pack filename, usually the only human-readable
distinguisher between duplicate charts. Keeper is labeled
"Preferred (auto)" vs "Preferred - your pick".
- Row click (or Enter/Space) = one-tap Set-preferred; "Reset to auto
pick" appears when the keeper is an explicit pick. Writes go through
the work-charts API and the drawer re-renders from the response; the
grid re-fetches in place since the representative may have flipped.
- Per-row Play (plays that exact chart) and Add-to-playlist (the picker
is z-[200], layering over the z-50 drawer).
- a11y: Tab focus-trap, Escape closes, ArrowUp/Down move focus between
rows (focus only - arrow-select would fire a preferred write per
keystroke), focus restored to the opener on close.
- Openers: the "N charts" chip opens the drawer directly; the card's
overflow menu gains "Charts (N)..." and "Play version >" (expands
inline; picking one plays it as a one-off - the keeper/headline does
not move). Tree rows ride the ungrouped artists endpoint, so the menu
resolves their work lazily via the new GET /api/chart/{fn}/work
({work_key, chart_count}) and slots a "Charts (N)..." entry in when
versions exist. A window.__fbOpenChartsDrawer global lets other views
open the drawer. Right-click is deferred: the open native card
context-menu PR should host that entry once both merge.
- tailwind.min.css rebuilt: carries the new utility classes from this
and the previous commit (the grouped-card chip tint was missing).
Split keys contain '#', so clients MUST URL-encode work_key in paths
(the v3 client does; a test documents the round-trip). 4 new endpoint
tests; 26/26 grouping+charts tests green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* v3 library: group toggle, split UI, the filter law + mastery-anchor rules — P5e
Completes the multi-chart grouping slice (design 7.1):
- Filter law under group=1: work-identity (artist/album/search) and
practice-state (favorites/mastery/tags/difficulty) predicates stay on
the representative row, while CHART-INTRINSIC predicates (format/
arrangements/stems/lyrics/tuning) now match if ANY member of the work
does - a song you own in Drop D is no longer hidden because your
preferred chart is E Standard. Intrinsic clauses moved to an
alias-aware builder and re-applied as a member EXISTS; identical in
query_page and query_stats so counts and the A-Z rail stay in
lockstep. A pure predicate - keyset paging is untouched (tested).
- Display-chart switch: when the representative itself doesn't match,
the row carries a display_chart override (the matching member). The
row stays the representative's - swapping rows wholesale would break
the (sort_value, filename) cursor - and the card renders/plays the
member while the accuracy badge and heart stay anchored on the
preferred chart.
- Mastery sort aggregates MAX across the group ("a song surfaces on any
chart you've touched"); OFFSET-paged, so cursor-safe. The
Recently-Added aggregate is deliberately deferred: mtime IS a keyset
sort, so its aggregate would need materializing into work_display.
- History-sticky auto-pick: most-played -> most-complete -> newest.
A newer/"more complete" import can't silently take the pick from the
chart your reps accrued on, and a one-off try of an alternate can't
out-rank a practiced incumbent; all-unplayed groups still pick by
completeness.
- Persisted "One card per song" toggle in the filter drawer (default
ON; OFF = one card per chart). A view mode: never counted in the
filter badge, never saved into collection rules, local provider only.
- Split escape hatch: "Split out" per drawer row gives a chart its own
card; the split card's overflow menu offers "Rejoin other versions"
(rows and the chart-work lookup now carry is_split).
- Mastery-anchor heads-up: after set-preferred the drawer shows a
one-line ambient note that practice history stays with each chart
(no toast - hearing-safe).
10 new filter-law tests; 38/38 grouping tests green. tailwind.min.css
rebuilt for the new utility classes.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* fix(v3): work-grouping — escape Charts-drawer meta (XSS), keep non-Latin titles distinct, guard mid-rebuild reads (PR #702 review)
- XSS: esc() the composed `meta` string in _chartRowHtml (arrangement/tuning
names come from untrusted feedpak metadata) before innerHTML; acc stays HTML.
- Non-Latin titles: _norm_token falls back to raw lowercased whitespace-collapsed
text when the NFKD+strip fold yields "" (CJK/Cyrillic/Greek/Arabic), so distinct
non-Latin titles keep distinct _work_key values instead of collapsing into one
bogus work. Latin names still hit the folded branch — behavior unchanged.
- Mid-rebuild reads: wrap the grouped representative SELECT in query_page and
query_stats under self._lock (nullcontext when ungrouped, so lazy reads stay
lock-free) so a reader can't observe work_display between rebuild_work_display's
DELETE and INSERT/commit. _ensure_work_display stays OUTSIDE the lock — it
self-locks the rebuild and self._lock is non-reentrant — so only the SELECT is
guarded (rebuild fully completes before the guarded SELECT runs).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* v3 library: artist aliases + Tidy-up merge UI — P4
Fixes the "ACDC vs AC/DC" split without touching a single file or row:
a never-purged artist_alias table (raw_name -> canonical_name) applied
at DISPLAY time. The scanner keeps writing whatever the pack says; one
alias row fixes every matching song.
- query_artists dedupes/groups/orders on the effective artist, with a
zero-cost fast path when no aliases exist; the artist filter expands
a canonical name to its raw variants (index-friendly, keyset-safe);
query_page re-labels row artists through the alias map.
- CRUD + merge API: list aliases, list raw artists (variants + counts
for the picker), set/merge/remove; a self-alias clears (= un-merge).
- "Tidy up artists..." in the filter drawer (local library only): a
searchable raw-variant checklist, merge-into-canonical, and a
current-merges list with per-row un-merge. The artist dropdown + tree
pick up canonical names with no dropdown code changes.
- Sort + A-Z rail stay on the RAW artist (keyset-safe): a cross-letter
alias shows its canonical label but buckets under the raw letter
until effective columns are materialized (the grouping engine's
work_key already resolves aliases when this table exists, so merged
artists group correctly there).
11 tests. tailwind.min.css regenerated (generated file - on a merge
conflict, re-run scripts/build-tailwind.sh).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* fix(v3): flatten transitive artist-alias chains + cycle guard so sequential merges unify (PR #705 review)
merge_artists looped set_artist_alias which stored one hop, so sequential
merges (ACDC->AC/DC then AC/DC->AC-DC) left a two-hop chain that the
single-hop effective_artist/grouping/filtering split into two groups. Add
_single_hop_canonical + _terminal_canonical (visited-set cycle break),
resolve the canonical to its terminal before storing, forward-flatten
existing rows that pointed at the raw name, and reject cycles (409). Batch
merge now runs under one lock + one commit for atomicity.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* v3 library: growth-edge "practice next" recommender — P3
The "Keep practicing" shelf stops being recency-only: a new GET
/api/library/practice-suggestions ranks started-but-unmastered songs by
difficulty-appropriateness x mastery-proximity (the growth edge - the
mid-difficulty, closest-to-mastery material where practice pays off
fastest), and the shelf sources it instead of filtering /api/stats/recent.
- Score = difficulty band fit (your 1-5 rating; unrated degrades to the
middle band so the shelf works before any ratings exist) x proximity
to the 0.9 mastery threshold. Read-only - never writes difficulty.
- A shelf click opens the closest-to-mastery arrangement.
- Per-arrangement difficulty and seed-from-authored intentionally NOT
faked: there is no authored/derived difficulty on songs yet (the
feedpak difficulty spec is unmerged) and the personal rating is
per-song - both revisit when that field lands.
9 endpoint tests. tailwind.min.css regenerated (generated file - on a
merge conflict, re-run scripts/build-tailwind.sh).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* fix(v3): deterministic tiebreak (filename) in practice-next ordering (PR #704 review)
Add r["filename"] as the final sort component so suggestions with equal
growth_score and equal/None last_played_at order deterministically instead
of by SQLite's unordered agg.items() scan.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* v3 library: song Details drawer + bulk edit — P2
Evolves the per-song editing surface from the legacy modal into a v3
slide-in Details drawer (the filter-drawer idiom, body-appended): catalog
fields (title/artist/album/year, written through the existing atomic
manifest writer) plus the P1 personal layer - your difficulty (1-5),
tags, and notes - with the heart staying the existing favorite system.
- Cards badge the personal layer at rest (difficulty pip + tag count,
top-right, fading on hover so the action buttons keep that corner);
un-annotated cards render byte-identical to before.
- Bulk edit from the select-mode batch bar: POST
/api/songs/user-meta/batch applies additive tag add/remove and a
leave/set/clear difficulty across the selection (mixed-state aware).
- The core card action relabels to "Details" and opens the drawer via a
feature-detected global, falling back to the legacy modal when the
drawer isn't mounted.
16 batch tests new; the P1 user-meta suite stays green.
tailwind.min.css rebuilt for the drawer's utility classes.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Nm7tHs1Yvjjtnnu4nzJgdN
* fix(v3): surface bulk-edit / save-details request failures instead of reporting success (PR #703 review)
Check the batch/write responses and show an fbNotify error (keeping selection and drawer) instead of unconditionally closing as success.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* v3 library: personal per-song metadata (user-difficulty / notes / tags) — P1
The local "your relationship to the song" layer: a per-song user-difficulty
(1–5, planning-only, distinct from the authored 1–10 difficulty bands),
free-form notes, and free-form practice tags. All kept OUT of the shared
feedpak file and OUT of the `songs` table, in two new never-clobbered tables
so a rescan's `INSERT OR REPLACE INTO songs` can't wipe them. Likes stay the
existing favorites heart — no rating column.
Backend only (the Details drawer + tag/difficulty filter UI come next).
Schema (additive, idempotent):
- song_user_meta(filename PK, user_difficulty INTEGER, notes TEXT, updated_at)
- song_tags(filename, tag, created_at, PK(filename, tag)) + idx on tag
API (DB-only — distinct from POST /api/song/{f}/meta, which writes catalog
fields back into the file):
- GET /api/song/{f}/user-meta → {user_difficulty, notes, tags}
- PUT /api/song/{f}/user-meta → partial update; user_difficulty (1–5 or
null), notes (string or null), tags (full-replace array). Tag removal is a
full-replace array rather than a DELETE sub-route because the greedy
DELETE /api/song/{filename:path} already owns every DELETE under /api/song
and would shadow it.
- GET /api/tags → tags in use with counts (for the filter UI)
Read path:
- query_page rows embed user_difficulty + tags (like `favorite`); notes stay
out of the list payload (per-song GET — they can be long).
- Read-time filters ?user_difficulty= and ?tags= threaded through _build_where
exactly like the mastery filter — EXISTS-style predicates, so keyset paging,
counts, and the A–Z rail are unaffected.
- delete_song purges both personal tables inside the existing lock.
Tags are normalized (trim + lowercase + collapse whitespace) so "Rock"/"rock"
don't split. New tests cover defaults, difficulty validation (rejects out-of-
range / non-integral / bool), notes trim, tag normalize/dedupe, grid embed,
both filters, never-clobber-on-rescan, and purge (21 tests). Neighboring
library tests (filters/keyset/providers/playlists/collections/stats) stay
green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* fix(v3): cap per-song tags at 50 to bound writes (PR #691 review)
set_song_tags capped each tag at 60 chars but not the number of tags,
so one PUT could write unbounded rows. Cap the normalized-unique tag
list to the first 50 after dedup. Adds a test asserting >50 stores 50.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
PR #530 set accept=".feedpak,.sloppak" on the shared upload input, but a
later index.html edit reverted the attribute to ".sloppak" only. The
client-side extension filter (app.js) and the server upload endpoint both
accept .feedpak, so the only effect was the OS file-picker hiding .feedpak
files from the dialog. Restore the dual filter.
Signed-off-by: Kris Anderson <topkoa@gmail.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Removes the Floating Button and Tuning Visibility settings sections and finishes retiring their still-live config: drops the disabledTunings menu filter and showFloatingButton gate from screen.js/ui.js and their persistence in routes.py (retired keys are stripped on write). Repositions the tuner panel opened from the v3 sidebar Plugins popover to anchor beside it via the host's stable plugin-control slot API (falling back to the popover id), clamped to the viewport so it can't open off-screen, and re-anchored on resize. Updates tuner config tests to the retired-key behavior; plugins/tuner 1.3.2 -> 1.3.3.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Disc radius 3.6->4.3, cymbal 3.0->3.6 (heights proportional) — hands-on
feedback said the gems were hard to read at the default camera. All
variant shapes key off these radii; accents still fit the lane gap.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
- BG_STYLES port (off/particles/lights/geometric; lights use the
pitch-class palette) mounted behind the scene; _bgGetAnalyser/
_bgReadBands (stems-first, one-shot #audio fallback, permanent-failure
latch); Ambience intensity + Audio-reactive settings; remounts on
style/intensity change
- Score-FX overlay canvas (drum_highway_3d pattern): +1 pops at the
scored key, ring pulse every 10-combo, milestone bursts at 25/50/100,
red wash on 3+ streak break (wrong notes AND swept misses); cleared
when idle, removed in teardown
- Tests: style id validation + FX defaults (30 total)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
- BG_STYLES port (off/particles/lights/geometric) into a renderOrder -1
group; _bgGetAnalyser/_bgReadBands (stems-first, one-shot #audio
fallback, permanent-failure latch, 5ms bands cache); Ambience
intensity slider + Audio-reactive toggle; remounts on style/intensity/
palette change and across kit-change scene rebuilds
- Score-FX overlay canvas (guitar drawScoreFx adapted to internal
scoring): +1 pops at the struck lane, ring pulse every 10-combo,
milestone bursts at 25/50/100, red wash on 3+ streak break; pooled,
cleared when idle, removed in teardown
- butterchurn/image/video deliberately out of scope
- Tests: style id validation + FX defaults (15 total)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* fix: correctly import and notate multi-staff (piano/keys) tracks from GP8
Fixes bass stave being dropped on import (bar-column enumeration bug)
and wrong hand-split heuristic in notation_lift for chords straddling
middle C.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Signed-off-by: byrongamatos <xasiklas@gmail.com>
* fix(gp-import): fold all grand-staff staves, per-stave tuning, playable hand-splits
Addresses review on #692 (topkoa):
- split_hands: only use the middle-C boundary when both resulting hands are
within HAND_SPLIT_SPAN_SEMITONES, else fall back to the largest-gap
heuristic — a hard middle-C split otherwise put a 19-semitone (unplayable)
span in one hand for bass-under-treble voicings (e.g. E2+B3 under an Em7
shape).
- Treat any multi-stave (grand-staff) track as keys end-to-end, so the
stave-0 and folded stave-1+ notes share one encoding and note_count (which
sums every stave column) matches what actually imports — closing the
phantom-count case for grand-staff instruments the name/program heuristics
miss (harp, celesta, marimba).
- Fold *every* extra stave (stave_columns[1:]), not just stave 1.
- Per-staff tuning fall-back to the track-level Tuning property so an untuned
staff never yields an empty pitch list (silent note loss); via a shared
_parse_tuning helper.
- Extract _collect_column_notes / _merge_lh_notes so the GPX LH/RH pair merge
and the GP8 grand-staff fold share one implementation and can't drift in
tie/timing/dedup handling.
- Rebuild filtered_to_raw from the already-computed stave_columns (one source
of truth for the counting rule) and drop the dead num_raw_tracks/raw_tracks.
Tests: grand-staff fold + bar-column offset (test_gp2notation.py); both
middle-C split cases (test_notation_lift.py). CHANGELOG updated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: byrongamatos <xasiklas@gmail.com>
---------
Signed-off-by: byrongamatos <xasiklas@gmail.com>
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
- Pooled spark bursts (PORTED highway_3d, pool 160) at the struck lane,
timing-colored via _timingHex/_classifyTiming (±50ms window, inner 40%
= on-time); streak-scaled counts (streakFx)
- Lane flashes resurrected as pooled additive gauss-tex quads at the hit
line (timing-colored; red for wrong hits); kick = full-width quad
- Kick pulse: camera dip + amber floor wash, exp decay, hitFx-scaled
- Approach highlight: lane stripes brighten toward their next note
- Open hi-hat: hh_open renders a warm ring around the gem (closes TODO);
orthogonal to accent/ghost/flam variants
- Settings: Hit sparks / Timing colours / Streak feedback toggles +
Hit feedback intensity slider (drum_h3d_bg_*, live-applying)
- All FX resources pooled/shared, disposed in BOTH teardown paths
- Tests: _classifyTiming boundaries + FX defaults (10 total)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
- setPixelRatio at last: DPR (cap 2; 1.25 when >1 instance) x host
adaptive bundle.renderScale — HiDPI displays were rendering at CSS
resolution and upscaling (soft/aliased)
- Bloom: port _bloomEnsure/_bloomDispose (UnrealBloomPass 0.65/0.5/0.82,
MSAA HalfFloat target, ACES<->None switch); hit-line, flames and
consume-glow benefit immediately; direct render is the degrade path
- First settings panel: settings.html (graphics category) with a live
Glow (bloom) toggle; FX scaffold (FX_DEFAULTS/readFxSettings/
window.keys3dSetFx, keys3d_bg_* keys, keys3d:settings event)
- Combo/accuracy/best-streak DOM HUD (drum_highway_3d pattern), gated on
a live MIDI session
- tests/fx_settings.test.js (3 tests; vm harness)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
- Consume host adaptive-quality bundle.renderScale; fold into DPR like
highway_3d (splitscreen proxy: >1 instance caps baseDPR at 1.25)
- Bloom: port _bloomEnsure/_bloomDispose (UnrealBloomPass 0.65/0.5/0.82,
MSAA HalfFloat target, ACES<->None tone-mapping switch); rebuilt across
the kit-change renderer recreation; direct render is the degrade path
- FX settings scaffold: FX_DEFAULTS + readFxSettings + drumH3dSetFx
(drum_h3d_bg_* keys), Graphics section in settings.html (bloom toggle,
default ON, live-applies)
- _applyLaneFlashes dead-code comment updated (visual consumer lands in
the hit-FX PR)
- __test export + tests/data_layer.test.js (vm harness, 8 tests) —
picked up by the CI glob from the bundling PR; README refreshed
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
The retrace after the label-swap fix showed getParameters unchanged
(~2.5s / ~4% throttled main thread) — the real driver is Three r158+'s
transparent-DoubleSide two-pass path: renderBufferDirect renders such
objects back side then front side, setting material.needsUpdate BOTH
times, i.e. a full getParameters/program-cache lookup twice per object
per frame, plus double draw calls. (Found by reading the two-pass
branch in the vendored three.module.min.js right next to the
getParameters call site.)
All 18 transparent DoubleSide materials in this renderer are flat
unlit quads — technique markers, sustain rails, chord frames, lane
planes, halo bars — where the two-pass self-occlusion ordering buys
nothing. Declare forceSinglePass: true on all of them.
Also corrects the _setLabelMap comment's churn attribution (that fix
removes the label-swap contribution; this one removes the dominant
source). Plugin 3.31.1 -> 3.31.2.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A 4x-CPU-throttled retrace (the honest weak-hardware proxy) surfaced
three residual per-frame costs; stack attribution pinned each:
- getParameters/getProgramCacheKey (~4% of main thread): every pooled
label sprite map swap set material.needsUpdate, bumping
material.version and forcing full program re-resolution next render.
Swapping between two non-null cached textures never changes the
compiled program (USE_MAP define unchanged) — new _setLabelMap()
helper only flags needsUpdate on a null<->texture transition, used at
all 7 swap sites.
- getBoundingClientRect (~1.2%): the 3D highway's per-frame canvas-size
self-check forced a layout read every frame. The CSS-box drift read
now runs every 10th frame (or when the wrap isn't pinned); the
backing-store comparison stays per-frame with cheap property reads
and forces an immediate box read + applySize when it fires.
- set textContent: the core 60 Hz HUD clock rewrote hud-time (and
getElementById'd it) every tick for a display that changes 1/s — now
write-on-change with a cached element ref.
(The remaining textContent writer in the trace is notedetect's
badges.js — external repo, to be filed there.)
tests/js: resize-reframe shape test updated for the hoisted _bsChanged
gate, incl. an assertion that the throttle can never delay the
backing-store path.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- isVisible() forces a fresh DOM sample (was serving the rAF loop's
throttled cache, contradicting its 'live DOM check' docstring).
- v3 chrome: reconcile the edge-driven overControls hover flag against
matches(':hover') on the throttled ~6 Hz tick — covers a missed
mouseleave (flag stuck true, transport never hides) and a re-created
#player-controls node with lost listeners.
- highway_3d pre-warm now also covers teachFg/teachSd label textures
and the technique sprite factories (mute X, hammer/pull triangles,
bend chevrons, slide arrows) per active-palette string colour, plus
a maintenance note tying new label styles to the warm list.
- Document that the visibility throttle's manual invalidations are
latency-only (periodic resample self-heals within ~10 frames), and
why highway_3d keeps its local lowerBoundT (downlevel hosts).
External-repo audit (finding 1): staffview, tabview, piano, drums,
keys_highway_3d, drum_highway_3d grepped — no cross-frame bundle
retention or bundle-identity checks found.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* Initial commit — clean relaunch
* Initial commit — clean relaunch
* Remove external game/format terminology from docs and code
Reword references to the external game and its proprietary file formats
in comments, docstrings, UI text and identifiers; no behaviour change.
* Repoint dead slopsmith URLs -> got-feedback
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(midi): consume core midi-input domain + session robustness fixes (#2)
* feat(midi): consume the core midi-input domain instead of private requestMIDIAccess (#881)
Route Web-MIDI device access through window.slopsmith.midiInput
(discover/list/select/open/close) rather than a private
navigator.requestMIDIAccess(). One shared device-access boundary with
piano/drums/onboarding; retires this plugin's private Web-MIDI. The
note-detection 'midi' exact-verdict provider role and the audio-input
source export are preserved (now backed by midi-input source data). Saved
pick ({id,name}) stays compatible — domain sourceId == the old MIDIInput.id.
Live listener is the domain handle's addListener/removeListener; the async
init/resume/pause gate and device-vanish handling are kept.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): stop exporting MIDI into the audio-input domain
Keys MIDI now lives in the dedicated midi-input domain (#881). Continuing to
register pseudonymized 'midi-input-N' sources into audio-input polluted audio
device pickers — notably the onboarding guitar audio-input dropdown showed MIDI
devices under cryptic names. Unregister any leftovers and register none.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): guard async connect races, denied-discovery latch, key consistency
Codex preflight findings on the midi-input consumption:
- Don't latch `_midiReady` when discover() resolves denied/unavailable — only
on a handled outcome — so reopening retries the permission prompt.
- Add a generation guard to async `_midiConnect`: a slower open() from an
earlier selection can resolve after a device/None switch and install a stale
handle/listener. Discard (and close) superseded opens.
- Carry the domain `logicalSourceKey` on the source/selection descriptor and use
it for select/open/close instead of synthesizing `web-midi::<id>`.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): invalidate pending opens on detach; clear phantom selection; manifest
Codex re-review (round 2):
- A detach driven by device removal (sources-changed) didn't advance
_midiConnectSeq, so a pending _midiConnect open could resume and install a
handle/listener for a now-gone source. Bump the generation inside _midiDetach
and capture myGen after it, so any later detach supersedes an in-flight open.
- If mi.open() yields no handle, _midiInput stayed set — a phantom connected
device the render loop's miss-sweeping would penalize. Clear it on the
no-handle and catch paths.
- plugin.json: replace the stale audio-input provider role (the removed
MIDI-into-audio-input export bridge) with the midi-input requester it now uses.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): make _midiResume idempotent under multiple live instances
Codex re-review (round 3): with the domain addListener API, a second live
renderer instance (splitscreen/overlapping lifetimes) calling _midiResume() while
already active could register the same listener again and double-deliver a MIDI
note to the focused instance (hit + duplicate misses). Return early when already
active rather than relying on the provider's Set-backed de-dup.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): release domain session on teardown; counter-free legacy source cleanup
Codex re-review (round 4):
- [P2] Last-instance teardown only detached the listener and never released the
midi-input domain session, leaking the requester ref. Add _midiReleaseSession()
(delegates to _midiDetach: close + null + generation bump) at the destroy site;
re-mount's _midiInit auto-connects from the saved pick.
- [P3] The legacy audio-input MIDI-source cleanup looped over the module-local
_aiRegisteredCount, which resets to 0 on an in-page upgrade so the prior build's
'keys-midi:input-N' entries were never unregistered. Iterate a fixed bound over
the known sourceId pattern instead (unregister of an absent source is a no-op).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): gate miss-sweep on live handle; don't open without a live renderer
Codex re-review (round 5):
- Miss-sweeping was gated on _midiInput, set as soon as a device is picked, but
the async mi.open() may still be pending (slow / permission prompt) — notes
passing during that window banked false misses. Gate on _midiHandle, truthy
only after a handle is opened and wired.
- A settings-only ensure-init (or a discover resolving after the last instance
was torn down) could open a midi-input session with no renderer to release it.
Gate mi.open() on _instances.size > 0; the pick is saved and a later mount
re-runs auto-connect.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): guard open-failure clobber + don't clobber saved device on unplug
Bringing keys-highway-3d in line with the piano/drums hardening:
- The _midiConnect catch cleared _midiInput unconditionally; a stale older open's
rejection could wipe a newer connect's _midiInput/_midiHandle (and leak the
handle). Only clear when myGen === _midiConnectSeq.
- The sources-changed recovery called _midiAutoConnect(), whose fallback persists
a substitute device — overwriting the user's saved pick on a transient
multi-device unplug. Parameterize _midiAutoConnect(allowFallback); recovery
passes false (reconnect the saved device only, never a persisted fallback).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): keep first-hotplug working; don't tear down a live session on re-init
Codex round-11:
- The unplug-recovery _midiAutoConnect(false) skipped ALL fallbacks, which broke
first-hotplug (open visualizer with no keyboard, plug one in → never picked →
no connect). Skip the substitute only when a saved pick exists but is absent
(preserve it); allow the fallback when nothing was ever picked.
- _midiInit re-ran _midiAutoConnect on every ready re-init (settings/splitscreen),
tearing down the live handle + releasing held keys. Only re-connect when there's
no live handle.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(midi): consume core midi-input domain + session robustness fixes (#2)
* feat(midi): consume the core midi-input domain instead of private requestMIDIAccess (#881)
Route Web-MIDI device access through window.slopsmith.midiInput
(discover/list/select/open/close) rather than a private
navigator.requestMIDIAccess(). One shared device-access boundary with
piano/drums/keys/onboarding. Saved pick ({id,name}, with legacy id-only
fallback) stays compatible — domain sourceId == the old MIDIInput.id. The
async in-flight init guard, _midiActive teardown gate, device-vanish handling,
and settings device APIs are preserved; the live listener is the domain
handle's addListener/removeListener. Degrades to no-MIDI when the domain
is absent.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): guard async connect races, denied-discovery latch, key consistency
Codex preflight findings on the midi-input consumption:
- Don't latch `_midiReady` when discover() resolves denied/unavailable — only
on a handled outcome — so reopening retries the permission prompt.
- Add a generation guard to async `_midiConnect`: a slower open() from an
earlier selection can resolve after a device/None switch and install a stale
handle/listener. Discard (and close) superseded opens.
- Carry the domain `logicalSourceKey` on the source/selection descriptor and use
it for select/open/close instead of synthesizing `web-midi::<id>`.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): invalidate pending opens on detach; clear phantom selection; manifest
Codex re-review (round 2):
- A detach driven by device removal (sources-changed) didn't advance
_midiConnectSeq, so a pending _midiConnect open could resume and install a
handle/listener for a now-gone source. Bump the generation inside _midiDetach
and capture myGen after it, so any later detach supersedes an in-flight open.
- If mi.open() yields no handle, _midiInput stayed set — a phantom connected
device the render loop's miss-sweeping would penalize. Clear it on the
no-handle and catch paths.
- Declare the midi-input requester capability (degrade-noop) in plugin.json.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): make _midiResume idempotent under multiple live instances
Proactively mirror the keys-highway-3d round-3 fix (same multi-instance lifecycle
and identical code): a second live renderer instance calling _midiResume() while
already active could double-register the listener and double-deliver a MIDI hit.
Return early when already active.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): release the domain session on final teardown
Mirror the keys-highway-3d round-4 fix: last-instance teardown only detached the
listener and never released the midi-input domain session, leaking the requester
ref. Add _midiReleaseSession() (delegates to _midiDetach: close + null +
generation bump) at the destroy site; re-mount's _midiInit auto-connects.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): gate miss-sweep on live handle; don't open without a live renderer
Codex re-review (round 5):
- [P2] drumH3dEnsureMidiInit (settings-only, no renderer) auto-connected and
opened a midi-input session that destroy()/_midiReleaseSession would never run
to release — held until reload. Gate mi.open() on _instances.size > 0; the pick
is saved and a later renderer mount re-runs auto-connect.
- Mirror the keys-highway-3d miss-sweep fix: gate accumulation on _midiHandle
(live wired session), not _midiInput (set before the async open resolves), so
notes during a pending open don't bank false misses.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): guard open-failure clobber + reconnect saved device on replug
Bringing drum-highway-3d in line with the piano/drums hardening:
- The _midiConnect catch cleared _midiInput unconditionally; a stale older open's
rejection could wipe a newer connect's _midiInput/_midiHandle. Only clear when
myGen === _midiConnectSeq.
- The sources-changed handler only refreshed the list (never reconnected), so a
replug of the saved device didn't reattach. Reconnect on sources-changed via
_midiAutoConnect(false) — saved device only, no fallback, so a transient unplug
can't switch to / persist another input.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): keep first-hotplug working; don't tear down a live session on re-init
Same two round-11 fixes as keys-highway-3d (shared structure): the recovery
_midiAutoConnect(false) now allows a fallback when nothing was ever picked
(first-hotplug) while still preserving a saved-but-absent pick; and _midiInit
only re-connects when there's no live handle.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): select by logicalSourceKey, not provider-local sourceId (#4)
Codex review of the midi-input migration: resolving a selection with
find(s => s.id === id) treats the provider-local sourceId as globally unique,
so when the core midi-input domain aggregates multiple providers two devices
sharing a sourceId can't be distinguished — the wrong device opens and unplug
detection can miss that the selected key vanished.
Thread the globally-unique logicalSourceKey through save/read, _midiConnect,
auto-connect reconnect, the public keysH3dSetMidiInput entry, and the
capability _aiOpen path, preferring the key and keeping the bare sourceId /
name only as a legacy fallback. Mirrors the already-merged drums/piano
migration (_midiResolveSaved). Self-reviewed against that reference.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(midi): select by logicalSourceKey, not provider-local sourceId (#3)
Codex review of the midi-input migration (same issue as keys-highway-3d):
resolving a selection with find(s => s.id === id) treats the provider-local
sourceId as globally unique, so when the core midi-input domain aggregates
multiple providers two devices sharing a sourceId can't be distinguished —
the wrong device opens and unplug detection can miss the vanished key.
Thread the globally-unique logicalSourceKey through save/read, _midiConnect,
the auto-connect reconnect, and the public drumH3dSetMidiInput entry,
preferring the key with the bare sourceId / name only as a legacy fallback.
Mirrors the already-merged drums/piano migration. Self-reviewed.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* chore: add plugin metadata (category, description, thumbnail) (#5)
* chore: add plugin metadata (category/description/thumbnail)
Refs got-feedback/feedBack#571
* chore: add placeholder thumbnail
Refs got-feedback/feedBack#571
* chore: add plugin metadata (category, description, thumbnail) (#4)
* chore: add plugin metadata (category/description/thumbnail)
Refs got-feedback/feedBack#571
* chore: add placeholder thumbnail
Refs got-feedback/feedBack#571
* fix(viz): register feedBackViz_<id> so the player viz picker lists this plugin (#5)
After the slopsmith->feedBack rename the factory was still registered as
window.slopsmithViz_<id>; the host viz picker only looks up
window.feedBackViz_<id> (no slopsmithViz alias exists) and silently skipped
this plugin. Add the feedBackViz_ global (aliased to the existing
slopsmithViz_ one for back-compat).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(viz): register feedBackViz_<id> so the player viz picker lists this plugin (#6)
After the slopsmith->feedBack rename the factory was still registered as
window.slopsmithViz_<id>; the host viz picker only looks up
window.feedBackViz_<id> (no slopsmithViz alias exists) and silently skipped
this plugin. Add the feedBackViz_ global (aliased to the existing
slopsmithViz_ one for back-compat).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(plugins): bundle drum_highway_3d + keys_highway_3d as in-tree core plugins
Import both 3D highway plugins from their standalone repos via git subtree
(history preserved), ahead of the visual-parity epic that ports the guitar
highway's polish to them.
- .gitignore: !plugins/drum_highway_3d/ + !plugins/keys_highway_3d/ exceptions
- keys plugin.json: "bundled": true (drum already had it)
- CI: JS test step gains 'plugins/*/tests/*.test.js' (+20 keys tests)
- static/tailwind.min.css regenerated (core build scans plugins/**)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(drum_highway_3d): narrow Auto-mode predicate so bundling can't steal guitar arrangements
has_drum_tab is pack-level; first-match-wins Auto order sorts this plugin
before highway_3d. Claim only drum arrangements, or packs nothing more
specific can render (Codex preflight P2).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* docs: CHANGELOG — note the deliberate Auto-predicate narrowing
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(drum_highway_3d): MIDI lifecycle — no false misses on mid-song connect, promote survivor on destroy
- _missSweepFloor exempts notes that passed before the device wired up
from the miss sweep (lowered on seek-back; cleared with scoring resets)
- destroy() promotes a surviving instance to _activeInstance so
splitscreen panel teardown doesn't drop all MIDI routing
(Codex preflight round 2, both pre-existing in the imported code)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Sin <deathlysin@outlook.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Static-analysis follow-ups to the trace-backed fixes; each is cheap
insurance on machines where the profiled headroom doesn't exist.
- highway.js: _makeBundle now mutates one persistent per-instance
object instead of allocating a fresh ~35-field bundle every rAF
frame (xN under splitscreen). Object identity is stable and
meaningless; array fields still swap reference on chart changes,
which field-identity caches rely on. Contract documented in both
CLAUDE.mds.
- highway.js: new bsearchTime (lower-bound on .time) windows the
default 2D renderer's beat-line scan (was O(all beats) per frame);
bundle.lowerBoundT / bundle.lowerBoundTime expose the searches to
custom viz so they stop reimplementing visible-window culling.
- highway_3d: localStorage 'h3d_full_sus' polled at ~1 Hz instead of
every frame (synchronous storage read on the hot path).
- highway_3d: drawLyrics caches the measureText row layout keyed on
(lyrics ref, line index, shown count, font size, width) — per-frame
work is now just drawing over cached widths.
- tests/js: bundle source-shape assertions widened to accept the
assignment form ([:=]) alongside the old object-literal form.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Trace showed frame spikes from Three.js first-use costs mid-song:
shader program compilation (getParameters/getProgramCacheKey) and lazy
texture uploads (texSubImage2D) whenever a chord name, section banner,
or fret label first appeared.
- ren.compile(scene, cam) after initScene (pools already warmed by
feedBack#226, board built, background mounted) so programs compile
during the load spinner.
- Pre-rasterise + GPU-upload (ren.initTexture) the deterministic txtMat
entries: fret numbers 0-24 in the noteFret/fretRow/ghostFret combos
the per-frame paths request.
- Chart-dependent labels (chord template names, section names) prewarm
once on the first draw() after each init, when bundle arrays are
guaranteed populated.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Chrome trace showed ~0.5s self-time in _isHighwayVisible (offsetParent
read every rAF frame forces style/layout recalc) and ~1.5s in the v3
player-chrome loop (matches(':hover') per frame, unconditional
textContent/width writes at 6 Hz -> ~1800 layout passes in 63s).
- highway.js: sample offsetParent every 10th frame, cached in between;
fresh sample forced on init/canvas-replace/resize/override-clear.
- player-chrome.js: hover tracked via mouseenter/mouseleave; Up-Next
refs cached, text written only on change (eta coarsened to 1s steps
beyond 10s), progress bar moved from width to scaleX (compositor-only).
- v3.css: bar fill uses transform-origin:left + scaleX transition.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(v3): sort the library by mastery (needs-practice / most-mastered)
Adds two sort options to the Songs library: "Needs practice first" (weakest
measured accuracy first) and "Most mastered first". Mastery = MAX(best_accuracy)
across a song's arrangements, from song_stats; because that's a separate table
it's a correlated subquery in the ORDER BY, so these sorts use OFFSET paging like
tuning/year. Unscored ("not started") songs always sort to the bottom in both
directions, so a large unpracticed library doesn't bury the songs you're
actually working on. Never the default.
Verified against a running server: scored songs at 0.90 / 0.30 plus unscored ->
ascending orders 0.30, 0.90, unscored; descending 0.90, 0.30, unscored.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* feat(v3): add the 3-state mastery filter (Mastered / In progress / Not started)
Complements the mastery sort with the gentler filter the charrette preferred: a
"Progress" facet in the Filters drawer with Mastered (>= 0.9), In progress
(attempted but < 0.9), and Not started (no score) -- multi-select, OR within the
set. Server-side via a correlated subquery on song_stats threaded through
_build_where / query_page (passed as a separate kwarg so query_artists /
query_stats are unaffected). Smart-collection providers ignore it (they define
their own filters).
Verified live: mastered -> the 0.90 song, in_progress -> the 0.30 song,
not_started -> the unscored songs, combined -> both scored (with correct totals).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Playing a playlist previously played one song and returned to the menu -- a
play-queue was never implemented. Add window.feedBack.playQueue (start / advance
/ hasNext / clear) and a "Play all" button on the playlist detail.
Advancing rides the same exit choke point as auto-exit and a results-card close:
song-end paths call window.closeCurrentSong() (the auto-exit grace timer and a
results screen's release()), so wrapping it plays the next track instead of
returning to the menu -- advancing AFTER the user dismisses a score card, not
through it. A user-initiated exit (Escape / the close button) uses the bareword
closeCurrentSong(), left untouched, so leaving the player still leaves and
abandons the queue. playSong gains a fromQueue guard (a manual play abandons a
stale queue) and closeCurrentSong clears the queue on a real close. Binds via
song:ended / the choke point, not the <audio> element, so it advances on the
desktop (JUCE) route too. The no-queue path is unchanged.
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Add a media-server-style "I dropped files in my folder, hit refresh" control to
the Songs toolbar. Refresh triggers an incremental /api/rescan and reuses the
existing /api/scan-status poll for live progress: a 3-state button (idle /
"Scanning..." while listing / "Scanning N/M" once counting) with a title tooltip
showing the current file + percent. A scan already running (the Settings buttons
or a background pass) is reflected on the button too. On completion it emits
library:changed so the grid reloads, and shows an honest, never-punishing
fbNotify toast (bottom-right, suppressed while in a song). No backend change --
same machinery the Settings rescan already drives. The precise "N added" count
is a follow-up.
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
The old add-to-playlist flow crammed every playlist into one uiPrompt label
("1. Foo 2. Bar 3. Baz ...") and asked the user to type a number -- unreadable
past a couple of playlists, and reported as a bug.
Replace it with openPlaylistPicker: a checkbox modal with membership pre-check (a
song already in a playlist shows checked; a multi-song selection shows an
indeterminate box when only some are in), an inline "+ New playlist" row, and a
search box once the list is long. Toggling adds/removes via the existing REST
(POST + DELETE .../songs/{filename}); only playlists the user actually touched
change. One shared function still feeds the per-card menu, the batch bar, and the
batch button, so the fix lands in all three. Escape / backdrop close; a
bottom-right fbNotify toast confirms.
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
The Songs library reset sort + filters to defaults on every visit; testers asked
it to remember their choice ("most players pick a preferred sort and leave it").
Persist sort, format, view, and the drawer filters to localStorage
('v3:songs-prefs'), restored once at first build. Cold start stays the neutral
Artist A-Z default. The search query and the artist/album drill-down are
deliberately NOT persisted (navigational). Single write point in reload() (where
every change already funnels); every restored value is validated against its
option list, so a removed/stale setting can never wedge the toolbar. Global (not
per-provider) for now.
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Working-tuning follow-ups:
- Mic-verify used _state.currentSongOffsets for the 'verified' stamp, but for a
MANUALLY-selected tuning (tuner opened off a song) that's a stale/different
song's tuning — so verify could mark the WRONG tuning verified. It now derives
the verified offsets from the tuning actually being checked (its target freqs;
the player's reference pitch cancels in the ratio), so 'verified' always
attaches to the tuning the player confirmed. Explicit offsets still win.
- Adds docs/working-tuning-on-device-tests.md: the checklist for the parts that
can't be covered headlessly — the auto-open/gate flow, both-directions prompts,
mic-verify detection, and the tuner-mic-vs-note_detect ASIO/exclusive-mode
contention flagged in the design charrette.
Test: mic-verify with no explicit offsets / no song context derives the correct
offsets (Drop-D). 55 tuner tests green.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(tuner): mic-verify — promote the working tuning assumed→verified via a per-string check (working-tuning PR 9b)
Adds the choreographed per-string mic verification the design reserved for
'verified' provenance (audio-engine's honesty rule — nothing else may claim it):
the player plays each string, and once every one reads in tune (±6 cents) and
holds stable for 8 frames, the tuner stamps the working tuning
provenance:'verified' + verifiedStrings via workingTuning.set.
- screen.js: a pure verify state machine (verifyStart/verifyFeed/verifyCancel/
verifyState, exposed on the tuner API) + the set-verified writer; cancels on close.
- ui.js: updateUI feeds each processed frame (matched string + cents) into the
session; a "Verify tuning" button + per-string progress + status, shown for a
selected (non-free) tuning.
Pairs with the 9a lifecycle: a 'verified' decays back to 'assumed' on the next
song load, so mic-verify is a per-session confidence boost, never a sticky claim.
Tests: tests/js/tuner_auto_open.test.js +4 (all-strings->verified, out-of-tune
never completes, streak resets on drift, API exposed / only it claims verified) —
33/33. The state machine is headless-verified with synthetic frames; the real
per-string mic detection + the button flow need an on-device pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* fix(tuner): mic-verify writes the confirmed tuning + no clobber + stricter streak (PR #670 review)
Review fixes for mic-verify (working-tuning PR 9b):
- 'verified' could attach to STALE offsets: _publishVerified stamped provenance
without writing offsets, so the slot's pre-tuning offsets got marked verified.
verifyStart(targets, offsets) now captures the confirmed tuning's offsets, and
_publishVerified writes offsets + stringCount + instrument + referencePitch +
verifiedStrings ATOMICALLY with provenance:'verified' into the selected slot
(and refuses to stamp verified with no concrete offsets).
- The assumed publish-on-clear immediately clobbered a just-earned 'verified':
disable() now skips it when a mic-verify wrote verified this session
(_verifiedPublished).
- The per-string streak could accumulate across silence / wrong-string frames.
verifyFeed now requires CONSECUTIVE in-tune frames: the one confirmed string
advances, every other unfinished string resets each frame.
- A mid-verify tuning change (song switch) left stale captured offsets; verify is
now cancelled in _syncCurrentTuning when the song tuning changes.
Tests: verify writes the confirmed offsets (not stale); source-guard for the
no-clobber path. 47 tuner + 77 tuner/capability tests green. Codex-reviewed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* feat(tuner): gate playback until you've tuned — hold autoplay + no-trap Skip/Back/Esc (working-tuning PR 4)
When the opt-in auto-open fires because a song needs a different tuning, playback
now WAITS behind the tuner instead of starting underneath it — the "tune before
you play" model. Built on a new generic core hook window.feedBack.holdAutoplay()
(mirrors holdAutoExit): the tuner claims the hold synchronously on song:loading
(beating the song:ready autostart) and releases it — or a 12s fail-open backstop
does — so a wedged plugin can never strand a song. Generation-guarded; manual
Play always wins.
No one-way trap:
- Skip = "I've tuned" -> plays and records the song's tuning as the instrument's
current working tuning (the explicit write-point PR 3 left as 'assumed').
- Back to library / Esc -> leave the song, record nothing (reuses requestExitSong;
Esc is the existing player shortcut).
- The in-panel x is dropped for an auto-open — Skip/Back/Esc are the dismiss
surface. This also keeps the write honest: Skip is the only on-player dismiss
that records, so leaving never falsely records a tuning.
Stacked on #660 (working-tuning PR 3). Core app.js gains only the generic hook
(a test asserts it never references the tuner's internals); shell-agnostic.
Needs a desktop smoke-test that the tuner mic doesn't contend with note_detect's
scoring input under ASIO/exclusive mode (per the design charrette).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* fix(tuner): backstop can't cut off tuning + gate race/token hardening (PR #666 review)
Review fixes for the autoplay gate:
- The 12s fail-open backstop could start playback UNDER a legitimately-open tuner
(a slow / mic-verify retune > 12s). holdAutoplay()'s release now carries a
.settle() that cancels the backstop; the tuner calls it once the tuner is
confirmed open (_gateClaimed), so the hold becomes deliberate and only a
dismiss / song switch releases it. (Fail-open still covers "claimed but wedged
before deciding".)
- The async song:ready handler could release a NEWER song's gate after its await
(global _gateClaimed, no guard). It now snapshots _autoOpenGeneration and bails
if a newer song took over.
- holdAutoplay guarded by song generation, not per-hold — a stale release from an
earlier hold could clear a later one. Each hold now mints a unique token that
release()/settle() must match.
Tests: source-level assertions for the token, settle(), the settle-on-open call,
and the song:ready gen-guard. 45 tuner+speed tests green. Codex-reviewed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
Hardens the host workingTuning capability (PR 1) with the "polish & safety"
lifecycle:
- Idempotent re-injection: a second load of the module no longer replaces the live
state with a fresh (empty) one — it early-returns once registered.
- Opt-in "launch tuning" default (setLaunchDefault/getLaunchDefault/clearLaunchDefault):
a per-instrument, localStorage-backed seed the player can opt into ("start me in
THIS tuning on app open"). Boot seeds from it when set, else /api/settings as before.
Off by default — a SEED only; the live tuning still resets on restart.
- Verified decay: on song:loading the current instrument's 'verified' provenance
decays to 'assumed' (offsets kept) — a per-string mic check is only trustworthy for
the context it was done in, so a stale 'verified' can never suppress a needed prompt.
Adds a state-machine smoke suite (tests/js/working_tuning_capability.test.js, 12/12):
defaults, per-instrument isolation, both-directions, verified-invalidation-on-retune,
decay-on-song-load, resetToDefault, launch-default set/seed/clear, idempotent
re-injection, the change event.
The opt-in UI + the mic-verify writer land with the tuner (PR 9b).
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* feat(v3): show the live working tuning on the instrument card (working-tuning PR 5)
The topbar instrument selector now surfaces the selected instrument's live
working tuning from the host workingTuning capability (PR 1): a compact label on
the card — dim while you're in your home tuning, amber once you've retuned — and,
in the dropdown, a "Now in: <tuning> <assumed/verified glyph>" banner with a
one-tap "Back to default" (resetToDefault). Switching instrument/strings calls
setCurrentInstrument so the card follows the right instrument (guitar's tuning vs
bass's, tracked separately); it re-renders on working-tuning-changed.
Names offsets via the shared window.displayTuningName resolver. Fully
feature-detected — without the workingTuning capability the card renders exactly
as before. v3-only, single file (static/v3/badges.js).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* fix(badges): consistent home-tuning check + gate working-context on save (PR #667 review)
Review fixes for the instrument-card working-tuning label:
- isHome used display-label equality where the home label was the raw settings
string ('Custom') but the working label came from displayTuningName(offsets) —
a real home tuning could show amber. Both now resolve through the same namer.
- The instrument pill moved the working-tuning context (setCurrentInstrument) even
when saveSettings() rejected the patch, desyncing the selector from the card.
saveSettings() now returns whether it was accepted; the pill only switches the
working context on success.
(No boot-time setCurrentInstrument: the host already seeds its current instrument
from /api/settings and emits working-tuning-changed on hydration — which this card
re-renders on — so pre-touching would only race/suppress that seed.)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* feat(v3): flag library songs by working-tuning match (working-tuning PR 6)
Each song's tuning chip in the v3 library grid is now coloured by whether your
CURRENT working tuning covers it: green = play it now, amber = needs a retune
(with a matching tooltip). Uses the tuner plugin's coverageReport (async), so it
runs as a post-paint decoration pass — chips render instantly, then colour a tick
later; a token cancels a superseded pass so scrolling stays snappy. Re-flags on
working-tuning-changed (retune / instrument swap / reset), no re-fetch.
Fully feature-detected: without the tuner coverage API + the host workingTuning
state, the chips render exactly as before. v3-only, single file (static/v3/songs.js).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* fix(tuner/library): correct bass matching + memoize player tuning (PR #668 review)
Review fixes for working-tuning PR 6 (library tuning-match chips):
- Bass songs were scored against the guitar tuning. The chip passed no arrangement
to coverageReport, so isBassArrangement fell back to guitar — a 4-string bass
drop-D read as guitar could FALSE-MATCH a drop-D guitar player (green). songCard
now flags a bass-only song (every arrangement name matches /\bbass\b/) with
data-tuning-bass, and decorateTuningChips passes arrangement 'Bass'/'Lead' so
coverage uses the right base pitches. Mixed guitar+bass songs → guitar (the
song-level tuning is the guitar one); least-wrong given one tuning per song.
- Per-chip /api/settings fetch storm. coverageReport()→_playerTuning() fetched
/api/settings once per visible chip per grid paint (~60). _playerTuning is now
memoized (the player's tuning is song-independent) so all callers share one read;
invalidated on instrument:changed / working-tuning-changed, with a 3s TTL so a
settings write that doesn't emit an event still heals. A transient fetch failure
is NOT cached (next read retries) — else one hiccup would freeze coverage.
Tests: player tuning shared across songs (one fetch); transient-failure retry
(fails without the fix). The prior #680 dedup test updated for the memoized behavior.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* Tuner: passive "different tuning" badge cue that names the retune
Building on the coverage check: when you enter a song your current
instrument doesn't cover, the topbar tuner badge gets an amber ring + a
tooltip naming the change (e.g. "retune B->A", or "the reference pitch"
for an A440 vs A432 mismatch). Advisory only -- it never auto-opens the
panel; recomputed on song:ready, cleared on song-load / leaving the
player.
Refactors the coverage check into a structured report
(window._tunerAutoOpen.coverageReport -> { covered, retune:[{from,to}],
reference, cantCover }); the boolean gate now wraps it. The cue is
CSS-free (inline ring + native tooltip, no Tailwind rebuild) and no-ops
when the tuner plugin is absent.
Touches static/v3/badges.js (cue) + plugins/tuner/screen.js (report).
v3-only. Stacked on #656 (issue E stage 2.5/3). The splitscreen-suppress
and no-usable-input guards move to E2 (the playback gate).
Tests: tests/js/tuner_auto_open.test.js (report names the strings,
reference mismatch, badge wiring).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* feat(tuner): read/write the live per-instrument working tuning — both-directions retune prompt (working-tuning PR 3) (#660)
The §4 coverage check compared each song against the player's fixed
instrument-profile tuning, so the tuner only ever prompted *away* from a "home"
tuning (E -> Drop C#) and stayed silent coming back (Drop C# -> E), even though
the player had physically retuned.
_playerTuning() now reads the host's live per-instrument working tuning
(window.feedBack.workingTuning, keyed by the selected instrument from
/api/settings) instead of re-deriving from the static settings tuning, so
coverage is measured against what the instrument is ACTUALLY in and prompts both
directions. On clearing an auto-opened tuner, _publishWorkingTuning() writes that
song's tuning as the instrument's live working tuning ('assumed' — PR 4's
explicit "I tuned / Skip" refines the write-point), so the next song is judged
against where the player now is.
Per-instrument (guitar vs bass tracked separately). Feature-detected: falls back
to the static /api/settings tuning when the working-tuning capability is absent,
so the 27 existing coverage tests are unchanged. Builds on PR 1 (host
workingTuning) + PR 2 (instrument->chart routing).
Tests: tests/js/tuner_auto_open.test.js — +2 (both-directions coverage via a live
Drop-D working tuning; publish-on-clear targets the right instrument slot); 29
pass total.
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(tuner): transactional open + fail-closed auto-open config (tuner-E #655 review) (#681)
Two review fixes for the auto-open opt-in+persist stage:
- enable() wasn't transactional. The panel (with the ×/Skip buttons) is shown
before `await _tunerAudio.start()`, and `_state.enabled` was only set after it.
A ×/Skip dismiss during that await hit disable() with wasEnabled=false, then
enable() completed and flipped enabled on — an enabled-but-hidden zombie. Guard
the open with an `_openGen` token bumped on every enable()/disable(); after the
audio-start await, bail if superseded instead of enabling. Closes#675.
- Config wasn't fail-closed. routes.py normalized the opt-in with
bool(data.get("autoOpenOnTuningChange", False)), so "false"/"0"/junk coerced to
True. Accept only a real JSON boolean. Closes#676.
Tests: tuner_auto_open.test.js (dismiss-mid-open stays disabled — fails without
the token guard), test_config.py (auto-open default-false + fail-closed on
non-bool). 34 JS + 24 config tests green.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(tuner): coverage stays conservative when the instrument is unknown (tuner-E #656 review) (#682)
_playerTuning() is documented as conservative ("missing data → not covered → still
prompt"), but when /api/settings carried no instrument identity (a fresh profile:
_default_settings() omits instrument/string_count/tuning) it invented guitar/6/440/
standard, so an unconfigured player was treated as 6-string E-standard and coverage
suppressed the auto-open (and badge cue) for matching songs. The post-#660 rewrite
only returned null when the whole fetch failed (!s), not when settings existed but
lacked an instrument.
Now return null unless there's a confident identity — any of instrument/string_count/
tuning in settings, or live working-tuning offsets. A configured standard guitar still
covers a standard song (no regression). Closes#677.
Tests: tuner_auto_open.test.js — empty-settings → not covered (fails without the fix);
configured standard guitar → still covered.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(tuner): badge coverage cue staleness + unknown-as-warning + dedupe (tuner-E #657 review) (#683)
Three review fixes for the passive "different tuning" badge cue stage:
- Stale async cue (#678): _refreshCoverageCue awaited coverageReport then wrote the
DOM unconditionally, so a slow /api/settings fetch could restore the previous
song's amber ring after song:loading / leaving the player. Add a monotonic token
bumped on every refresh and both clear paths; apply the awaited report only if the
token still matches.
- "Unknown" rendered as "needs retune" (#679): the plugin returns a conservative
all-false report on a fetch hiccup; the cue painted that as an amber "retune the
reference pitch" ring. Collapse a no-signal report (not covered, no reference /
retune / cantCover) to null (no cue) via _meaningfulReport(). A genuine not-covered
report always carries reference / retune / cantCover, so real cues are preserved.
- Duplicate /api/settings fetch (#680): the auto-open gate and the badge cue both
call coverageReport() per song:ready. Cache the coverage promise per song (keyed by
session + tuning + centOffset) so they share one fetch; invalidate on song:loading,
instrument:changed, and working-tuning-changed so it can't go stale within a song.
Tests: tuner_auto_open.test.js — concurrent reports share one fetch, a new song
refetches (fails without the cache). 34 JS tests green. Codex-reviewed.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: Byron Gamatos <xasiklas@gmail.com>
With the opt-in auto-open on, only prompt when the player's current
physical tuning doesn't already cover the song. FeedBack is tune-to-song
(the highway draws tab in the song's tuning), so the check aligns the
song's open-string tuning string-for-string against the player's
instrument:
- An 8-string F# player gets no prompt for a 6-/7-string standard song
(its top strings already match those tunings).
- A song needing an open string the player lacks (e.g. a Drop-A
7-string's low A on an F# 8-string) still prompts.
- A whole-instrument reference difference (A440 vs A432, or an
octave-down centOffset, previously ignored) also prompts.
Reads the player's instrument from core /api/settings (the v3 instrument
selector, a stable physical reference); conservative fallback (prompt)
when undeclared or unavailable, so a real retune is never silently
skipped. v3-only. All in plugins/tuner/screen.js; no core changes.
Stacked on #655. Follow-up E1.6: a passive badge cue that names the
strings to retune, plus splitscreen / no-usable-input guards.
Tests: tests/js/tuner_auto_open.test.js (covered/uncovered, the Drop-A
case, reference mismatch, direct contiguous alignment).
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
The tuner self-closes on song:play; autoplay fires it right after a song
switch, so an auto-opened tuner flashed shut ~1s later. An arrangement
switch (which never arms autoplay) instead persisted — the opposite
tester reports, and not the mic.
- New opt-in setting autoOpenOnTuningChange (tuner Settings, default OFF)
- An auto-opened tuner persists: it ignores the autoplay song:play, stray
outside-clicks, and same-screen re-emits, closing only via the new
in-panel x / Skip buttons or leaving the song. A manual open keeps the
classic click-away / play-to-close behaviour.
- Adds the panel's first in-box close (x + contextual Skip).
- All in the tuner plugin; no core app.js changes.
Default (opt-in vs opt-out) is teed up for Byron to flip one boolean.
Staged follow-ups: E1.5 = instrument-coverage smart prompting + badge
cue; E2 = holdAutoplay gate.
Tests: tests/js/tuner_auto_open.test.js (opt-in gate, persist mode,
play/click-proofing).
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
The Reset handler forced base.enabled = true after copying _ASPECT_DEFAULTS
(where enabled is false) — a leftover from when enabled controlled panel
visibility. Visibility is now independent (Shift+A / ×), so drop the override
and let Reset restore the defaults exactly.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
When a song loads without an explicit arrangement, highway_ws now reads the
player's selected `instrument` from config.json (the same file it already reads
for the default-arrangement preference) and picks the arrangement that matches:
bass -> the Bass part. Guitar — and any unknown/future instrument (drums, keys)
— falls through to the existing preference/most-notes default, which already
lands on a guitar part.
Previously the instrument selector only fed the tuner, so a bass player was
handed the default Lead/guitar chart, and the working-tuning coverage check then
compared a 4-string bass against a 6-string part (always "can't cover"). This is
the instrument->chart routing the working-tuning series leans on.
Server-only (every launch path flows through the WS, so no client change). An
explicit arrangement request always wins, so only the default part chosen on
load changes. Tests: tests/test_highway_ws_instrument_routing.py (bass->Bass,
guitar->default, explicit-wins) — 3 new, existing highway WS tests still green.
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
camUpdate registered every pane each frame regardless of whether the tuner
had ever been opened, so window.__h3dAspectPanes could grow unbounded (prune
runs only while the panel is open) and it ran even for users who never opt
in. Gate _aspectRegisterPane behind __h3dAspectPanelOpen (same gate as the
readout). The pane key is still resolved every frame so saved overrides keep
applying; only the picker bookkeeping is deferred until the panel is open.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
- Memoize _resolveTuneFor per pane, invalidated by a revision bumped on every
tune mutation (all writes funnel through _aspectPersist). Panes with an
override no longer rebuild the merged object every frame; panes without one
still return the base directly.
- _aspectNowMs falls back to Date.now() when the Performance API is absent, so
pane/readout pruning still works in older/borrowed contexts.
- _setAspectPanelVisible prunes stale panes before the first dropdown build, so
panes from a prior song/split don't flash until the first RAF tick.
- Rename _abShortcutRegistered/_registerAspectAbShortcut to
_tunerShortcutRegistered/_registerTunerShortcut — the shortcut opens/closes
the tuner now, it isn't an A/B toggle.
- Fix a stale 'pane1' example in a comment (keys are 'arr:<name>'/'pane:<uid>').
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
When only one pane is live the Target row is hidden, but _aspectEditTarget
could remain a specific pane key — silently routing edits into a hidden
(and persistent arr:*) override in single-player. Reset the edit target to
"" in _aspectBuildTargets whenever the row is hidden (or the selected pane
is gone), so single-pane edits always go to the shared base.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
- Set type="button" on the × close control and the Reset/Copy buttons so
they can never act as submit if the panel is ever nested in a <form>.
- Add aria-label="Target pane" to the Target <select> so screen readers can
identify the control.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
Three fixes from PR review of the per-pane tuner:
- Sync no longer writes back. _syncAspectPanel dispatches synthetic input
events to refresh slider labels; guard those with _aspectSyncing so the
slider handler skips the write. Previously opening/switching a target
populated a full override for every field (defeating sparse inherit) and
spammed localStorage.
- Unchecking "Override held hFOV" on a pane target now clears the override
key (via _aspectClearVal) so the pane re-inherits the base value, instead
of pinning hfovDeg:null in the override. On the base target it still sets
the explicit auto (null).
- _aspectPrunePanes now prunes the matching __h3dAspectReadout slot and drops
a dangling __last, so the readout cache can't grow unbounded as songs and
arrangements churn.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
The Target picker disappeared in split because it keyed panes off the
external splitscreen panel index (panelIndexFor), which isn't always
available — both panes then collapsed to a single 'main' key and the
one-pane row-hide kicked in.
Key panes by arrangement name instead ('arr:Bass'): distinct between split
panes AND stable across songs, with no dependency on the split plugin. A
per-instance id ('pane:N') is the fallback when a pane has no arrangement.
Only arr:* overrides persist to localStorage (instance-id fallback keys are
session-only, so they can't leak a new key each reload). This also gives
nicer semantics — a pane's framing follows its arrangement into the next
song.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
Per-pane overrides were keyed by an ephemeral per-instance id, so leaving a
song and opening another rebuilt the renderer with a new id and the pane's
framing was lost.
Key overrides by the durable split slot again ('main' | 'panel<idx>', via
_bgPanelKey) so the same slot means the same pane across songs, and persist
__panels to localStorage. Keep the anti-flicker fixes that were the actual
cause of the earlier dropdown churn (prune stale panes, rebuild only on a
pane-set change, never rebuild while the select is focused). The slot key is
latched to the last real slot so a transient null from panelIndexFor during
a song/layout transition can't flip it to 'main' and drop the override for a
frame; it resets in destroy() for instance reuse in another slot.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
* feat(core): host per-instrument workingTuning capability + read-API/event (working-tuning PR 1)
Introduce window.feedBack.workingTuning — the live, host-authoritative current
instrument tuning (offsets + string-count + reference pitch + assumed/verified
provenance), distinct from any one song's tuning and from a soft opt-in default.
It's the single source of truth the highway, library, and plugins (tuner,
Virtuoso, minigames) will read so a retune or instrument swap is reflected
app-wide instead of being re-derived per surface.
PER-INSTRUMENT: state is a map keyed by `${instrument}-${stringCount}` (e.g.
guitar-6 / bass-4, the selector's key) — your guitar's tuning and your bass's are
kept separately; get() returns the selected instrument's, and switching the
selector surfaces that instrument's own remembered tuning. You only ever deal
with the one you've picked.
Modeled on the shipped `tuning` capability + the `feedBack.theme` read-API:
synchronous get(instrument?), set(state,{provenance,instrument}) mutator,
setCurrentInstrument(), resetToDefault(), and a `working-tuning-changed` event
that fires on change and once on hydration (carrying which instrument changed).
In-memory, seeded from /api/settings, reset-on-restart. Registered as a separate
`working-tuning` exclusive-owner capability (tuner = sole writer, others read).
Foundation only — pure plumbing, nothing writes to it yet and no behavior
changes. The tuner becomes the writer (and the gate's E->C# asymmetry is fixed)
in a later PR.
Frontend-only: new static/capabilities/working-tuning.js, loaded from
static/index.html + static/v3/index.html. Per-instrument state machine verified
by a stubbed node harness (separate guitar/bass slots, selector switch, isolated
writes, verified stamp, reset, defensive copies, capability registration).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* fix(working-tuning): resolve review P1/P2s + add behavioral test harness
Addresses the manual + Codex review of PR 1 (working-tuning foundation).
P1 — named tunings were dropped by the boot seed: /api/settings.tuning may be a
name ("Drop D") OR an offsets list, but the seed only handled the list and stored
offsets:null for names. The seed now resolves a name to per-string semitone offsets
via /api/tunings (ratio vs Standard; reference pitch cancels).
P1 — async seed could clobber state a consumer had already written: _seedFromSettings
resolves after boot and used to overwrite _currentKey/_byInstrument unconditionally.
It now bails when state was already _touched (and re-checks after the /api/tunings
leg), so an explicit set()/setCurrentInstrument()/resetToDefault() before hydration
wins. Hydration still fires.
P1/P2 — shallow copy leaked live nested arrays: get() and set() now clone offsets and
verifiedStrings on both ingress and egress, honouring the "readers can't mutate live
state" contract.
P2 — provenance/verification state machine made coherent by construction:
verified <=> verifiedStrings is an array AND verifiedAt is a finite number. A tuning
change invalidates prior verification unless a fresh bundle is supplied; a "verified"
claim with no strings or a null/absent timestamp is repaired (assumed / stamped now).
P2 — bare-instrument writes targeted a hard-coded default string count: _keyOfResolved()
resolves an omitted string count against the current selection (same instrument), so
set({instrument:'bass'}) / set({stringCount:5}) hit the selected bass-5, not bass-4.
Test — adds tests/js/working_tuning.test.js (the harness the PR described but did not
commit): 11 behavioral cases over a stubbed window — registration, per-instrument
isolation + selector switch, defensive copies, the verification invariant, bare-key
routing, named + offsets-list seeding, and the boot-race guard. Full tests/js suite:
no new failures (the 12 pre-existing branch failures are unrelated).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
The Target dropdown keyed panes off feedBackSplitscreen.panelIndexFor,
which can return the focused index for any canvas — so both split panes'
keys ping-ponged, rebuilding the <select> every frame (flicker) and
listing wrong/duplicate entries. The registry also never dropped panes
from a prior song or a closed split.
- Key each pane by a stable per-renderer-instance id (_paneUid, assigned
once in init) instead of the split panel index.
- Prune panes not reported within ~1.5s (song change / split teardown).
- Mark the dropdown dirty only when the pane SET changes, not on every
per-frame re-report, and skip rebuilding while the <select> is focused.
- Hide the Target row entirely when there's a single pane.
- Label panes by arrangement name, falling back to "Pane N".
Per-pane overrides are now session-only (keyed by ephemeral instance ids),
so they're no longer persisted to localStorage.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
Two usability gaps in the wide-pane framing tuner:
- No way to dismiss the panel. Add a × close button to the header and make
the Shift+A shortcut open/close the panel (reveal/dismiss). The A/B
enabled toggle now lives as a checkbox in the panel, so closing the panel
no longer changes the framing state.
- Edits hit every split pane at once. Add a Target selector (All panes, or a
specific pane labelled by its arrangement, e.g. "Panel 1 — Rhythm"). Per-
pane edits write a sparse override map (__panels[key]); each renderer
resolves the shared base with its own pane's overrides laid on top via
_resolveTuneFor(paneKey), so one pane can be framed independently. Reset on
a pane clears its override (re-inherits the base); Copy exports the resolved
values for the selected target. The live readout is keyed per pane.
Panes are discovered from the existing per-panel key (_bgPanelKey /
feedBackSplitscreen.panelIndexFor) and self-register each frame for the
picker. Overrides persist to localStorage alongside the base.
Tests extended in tests/js/highway_3d_wide_fov.test.js.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
The v2 control bar exposes a "Min res" selector next to Quality that sets
highway.setMinRenderScale — capping how far the load-adaptive resolution
scaler (feedBack#654) may downscale, or disabling it entirely (Full). The
v3 UI only ported the Quality selector, so v3 users had no way to stop the
highway auto-downscaling to as low as quarter-res on heavy scenes / weak-
GPU launches — pixelated even at Quality = HD, with no workaround (worse
than v2).
Add the Min res row to the v3 viz/quality rail popover, under Quality,
mirroring the v2 control (same options, handler, title, aria). The handler,
the setMinRenderScale/getMinRenderScale API, and the shared app.js init
that syncs the selector's value (guarded by element id) all already exist —
only the v3 markup was missing.
Fixes#662
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* Add aspect-aware framing for ultra-wide highway panes
On a top/bottom 2-player split each 3D highway pane is full-width /
half-height (~32:9). The camera's vertical FOV was locked at a single
value, so at that aspect the horizontal cone ballooned past 130deg and
squeezed the fixed-width neck into a thin central sliver with large dead
margins on either side.
Add a "horizontal-FOV-hold" path: past a configurable start aspect the
effective vertical FOV is lowered so the horizontal cone stays roughly
constant, letting the neck fill a wide pane. At/under the start aspect it
is an exact no-op, so normal ~16:9 single-player and most 2x2 panes are
unchanged. Optional pose nudges (height / dolly / pitch / look-depth)
further flatten the view toward a low, immersive angle.
Everything is driven by a runtime bridge (window.__h3dAspectTune) with a
live tuner panel (Shift+A in the player) exposing every knob plus a live
aspect/FOV readout, localStorage persistence, and a Copy button. Toggling
the feature off restores the exact prior framing, so it doubles as an A/B
control. Shipped on by default for wide panes for testing feedback.
Source-pinned by tests/js/highway_3d_wide_fov.test.js.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: topkoa <topkoa@gmail.com>
* fix(highway_3d): ship wide-pane framing default-OFF with a coherent config
Review fixes for the aspect-aware framing. The first cut shipped
_ASPECT_DEFAULTS = { enabled:true, baseVfov:30, blend:0, minVfovDeg:36 },
which contradicted the PR's own "default off → byte-for-byte prior behaviour"
claim:
- enabled:true made the tune active for everyone, and baseVfov:30 forced every
pane's vertical fov from 70° to 30° (normal single-player/2x2 panes included —
a drastic global zoom, not the advertised no-op).
- blend:0 collapsed the Hor+ math back to base, so the actual horizontal-FOV-
hold did nothing even on wide panes — the only net effect was the zoom.
- minVfovDeg:36 > baseVfov:30 was an inverted floor (clamped wide panes UP to
36° rather than flooring a real reduction).
New defaults: { enabled:false, baseVfov:BASE_VFOV(70), blend:1,
minVfovDeg:HORPLUS_MIN_VFOV(28) }. Now:
- OFF by default → camUpdate passes a null tune → effectiveVfov returns
BASE_VFOV → exact no-op on every pane (verified: 70° at 16:9 and 32:9).
- When a tester enables it (Shift+A), baseVfov==BASE_VFOV keeps normal/≤start
panes at 70° (still a no-op there) and blend:1 makes the hold actually engage
on genuinely wide panes (47.7° at 32:9, flooring toward 28° as aspect grows).
- minVfovDeg < baseVfov is a real floor.
Also bumps the localStorage key (h3d_aspect_tune → h3d_aspect_tune2) so a
machine that persisted the old broken default gets the corrected one, and adds
source-pin tests guarding default-off + the coherent base/blend/floor so this
can't silently regress to default-on again. The pose-nudge values are left as
the author's in-progress wide-pane look (dormant until enabled). 110/110 tests
pass; node --check clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Signed-off-by: topkoa <topkoa@gmail.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
The persistent top-right "Up Next" pill showed the upcoming section name
and a countdown ("in 12.3s") but no at-a-glance sense of how far through
the current section the song is. Add a thin progress bar directly under
the existing text that fills as the current section elapses toward the
next, reaching full when the section flips.
The text row is wrapped unchanged in a flex row and the pill stacks the
bar beneath it; nothing else about the pill's content or styling changes.
Progress is computed in updateUpNext() as the fraction elapsed between the
previous section boundary (last section at/before now, else song start)
and the next section. The fill uses the same gradient as the section name
for visual cohesion.
Signed-off-by: topkoa <topkoa@gmail.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
The v3 home tuner card runs continuously, and every tuner visualization
drove a self-rescheduling 60fps requestAnimationFrame loop that never
stopped — pinning a renderer core even on a silent home screen with the
needle/strobe at rest.
Make each viz idle its loop once there's nothing left to animate, and
re-kick it from update() on the next reading that actually moves it:
- analogue-gauge: stop when the needle + drum strip have settled on their
targets (|target-current| below a sub-visible epsilon); restart when a
new reading moves the target.
- strobe / mace-fx-iii / chef-mt3: stop when there's no live signal and the
strobe drift (and glow fade) have fully decayed; restart on the next note.
- toilet-tuner: stop when silent and the plunger has eased back to centre;
restart on the next reading (guarded so repeated no-signal updates don't
re-kick a parked loop).
Active tuning is unchanged — the loop runs whenever a note is sounding or
the indicator is still moving. Bumps tuner 1.3.1 -> 1.3.2.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs: host theme contract proposal (prevent features carving into one theme)
Charrette output after a plugin UI feature (note_detect results card) was built
against only the default skin and broke on the others — the colours adapted via
tokens but the visual *devices* (glow ring, gradient) did not, because themes are
design languages, not palettes, and nothing governs whether a theme does glow.
Proposes a host theme contract: always-present semantic role tokens (incl. the
missing on-accent + focus-ring), intent-named capability recipe slots where "off"
is legal (an EMPHASIS recipe + an ACCENT-TEXT recipe), a window.feedBack.theme
read/capability API + theme:changed event, a derive-surfaces-from-host
reconciliation rule, accessibility baked in, and a skin-matrix verification gate.
All additive + feature-detected.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* docs(host-theme-contract): make the proposal normative (review fixes)
Address the review of #645 (manual + Codex) — the doc was a sound design
sketch but not yet the precise *contract* it claims to be:
- P1 Token contract pinned: one public namespace `--fb-*` written on :root by a
host-owned contract stylesheet (present themed-or-not); `--fbv-*` explicitly
demoted to internal Tailwind-override plumbing (plugins must not read it).
Added a normative role table + value grammar (colour roles = `r g b` triplets
consumed via `rgb(var(--fb-x))`; recipe slots = full CSS device values). The
§5 example now uses `--fb-*` throughout (was bare `--accent`/`--emph-*`).
- P1 Invisible-text bug removed from the spec: `--fb-acc-text-fill` is the one
slot where `none` is ILLEGAL (always a real paint, defaulting to the solid
accent); the example feature-detects `background-clip: text` and keeps a solid
`color` base, so the accuracy number can never render transparent — honouring
the DoD "a device stays legible when its slot resolves to none".
- P1 capabilities() booleans removed: they contradicted "never branch on
glowy?" and were too lossy for canvas. The JS API is now CSS/DOM-forbidden and
exposes RESOLVED token values (`get().tokens`) for canvas/WebGL renderers only.
- P1 Physical home decided: a static `theme-contract.css` (outside the prebuilt
Tailwind artifact, so no tailwind-fresh CI churn) holds the :root `--fb-*`
defaults + the single focus-visible + reduced-motion rules; existing v3.css
focus/motion rules are a tracked reconciliation, not day-one magic.
- P2 Full on-fill family (`--fb-on-accent/-good/-warn/-bad`) + good/warn/bad ↔
existing good/mid/low mapping; `theme:changed` lifecycle pinned (get() sync +
valid pre-apply, event after commit + once on hydration, plugins read on
mount); same-document-light-DOM scope + shadow/iframe bridge stated;
prefersReducedMotion() named the single JS motion gate.
Resolved open questions folded into the body; the two genuine ones (skins-as-
host-themes, component-recipe bundles) remain.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
The heat-coloured fret-number row is drawn as a band BELOW the board
(sY(lowest) - S_GAP*1.4), but camUpdate's self-correcting framing only
anchors the board CENTRE to the lower third of the screen and reserves no
headroom for that row. So a tight zoom on a centred active span (worst
mid-neck; fine at either end of the neck) pushes the numbers past the
bottom edge -- which is why testers saw it "only when centered" and "not
every song." Tilt can't fix it (it would only trade a bottom clip for a
top clip); the vertical-extent problem at tight zoom needs camera distance.
Add a fret-row fit guard: project the row band with the final camera and,
when it falls below FRET_ROW_FIT_NDC_MIN, raise a capped, hysteretic
_fretRowFitBoost applied to the curDist lerp target (the span-driven
tgtDist still owns zooming IN). The boost rises promptly (proportional to
the deficit), relaxes lazily past a deadband, and is capped at
FRET_ROW_FIT_BOOST_MAX (+60%) so the zoom can't pop or hunt. It cooperates
with the tilt loop (pull-back shrinks the scene, tilt keeps the centre
anchored) and yields entirely to the Camera Director free-cam. Surgical:
passages where the row is already visible never trigger it, so framing is
unchanged everywhere it already worked.
plugin.json 3.30.0 -> 3.30.2 (screen.js cache-buster; 3.30.1 is taken by the
FPS-counter PR). Tests: tests/js/highway_3d_camera_framing.test.js
(guard constants, the boosted curDist lerp, the projected-row hysteresis,
free-cam yield).
Fixes#632
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(v3): dismiss Section Practice popover when another player popover opens
The Section Practice popover (Songs > Song > Practice pill) stayed open
when the user then clicked a v3 player-rail icon (Plugins, Audio, …),
leaving two popovers stacked on top of each other. Reported on 0.3.0
(macOS) and still reproducing in the 2026-06-28 build.
Root cause: the popover's outside-click dismiss was bound in the
bubbling phase, but the v3 rail's icon buttons call e.stopPropagation()
in their click handler (player-chrome.js wireRail), which kills bubbling
before the click reaches document. So the dismiss listener never fired
for a rail-icon click and the popover was orphaned open.
Fix: bind the outside-click dismiss in the capture phase, which runs
before the target's handler so stopPropagation() can't swallow it. This
mirrors the audio mixer popover (audio-mixer.js), which already
dismisses outside-clicks via capture-phase listeners for exactly this
reason. Esc handling stays in the bubble phase (no rail handler stops
keydown propagation, and capturing it would reorder it ahead of the
player's Escape-to-exit handling).
Shared app.js code, so v2 is covered too; v2 has no stopPropagation rail,
so its outside-click dismiss behaviour is unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* chore(#638): add CHANGELOG entry + capture-phase regression test
Review follow-ups for the Section Practice popover dismiss fix:
- CHANGELOG [Unreleased] → Fixed entry (repo workflow requires one).
- tests/js/section_practice_dismiss.test.js pins the fix: the outside-click
dismiss binds in the CAPTURE phase (so a rail icon's stopPropagation can't
swallow it), exactly one capture binding (Escape keydown stays bubble-phase),
and the #section-practice-control containment guard (no self-close). A revert
to bubble-phase fails the test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* fix(v3): reject accidental text-selection of UI chrome (user-select policy)
Dragging/double-clicking across the v3 UI marquee-highlighted buttons, labels,
the sidebar, transport, and the note-highway HUD — looks broken (reported Mac +
Windows). Default the v3 shell to user-select:none on html, then opt CONTENT
back in. Decided by a 4-lens panel (UX / a11y / dev-ops / plugin-ecosystem);
their guardrails are baked in:
- Form fields ALWAYS re-enabled (input/textarea/select/[contenteditable]) so the
caret + IME composition never break. No `* { user-select:none }` (WebKit input
bug 82692).
- Plugin screens (.screen[id^="plugin-"]) stay selectable BY INHERITANCE (no `*`,
so a plugin's own non-select chrome still wins) — a plugin's copyable text
(lyrics, chords, results), including community/out-of-tree plugins that never
adopt the class, isn't silently locked.
- Core read-only content opts back in by CONTAINER via a hand-authored
`.fb-selectable` (not a Tailwind utility — so runtime-installed plugins get it
too): the whole Settings panel (paths, device names, version, diagnostics,
About) and the now-playing song metadata. Answers the open "keep settings
copyable?" question: yes, at the container.
Cosmetic only — never used to lock copy-worthy text (errors/IDs/paths/versions/
metadata stay selectable; WCAG 2.2 allows copy-paste as a mechanism). v3-only
(v2 unchanged; v3.css loads only on /v3); plain CSS, no Tailwind rebuild; no
desktop/Electron changes (standard OS-framed window). `.fb-selectable` is
documented in CLAUDE.md for plugin authors.
Tests: tests/js/v3_user_select_policy.test.js (html default, form-field
re-enable, plugin-screen carve without `*`, .fb-selectable, container opt-ins,
and the no-`*`-rule guardrail).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* fix(v3): address review of the user-select policy (#637)
Review (manual + Codex) of the v3 text-selection policy:
- P1 (real bug): the now-playing HUD metadata opted into `.fb-selectable` but
its `#player-hud` parent is `pointer-events: none`, so the mouse could never
reach the text to select it — the opt-in was inert. Add `pointer-events-auto`
to the metadata block (verified in-browser: user-select:text + pointer-
events:auto, while the HUD parent stays pointer-events:none).
- Coverage: the PR's a11y guardrail promised copyable text stays selectable
"incl. in modals/toasts", but only Settings + the HUD were opted in. Blanket-
opt the focused copyable surfaces back in by selector — `.feedBack-modal`,
`[role="dialog"]`, `#fb-notify-stack`, `#v3-fb-toast`, `#scan-banner` — so
errors / IDs / paths / file names in dialogs, toasts, and the scan banner stay
copyable. These are focused panels, not dense card lists, so re-enabling
selection there can't recreate the across-cards marquee mess.
(Deliberately NOT opting in the library grid / dashboard / profile card lists:
making dense card text selectable would reintroduce exactly that marquee mess
on a drag — copy song metadata from the now-playing HUD / Settings instead.)
- Test (P3): assert the selectable rule's selectors order-independently, cover
the new modal/toast/banner surfaces, and check the HUD block carries BOTH
fb-selectable and pointer-events-auto (class-order independent).
Verified in a real browser (chromium): html=none, sidebar chrome=none, input=
text, Settings=text, HUD meta=text+pointer-events:auto, dialog/modal=text.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
The v3 Songs grid appended every scrolled page and never released nodes,
so card-node count grew unbounded with scroll depth (24 → 624 → 2001 for a
2000-song library). Replace it with a windowed/recycled render: only the
visible window (± overscan) is in the DOM while a #v3-songs-gridsizer
element sized to ceil(total/cols)*rowH gives the scrollbar full-library
geometry; #v3-songs-grid is absolutely positioned to the first visible row.
- state.songs is a sparse, absolutely-indexed store filled a page at a time
by ensureWindow(): the stage-1 keyset cursor for contiguous forward scroll
(O(page)), OFFSET page= for jumps/restore/non-keyset providers. _loadPage
shares an in-flight promise per page and an epoch guard discards a stale
fetch that lands after a reset.
- A–Z rail seeks directly via sort_letters cumulative counts (O(1), no
page-through); bounded scan fallback for legacy providers without it.
- Snapshot/restore is now scrollTop-based (geometry is stable). Select mode,
accuracy badges, ⋮ menu, plugin card actions, and tree/folder coexistence
survive cards recycling; renderWindow re-renders when select mode toggles.
- Plugins get window.v3Songs.visibleCards() + a v3:library-window-rendered
event instead of assuming all cards are present (highway-stutter lesson).
Verified in a browser against a seeded 2001-song library: DOM bounded to
~60 nodes while the count reads "2001 songs", rail jump lands on the target
row, selection survives recycling, scroll-restore exact. Codex-reviewed
(3 findings fixed: stale-fetch epoch guard, await-in-flight page promise,
select-mode resync on cached re-entry).
Frontend-only. Tests: tests/browser/v3-grid-virtualization.spec.ts pins the
bounded-DOM invariant + direct rail jump; tests/js/v3_az_rail.test.js and
v3_songs_scroll.test.js updated to the new wiring.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Implements feedBack#636 item 2 (P1) — saved library filters that stay live,
the homelab primitive FeedBack was missing (Plex smart collections / Navidrome
.nsp / *arr custom filters).
A collection is a saved /api/library query surfaced as a registered library
provider, so it appears in the v3 source picker and inherits the whole Songs
UI (paging, stats, A–Z rail, art) with no new screen.
- Storage reuses the playlist subsystem: a `playlists.rules` JSON column
(additive, idempotent migration). A row with rules != NULL is a smart
collection; list_playlists + get_playlist filter `rules IS NULL`, so
collections are excluded from the manual-playlist list and read-only to
every playlist mutation that gates on get_playlist.
- SmartCollectionProvider (kind="local" — matched songs are local rows, so the
client's play/art paths stay on the local branch) delegates query_page/
query_stats/query_artists to the local DB with the stored rules applied;
tuning_names/get_art delegate straight through. Registered via a boot scan +
on create/update (replace=True) / delete.
- Rules mirror the raw /api/library query params; `_sanitize_collection_rules`
drops unknown keys and is applied at API ingress AND on provider load, so a
hand-edited / imported bad value can't crash a query.
- API: GET/POST/PUT/DELETE /api/collections. Frontend: a "+ Save as
collection" action in the v3 filter drawer (local provider + active filters
only) that names the current filter set and switches to it.
Reviewed by Codex; 3 findings fixed (local-kind playback path, save gated to
local provider, re-sanitize persisted rules).
Tests: tests/test_collections_api.py (CRUD, provider filtering, restart
re-registration, kind=local, corrupt-rule tolerance, playlist isolation),
tests/js/v3_collections.test.js.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Stage 1 of the virtualized-grid project (got-feedback/feedBack#636 item 3):
the data layer the DOM-recycling render window will build on, plus a latent
paging bug fixed on the way.
- Every grid sort now appends a unique `filename` tiebreak → a TOTAL order.
Without it, rows with an equal sort key (e.g. two songs by the same artist)
could be skipped or duplicated across OFFSET pages.
- query_page gains an opaque `after` keyset cursor: when supplied and the sort
can keyset (artist[-desc], title[-desc], recent), the page is fetched with a
WHERE-seek instead of OFFSET — O(page), independent of depth. The seek is
NULL-aware (NULLs first in ASC / last in DESC) so it's EXACTLY OFFSET-
equivalent; the legacy `dir=desc` shape is canonicalized so its cursor seeks
the right direction. Unknown/compound sorts + bad cursors fall back to OFFSET.
- /api/library exposes `after` + `next_cursor`. Only the true local provider is
handed a cursor (a collection may pin a different sort; remote don't keyset),
so both page by OFFSET safely.
- Composite (artist NOCASE, filename) / (title NOCASE, filename) /
(mtime, filename) indexes cover the order; `after` added to the optional
provider kwargs so legacy providers drop it.
Codex-reviewed; 3 findings fixed (dir=desc canonicalization, NULL-key seek,
cursor only for the local provider).
Tests: tests/test_library_keyset.py (keyset==OFFSET parity for 5 sorts, stable
tiebreak on equal keys, dir=desc, NULL sort keys, bad-cursor + compound-sort
fallback).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Closes feedBack#636 item 4 — the *arr "Wanted/Monitored" analogue FeedBack
was missing. A wishlist entry is a song the user does NOT own yet, so unlike
a playlist (which references owned local songs by filename) it can't reuse the
playlist subsystem; it lives in a new `wanted` table keyed by descriptive
identity (artist, title, source, source_ref, note, created_at).
- New table + a UNIQUE index on (artist NOCASE, title NOCASE, source,
source_ref); additive + idempotent (CREATE … IF NOT EXISTS).
- MetadataDB.add_wanted (INSERT OR IGNORE + re-select under the write lock,
so a re-run of an ownership-diff returns the existing row, never a dup),
list_wanted (newest first), remove_wanted, count_wanted.
- Routes GET/POST/DELETE /api/wanted. POST requires artist or title and
defaults source to "manual"; idempotent on identity so producers (the
find_more ownership-diff, or a manual add) can re-post freely.
This is the core persistence primitive the charrette flagged as the missing
piece; the consuming UI lives in the producing plugin (find_more / the_daily).
Tests: tests/test_wanted_api.py (round-trip, identity idempotency incl.
case-insensitive, distinct source_ref, ordering, validation, additive schema).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Closes the dev-ops lens's #1 finding from the library charrette
(got-feedback/feedBack#636 item 1): scores, favorites, playlists, and play
history — the only library state a rescan can't rebuild — were absent from
the settings backup. Now GET /api/settings/export carries an additive
`core_server_files` section:
- a CONSISTENT snapshot of web_library.db via the SQLite online-backup API
(a complete single file even while the server runs; taken under the
MetadataDB write lock), base64-encoded;
- custom playlist covers + avatar (CONFIG_DIR/playlist_covers, /avatars),
walked with the existing _walk_export_paths machinery.
Restore is DB-safe:
- POST /api/settings/import STAGES the DB to web_library.db.restore (never
over the live, open file); _apply_pending_db_restore swaps it in at the
next startup BEFORE the connection opens, clearing stale -wal/-shm so a
stale WAL can't be replayed onto the restored file. Response sets
`restart_required` + a warning; custom art applies immediately.
- The staged DB is integrity-checked (open + PRAGMA quick_check) at import
AND again at startup before the live DB is touched — a corrupt/truncated
restore is refused/discarded and the live DB is left intact, so a bad
bundle can never brick startup or lose data.
- Export hard-fails (500) if the snapshot can't be produced (no silent
DB-less backup); a partial import disarms its own staged restore.
Backward-compatible: older servers ignore the new section; a bundle without
it imports as before. Known gap: custom uploaded *song* art is still
commingled with the rebuildable thumbnail cache in art_cache/, so it isn't
bundled yet (tracked follow-up on #636).
Tests: tests/test_settings_export_library_db.py (snapshot consistency,
staged-not-live restore, sidecar clearing, corrupt-DB refusal at import +
startard, traversal rejection, export hard-fail, disarm-on-failure, full
round-trip).
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(v3 library): practice-aware home — Repertoire meter + "Keep practicing" shelf
The Songs page opened cold into a flat sorted grid. This adds a practice-aware
front door on the unfiltered grid, built entirely from data already on hand
(no new endpoints, no new stored state):
- Repertoire meter — "Repertoire: N of M songs · K in progress" + a bar,
counting songs at/above the same mastery threshold the green accuracy badge
uses (>= 0.9 best accuracy) over the unfiltered library total. Reads
state.accuracy (/api/stats/best, already loaded for the card badges) and the
unfiltered /api/library/stats total.
- "Keep practicing" shelf — a horizontal row of recently-played, not-yet-
mastered songs (newest first, click to play). Reads /api/stats/recent.
Both show ONLY on the grid view when not searching/filtering/selecting (the
front-door context), refresh after a song is scored (applyScoreRefresh), and
collapse on an empty library. Soft-gamification only: descriptive encouragement
(goal-gradient / endowed-progress), never content-gating, decay, or nagging —
the practice-accuracy "continue" rail a media server can't do.
Frontend-only: static/v3/songs.js (renderLibraryHome / _repertoireCounts /
libHomeVisible, wired through reload() + applyScoreRefresh), static/v3/v3.css.
Came out of the library design charrette (UX + gamification lenses' top pick).
Stacked on the A–Z rail branch (feat/v3-library-az-rail) since both touch
static/v3/songs.js; merge that PR first (or retarget).
Tests: tests/js/v3_keep_practicing.test.js (threshold, front-door gating,
shelf filter, denominator, render/reload/score-refresh wiring, click-to-play).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* fix(v3 library): correct practice-aware home for review P1/P2/P3
Addresses the PR #635 review findings (manual + Codex):
P1 correctness
- Gate the Repertoire meter + "Keep practicing" shelf to the LOCAL
provider (libHomeVisible). They read local practice stats
(state.accuracy / /api/stats/recent); on a remote provider they mixed a
local mastered count with a remote song total (e.g. "85 of 80") and the
shelf played local files while browsing a remote library.
- Shelf now gates on the per-SONG best (state.accuracy[filename] = MAX
across arrangements, what the green badge shows) and dedupes by filename,
instead of the per-arrangement recents row — so a "keep practicing" card
can no longer show a green "mastered" badge, and a song can't appear twice.
P2 robustness
- renderLibraryHome fetches /api/library/stats + /api/stats/recent together
(Promise.all) and a _homeToken generation guard discards a stale render
so a slow response can't repaint a home the grid already moved past.
P3 polish
- accuracyBadge references MASTERY_ACCURACY instead of a bare 0.9, so the
badge and the meter/shelf can't drift from "the same mastery threshold".
Tests updated (v3_keep_practicing.test.js): provider gating, per-song
deduped shelf, Promise.all + token.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
* feat(v3 library): A–Z fast-scroll jump rail on the Songs grid
Adds a vertical letter rail (Plex/Radarr/iOS-contacts pattern) pinned to the
right edge next to the scrollbar so you can jump the library to a starting
letter — tap, drag-to-scrub with a live letter bubble, or arrow-key between
letters. The classic (v2) tree already had letter selection; this brings the
new v3 grid to parity (it was the gap behind the "alphabetical scroll
selection next to the scrollbar" idea).
It shows ONLY for the grid view + alphabetical (artist/title) sorts, and only
offers letters present in the current sort AND filter set, so a tap always
lands on a real card (absent letters are dimmed + non-interactive). The grid
is forward-only, server-paged infinite scroll with no virtualization, so a
jump pages through to the target card then scrolls to it; a token guards
overlapping jumps (drag) so the newest wins. A keyset-seek + virtualized
window is the scaling follow-up for very large libraries.
Backend: /api/library/stats gains an optional `sort` param and an additive
`sort_letters` map — songs-per-first-letter of the ACTIVE sort column (artist
or title), filter-synced — so the rail's present-letters match the grid's real
order. The legacy `letters` (distinct-artist) field is unchanged, so the
dashboard + classic tree are unaffected. `sort` is dropped for providers whose
query_stats predates it (existing kwarg-filter), so third-party library
providers keep working (rail simply falls back / hides).
Frontend: static/v3/songs.js (refreshRail / jumpToLetter / pointer-drag +
keyboard, cards tagged data-letter), static/v3/v3.css (.v3-azrail + bubble).
Tests: tests/test_library_filters.py (sort_letters artist/title, song-vs-
distinct-artist counting), tests/test_library_providers.py (sort forwarded),
tests/js/v3_az_rail.test.js (gating, data-letter, load-through, drag/keys).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
* fix(v3 library): harden A–Z jump rail (review P2/P3)
Addresses the PR #634 review findings (manual + Codex):
P2 correctness
- refreshRail prefers the active-sort `sort_letters`; falls back to the
artist-based `letters` only on an artist sort, and hides the rail on a
title sort when a legacy provider returns none (was mislabeling letters).
- reload() bumps `_jumpToken` so an in-flight letter jump can't scroll a
grid that's being rebuilt from page 0.
- songBucket no longer trims, matching the server SQL + grid ORDER BY raw
first-char bucketing (a leading-space title now buckets under '#' on both
sides).
P3 polish
- Paging guard is total-derived (ceil(total/PAGE_SIZE)+2) instead of a
magic 4000, keeping large libraries reachable while still bounded.
- Roving tabindex: only the first present letter is tabbable; arrow keys
move it. Removes up to 27 page tab stops.
- `sort_letters` is computed only when the caller opts in
(want_sort_letters / route `sort_letters=1`); the dashboard + v2 tree
skip the extra GROUP BY. Added sort + want_sort_letters to the optional
provider-kwargs so non-introspectable legacy providers drop them.
- _railToken supersedes stale refreshRail responses; hide the rail when no
letters are present instead of rendering disabled buttons.
Tests updated accordingly (v3_az_rail.test.js, test_library_filters.py).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: byrongamatos <xasiklas@gmail.com>
The on-highway FPS readout (Settings -> Graphics -> 3D Highway -> Show FPS
counter) is pinned to the top-right of the highway overlay -- the same
corner the v3 player chrome stacks its persistent Up Next pill and
live-performance HUD into, on a higher layer that paints over the canvas.
So the readout sat behind that chrome and couldn't be read, exactly when a
tester turned it on to judge performance (and because the pill is default-on
it covered the counter regardless of the separate "Up Next won't turn off"
report).
Keep it top-right (where testers look) but drop it just below whichever of
that chrome is showing: measure the lowest visible top-right v3 HUD element
(#v3-upnext / #v3-live-performance-hud / #hud-time) and floor the FPS box's
Y beneath it. Element refs are resolved once and cached (no per-frame
querySelector, per the plugin perf rules) and only read while the counter is
actually drawn; gated on window.feedBack.uiVersion === 'v3' so classic v2 is
byte-for-byte unaffected. Bump plugin version 3.30.0 -> 3.30.1 (the screen.js
cache-buster).
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Second half of feedBack-desktop#46. The desktop app monitors through an
in-app amp-sim/tone chain that, once loaded, auto-restores every launch —
an idle high-gain amp on the input is a constant distorted buzz, and the
dry-only monitor mute can't silence it. This adds the "own-rig first"
opt-in so players using their own external amp/rig never get a processed
monitor in the first place.
Core changes:
- New `use_amp_sims` setting (default OFF / own-rig first): GET default,
POST boolean validation, and resettable key — mirroring achievements_enabled.
- Onboarding wizard: a DESKTOP-ONLY step ("How do you want to hear
yourself?") between instrument paths and the calibration challenge. The
web build has no native amp sims, so the step is skipped there (5 steps
on web, 6 on desktop) — gated on window.feedBackDesktop, dot count and
setStep bounds are derived from it. Ticking "Use in-app amp simulations"
POSTs use_amp_sims; default unticked.
The desktop renderer consumes this setting to gate its saved-tone-chain
restore (feedback-desktop PR, paired).
Verified by booting core locally and walking the wizard with Playwright:
web shows 5 dots/no amp step, desktop shows 6 dots, the amp step is
reachable, calibration stays the final "Play it now" step, ticking the box
persists use_amp_sims=true, and there are no page errors. Server-side
GET default / POST validation / reset confirmed via curl.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Surface the feedpak manifest `authors` list (spec §5.4) on the highway: a credits card ("Charted by Azure") shown over the highway when a song loads, riding the count-in / a ~3s hold and dismissed when playback starts. Gated to fresh feedpak plays only (minigames, loose/archive, arrangement switches, seeks, replays excluded). Includes a 12s backstop so the overlay never lingers if playback fails to start.
Closes#628. Reviewed by Codex (3 passes, converged). Verified locally: pytest 9/9, node --test 23/23, headless-browser end-to-end.
The onboarding audio picker de-duped the device list by display LABEL. On
Windows the engine enumerates one interface once per host API (ASIO /
Windows Audio / DirectSound) with the same name, so the variants collapsed
to a single choice — silently keeping whichever sorted first, often not the
low-latency ASIO one the player wants. It could also drop the variant that
was actually `selected`.
The audio-input capability already collapses true duplicates by
logicalSourceKey (_visibleInputSources), and these variants each have a
DISTINCT key, so the wizard's extra label-collapse was redundant for real
dupes and destructive for the variants. Removed it; the picker now lists
every selectable input.
Pairs with feedBack-desktop's change to tag each source label with its
driver type ("Focusrite (ASIO)" vs "(Windows Audio)") so the now-distinct
entries are legible.
Claude-Session: https://claude.ai/code/session_01QbexxfTt8q2tAn436MqGWF
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
feedpak-spec §8 (FEP #3 / PR #13) allows .jsonc files (JSON with C-style
// line and /* */ block comments) anywhere .json is specified. A Reader MUST
strip comments before parsing. Core's sloppak/feedpak readers parsed every
side file with bare json.loads, so a .jsonc arrangement / notation / drum_tab
/ song_timeline / lyrics / keys would fail to load.
- lib/jsonc.py (new): shared parse_jsonc(text) + load_json(path). String-aware
regex (mirrors the spec reference validator in feedpak-spec/tools/validate.py)
— keeps comment-like text inside JSON string literals. load_json auto-detects
.jsonc by suffix; plain .json goes straight through json.loads.
- lib/sloppak.py: import load_json; replace the 6 json.loads(...read_text...)
side-file read sites (arrangement, notation, drum_tab, song_timeline,
lyrics, keys) with load_json(<path>). Removed the now-unused `import json`.
- scripts/lift_keys_notation.py: import load_json; replace the 3 read sites
(song_timeline, arrangement beats fallback, arrangement lift). `import json`
stays (json.dumps write at the notation sidecar emit).
Additive (MINOR) change: older readers parse .jsonc as plain JSON and ignore
comments via the spec's forward-compatibility rules, so no existing pack needs
regeneration.
Tests: tests/test_sloppak_jsonc_load.py (16 tests) — parse_jsonc unit cases
(line/block/multiline/string-boundary/malformed/plain), and end-to-end loads
for all 6 side-file types via .jsonc with comments, plus the lift helper
reading .jsonc song_timeline + .jsonc arrangement beats, plus the
string-boundary preservation rule through the full loader. 122 sloppak/lift
tests pass.
2026-06-20 14:10:04 -07:00
460 changed files with 110078 additions and 38580 deletions
<!-- What does this PR do, and why? Link the issue it addresses. -->
## feedpak surface
<!-- The feedpak spec is sacrosanct: the spec defines the format, this app implements it.
Delete this section ONLY if your change doesn't touch how the app reads or writes packs. -->
- [ ] This PR does **not** change how the app reads/writes feedpaks (manifest keys, pack files, folder layout)
- [ ] …or it does, and the spec change landed first via the [FEP process](https://github.com/got-feedback/feedpak-spec/blob/main/CONTRIBUTING.md) — FEP / spec PR: `got-feedback/feedpak-spec#___` (once it merges, re-run this PR's checks and the gate goes green)
All notable changes to this project will be documented in this file.
The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
## [Unreleased]
### Added
- **Playlists get content-dependent covers + custom art.** Playlist cards were a tiny `🎵` emoji on an empty square. Now a playlist's cover reflects its contents: **empty → the icon**, **a few songs → the first song's album art**, **4+ songs → a 2×2 art mosaic**. You can also **upload a custom cover** (a "Cover" button in the playlist detail view → image picker; "Remove cover" reverts to the content view). `MetadataDB.list_playlists()` now returns each playlist's first few song `art_urls`; `GET /api/playlists` and `GET /api/playlists/{id}` add `cover_url` when a custom cover exists. New routes `POST` / `GET` / `DELETE /api/playlists/{id}/cover` store a small PNG thumbnail under `CONFIG_DIR/playlist_covers/` (PIL-converted, like song-art upload); the cover is removed with the playlist. Frontend: `playlistCoverHtml(p)` in `static/v3/playlists.js`. Tests: `tests/test_playlists_api.py` (art_urls + cover roundtrip / reject-non-image / delete-cleanup), `tests/js/v3_playlist_cover.test.js`.
- **v3 Songs: "Add to playlist" is now on each song's ⋮ "More" menu.** Previously a song could only be added to a playlist through select-mode (the checkbox → batch bar). The per-card overflow menu now has an **Add to playlist** row that targets that one song, reusing the same picker (choose a listed number or type a new name to create it). The select-mode batch flow and the single-song menu now share one extracted `addFilenamesToPlaylist(filenames)` helper in `static/v3/songs.js` (both grid and tree rows, since they share `openCardMenu`). Tests: `tests/js/v3_add_to_playlist_menu.test.js`.
- **Resume where you left off — leaving a song now snapshots your place so an exit is recoverable, not a restart-from-zero.** Exiting the player (`showScreen()` teardown, before audio unload) writes `{song, arrangement, position, speed}` to `localStorage` (`feedBack.resumeSession`), and a non-blocking **"Resume practice"** pill offers it back on the next non-player screen (and on the next app launch). Clicking Resume re-enters the song, restores the arrangement + playback speed, and seeks to the saved position via the existing `_audioSeek` funnel; `playSong()` gains a `{ resume: {position, speed} }` option that arms a `song:ready`-consumed restore instead of the normal autostart, so the two never fight over playback. The snapshot is deliberately conservative — ignored for a song you barely started (< 3s) or had basically finished (within 5s of the end), cleared on natural song-end and once consumed, and expired after 24h. The pill is self-contained (inline-styled, body-appended, works identically in the classic and v3 shells with no Tailwind rebuild), never blocks, and a dismiss forgets the current snapshot for the session. This pairs with the Escape focus fix: now that Escape reliably leaves regardless of focus, an *accidental* exit is one tap to undo. Public surface: `window.resumeLastSession()` / `window.feedBack.resumeLastSession`. (The broader nav-state work — returning to a song after wandering into Settings → Tone Builder — is a separate, larger track; this lands the player-session slice.) Tests: `tests/browser/resume-session.spec.ts` (snapshot guards, staleness, pill show/hide/dismiss, resume consumption).
- **Optional "Ask before leaving a song" confirm (Gameplay tab, default OFF).** A new client-only toggle (`confirmExitSong` in `localStorage`, in the v3 Gameplay settings + the Gameplay "Reset" set) for players who want a guard against an accidental exit. **Off by default — Escape leaves instantly, zero change for everyone else.** When on, a *user-initiated* exit (the player-scope Escape shortcut, or the player's ✕) opens a small true-modal confirm instead of leaving; auto-exit on song-end and a results screen's own Close are unaffected (they call `closeCurrentSong()` directly, which stays the unguarded actual-exit). The confirm honors the team's refined asks: **opening it pauses the song** (so it isn't running or being scored behind the prompt) and **Stay resumes exactly what was paused**; **Escape = Stay** — the dialog's capture-phase handler *dismisses* it (now consistent with every other modal and the generic `_confirmDialog`'s Esc=cancel), so a second Escape returns you to the (resumed) song rather than leaving; **Space/Enter (or click) Leave** by natively activating the default-focused "Leave" button ("just get me out"). Pause/resume run through the canonical `togglePlay()` path (HTML5 + `_juceMode`), guarded so a count-in, an already-paused song, or a teardown/seek/end behind the modal can't mis-resume. It's a real modal (`role="dialog" aria-modal="true"` / `.feedBack-modal`) with **Tab trapped inside it** and a **backdrop click that also Stays**, so the Escape/Space focus carve-outs treat it as a trap and don't fire player-back / play-pause behind it. The player Escape shortcut and the v3 ✕ route through a shared `window.requestExitSong()` gate (the ✕ also becomes origin-aware, matching Escape). Tests: `tests/browser/exit-confirm.spec.ts` (default-off instant exit, confirm-on opens + stays, second-Escape stays, backdrop stays, Stay/Leave, Enter-leaves); the audio pause/resume is verified manually on web + desktop (the mock song has no backing track).
- **Folder Library — a bundled core plugin (`plugins/folder_library/`) that browses the DLC library by its on-disk folder tree.** Surfaces top-level folders → subfolders → songs (root-level songs land in `(Unsorted)`), with in-app folder management (create / rename / delete nested folders), song moves via dialog or drag-and-drop, and sort/filter that mirrors the host library's filter state. Wired into both the classic (v2) library toolbar and the v3 Songs page as a third **Folders** view alongside grid/tree; the plugin's `screen.js` is loaded once by the host and reused (idempotent IIFEs). Supersedes the former standalone "Folder Organizer" community plugin (removed from the README list). Backend (`routes.py`) registers `/api/plugins/folder_library/{tree,folder/create,folder/rename,folder/delete,song/move}`; **all filesystem mutations are confined to `DLC_DIR` and validated against path traversal** (per-segment name validation plus a resolved-containment check on `song/move`), and folder deletion relocates every song — de-duplicating colliding names — so a name clash never destroys a song. A two-level cache keeps re-opening folders fast. Tests: `tests/plugins/folder_library/test_routes.py` (path-safety helpers + move-traversal and delete-no-data-loss end-to-end).
- **Full-screen (immersive) plugin screens — opt-in via `"fullscreen": true` in `plugin.json`.** DAW-style plugin UIs (e.g. a practice studio) need the whole viewport, not a scrolling content page below the topbar — embedded in the v3 shell they get cut off at the bottom with dead space up top. A plugin can now declare a top-level `"fullscreen": true`; `plugins/__init__.py` surfaces it as the `fullscreen` boolean on `/api/plugins` (mirroring the `settings_category` plumbing). When such a plugin's screen is active, `static/v3/shell.js` toggles `html.fb-immersive` from `syncActive()` (so it tracks every navigation incl. deep-link), and `static/v3/v3.css` hides the topbar, collapses the sidebar to a functional **icon rail** (kept reachable — Escape is bound only on player/settings scopes, so a fully-hidden sidebar would trap the user), and lets the active plugin screen fill `#v3-main`. Mirrors the existing `ss-follower-pre` chrome-hide pattern. Additive + opt-in: plugins without the flag are unaffected. Tests: `tests/test_plugins.py::test_fullscreen_flag_parsed_from_manifest`.
- **Achievements wall sync — background drain worker (epic PR3, client side).** The bundled `achievements` plugin gains a dead-letter sync worker that POSTs queued Feat unlocks (and removals) to the hosted **feedback-achievements** wall service (separate repo). Idle unless a wall URL is configured (`FEEDBACK_ACHIEVEMENTS_WALL_URL`); uses `requests` with the baked-in client-token header, mirroring `lib/lyrics_transcribe`'s outbound pattern (explicit timeout, no raise on non-2xx). **Dead-letter, never drop** (pure `engine.drain_decision`): network error / `429` / `5xx` → keep `pending` (retry); other `4xx` → `dead_letter` (diagnosable, replayable); `2xx` → delete on server ack. A row leaves the queue only on ack or a user opt-out. `remove-me` now enqueues a wall removal keyed by the reused `player_hash`. Verified by an end-to-end staging round-trip (earn a Feat → drains onto the wall with name + short hash → `remove-me` → wall empties) with **no IP** in tables or access logs. Tests: `tests/plugins/achievements/test_sync.py` (decision table + ack/retry/dead-letter retention + four-field payload on the wire). The hosted service itself (FastAPI + SQLite-on-disk, Feats-only, hidden-until-first-global-unlock, profanity filter, in-memory rate limit, Render blueprint, migration tool) lives in the new `feedback-achievements` repo.
- **Achievements wall — opt-in, privacy controls & data-minimization gate (epic PR2).** Sharing earned **Feats** on the (forthcoming) public wall is strictly opt-in. A new **onboarding step** (`static/v3/profile.js`, inserted after song-directory / before instrument paths — the wizard is now five steps) presents a plain-language card: it publishes only your display name and the Feats you earn, never songs/skills/scores, and is **off by default**. The bundled plugin's Settings panel (`plugins/achievements/settings.html`, mounted under the **System** tab via `settings.category`) carries the same toggle plus a **"Remove me from the wall"** button (`POST /api/plugins/achievements/remove-me` — wipes local synced state offline + enqueues a wall removal). Core adds `achievements_enabled` (bool, default `false`) to `_default_settings()` + the `/api/settings` validation block + `_RESETTABLE_SETTINGS_KEYS` in `server.py`, mirrored to `localStorage` in `app.js loadSettings()`. **Data-minimization contract (binding, code-enforced):** every outbound payload is built by a single explicit-dict serializer (`engine.build_wall_payload`, never `dict(row)`/`**model`) whose key-set is **exactly**`{display_name, player_hash, achievement_id, unlocked_at}` with `achievement_id` always a **Feat** id — a unit test asserts the four-field set and goes red on a fifth. Enqueue is doubly gated: it happens only when opted-in **and** a profile identity (name + the reused `player_hash`) exists; **competency unlocks never enqueue** (integration law). Tests: `tests/plugins/achievements/test_datamin.py` (key-set, opt-out/identity/competency gating) + `tests/test_settings_api.py` (flag persists/validates/resettable).
- **Achievements & Feats of Power — local engine + tabbed Profile (epic PR1).** The Profile screen (`static/v3/profile.js`) becomes **tabbed** exactly like the v3 Settings page (`.fb-tabbar` / `.fb-tab[data-tab]` / `.fb-tabpanel[data-tab]`, active-tab persisted in `localStorage 'v3-profile-tab'`): a **Profile** (main) tab carrying the existing header + best-scores cards plus a new **Feats of Power** trophy shelf mount (`#v3-profile-feats-slot`, earned-only / hidden-until-earned), and an **Achievements** tab with a plugin mount (`#v3-profile-achievements-mount`) + `[data-empty-for]` empty note. Core dispatches a new **`v3:profile-rendered`** event after every render (mirrors `v3:settings-rendered`) so the plugin re-injects on each profile entry. A new bundled **`plugins/achievements/`** plugin owns the engine: SQLite under `<config_dir>/achievements/achievements.db` (`unlocks` / `counters` / `comp_ledger` / `sync_queue`), pure threshold/criterion math in the testable sibling `engine.py` (P-V), and routes under `/api/plugins/achievements/` (`activity`, `report-unlock`, `report-criterion`, `catalog`, `earned`, `feats`, `remove-me`). **Two surfaces, one engine, structurally separated (integration law):****Feats** (activity/volume — Note Hunter, Marathon, Untouchable, Road Warrior, Time Served, Encore, two 🥚 secrets) read activity counters only, evaluated from a batched `song:ended` activity POST (notes only when **notedetect** is present — graceful degradation, no fake progress); **competency Achievements** (baseline: First Steps / Ascendant / Steady Hands / Renaissance + per-instrument Apprentice·Journeyman·Master / Personal Best / Challenger) are evaluated from **progression events only** and never re-derived from activity. The Achievements catalogue is always shown (locked = greyed), grouped by a secondary pill row over the **real progression paths** (Global / Guitar / Bass / Drums / Keys — auto-extends to new paths) with a per-category "X / Y earned" badge, defaulting to the player's primary path. Source plugins contribute their own competency defs and report unlocks through a versioned **`window.feedBack.achievements`** API (`register`/`registerAll`/`unlock`/`progress`), load-order-safe via the `window.__feedBackAchievementsPending` queue + an `achievements:ready` event (minigames pending-queue pattern); an absent source contributes nothing (no dead greyed rows). Opt-in publishing to a hosted Feats wall, the Settings privacy toggle, and the data-minimization gate land in epic PR2/PR3. Tests: `tests/plugins/achievements/test_engine.py` + `test_routes.py` (incl. the integration-law assertion that a competency unlock never reaches the Feats shelf).
- **v3 settings page redesigned as a tabbed, card-row layout.** The single long scrolling settings screen becomes a horizontal tab bar (Gameplay / Audio / Graphics / Keybinds / Progression / Mic / Plugins / System) over card rows — each a leading icon + title + description with the control (toggle/dropdown/slider) on the right, plus a per-category "Reset" action. The markup lives in `static/v3/index.html` (so existing element ids keep hydrating through the unchanged `app.js``loadSettings()`/`persistSetting()` path); a new `static/v3/settings.js` owns tab switching + active-tab persistence (`localStorage 'v3-settings-tab'`), the per-category reset, and a read-only **Keybinds** reference built from the live shortcut registry (`window.getAllShortcuts()`); styling is plain CSS in `static/v3/v3.css` (no Tailwind rebuild). **Plugins choose their settings tab** via a new optional `settings.category` field in `plugin.json` (`plugins/__init__.py` surfaces it as `settings_category`; `app.js` mounts each plugin's `<details>` panel into `#plugin-settings-<category>`, falling back to the generic Plugins tab) — `highway_3d` ships `category: "graphics"`; the out-of-repo notedetect/progression plugins should declare `"mic"` / `"progression"`. **New gameplay settings:****Countdown before song** (a four-beat count-in before playback, wired end-to-end via the existing count-in engine + the song-start autostart path; key `countdown_before_song`, default off); **Miss penalty** (`miss_penalty`) and **Fail behavior** (`fail_behavior`) are persisted now but not yet consumed by scoring (shown with a "Not yet active" badge). "Note highway speed" surfaces the existing `master_difficulty` and stays in sync with the player-popover difficulty slider. New `POST /api/settings/reset` clears chosen keys back to defaults. Tests: `tests/test_settings_api.py` (new keys + reset), `tests/test_plugins.py::test_settings_category_parsed_from_manifest`, `tests/browser/settings-tabbed.spec.ts`.
- **Full-mix audio exposed alongside stems for the stem mixer's auto-switch.** `lib/sloppak.py::load_song` now parses the optional manifest `original_audio:` key (the single pre-separation mixdown, e.g. `original/full.ogg`) into a new `LoadedSloppak.original_audio` field, with the same path-traversal guard and permissive "missing → disabled" posture as the `drum_tab` loader. The highway WS `song_info` frame additively carries three new fields next to `stems`: `original_audio_url` (served by the existing `/api/sloppak/{filename}/file/{rel_path}` endpoint, `None` when the pack ships stems only), `has_original_audio`, and `has_stems` (mirroring the `has_drum_tab`/`has_keys` flag convention). The stems plugin consumes `original_audio_url` to play the untouched single file while every stem slider is at unity and switch to the separate stems the moment one drops below 100%. **Migration notes:** the `song_info` message shape is a stable contract — these are purely additive; all existing fields are unchanged. `audio_url` still points at stem[0] when stems exist (it is only the degraded native fallback); the one behavioural change is that a stem-less, full-mix-only sloppak now sets `audio_url` to the full mix instead of emitting `audio_error`, so it plays natively.
- **Autoplay & auto-exit — a global "click it, it plays; finish, you're back at the menu" option (default ON).** New single Settings toggle (`autoplayExit` in `localStorage`, surfaced in both the v3 and classic settings screens; absence of the key = enabled) that closes the friction at both ends of the play loop. **Autoplay:**`playSong()` previously loaded a chart paused, requiring a Play press; a one-shot flag armed per fresh load is now consumed by the next `song:ready` (highway.js) to auto-start via the existing `togglePlay()` path (HTML5 + `_juceMode` + count-in). Arrangement switches / seeks reuse the same `song:ready` event but never arm the flag, so they don't auto-restart. **Auto-exit:** on `song:ended`, core returns to the launching menu after a short grace delay — unless a visible full-screen results/dialog overlay is on top (detected via `[role=dialog][aria-modal]` / `.fixed.inset-0` with a `getClientRects()` visibility test that works for `position:fixed`), in which case the return is deferred so that score screen's own Close button (calling `window.closeCurrentSong()`) drives the exit. A plugin can also defer explicitly via the new `window.feedBack.holdAutoExit()` (called synchronously from its own `song:ended` handler — core's listener runs first). Both paths mean **no external plugin PR is required** for a results screen to be respected. **Context-aware destination:** the player's remembered origin (`_playerOriginScreen`) now honours any real launch screen instead of clamping to library/home/favorites, and a one-shot `window.feedBack.setReturnScreen(id)` override lets the lessons catalog (`static/v3/lessons.js`) send a finished lesson back to the lessons screen — not the song library — even though the external tutorials plugin owns the `playSong` call. Also exposes a read-only `window.feedBack.autoplayExit` getter for plugins. Songs and lessons share the same `playSong` → highway path, so both inherit the behaviour. Core-only (`static/app.js`, `static/v3/lessons.js`, both `index.html`s); the end-of-song score screen itself remains a plugin. Optional polish (not required — the overlay heuristic already covers it): external scoring/note-detection plugins (e.g. SlopScale) may call `holdAutoExit()` + `closeCurrentSong()` for an exact, heuristic-free handoff.
- **"Song Editor" promoted to a first-class v3 sidebar item.** The editor
plugin (`id: editor`) now gets its own dedicated sidebar entry — under the
Library group, just below Songs — via the existing `PROMOTED_PLUGINS`
mechanism in `static/v3/shell.js`, instead of being reachable only through
the generic Plugins gallery. Gated on the plugin actually being installed
(`renderPromotedNav` checks `/api/plugins`), so it appears only when the
editor is loaded. The displayed label comes from the plugin's manifest
`nav.label`.
- **Guitar Pro → notation importer (`lib/gp2notation.py`)** (feedBack#825 WS4b, epic #828). Piano/keys tracks imported from Guitar Pro (GPIF: `.gpx` GP6 / `.gp` GP7-8) now produce real Sloppak Notation Format data (sloppak-spec §5.3) alongside the `midi = string*24 + fret` guitar wire encoding. `gp2rs_gpx.convert_file` writes a `<stem>.notation.json` sidecar next to each keys arrangement XML (best-effort — a notation bug never breaks the RS-XML conversion), and `gp2notation.attach_notation_to_sloppak()` is the assembly-side helper that renames it into `notation_<id>.json` + adds the per-arrangement `notation:` manifest sub-key. Voice→staff routing salvages the logic from PR #703 (whose `stf` wire-field approach this supersedes): GP voice position 0 → `rh` staff (`G2`), positions ≥ 1 → `lh` (`F4`); a forced-LH track (the merged `Piano LH` partner from `_find_piano_pairs`, or a standalone track named `… LH`) routes everything to `lh` — preserving authored hand crossings instead of inferring hands from pitch. Emits measures with absolute `t` from the bar-indexed tempo map, change-only `ts`/`tempo`/`ks`, and `beat_groups` for compound/irregular meters (6/8 → `[3,3]`, 9/8 → `[3,3,3]`, 5/8 → `[2,3]`, 7/8 → `[2,2,3]` — cf. the feedBack#261 denominator pitfalls); beats carry `dur`/`dot`/`tu`/`rest` from GP rhythms and notes carry absolute `midi` (String+Fret resolves via the string template's concert pitches, Tone+Octave via `(octave+1)*12 + step`) with `tied` continuations kept as real beats (engraving needs the tied notehead — unlike the RS-XML walk, which drops them and extends sustain). Timing reuses the `gp2rs_gpx` machinery (bar-indexed tempo map, per-beat rhythm durations, `_note_midi`) so notation lines up with the RS XML the highway plays — with one deliberate divergence: double dots advance time ×1.75 (vs. the RS-XML walk's single-dot ×1.5 approximation) so a written `dot: 2` agrees with the emitted beat times; sharing the walk itself stays tracked in feedBack#618. Tests: `tests/test_gp2notation.py`.
- **Legacy keys → notation lifter (`scripts/lift_keys_notation.py`)** (feedBack#825 WS4c, epic #828). One-time batch converter that lifts existing **directory-form** piano/keys sloppaks from the legacy guitar wire encoding (`midi = s*24 + f`) into real Sloppak Notation Format files (sloppak-spec §5.3). Candidates are arrangements whose name matches `\b(keys|piano|keyboard|synth)\b` (case-insensitive); each gets a `notation_<id>.json` plus the per-arrangement `notation:` manifest sub-key. Measures derive from the song-level `beats` downbeats (`measure >= 0`; `song_timeline.json` preferred, first-arrangement fallback), with per-measure tempo from downbeat spacing (emitted only on a > 1 BPM change). Durations come from the wire sustain (`sus`, legacy `l` alias) when present, else the gap to the next onset in the same hand — quantized to the nearest plain/single-dotted `{1,2,4,8,16,32}` denominator at the local tempo, floored at a 32nd. Hands are split heuristically: onsets within 10 ms form a group; a group spanning > 12 semitones splits at its largest internal interval gap (low side → `lh`), otherwise the whole group goes by mean pitch vs middle C — single-staff output when everything lands on one hand. Idempotent (arrangements already carrying `notation:` are skipped; an orphan `notation_<id>.json` without the manifest key is refused, not overwritten) with `--dry-run` support; every payload is checked via `notation.validate_notation` before write. Honest caveat: the manifest is round-tripped through PyYAML (`safe_load` + `safe_dump(sort_keys=False)`) — key order survives, YAML comments/custom formatting do not (the script warns when comments are present). Zip-form `.sloppak` files are reported and skipped. Tests: `tests/test_lift_keys_notation.py`.
- **Notation schema v1 freeze — completeness batch** (feedBack#822, epic #828). Adds the low-hanging-fruit fields ahead of content production: top-level credits `rights`/`lyricist`/`arranger`; measure `pickup` (anacrusis); beat `arp` (arpeggiate), `ferm` (fermata), and **typed grace notes** — `grace: "a"` (acciaccatura, MusicXML `grace/@slash=yes`) / `"p"` (appoggiatura); note `stem` (`"up"`/`"down"` force). Pedal is settled as the existing `spd`/`sph`/`spu` trio with a documented MusicXML `<pedal start|change|stop>` mapping — no separate `ped` field. A new "v1 non-features" spec subsection pins the accepted limitations (microtonal, figured bass, mid-measure key/time/clef changes, `ott`/`barline`/ornaments/`trem`/`glis`) as additive-v1.x territory. `lib/notation.py` gains the `GRACE_TYPES`, `STEM_DIRECTIONS`, and `DYNAMICS` vocabularies; the validator stays permissive by design.
- **Notation format — standard musical notation as a first-class sloppak type.** Promotes keys, piano, violin, and any other staff-notation instrument out of the guitar wire format and into their own data structure, following the same promotion path used for drums (feedBack#344). New `lib/notation.py` defines the canonical vocabulary (`CLEFS`, `DURATIONS`, `SCHEMA_VERSION`), a permissive `validate_notation()` check, and `measures_to_wire()` / `measure_to_wire()` wire helpers. `lib/sloppak.py::load_song` reads a new per-arrangement `notation:` sub-key from each arrangement entry in the manifest (Option B: per-arrangement, not song-wide), applies path-traversal guards, validates the parsed JSON via `validate_notation()`, and surfaces all notation payloads on `LoadedSloppak.notation_by_id` (a `dict[str, dict]` keyed by arrangement id). A failed or missing notation file for one arrangement does not abort or skip the arrangement itself — partial-failure isolation mirrors the drum tab loader. `file:` is now optional when `notation:` is present: the loader creates a minimal stub arrangement so a notation-only arrangement entry does not require a guitar wire format JSON. `/ws/highway/{filename}` gains two new message types — `notation_info` (staves, instrument, total measure count) and chunked `notation_measures` (32 measures per chunk) — streamed after `sections` and before `anchors`; `song_info` carries a new `has_notation: bool` flag so viz pickers can auto-activate the notation plugin regardless of arrangement name. The notation file schema is measure-structured (`measure → staff → voice → beat → note`), uses MIDI for pitch (no string/fret/tuning indirection), and carries the full set of effects that alphaTab can render. See `docs/sloppak-spec.md` §5.3 for the full schema. Open questions resolved per the piano/keys epic (feedBack#828 / #822): Option B (per-arrangement `notation:` sub-key) and `file:`-optional-when-`notation:`-present are the endorsed design.
- **`song_timeline.json` — beats and sections as a top-level file.** A new optional top-level file pointed at by a new manifest key (`song_timeline: song_timeline.json`) provides the correct home for song-wide beats and sections, replacing the legacy convention of embedding them in the first arrangement JSON. The loader in `lib/sloppak.py` reads and validates the file (must be a dict with `beats` and `sections` as lists), clears and repopulates `Song.beats` / `Song.sections` from it when present, and stores the raw dict on `LoadedSloppak.song_timeline`. The existing arrangement-JSON fallback is fully preserved: all existing sloppaks that omit `song_timeline:` continue to load without any change. This is a prerequisite for notation-only sloppaks, which may have no arrangement JSON at all and therefore no carrier for beats/sections data. New sloppaks should put beats/sections in `song_timeline.json` only. See `docs/sloppak-spec.md` §2 and §5.3.
- **`note-detection` capability domain promoted — control plane (spec 009)** (feedBack#727/#728, epic #828). New core host `static/capabilities/note-detection.js`: provider registry (kinds `midi`/`engine`/`js`, primitives `pitch.estimate`/`verify.target`), requester-owned context-scoped detection bindings (`open-binding`/`close-binding`/`set-target`/`clear-target` — each binding carries its own redacted tuning context, independent of the host's loaded song, per spec-009 FR-003), and hit/miss/verdict observability events (consumers own judgment). The legacy chart-coupled `highway.setNoteStateProvider` surface keeps working and is wrapped for compatibility-shim hit accounting. Diagnostics (`feedBack.note_detection_capability.v1`) carry provider/binding summaries and bounded outcomes — no raw audio, device labels, or song identity. Migrating the chart path, Step Mode verify, minigames YIN, and the engine verifier onto bindings is the remainder of the spec-009 slice.
- **`visualization` capability domain promoted (cap:6)** (feedBack#828). New core host `static/capabilities/visualization.js` registers a provider-coordinator owning the highway renderer surface: commands `inspect` / `list-providers` / `select-renderer` / `clear-renderer` (selection delegates to the existing picker so persistence, WebGL2 gating, and fallback stay single-sourced), events `providers-refreshed` / `renderer-changed` / `renderer-ready` / `renderer-failed`. Legacy discovery (`type: "visualization"` manifests, `window.feedBackViz_*` globals) keeps working unchanged and is accounted as compatibility shims with hit counts. `static/app.js` attributes every renderer change (auto-match / user-select / fallback) and auto-match outcomes into the domain. Diagnostics (`feedBack.visualization_capability.v1`) carry provider ids/labels/context types, active renderer + selection source, last auto-match outcome, and last failure — no song filenames/titles. Per-panel (splitscreen) selection is a tracked follow-up.
- **Viz picker routes notation arrangements** (feedBack#826, epic #828). `window.feedBack.currentSong` gains `hasNotation` (sibling of `hasDrumTab`) from the `song_info` frame's `has_notation` flag, so notation viz plugins (Staff View, Keys Highway 3D) can gate `matchesArrangement` on data presence instead of arrangement-name heuristics. When a notation-only arrangement (no wire notes — `file:` omitted per sloppak-spec §5.3) falls through Auto with no notation plugin installed, the built-in highway still takes the canvas but the Auto label reads "no notation view installed" and a one-shot dismissable hint points at the visualization picker — never a silently blank board.
- **Keys instrument path in progression** (feedBack#828). New `data/progression/paths/keys.json` (5 levels / 15 challenges at parity with the guitar path) plus keys-flavoured daily/weekly quest pool entries (`d.keys-one` "Ivory Tower", `w.keys-three` "Grand Recital"). `lib/progression.py::instrument_for_arrangement()` now attributes `type: piano|keys` arrangements — and names matching `keys`/`piano`/`keyboard`/`synth` on a word boundary — to the new `keys` instrument, so scored keys runs advance the path automatically. Purely content + attribution: no schema or API changes.
- **v3 library: exact artist/album filters + scroll/page-depth restore** (feedBack#857). The v3 Songs toolbar gains Artist and Album dropdowns (Album populates from the selected artist and stays disabled until one is chosen), backed by new exact, case-insensitive (`COLLATE NOCASE`) `artist` / `album` query params threaded through `MetadataDB._build_where` → `query_page` / `query_artists` / `query_stats` and the `/api/library`, `/api/library/artists`, `/api/library/stats` endpoints (the free-text `q` search stays fuzzy and composes with the exact filters). The artist/album catalog is fetched independently of the active artist/album selection so the dropdowns always list the full set for the current provider/search. The toolbar is now sticky so filter controls stay reachable when browsing deep libraries, and returning from the player restores the previous scroll position **and** the loaded infinite-scroll page depth via a `sessionStorage` snapshot keyed by a filter/sort/view state hash (invalidated whenever those change, so a filter change still resets to the top). Tests: `tests/test_library_filters.py` (backend artist/album filters), `tests/js/v3_songs_scroll.test.js` (state-hash + snapshot helpers).
### Fixed
- **v3 Songs grid now refreshes after a Settings rescan / DLC-folder change — no app restart needed.** On a fresh install, pointing at a DLC folder in Settings and running a scan left the Songs section empty until a restart (the scan *did* populate the library — `_background_scan` re-reads `config.json` fresh — but the v3 grid never reloaded). The Settings **Rescan / Full Rescan** handlers only refreshed the classic (v2) library via `loadLibrary()`; the v3 grid (`static/v3/songs.js`) had no listener for a scan it didn't start itself (only its own upload path self-refreshed via `watchUploadScan`), so its cached, pre-DLC (empty) DOM/snapshot survived a sidebar return until a full reload. The rescan handlers now emit a **`library:changed`** event (`static/app.js`); the v3 grid listens and **reloads if it's the active screen, else marks itself dirty** so the next entry does a full re-fetch instead of restoring the stale snapshot (a `_libraryDirty` short-circuit ahead of every cached-DOM fast-path in `onV3SongsScreenEnter`). Tests: `tests/js/v3_library_refresh.test.js` (the emit + the reload/dirty wiring).
- **Edit Metadata modal: the Year is now editable.** You could set a year when authoring a pak but the Songs → Edit Metadata modal had no Year field, so it could never be changed afterward. The backend (`POST /api/song/<f>/meta`) already accepted and normalized `year` (writes it into the file via `songmeta`, survives a rescan) — only the UI omitted it. Added a **Year** input to `openEditModal()` (populated from the song's existing year) and included `year` in `saveEditModal()`'s POST body (`static/app.js`). Both the v3 card menu and the legacy edit button already pass the year through, so both surfaces get the field.
- **Edit Metadata modal no longer closes when a click-drag is released on the backdrop.** Selecting text inside a field and releasing the mouse past the modal's edge dismissed the form without warning (the `click` event's target resolved to the backdrop), discarding the edit. Backdrop dismissal now requires the **mousedown to have started on the backdrop** too — tracked per-modal and decided by a new pure `_editModalShouldClose(clickTarget, modalEl, downOnBackdrop)` helper (`static/app.js`). Cancel / ✕ still close on a normal click. Tests: `tests/js/edit_metadata_modal.test.js` (year in the POST body + the backdrop-close decision table).
- **Built-in diagnostic sloppak rebranded "Slopsmith" → "FeedBack" in the song name.** PR #586 renamed the file to `feedBack-diagnostic-basic-guitar.sloppak` but never regenerated the archive, so the manifest inside still carried `title: Slopsmith Diagnostic — Basic Guitar` / `artist: Slopsmith` (and the same heading in `DIAGNOSTIC.md`) — the stale name testers saw in the library/player and the onboarding calibration step, even though the build script, server, and docs all already say "FeedBack Diagnostic — Basic Guitar". Regenerated `docs/diagnostics/feedBack-diagnostic-basic-guitar.sloppak` from `docs/diagnostics/build_diagnostic_basic_guitar.py` so the committed artifact matches its source generator (title/artist/heading now "FeedBack"; chart, stem, and `diagnostic:` metadata unchanged). No code change — the rename in #586 just needed the rebuild.
- **v3 song/lesson accuracy badges now refresh on the first return from a song — no restart needed.** PR #574 added a `stats:recorded` → in-place badge repaint, but the repaint never matched a card. The event (like `song:loading`) carries the filename **`encodeURIComponent`'d** — exactly as `playCard` hands it to `playSong` (the highway WS `decodeURIComponent`s it back) — whereas library cards key on the **decoded**`localFilename` (`data-fn`), and `/api/stats/best` is server-canonicalized to that same decoded key (`server.py``_canonical_song_filename`). So `repaintAccuracy`'s `data-fn !== key` check rejected every card and `state.accuracy[encoded]` was `undefined`, leaving the just-earned badge stale until a full `render()` (app restart / search / re-enter the screen) — which is why it "came back after a restart." `static/v3/songs.js` now decodes the `stats:recorded` filename back into the card / `state.accuracy` key space via a small `decFn` helper before marking dirty and repainting (idempotent for already-decoded names; falls back to the original on malformed input so a real filename containing a literal `%` is never corrupted), so both the immediate repaint and the `onV3SongsScreenEnter` deferred path land on the right card. Tests: `tests/js/v3_songs_score_badge_refresh.test.js`.
- **Escape now exits a song (and leaves Settings) even when a transport/rail control button holds keyboard focus.** Clicking a player control (Play / FF / RW / Restart) left that `<button>` focused, and `_shortcutDispatchBlocked()` in `static/app.js` treats any focused `INPUT/SELECT/TEXTAREA/BUTTON` as an "interactive control" and bails before the shortcut registry runs — so the player-scope `Escape → Back` shortcut never fired until the user clicked empty canvas to blur the control ("Escape in song not consistent"). Space already had a player-screen carve-out (#593) that let it fire through a focused control; Escape did not. Generalized that carve-out to Escape, scoped to the player **and** settings screens (both register an `Escape = Back` shortcut, and settings had the identical latent bug). The earlier guards are preserved and still win: text inputs are exempted first (Escape there clears/blurs the field), the Section Practice popover already claims Escape before the carve-out, and a true modal layered over the screen (`[role="dialog"][aria-modal="true"]` / `.feedBack-modal`) still traps Escape so it closes the modal rather than ejecting past it. Escape becomes a reliable, focus-independent "Back" — making it monotonic groundwork for an optional exit-confirm. Plugins that register a player-scope `Escape` shortcut benefit identically (they were broken the same way). Tests: `tests/browser/keyboard-shortcuts.spec.ts` (focused-button repro, text-input no-exit, no-escape-past-modal, Section Practice popover, settings twin-bug).
- **The v3 "Up Next" pill can now be turned off — new "Show 'Up Next'" gameplay toggle (default ON).** The v0.3.0 player chrome's persistent upcoming-section pill (`#v3-upnext`, drawn by `static/v3/player-chrome.js`'s `updateUpNext()`) shipped with no off switch, so it always showed during playback whenever a section was upcoming — overlapping the top-right FPS HUD and ignoring the 3D-highway "Show 'Up Next' section card" checkbox (a *different*, in-canvas widget that was demoted to default-off precisely because this pill is the canonical readout). Users reading the pill as the same setting saw "disabled in settings but still there." Adds a real core toggle following the `autoplayExit` idiom: a client-only `showUpNext``localStorage` pref (absence = enabled), a **Show "Up Next"** switch in the Gameplay settings tab (`static/v3/index.html`), reader/writer + `loadSettings()` hydration + a read-only `window.feedBack.showUpNext` getter in `static/app.js`, and a gate at the top of `updateUpNext()` that hides the pill when off. Disabling mid-playback hides it immediately; re-enabling re-shows it on the next chrome tick (~6 Hz). Added to `RESET_MAP.gameplay.local` in `static/v3/settings.js` so the Gameplay "Reset" restores the default-on state. Default ON = zero change for existing users. No Tailwind rebuild (plain markup + existing classes).
- **v3 list/tree view brought to parity with the grid: select mode, parts chips, and song actions — plus a stale-CSS Docker fix.** Re-lands a previously-reverted change. **Frontend (`static/v3/songs.js`):** entering select mode no longer collapses the tree — `loadTree()` now captures the expanded artist groups (`details[open]` keyed by `data-artist`) before the "Loading…" wipe and restores them on rebuild, so toggling select mode (which re-renders via `reload()`) keeps groups open and selection usable; tree rows gain a display-only checkbox + selection ring, the same fav / save-for-later / overflow-menu cluster as the grid card (always shown, all bound by `wireCards()`), and a capture-phase select guard mirroring the grid so clicking a row or arrangement chip in select mode selects instead of playing (`<summary>` headers sit outside `[data-fn]`, so native expand/collapse is untouched). **Docker fix (`static/tailwind.min.css`):** the committed Tailwind stylesheet was stale — `.sm\:flex` (and the other utilities behind #582's `hidden sm:flex` arrangement chips and the new action cluster) were never compiled in, so they rendered `display:none` on the Docker build (which serves the committed CSS as-is; Desktop rebuilds from source so it looked fine). Regenerated with the pinned `tailwindcss@3.4.19` via `scripts/build-tailwind.sh` so Docker matches Desktop and #582's chips render on every Docker deploy. Regression tests: `tests/browser/v3-tree-select.spec.ts`.
- **Space bar now plays/pauses on the player screen even when a sidebar nav link or rail button has focus.** When any `<button>` in the player rail (viz, audio, mixer, lyrics, plugins, advanced), a sidebar nav link, or a popover control held keyboard focus, pressing Space was swallowed by `_shortcutDispatchBlocked` → `_isInsideInteractiveControl` (which treats `BUTTON`/`A` as interactive), so the Space shortcut never reached the dispatcher and `togglePlay()` never ran. `_shortcutDispatchBlocked` (`static/app.js`) now extends the same carve-out already used for the Section Practice bar: while the player screen is active, Space is always routed through the shortcut system — the dispatcher calls `e.preventDefault()` before invoking the handler, so the focused element does not also activate. Text inputs (`_isTextInput`) remain exempted first, so typing space in a search/input field still works normally, and focus inside a true modal dialog (`role="dialog" aria-modal="true"` / `.feedBack-modal`) layered over the player is also exempted so Space reaches the modal's focused control (e.g. its Close button) instead of toggling playback behind it — non-modal player popovers/toasts (loop A/B, arrangement pin) stay covered. Regression tests in `tests/browser/keyboard-shortcuts.spec.ts` cover the focused-rail-button play/pause, the text-input exemption, and the modal-dialog exemption.
- **A song's accuracy badge now updates on its library card right after you play it — no restart needed.** The v3 library (`static/v3/songs.js`) loaded the best-accuracy map (`/api/stats/best`) once into `state.accuracy` at render time and only ever refreshed it on a full re-render; the play→return flow takes the screen-entry fast-path that restores the cached grid DOM without re-fetching, so a just-earned score stayed invisible until the next restart re-ran `render()`. The `stats-recorder` now emits a `stats:recorded` event (carrying `filename`/`arrangement`) once the scored `POST /api/stats` resolves on the server — the correct moment, since `song:stop` fires before the POST completes. `songs.js` listens: if the library is the active screen it re-fetches `/api/stats/best` and patches the affected card/row badge in place; otherwise it marks the filename dirty and `onV3SongsScreenEnter` applies it on return (a failed fetch keeps the entry dirty to retry). Badge markup was factored into a shared `accuracyBadge(filename, variant)` (grid pill + tree-row percentage, both tagged `.fb-acc-badge`) so the in-place `repaintAccuracy` can find and replace them without a full list re-render (scroll/pagination preserved). The old empty `song:stop` "refresh lazily next render" placeholder is replaced.
- **Changing Settings → 3D Highway → Fret spacing no longer ejects you to the home screen.** The `highway_3d` plugin's `h3dSetFretSpacing` was the lone 3D-highway setting that called `location.reload()` to apply — and since the SPA boots with `#home` as the active screen (`index.html``.screen.active`), the reload dropped the user out of Settings onto the homescreen. It now applies live like every other 3D-highway setting: it rebinds the module-scope `_h3dFretUniform` flag (so panels mounted later this session pick up the new mode), recomputes the two `fretX`-derived scalars that were baked at init (`_fretLabelScaleRefW` for fret-label sprite scaling, `FRET_WIDTH_MID` for camera hysteresis), and broadcasts a `fretSpacing` change over the existing `_bgEmitChange` pub-sub so every mounted panel rebuilds its board via `buildBoard()`. Per-frame note geometry already reads `fretX` live and needs no rebuild. No page reload, so the Settings screen stays put. Source-level regression tests in `tests/js/highway_3d_fret_spacing.test.js` now pin the no-reload / live-rebuild behavior.
- **v3 library scroll-restore no longer breaks the classic v2 UI or drops off-screen searches** (feedBack#857). Two regressions in the scroll-restore work above: (1) `playSong` remapped `home`-launched songs to return to the `#v3-songs` screen unconditionally, but `static/app.js` is shared with the v2 UI (served at `/v2` / `FEEDBACK_UI=v2`) where that screen does not exist — Esc-from-player then called `showScreen('v3-songs')`, which threw on the missing element and stranded the user on a blank screen with playback still running; the remap now applies only when `#v3-songs` is present. (2) The Songs screen-entry fast-path skips reloading to preserve scroll, but the global topbar search routed through it, so once Songs had been visited, searching from another screen navigated there without applying the new query; the screen now tracks the state hash each fetch reflects and refetches when it has drifted, keeping the scroll-preserving no-op only when nothing changed.
- **An active custom highway renderer is no longer starved of `draw()` when it hides the canvas** (#819). The per-frame draw gate in `static/highway.js` bailed on `if (!_lastVisible) return`, which conflated two different "hidden" states: a genuine off-screen canvas (`offsetParent === null` — navigate-away / `display:none` splitscreen panel, #246) versus a renderer-set *override-hide* (`setVisible(false)`, where an opaque overlay covers the canvas but the active renderer keeps painting its own surface). The gate now only pauses everything for the off-screen case (and still pauses the default 2D renderer on an override-hide); the **active custom renderer** keeps receiving `draw()` through its own override-hide. The `highway:visibility` event still fires before the gate, so sibling overlay renderers (e.g. 3D Highway's `.h3d-wrap`) still pause. This is the core-side root cause behind the Tab View cursor freezing in single-player (feedBack#734; worked around plugin-side in feedBack-plugin-tabview#25).
- **Screensaver no longer kicks in during windowed-mode playback** (#686). While a song is playing, `static/app.js` now holds a [Screen Wake Lock](https://developer.mozilla.org/en-US/docs/Web/API/Screen_Wake_Lock_API) (`navigator.wakeLock.request('screen')`) so the OS display/screensaver stays awake even though only audio + the highway animation are active and the keyboard/mouse are idle. The lock is acquired on `song:play`/`song:resume` and released on `song:pause`/`song:ended`/`song:stop` (kept only while actually playing), and re-acquired on `visibilitychange` when the tab refocuses (the API auto-releases a lock whenever the page is hidden). Both the HTML5 `<audio>` and JUCE desktop playback paths emit the same `song:*` events, so the fix covers both. In feedBack-desktop (Electron), where `navigator.wakeLock` is unreliable, it also drives a native `powerSaveBlocker` bridge via the optional `window.feedBackDesktop.power.setScreenAwake` hook when present; both calls degrade silently where unsupported. Note: the browser Wake Lock API is secure-context only, so in a plain browser this is active on `localhost` / HTTPS only — a session opened over plain HTTP to a LAN IP (e.g. `http://192.168.1.100:8000`) won't keep the screen awake; front it with HTTPS or use the desktop app (see README → reverse-proxy notes).
### Changed
- **Practice plugin first-class sidebar slot now points at Virtuoso.** The bundled practice plugin was rebranded/re-homed from the SlopScale fork (`id: slopscale`) to `got-feedback/feedBack-plugin-virtuoso` (`id: virtuoso`); the desktop bundle swap is feedBack-desktop#31. `static/v3/shell.js` still promoted `slopscale`, whose id no longer ships — so `renderPromotedNav()` (gated on the plugin appearing in `/api/plugins`) would have found no match and the dedicated sidebar slot would have gone dark, dropping Virtuoso to the generic Plugins gallery. Update the NAV entry + `PROMOTED_PLUGINS` slot `slopscale` → `virtuoso` (`screen: plugin-virtuoso`, label "Virtuoso - Practice", same FeedBarcade anchor + `target` icon) so the practice plugin keeps its first-class entry. Also clear the now-dead `slopscale` id from the Plugins-gallery curated category map (`static/v3/plugins-page.js`) and add `virtuoso: 'practice'` as a defensive fallback (the manifest's `category: "practice"` is authoritative, so it lands on the practice board regardless), and refresh the stale SlopScale references in `README.md` + `docs/plugin-capability-inventory.md`. Must land with the bundle swap or the practice plugin regresses in the UI.
- **3D highway: realistic curved metal frets.** The fret wires are now bowed `TubeGeometry` (the middle strings push away from the camera so the row of frets reads as wrapping a cylindrical neck — a depth cue) rendered with a lit `MeshStandardMaterial` instead of the old flat, straight `MeshBasicMaterial` boxes, so the scene's ambient + directional light glints across the rounded surface for a polished-steel look. The existing per-frame highlight is preserved unchanged: frets inside the active anchor still turn gold (`0xD8A636`), which under the metallic shading reads as brass. Metalness is kept moderate (0.4, not full-metal) because the scene has no envMap — a PBR full-metal surface would reflect black — with a dim emissive floor so frets stay legible down the fogged neck. Backported from the `highway_babylon` plugin's "hit-zone fret bars". All knobs (`FRET_BOW_DZ`, metalness/roughness/emissive) are tunable constants. `plugins/highway_3d` v3.25.0.
- **3D highway: section + tone HUD cards now default OFF.** The v0.3.0 player chrome carries a persistent "Up Next" pill, making the in-canvas section card redundant by default (it doubled the readout, feedBack feedback); the tone HUD follows the same less-is-more default. Both remain available in Settings → 3D Highway (visibility/position/size unchanged); users who previously toggled either explicitly keep their stored preference — only the untouched default flips. `plugins/highway_3d` v3.24.1.
- **Perf**: replace runtime Tailwind Play CDN with a prebuilt static stylesheet (`static/tailwind.min.css`). The Play CDN's runtime JIT scanned the DOM ~1.8x/sec on the main thread (~37 ms blocking spans), dropping ~26% of frames in long playback sessions with the 3D highway as default. Theme extensions (dark/accent/gold colors, Inter font) move to `tailwind.config.js`; regen via `bash scripts/build-tailwind.sh`. No runtime build step — the generated CSS is committed. Fixes feedBack-desktop#110.
- **Perf**: reduce per-frame allocations in the 2D highway chord + lyric render paths. `_ensureChordRenderCache` now also caches `sortedNotes` / `nonZeroNotes` / `nonZeroFrets` / `allMuted` / `hasMultipleNotes` (computed once per chord, invalidated on `src` / `_inverted` / `chordTemplates` change — the third key catches a stale `isOpen`-derived classification when the WS `chord_templates` message lands after the final `chords` chunk), so `drawChords` no longer re-sorts / re-filters / spreads min-max per visible chord per frame. The in-chord unison bend classification is folded inline (no `chordPositions.filter`× 2 per frame). `drawLyrics` memoizes `ctx.measureText` results in a two-level `Map<fontSize, Map<text, width>>` so cache hits don't allocate a composite string key. Lit-sustain shimmer in `drawSustains` swaps the 4 per-note-per-frame `Math.random()` calls for a 64-entry precomputed jitter LUT (xorshift32-seeded — the LUT contents are reload-stable and test-reproducible; rendered shimmer is deterministic per `createHighway()` instance, since the seed includes that instance's `_frameIdx`) indexed by `(frameIdx + n.s + ⌊n.t·60⌋)`, visually indistinguishable and allocation-free.
- **Perf**: the load-adaptive render scale (`_adaptRenderScale`, #654) no longer visibly hunts up/down on passages that hover near the frame budget (testers saw "quality going up and down" with the 3D highway). Downscaling stays prompt to protect the frame rate, but **upscaling is now lazy**: a smaller step (×1.06 vs ×1.1) on a longer cooldown (`_AUTO_UPSCALE_COOLDOWN_MS` 2500 ms vs the 600 ms adjust cooldown), reset on any downscale, and gated by a predictive guard — it only upscales when the projected cost *after* the step (≈ cost × step², since draw cost tracks the pixel count) still clears the high budget. The scale therefore settles just inside the 7–12 ms deadband instead of oscillating across it. No new public API; the `_autoScaleMin` "Min res" floor is unchanged.
### Removed
- **`c` library hotkey ("Convert to .sloppak") removed from core.** Core hardcoded a plugin-specific shortcut: a documentation-only `registerShortcut({ key: 'c', scope: 'library' })` no-op plus a `c → button.sloppak-convert-btn` entry in the library keydown handler that fired the Sloppak Converter plugin's button. Per the plugins-own-their-behavior principle, core no longer ships this hotkey — the convert button still works by click, and the Sloppak Converter plugin can register its own `c` shortcut via `window.registerShortcut()` if keyboard access is wanted. The `f` (favorite) and `e` (edit) library hotkeys, which drive core buttons, are unchanged. Help-modal/registry tests in `tests/browser/keyboard-shortcuts.spec.ts` updated to drop the `c` assertions.
### Added
- **Player progression: Mastery Rank, instrument-path challenges, daily/weekly quests, Decibels currency, cosmetics shop (spec 010).** Onboarding gains two steps: pick one or more **instrument paths** (Guitar / Bass / Drums — data-driven, more can ship as content) and a **calibration challenge** offer (play the bundled FeedBack Diagnostic with note detection at 100% accuracy — or skip; either way you reach **Mastery Rank 1**, and a skipped calibration can still be completed later from the Progress screen). Each path levels by completing a content-defined number of **challenges** (any order) from that level's set; Mastery Rank = onboarding rank + the sum of path levels, starting at 0 on a fresh install. The existing unified XP backend is untouched but the frontend renames it to **Decibels (dB)** — a spendable currency earned ONLY by playing (songs, FeedBarcade rounds, quest rewards; no real-money path exists or may be added) — with spend tracked in a separate wallet so lifetime earnings stay monotonic. Rotating **daily/weekly quests** (deterministic per period, lazy instantiation, local-midnight / Monday resets) award dB and feed `quest_completed` challenges. A new **Progress** screen (rank hero, per-path checklists, quest countdowns, add-a-path) and **Shop** screen (themes via CSS-variable swaps under `html[data-fb-theme]`, avatar frames; atomic balance-checked purchases — 402 on insufficient dB, 409 on re-buy) join the v3 nav, and the topbar badge now shows Rank + challenge-set progress + dB balance. All definitions live in `data/progression/` JSON (paths/levels/challenges, quest pools, shop catalog) — adding a rank, challenge, quest, or cosmetic is a content edit + restart, never code; invalid content degrades to logged warnings. New tables (additive + idempotent): `progression_state`, `player_paths`, `challenge_progress`, `quest_state`, `wallet`, `shop_owned`, `shop_equipped`. New endpoints: `GET /api/progression`, `POST /api/progression/paths|onboarding|events` (events whitelists `minigame_run`; `song_completed` stays server-derived inside `POST /api/stats`, which now resolves the instrument server-side and reports an additive `progression` outcome key), `GET /api/shop`, `POST /api/shop/buy|equip`; equipped cosmetics ride along on `GET /api/profile`. A new **`progression` capability domain** (core-owned, kind: command, safety: safe — `inspect`, `record-event`, `list-shop`, `buy-item`/`equip-item` gated on user action) emits `challenge-completed`/`quest-completed`/`path-level-up`/`rank-changed`/`db-changed`/`calibration-completed`/`cosmetic-equipped`, mirrored as `progression:*` window events, with a redaction-safe diagnostics contributor; backend plugins get the symmetric `record_progression_event` context hook (the bundled minigames hub reports runs through it, guarded for standalone). Spec: `specs/010-progression-domain/`. Tests: `tests/test_progression.py`, `tests/test_progression_api.py`. **Migration notes:** existing XP totals carry over as lifetime dB (balance = lifetime − spent); resetting a per-source XP contribution (e.g. a minigames profile reset) after spending can clamp the spendable balance to 0 until new dB is earned; drums-path v1 content uses currently-satisfiable goals (arcade rounds, quests, any-instrument plays) until drums scoring lands.
- **3D highway score FX (notedetect game-scoring layer).** The bundled `highway_3d` renderer now visualizes the scoring layer shipped in feedBack-plugin-notedetect ≥1.13: floating **"+N" score pops** above each judged gem (sourced from the note-state provider's new `{ points, mult, popKey }` verdict fields — chord members share the chord-level `popKey`, so a chord pops once, not once per string), plus session-level FX from the new `notedetect:fx` event — a particle burst at the strike line on streak milestones (25/50/every 100), an expanding ring pulse on multiplier tier-ups (×2/×3/×4), and a brief red wash when a ≥10 streak breaks. Colors and the pop font follow the user's notedetect scoring-UI skin (`feedBack_notedetect_skin`: neon/esports/metal, refreshed live on the `notedetect:skin` bus event). Everything renders on the existing 2D overlay canvas from fixed-size slot pools — no Three.js geometry, no text-sprite cache traffic, near-zero cost when idle — and degrades to a silent no-op with older notedetect builds (the new fields/events simply never arrive). Splitscreen panels scope FX to their own detector instance via the bubbling per-panel `notedetect:fx` dispatch. `plugins/highway_3d` v3.24.1.
- **3D highway: slide direction arrows + gem-follow animation.** Slide notes now show a ›/‹ arrow indicating which way the slide goes — on the note/gem itself, as an early preview on the neck before the note arrives, and (optionally) chained further ahead for multi-leg slides — each independently toggleable in Settings → 3D Highway (`slideArrowApproachVisible`, `slideArrowNeckVisible`, `slideArrowChainPreviewVisible`). The note gem also now visually glides from its starting fret to the slide's destination over the note's sustain and holds there through the brief post-sustain linger, instead of snapping back to the starting fret — most noticeable on unpitched "slide to nothing" notes. `plugins/highway_3d` v3.25.2.
- **3D highway: up to 3 upcoming-note ghost previews per string, with fade-in/grow.** Each string now previews up to 3 upcoming notes (was 1) on a fixed 0.6 s fade-in/grow ramp, so tight same-string runs no longer pop in at full size right before impact and the player can read note order ahead of time. `isBlocked` (the pre-impact ghost suppression in a note's last 150 ms) is now scoped to chord notes only — for lead notes it had been blinking the ghost out right before each sustained note in dense runs. (Slide notes stay excluded too, per the slide-arrow work above, since their gem glides off the start fret.) `plugins/highway_3d` v3.26.0.
- **Enable/disable plugins from the v3 Pedalboard (footswitch backend).** Every `/api/plugins` entry now carries an `enabled` boolean (default `true`), and a new `POST /api/plugins/{plugin_id}/enabled` endpoint (`{"enabled": <bool>}` → `{"id", "enabled"}`) persists the choice to `CONFIG_DIR/plugin_state.json` (only non-default `enabled:false` entries are stored; a missing/corrupt file is tolerated and never crashes startup). The loader **skips disabled plugins at startup** — no requirements install, no `routes.setup()`, no screen/nav/capabilities — while still surfacing them in `/api/plugins` as a disabled entry (`status:"disabled"`, `enabled:false`) so the UI can show an "off" pedal you can switch back on. Toggling persists immediately and flips the in-memory flag so the next `/api/plugins` reflects it at once (a runtime-disabled plugin's already-mounted routes/screen remain until the next restart; re-enabling a startup-skipped plugin mounts on restart). A disabled plugin is **excluded from the capability pipeline** — its capability metadata is emptied in `/api/plugins`. Guard rails keep `capability_inspector` and `app_tour_*` always enabled (disable → `400`); unknown id → `404`; missing/non-boolean `enabled` → `400`. Backend only; the v3 Pedalboard frontend consumes this contract. Docs: `docs/plugin-v3-ui.md`.
- **fee[dB]ack v0.3.0 rebrand + UI redesign (opt-in, isolated).** The visible product is being renamed from **FeedBack** to **fee[dB]ack** (the `[dB]` is a decibel pun on the practice "feedback" loop) alongside a full dashboard-style UI redesign. **Rebrand scope is the app + docs wordmark only** — the repository, Python package, ghcr Docker image, `CONFIG_DIR`, and `FEEDBACK_*` env vars all keep the `feedBack` name, so existing deployments and data are unaffected. The redesigned UI is additive and served behind a feature flag: `FEEDBACK_UI=v3` flips the `/` route to the new `static/v3/` shell, and `GET /v3` always serves it; the default `/` stays byte-identical to 0.2.9 until 0.3.0 flips the default. This release adds the `static/v3/` scaffold (navy app shell + styled fee[dB]ack wordmark, brand SVG + favicon + PWA manifest with 192/512 icons), an additive `fb` Tailwind color palette (legacy `dark`/`accent`/`gold` retained) with `static/v3/**` in the content globs, and the regenerated `static/tailwind.min.css`. Vanilla JS, prebuilt Tailwind, no Play CDN (Principle II). Shell wiring, screens, profile/scoring backends, and capability-runtime integration land in subsequent v0.3.0 changes.
- **fee[dB]ack v0.3.0 app shell (sidebar + topbar + routing).** The v3 shell (`static/v3/index.html`) is now a re-chromed copy of the legacy app: the new left **sidebar** (HOME / LIBRARY groups) and **topbar** (secondary nav, search, Support, badge-cluster mount points) replace the hidden legacy navbar, and new `#v3-*` screens (dashboard/plugins/profile/playlists/saved) are added — while all legacy screens (`#home` library, `#favorites`, `#settings`, `#player`, `#audio`, plugin nav containers) are kept verbatim so `static/app.js` boots **unmodified** and the whole engine (player/highway, plugin loader, capabilities, audio, library, settings) is reused as-is. Navigation is the shared `window.showScreen` across `#v3-*`, reused legacy, and `#plugin-*` screens, with a responsive hamburger and a `localStorage`/`v3:`-namespaced shell. Plugin nav is mirrored into the sidebar from `/api/plugins` (UI placement is a deferred capability domain, so this uses the legacy loader, not capability dispatch). `static/v3/shell.js` wraps `window.showScreen` via the idempotent rehydration pattern to keep sidebar/topbar active-state in sync.
- **fee[dB]ack v0.3.0 player profile + first-run onboarding + unified XP + streak.** Adds a single-user core **profile** (`profile`/`profile_progress`/`xp_profile` tables in `web_library.db`, additive + idempotent): display name + avatar, a stable `player_hash` (SHA-256 of the first name + a once-generated salt — stable across later renames; a future-leaderboard label, never auth), and a **streak** (any session on a calendar day keeps it; a missed day resets to 1). New endpoints: `GET/POST /api/profile`, `POST /api/profile/avatar` (base64, re-encoded to a ≤512px PNG under `CONFIG_DIR/avatars/`), `GET /api/profile/avatar/{name}` (safe-joined), `GET /api/profile/avatars` (bundled defaults under `static/v3/avatars/`), `GET /api/profile/progress` (one call for the badge), and `POST /api/xp/award`. **Unified XP:**`lib/xp.py` is the single XP curve (same math the minigames plugin shipped); the core `xp_profile` store is the one source of truth the profile badge reads, exposed to plugins via `context["award_xp"]`/`get_xp_progress`/`seed_xp`. The bundled **minigames** plugin now delegates XP to the core store (seeding once from its existing `profile.json` so earned levels carry over) — so song-play, minigames, and tutorials all feed one level. Frontend: a blocking first-run onboarding overlay (name + avatar grid + upload), the topbar profile badge (avatar, 🔥 streak, level + XP bar), and the `#v3-profile` screen. Tests: `tests/test_xp.py`, `tests/test_profile_api.py`.
- **fee[dB]ack v0.3.0 song-stats store (best score + accuracy, plays, resume position).** A core `song_stats` table (`web_library.db`, additive + idempotent, PK `(filename, arrangement)`) records per-song/arrangement best/last score + accuracy, play count, and last position. Endpoints: `POST /api/stats` (scored session → `plays += 1`, `best_*` = max, `last_*` = new, plus **unified-XP award**`xp_for_run(score)` and a **streak** bump, both behind try/except so a side-effect failure never drops the stat write; or position-only `lastPlayPosition` → resume touch with no `plays` change), `GET /api/stats/{filename}` (aggregated across arrangements), `GET /api/stats/recent` (joined to song title/artist/art for "Jump back in"). Scoring stays frontend-driven: `static/v3/stats-recorder.js` tallies the `note:hit`/`note:miss` events the optional `feedBack-plugin-notedetect` already emits (and also accepts an explicit `note_detect:session-ended` summary), then POSTs on song end; it also persists resume position on pause/stop. No note-detect edit required — note-detection is a deferred capability domain, so the recorder uses those legacy events and degrades to "no accuracy" when the plugin isn't installed. Score/accuracy math is shared with the server via `lib/song_score.py`. Tests: `tests/test_song_score.py`, `tests/test_song_stats_api.py`.
- **fee[dB]ack v0.3.0 playlists, Saved for Later, and Continue-Playing.** Core playlist management (`playlists` + `playlist_songs` tables in `web_library.db`, additive + idempotent): create/rename/delete, add/remove/reorder songs, plus a reserved **Saved for Later** system playlist (created on first use; protected from rename/delete). Endpoints: `GET/POST /api/playlists`, `GET/PATCH/DELETE /api/playlists/{id}`, `POST /api/playlists/{id}/songs`, `DELETE /api/playlists/{id}/songs/{filename}`, `POST /api/playlists/{id}/reorder`, `POST /api/saved/toggle`, and `GET /api/session/continue` (derives the resume song + last position from `song_stats`, no new table). Frontend `static/v3/playlists.js` renders the `#v3-playlists` list + detail (drag-reorder, play, remove) and `#v3-saved`, and exposes `window.v3Saved.toggle()` for a "Save for later" affordance on song cards. Favorites reuse the existing favorites screen/API. Core REST, no capability domain. Tests: `tests/test_playlists_api.py`.
- **fee[dB]ack v0.3.0 Dashboard / Home.** The `#v3-home` dashboard (matching the v0.3.0 design target) composes the new backends: a "Welcome back, {name}!" banner with a patch-notes link (`/api/version`), a hero card (Start Playing / Create Lobby), a **Continue-Playing** card (`/api/session/continue` → art, tuning chip, 4-segment progress; click resumes via `playSong` + best-effort seek), a stats row (audio-routing widget placeholder until prompt 18, library count from `/api/library/stats`, plugins count from `/api/plugins` where `status==="ready"`), and a **Recently Played** grid (`/api/stats/recent`) with per-song accuracy badges (good/mid/low ramp). Each widget fetches + renders independently and degrades gracefully (missing/empty endpoint → placeholder, never blocks first paint). `static/v3/dashboard.js`; re-renders on return to Home and on profile update.
- **fee[dB]ack v0.3.0 tuner + instrument topbar badges.** The topbar gains an **instrument selector** (guitar/bass + string count + tuning + reference pitch) persisted via additive `/api/settings` fields (`reference_pitch` clamped 430–450, `instrument`, `string_count` 4–8, `tuning` name or semitone offsets); changing it emits `instrument:changed` so the note_detect scorer can re-tune (consumed once the external plugin adopts it). A **live tuner badge** stays idle until the user enables the mic (explicit gesture; `getUserMedia`), then shows a YIN note readout with a cents needle (green within ±5¢) using a new dependency-free `static/v3/tuner-core.js` (YIN + frequency→note/cents, honoring the reference pitch); clicking opens the full `feedBack-plugin-tuner` screen when installed. CPU-friendly (~20 Hz, paused when the tab ishidden, respects `prefers-reduced-motion`). Tests: `tests/test_settings_instrument.py`, `tests/js/tuner_core.test.js`.
- **fee[dB]ack v0.3.0 audio-routing widget (dashboard).** The dashboard's audio stat tile now reads the live audio session **through the capability runtime** — `audio-mix inspect` (route + faders + required kinds), `audio-input list-sources` (selected/available input), `audio-monitoring inspect` — and renders **Audio Input → VST/NAM/IR → Audio Output** with per-node state dots and a Connected/Not Connected line. It never touches `audio-mixer.js` internals or `nam_tone` routes directly; "Not Connected" is the honest browser default (no native route), and it degrades on `no-owner`/`no-handler`/`failed` or absent capabilities. Refreshes on `instrument:changed`, play/stop, capability audio events, and each Home visit. `static/v3/audio-routing.js`.
- **fee[dB]ack v0.3.0 Plugins page.** The `#v3-plugins` screen renders the enriched `/api/plugins`: a "{N} active" header (`status==="ready"`), a card grid per plugin (icon, name, version, status pill with the error on failed, capability summary badges — declared domains / validation warnings / unsupported versions / shim hits / bundled / type), an **Open →** action that navigates to the plugin's injected `#plugin-<id>` screen, and All/Bundled/Visualizations filters. Surfaces a deep-link to the bundled **Capability Inspector** rather than re-implementing the graph. No new backend. `static/v3/plugins-page.js`.
- **fee[dB]ack v0.3.0 Songs / Library screen (`#v3-songs`).** A native vanilla-JS library browser over the existing `/api/library*` endpoints: provider selector (via the `library` capability, not DOM scraping), grid + tree views, sort, format filter, a tri-state filter drawer (arrangements / stems / lyrics / tunings), topbar-driven search (debounced), infinite scroll, fb song cards with **accuracy badges** (good/mid/low ramp, batched via a new `GET /api/stats/best`), favorite + save-for-later affordances, and upload (reuses the existing uploader). The "Songs" sidebar nav now opens this screen. No regression to `/api/library*`. `static/v3/songs.js`.
- **`ui.library-card-injection` capability + native song-card actions (fee[dB]ack v0.3.0).** New core capability (`static/capabilities/library-card-actions.js`, owner `core.ui.library-card-injection`, exposed as `window.feedBack.libraryCardActions`) lets plugins **register** per-song library-card actions (id, label, placement, applicability, enabled state, run handler) with `action-registered`/`action-result` events — replacing the legacy `.song-card` DOM-injection pattern (roadmap domain #9, now delivered as a frontend host). The native Songs grid renders registered actions in each card's "⋮" menu; the built-in **Edit metadata** and **Convert to E Standard (retune)** actions ship through it (`static/v3/card-actions-core.js`, calling the existing `openEditModal`/`retuneSong` globals). Songs cards also gain **arrangement chips** (play a specific arrangement) and a **multi-select** mode with batch **Add to playlist** / **Save for Later**. Recipe in `docs/capability-recipes.md`; tests in `tests/js/library_card_actions.test.js`. Migrating the external card-action plugins (Sloppak Converter, Find More, editor) onto `register(...)` is a follow-up.
- **`centOffset` exposed via `getSongInfo()`** — the arrangement `<centOffset>` field (float, cents) is now parsed from all chart sources (loose folder XML, sloppak wire format) and sent as `centOffset` in the `song_info` WebSocket message. Plugins can read `getSongInfo().centOffset` to obtain the arrangement's pitch-shift offset — commonly `-1200.0` for extended-range bass (one octave down) or a small non-zero value for true-tuned content (e.g. A443 ≈ +11.8 cents). Defaults to `0.0` when absent.
- **`highway.getPhrases()` and `highway.getMastery()` public plugin API** — exposes phrase timing windows (`[{ index, start_time, end_time, max_difficulty }]`) and the current mastery slider value (`0..1`) as documented, stable plugin API. Both values were already in memory and reachable via internal names; this surfaces them with intent so plugins can implement section-aware logic (e.g. tracking accuracy per phrase, suppressing difficulty changes during a hard solo) without reaching into undocumented internals. Returns `null` when the song has no phrase data (GP imports, single-difficulty charts). Pair with the existing `hasPhraseData()` to gate phrase-aware code paths.
- **Tailwind freshness guard + wider plugin scan.** A new `tailwind-fresh` CI job (`.github/workflows/tests.yml`) rebuilds `static/tailwind.min.css` with the pinned `tailwindcss@3.4.19` and hard-fails on any diff, so the committed prebuilt stylesheet can no longer silently lag source (after PR #411 removed the runtime Play CDN, a stale file shipped unstyled elements with no guard). The `tailwind.config.js` plugin content glob is widened to `./plugins/**/*.{js,html}`, which also scans non-`screen.js` plugin JS (e.g. `plugins/app_tour_*/script.js`) that was previously invisible to the build. Regenerating under the wider glob is a no-op for runtime behaviour — it only adds classes that were already used in source. Groundwork for the plugin `styles` capability (constitution 1.1.0, Principle II): runtime-installed plugins ship their own compiled CSS rather than relying on core's build-time scan.
- **Plugin capability pipelines** — adds the first versioned capability coordination layer for plugin authors and support tooling. `/api/plugins` now exposes validated capability declarations, validation warnings, unsupported-version metadata, UI/runtime domain declarations, and compatibility shim summaries for legacy `nav` / `screen` / `settings` / `routes` / visualization surfaces. The browser runtime now tracks manifest participants separately from live handlers, explicit dispatch outcomes (`no-owner`, `no-handler`, `unsupported-command`, `incompatible-version`), claim lifecycle cleanup, manual override precedence, deterministic ownership conflicts, multi-provider ordering, shim hit counts, and a redaction-safe diagnostics snapshot capped at 64 KB. A bundled Capability Inspector plugin shows the live graph, and new docs cover the manifest schema, recipes, safety matrix, lifecycle cleanup, and diagnostics contract.
- **Audio graph/session capability slice** — promotes `audio-mix`, `audio-input`, `audio-monitoring`, and coordinated `stems` diagnostics into the capability runtime. The new audio session host records song route/fader state, redaction-safe input sources, monitoring lifecycle outcomes, stem automation claims/overrides/orphans, and compatibility bridge hits for legacy faders, song volume, Stems master volume, 3D Highway analyser taps, audio startup barriers, and input source handoffs. `core.audio.session` coordinates `stems` without replacing the Stems plugin as the owner of actual stem playback/state.
- **Audio-mix control plane** — makes `audio-mix` the player mixer source of truth. Core now exposes `list-faders`, `get-fader-value`, `set-fader-value`, `inspect-route`, and `inspect-analyser` through the capability runtime, routes native and compatibility-backed fader provider operations with a 2-second timeout, reports committed values back to the mixer UI, suppresses matching legacy faders when a native participant owns the same logical control, and expands audio-session diagnostics/Capability Inspector rendering for fader availability, source modes, bridge hits, route/analyser summaries, and timeout failures.
- **Audio-input control plane** — makes `audio-input` the redaction-safe source of truth for instrument input discovery and lifecycle. Core now exposes `list-sources`, `select-source`, `open-source`, and `close-source` through the capability runtime, persists selected logical sources, keeps inspect/list/select prompt-free, routes provider `source.open`/`source.close` operations with bounded outcomes, shares compatible open sessions across requesters, suppresses compatibility-backed duplicate sources when a native provider owns the same logical key, and expands audio-session diagnostics/Capability Inspector rendering for selected input, open sessions, bridge hits, storage status, and permission/device failures without exposing raw device labels or live audio handles.
- **Audio-monitoring control plane** — makes `audio-monitoring` the shared live-monitoring coordinator. Core now exposes provider registration/list/selection, explicit user-action `start`, requester-counted `stop`, prompt-free `inspect`/`monitoring.status`, and `set-direct-monitor` through the capability runtime. Monitoring starts integrate with selected `audio-input` readiness, background requesters can only attach to active compatible sessions, active sessions survive song/playback stops without auto-resuming after reload, native providers suppress compatibility-backed legacy monitor paths, and diagnostics/Capability Inspector now show providers, sessions, requesters, direct-monitor state, bridge hits, and distinct safe outcomes (`provider-selection-required`, `user-action-required`, `incompatible`, `unavailable`, `stopped`, etc.) without exposing raw audio/device data.
- **Playback control plane** — promotes `playback` to an active core capability domain for song transport, timing, loop, route, requester/observer, bridge, and diagnostics state. Core now exposes `inspect`, user-authorized `start`, `pause`, `resume`, `stop`, `seek`, `set-loop`, and `clear-loop` through the capability runtime while `static/app.js` keeps raw `<audio>`/JUCE handles private behind a redaction-safe adapter. Playback diagnostics use pseudonymous targets in exported bundles, local display labels only in the Capability Inspector, bounded recent outcomes/events, and bridge accounting for `window.playSong`, legacy `song:*` events, `window.feedBack` transport helpers, loop helpers, and browser/native route handoff.
- **3D highway — Tone HUD, fret dividers, chord-diagram toggle, FPS counter.** The bundled `plugins/highway_3d` gains an amber **Tone-change HUD** (shows the active tone and the next scheduled tone change; position / size / visibility configurable in settings), a **fret-dividers** toggle (vertical dividers on the highway, on by default, via `h3dBgSetFretDividersVisible`), a **chord-diagram visibility** toggle (`h3dBgSetChordDiagramVisible`), and an **FPS counter** setting migrated to `BG_DEFAULTS.fpsVisible` (drops the legacy `h3d_showFps` localStorage key). Chord-diagram position is restricted to `tl`/`tr`; legacy `bl`/`br` values are coerced on load. Perf: accent-halo shell descriptors are pre-built per string in `initScene()` and the chord-verdict cache key is encoded as a number, eliminating per-frame allocations in the `drawNote()` and chord hot paths.
- **Sloppak assembly preserves a short preview clip.** When a source chart carries a separate short browser-preview audio clip alongside the full song, the sloppak assembler now decodes it to `preview.ogg` at the sloppak root and records it under a new top-level `preview:` manifest key (POSIX relpath, same shape as `lyrics`/`cover`). A failed preview decode is logged at debug and skipped without aborting the overall build. Sources with no separate preview are unaffected. Older sloppak readers ignore the unknown `preview` key, so the change is purely additive (sloppak-spec.md §5.5 backward-compat). Documented in `docs/sloppak-spec.md` §2 alongside the other optional top-level keys. Enables [`feedBack-plugin-song-preview`](https://github.com/got-feedback/feedBack-plugin-song-preview) to render hover-to-listen previews for sloppaks without seeking into the full audio.
- **Generic plugin asset route** — `GET /api/plugins/{plugin_id}/assets/{path}` serves arbitrary static files a plugin bundles under its own `assets/` directory (AudioWorklet modules, WASM, images, etc.), so plugins can self-host browser-fetchable assets without a CDN (Principle II). Containment is enforced by `lib/safepath.safe_join` against `<plugin>/assets/`, so `..` traversal, absolute paths, and NUL bytes cannot escape `assets/` to reach a plugin's Python modules. `.js` is served as `application/javascript`. First consumer: the stems plugin's pitch-preserving time-stretch worklet.
- **Minigames framework — bundled as a core plugin (`plugins/minigames/`).** Promotes the upstream [`feedBack-plugin-minigames`](https://github.com/got-feedback/feedBack-plugin-minigames) repo into the core bundle so every FeedBack install gets the framework out of the box (same promotion path used for `highway_3d`). The plugin adds a top-level **Minigames** nav link (alongside Library / Favorites / Upload — not buried in the Plugins dropdown), a library-style card grid of installed minigame plugins, and a shared profile layer (XP, level, per-game leaderboards, cross-minigame unlocks) persisted under `CONFIG_DIR/minigames/` and opted into the settings export. Other plugins that want to ship a minigame add a `minigame` block to their `plugin.json` and call `window.feedBackMinigames.register(spec)`; the SDK exposes scoring (`createContinuous` runs a self-contained YIN tracker; `createDiscrete` / `createChord` wrap `note_detect`'s `createNoteDetector`), HUD primitives, run persistence, and a scheduler so individual minigames don't need their own DSP or backend. Backend endpoints live under `/api/plugins/minigames/{runs,profile,registry}`. The framework is plugin-shaped (not core code) per Principle III, but bundled so it ships with every install. First consumer: [`feedBack-plugin-flappy-bend`](https://github.com/got-feedback/feedBack-plugin-flappy-bend), shipped separately.
- **Alpha-build heads-up banner** — when `/api/version` reports a version string containing "alpha" (case-insensitive), an amber banner appears at the top of the library section warning users that the build is in active development and may have bugs or breaking changes. The banner stays hidden on stable / beta / RC builds. No persistence or dismiss state — it's a passive notice, not a modal.
- **Drum vocabulary expanded to 18 pieces** — adds `stack` (MIDI 30, from GM's extended-percussion range, unused by real drum-kit MIDIs) and `bell` (MIDI 80 "Mute Triangle", also unused in real drum-kit MIDIs) to `lib/drums.py` PIECES. Inserted in the iteration order so the editor / highway lane ordering is *hi-hat → stack → crash → … → ride bell → bell*. Both are cymbals; default shape `circle_jagged` (stack) / `circle_dot` (bell). Old drum tabs round-trip unchanged — the schema is permissive and existing piece-ids are untouched.
- **GP / MIDI drum import surfaces unmapped notes** — `convert_drum_track_to_drumtab` (`lib/gp2rs.py`) and `convert_drum_track_from_midi` (`lib/midi_import.py`) gain an optional keyword-only `out_unmapped` parameter. Callers that pass an empty dict receive a per-MIDI record of every silently-skipped percussion note (`{midi: {"count": int, "times": [float, ...]}}`, times capped at 100 samples per note). This lets the editor plugin show a warning + manual-mapping UI on import instead of silently dropping unmapped notes. Default behavior unchanged for callers that don't opt in.
- **Drum support from scratch** — drums are now a first-class arrangement type with their own JSON payload on disk and their own WS stream to the highway. New `lib/drums.py` defines the closed piece-id vocabulary (kick, snare, snare_xstick, hh_closed/open/pedal, tom_hi/mid/low/floor, crash_l/r, splash, china, ride, ride_bell), default GM-MIDI mappings, three preset lane configurations, and a permissive `drum_tab.json` validator. `lib/sloppak.py::load_song` reads the manifest's optional top-level `drum_tab:` key, parses + validates the JSON, and surfaces it on `LoadedSloppak.drum_tab`; the load stays permissive so a missing or malformed tab silently disables drums rather than failing the sloppak load. `/ws/highway/{filename}` gains two new message types — `drum_tab` (metadata + kit legend) and chunked `drum_hits` (500 hits per frame, same chunking as notes) — exposed to renderers via `bundle.drumTab`. `song_info` carries a `has_drum_tab` flag so viz pickers can auto-activate the drums highway regardless of which guitar arrangement is selected. `lib/gp2rs.py::convert_drum_track_to_drumtab` converts a Guitar Pro drum track to a `drum_tab.json` dict, preserving velocity verbatim, mapping hi-hat openness through the canonical piece-ids, and flagging flam / ghost / cymbal-choke articulations from GP effects. `lib/midi_import.py` gains `list_drum_tracks` + `convert_drum_track_from_midi` (channel-9 only) with heuristic flam-collapse (≤30 ms same-piece) and choke detection (cymbal note-off ≤120 ms). `docs/sloppak-spec.md` §5.3 promotes drum_tab from worked-example to canonical with the closed piece-id table and wire format. Sloppaks without a drum_tab are unaffected; legacy drums-as-guitar-notes sloppaks keep playing via the drums plugin's fallback decoder.
- **Loose folder support** — a directory containing an audio file + arrangement XMLs, with optional `manifest.json` and album art, is now discovered, indexed, and playable as a first-class library format alongside Sloppak. The scanner walks `DLC_DIR` for non-preview audio files and treats each parent directory that also contains XMLs as a loose song. Metadata follows a `manifest.json` → XML tags → folder-name priority chain (see `lib/loosefolder.py`). Songs are tagged `format: "loose"`, render an amber `FOLDER` badge in the library, and are filterable via the new "Folder" option in the format dropdown. Audio uses the shared vgmstream/`convert_wem` pipeline, cached under `AUDIO_CACHE_DIR`. The chart `<offset>` from the first non-vocals XML is now propagated to the frontend via `song_info.offset` and applied in `highway.setTime()` so loose folders authored against non-silence-padded audio stay in sync. Pairs with the companion `feedBack-plugin-loosefolder` plugin which adds an in-player Fix Sync UI for nudging and saving offset corrections.
- Highway note-state hook (#254). New `highway.setNoteStateProvider(fn)` lets a scorer plugin publish a per-note judgment (`'hit'` / `'active'` for a sustain currently held correctly / `'miss'`, or `{ state, alpha, color }`) so the renderer lights up the **gem itself** on a correct hit and keeps a sustain trail glowing while it's still being played right — instead of a separate overlay ring near the note. The built-in 2D highway honors it in `drawNote` / `drawSustains` / the chord-frame path (bright string colour + additive halo on hits, bright vs dim sustain trail, faint red wash on misses); the bundled 3D highway reads the same data via `bundle.getNoteState` (bright string-tinted outline + bright body + glowing sustain on hit/active, red outline + suppressed body on miss). Custom renderers opt in by calling `bundle.getNoteState(note, chartTime)`. note_detect registers the provider (and still owns its HUD / diagnostic miss markers / "currently detected" indicator); renderers that ignore the hook simply don't light gems. On a confirmed hit/active the renderers add a contained "sparkle/sizzle" on the note — the 2D highway: additive throbbing halo + flickering hot core + crackling spark lines (+ an expanding shockwave ring on a fresh strike) on the gem and a glowing/jittery sustain trail; the 3D highway: a few twinkling bright dots and short crackling arc segments hugging the note's rectangle (no bloom past the note), drawn on its overlay and projected through the camera so they ride the note. Also adds `highway.isDefaultRenderer()` so overlays that position with the 2D-highway helpers (`project` / `fretX`) can skip rendering when a custom renderer is active — fixes note_detect's miss markers appearing in random places over the 3D highway. New 3D-highway setting **Show note preview on the fretboard** (on by default) toggles the board-projection ghost — the translucent preview of the upcoming note on the fretboard surface. (Note: the companion change in the **note_detect plugin repo** turns its full-screen green/red edge flash off by default and adds a toggle to re-enable it — ships separately with note_detect, not in this feedBack release.)
- Diagnostic bundle export (#166). New "Export Diagnostics" + "Preview Bundle" buttons in Settings produce a single redacted zip combining server logs (tail of `LOG_FILE`), system info (Python/OS/version), hardware probe (CPU model + cores + freq + RAM, GPU via `nvidia-smi`/`rocm-smi`/`system_profiler`, container/Electron/bare runtime detection), full plugin inventory with git SHA + remote URL (read directly from `.git/HEAD` so it works in minimal runtime images without `git` installed) + orphan/failed-to-load detection, the browser console transcript (all levels: log/info/warn/error/debug + window.onerror + unhandledrejection, 500-entry ring buffer), browser hardware (WebGL/WebGPU adapter info, navigator + userAgentData), filtered localStorage, and per-plugin contributed diagnostics. Top-level `manifest.json` lists every file with its versioned schema id (`system.hardware.v1`, `client.console.v1`, etc.) so AI agents can dispatch by schema. Redaction is on by default: DLC paths, song filenames (`<song:HASH8>` stable per-bundle), IPv4/IPv6 addresses, bearer tokens, and `key=`/`token=` query strings are replaced. Plugins opt their backend diagnostics in via a new `diagnostics` manifest field (`server_files` allowlist mirroring `settings.server_files` semantics, plus an optional `callable: "<module>:<function>"` resolved lazily via `load_sibling`). Frontend plugins push diagnostics via `window.feedBack.diagnostics.contribute(plugin_id, payload)`. Three new endpoints: `POST /api/diagnostics/export`, `GET /api/diagnostics/preview`, `GET /api/diagnostics/hardware`. Full bundle format spec in `docs/diagnostics-bundle-spec.md`.
- Structured logging bootstrap (phase 1 of #155). Three new environment variables control server log output: `LOG_LEVEL` (default `INFO`), `LOG_FORMAT` (`text` for coloured console, `json` for one-JSON-object-per-line suitable for Loki/ELK/Promtail), and `LOG_FILE` (optional path, rotated at 10 MB with 5 backups). HTTP responses now include a `X-Request-ID` correlation header (via `CorrelationIdMiddleware`); the same request ID appears as `request_id` in structured log lines emitted via the stdlib `logging` / `structlog` APIs during that request.
- Structured logging migration completed (phase 2 of #155, #159, #242). The 42 `print()` calls and 6 `traceback.print_exc()` calls across `server.py` and `lib/` have been migrated to levelled `feedBack.*` loggers. Silent `except: pass` blocks in `lib/sloppak.py` and `lib/sloppak_convert.py` now surface as `log.warning` / `log.debug` with the exception attached. WebSocket handlers (`highway_ws`, `ws_retune`) bind a `ws_conn_id` contextvar at accept time so every log line within a session carries a connection ID. A CI grep guard in `.github/workflows/tests.yml` fails the build if either `print(` or `traceback.print_exc(` reappears in `server.py` or `lib/`.
- **Lyrics Karaoke plugin** — end-to-end karaoke setup for Sloppak songs in one workflow. The setup screen shows a per-song checklist (vocals stem / synced lyrics / per-syllable pitch) and a single "Build Karaoke" button that runs whatever's missing: Whisper alignment of pasted lyric text against the vocals stem, then `librosa.pyin` pitch extraction. Both artifacts persist inside the Sloppak (`lyrics.json`, `vocal_pitch.json`). In the player, a "Karaoke" toggle swaps the text-lyrics overlay for a horizontal pitch ribbon (one bar per syllable, vertically positioned by pitch, sweeping playhead).
- Settings export/import (#113). Two buttons on the Settings page bundle server config, browser localStorage, and opted-in plugin server-side files into a single versioned JSON file for backup, migration, or sharing a calibrated setup. Server-side import is all-or-nothing for safety-critical failures: phase-1 validates the entire bundle (schema, path-traversal, encoding) before any disk writes; phase-2 commits each file via temp+rename. Plugin-state mismatches between export and import are handled leniently: files referenced for a plugin that isn't loaded are skipped with a warning, files referenced for a plugin whose manifest no longer declares them are skipped with a warning, and localStorage is merged (not cleared) so first-run defaults from plugins installed after the export are preserved. Path-traversal, absolute paths, schema mismatch, and decode failures remain hard refusals. Plugins opt their server-side files in by declaring `settings.server_files` in `plugin.json` (list of relpaths under `CONFIG_DIR`; trailing `/` denotes a directory).
- Library filtering by parts present or missing (#129, #69). New right-side Filters drawer (single button next to the format/sort row, with active-filter count badge and dismissible chips below) lets you require or exclude arrangements (Lead/Rhythm/Bass/Combo), specific stems on Sloppaks (drums/bass/vocals/piano/other), lyrics, and tuning. Multi-select within an axis is OR (Lead OR Rhythm); cross-axis is AND. State persists across reloads. New endpoint `GET /api/library/tuning-names` returns distinct tunings present in the library, ordered by musical distance.
- Sort library by year (#128). Two new options in the sort dropdown: "Year (newest)" and "Year (oldest)". Songs without a year are pushed to the bottom for both directions.
- **`highway.getLyrics()` accessor.** `createHighway()` now exposes the parsed timed lyric syllables (`[{t, d, w}]`) via `getLyrics()`, mirroring `getBeats()`/`getSections()`, so overlay plugins can render karaoke without opening a second highway WebSocket. Pure accessor; no behavior change.
### Changed
- **Perf (3D highway, feedBack#226)**: pre-warm `plugins/highway_3d/screen.js` object pools at board init. Previously the pool factory grew lazily on first `.get()` past the high-water mark, allocating a fresh `T.Mesh` mid-rAF on dense 7/8-string charts and stalling those frames; the meshes were then permanently added to `noteG` (the pool only hides on `reset()`, never removes), bloating the scene graph for the rest of the session. Pre-warming spends the cost up front. Fold the per-frame `updateStringHighlights()` per-string loop with the post-call `mGlow`/`mAccentCore` emissive writes — one walk over the per-string scratch arrays instead of two. Replace `longestConsecutiveRun`'s per-call array allocations with a `{start, len}` index pair (trades two per-call sub-array allocations for one small 2-key object — net reduction in per-visible-chord allocation churn). Opt-in perf bench harness via `?h3dbench=1` URL param: `console.log` p50/p95/max for six update() segments every 5 seconds; when the URL flag is absent the mark helpers are bound to empty functions at renderer-instance init (each `createHighway()` panel re-checks the flag), so the hot-path calls are no-ops with negligible overhead (typically JIT-inlined).
- **License**: Relicensed to AGPL-3.0-only. Prior versions claimed MIT in the README, but the bundled desktop build statically links JUCE 8 (AGPL-3.0), so AGPL terms have effectively governed the desktop distribution since JUCE was added. AGPL-3.0-only is now the canonical license for the project — see [LICENSE](LICENSE) and [CONTRIBUTING.md](CONTRIBUTING.md) (DCO sign-off + plugin licensing policy). Bundled and vendored third-party code keeps its original license.
- Tuning sort is now ordered by musical distance from E Standard (#22) instead of alphabetical: E Standard first, then Drop D / F Standard at distance 2, then Eb Standard / F# Standard at distance 6, etc. Within a magnitude tier, down-tuned variants come before up-tuned, then alphabetical.
- Settings page restructured into separate "FeedBack" (core) and "Plugins" sections, with each plugin's settings rendered as a collapsible panel (collapsed by default). "Plugin Updates" moved into the Plugins section.
- **Lyrics Sync** is now a redirect stub. Its alignment + save endpoints moved into the new Lyrics Karaoke plugin alongside the pitch extraction. Existing nav entries and bookmarks land on a "moved" page that auto-redirects to the merged plugin.
### Security
- **Path traversal in archive extractors and library path resolution.** `lib/sloppak.py::_unpack_zip` and `server.py::_resolve_dlc_path` previously concatenated attacker-controlled entry names or filenames directly onto the extraction or library directory, so a crafted sloppak zip member or library filename with `..` segments, an absolute path, or backslash separators could write or read outside the intended directory. Any code path that unpacks a user-supplied archive (library upload, click-to-play, retune) or resolves a library path was reachable. Both locations now delegate to a new `lib/safepath.py::safe_join` helper that resolves each destination once and rejects entries that don't fall under the target directory; rejected entries are logged and skipped, the rest of the archive still extracts. The stem-split paths in `lib/sloppak_convert.py::split_stems` and `scripts/split_stems.py` previously called `ZipFile.extractall()` directly on user-supplied sloppaks; both now delegate to the same hardened `lib/sloppak.py::_unpack_zip` so every sloppak-unzip site in the codebase shares one containment guarantee. Tests in `tests/test_archive_traversal.py` and `tests/test_safepath.py` pin the contract for `../`, deep traversal, absolute paths, mixed `subdir/../../` forms, Windows-style separators, NUL bytes, names that resolve to the unpack root, and symlinked roots.
### Fixed
- E Standard retune now stays metadata-consistent across a chart's arrangement files (feedBack-plugin-notedetect#50). Previously the retune path could shift the audio and update manifests while leaving some arrangement metadata untouched, so `load_song()` later exposed the *original* tuning at runtime. `lib/retune.py` now updates every arrangement's tuning metadata consistently before applying the E Standard tuning, and raises on a partial update instead of silently packing split tuning metadata. EStd files generated before this fix should be re-converted so their metadata is consistent.
- Keyboard shortcut help now opens from the Player/3D Highway context when Linux/Electron reports Shift+Slash as `key="/"`, including while player controls such as the visualization picker are focused (#598).
- 3D Highway left-handed mode now has regression coverage for fret-axis mirroring, board rebuilds on runtime lefty changes, and mirrored camera state including the lookahead target and shoulder offset; the maintainer guide no longer claims the renderer ignores `bundle.lefty` (#321).
- Chord-level `fretHandMute` is now parsed into each note's `fret_hand_mute` (wire `fhm`) instead of being folded into `mute` (`mt`), matching `_parse_note` and preserving wire-format fidelity for both the template-expanded (synthetic-note) and explicit-`chordNote` paths. The 3D highway renders the fret-hand-mute X for `mt`*or*`fhm` notes, so the muted-chord overlay still shows. Also fixes the per-note fret-connector label vanishing exactly at the hit line (the fade now holds full opacity through `dt = 0`).
-`gp2rs` now respects the time-signature denominator when emitting ebeat subdivisions, fixing misaligned beat grids in 6/8 and other non-quarter-note meters.
- Settings dropdowns (Default Arrangement, Platform Filter) now persist immediately when changed. Previously a dropdown selection was only written to `config.json` when an unrelated "Save" button (Library Folder or Demucs Server) was clicked, so picking a default arrangement and navigating away silently discarded it. Both `<select>` controls now POST the single changed field on `change` via the partial-merge `/api/settings` endpoint, matching the auto-save behaviour of the A/V Sync Offset and mastery sliders. The text inputs (Library Folder Path, Demucs Server URL) keep their explicit Save buttons. Autosaves are sent through a single client-side queue (one request in flight at a time, in selection order), and the `POST /api/settings` handler now serializes its read-modify-write of `config.json` under a lock so concurrent partial updates can no longer overwrite each other and drop a key. The config write is also atomic (temp + rename) and `/api/settings/import` shares the same lock, so readers never observe a half-written file and a settings import can't race a concurrent partial save.
- Demucs stem split failing on Windows desktop with `OSError: Could not load this library: libtorchcodec_core4.dll` or `ImportError: TorchCodec is required for save_with_torchcodec`. The demucs subprocess now bootstraps a `torchaudio.save` → `soundfile.write` shim before importing demucs, sidestepping the torchcodec dependency entirely. The override stays in place across torchaudio versions — soundfile's WAV writes are behaviorally equivalent for demucs's float32 outputs.
- Splitscreen pop-out windows briefly flashed the library/song grid before showing the popped panel. A popup loads the full app (whose default screen, `#home`, is the library) and only swaps to the player once the splitscreen plugin loads; app init now detects `?ssFollower=1` and switches to the player screen up front, so the popup shows player chrome the whole time.
- Sloppak assembly dropped all tone data — affected sloppaks showed no signal chain in the Tones plugin and no tone-change markers on the highway. The assembler (`lib/sloppak_convert.py`) now lifts each arrangement's tones from the source chart via the new `lib/tones.py` helper and embeds them inline in the arrangement JSON under a `tones` key (`base`, `changes`, `definitions` — see `docs/sloppak-spec.md` §3.9). The highway WebSocket reads `base`/`changes` for sloppaks, and the Tones plugin (≥ 1.1.0) reads `definitions` to render the gear chain. Sloppaks built before this release carry no tone data and must be rebuilt from their source chart to gain it.
- Tab View (feedBack-plugin-tabview ≥ 3.0.1): the bottom row of tablature was permanently hidden behind the player controls bar (#336). The overlay reserved 60px at the *top* (clearing the transparent HUD) and extended all the way to the bottom of `#player`, where the opaque `#player-controls` (z-index 10) drew over the last row. The overlay now measures `#player-hud` and `#player-controls` dynamically and insets both edges; a `ResizeObserver` on the controls bar re-runs the inset when it wraps to a second row on narrow viewports.
- Tab View (feedBack-plugin-tabview ≥ 3.0.1): the cursor highlight led playback by roughly one beat (#336). alphaTab snaps `tickPosition` to the start of the *next* beat, so the cursor would race ahead by 500ms+ at typical tempos. The plugin now sends `tickPosition` one beat earlier so the snap lands on the current beat, and the highlight overlay tracks the bar cursor (`.at-cursor-bar`) instead of the next-beat cursor (`.at-cursor-beat`).
### Migration notes
- **Constitution amended to 1.1.0 (Principle II — Vanilla Frontend).** Prebuilt Tailwind (`static/tailwind.min.css`) is now codified as non-negotiable: no Play CDN / runtime CSS JIT anywhere, core or plugin. Plugin authors: a plugin that uses Tailwind classes not guaranteed in core — especially arbitrary values like `w-[37px]` — MUST ship its own compiled stylesheet via the new `styles` manifest key, built with `corePlugins.preflight = false`. Plugins that use only core-guaranteed utilities, or that ship no Tailwind at all, need no change. Contributors: after adding any Tailwind class to core or a bundled plugin, run `bash scripts/build-tailwind.sh` and commit the regenerated CSS, or the `tailwind-fresh` CI job fails.
- The library filters depend on three new columns (`stem_ids`, `tuning_name`, `tuning_sort_key`) that are populated as songs are scanned. If filters look empty after upgrading, run **Settings → Full Rescan** to repopulate; alternatively the periodic background rescan picks them up over time.
## [0.2.4] - 2026-04-22
### Added
- Version badge in navbar (`/api/version` endpoint + `VERSION` file)
-`CHANGELOG.md` and semantic versioning
- Step Mode plugin
-`gp2midi` improvements and expanded test coverage
- Note Detection plugin factory-pattern refactor with multi-instance/splitscreen support
- Per-panel note detection in Split Screen plugin with M/L/R channel routing for multi-input interfaces
### Fixed
-`SLOPPAK_CACHE_DIR` moved to `CONFIG_DIR` for AppImage compatibility
- Improved error message when plugin requirements fail to install
# Changelog
All notable changes to this project will be documented in this file.
The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
## [Unreleased]
### Added
- **Opt-in career venue packs (#122)** — higher-tier venue crowd media
(`club`, `arena`) is no longer bundled; the app downloads each pack on demand
from its release when you reach the venue (sha256-verified), keeping the
starter `bar` venue bundled for offline play. Trims ~678 MB from the desktop
download; an unpublished pack shows "coming soon" and plays on the standard
stage until its release lands.
- **Session-sync relay WebSocket — `/ws/sync/{session_id}` (#1030).** A
deliberately dumb JSON fan-out room: a text frame received from one client is
forwarded verbatim to every other client on the same session id; the server
interprets nothing (message schemas are owned by consumers). Rooms are created
on first join and garbage-collected when the last socket leaves — no history,
no replay, no persistence, so a host that crashes and rejoins the same id
resumes publishing to reconnecting subscribers with no server-side
coordination. Session ids are client-generated (`[A-Za-z0-9_-]{4,64}`); DoS
hygiene for a LAN-exposed port via frame-size (16 KB), per-room (16 sockets),
`role="dialog"` + `aria-modal` with focus moved to the close button on open
and restored on close), and a corrupt stored seen-badges value (e.g. a stray
`"null"`) can no longer throw on every passport refresh.
- **The packaged desktop app could not start (`ModuleNotFoundError: No module named
'appstate'`).** feedback-desktop's `scripts/bundle-slopsmith.sh` copies a *hardcoded
list* of core files into the app bundle — `server.py`, `VERSION`, `lib/`, `data/`,
`static/`, `plugins/__init__.py`. The root-level `appstate.py` and `routers/` added in
R3 shipped correctly in Docker and passed every test, and were then silently dropped
from the packaged app, which died at startup. Both now live under **`lib/`** — the one
core directory the Dockerfile (`COPY lib/`), `docker-compose.yml`, and the desktop
bundler (`cp -r lib`) all copy wholesale, and that all three put on `sys.path` (on
Windows via the embeddable-Python `._pth`, where `PYTHONPATH` is ignored). This needs no
change in feedback-desktop and no new release to take effect. Placing them there is also
correct under Principle V: with the injection seam, `appstate.py` constructs nothing and
does no import-time IO, and a route module only builds an `APIRouter`. The
`Dockerfile` / `.dockerignore` / `docker-compose.yml` entries added for the root layout
are reverted. New `tests/test_packaging.py` walks `server.py`'s module-level imports and
fails if any first-party module resolves outside a directory the packagers copy, so the
next root-level module can't ship broken.
### Added
- **Perf harness now measures 2D-highway frame time (R3c gate).** `scripts/perf-baseline.mjs` gains a `--song` mode that reports per-frame draw-cost p50/p95/p99 (draw-tagged via `highway.addDrawHook`), the metric that gates the `highway.js` split. Maintainer/CI-only; baseline recorded in `docs/perf-baseline.md`.
- **`routers/` — extracting `server.py`'s route layer, cheapest-first (R3).** Each PR moves a cohesive route group into a `fastapi.APIRouter` under `lib/routers/`, mounted with `app.include_router(...)` at its original site (FastAPI matches in registration order; the full route table stays byte-identical). Bodies are verbatim — only the decorator receiver (`@app` → `@router`) and singleton reads (`meta_db` → `appstate.meta_db`, resolved at call time) change. So far: `audio_effects` (5), `artist_aliases` (5), `loops` (3), `playlists` (12 + covers), `ws_highway` (the 902-line highway chart WebSocket), `chart` (split/unsplit/work/fileinfo — unblocked by the DLC-path substrate), `library_extras`, `wanted`, `shop`, `progression`, `profile`, `stats` (the `/api/stats/{path}` catch-all stays registered last so it can't shadow `/recent` `/best` `/top`), `version` (`/api/version`; VERSION-file lookup adjusted for the router subdir depth), `art` (the `/api/song/{f}/art*` serve/cover-search/candidates/upload/url + `/api/art/{f}/override` routes; the shared `_song_pack_art_exists`/`_art_override_paths`/`_art_safe_name` helpers stay in `server.py` for the song/delete routes and are reached through the `appstate` seam, the CAA/release transport as `enrichment.X`), and `settings` (`GET`/`POST /api/settings`, `/reset`, and the two-phase atomic export/import bundle `/api/settings/export|import`; the shared `_default_settings` builder stays in `server.py` and is reached through the `appstate` seam), and `song` (upload/delete + the metadata write-back, user-meta, overrides, gap-fill, and per-song info routes; the scan/ingest helpers stay in `server.py` and are reached through new `appstate` seams — `kick_scan`, `invalidate_song_caches`, `stat_for_cache`, and a `scan_status()` getter — the `get_song_info` catch-all mounts after the art routes so it can't shadow them), and `library` + collections (the provider list/art/sync endpoints, the library query surface, and collection CRUD → `lib/routers/library.py`; the `LibraryProviderRegistry`/`LocalLibraryProvider`/`SmartCollectionProvider` classes + shared query/collection helpers move to `lib/library_registry.py`, and the registry instance + local provider ride the `appstate` seam — server.py still constructs the singleton and exposes `register_library_provider`/`unregister_library_provider` to plugins via `plugin_context` unchanged), and the `enrichment` route handlers (`/api/enrichment/*`: status, kick/cancel, per-song state, the Match-Review queue, and AcoustID identify → `lib/routers/enrichment.py`; the engine already lives in `lib/enrichment.py` and is reached as `enrichment.X`), and `media` (the file-serving routes — song audio `/audio/{f}`, the local-audio-path resolver `/api/audio-local-path`, and raw sloppak-member serving `/api/sloppak/{f}/file/{rel}` → `lib/routers/media.py`; the cache/static path seams were already in `appstate`), and `artist` (the artist page + external-links payload `/api/artist/{name}/page|links|links/refresh` → `lib/routers/artist.py`; MB link enrichment reached as `enrichment.X`), and `diagnostics` (`/api/diagnostics/export|preview|hardware`; the plugins-root lookup adjusted for the router subdir depth, `_running_version` reached through the `appstate` seam, pure payload-cap helpers re-exported for the `server._diag_*` tests), and `tunings` (`/api/tunings`; the pure `config.json` reader moved to `lib/appconfig.py`, the tuning-provider registry read through the `appstate` seam so plugin-contributed tunings still merge). The DLC library-path resolution (`_get_dlc_dir`, pure `_resolve_dlc_path`) moved to `lib/dlc_paths.py`, reading paths through the seam; `config_dir`/`dlc_dir`/`dlc_dir_env` now ride the `appstate` seam (env-derived, so the pop-and-reimport fixtures reconfigure it for free), and the shared request-field sanitizer `_clean_str` moved to `lib/reqfields.py`. The next cut is picked by a dependency-closure scan that ranks groups by how many `monkeypatch.setattr(server, …)` targets they'd drag along.
- **`routers/` — the first extracted route module (R3).** The five audio-effects mapping
endpoints move out of `server.py` into `lib/routers/audio_effects.py` as a
`fastapi.APIRouter`, mounted with `app.include_router(...)` **at the point in the file
where they used to be defined** — FastAPI matches routes in registration order, so the
mount site preserves it. Verified: the full 143-route table (paths, methods, *and*
order) is byte-for-byte identical to `main`. Bodies are verbatim; the only edits are
the decorator receiver (`@app.get` → `@router.get`) and the singleton read
(`audio_effect_mappings` → `appstate.audio_effect_mappings`, a module attribute
resolved at call time). This proves the seam from #833 under a real consumer, including
the second slot. The `_demo_mode_guard` middleware still blocks all four moved write
routes with 403, and `Query(...)` validation still 422s — both checked against a running
server. `server.py`: **9,445 → 9,386 lines**.
- **`lib/enrichment.py` — the metadata-enrichment subsystem leaves `server.py` (R3, move-only).** MusicBrainz / Cover-Art-Archive / AcoustID transport, the match-scorer glue, and the background enrichment worker (~930 lines, 61 defs) move out as one cohesive unit. Bodies are verbatim; the only changes are seam reads — `meta_db`/`config_dir`/`sloppak_cache_dir`/`art_cache_dir` and the two shared art helpers (`song_pack_art_exists`, `art_override_paths`, which stay in `server.py` for the art/delete routes) are reached through `appstate` at call time, and the User-Agent VERSION lookup is corrected for the module's new depth. `server.py` drives the worker through the module (`import enrichment`; the routes + scan lifecycle call `enrichment.X`); tests that faked the network on `server` now patch the same names on `enrichment` (module attribute resolved at call time, so one `setattr` reaches both the routes and the worker's internal callers). Acyclic — `enrichment` imports no `server`. Route table byte-identical; full suite green. `server.py`: 6,917 → 5,988.
- **`appstate.py` — the router seam (R3).** Route modules moving out of `server.py`
need `meta_db` and friends but must not `import server`, or the import graph goes
circular the moment `server` imports them back. So `server.py` keeps *constructing*
its singletons and now **injects** them once — `appstate.configure(meta_db=…,
audio_effect_mappings=…)` — and a router reads them back as module attributes at call
time (`import appstate; appstate.meta_db.…`). This is the Python analogue of the
frontend refactor's injected `configureX({…})` seams and of the plugin
so the module does no IO at import. The singleton stays in `server.py`; no route,
no test and no `monkeypatch.setattr(server, …)` target moves. `server.py`:
**9,705 → 9,433 lines**.
- **`MetadataDB` moved out of `server.py` into `lib/metadata_db.py` (R3, move-only).**
The library metadata cache — the `MetadataDB` class (4,018 lines) plus the query
helpers it owns (keyset paging cursors, the tuning grouping key, smart-arrangement
naming, tag normalisation, the startup DB-restore swap) — now lives in its own flat
`lib/` module. `server.py` drops from **14,037 → 9,705 lines** and keeps the
`meta_db` singleton, so `server.meta_db` and `server.app` resolve exactly as before
and every route is untouched. The only non-verbatim change is the seam that lets the
class leave `server.py`: `MetadataDB.__init__` now takes `config_dir` explicitly
(`meta_db = MetadataDB(CONFIG_DIR)`) instead of reading the module-level `CONFIG_DIR`,
which also means `lib/metadata_db.py` performs no IO at import (Principle V). Logging
still goes through the `feedBack.server` logger, so existing log filters and `caplog`
assertions resolve to the same logger object. `tests/test_settings_export_library_db.py`
now imports `_apply_pending_db_restore` from `metadata_db` (the test moved with its
subject); no other test changed. Every moved block is byte-identical to its
`server.py` original.
### Added
- **Plugins can ship an ES-module `src/` tree (module-migration rails, R0).** The host gains three things so a plugin can move off a single global-scope `screen.js` IIFE onto native ES modules with **no build step**: (1) a new sandboxed `GET /api/plugins/{id}/src/{path}` route that serves a plugin's `src/` source subtree, containment-checked by the same `safe_join` guard as `assets/` (traversal/absolute/NUL → 404); (2) the live-edit cache contract — `Cache-Control: no-cache` + a weak mtime/size `ETag` + `If-None-Match`→`304` — applied to `src/`, `screen.js`, and `assets/` (previously `screen.js` sent no cache headers and `assets/` emitted an ETag but never revalidated), so an edited module reloads on refresh while unchanged ones `304`; and (3) `scriptType`/`minHost` passthrough from `plugin.json` to `/api/plugins`, with the loader injecting a plugin that declares `"scriptType":"module"` as `<script type="module">` (its screen.js becomes `import './src/main.js'`). A `<script type=module>` fires its load event only after its whole static-import graph evaluates, preserving the loader's completion-by-`onload` + `_loadingPluginId` contract. Classic plugins are unaffected; `minHost` is passthrough-only for now (enforcement deferred). Tests: `tests/test_plugin_src_route.py` (serve/media-type/traversal/304/no-stale-304/screen.js+assets conditional), `tests/js/plugin_loader_script_type.test.js` (guarded module injection).
- **Module-migration governance & rails (R0).** Constitution amended to **v1.2.0**: Principle II now names native ES modules as a first-class, *build-free* extension mechanism (the `scriptType:"module"` load path, both plugins and — over time — core's `static/js/`), keeping the no-bundler/no-transpiler/source-served rule intact; Operating Constraints gains a "Module load contract" clause (a `<script type=module>` load event awaits the whole static-import graph, so completion-by-`onload` is preserved; per-visit re-init comes from the `screen:changed` event, not screen.js re-execution). Mirrored into `CLAUDE.md`. New `docs/plugin-modules.md` (the migration playbook — layering, import-time purity, `import.meta.url` assets, the ETag live-edit loop) and `docs/size-exemptions.md` (the signed 1,500-line size-norm register; Byron signs core/bundled rows, Christian the authored virtuoso row). Adds a **maintainer/CI-only** ESLint gate (`eslint.config.js` + a `lint` CI job): `max-lines` warns at 1,500 as a non-blocking ratchet (ceilings for exempt files mirror the register), and `import-x/no-unresolved` + `import-x/no-cycle` hard-error on ES-module graphs — dormant until module code lands, never on the serve/Docker path.
- **Perf-baseline harness (R0).** `scripts/perf-baseline.mjs` (maintainer-only, Playwright-driven) captures server p50/p95/p99 latency, cold boot-to-interactive, JS-heap after an idle soak, and the injected plugin-script shape (how many are `type=module`), so every refactor phase can be checked for "screen-entry and frame-time no worse." Methodology + the R0 baseline live in `docs/perf-baseline.md`; playback frame-time and chart-loaded screen-entry rows need a seeded library and are re-taken per environment.
- **Sort and filter the library by your personal difficulty rating — now visible at a glance, not just in the edit drawer.** `song_user_meta.user_difficulty` (the 1–5 planning rating, settable manually or seeded by a plugin like the community `difficulty_tagger`) already existed but was only readable by opening a song's per-song details drawer. The library API gains `sort=difficulty` / `sort=difficulty-desc` — a correlated subquery over `song_user_meta`, following the same unrated-rows-sort-to-the-bottom-in-both-directions pattern as the existing `mastery` sort — and library cards now show the rating as a `◆N` badge (v2 grid/tree views and the v3 grid alike), next to the tuning and lyrics badges. The classic tree view's `query_artists` batch-attaches `user_difficulty` the same way `query_page` already did for the grid, so the badge actually renders there too instead of staying dark. Tests: `tests/test_library_filters.py::test_difficulty_sort_pushes_unrated_to_bottom`, `tests/test_library_filters.py::test_tree_view_songs_carry_user_difficulty`.
- **`lib/midi_import.py`: `convert_midi_tempo_map` — MIDI imports can finally carry
their bars.** The keys/drums note converters always computed a tempo-aware
tick→seconds map internally (to bake note times to absolute seconds) and then threw
it away — and never read `time_signature` meta at all — so every MIDI import landed
with no measures and an implied 4/4 regardless of what the file said. The new helper
extracts the whole grid: `tempos` (`{time, bpm}`), `time_signatures` (`{time,
ts:[num,den]}`, the song-timeline sidecar shape), and a full `beats` grid on the
editor's row shape (numbered downbeats with a `den` hint, `-1` interior beats,
eighth-note rows in 6/8 etc.). Event scope mirrors the existing tick map — SMF
type 0/1 merge meta across tracks, type 2 reads only the chosen track (independent
timelines must never share a grid); mid-bar signature events apply at the next bar
boundary; times are computed from absolute ticks through the cumulative tempo table
and rounded once at emit, so rounding error never accumulates with song length.
Consumed by the editor's upcoming multitrack MIDI import (tempo-seed dialog). Tests:
`tests/test_midi_tempo_map.py`.
### Fixed
- **Tuner: opening the player screen no longer throws `NotFoundError` and aborts the player render (feedBack#800).** `injectPlayerButton()` anchored the injected Tuner button with `controls.querySelector('button:last-child')`, which — unlike a `:scope`-scoped query — can match a **nested** button that is not a direct child of `#player-controls`. `controls.insertBefore(btn, nestedButton)` then throws `NotFoundError` (the reference node must be a direct child), and because the injection runs from the tuner's `screen:changed` → player handler, the throw propagated out of the player-screen transition and stalled its render (surfaced by a headless render of a notation arrangement; the v3 path was already safe via the plugin-control slot, only the classic path had the bad anchor). The anchor is now `:scope > button:last-of-type` (a direct child only) with a `parentNode === controls` guard before `insertBefore`, falling back to `appendChild`. `plugins/tuner` → 1.3.4. Tests: `tests/plugins/tuner/js/inject_player_button.test.js` (nested-last-button repro, direct-child insert, no-button append, idempotency, v3 slot path).
- **Auto-sync: DTW step constraint — riff-based songs no longer produce garbage sync points.** `librosa.sequence.dtw`'s default step pattern allows unbounded horizontal/vertical path runs, and on music with long self-similar chroma stretches (riff-driven stoner/doom, drone sections) the flat cost surface let the warping path collapse — minutes of score mapped onto a single audio frame, so the per-bar warp imported charts wildly out of sync while reporting success (observed on a real 138 BPM tab: effective displayed tempo 159 BPM, three sync points sharing one audio timestamp). `_dtw_align` now uses the standard music-sync slope-constrained step pattern (`[[1,1],[1,2],[2,1]]`, local tempo ratio bounded to 0.5x–2x), which makes the degenerate path impossible, with a fallback to unconstrained steps when the global length ratio makes the constrained pattern infeasible (e.g. a tab aligned against a full-concert video). Validated on the failing song: coarse points track the recording 1:1, refined downbeats land on onset peaks at 3.3x background energy.
### Added
- **3D Keys Highway: key layout modes, lane-color opacity & octave lines.** A new **Highway layout** settings section rebuilds how sharps/flats and lanes draw on the 3D piano highway. **Sharps & flats layout** (`keys3d_bg_sharpMode`) picks between **floating** (the original raised-sharp look), **flat** (one plane, zero-overlap piano-shaped tiled lanes with the naturals evened out), and **realistic** (one plane, bars sized like the physical keys) — default **realistic**; the geometry lives in pure, unit-tested `laneSpanFlat()`/`laneSpanReal()` helpers. **Lane color opacity** (`keys3d_bg_laneOpacity`, 0–1, default 0) fades the pitch-class lane tint from full vivid color down to a dark floor with guide lines only at the key-block boundaries (E→F and each octave); the lane strips, per-lane separators and block lines crossfade with the value. **Octave separators** (`keys3d_bg_octaveGaps`, default on) and **Octave line contrast** (`keys3d_bg_octaveContrast`, 0–1) control the B→C octave divider, which auto-shifts from a dark to a bright layer as lane opacity fades. Settings re-read on init and apply on the next chart build. `plugins/keys_highway_3d` → 0.2.0. Tests: `plugins/keys_highway_3d/tests/fx_settings.test.js` (new defaults, sharp-mode setting, lane-geometry tiling/evening for flat, uniform/overlap for realistic, and an active-range boundary case where a white key's edge stays untrimmed when its neighboring sharp falls outside the active range).
- **Unmapped-percussion capture now records velocities alongside times.** Both drum converters' opt-in `out_unmapped` reporting (`lib/midi_import.py` `convert_drum_track_from_midi`, `lib/gp2rs.py` `convert_drum_track_to_drumtab`) gain an index-aligned `velocities` list next to `times`, carrying each dropped note's real dynamics (MIDI velocity verbatim; GP velocity with the same 1–127 gate as mapped hits, falling back to the 100 import default). This lets a hand-mapping UI (the editor's unmapped-notes dialog) restore mapped notes at their source dynamics instead of flattening everything to `v:100`. The GP path's chronological sort now reorders times and velocities in lockstep so multi-voice measures can't silently reassign dynamics. Additive — callers that ignore the new key are unaffected. Tests: `tests/test_midi_import_drums.py`, `tests/test_gp2rs_drums.py`.
- **Handedness (left-handed) is now a first-class choice in the instrument selector — and surfaced during onboarding.** Left-handed players could already mirror the highway, but only via a buried Settings toggle they had to find *after* setup — so a lefty hit the tour, the tuner and calibration all right-handed first. The v3 instrument badge popover now has a **Handedness: Right / Left** row alongside Instrument / Strings / Tuning (all player-orientation choices), writing the same `lefty` preference (`highway.setLefty` when a live highway exists, else the `lefty` localStorage key it reads on init; the Settings checkbox stays in sync). The first-run tour's "Choose your instrument" step — which runs **before** the tuner/audio-calibration steps — now calls it out so lefties flip it up front. Frontend-only, additive: `static/v3/badges.js`, `static/v3/onboarding-tour.js`. Tests: `tests/js/badges_handedness.test.js`.
- **"Colorblind (deuteranope)" highway string-color preset.** Adds a one-click preset to the shared "Highway String Colors" picker, sitting next to the existing Okabe–Ito "Colorblind-friendly" preset — contributed by a deuteranopic player who found the Okabe–Ito set still hard to separate. It retunes the six main strings (red / yellow-green / blue / orange / teal / deep-purple) and keeps that set's 7/8-string colors, and applies to **both** the 2D and 3D highways via the shared picker. Frontend-only, additive: `static/app.js` (`HWC_PRESETS`).
- **`lib/gp_autosync.py`: piecewise time-warp helpers + a working `refine_sync()`.** `auto_sync()` has always computed per-bar sync points (DTW), but consumers could only apply the scalar bar-1 `audio_offset`, so any tempo difference between the recording and the tab's authored tempo accumulated audibly over the song. New librosa-free helpers expose the full mapping: `bar_start_times(gp_path)` (per-bar score times on the same axis as the sync points — GPIF bar-resolution map for `.gp`/`.gpx`, per-tick integration for GP3/4/5), `build_warp_anchors(points, bar_starts)` (strictly-monotonic `(score, audio)` anchor pairs), `warp_time(t, anchors)` (piecewise-linear map with edge-slope extrapolation for count-ins/tails), `warp_song_times(song, warp)` (retimes a `lib.song.Song` in place: notes + sustains, chords, beats, sections, anchors, handshapes, per-phrase difficulty levels, tone changes, tempo overrides), and `gp_has_expandable_repeats(gp_path)` (detects GP3/4/5 repeat/volta/direction markup whose playback expansion the as-written sync points cannot map — callers fall back to offset-only sync). Also implements `refine_sync()`, which the editor plugin's refine-sync endpoint has imported since the snapshot but which never existed in core (the Refine button 500'd): it densifies the coarse DTW points to every Nth bar and re-times each with a local onset phase sweep (sweep radius clamped under half a beat so periodic material can't lock a full beat off; short scoring grid + median residual snap). Synthetic click-track validation: ~13ms mean / ~40ms max error from ±180ms coarse input across 117–123 BPM recordings of a 120 BPM tab. Tests: `tests/test_gp_autosync_warp.py`.
- **Playlist shuffle.** The v3 playlist detail page gains a crossing-arrows shuffle toggle next to Play all / Play album. When on, `playQueue.start` Fisher-Yates-shuffles the queue once at start (on a copy — the stored playlist order is untouched), swapping any per-slot album arrangements in lockstep so each slot keeps its pinned arrangement. The preference is global and persists in `localStorage` (`v3PlaylistShuffle`). Tests: `tests/js/play_queue_shuffle.test.js`.
### Changed
- **Player frame-time hotspots removed (trace-backed) + weak-hardware hardening.** A Chrome performance trace of a 3D-highway session surfaced two core per-frame layout-thrash sources, now fixed: the highway's visibility check read `canvas.offsetParent` every rAF frame (forces style/layout recalc — now sampled every 10th frame with a cached value, force-refreshed on init/canvas-replace/resize/override-clear), and the v3 player chrome loop called `matches(':hover')` per frame and unconditionally rewrote the Up-Next pill's `textContent`/bar width at 6 Hz (now hover-tracked via mouseenter/mouseleave, DOM writes only on value change, progress bar moved from `width` to compositor-only `scaleX`). The 3D highway pre-warms shader programs (`ren.compile`) and deterministic label textures at init — and chart-dependent chord/section label textures on first draw — so first-appearance shader-compile/texture-upload frame spikes move into the load spinner. For weaker hardware: the per-frame renderer bundle is now a single reused object instead of a fresh ~35-field allocation per frame (object identity is stable and meaningless; array fields still swap reference on chart changes), custom viz get `bundle.lowerBoundT`/`bundle.lowerBoundTime` binary-search helpers for visible-window culling, the default 2D highway's beat lines no longer scan every beat in the song per frame, and the 3D highway stops reading `localStorage` per frame (1 Hz poll) and caches its lyrics text-measurement layout per displayed line instead of re-measuring every syllable every frame. A second, throttled-CPU trace pass additionally removed: shader-program re-resolution churn from label texture swaps (`material.needsUpdate` is now only set on a null↔texture transition — swapping between two cached label textures never changes the compiled program), the 3D highway's per-frame `getBoundingClientRect` layout read in its canvas-size self-check (now every 10th frame, still immediate on backing-store change), and the core 60 Hz HUD clock rewriting `textContent` on every tick (now write-on-change, ~1/s). The dominant residual — steady `getParameters` shader-program re-resolution (~4% of throttled main thread) — turned out to be Three r158+'s transparent-DoubleSide two-pass rendering, which sets `material.needsUpdate` twice per object per frame; all 18 of the 3D highway's transparent DoubleSide materials are flat unlit quads (labels, rails, chord frames, lanes), so they now declare `forceSinglePass: true`, eliminating the recompile churn and halving those objects' draw calls.
### Fixed
- **`playback.loop-api` bridge no longer fires dozens of times per second.** Every `window.feedBack.getLoop()` call recorded a full bridge hit — compat-shim bookkeeping, a `playback:bridge-hit` event, and a diagnostics snapshot rebuild + stringify — so a plugin polling loop state from a HUD tick (note_detect at ~30 Hz) flooded the capability inspector and burned main-thread time even with no song playing. `_recordPlaybackBridge` now throttles per bridge/surface (5 s window): bridge hits are a "surface still in use" signal, not a call counter. The manual A/B loop buttons (`setLoopEnd`) also now emit the same `loop-set` transport event as `setLoop()`, so plugins can react to loop changes via `playback:loop-set` / `playback:loop-cleared` events instead of polling `getLoop()`.
- **3D Highway: recover from a WebGL context loss instead of crashing on alt-tab.** Switching the active window / alt-tabbing away from the app (most often on Windows) can trigger a GPU context reset; the 3D highway's WebGL renderer had **no `webglcontextlost` handler**, so a lost context was left to escalate into a render-process crash — matching the intermittent "randomly crashes when I change windows" desktop reports. The renderer now binds `webglcontextlost`/`webglcontextrestored` on its own WebGL canvas (`ren.domElement`): the loss is `preventDefault()`'d so the browser keeps the context restorable, `draw()` bails while the context is down so no GL work runs on a dead context, and on restore the viewport is re-applied and rendering resumes (Three re-uploads scene resources on the next frame). Listeners are torn down with the renderer. `plugins/highway_3d` → 3.31.3. Tests: `tests/js/highway_3d_context_loss.test.js`. (The sibling `keys_highway_3d` / `drum_highway_3d` renderers share the same gap — tracked as a follow-up in their repos.)
- **Guitar Pro 6 (`.gpx`) import no longer fails on every real file.** The GPX BCFS container reader (`lib/gp2rs_gpx.py`) rejected any file whose final sector wasn't a full `0x1000` block — but a real `.gpx`'s BCFZ-declared decompressed size isn't sector-aligned, so the last (small) container file always lands in a partial trailing sector. The bounds check *raised* `GPX BCFS sector pointer out of range (malformed file)` instead of clamping the tail read, so `_load_gpif` threw before `score.gpif` could be extracted and **no GP6 file could be imported into the song editor** (both real test files failed identically — this wasn't file-specific). GP7/GP8 `.gp` files were unaffected — they take the ZIP path, not BCFS, which is why prior GP-import work didn't surface it. The reader now **clamps the final sector read to the buffer end** (the per-file size field trims the padding anyway), matching canonical GPX readers (alphaTab / PyGuitarPro); a sector whose *start* is past the end still raises, preserving the malformed-file guard. Verified against two real GP6 files — both now unpack to valid GPIF with all tracks. Tests: `tests/test_gp2rs_gpx.py` (partial-final-sector round-trip, multi-file container, sector-aligned baseline, and the preserved out-of-range guard).
- **v3 Songs grid: fixed the scroll stutter that "skips every so many scrolls," up or down.** The virtualized grid rebuilt its **entire** visible window (`grid.innerHTML = …` + a full `wireCards` pass) every time it slid by one row, so each row-boundary crossing was a heavy synchronous frame that stalled the main thread and buffered held-arrow key-repeats into a visible lurch (a tester's "super fast for a second then slowed down") at fixed scroll offsets — in **both directions and regardless of whether the page was already loaded** (the cost was DOM teardown, not fetching, which is why scrolling back up over cached songs hitched too). `renderWindow()` now **reconciles the window in place**: it reuses the card nodes that stay on-screen and builds only the row that enters/leaves (~6 nodes per slide instead of ~60), keyed by absolute index with a real-vs-skeleton + select-mode signature so hole-fills (after a page fetch) and select-mode toggles still rebuild exactly the nodes that changed. `wireCards`'s `data-wired` guard then wires only the freshly-built nodes, so per-slide listener churn drops with it. Follow-up to the stage-2 virtualized grid (got-feedback/feedBack#636 item 3). Frontend-only: `static/v3/songs.js`. Tests: `tests/js/v3_songs_window_recycle.test.js` (window stays `[start,end)` contiguous + in-window node identity reused across a down-then-up scroll; select-mode toggle and rail-seek jump rebuild correctly).
- **Starter content seeds again (and now ships The Adicts' "Ode to Joy").** `_BUILTIN_STARTER_SOURCES` still listed `beethoven-ode_to_joy.feedpak` after that pack was deleted, and never wired up its replacement `the_adicts-ode-to-joy_vst_cover.feedpak` that landed on disk. The listed-but-missing file made the all-present gate never fire, so **no** starter songs seeded on first run. Synced the manifest to what's on disk (Für Elise, Star Spangled Banner, The Adicts' Ode to Joy). Tests: `tests/test_builtin_starter_seed.py` (the present/unlisted guards were red on `main`).
- **Edit Metadata now writes into `.feedpak` files, not just legacy `.sloppak` ones.** `lib/songmeta.py`'s suffix gate predated the format rename — core reads both suffixes everywhere else (`sloppak.SONG_EXTS`), but the metadata writer only dispatched on `.sloppak`, so editing a zip-form `.feedpak`'s title/artist/album/year silently fell back to a DB-only update. That looked fine until the next **full library rescan** re-derived metadata from the file and reverted the edit (directory-form packages were unaffected — they dispatch on manifest presence, not suffix). The gate now accepts both package suffixes. Tests: `tests/test_songmeta.py` `TestWriteSongMetadata` (both zip suffixes, mixed-case suffix, directory form, unknown-suffix fallback).
- **3D Drum & Keys highways now re-frame on fullscreen/layout drift under splitscreen.** The guitar/bass `highway_3d` self-detects when its panel canvas changes size and re-runs `applySize()` every frame, because the splitscreen host overrides `hw.resize` and never calls `renderer.resize()`. The drum and keys highways lacked that fallback — they only re-framed when the host explicitly called `resize(w, h)` — so their panels stayed framed for the pre-fullscreen size while the guitar/bass panels adapted (visible as a too-small, off-center highway after maximizing a split-screen session). Both draw loops now port `highway_3d`'s per-frame drift check: they re-apply on backing-store change (`canvas.width/height`) AND on CSS-box drift (`clientWidth/clientHeight` vs the last applied logical size, throttled to every 10th frame), and reset the tracking in `destroy()` so a reused instance re-frames on the next song. `plugins/drum_highway_3d` → 0.3.1, `plugins/keys_highway_3d` → 0.1.1. Tests: `tests/js/drum_keys_highway_3d_resize_reframe.test.js`.
- **Tuner: finished the "remove unused settings" cleanup and fixed the sidebar panel position.** The Floating Button and Tuning Visibility settings sections were removed, but their config was still live: `disabledTunings` still filtered the tuner menu (with no UI left to re-enable a hidden tuning — a one-way trap) and `showFloatingButton` still gated the floating launcher. Both are now fully retired — the enforcement paths in `plugins/tuner/screen.js`/`utils/ui.js` and the persistence in `plugins/tuner/routes.py` are gone (and `routes.py` strips the retired keys on write, so stale values are purged). The tuner panel opened from the v3 sidebar Plugins rail popover now anchors beside it via the host's stable plugin-control slot API (falling back to the popover id), is **clamped to the viewport** so it can't open off the right/bottom edge on narrow/short windows, and re-anchors on window resize. `plugins/tuner` → 1.3.3.
### Added
- **Host theme read surface — `window.feedBack.theme` + always-present `--fbv-*` tokens — so a plugin feature can render correctly under any theme instead of binding to whichever one the developer happened to see.** The cosmetics applier (`static/v3/theme-core.js`) previously only *applied* themes and emitted `--fbv-*` vars **only while a theme was equipped** (nothing for a plugin to read in the default, un-themed state) with no read/capability API — so plugins reinvented their own theming and a new visual *device* (a glow, a gradient) silently bound to one look. It now: (1) emits the default `fb` palette as **always-present `--fbv-*` on `:root`** (additive — the un-themed look is unchanged; the `fb-*` utilities still use their compiled defaults; this only hands plugins a stable host token to read + derive surfaces from), plus two keystone ROLES the palette lacked — **`on-accent`** (a foreground legible *on* the accent fill — the missing piece behind white-on-accent contrast bugs) and **`focus-ring`**; (2) adds **`window.feedBack.theme`** — `get()` → `{id, isThemed, tokens}`, **`capabilities()`** → `{glow, gradients, motion}` (the device-affordance signal a feature reads to choose a glow vs. a solid device; recolor-only themes report defaults, a theme may opt out via a `capabilities` block in its payload, and `motion` is additionally reduced-motion-gated), and **`prefersReducedMotion()`** (one central matchMedia wrapper); and (3) emits a normalized **`theme:changed`** event (`{id, isThemed, tokens, capabilities}`) from the single `apply()` chokepoint. All additive + feature-detected (the apply side stays on `window.v3Theme`; the read surface is attached defensively so it survives the `feedBack` bus being (re)built by `capabilities.js` regardless of load order). First slice of the host theme contract (got-feedback/feedBack#644) — the framework fix so a plugin UI feature can't accidentally carve itself into a single theme; see `docs/host-theme-contract.md`. Verified by a headless render (apply/unequip intact, defaults present + restored, capability opt-out honored, event payload correct). Tests: `tests/js/v3_theme_read_api.test.js`.
- **3D Keys Highway: audio-reactive background ambience + score effects (parity slice 4 — completes the keys side of the visual-parity epic).** The gradient sky behind the highway gains the guitar's **background ambience styles** — drifting **Particles**, pitch-class-colored pulsing **Stage lights**, wireframe **Geometric** shapes, or Off — driven by the shared audio-analyser bridge (stems-first on sloppaks, one-shot `#audio` fallback; bass/mid/treble bands, 5 ms cache) with an **Ambience intensity** slider and an **Audio-reactive** toggle. And the score talks back: a **score-FX overlay** canvas draws rising **“+1” pops** off each scored key, an **expanding ring every 10-combo tier**, **milestone bursts** at 25/50/100 streak, and a brief **red flicker** when a 3+ streak breaks (wrong notes and swept misses both count). All settings-gated and on by default (`keys3d_bg_*`), pooled, cleared when idle, torn down with the scene. Butterchurn/image/video remain out of scope. The `#audio` analyser tap is also **shared across visualizers** now (`window.__feedBackAudioTap` — whoever taps first publishes, everyone else adopts, `highway_3d` included), so switching between or splitting the guitar/drum/keys highways can't strand a permanently non-reactive backdrop, and the tap is never created before the page has user activation (a suspended AudioContext would silence live playback). Tests: bg-style id validation + FX defaults (30 total).
- **3D Keys Highway: the anti-plastic pass — lacquered note gems, glossy piano-black keys, a studio environment, scene themes and a gradient sky (parity slice 3).** The note gems move to `MeshPhysicalMaterial` with a full **clearcoat** (roughness 0.32, clearcoat 1.0/0.18, envMapIntensity 0.9): a sharp lacquer highlight over the colored body instead of the old dead matte surface — glass, not plastic. What sells it is **image-based lighting**: the same procedural PMREM "studio" environment as the drum highway (dark room + cool overhead / warm+cool side light strips, no addon dependency) feeds `scene.environment`, so the **black keys finally read as glossy piano black** (roughness 0.55 → 0.22, envMapIntensity 1.3) with visible strip reflections, whites keep an ivory sheen (0.42/0.55), and the highway floor gets a stage sheen (roughness 0.9 → 0.55, metalness 0.15). The flat background becomes a **vertical gradient** (lighter above the horizon → theme color → darker toward the keyboard), and the guitar highway's **11 scene themes** arrive (same names/values — your look carries across instruments; `default` preserves the original keys palette; pitch-class note/key colors are never themed — themes own the scene, Synthesia colors own the notes). Plus **Cinematic lighting** (ambient 0.55/key 1.3, on by default) and a **Glow strength** slider multiplying the note glow, key approach-glow and the sustain consume-flash (0.5 = stock). All live-applying from the Graphics settings (`keys3d_bg_theme` + `keys3d_bg_*`); the PMREM target and gradient texture are disposed with the scene. Tests: theme-table id parity + default-look preservation + fallbacks (28 total).
- **3D Drum Highway: audio-reactive background ambience + score effects (parity slice 4 — completes the drum side of the visual-parity epic).** The empty fog band behind the kit gains the guitar highway's **background ambience styles**: drifting **Particles**, palette-colored pulsing **Stage lights**, and slowly-tumbling wireframe **Geometric** shapes (plus Off) — driven by the same audio-analyser bridge the guitar uses (prefers the stems plugin's per-song analyser on sloppaks, falls back to a one-shot `#audio` tap; bass/mid/treble bands with a 5 ms cache), with an **Ambience intensity** slider and an **Audio-reactive** toggle (off = the styles animate on time only; a permanently-tapped `#audio` in a mixed split degrades the same way). The guitar's butterchurn/image/video styles are deliberately out of scope (vendored megabytes / upload plumbing; the style enum is extensible). And your combo finally talks back: a **score-FX overlay** (2D canvas over the WebGL scene, guitar `drawScoreFx` adapted to this plugin's internal scoring) draws rising **“+1” pops** off each scored lane, an **expanding ring pulse every 10-combo tier**, **milestone particle bursts** at 25/50/100 streak, and a brief **red flicker** when a 3+ streak breaks. Everything is settings-gated (Background ambience dropdown + intensity + reactive, Score effects toggle — all on by default, live-applying, `drum_h3d_bg_*` keys), pooled (zero per-frame allocation), and torn down with the scene across kit changes. Tests: bg-style id validation + FX defaults (15 total).
- **3D Drum Highway: real materials + scene themes (parity slice 3).** The scene gets **image-based lighting**: a procedural PMREM "studio" environment (dark room + three emissive light strips — cool overhead key, warm/cool side fills; no vendored-addon dependency) feeds `scene.environment`, so the cymbals' metalness **finally reads as metal** (retuned to roughness 0.2 / metalness 0.85 / envMapIntensity 1.2 — the old 0.7-metalness look was matte because there was nothing to reflect), drumheads get a satin sheen, and the floor (roughness 0.95 → 0.7) catches the strips without turning into a mirror. **Scene themes arrive** — the same 11 theme names as the guitar highway (Midnight, Charcoal, Deep Purple, Forest, Warm Slate, Deep Focus, Deep Sea, Cathode, Cathode Green, Hearth) retint the background/fog, floor and lane stripes so your look carries across instruments; `default` preserves the original drum palette byte-for-byte, and piece colours stay with the existing Palette picker (themes own the scene, palettes own the kit). Plus **Cinematic lighting** (dimmer ambient / stronger key, on by default), a **Glow strength** slider (0–1, 0.5 = stock) multiplying every emissive base — notes, hit line, snare wires — and a **Lane vibrancy** slider driving stripe/halo/ghost-ring strength (the hit-FX approach highlight stacks on top). Everything applies live from the plugin settings (`drum_h3d_bg_theme` + `drum_h3d_bg_*` keys); the PMREM render target is rebuilt across kit-change renderer recreation and disposed in both teardown paths — and the floor/hit-bar geometry+materials that previously leaked on every kit change are now tracked and disposed too. Tests: theme-table parity with the guitar ids + default-look preservation + fallbacks (13 total).
- **3D Drum Highway: hit FX — sparks, timing-colored lane flashes, kick camera pulse, approach glow, and open hi-hat notation (parity slice 2).** Striking a pad now *feels* struck: a pooled additive **spark burst** fires at the lane (ported from the guitar highway's Points-cloud system, pool 160), colored by **timing** — on-time green, early cyan, late amber (same `_timingHex` vocabulary as `highway_3d`, classified against the ±50 ms hit window with the inner 40% reading as on-time); with **Streak feedback** on, bursts grow with your combo. The **lane flash** feedback that was removed when note-recoloring landed is resurrected properly: pooled additive quads with a soft gaussian falloff light up the struck lane at the hit line (timing-colored; red for wrong-pad hits), and a **kick** hit fires triple amber bursts across the bar plus a subtle **camera dip + amber floor wash** that decays exponentially. Lanes also glow ahead of time: each stripe brightens as its next note approaches the hit line, so the eye is led to where the next hit lands. **Open hi-hat finally renders distinctly** — `hh_open` chart hits get a thin warm ring around the cymbal gem (standard notation's "o"), closing the long-standing TODO; the flag is orthogonal to accents/ghosts/flams so combined cues stack. All of it is settings-gated (Graphics → Hit sparks / Timing colours / Streak feedback / a 0–1 **Hit feedback intensity** slider driving flashes, approach glow and the kick pulse; everything on by default, `drum_h3d_bg_*` keys, live-applying) and GPU-frugal: every new visual is pooled or shares geometry/materials — zero per-note allocation on top of the per-frame notes rebuild, all registered in both dispose paths (kit-change renderer recreation included). Tests: timing-classifier boundaries + FX defaults added to `plugins/drum_highway_3d/tests/data_layer.test.js` (10 total).
- **3D Keys Highway: hit FX — vibrant note gems, timing-colored sparks, and a hit-line that reacts to your playing (parity slice 2).** The washed-out note look is gone: gem opacity is now driven by a **Note vibrancy** slider (default 0.85 → opacity 0.92, up from a fixed 0.8; lane guides scale with it too, live-applying without a chart rebuild) and the resting emissive glow rises 0.08 → 0.22, so the falling notes finally read saturated against the dark floor. Scored key presses fire a pooled additive **spark burst** at the struck key (guitar-highway port, pool 96) **colored by timing** — on-time green, early cyan, late amber, classified against the ±100 ms window with the inner 40% reading as on-time (the timing delta is recovered from the matched note's key, so `judgeHit`'s tested contract is untouched); the per-pitch-class flame sprite keeps its identity color so pitch and timing stay separate signals. With **Streak feedback** on, bursts grow with the combo. The **hit line kicks brighter** for a beat on every scored press (exponential decay, scaled by a 0–1 **Hit feedback intensity** slider). All new controls live in the plugin's Graphics settings (on by default, `keys3d_bg_*` keys, live-applying), and the spark pool is disposed with the scene like every other GPU resource. Tests: timing-classifier boundaries, the noteKey time round-trip that the delta recovery relies on, and the new FX defaults (26 total).
- **3D Drum Highway: bloom glow + adaptive-resolution support — the first slice of visual parity with the guitar highway.** The drum highway now renders through the same post-processing path as `highway_3d`: an `UnrealBloomPass` (strength 0.65, radius 0.5, threshold 0.82 — high, so only emissive/bright surfaces bleed) on a multisampled HalfFloat target with ACES filmic tone mapping, so the white hit-line bar and proximity-lit notes get a real glow instead of a flat emissive tint. **On by default**, with a new **Graphics → "Glow (bloom)"** toggle in the plugin settings (`drum_h3d_bg_bloom`, applies live, no reload); if the vendored postprocessing addons can't load (older self-hosted core), the plugin silently falls back to the direct render path. The plugin also now honors the host's **adaptive render scale** (`bundle.renderScale` — the Quality/"Min res" controls that the guitar highway already respected), multiplying it into the device pixel ratio, and caps DPR at 1.25 when more than one viz instance is live (splitscreen) so two panels don't double the GPU fill cost. Groundwork for the rest of the parity series: an FX-settings scaffold (`FX_DEFAULTS`/`readFxSettings`/`window.drumH3dSetFx`, `drum_h3d_bg_*` localStorage keys) that the sparks/themes/backgrounds PRs extend, plus a first node test suite for the plugin (`plugins/drum_highway_3d/tests/data_layer.test.js` — vm-loaded like the keys plugin's, covering the hit-variant precedence, the Auto-mode steal-guard predicate, and FX defaults; 8 tests, runs in CI via the `plugins/*/tests/*.test.js` glob).
- **3D Keys Highway: sharp HiDPI rendering, bloom glow, a live combo HUD, and a graphics settings panel — the first slice of visual parity with the guitar highway.** The biggest single fix is resolution: the plugin never called `setPixelRatio`, so on HiDPI/retina displays (and Windows display scaling) it rendered at CSS resolution and was upscaled — soft and aliased. It now multiplies the device pixel ratio (capped at 2, or 1.25 when two viz panels are live in splitscreen) with the host's **adaptive render scale** (`bundle.renderScale`, the Quality/"Min res" controls), exactly like `highway_3d`. On top of that: the same **bloom** post-processing path as the guitar highway (UnrealBloomPass 0.65/0.5/0.82 on a multisampled HalfFloat target + ACES tone mapping — the cyan hit-line, hit flames and the sustain "consume" glow finally bleed light instead of reading flat), **on by default** with a graceful direct-render fallback when the vendored addons can't load. The plugin gains its first **settings panel** (`settings.html`, Settings → graphics category, `"settings"` block in plugin.json) with a live-applying "Glow (bloom)" toggle (`keys3d_bg_bloom`), plus the FX scaffold (`FX_DEFAULTS`/`readFxSettings`/`window.keys3dSetFx`, `keys3d_bg_*` keys) the later parity PRs extend. And the score state the plugin was already tracking is finally visible: a **combo / accuracy / best-streak HUD** overlay (drum-highway pattern), shown only while a MIDI keyboard session is wired so it never renders a frozen 0× combo. Tests: `plugins/keys_highway_3d/tests/fx_settings.test.js` (defaults, localStorage overrides + type coercion, setter persist/dispatch/unknown-key guard; 3 tests alongside the existing 20).
- **The 3D Drum Highway and 3D Keys Highway are now bundled core plugins** (`plugins/drum_highway_3d/`, `plugins/keys_highway_3d/`), imported from their former standalone repos (`feedBack-plugin-drum-highway-3d`, `feedBack-plugin-keys-highway-3d`, now archived) via `git subtree` so their history is preserved. They join the other in-tree plugins-as-plugins: the loader treats them identically to user-installed ones, both are marked `"bundled": true` in their manifests, and `.gitignore` gains the matching `!plugins/<id>/` exceptions. This puts all three 3D highways (guitar, drums, keys) in one repo ahead of a visual-parity pass that ports the guitar highway's polish (bloom, sparks, themes, reactive backgrounds) to the other two — shared helper code and theme tables can now be reviewed and kept in sync in a single place. The keys plugin's existing node test suite is wired into CI (the JS test step gains a `plugins/*/tests/*.test.js` glob, +20 tests), and `static/tailwind.min.css` is regenerated since the core Tailwind build scans `plugins/**`. One deliberate behavior change ships with the bundling: the drum highway's Auto-mode predicate is **narrowed** (it used to claim any pack with `has_drum_tab` — a pack-level flag — which, now that the plugin ships to everyone and sorts before `highway_3d` in first-match-wins Auto order, would have stolen full-band packs from the guitar highway even on Lead/Bass arrangements; it now claims only drum arrangements, or packs nothing more specific can render). Picking the drum highway manually from the viz picker is unchanged.
- **The tuner now tracks what tuning your instrument is *actually* in, so it prompts you to retune in BOTH directions — down to a song's tuning, and back up when the next song needs it.** The coverage check used to compare each song against your fixed instrument-profile tuning, so it only ever prompted you *away* from "home" (e.g. E → Drop C#) and stayed silent coming back (Drop C# → E), even though you'd physically retuned. It now reads the host's live **per-instrument working tuning** (`window.feedBack.workingTuning`) — what your selected instrument is currently in — so coverage is measured against your *actual* tuning and fires both ways. When you clear an auto-opened tuner, the tuner publishes that song's tuning as your instrument's live working tuning (`assumed` — an explicit "I tuned / Skip" refines it in a later PR), so the next song is judged against where you now are. **Per-instrument** — your guitar's and bass's tunings are tracked separately (keyed like the selector), so switching instruments uses the right one. Feature-detected: on a host without the working-tuning capability it falls back to the static `/api/settings` tuning (today's behavior). `plugins/tuner/screen.js` (`_playerTuning` reads `workingTuning` keyed by the selected instrument; `_publishWorkingTuning` writes on clear). Builds on the host `workingTuning` foundation (PR 1 of the series) + the instrument→chart routing (PR 2). Tests: `tests/js/tuner_auto_open.test.js` (both-directions coverage via a live Drop-D working tuning; publish-on-clear targets the right instrument slot) — 29 pass.
- **`.jsonc` support for feedpak data files** (feedpak-spec §8, FEP #3 / PR #13). Hand-edited packs may now use the `.jsonc` extension (JSON with C-style `//` line and `/* */` block comments) for any data file the manifest points at — arrangements, notation sidecars, `drum_tab`, `song_timeline`, `lyrics`, and `keys`. New shared `lib/jsonc.py` provides `parse_jsonc(text)` + `load_json(path)` (auto-detects `.jsonc` by suffix, string-aware so comment-like text inside JSON string values is preserved) and is now used by every reader in `lib/sloppak.py` (six side-file sites) and `scripts/lift_keys_notation.py` (three arrangement / song_timeline read sites). The strip regex mirrors the reference validator in `feedpak-spec/tools/validate.py`. This is an additive (MINOR) change: `.jsonc` is opt-in, so any pack that keeps its data files as `.json` is unaffected and needs no regeneration. Note that a `.jsonc` file containing real comments only loads on a reader that implements §8 — a pre-this-change reader calls bare `json.loads` and fails on the comments rather than ignoring them, so don't hand out `.jsonc` packs to older hosts. Tests: `tests/test_sloppak_jsonc_load.py` (covers all six side-file types, the lift helper, and the string-boundary preservation rule end-to-end).
- **The highway now loads the part that matches your selected instrument — a bass player gets the Bass arrangement, not the default Lead/guitar chart.** When you open a song without an explicit arrangement, the WebSocket handler (`server.py` `highway_ws`) reads your selected `instrument` from `config.json` (the same file it already reads for your default-arrangement preference) and routes to the matching part: **bass → the Bass arrangement**; guitar — and any unknown/future instrument (drums, keys) — falls through to the existing preference/most-notes default, which already lands on a guitar part. Previously the instrument selector only fed the tuner, so a bass player was handed a guitar chart (and a tune/coverage check then compared a 4-string bass against a 6-string part). An **explicit arrangement request always wins** (a manual arrangement switch is untouched), and a bass player's saved default-arrangement preference is still honored **within** the bass parts (so a preferred `Bass 2` / `Alt. Bass` wins over the canonical Bass), so this only changes the *default* part chosen on load. Server-only — every launch path already flows through the WS, so there's no client change. This is the instrument↔chart-routing piece the working-tuning series leans on (otherwise coverage compares across instruments). Tests: `tests/test_highway_ws_instrument_routing.py` (bass→Bass, bass-honors-pref, bass-no-bass-part→guitar, guitar→default, explicit-wins).
- **Host "working tuning" — a live, app-wide record of what tuning your instrument is *actually* in right now (foundation; no behavior change yet).** Introduces `window.feedBack.workingTuning`, a host-owned, session-lived state distinct from any one song's tuning and from a soft opt-in default: the offsets + string-count + reference pitch the player's instrument is currently in, plus an `assumed`/`verified` provenance flag. It's **per-instrument** — your guitar's current tuning and your bass's are kept *separately* (keyed like the instrument selector, e.g. `guitar-6` / `bass-4`), so switching instruments surfaces that instrument's own remembered tuning and you only ever deal with the one you've selected. It exists so a retune — or an instrument swap mid-session — is reflected **everywhere** (the highway, the library/song-picker, and plugins like the tuner, Virtuoso, and the minigames) instead of being re-derived per surface or wrongly assumed from a fixed profile. Modeled on the shipped `tuning` capability + the `feedBack.theme` read-API: a **synchronous `get(instrument?)`** (returns the selected instrument's state, defaulting to the seed until known), a `set(state, {provenance, instrument})` mutator (the tuner becomes the sole writer in the next change), `setCurrentInstrument()` for the selector, `resetToDefault()`, and a `working-tuning-changed` event that fires on every change **and once on hydration** (carrying which instrument changed) so a late-mounting consumer is never stuck on stale state. State is **in-memory, seeded from `/api/settings` on boot and reset on restart** — a stale "you're in drop-A" assumption is worse than re-asking. Registered as a separate `working-tuning` **exclusive-owner** capability (tuner = writer, the rest = requesters). This is the foundation (plumbing only — nothing writes to it yet) of the working-tuning series, which fixes the tuner gate only ever prompting *away from* a fixed "home" tuning (never back) and makes the current tuning a first-class signal the whole app shares. Offsets use the same per-string semitone vocabulary as song tunings, so fully custom/extended tunings (e.g. a drop-A 8-string) are first-class. Frontend-only: new `static/capabilities/working-tuning.js`, loaded from `static/index.html` + `static/v3/index.html`.
- **The v3 Songs grid is now DOM-virtualized — card-node count stays bounded no matter how big the library is or how far you scroll.** The grid used to append every scrolled page and never let go, so a 2000-song library grew the DOM from 24 → 624 → 2001 card nodes as you scrolled (layout/memory cost scaling with depth). It now renders only the **visible window** of cards (± a small overscan); a sizer element sized to the whole library (`ceil(total/cols) × rowH`) gives the scrollbar its full geometry while the grid is absolutely positioned to the first visible row. `state.songs` is a sparse, absolutely-indexed store fetched a page at a time on demand — using the stage-1 **keyset cursor** for contiguous forward scroll (O(page)) and falling back to `OFFSET page=` for jumps/restore/non-keyset providers (collections, remote). Verified bounded (~60 nodes for a 2001-song library while the count still reads "2001 songs"). The **A–Z rail now seeks directly**: `sort_letters` gives a letter's first-row index (cumulative of prior buckets), converted to a scrollTop in O(1) — no more paging through every intervening row (a bounded forward scan covers the rare legacy provider without `sort_letters`). Select-mode selections, accuracy badges, the ⋮ card menu, plugin card actions, scroll-restore (now scrollTop-based, since geometry is stable), and the tree/folder views all survive cards leaving and re-entering the DOM. Plugins that decorate cards get a stable `window.v3Songs.visibleCards()` accessor + a `v3:library-window-rendered` event instead of assuming every card is present (the highway-stutter lesson). Stage 2 of the virtualized-grid project (got-feedback/feedBack#636 item 3), building on the stage-1 keyset data layer below. Frontend-only: `static/v3/songs.js`, `static/v3/v3.css`. Tests: `tests/browser/v3-grid-virtualization.spec.ts` (bounded-DOM invariant across a 2001-song scroll + direct rail jump), updated `tests/js/v3_az_rail.test.js` + `tests/js/v3_songs_scroll.test.js`.
- **Keyset (cursor) pagination for the library grid — the data layer for an upcoming virtualized grid, and a latent paging bug fixed along the way.** Every library sort now carries a unique `filename` tiebreak, making the order **total** — which fixes a latent bug where rows sharing a sort key (e.g. two songs by the same artist) could be skipped or duplicated across `OFFSET` pages. `GET /api/library` gains an opaque `after` cursor + a `next_cursor` in the response: passing the cursor back fetches the next page with a **WHERE-seek** instead of `OFFSET`, so deep paging is O(page) regardless of depth. The seek is NULL-aware and exactly `OFFSET`-equivalent (verified across artist/title/recent, ascending + descending, including the legacy `dir=desc` shape and NULL sort keys); unknown/compound sorts and bad cursors fall back to `OFFSET`, and only the local provider is handed a cursor (collections/remote page by `OFFSET`). New composite `(artist NOCASE, filename)` / `(title NOCASE, filename)` / `(mtime, filename)` indexes cover the order. This is stage 1 of the virtualized-grid project (got-feedback/feedBack#636 item 3); the DOM-recycling render window builds on it next. Tests: `tests/test_library_keyset.py` (keyset==OFFSET parity, stable tiebreak, dir=desc, NULL keys, cursor fallback).
- **Smart collections — save a set of library filters as a live, auto-updating source.** A collection is a saved `/api/library` query (e.g. "Drop-D tunings", "sloppak only", "recently added") that stays live: it's registered as a **library provider**, so it shows up in the v3 Songs source picker and inherits the whole grid UI — paging, stats, the A–Z rail, art — for free, with **no new screen**. Storage reuses the playlist subsystem (a `playlists.rules` JSON blob = a smart collection; membership is the live filter result, not stored songs, and collections are excluded from the manual-playlist list + read-only to playlist mutations). New `GET`/`POST`/`PUT`/`DELETE /api/collections`; a per-collection `SmartCollectionProvider` delegates `query_page`/`query_stats`/`query_artists` to the local DB with the stored rules applied; providers are re-registered from a boot scan so collections survive a restart. Rules mirror the raw `/api/library` query params (unknown keys dropped, never 500). Frontend: a "+ Save as collection" action in the v3 filter drawer (shown when filters are active) names the current filter set and switches to it. The charrette's "the homelab primitive FeedBack was missing" pick (got-feedback/feedBack#636 item 2); richer rule fields (accuracy, genre, difficulty) follow as the metadata work lands. Tests: `tests/test_collections_api.py`, `tests/js/v3_collections.test.js`.
- **The settings backup now includes your library database + custom art — your scores, favorites, playlists, and play history are no longer the one thing a backup can't save.** `GET /api/settings/export` gains an additive `core_server_files` section carrying a **consistent snapshot of `web_library.db`** (taken via the SQLite online-backup API, so it's a complete single file even while the server is running) plus any custom **playlist covers** and **avatar** (`CONFIG_DIR/playlist_covers/`, `CONFIG_DIR/avatars/`). On `POST /api/settings/import` the database is **staged** to `web_library.db.restore` rather than written over the live, open DB; it's swapped in at the next startup (`_apply_pending_db_restore`, before the connection opens), which also clears the old WAL sidecars so a stale `-wal` can't be replayed onto the restored file — the import response sets `restart_required: true` and warns accordingly. Custom art is written immediately. The bundle stays backward-compatible (older servers ignore the new section). Came out of the library design charrette (dev-ops lens's top "protect irreplaceable data" pick, got-feedback/feedBack#636). _Known gap:_ custom uploaded **song** art is still commingled with the rebuildable thumbnail cache in `art_cache/`, so it isn't bundled yet (a tracked follow-up). Tests: `tests/test_settings_export_library_db.py` (snapshot consistency, staged-not-live restore, sidecar clearing, traversal rejection, full round-trip).
- **A persisted wishlist — keep a list of songs you want but don't own yet.** New `wanted` table + `GET`/`POST`/`DELETE /api/wanted` give FeedBack the *arr-style "Wanted/Monitored" primitive it was missing: an entry is a *not-owned* song (artist/title/source/source_ref/note), so it lives in its own table rather than the playlist subsystem (which references owned local files). The API is idempotent on identity (case-insensitive artist+title, plus source+source_ref), so a producer — the `find_more` ownership-diff, or a manual add — can re-post without duplicating. Newest-first. Backend primitive for the charrette's wishlist finding (got-feedback/feedBack#636 item 4); the consuming UI lives in the producing plugin. Tests: `tests/test_wanted_api.py`.
- **Practice-aware library home — a "Repertoire" meter + a "Keep practicing" shelf on the v3 Songs page.** The library opened cold into a flat sorted grid; now the unfiltered grid front door leads with two practice-aware surfaces built entirely from data already on hand (no new endpoints or stored state). A **Repertoire meter** shows how much of your library you can actually play — *"Repertoire: 12 of 80 songs · 7 in progress"* with a progress bar — counting songs at or above the same mastery threshold the green accuracy badge uses (≥ 90% best accuracy) over the unfiltered library total. A **"Keep practicing" shelf** is a horizontal row of your recently-played-but-not-yet-mastered songs (newest first, click to play) — the practice-accuracy-driven "continue" rail a media server can't do. Both reuse `/api/stats/best` (already loaded for the card badges) + `/api/stats/recent`; they show **only** on the grid view when you aren't searching/filtering/selecting, refresh after a song is scored, and collapse to nothing on an empty library. Soft-gamification only — descriptive encouragement (goal-gradient / endowed-progress), never content-gating, decay, or nagging. Frontend-only: `static/v3/songs.js` (`renderLibraryHome`/`_repertoireCounts`), `static/v3/v3.css`. Came out of the library design charrette (the UX + gamification lenses' top pick). Tests: `tests/js/v3_keep_practicing.test.js`.
- **A–Z fast-scroll rail on the v3 Songs grid.** A vertical letter rail (Plex/Radarr/iOS-contacts pattern) pinned to the right edge next to the scrollbar lets you jump the library to a starting letter — tap a letter, drag to scrub with a live letter bubble, or arrow-key between letters. It shows **only** for the grid view + alphabetical (artist/title) sorts, and only offers letters actually present in the current sort **and filter set**, so a tap always lands on a real card (absent letters are dimmed + non-interactive). Because the grid is forward-only, server-paged infinite scroll, a jump pages through to the target card and scrolls to it (a newer jump supersedes an in-flight one); a keyset-seek + virtualized window is the noted scaling follow-up for very large libraries. Backend: `/api/library/stats` now accepts `sort` and returns an additive `sort_letters` map (songs-per-first-letter of the active sort column — artist or title), filter-synced; the legacy `letters` (distinct-artist) field is unchanged for the dashboard + classic tree. Frontend: `static/v3/songs.js` (`refreshRail`/`jumpToLetter`, cards tagged with `data-letter`), `static/v3/v3.css` (`.v3-azrail`). The classic (v2) tree already had letter selection; this brings the new grid to parity. Tests: `tests/test_library_filters.py` (sort_letters artist/title + song-vs-artist counting), `tests/test_library_providers.py` (sort forwarded to providers), `tests/js/v3_az_rail.test.js`.
- **Playlists get content-dependent covers + custom art.** Playlist cards were a tiny `🎵` emoji on an empty square. Now a playlist's cover reflects its contents: **empty → the icon**, **a few songs → the first song's album art**, **4+ songs → a 2×2 art mosaic**. You can also **upload a custom cover** (a "Cover" button in the playlist detail view → image picker; "Remove cover" reverts to the content view). `MetadataDB.list_playlists()` now returns each playlist's first few song `art_urls`; `GET /api/playlists` and `GET /api/playlists/{id}` add `cover_url` when a custom cover exists. New routes `POST` / `GET` / `DELETE /api/playlists/{id}/cover` store a small PNG thumbnail under `CONFIG_DIR/playlist_covers/` (PIL-converted, like song-art upload); the cover is removed with the playlist. Frontend: `playlistCoverHtml(p)` in `static/v3/playlists.js`. Tests: `tests/test_playlists_api.py` (art_urls + cover roundtrip / reject-non-image / delete-cleanup), `tests/js/v3_playlist_cover.test.js`.
- **v3 Songs: "Add to playlist" is now on each song's ⋮ "More" menu.** Previously a song could only be added to a playlist through select-mode (the checkbox → batch bar). The per-card overflow menu now has an **Add to playlist** row that targets that one song, reusing the same picker (choose a listed number or type a new name to create it). The select-mode batch flow and the single-song menu now share one extracted `addFilenamesToPlaylist(filenames)` helper in `static/v3/songs.js` (both grid and tree rows, since they share `openCardMenu`). Tests: `tests/js/v3_add_to_playlist_menu.test.js`.
- **Resume where you left off — leaving a song now snapshots your place so an exit is recoverable, not a restart-from-zero.** Exiting the player (`showScreen()` teardown, before audio unload) writes `{song, arrangement, position, speed}` to `localStorage` (`feedBack.resumeSession`), and a non-blocking **"Resume practice"** pill offers it back on the next non-player screen (and on the next app launch). Clicking Resume re-enters the song, restores the arrangement + playback speed, and seeks to the saved position via the existing `_audioSeek` funnel; `playSong()` gains a `{ resume: {position, speed} }` option that arms a `song:ready`-consumed restore instead of the normal autostart, so the two never fight over playback. The snapshot is deliberately conservative — ignored for a song you barely started (< 3s) or had basically finished (within 5s of the end), cleared on natural song-end and once consumed, and expired after 24h. The pill is self-contained (inline-styled, body-appended, works identically in the classic and v3 shells with no Tailwind rebuild), never blocks, and a dismiss forgets the current snapshot for the session. This pairs with the Escape focus fix: now that Escape reliably leaves regardless of focus, an *accidental* exit is one tap to undo. Public surface: `window.resumeLastSession()` / `window.feedBack.resumeLastSession`. (The broader nav-state work — returning to a song after wandering into Settings → Tone Builder — is a separate, larger track; this lands the player-session slice.) Tests: `tests/browser/resume-session.spec.ts` (snapshot guards, staleness, pill show/hide/dismiss, resume consumption).
- **Optional "Ask before leaving a song" confirm (Gameplay tab, default OFF).** A new client-only toggle (`confirmExitSong` in `localStorage`, in the v3 Gameplay settings + the Gameplay "Reset" set) for players who want a guard against an accidental exit. **Off by default — Escape leaves instantly, zero change for everyone else.** When on, a *user-initiated* exit (the player-scope Escape shortcut, or the player's ✕) opens a small true-modal confirm instead of leaving; auto-exit on song-end and a results screen's own Close are unaffected (they call `closeCurrentSong()` directly, which stays the unguarded actual-exit). The confirm honors the team's refined asks: **opening it pauses the song** (so it isn't running or being scored behind the prompt) and **Stay resumes exactly what was paused**; **Escape = Stay** — the dialog's capture-phase handler *dismisses* it (now consistent with every other modal and the generic `_confirmDialog`'s Esc=cancel), so a second Escape returns you to the (resumed) song rather than leaving; **Space/Enter (or click) Leave** by natively activating the default-focused "Leave" button ("just get me out"). Pause/resume run through the canonical `togglePlay()` path (HTML5 + `_juceMode`), guarded so a count-in, an already-paused song, or a teardown/seek/end behind the modal can't mis-resume. It's a real modal (`role="dialog" aria-modal="true"` / `.feedBack-modal`) with **Tab trapped inside it** and a **backdrop click that also Stays**, so the Escape/Space focus carve-outs treat it as a trap and don't fire player-back / play-pause behind it. The player Escape shortcut and the v3 ✕ route through a shared `window.requestExitSong()` gate (the ✕ also becomes origin-aware, matching Escape). Tests: `tests/browser/exit-confirm.spec.ts` (default-off instant exit, confirm-on opens + stays, second-Escape stays, backdrop stays, Stay/Leave, Enter-leaves); the audio pause/resume is verified manually on web + desktop (the mock song has no backing track).
- **Folder Library — a bundled core plugin (`plugins/folder_library/`) that browses the DLC library by its on-disk folder tree.** Surfaces top-level folders → subfolders → songs (root-level songs land in `(Unsorted)`), with in-app folder management (create / rename / delete nested folders), song moves via dialog or drag-and-drop, and sort/filter that mirrors the host library's filter state. Wired into both the classic (v2) library toolbar and the v3 Songs page as a third **Folders** view alongside grid/tree; the plugin's `screen.js` is loaded once by the host and reused (idempotent IIFEs). Supersedes the former standalone "Folder Organizer" community plugin (removed from the README list). Backend (`routes.py`) registers `/api/plugins/folder_library/{tree,folder/create,folder/rename,folder/delete,song/move}`; **all filesystem mutations are confined to `DLC_DIR` and validated against path traversal** (per-segment name validation plus a resolved-containment check on `song/move`), and folder deletion relocates every song — de-duplicating colliding names — so a name clash never destroys a song. A two-level cache keeps re-opening folders fast. Tests: `tests/plugins/folder_library/test_routes.py` (path-safety helpers + move-traversal and delete-no-data-loss end-to-end).
- **Full-screen (immersive) plugin screens — opt-in via `"fullscreen": true` in `plugin.json`.** DAW-style plugin UIs (e.g. a practice studio) need the whole viewport, not a scrolling content page below the topbar — embedded in the v3 shell they get cut off at the bottom with dead space up top. A plugin can now declare a top-level `"fullscreen": true`; `plugins/__init__.py` surfaces it as the `fullscreen` boolean on `/api/plugins` (mirroring the `settings_category` plumbing). When such a plugin's screen is active, `static/v3/shell.js` toggles `html.fb-immersive` from `syncActive()` (so it tracks every navigation incl. deep-link), and `static/v3/v3.css` hides the topbar, collapses the sidebar to a functional **icon rail** (kept reachable — Escape is bound only on player/settings scopes, so a fully-hidden sidebar would trap the user), and lets the active plugin screen fill `#v3-main`. Mirrors the existing `ss-follower-pre` chrome-hide pattern. Additive + opt-in: plugins without the flag are unaffected. Tests: `tests/test_plugins.py::test_fullscreen_flag_parsed_from_manifest`.
- **Achievements wall sync — background drain worker (epic PR3, client side).** The bundled `achievements` plugin gains a dead-letter sync worker that POSTs queued Feat unlocks (and removals) to the hosted **feedback-achievements** wall service (separate repo). Idle unless a wall URL is configured (`FEEDBACK_ACHIEVEMENTS_WALL_URL`); uses `requests` with the baked-in client-token header, mirroring `lib/lyrics_transcribe`'s outbound pattern (explicit timeout, no raise on non-2xx). **Dead-letter, never drop** (pure `engine.drain_decision`): network error / `429` / `5xx` → keep `pending` (retry); other `4xx` → `dead_letter` (diagnosable, replayable); `2xx` → delete on server ack. A row leaves the queue only on ack or a user opt-out. `remove-me` now enqueues a wall removal keyed by the reused `player_hash`. Verified by an end-to-end staging round-trip (earn a Feat → drains onto the wall with name + short hash → `remove-me` → wall empties) with **no IP** in tables or access logs. Tests: `tests/plugins/achievements/test_sync.py` (decision table + ack/retry/dead-letter retention + four-field payload on the wire). The hosted service itself (FastAPI + SQLite-on-disk, Feats-only, hidden-until-first-global-unlock, profanity filter, in-memory rate limit, Render blueprint, migration tool) lives in the new `feedback-achievements` repo.
- **Achievements wall — opt-in, privacy controls & data-minimization gate (epic PR2).** Sharing earned **Feats** on the (forthcoming) public wall is strictly opt-in. A new **onboarding step** (`static/v3/profile.js`, inserted after song-directory / before instrument paths — the wizard is now five steps) presents a plain-language card: it publishes only your display name and the Feats you earn, never songs/skills/scores, and is **off by default**. The bundled plugin's Settings panel (`plugins/achievements/settings.html`, mounted under the **System** tab via `settings.category`) carries the same toggle plus a **"Remove me from the wall"** button (`POST /api/plugins/achievements/remove-me` — wipes local synced state offline + enqueues a wall removal). Core adds `achievements_enabled` (bool, default `false`) to `_default_settings()` + the `/api/settings` validation block + `_RESETTABLE_SETTINGS_KEYS` in `server.py`, mirrored to `localStorage` in `app.js loadSettings()`. **Data-minimization contract (binding, code-enforced):** every outbound payload is built by a single explicit-dict serializer (`engine.build_wall_payload`, never `dict(row)`/`**model`) whose key-set is **exactly** `{display_name, player_hash, achievement_id, unlocked_at}` with `achievement_id` always a **Feat** id — a unit test asserts the four-field set and goes red on a fifth. Enqueue is doubly gated: it happens only when opted-in **and** a profile identity (name + the reused `player_hash`) exists; **competency unlocks never enqueue** (integration law). Tests: `tests/plugins/achievements/test_datamin.py` (key-set, opt-out/identity/competency gating) + `tests/test_settings_api.py` (flag persists/validates/resettable).
- **Achievements & Feats of Power — local engine + tabbed Profile (epic PR1).** The Profile screen (`static/v3/profile.js`) becomes **tabbed** exactly like the v3 Settings page (`.fb-tabbar` / `.fb-tab[data-tab]` / `.fb-tabpanel[data-tab]`, active-tab persisted in `localStorage 'v3-profile-tab'`): a **Profile** (main) tab carrying the existing header + best-scores cards plus a new **Feats of Power** trophy shelf mount (`#v3-profile-feats-slot`, earned-only / hidden-until-earned), and an **Achievements** tab with a plugin mount (`#v3-profile-achievements-mount`) + `[data-empty-for]` empty note. Core dispatches a new **`v3:profile-rendered`** event after every render (mirrors `v3:settings-rendered`) so the plugin re-injects on each profile entry. A new bundled **`plugins/achievements/`** plugin owns the engine: SQLite under `<config_dir>/achievements/achievements.db` (`unlocks` / `counters` / `comp_ledger` / `sync_queue`), pure threshold/criterion math in the testable sibling `engine.py` (P-V), and routes under `/api/plugins/achievements/` (`activity`, `report-unlock`, `report-criterion`, `catalog`, `earned`, `feats`, `remove-me`). **Two surfaces, one engine, structurally separated (integration law):** **Feats** (activity/volume — Note Hunter, Marathon, Untouchable, Road Warrior, Time Served, Encore, two 🥚 secrets) read activity counters only, evaluated from a batched `song:ended` activity POST (notes only when **notedetect** is present — graceful degradation, no fake progress); **competency Achievements** (baseline: First Steps / Ascendant / Steady Hands / Renaissance + per-instrument Apprentice·Journeyman·Master / Personal Best / Challenger) are evaluated from **progression events only** and never re-derived from activity. The Achievements catalogue is always shown (locked = greyed), grouped by a secondary pill row over the **real progression paths** (Global / Guitar / Bass / Drums / Keys — auto-extends to new paths) with a per-category "X / Y earned" badge, defaulting to the player's primary path. Source plugins contribute their own competency defs and report unlocks through a versioned **`window.feedBack.achievements`** API (`register`/`registerAll`/`unlock`/`progress`), load-order-safe via the `window.__feedBackAchievementsPending` queue + an `achievements:ready` event (minigames pending-queue pattern); an absent source contributes nothing (no dead greyed rows). Opt-in publishing to a hosted Feats wall, the Settings privacy toggle, and the data-minimization gate land in epic PR2/PR3. Tests: `tests/plugins/achievements/test_engine.py` + `test_routes.py` (incl. the integration-law assertion that a competency unlock never reaches the Feats shelf).
- **v3 settings page redesigned as a tabbed, card-row layout.** The single long scrolling settings screen becomes a horizontal tab bar (Gameplay / Audio / Graphics / Keybinds / Progression / Mic / Plugins / System) over card rows — each a leading icon + title + description with the control (toggle/dropdown/slider) on the right, plus a per-category "Reset" action. The markup lives in `static/v3/index.html` (so existing element ids keep hydrating through the unchanged `app.js` `loadSettings()`/`persistSetting()` path); a new `static/v3/settings.js` owns tab switching + active-tab persistence (`localStorage 'v3-settings-tab'`), the per-category reset, and a read-only **Keybinds** reference built from the live shortcut registry (`window.getAllShortcuts()`); styling is plain CSS in `static/v3/v3.css` (no Tailwind rebuild). **Plugins choose their settings tab** via a new optional `settings.category` field in `plugin.json` (`plugins/__init__.py` surfaces it as `settings_category`; `app.js` mounts each plugin's `<details>` panel into `#plugin-settings-<category>`, falling back to the generic Plugins tab) — `highway_3d` ships `category: "graphics"`; the out-of-repo notedetect/progression plugins should declare `"mic"` / `"progression"`. **New gameplay settings:** **Countdown before song** (a four-beat count-in before playback, wired end-to-end via the existing count-in engine + the song-start autostart path; key `countdown_before_song`, default off); **Miss penalty** (`miss_penalty`) and **Fail behavior** (`fail_behavior`) are persisted now but not yet consumed by scoring (shown with a "Not yet active" badge). "Note highway speed" surfaces the existing `master_difficulty` and stays in sync with the player-popover difficulty slider. New `POST /api/settings/reset` clears chosen keys back to defaults. Tests: `tests/test_settings_api.py` (new keys + reset), `tests/test_plugins.py::test_settings_category_parsed_from_manifest`, `tests/browser/settings-tabbed.spec.ts`.
- **Full-mix audio exposed alongside stems for the stem mixer's auto-switch.** `lib/sloppak.py::load_song` now parses the optional manifest `original_audio:` key (the single pre-separation mixdown, e.g. `original/full.ogg`) into a new `LoadedSloppak.original_audio` field, with the same path-traversal guard and permissive "missing → disabled" posture as the `drum_tab` loader. The highway WS `song_info` frame additively carries three new fields next to `stems`: `original_audio_url` (served by the existing `/api/sloppak/{filename}/file/{rel_path}` endpoint, `None` when the pack ships stems only), `has_original_audio`, and `has_stems` (mirroring the `has_drum_tab`/`has_keys` flag convention). The stems plugin consumes `original_audio_url` to play the untouched single file while every stem slider is at unity and switch to the separate stems the moment one drops below 100%. **Migration notes:** the `song_info` message shape is a stable contract — these are purely additive; all existing fields are unchanged. `audio_url` still points at stem[0] when stems exist (it is only the degraded native fallback); the one behavioural change is that a stem-less, full-mix-only sloppak now sets `audio_url` to the full mix instead of emitting `audio_error`, so it plays natively.
- **Autoplay & auto-exit — a global "click it, it plays; finish, you're back at the menu" option (default ON).** New single Settings toggle (`autoplayExit` in `localStorage`, surfaced in both the v3 and classic settings screens; absence of the key = enabled) that closes the friction at both ends of the play loop. **Autoplay:** `playSong()` previously loaded a chart paused, requiring a Play press; a one-shot flag armed per fresh load is now consumed by the next `song:ready` (highway.js) to auto-start via the existing `togglePlay()` path (HTML5 + `_juceMode` + count-in). Arrangement switches / seeks reuse the same `song:ready` event but never arm the flag, so they don't auto-restart. **Auto-exit:** on `song:ended`, core returns to the launching menu after a short grace delay — unless a visible full-screen results/dialog overlay is on top (detected via `[role=dialog][aria-modal]` / `.fixed.inset-0` with a `getClientRects()` visibility test that works for `position:fixed`), in which case the return is deferred so that score screen's own Close button (calling `window.closeCurrentSong()`) drives the exit. A plugin can also defer explicitly via the new `window.feedBack.holdAutoExit()` (called synchronously from its own `song:ended` handler — core's listener runs first). Both paths mean **no external plugin PR is required** for a results screen to be respected. **Context-aware destination:** the player's remembered origin (`_playerOriginScreen`) now honours any real launch screen instead of clamping to library/home/favorites, and a one-shot `window.feedBack.setReturnScreen(id)` override lets the lessons catalog (`static/v3/lessons.js`) send a finished lesson back to the lessons screen — not the song library — even though the external tutorials plugin owns the `playSong` call. Also exposes a read-only `window.feedBack.autoplayExit` getter for plugins. Songs and lessons share the same `playSong` → highway path, so both inherit the behaviour. Core-only (`static/app.js`, `static/v3/lessons.js`, both `index.html`s); the end-of-song score screen itself remains a plugin. Optional polish (not required — the overlay heuristic already covers it): external scoring/note-detection plugins (e.g. SlopScale) may call `holdAutoExit()` + `closeCurrentSong()` for an exact, heuristic-free handoff.
- **"Song Editor" promoted to a first-class v3 sidebar item.** The editor
plugin (`id: editor`) now gets its own dedicated sidebar entry — under the
Library group, just below Songs — via the existing `PROMOTED_PLUGINS`
mechanism in `static/v3/shell.js`, instead of being reachable only through
the generic Plugins gallery. Gated on the plugin actually being installed
(`renderPromotedNav` checks `/api/plugins`), so it appears only when the
editor is loaded. The displayed label comes from the plugin's manifest
`nav.label`.
- **Guitar Pro → notation importer (`lib/gp2notation.py`)** (feedBack#825 WS4b, epic #828). Piano/keys tracks imported from Guitar Pro (GPIF: `.gpx` GP6 / `.gp` GP7-8) now produce real Sloppak Notation Format data (sloppak-spec §5.3) alongside the `midi = string*24 + fret` guitar wire encoding. `gp2rs_gpx.convert_file` writes a `<stem>.notation.json` sidecar next to each keys arrangement XML (best-effort — a notation bug never breaks the RS-XML conversion), and `gp2notation.attach_notation_to_sloppak()` is the assembly-side helper that renames it into `notation_<id>.json` + adds the per-arrangement `notation:` manifest sub-key. Voice→staff routing salvages the logic from PR #703 (whose `stf` wire-field approach this supersedes): GP voice position 0 → `rh` staff (`G2`), positions ≥ 1 → `lh` (`F4`); a forced-LH track (the merged `Piano LH` partner from `_find_piano_pairs`, or a standalone track named `… LH`) routes everything to `lh` — preserving authored hand crossings instead of inferring hands from pitch. Emits measures with absolute `t` from the bar-indexed tempo map, change-only `ts`/`tempo`/`ks`, and `beat_groups` for compound/irregular meters (6/8 → `[3,3]`, 9/8 → `[3,3,3]`, 5/8 → `[2,3]`, 7/8 → `[2,2,3]` — cf. the feedBack#261 denominator pitfalls); beats carry `dur`/`dot`/`tu`/`rest` from GP rhythms and notes carry absolute `midi` (String+Fret resolves via the string template's concert pitches, Tone+Octave via `(octave+1)*12 + step`) with `tied` continuations kept as real beats (engraving needs the tied notehead — unlike the RS-XML walk, which drops them and extends sustain). Timing reuses the `gp2rs_gpx` machinery (bar-indexed tempo map, per-beat rhythm durations, `_note_midi`) so notation lines up with the RS XML the highway plays — with one deliberate divergence: double dots advance time ×1.75 (vs. the RS-XML walk's single-dot ×1.5 approximation) so a written `dot: 2` agrees with the emitted beat times; sharing the walk itself stays tracked in feedBack#618. Tests: `tests/test_gp2notation.py`.
- **Legacy keys → notation lifter (`scripts/lift_keys_notation.py`)** (feedBack#825 WS4c, epic #828). One-time batch converter that lifts existing **directory-form** piano/keys sloppaks from the legacy guitar wire encoding (`midi = s*24 + f`) into real Sloppak Notation Format files (sloppak-spec §5.3). Candidates are arrangements whose name matches `\b(keys|piano|keyboard|synth)\b` (case-insensitive); each gets a `notation_<id>.json` plus the per-arrangement `notation:` manifest sub-key. Measures derive from the song-level `beats` downbeats (`measure >= 0`; `song_timeline.json` preferred, first-arrangement fallback), with per-measure tempo from downbeat spacing (emitted only on a > 1 BPM change). Durations come from the wire sustain (`sus`, legacy `l` alias) when present, else the gap to the next onset in the same hand — quantized to the nearest plain/single-dotted `{1,2,4,8,16,32}` denominator at the local tempo, floored at a 32nd. Hands are split heuristically: onsets within 10 ms form a group; a group spanning > 12 semitones splits at its largest internal interval gap (low side → `lh`), otherwise the whole group goes by mean pitch vs middle C — single-staff output when everything lands on one hand. Idempotent (arrangements already carrying `notation:` are skipped; an orphan `notation_<id>.json` without the manifest key is refused, not overwritten) with `--dry-run` support; every payload is checked via `notation.validate_notation` before write. Honest caveat: the manifest is round-tripped through PyYAML (`safe_load` + `safe_dump(sort_keys=False)`) — key order survives, YAML comments/custom formatting do not (the script warns when comments are present). Zip-form `.sloppak` files are reported and skipped. Tests: `tests/test_lift_keys_notation.py`.
- **Notation schema v1 freeze — completeness batch** (feedBack#822, epic #828). Adds the low-hanging-fruit fields ahead of content production: top-level credits `rights`/`lyricist`/`arranger`; measure `pickup` (anacrusis); beat `arp` (arpeggiate), `ferm` (fermata), and **typed grace notes** — `grace: "a"` (acciaccatura, MusicXML `grace/@slash=yes`) / `"p"` (appoggiatura); note `stem` (`"up"`/`"down"` force). Pedal is settled as the existing `spd`/`sph`/`spu` trio with a documented MusicXML `<pedal start|change|stop>` mapping — no separate `ped` field. A new "v1 non-features" spec subsection pins the accepted limitations (microtonal, figured bass, mid-measure key/time/clef changes, `ott`/`barline`/ornaments/`trem`/`glis`) as additive-v1.x territory. `lib/notation.py` gains the `GRACE_TYPES`, `STEM_DIRECTIONS`, and `DYNAMICS` vocabularies; the validator stays permissive by design.
- **Notation format — standard musical notation as a first-class sloppak type.** Promotes keys, piano, violin, and any other staff-notation instrument out of the guitar wire format and into their own data structure, following the same promotion path used for drums (feedBack#344). New `lib/notation.py` defines the canonical vocabulary (`CLEFS`, `DURATIONS`, `SCHEMA_VERSION`), a permissive `validate_notation()` check, and `measures_to_wire()` / `measure_to_wire()` wire helpers. `lib/sloppak.py::load_song` reads a new per-arrangement `notation:` sub-key from each arrangement entry in the manifest (Option B: per-arrangement, not song-wide), applies path-traversal guards, validates the parsed JSON via `validate_notation()`, and surfaces all notation payloads on `LoadedSloppak.notation_by_id` (a `dict[str, dict]` keyed by arrangement id). A failed or missing notation file for one arrangement does not abort or skip the arrangement itself — partial-failure isolation mirrors the drum tab loader. `file:` is now optional when `notation:` is present: the loader creates a minimal stub arrangement so a notation-only arrangement entry does not require a guitar wire format JSON. `/ws/highway/{filename}` gains two new message types — `notation_info` (staves, instrument, total measure count) and chunked `notation_measures` (32 measures per chunk) — streamed after `sections` and before `anchors`; `song_info` carries a new `has_notation: bool` flag so viz pickers can auto-activate the notation plugin regardless of arrangement name. The notation file schema is measure-structured (`measure → staff → voice → beat → note`), uses MIDI for pitch (no string/fret/tuning indirection), and carries the full set of effects that alphaTab can render. See `docs/sloppak-spec.md` §5.3 for the full schema. Open questions resolved per the piano/keys epic (feedBack#828 / #822): Option B (per-arrangement `notation:` sub-key) and `file:`-optional-when-`notation:`-present are the endorsed design.
- **`song_timeline.json` — beats and sections as a top-level file.** A new optional top-level file pointed at by a new manifest key (`song_timeline: song_timeline.json`) provides the correct home for song-wide beats and sections, replacing the legacy convention of embedding them in the first arrangement JSON. The loader in `lib/sloppak.py` reads and validates the file (must be a dict with `beats` and `sections` as lists), clears and repopulates `Song.beats` / `Song.sections` from it when present, and stores the raw dict on `LoadedSloppak.song_timeline`. The existing arrangement-JSON fallback is fully preserved: all existing sloppaks that omit `song_timeline:` continue to load without any change. This is a prerequisite for notation-only sloppaks, which may have no arrangement JSON at all and therefore no carrier for beats/sections data. New sloppaks should put beats/sections in `song_timeline.json` only. See `docs/sloppak-spec.md` §2 and §5.3.
- **`note-detection` capability domain promoted — control plane (spec 009)** (feedBack#727/#728, epic #828). New core host `static/capabilities/note-detection.js`: provider registry (kinds `midi`/`engine`/`js`, primitives `pitch.estimate`/`verify.target`), requester-owned context-scoped detection bindings (`open-binding`/`close-binding`/`set-target`/`clear-target` — each binding carries its own redacted tuning context, independent of the host's loaded song, per spec-009 FR-003), and hit/miss/verdict observability events (consumers own judgment). The legacy chart-coupled `highway.setNoteStateProvider` surface keeps working and is wrapped for compatibility-shim hit accounting. Diagnostics (`feedBack.note_detection_capability.v1`) carry provider/binding summaries and bounded outcomes — no raw audio, device labels, or song identity. Migrating the chart path, Step Mode verify, minigames YIN, and the engine verifier onto bindings is the remainder of the spec-009 slice.
- **`visualization` capability domain promoted (cap:6)** (feedBack#828). New core host `static/capabilities/visualization.js` registers a provider-coordinator owning the highway renderer surface: commands `inspect` / `list-providers` / `select-renderer` / `clear-renderer` (selection delegates to the existing picker so persistence, WebGL2 gating, and fallback stay single-sourced), events `providers-refreshed` / `renderer-changed` / `renderer-ready` / `renderer-failed`. Legacy discovery (`type: "visualization"` manifests, `window.feedBackViz_*` globals) keeps working unchanged and is accounted as compatibility shims with hit counts. `static/app.js` attributes every renderer change (auto-match / user-select / fallback) and auto-match outcomes into the domain. Diagnostics (`feedBack.visualization_capability.v1`) carry provider ids/labels/context types, active renderer + selection source, last auto-match outcome, and last failure — no song filenames/titles. Per-panel (splitscreen) selection is a tracked follow-up.
- **Viz picker routes notation arrangements** (feedBack#826, epic #828). `window.feedBack.currentSong` gains `hasNotation` (sibling of `hasDrumTab`) from the `song_info` frame's `has_notation` flag, so notation viz plugins (Staff View, Keys Highway 3D) can gate `matchesArrangement` on data presence instead of arrangement-name heuristics. When a notation-only arrangement (no wire notes — `file:` omitted per sloppak-spec §5.3) falls through Auto with no notation plugin installed, the built-in highway still takes the canvas but the Auto label reads "no notation view installed" and a one-shot dismissable hint points at the visualization picker — never a silently blank board.
- **Keys instrument path in progression** (feedBack#828). New `data/progression/paths/keys.json` (5 levels / 15 challenges at parity with the guitar path) plus keys-flavoured daily/weekly quest pool entries (`d.keys-one` "Ivory Tower", `w.keys-three` "Grand Recital"). `lib/progression.py::instrument_for_arrangement()` now attributes `type: piano|keys` arrangements — and names matching `keys`/`piano`/`keyboard`/`synth` on a word boundary — to the new `keys` instrument, so scored keys runs advance the path automatically. Purely content + attribution: no schema or API changes.
- **v3 library: exact artist/album filters + scroll/page-depth restore** (feedBack#857). The v3 Songs toolbar gains Artist and Album dropdowns (Album populates from the selected artist and stays disabled until one is chosen), backed by new exact, case-insensitive (`COLLATE NOCASE`) `artist` / `album` query params threaded through `MetadataDB._build_where` → `query_page` / `query_artists` / `query_stats` and the `/api/library`, `/api/library/artists`, `/api/library/stats` endpoints (the free-text `q` search stays fuzzy and composes with the exact filters). The artist/album catalog is fetched independently of the active artist/album selection so the dropdowns always list the full set for the current provider/search. The toolbar is now sticky so filter controls stay reachable when browsing deep libraries, and returning from the player restores the previous scroll position **and** the loaded infinite-scroll page depth via a `sessionStorage` snapshot keyed by a filter/sort/view state hash (invalidated whenever those change, so a filter change still resets to the top). Tests: `tests/test_library_filters.py` (backend artist/album filters), `tests/js/v3_songs_scroll.test.js` (state-hash + snapshot helpers).
### Fixed
- **GP8 multi-staff (piano/keys) tracks now import both hands — the bass stave was being silently dropped, and hand-splits landed on the wrong hand.** A GP8 grand-staff keyboard part is one `<Track>` with two `<Staff>` entries, and `MasterBar/Bars` lists one bar id per **stave**, not per track (`lib/gp2rs_gpx.py`). Two bugs fell out of assuming one stave per track: (1) the bar-column lookup used a raw `enumerate(Tracks)` index, so every track *after* a multi-stave track read the wrong column; (2) the string-tuning parse scanned all `.//Property` descendants and let the last stave's `<Tuning>` overwrite the first, so a treble note indexed against the 5-entry bass tuning fell out of range in `_note_midi` and was **dropped without a trace**. The importer now advances a bar-column counter by each track's stave count, reads tuning **per stave** (with a per-staff fall-back to the track-level property so an untuned staff never yields empty pitches), and folds **every** extra stave's notes into the arrangement (not just stave 1), keeping the `note_count` import-preview honest. A grand-staff track is now classified as keys end-to-end so the stave-0 and folded stave-1+ notes share one encoding. Separately, `notation_lift.split_hands` no longer forces a hard middle-C split when doing so produces a physically unplayable hand (e.g. a bass note under an Em7-shape voicing dipping below C4 would put a 19-semitone span in one hand) — it uses the middle-C boundary only when both resulting hands are within `HAND_SPLIT_SPAN_SEMITONES`, else falls back to the largest-gap heuristic. The GPX LH/RH pair merge and the GP8 stave fold now share one `_collect_column_notes` / `_merge_lh_notes` pair so the two formats can't drift in tie/timing/dedup handling. Companion editor change: got-feedback/feedBack-plugin-editor#38. Tests: `tests/test_gp2notation.py` (grand-staff fold + bar-column offset), `tests/test_notation_lift.py` (both middle-C split cases). Follow-up: `lib/gp_autosync.py` still carries the pre-fix bar-column + tuning logic (CLI/tests only, no production caller).
- **Tuner auto-open is now opt-in and persists instead of flashing open-then-shut.** When you entered a song (or switched arrangement) whose tuning differed from the last, the tuner auto-opened and — for some testers — vanished ~1s later (reported macOS+Windows, 0.3.0). Root cause: the tuner closes itself on `song:play` (`plugins/tuner/utils/ui.js` — you don't tune while playing), so a **song switch** fired autoplay → `song:play` → the just-auto-opened tuner closed; an **arrangement switch** (which never arms autoplay) had no `song:play`, so it stayed open — exactly why two testers saw opposite behaviour (it wasn't the mic). Now: (1) the feature is a **new opt-in setting** ("Auto-open on tuning change", in the tuner's Settings panel, persisted as `autoOpenOnTuningChange`, **default OFF**); (2) an **auto**-opened tuner *persists* — it ignores the autoplay `song:play`, stray outside-clicks, and same-screen re-emits, closing only via the new in-panel **`×`** / **"Skip"** buttons or when you leave the song. A *manually* opened tuner keeps its classic click-away / play-to-close behaviour. The panel previously had no in-box close at all; this adds one (`×` + contextual Skip). All in the tuner plugin (`routes.py` config, `screen.js` gate + persist, `utils/ui.js` buttons + `song:play` guard, `settings.html` toggle) — **no core `app.js` changes**. Tests: `tests/js/tuner_auto_open.test.js` (opt-in gate, `{ auto: true }` persist mode, play/click-proofing). **Default (opt-in vs opt-out) is teed up for Byron to decide — flip one boolean.**
- **Tuner auto-open is now tuning-coverage-aware — extended-range players aren't nagged for songs their instrument already covers.** With the opt-in auto-open on, it now prompts only when your **current physical tuning** (from your instrument selection in Settings) doesn't already cover the song. FeedBack is tune-to-song — the highway draws tab in the song's tuning — so the check aligns the song's open-string tuning string-for-string against your instrument: an **8-string F♯-standard** player gets **no** prompt for a 6- or 7-string standard song (its top strings already match those tunings), while a song needing an open string you don't have (e.g. a **Drop-A 7-string**, whose low A isn't an open string on an F♯ 8-string) **still** prompts. A whole-instrument reference difference also prompts — A440 vs A432, or an octave-down `centOffset` (which the auto-open now accounts for; it was previously ignored). The player's instrument is read from core **`/api/settings`** (the v3 instrument selector — a stable physical reference, not the tuner's song-tracking selection); when nothing's declared or the lookup is unavailable it falls back to a conservative prompt, so a real retune is never silently skipped. **v3-only** (the instrument selector is v3). All in the tuner plugin (`plugins/tuner/screen.js`) — **no core changes**. Tests: `tests/js/tuner_auto_open.test.js` (covered vs uncovered, the Drop-A case, reference-pitch mismatch, contiguous alignment). _Follow-up (E1.6): a passive "different tuning" badge cue that names the string(s) to retune, plus the splitscreen / no-usable-input guards._
- **The tuner badge now passively flags when a song needs a different tuning — and names the retune.** Building on the coverage check: when you enter a song your current instrument doesn't cover, the topbar tuner badge gets an amber ring and a tooltip that **names the change** — e.g. *"retune B→A"* for a Drop-A song on an F♯ 8-string, or *"the reference pitch"* for an A440-vs-A432 mismatch. It's purely **advisory** (it never auto-opens the panel — tap the badge to tune), recomputed on `song:ready` and cleared when a new song loads or you leave the player. The retune diff comes from the tuner plugin's coverage report (`window._tunerAutoOpen.coverageReport` → `{ covered, retune: [{ from, to }], reference, cantCover }`); the cue is CSS-free (an inline ring + native tooltip — no Tailwind rebuild) and no-ops when the tuner plugin isn't installed. **v3-only.** Touches `static/v3/badges.js` (the cue) + `plugins/tuner/screen.js` (the report). Tests: `tests/js/tuner_auto_open.test.js` (the report names the strings; reference mismatch; the badge wiring). _(The splitscreen-suppress and no-usable-input guards move to the playback-gate stage, where they matter for its no-trap rule.)_
- **Tuner auto-open can now gate playback until you've tuned — the "tune before you play" model — via a new core `holdAutoplay()` hook.** With the opt-in auto-open on, when a song needs a retune the tuner opens and **playback waits** for your choice — **Skip** (you've tuned → play, and record the song's tuning as your instrument's current working tuning), **Back to library** / **Esc** (leave the song; a gated retune is never a one-way trap), or press **Play** (always wins). For an auto-open the in-panel **×** is dropped — Skip / Back to library / Esc are its dismiss surface. Previously the song played with the tuner overlaid; now it holds — which also definitively kills the original flash, since autoplay's `song:play` can't fire while playback is held. Implemented as a small **core hook** `window.feedBack.holdAutoplay()` (mirrors the existing `holdAutoExit()`): a plugin claims it **synchronously on `song:loading`** (so it beats the `song:ready` autostart), and `release()` — or a **12-second fail-open backstop** — runs the deferred start. **Generation-guarded** (a new song invalidates a stale hold) and **fail-open** (a wedged or crashed plugin can never permanently strand a song); **manual Play always wins** (it doesn't flow through the autostart path). The tuner claims the gate only when the feature is on, and **releases it the instant** it decides not to open (song already covered / tuning unchanged) or when you Skip. Touches core `static/app.js` (the hook + an autostart refactor) and the tuner plugin (`plugins/tuner/screen.js` — the claim/release; `plugins/tuner/utils/ui.js` — the Skip / Back-to-library buttons, × dropped on auto-open); the hook is generic and shell-agnostic (a test asserts `app.js` still doesn't reference the tuner's internals). Tests: `tests/js/tuner_auto_open.test.js` (claim on `song:loading`, release on dismiss, feature-off no-claim, the core hook + fail-open backstop, the Skip / Back-to-library / Esc escape-hatch) + a `speed_reset.test.js` stub. ⚠️ **Needs a manual smoke-test before shipping** — this is a core playback change; verify on desktop that the tuner mic doesn't contend with note_detect's scoring input (ASIO/exclusive mode), per the design charrette.
- **v3 Songs List View: favoriting a song now turns the heart red immediately (no re-search needed).** In the tree / "List View" (Songs → List → expand an artist), clicking the heart flipped the glyph ♡→♥ but it stayed dim grey until you re-searched the library — reported on macOS + Windows, open since 0.3.0 / 2026-06-25. One shared `wireCards()` `[data-fav]` handler (`static/v3/songs.js`) serves both the grid card and the List-View row, but the two render with different idle colours — grid `text-white`, List View `text-fb-textDim` — and the handler only ever removed the grid's `text-white`. So in List View the row kept `text-fb-textDim` alongside the freshly-added `text-fb-accent`, and the dim class won by CSS source order (glyph changed, colour didn't). Each heart now declares its idle colour via a `data-fav-idle` attribute and the handler swaps exactly that class, so only one colour class is ever present; the handler also writes the new state back onto the in-memory song model so a re-render / virtualized-grid recycle agrees instead of reverting. Tests: `tests/js/v3_favorites_toggle.test.js`.
- **v3 Songs A–Z rail: taps now land reliably, a drag releases exactly on the let-go letter, and the rail is large enough to hit on hi-res displays.** Follow-up to the rail's debut (#634); three bugs reported on macOS + Windows (0.3.0, 2026-06-29): a tap often did nothing ("clicked O, nothing happened"), a drag "got you kind of there but where you release isn't where you get sent," and the rail was "way too small" at 1440p and didn't scale with resolution. Root causes & fixes, all in `static/v3/songs.js` + `static/v3/v3.css` (`bindRailOnce`/`jumpToLetter`/`.v3-azrail`): (1) **taps** — `pointerdown` calls `setPointerCapture`, after which the browser **retargets the follow-up `click` to the rail container**, so the click handler's `closest('.v3-azrail-letter')` resolved `null` and a plain tap (no `pointermove`) had no other path → no-op. The jump is now driven from `pointerdown` itself (seek on press); the `click` handler is reduced to **keyboard activation only** (`e.detail === 0`, Enter/Space). (2) **drag precision** — every letter crossed fired `jumpToLetter` with `behavior:'smooth'`; stacked smooth-scroll animations over the virtualized grid lagged and settled short of the release. `jumpToLetter(letter, smooth)` now scrolls **instantly while scrubbing** (`'auto'`) and only animates discrete taps/keyboard jumps, so the grid tracks the finger and the release lands on the let-go letter. (3) **size** — the letters were a fixed `.62rem` glued at `right:2px` (~13px-tall target on the screen edge); they now scale with the viewport (`clamp(.72rem, 1.4vh, 1.05rem)`), sit off the edge with taller/wider equal-width hit targets and a hover/active highlight so the scrub target is visible. Keyboard arrow-nav + the present-letter gating are unchanged. Reported by =Scr4tch= and MajorMokoto. Tests: `tests/js/v3_az_rail.test.js` (pointerdown-seek, keyboard-only click guard, instant-vs-smooth scroll).
- **v3 player: opening another rail popover now closes the Section Practice popover (no more two stacked popovers).** Opening the **Practice** pill's popover and then clicking a different player-rail icon (e.g. **Plugins**) left the Practice popover open underneath the new one — looked broken (reported on macOS, 0.3.0 / 2026-06-28). The rail icons call `e.stopPropagation()` in their click handler (`static/v3/player-chrome.js`), which killed bubbling before it reached the Practice popover's outside-click dismiss bound on `document`. The dismiss (`_installSectionPracticeDismiss` in `static/app.js`) now binds in the **capture phase**, which runs before the target's handler so a descendant's `stopPropagation()` can't swallow it — mirroring how the audio-mixer popover already dismisses. Esc handling stays bubble-phase (the player's Escape-to-exit ordering is unchanged). v2 shares `app.js` and is only hardened (no rail `stopPropagation` there). Tests: `tests/js/section_practice_dismiss.test.js`.
- **v3 UI no longer lets you accidentally text-select the chrome.** Dragging or double-clicking across the interface used to marquee-highlight buttons, labels, the sidebar, the transport, and the note-highway HUD — which looks broken (reported on Mac + Windows). The v3 shell now defaults to `user-select: none` on `html` (`static/v3/v3.css`), then opts *content* back in — so chrome is non-selectable but the text you actually copy still works. Decided by a 4-lens panel (UX / accessibility / dev-ops / plugin-ecosystem); the guardrails are deliberate: **form fields are always re-enabled** (never break the caret / IME — no `* { user-select:none }`, which trips a WebKit input bug); **plugin screens (`.screen[id^="plugin-"]`) stay selectable by default** so a plugin's copyable text (lyrics, chord names, results) — including community plugins that don't know about this — isn't silently locked; and **core read-only content opts back in by container** via a new hand-authored **`.fb-selectable`** class — applied to the whole **Settings** panel (paths, device names, version, diagnostics, About — answering "is settings still copyable?": yes), the **now-playing song metadata** (with `pointer-events` re-enabled so the HUD text is actually reachable), and the focused **modals / dialogs / toasts / scan banner** that carry copyable errors, IDs, paths, and file names. It's cosmetic only (it protects nothing) and never used to lock copy-worthy text — errors, IDs, paths, versions, and metadata stay selectable per WCAG 2.2 (copy-paste as a permitted mechanism). Dense card lists (library grid, dashboard, profile) stay non-selectable by design — making them selectable would reintroduce the marquee-mess across cards. **v3-only** (v2 unchanged); plain CSS, no Tailwind rebuild; no desktop changes (standard OS-framed window). Plugin authors: `.fb-selectable` is documented in `CLAUDE.md` for re-enabling copyable content rendered outside a plugin screen. Tests: `tests/js/v3_user_select_policy.test.js`.
- **Input-setup wizard no longer collapses an audio device's driver-type variants into one entry.** On Windows the desktop engine enumerates the same interface once per host API (ASIO / Windows Audio / DirectSound), and the wizard's audio picker (`plugins/input_setup/screen.js`) de-duped the source list by display **label** — so the variants (which share a name) collapsed to a single choice, silently keeping whichever sorted first (often *not* the low-latency ASIO one the player wants). The audio-input capability already collapses true duplicates by `logicalSourceKey` (`_visibleInputSources` in `static/capabilities/audio-session.js`), and the variants each have a **distinct** key, so the wizard's extra label-collapse was redundant for real dupes and destructive for these — it also could drop the variant that was actually `selected`. Removed it; the picker now lists every selectable input. Pairs with feedBack-desktop's change to label each source with its driver type (e.g. "Focusrite (ASIO)") so the now-distinct entries are legible.
- **3D Highway FPS counter no longer hides behind the v3 "Up Next" pill.** The on-highway FPS readout (Settings → Graphics → 3D Highway → Show FPS counter) is pinned to the top-right of the highway overlay — the same corner the v3 player chrome stacks its persistent **Up Next** pill and live-performance HUD into, on a higher layer that paints over the canvas. So the readout sat *behind* that chrome and couldn't be read — precisely when a tester had turned it on to judge performance (it also made the separate "Up Next won't turn off" complaint worse, since the default-on pill covered the counter regardless). The counter now stays top-right but drops just **below** whichever of that chrome is showing: `highway_3d`'s `screen.js` measures the lowest visible top-right v3 HUD element (`#v3-upnext` / `#v3-live-performance-hud` / `#hud-time`) and floors the FPS box's Y beneath it. Element refs are resolved once and cached (no per-frame `querySelector`, per the plugin perf rules) and only consulted while the counter is actually drawn; gated on `window.feedBack.uiVersion === 'v3'` so the classic (v2) UI is byte-for-byte unaffected. `plugins/highway_3d/plugin.json` version → `3.30.1` (cache-buster). (For reading raw perf numbers unobstructed, the core perf HUD — `localStorage.highwayPerfHud='1'` — still renders above all chrome and additionally shows the adaptive render-scale.)
- **3D Highway fret-number row no longer clips off the bottom edge when the camera zooms in on a centred span.** The heat-coloured fret-number row is drawn as a band *below* the board (`sY(lowest) − S_GAP*1.4`), but the camera's self-correcting framing only anchors the board **centre** to the lower third of the screen — it reserved no headroom for that row. So a tight zoom on a centred active span (worst around mid-neck; fine when the span sits at either end of the neck, which is why testers saw it "only when centered" and "not every song") dropped the numbers past the bottom edge. Tilt can't fix it there (it would only trade a bottom clip for a top clip), so `camUpdate()` now **dollies the camera back just enough to bring the row back into frame**: it projects the row band with the final camera and, when it falls below a safe NDC line (`FRET_ROW_FIT_NDC_MIN`), raises a capped, hysteretic `_fretRowFitBoost` applied to the `curDist` lerp target (the span-driven zoom still owns zooming *in*). The boost rises promptly (proportional to the deficit), relaxes lazily past a deadband, and is capped (`FRET_ROW_FIT_BOOST_MAX`, +60%) so the zoom can't pop or hunt; it cooperates with the tilt loop (pull-back shrinks the scene, tilt keeps the centre anchored) and yields entirely to the Camera Director free-cam. Surgical: passages where the row is already visible never trigger it, so framing is unchanged everywhere it already worked. `plugins/highway_3d/plugin.json` version → `3.30.2` (cache-buster). Tests: `tests/js/highway_3d_camera_framing.test.js` (guard constants, the boosted `curDist` lerp, the projected-row hysteresis, free-cam yield).
- **v3 Songs grid now refreshes after a Settings rescan / DLC-folder change — no app restart needed.** On a fresh install, pointing at a DLC folder in Settings and running a scan left the Songs section empty until a restart (the scan *did* populate the library — `_background_scan` re-reads `config.json` fresh — but the v3 grid never reloaded). The Settings **Rescan / Full Rescan** handlers only refreshed the classic (v2) library via `loadLibrary()`; the v3 grid (`static/v3/songs.js`) had no listener for a scan it didn't start itself (only its own upload path self-refreshed via `watchUploadScan`), so its cached, pre-DLC (empty) DOM/snapshot survived a sidebar return until a full reload. The rescan handlers now emit a **`library:changed`** event (`static/app.js`); the v3 grid listens and **reloads if it's the active screen, else marks itself dirty** so the next entry does a full re-fetch instead of restoring the stale snapshot (a `_libraryDirty` short-circuit ahead of every cached-DOM fast-path in `onV3SongsScreenEnter`). Tests: `tests/js/v3_library_refresh.test.js` (the emit + the reload/dirty wiring).
- **Edit Metadata modal: the Year is now editable.** You could set a year when authoring a pak but the Songs → Edit Metadata modal had no Year field, so it could never be changed afterward. The backend (`POST /api/song/<f>/meta`) already accepted and normalized `year` (writes it into the file via `songmeta`, survives a rescan) — only the UI omitted it. Added a **Year** input to `openEditModal()` (populated from the song's existing year) and included `year` in `saveEditModal()`'s POST body (`static/app.js`). Both the v3 card menu and the legacy edit button already pass the year through, so both surfaces get the field.
- **Edit Metadata modal no longer closes when a click-drag is released on the backdrop.** Selecting text inside a field and releasing the mouse past the modal's edge dismissed the form without warning (the `click` event's target resolved to the backdrop), discarding the edit. Backdrop dismissal now requires the **mousedown to have started on the backdrop** too — tracked per-modal and decided by a new pure `_editModalShouldClose(clickTarget, modalEl, downOnBackdrop)` helper (`static/app.js`). Cancel / ✕ still close on a normal click. Tests: `tests/js/edit_metadata_modal.test.js` (year in the POST body + the backdrop-close decision table).
- **Built-in diagnostic sloppak rebranded "Slopsmith" → "FeedBack" in the song name.** PR #586 renamed the file to `feedBack-diagnostic-basic-guitar.sloppak` but never regenerated the archive, so the manifest inside still carried `title: Slopsmith Diagnostic — Basic Guitar` / `artist: Slopsmith` (and the same heading in `DIAGNOSTIC.md`) — the stale name testers saw in the library/player and the onboarding calibration step, even though the build script, server, and docs all already say "FeedBack Diagnostic — Basic Guitar". Regenerated `docs/diagnostics/feedBack-diagnostic-basic-guitar.sloppak` from `docs/diagnostics/build_diagnostic_basic_guitar.py` so the committed artifact matches its source generator (title/artist/heading now "FeedBack"; chart, stem, and `diagnostic:` metadata unchanged). No code change — the rename in #586 just needed the rebuild.
- **v3 song/lesson accuracy badges now refresh on the first return from a song — no restart needed.** PR #574 added a `stats:recorded` → in-place badge repaint, but the repaint never matched a card. The event (like `song:loading`) carries the filename **`encodeURIComponent`'d** — exactly as `playCard` hands it to `playSong` (the highway WS `decodeURIComponent`s it back) — whereas library cards key on the **decoded** `localFilename` (`data-fn`), and `/api/stats/best` is server-canonicalized to that same decoded key (`server.py` `_canonical_song_filename`). So `repaintAccuracy`'s `data-fn !== key` check rejected every card and `state.accuracy[encoded]` was `undefined`, leaving the just-earned badge stale until a full `render()` (app restart / search / re-enter the screen) — which is why it "came back after a restart." `static/v3/songs.js` now decodes the `stats:recorded` filename back into the card / `state.accuracy` key space via a small `decFn` helper before marking dirty and repainting (idempotent for already-decoded names; falls back to the original on malformed input so a real filename containing a literal `%` is never corrupted), so both the immediate repaint and the `onV3SongsScreenEnter` deferred path land on the right card. Tests: `tests/js/v3_songs_score_badge_refresh.test.js`.
- **Escape now exits a song (and leaves Settings) even when a transport/rail control button holds keyboard focus.** Clicking a player control (Play / FF / RW / Restart) left that `<button>` focused, and `_shortcutDispatchBlocked()` in `static/app.js` treats any focused `INPUT/SELECT/TEXTAREA/BUTTON` as an "interactive control" and bails before the shortcut registry runs — so the player-scope `Escape → Back` shortcut never fired until the user clicked empty canvas to blur the control ("Escape in song not consistent"). Space already had a player-screen carve-out (#593) that let it fire through a focused control; Escape did not. Generalized that carve-out to Escape, scoped to the player **and** settings screens (both register an `Escape = Back` shortcut, and settings had the identical latent bug). The earlier guards are preserved and still win: text inputs are exempted first (Escape there clears/blurs the field), the Section Practice popover already claims Escape before the carve-out, and a true modal layered over the screen (`[role="dialog"][aria-modal="true"]` / `.feedBack-modal`) still traps Escape so it closes the modal rather than ejecting past it. Escape becomes a reliable, focus-independent "Back" — making it monotonic groundwork for an optional exit-confirm. Plugins that register a player-scope `Escape` shortcut benefit identically (they were broken the same way). Tests: `tests/browser/keyboard-shortcuts.spec.ts` (focused-button repro, text-input no-exit, no-escape-past-modal, Section Practice popover, settings twin-bug).
- **The v3 "Up Next" pill can now be turned off — new "Show 'Up Next'" gameplay toggle (default ON).** The v0.3.0 player chrome's persistent upcoming-section pill (`#v3-upnext`, drawn by `static/v3/player-chrome.js`'s `updateUpNext()`) shipped with no off switch, so it always showed during playback whenever a section was upcoming — overlapping the top-right FPS HUD and ignoring the 3D-highway "Show 'Up Next' section card" checkbox (a *different*, in-canvas widget that was demoted to default-off precisely because this pill is the canonical readout). Users reading the pill as the same setting saw "disabled in settings but still there." Adds a real core toggle following the `autoplayExit` idiom: a client-only `showUpNext` `localStorage` pref (absence = enabled), a **Show "Up Next"** switch in the Gameplay settings tab (`static/v3/index.html`), reader/writer + `loadSettings()` hydration + a read-only `window.feedBack.showUpNext` getter in `static/app.js`, and a gate at the top of `updateUpNext()` that hides the pill when off. Disabling mid-playback hides it immediately; re-enabling re-shows it on the next chrome tick (~6 Hz). Added to `RESET_MAP.gameplay.local` in `static/v3/settings.js` so the Gameplay "Reset" restores the default-on state. Default ON = zero change for existing users. No Tailwind rebuild (plain markup + existing classes).
- **v3 list/tree view brought to parity with the grid: select mode, parts chips, and song actions — plus a stale-CSS Docker fix.** Re-lands a previously-reverted change. **Frontend (`static/v3/songs.js`):** entering select mode no longer collapses the tree — `loadTree()` now captures the expanded artist groups (`details[open]` keyed by `data-artist`) before the "Loading…" wipe and restores them on rebuild, so toggling select mode (which re-renders via `reload()`) keeps groups open and selection usable; tree rows gain a display-only checkbox + selection ring, the same fav / save-for-later / overflow-menu cluster as the grid card (always shown, all bound by `wireCards()`), and a capture-phase select guard mirroring the grid so clicking a row or arrangement chip in select mode selects instead of playing (`<summary>` headers sit outside `[data-fn]`, so native expand/collapse is untouched). **Docker fix (`static/tailwind.min.css`):** the committed Tailwind stylesheet was stale — `.sm\:flex` (and the other utilities behind #582's `hidden sm:flex` arrangement chips and the new action cluster) were never compiled in, so they rendered `display:none` on the Docker build (which serves the committed CSS as-is; Desktop rebuilds from source so it looked fine). Regenerated with the pinned `tailwindcss@3.4.19` via `scripts/build-tailwind.sh` so Docker matches Desktop and #582's chips render on every Docker deploy. Regression tests: `tests/browser/v3-tree-select.spec.ts`.
- **Space bar now plays/pauses on the player screen even when a sidebar nav link or rail button has focus.** When any `<button>` in the player rail (viz, audio, mixer, lyrics, plugins, advanced), a sidebar nav link, or a popover control held keyboard focus, pressing Space was swallowed by `_shortcutDispatchBlocked` → `_isInsideInteractiveControl` (which treats `BUTTON`/`A` as interactive), so the Space shortcut never reached the dispatcher and `togglePlay()` never ran. `_shortcutDispatchBlocked` (`static/app.js`) now extends the same carve-out already used for the Section Practice bar: while the player screen is active, Space is always routed through the shortcut system — the dispatcher calls `e.preventDefault()` before invoking the handler, so the focused element does not also activate. Text inputs (`_isTextInput`) remain exempted first, so typing space in a search/input field still works normally, and focus inside a true modal dialog (`role="dialog" aria-modal="true"` / `.feedBack-modal`) layered over the player is also exempted so Space reaches the modal's focused control (e.g. its Close button) instead of toggling playback behind it — non-modal player popovers/toasts (loop A/B, arrangement pin) stay covered. Regression tests in `tests/browser/keyboard-shortcuts.spec.ts` cover the focused-rail-button play/pause, the text-input exemption, and the modal-dialog exemption.
- **A song's accuracy badge now updates on its library card right after you play it — no restart needed.** The v3 library (`static/v3/songs.js`) loaded the best-accuracy map (`/api/stats/best`) once into `state.accuracy` at render time and only ever refreshed it on a full re-render; the play→return flow takes the screen-entry fast-path that restores the cached grid DOM without re-fetching, so a just-earned score stayed invisible until the next restart re-ran `render()`. The `stats-recorder` now emits a `stats:recorded` event (carrying `filename`/`arrangement`) once the scored `POST /api/stats` resolves on the server — the correct moment, since `song:stop` fires before the POST completes. `songs.js` listens: if the library is the active screen it re-fetches `/api/stats/best` and patches the affected card/row badge in place; otherwise it marks the filename dirty and `onV3SongsScreenEnter` applies it on return (a failed fetch keeps the entry dirty to retry). Badge markup was factored into a shared `accuracyBadge(filename, variant)` (grid pill + tree-row percentage, both tagged `.fb-acc-badge`) so the in-place `repaintAccuracy` can find and replace them without a full list re-render (scroll/pagination preserved). The old empty `song:stop` "refresh lazily next render" placeholder is replaced.
- **Changing Settings → 3D Highway → Fret spacing no longer ejects you to the home screen.** The `highway_3d` plugin's `h3dSetFretSpacing` was the lone 3D-highway setting that called `location.reload()` to apply — and since the SPA boots with `#home` as the active screen (`index.html` `.screen.active`), the reload dropped the user out of Settings onto the homescreen. It now applies live like every other 3D-highway setting: it rebinds the module-scope `_h3dFretUniform` flag (so panels mounted later this session pick up the new mode), recomputes the two `fretX`-derived scalars that were baked at init (`_fretLabelScaleRefW` for fret-label sprite scaling, `FRET_WIDTH_MID` for camera hysteresis), and broadcasts a `fretSpacing` change over the existing `_bgEmitChange` pub-sub so every mounted panel rebuilds its board via `buildBoard()`. Per-frame note geometry already reads `fretX` live and needs no rebuild. No page reload, so the Settings screen stays put. Source-level regression tests in `tests/js/highway_3d_fret_spacing.test.js` now pin the no-reload / live-rebuild behavior.
- **v3 library scroll-restore no longer breaks the classic v2 UI or drops off-screen searches** (feedBack#857). Two regressions in the scroll-restore work above: (1) `playSong` remapped `home`-launched songs to return to the `#v3-songs` screen unconditionally, but `static/app.js` is shared with the v2 UI (served at `/v2` / `FEEDBACK_UI=v2`) where that screen does not exist — Esc-from-player then called `showScreen('v3-songs')`, which threw on the missing element and stranded the user on a blank screen with playback still running; the remap now applies only when `#v3-songs` is present. (2) The Songs screen-entry fast-path skips reloading to preserve scroll, but the global topbar search routed through it, so once Songs had been visited, searching from another screen navigated there without applying the new query; the screen now tracks the state hash each fetch reflects and refetches when it has drifted, keeping the scroll-preserving no-op only when nothing changed.
- **An active custom highway renderer is no longer starved of `draw()` when it hides the canvas** (#819). The per-frame draw gate in `static/highway.js` bailed on `if (!_lastVisible) return`, which conflated two different "hidden" states: a genuine off-screen canvas (`offsetParent === null` — navigate-away / `display:none` splitscreen panel, #246) versus a renderer-set *override-hide* (`setVisible(false)`, where an opaque overlay covers the canvas but the active renderer keeps painting its own surface). The gate now only pauses everything for the off-screen case (and still pauses the default 2D renderer on an override-hide); the **active custom renderer** keeps receiving `draw()` through its own override-hide. The `highway:visibility` event still fires before the gate, so sibling overlay renderers (e.g. 3D Highway's `.h3d-wrap`) still pause. This is the core-side root cause behind the Tab View cursor freezing in single-player (feedBack#734; worked around plugin-side in feedBack-plugin-tabview#25).
- **Screensaver no longer kicks in during windowed-mode playback** (#686). While a song is playing, `static/app.js` now holds a [Screen Wake Lock](https://developer.mozilla.org/en-US/docs/Web/API/Screen_Wake_Lock_API) (`navigator.wakeLock.request('screen')`) so the OS display/screensaver stays awake even though only audio + the highway animation are active and the keyboard/mouse are idle. The lock is acquired on `song:play`/`song:resume` and released on `song:pause`/`song:ended`/`song:stop` (kept only while actually playing), and re-acquired on `visibilitychange` when the tab refocuses (the API auto-releases a lock whenever the page is hidden). Both the HTML5 `<audio>` and JUCE desktop playback paths emit the same `song:*` events, so the fix covers both. In feedBack-desktop (Electron), where `navigator.wakeLock` is unreliable, it also drives a native `powerSaveBlocker` bridge via the optional `window.feedBackDesktop.power.setScreenAwake` hook when present; both calls degrade silently where unsupported. Note: the browser Wake Lock API is secure-context only, so in a plain browser this is active on `localhost` / HTTPS only — a session opened over plain HTTP to a LAN IP (e.g. `http://192.168.1.100:8000`) won't keep the screen awake; front it with HTTPS or use the desktop app (see README → reverse-proxy notes).
### Changed
- **Practice plugin first-class sidebar slot now points at Virtuoso.** The bundled practice plugin was rebranded/re-homed from the SlopScale fork (`id: slopscale`) to `got-feedback/feedBack-plugin-virtuoso` (`id: virtuoso`); the desktop bundle swap is feedBack-desktop#31. `static/v3/shell.js` still promoted `slopscale`, whose id no longer ships — so `renderPromotedNav()` (gated on the plugin appearing in `/api/plugins`) would have found no match and the dedicated sidebar slot would have gone dark, dropping Virtuoso to the generic Plugins gallery. Update the NAV entry + `PROMOTED_PLUGINS` slot `slopscale` → `virtuoso` (`screen: plugin-virtuoso`, label "Virtuoso - Practice", same FeedBarcade anchor + `target` icon) so the practice plugin keeps its first-class entry. Also clear the now-dead `slopscale` id from the Plugins-gallery curated category map (`static/v3/plugins-page.js`) and add `virtuoso: 'practice'` as a defensive fallback (the manifest's `category: "practice"` is authoritative, so it lands on the practice board regardless), and refresh the stale SlopScale references in `README.md` + `docs/plugin-capability-inventory.md`. Must land with the bundle swap or the practice plugin regresses in the UI.
- **3D highway: realistic curved metal frets.** The fret wires are now bowed `TubeGeometry` (the middle strings push away from the camera so the row of frets reads as wrapping a cylindrical neck — a depth cue) rendered with a lit `MeshStandardMaterial` instead of the old flat, straight `MeshBasicMaterial` boxes, so the scene's ambient + directional light glints across the rounded surface for a polished-steel look. The existing per-frame highlight is preserved unchanged: frets inside the active anchor still turn gold (`0xD8A636`), which under the metallic shading reads as brass. Metalness is kept moderate (0.4, not full-metal) because the scene has no envMap — a PBR full-metal surface would reflect black — with a dim emissive floor so frets stay legible down the fogged neck. Backported from the `highway_babylon` plugin's "hit-zone fret bars". All knobs (`FRET_BOW_DZ`, metalness/roughness/emissive) are tunable constants. `plugins/highway_3d` v3.25.0.
- **3D highway: section + tone HUD cards now default OFF.** The v0.3.0 player chrome carries a persistent "Up Next" pill, making the in-canvas section card redundant by default (it doubled the readout, feedBack feedback); the tone HUD follows the same less-is-more default. Both remain available in Settings → 3D Highway (visibility/position/size unchanged); users who previously toggled either explicitly keep their stored preference — only the untouched default flips. `plugins/highway_3d` v3.24.1.
- **Perf**: replace runtime Tailwind Play CDN with a prebuilt static stylesheet (`static/tailwind.min.css`). The Play CDN's runtime JIT scanned the DOM ~1.8x/sec on the main thread (~37 ms blocking spans), dropping ~26% of frames in long playback sessions with the 3D highway as default. Theme extensions (dark/accent/gold colors, Inter font) move to `tailwind.config.js`; regen via `bash scripts/build-tailwind.sh`. No runtime build step — the generated CSS is committed. Fixes feedBack-desktop#110.
- **Perf**: reduce per-frame allocations in the 2D highway chord + lyric render paths. `_ensureChordRenderCache` now also caches `sortedNotes` / `nonZeroNotes` / `nonZeroFrets` / `allMuted` / `hasMultipleNotes` (computed once per chord, invalidated on `src` / `_inverted` / `chordTemplates` change — the third key catches a stale `isOpen`-derived classification when the WS `chord_templates` message lands after the final `chords` chunk), so `drawChords` no longer re-sorts / re-filters / spreads min-max per visible chord per frame. The in-chord unison bend classification is folded inline (no `chordPositions.filter` × 2 per frame). `drawLyrics` memoizes `ctx.measureText` results in a two-level `Map<fontSize, Map<text, width>>` so cache hits don't allocate a composite string key. Lit-sustain shimmer in `drawSustains` swaps the 4 per-note-per-frame `Math.random()` calls for a 64-entry precomputed jitter LUT (xorshift32-seeded — the LUT contents are reload-stable and test-reproducible; rendered shimmer is deterministic per `createHighway()` instance, since the seed includes that instance's `_frameIdx`) indexed by `(frameIdx + n.s + ⌊n.t·60⌋)`, visually indistinguishable and allocation-free.
- **Perf**: the load-adaptive render scale (`_adaptRenderScale`, #654) no longer visibly hunts up/down on passages that hover near the frame budget (testers saw "quality going up and down" with the 3D highway). Downscaling stays prompt to protect the frame rate, but **upscaling is now lazy**: a smaller step (×1.06 vs ×1.1) on a longer cooldown (`_AUTO_UPSCALE_COOLDOWN_MS` 2500 ms vs the 600 ms adjust cooldown), reset on any downscale, and gated by a predictive guard — it only upscales when the projected cost *after* the step (≈ cost × step², since draw cost tracks the pixel count) still clears the high budget. The scale therefore settles just inside the 7–12 ms deadband instead of oscillating across it. No new public API; the `_autoScaleMin` "Min res" floor is unchanged.
### Removed
- **`c` library hotkey ("Convert to .sloppak") removed from core.** Core hardcoded a plugin-specific shortcut: a documentation-only `registerShortcut({ key: 'c', scope: 'library' })` no-op plus a `c → button.sloppak-convert-btn` entry in the library keydown handler that fired the Sloppak Converter plugin's button. Per the plugins-own-their-behavior principle, core no longer ships this hotkey — the convert button still works by click, and the Sloppak Converter plugin can register its own `c` shortcut via `window.registerShortcut()` if keyboard access is wanted. The `f` (favorite) and `e` (edit) library hotkeys, which drive core buttons, are unchanged. Help-modal/registry tests in `tests/browser/keyboard-shortcuts.spec.ts` updated to drop the `c` assertions.
### Added
- **Player progression: Mastery Rank, instrument-path challenges, daily/weekly quests, Decibels currency, cosmetics shop (spec 010).** Onboarding gains two steps: pick one or more **instrument paths** (Guitar / Bass / Drums — data-driven, more can ship as content) and a **calibration challenge** offer (play the bundled FeedBack Diagnostic with note detection at 100% accuracy — or skip; either way you reach **Mastery Rank 1**, and a skipped calibration can still be completed later from the Progress screen). Each path levels by completing a content-defined number of **challenges** (any order) from that level's set; Mastery Rank = onboarding rank + the sum of path levels, starting at 0 on a fresh install. The existing unified XP backend is untouched but the frontend renames it to **Decibels (dB)** — a spendable currency earned ONLY by playing (songs, FeedBarcade rounds, quest rewards; no real-money path exists or may be added) — with spend tracked in a separate wallet so lifetime earnings stay monotonic. Rotating **daily/weekly quests** (deterministic per period, lazy instantiation, local-midnight / Monday resets) award dB and feed `quest_completed` challenges. A new **Progress** screen (rank hero, per-path checklists, quest countdowns, add-a-path) and **Shop** screen (themes via CSS-variable swaps under `html[data-fb-theme]`, avatar frames; atomic balance-checked purchases — 402 on insufficient dB, 409 on re-buy) join the v3 nav, and the topbar badge now shows Rank + challenge-set progress + dB balance. All definitions live in `data/progression/` JSON (paths/levels/challenges, quest pools, shop catalog) — adding a rank, challenge, quest, or cosmetic is a content edit + restart, never code; invalid content degrades to logged warnings. New tables (additive + idempotent): `progression_state`, `player_paths`, `challenge_progress`, `quest_state`, `wallet`, `shop_owned`, `shop_equipped`. New endpoints: `GET /api/progression`, `POST /api/progression/paths|onboarding|events` (events whitelists `minigame_run`; `song_completed` stays server-derived inside `POST /api/stats`, which now resolves the instrument server-side and reports an additive `progression` outcome key), `GET /api/shop`, `POST /api/shop/buy|equip`; equipped cosmetics ride along on `GET /api/profile`. A new **`progression` capability domain** (core-owned, kind: command, safety: safe — `inspect`, `record-event`, `list-shop`, `buy-item`/`equip-item` gated on user action) emits `challenge-completed`/`quest-completed`/`path-level-up`/`rank-changed`/`db-changed`/`calibration-completed`/`cosmetic-equipped`, mirrored as `progression:*` window events, with a redaction-safe diagnostics contributor; backend plugins get the symmetric `record_progression_event` context hook (the bundled minigames hub reports runs through it, guarded for standalone). Spec: `specs/010-progression-domain/`. Tests: `tests/test_progression.py`, `tests/test_progression_api.py`. **Migration notes:** existing XP totals carry over as lifetime dB (balance = lifetime − spent); resetting a per-source XP contribution (e.g. a minigames profile reset) after spending can clamp the spendable balance to 0 until new dB is earned; drums-path v1 content uses currently-satisfiable goals (arcade rounds, quests, any-instrument plays) until drums scoring lands.
- **3D highway score FX (notedetect game-scoring layer).** The bundled `highway_3d` renderer now visualizes the scoring layer shipped in feedBack-plugin-notedetect ≥1.13: floating **"+N" score pops** above each judged gem (sourced from the note-state provider's new `{ points, mult, popKey }` verdict fields — chord members share the chord-level `popKey`, so a chord pops once, not once per string), plus session-level FX from the new `notedetect:fx` event — a particle burst at the strike line on streak milestones (25/50/every 100), an expanding ring pulse on multiplier tier-ups (×2/×3/×4), and a brief red wash when a ≥10 streak breaks. Colors and the pop font follow the user's notedetect scoring-UI skin (`feedBack_notedetect_skin`: neon/esports/metal, refreshed live on the `notedetect:skin` bus event). Everything renders on the existing 2D overlay canvas from fixed-size slot pools — no Three.js geometry, no text-sprite cache traffic, near-zero cost when idle — and degrades to a silent no-op with older notedetect builds (the new fields/events simply never arrive). Splitscreen panels scope FX to their own detector instance via the bubbling per-panel `notedetect:fx` dispatch. `plugins/highway_3d` v3.24.1.
- **3D highway: slide direction arrows + gem-follow animation.** Slide notes now show a ›/‹ arrow indicating which way the slide goes — on the note/gem itself, as an early preview on the neck before the note arrives, and (optionally) chained further ahead for multi-leg slides — each independently toggleable in Settings → 3D Highway (`slideArrowApproachVisible`, `slideArrowNeckVisible`, `slideArrowChainPreviewVisible`). The note gem also now visually glides from its starting fret to the slide's destination over the note's sustain and holds there through the brief post-sustain linger, instead of snapping back to the starting fret — most noticeable on unpitched "slide to nothing" notes. `plugins/highway_3d` v3.25.2.
- **3D highway: up to 3 upcoming-note ghost previews per string, with fade-in/grow.** Each string now previews up to 3 upcoming notes (was 1) on a fixed 0.6 s fade-in/grow ramp, so tight same-string runs no longer pop in at full size right before impact and the player can read note order ahead of time. `isBlocked` (the pre-impact ghost suppression in a note's last 150 ms) is now scoped to chord notes only — for lead notes it had been blinking the ghost out right before each sustained note in dense runs. (Slide notes stay excluded too, per the slide-arrow work above, since their gem glides off the start fret.) `plugins/highway_3d` v3.26.0.
- **Enable/disable plugins from the v3 Pedalboard (footswitch backend).** Every `/api/plugins` entry now carries an `enabled` boolean (default `true`), and a new `POST /api/plugins/{plugin_id}/enabled` endpoint (`{"enabled": <bool>}` → `{"id", "enabled"}`) persists the choice to `CONFIG_DIR/plugin_state.json` (only non-default `enabled:false` entries are stored; a missing/corrupt file is tolerated and never crashes startup). The loader **skips disabled plugins at startup** — no requirements install, no `routes.setup()`, no screen/nav/capabilities — while still surfacing them in `/api/plugins` as a disabled entry (`status:"disabled"`, `enabled:false`) so the UI can show an "off" pedal you can switch back on. Toggling persists immediately and flips the in-memory flag so the next `/api/plugins` reflects it at once (a runtime-disabled plugin's already-mounted routes/screen remain until the next restart; re-enabling a startup-skipped plugin mounts on restart). A disabled plugin is **excluded from the capability pipeline** — its capability metadata is emptied in `/api/plugins`. Guard rails keep `capability_inspector` and `app_tour_*` always enabled (disable → `400`); unknown id → `404`; missing/non-boolean `enabled` → `400`. Backend only; the v3 Pedalboard frontend consumes this contract. Docs: `docs/plugin-v3-ui.md`.
- **fee[dB]ack v0.3.0 rebrand + UI redesign (opt-in, isolated).** The visible product is being renamed from **FeedBack** to **fee[dB]ack** (the `[dB]` is a decibel pun on the practice "feedback" loop) alongside a full dashboard-style UI redesign. **Rebrand scope is the app + docs wordmark only** — the repository, Python package, ghcr Docker image, `CONFIG_DIR`, and `FEEDBACK_*` env vars all keep the `feedBack` name, so existing deployments and data are unaffected. The redesigned UI is additive and served behind a feature flag: `FEEDBACK_UI=v3` flips the `/` route to the new `static/v3/` shell, and `GET /v3` always serves it; the default `/` stays byte-identical to 0.2.9 until 0.3.0 flips the default. This release adds the `static/v3/` scaffold (navy app shell + styled fee[dB]ack wordmark, brand SVG + favicon + PWA manifest with 192/512 icons), an additive `fb` Tailwind color palette (legacy `dark`/`accent`/`gold` retained) with `static/v3/**` in the content globs, and the regenerated `static/tailwind.min.css`. Vanilla JS, prebuilt Tailwind, no Play CDN (Principle II). Shell wiring, screens, profile/scoring backends, and capability-runtime integration land in subsequent v0.3.0 changes.
- **fee[dB]ack v0.3.0 app shell (sidebar + topbar + routing).** The v3 shell (`static/v3/index.html`) is now a re-chromed copy of the legacy app: the new left **sidebar** (HOME / LIBRARY groups) and **topbar** (secondary nav, search, Support, badge-cluster mount points) replace the hidden legacy navbar, and new `#v3-*` screens (dashboard/plugins/profile/playlists/saved) are added — while all legacy screens (`#home` library, `#favorites`, `#settings`, `#player`, `#audio`, plugin nav containers) are kept verbatim so `static/app.js` boots **unmodified** and the whole engine (player/highway, plugin loader, capabilities, audio, library, settings) is reused as-is. Navigation is the shared `window.showScreen` across `#v3-*`, reused legacy, and `#plugin-*` screens, with a responsive hamburger and a `localStorage`/`v3:`-namespaced shell. Plugin nav is mirrored into the sidebar from `/api/plugins` (UI placement is a deferred capability domain, so this uses the legacy loader, not capability dispatch). `static/v3/shell.js` wraps `window.showScreen` via the idempotent rehydration pattern to keep sidebar/topbar active-state in sync.
- **fee[dB]ack v0.3.0 player profile + first-run onboarding + unified XP + streak.** Adds a single-user core **profile** (`profile`/`profile_progress`/`xp_profile` tables in `web_library.db`, additive + idempotent): display name + avatar, a stable `player_hash` (SHA-256 of the first name + a once-generated salt — stable across later renames; a future-leaderboard label, never auth), and a **streak** (any session on a calendar day keeps it; a missed day resets to 1). New endpoints: `GET/POST /api/profile`, `POST /api/profile/avatar` (base64, re-encoded to a ≤512px PNG under `CONFIG_DIR/avatars/`), `GET /api/profile/avatar/{name}` (safe-joined), `GET /api/profile/avatars` (bundled defaults under `static/v3/avatars/`), `GET /api/profile/progress` (one call for the badge), and `POST /api/xp/award`. **Unified XP:** `lib/xp.py` is the single XP curve (same math the minigames plugin shipped); the core `xp_profile` store is the one source of truth the profile badge reads, exposed to plugins via `context["award_xp"]`/`get_xp_progress`/`seed_xp`. The bundled **minigames** plugin now delegates XP to the core store (seeding once from its existing `profile.json` so earned levels carry over) — so song-play, minigames, and tutorials all feed one level. Frontend: a blocking first-run onboarding overlay (name + avatar grid + upload), the topbar profile badge (avatar, 🔥 streak, level + XP bar), and the `#v3-profile` screen. Tests: `tests/test_xp.py`, `tests/test_profile_api.py`.
- **fee[dB]ack v0.3.0 song-stats store (best score + accuracy, plays, resume position).** A core `song_stats` table (`web_library.db`, additive + idempotent, PK `(filename, arrangement)`) records per-song/arrangement best/last score + accuracy, play count, and last position. Endpoints: `POST /api/stats` (scored session → `plays += 1`, `best_*` = max, `last_*` = new, plus **unified-XP award** `xp_for_run(score)` and a **streak** bump, both behind try/except so a side-effect failure never drops the stat write; or position-only `lastPlayPosition` → resume touch with no `plays` change), `GET /api/stats/{filename}` (aggregated across arrangements), `GET /api/stats/recent` (joined to song title/artist/art for "Jump back in"). Scoring stays frontend-driven: `static/v3/stats-recorder.js` tallies the `note:hit`/`note:miss` events the optional `feedBack-plugin-notedetect` already emits (and also accepts an explicit `note_detect:session-ended` summary), then POSTs on song end; it also persists resume position on pause/stop. No note-detect edit required — note-detection is a deferred capability domain, so the recorder uses those legacy events and degrades to "no accuracy" when the plugin isn't installed. Score/accuracy math is shared with the server via `lib/song_score.py`. Tests: `tests/test_song_score.py`, `tests/test_song_stats_api.py`.
- **fee[dB]ack v0.3.0 playlists, Saved for Later, and Continue-Playing.** Core playlist management (`playlists` + `playlist_songs` tables in `web_library.db`, additive + idempotent): create/rename/delete, add/remove/reorder songs, plus a reserved **Saved for Later** system playlist (created on first use; protected from rename/delete). Endpoints: `GET/POST /api/playlists`, `GET/PATCH/DELETE /api/playlists/{id}`, `POST /api/playlists/{id}/songs`, `DELETE /api/playlists/{id}/songs/{filename}`, `POST /api/playlists/{id}/reorder`, `POST /api/saved/toggle`, and `GET /api/session/continue` (derives the resume song + last position from `song_stats`, no new table). Frontend `static/v3/playlists.js` renders the `#v3-playlists` list + detail (drag-reorder, play, remove) and `#v3-saved`, and exposes `window.v3Saved.toggle()` for a "Save for later" affordance on song cards. Favorites reuse the existing favorites screen/API. Core REST, no capability domain. Tests: `tests/test_playlists_api.py`.
- **fee[dB]ack v0.3.0 Dashboard / Home.** The `#v3-home` dashboard (matching the v0.3.0 design target) composes the new backends: a "Welcome back, {name}!" banner with a patch-notes link (`/api/version`), a hero card (Start Playing / Create Lobby), a **Continue-Playing** card (`/api/session/continue` → art, tuning chip, 4-segment progress; click resumes via `playSong` + best-effort seek), a stats row (audio-routing widget placeholder until prompt 18, library count from `/api/library/stats`, plugins count from `/api/plugins` where `status==="ready"`), and a **Recently Played** grid (`/api/stats/recent`) with per-song accuracy badges (good/mid/low ramp). Each widget fetches + renders independently and degrades gracefully (missing/empty endpoint → placeholder, never blocks first paint). `static/v3/dashboard.js`; re-renders on return to Home and on profile update.
- **fee[dB]ack v0.3.0 tuner + instrument topbar badges.** The topbar gains an **instrument selector** (guitar/bass + string count + tuning + reference pitch) persisted via additive `/api/settings` fields (`reference_pitch` clamped 430–450, `instrument`, `string_count` 4–8, `tuning` name or semitone offsets); changing it emits `instrument:changed` so the note_detect scorer can re-tune (consumed once the external plugin adopts it). A **live tuner badge** stays idle until the user enables the mic (explicit gesture; `getUserMedia`), then shows a YIN note readout with a cents needle (green within ±5¢) using a new dependency-free `static/v3/tuner-core.js` (YIN + frequency→note/cents, honoring the reference pitch); clicking opens the full `feedBack-plugin-tuner` screen when installed. CPU-friendly (~20 Hz, paused when the tab is hidden, respects `prefers-reduced-motion`). Tests: `tests/test_settings_instrument.py`, `tests/js/tuner_core.test.js`.
- **fee[dB]ack v0.3.0 audio-routing widget (dashboard).** The dashboard's audio stat tile now reads the live audio session **through the capability runtime** — `audio-mix inspect` (route + faders + required kinds), `audio-input list-sources` (selected/available input), `audio-monitoring inspect` — and renders **Audio Input → VST/NAM/IR → Audio Output** with per-node state dots and a Connected/Not Connected line. It never touches `audio-mixer.js` internals or `nam_tone` routes directly; "Not Connected" is the honest browser default (no native route), and it degrades on `no-owner`/`no-handler`/`failed` or absent capabilities. Refreshes on `instrument:changed`, play/stop, capability audio events, and each Home visit. `static/v3/audio-routing.js`.
- **fee[dB]ack v0.3.0 Plugins page.** The `#v3-plugins` screen renders the enriched `/api/plugins`: a "{N} active" header (`status==="ready"`), a card grid per plugin (icon, name, version, status pill with the error on failed, capability summary badges — declared domains / validation warnings / unsupported versions / shim hits / bundled / type), an **Open →** action that navigates to the plugin's injected `#plugin-<id>` screen, and All/Bundled/Visualizations filters. Surfaces a deep-link to the bundled **Capability Inspector** rather than re-implementing the graph. No new backend. `static/v3/plugins-page.js`.
- **fee[dB]ack v0.3.0 Songs / Library screen (`#v3-songs`).** A native vanilla-JS library browser over the existing `/api/library*` endpoints: provider selector (via the `library` capability, not DOM scraping), grid + tree views, sort, format filter, a tri-state filter drawer (arrangements / stems / lyrics / tunings), topbar-driven search (debounced), infinite scroll, fb song cards with **accuracy badges** (good/mid/low ramp, batched via a new `GET /api/stats/best`), favorite + save-for-later affordances, and upload (reuses the existing uploader). The "Songs" sidebar nav now opens this screen. No regression to `/api/library*`. `static/v3/songs.js`.
- **`ui.library-card-injection` capability + native song-card actions (fee[dB]ack v0.3.0).** New core capability (`static/capabilities/library-card-actions.js`, owner `core.ui.library-card-injection`, exposed as `window.feedBack.libraryCardActions`) lets plugins **register** per-song library-card actions (id, label, placement, applicability, enabled state, run handler) with `action-registered`/`action-result` events — replacing the legacy `.song-card` DOM-injection pattern (roadmap domain #9, now delivered as a frontend host). The native Songs grid renders registered actions in each card's "⋮" menu; the built-in **Edit metadata** and **Convert to E Standard (retune)** actions ship through it (`static/v3/card-actions-core.js`, calling the existing `openEditModal`/`retuneSong` globals). Songs cards also gain **arrangement chips** (play a specific arrangement) and a **multi-select** mode with batch **Add to playlist** / **Save for Later**. Recipe in `docs/capability-recipes.md`; tests in `tests/js/library_card_actions.test.js`. Migrating the external card-action plugins (Sloppak Converter, Find More, editor) onto `register(...)` is a follow-up.
- **`centOffset` exposed via `getSongInfo()`** — the arrangement `<centOffset>` field (float, cents) is now parsed from all chart sources (loose folder XML, sloppak wire format) and sent as `centOffset` in the `song_info` WebSocket message. Plugins can read `getSongInfo().centOffset` to obtain the arrangement's pitch-shift offset — commonly `-1200.0` for extended-range bass (one octave down) or a small non-zero value for true-tuned content (e.g. A443 ≈ +11.8 cents). Defaults to `0.0` when absent.
- **`highway.getPhrases()` and `highway.getMastery()` public plugin API** — exposes phrase timing windows (`[{ index, start_time, end_time, max_difficulty }]`) and the current mastery slider value (`0..1`) as documented, stable plugin API. Both values were already in memory and reachable via internal names; this surfaces them with intent so plugins can implement section-aware logic (e.g. tracking accuracy per phrase, suppressing difficulty changes during a hard solo) without reaching into undocumented internals. Returns `null` when the song has no phrase data (GP imports, single-difficulty charts). Pair with the existing `hasPhraseData()` to gate phrase-aware code paths.
- **Tailwind freshness guard + wider plugin scan.** A new `tailwind-fresh` CI job (`.github/workflows/tests.yml`) rebuilds `static/tailwind.min.css` with the pinned `tailwindcss@3.4.19` and hard-fails on any diff, so the committed prebuilt stylesheet can no longer silently lag source (after PR #411 removed the runtime Play CDN, a stale file shipped unstyled elements with no guard). The `tailwind.config.js` plugin content glob is widened to `./plugins/**/*.{js,html}`, which also scans non-`screen.js` plugin JS (e.g. `plugins/app_tour_*/script.js`) that was previously invisible to the build. Regenerating under the wider glob is a no-op for runtime behaviour — it only adds classes that were already used in source. Groundwork for the plugin `styles` capability (constitution 1.1.0, Principle II): runtime-installed plugins ship their own compiled CSS rather than relying on core's build-time scan.
- **Plugin capability pipelines** — adds the first versioned capability coordination layer for plugin authors and support tooling. `/api/plugins` now exposes validated capability declarations, validation warnings, unsupported-version metadata, UI/runtime domain declarations, and compatibility shim summaries for legacy `nav` / `screen` / `settings` / `routes` / visualization surfaces. The browser runtime now tracks manifest participants separately from live handlers, explicit dispatch outcomes (`no-owner`, `no-handler`, `unsupported-command`, `incompatible-version`), claim lifecycle cleanup, manual override precedence, deterministic ownership conflicts, multi-provider ordering, shim hit counts, and a redaction-safe diagnostics snapshot capped at 64 KB. A bundled Capability Inspector plugin shows the live graph, and new docs cover the manifest schema, recipes, safety matrix, lifecycle cleanup, and diagnostics contract.
- **Audio graph/session capability slice** — promotes `audio-mix`, `audio-input`, `audio-monitoring`, and coordinated `stems` diagnostics into the capability runtime. The new audio session host records song route/fader state, redaction-safe input sources, monitoring lifecycle outcomes, stem automation claims/overrides/orphans, and compatibility bridge hits for legacy faders, song volume, Stems master volume, 3D Highway analyser taps, audio startup barriers, and input source handoffs. `core.audio.session` coordinates `stems` without replacing the Stems plugin as the owner of actual stem playback/state.
- **Audio-mix control plane** — makes `audio-mix` the player mixer source of truth. Core now exposes `list-faders`, `get-fader-value`, `set-fader-value`, `inspect-route`, and `inspect-analyser` through the capability runtime, routes native and compatibility-backed fader provider operations with a 2-second timeout, reports committed values back to the mixer UI, suppresses matching legacy faders when a native participant owns the same logical control, and expands audio-session diagnostics/Capability Inspector rendering for fader availability, source modes, bridge hits, route/analyser summaries, and timeout failures.
- **Audio-input control plane** — makes `audio-input` the redaction-safe source of truth for instrument input discovery and lifecycle. Core now exposes `list-sources`, `select-source`, `open-source`, and `close-source` through the capability runtime, persists selected logical sources, keeps inspect/list/select prompt-free, routes provider `source.open`/`source.close` operations with bounded outcomes, shares compatible open sessions across requesters, suppresses compatibility-backed duplicate sources when a native provider owns the same logical key, and expands audio-session diagnostics/Capability Inspector rendering for selected input, open sessions, bridge hits, storage status, and permission/device failures without exposing raw device labels or live audio handles.
- **Audio-monitoring control plane** — makes `audio-monitoring` the shared live-monitoring coordinator. Core now exposes provider registration/list/selection, explicit user-action `start`, requester-counted `stop`, prompt-free `inspect`/`monitoring.status`, and `set-direct-monitor` through the capability runtime. Monitoring starts integrate with selected `audio-input` readiness, background requesters can only attach to active compatible sessions, active sessions survive song/playback stops without auto-resuming after reload, native providers suppress compatibility-backed legacy monitor paths, and diagnostics/Capability Inspector now show providers, sessions, requesters, direct-monitor state, bridge hits, and distinct safe outcomes (`provider-selection-required`, `user-action-required`, `incompatible`, `unavailable`, `stopped`, etc.) without exposing raw audio/device data.
- **Playback control plane** — promotes `playback` to an active core capability domain for song transport, timing, loop, route, requester/observer, bridge, and diagnostics state. Core now exposes `inspect`, user-authorized `start`, `pause`, `resume`, `stop`, `seek`, `set-loop`, and `clear-loop` through the capability runtime while `static/app.js` keeps raw `<audio>`/JUCE handles private behind a redaction-safe adapter. Playback diagnostics use pseudonymous targets in exported bundles, local display labels only in the Capability Inspector, bounded recent outcomes/events, and bridge accounting for `window.playSong`, legacy `song:*` events, `window.feedBack` transport helpers, loop helpers, and browser/native route handoff.
- **3D highway — Tone HUD, fret dividers, chord-diagram toggle, FPS counter.** The bundled `plugins/highway_3d` gains an amber **Tone-change HUD** (shows the active tone and the next scheduled tone change; position / size / visibility configurable in settings), a **fret-dividers** toggle (vertical dividers on the highway, on by default, via `h3dBgSetFretDividersVisible`), a **chord-diagram visibility** toggle (`h3dBgSetChordDiagramVisible`), and an **FPS counter** setting migrated to `BG_DEFAULTS.fpsVisible` (drops the legacy `h3d_showFps` localStorage key). Chord-diagram position is restricted to `tl`/`tr`; legacy `bl`/`br` values are coerced on load. Perf: accent-halo shell descriptors are pre-built per string in `initScene()` and the chord-verdict cache key is encoded as a number, eliminating per-frame allocations in the `drawNote()` and chord hot paths.
- **Sloppak assembly preserves a short preview clip.** When a source chart carries a separate short browser-preview audio clip alongside the full song, the sloppak assembler now decodes it to `preview.ogg` at the sloppak root and records it under a new top-level `preview:` manifest key (POSIX relpath, same shape as `lyrics`/`cover`). A failed preview decode is logged at debug and skipped without aborting the overall build. Sources with no separate preview are unaffected. Older sloppak readers ignore the unknown `preview` key, so the change is purely additive (sloppak-spec.md §5.5 backward-compat). Documented in `docs/sloppak-spec.md` §2 alongside the other optional top-level keys. Enables [`feedBack-plugin-song-preview`](https://github.com/got-feedback/feedBack-plugin-song-preview) to render hover-to-listen previews for sloppaks without seeking into the full audio.
- **Generic plugin asset route** — `GET /api/plugins/{plugin_id}/assets/{path}` serves arbitrary static files a plugin bundles under its own `assets/` directory (AudioWorklet modules, WASM, images, etc.), so plugins can self-host browser-fetchable assets without a CDN (Principle II). Containment is enforced by `lib/safepath.safe_join` against `<plugin>/assets/`, so `..` traversal, absolute paths, and NUL bytes cannot escape `assets/` to reach a plugin's Python modules. `.js` is served as `application/javascript`. First consumer: the stems plugin's pitch-preserving time-stretch worklet.
- **Minigames framework — bundled as a core plugin (`plugins/minigames/`).** Promotes the upstream [`feedBack-plugin-minigames`](https://github.com/got-feedback/feedBack-plugin-minigames) repo into the core bundle so every FeedBack install gets the framework out of the box (same promotion path used for `highway_3d`). The plugin adds a top-level **Minigames** nav link (alongside Library / Favorites / Upload — not buried in the Plugins dropdown), a library-style card grid of installed minigame plugins, and a shared profile layer (XP, level, per-game leaderboards, cross-minigame unlocks) persisted under `CONFIG_DIR/minigames/` and opted into the settings export. Other plugins that want to ship a minigame add a `minigame` block to their `plugin.json` and call `window.feedBackMinigames.register(spec)`; the SDK exposes scoring (`createContinuous` runs a self-contained YIN tracker; `createDiscrete` / `createChord` wrap `note_detect`'s `createNoteDetector`), HUD primitives, run persistence, and a scheduler so individual minigames don't need their own DSP or backend. Backend endpoints live under `/api/plugins/minigames/{runs,profile,registry}`. The framework is plugin-shaped (not core code) per Principle III, but bundled so it ships with every install. First consumer: [`feedBack-plugin-flappy-bend`](https://github.com/got-feedback/feedBack-plugin-flappy-bend), shipped separately.
- **Alpha-build heads-up banner** — when `/api/version` reports a version string containing "alpha" (case-insensitive), an amber banner appears at the top of the library section warning users that the build is in active development and may have bugs or breaking changes. The banner stays hidden on stable / beta / RC builds. No persistence or dismiss state — it's a passive notice, not a modal.
- **Drum vocabulary expanded to 18 pieces** — adds `stack` (MIDI 30, from GM's extended-percussion range, unused by real drum-kit MIDIs) and `bell` (MIDI 80 "Mute Triangle", also unused in real drum-kit MIDIs) to `lib/drums.py` PIECES. Inserted in the iteration order so the editor / highway lane ordering is *hi-hat → stack → crash → … → ride bell → bell*. Both are cymbals; default shape `circle_jagged` (stack) / `circle_dot` (bell). Old drum tabs round-trip unchanged — the schema is permissive and existing piece-ids are untouched.
- **GP / MIDI drum import surfaces unmapped notes** — `convert_drum_track_to_drumtab` (`lib/gp2rs.py`) and `convert_drum_track_from_midi` (`lib/midi_import.py`) gain an optional keyword-only `out_unmapped` parameter. Callers that pass an empty dict receive a per-MIDI record of every silently-skipped percussion note (`{midi: {"count": int, "times": [float, ...]}}`, times capped at 100 samples per note). This lets the editor plugin show a warning + manual-mapping UI on import instead of silently dropping unmapped notes. Default behavior unchanged for callers that don't opt in.
- **Drum support from scratch** — drums are now a first-class arrangement type with their own JSON payload on disk and their own WS stream to the highway. New `lib/drums.py` defines the closed piece-id vocabulary (kick, snare, snare_xstick, hh_closed/open/pedal, tom_hi/mid/low/floor, crash_l/r, splash, china, ride, ride_bell), default GM-MIDI mappings, three preset lane configurations, and a permissive `drum_tab.json` validator. `lib/sloppak.py::load_song` reads the manifest's optional top-level `drum_tab:` key, parses + validates the JSON, and surfaces it on `LoadedSloppak.drum_tab`; the load stays permissive so a missing or malformed tab silently disables drums rather than failing the sloppak load. `/ws/highway/{filename}` gains two new message types — `drum_tab` (metadata + kit legend) and chunked `drum_hits` (500 hits per frame, same chunking as notes) — exposed to renderers via `bundle.drumTab`. `song_info` carries a `has_drum_tab` flag so viz pickers can auto-activate the drums highway regardless of which guitar arrangement is selected. `lib/gp2rs.py::convert_drum_track_to_drumtab` converts a Guitar Pro drum track to a `drum_tab.json` dict, preserving velocity verbatim, mapping hi-hat openness through the canonical piece-ids, and flagging flam / ghost / cymbal-choke articulations from GP effects. `lib/midi_import.py` gains `list_drum_tracks` + `convert_drum_track_from_midi` (channel-9 only) with heuristic flam-collapse (≤30 ms same-piece) and choke detection (cymbal note-off ≤120 ms). `docs/sloppak-spec.md` §5.3 promotes drum_tab from worked-example to canonical with the closed piece-id table and wire format. Sloppaks without a drum_tab are unaffected; legacy drums-as-guitar-notes sloppaks keep playing via the drums plugin's fallback decoder.
- **Loose folder support** — a directory containing an audio file + arrangement XMLs, with optional `manifest.json` and album art, is now discovered, indexed, and playable as a first-class library format alongside Sloppak. The scanner walks `DLC_DIR` for non-preview audio files and treats each parent directory that also contains XMLs as a loose song. Metadata follows a `manifest.json` → XML tags → folder-name priority chain (see `lib/loosefolder.py`). Songs are tagged `format: "loose"`, render an amber `FOLDER` badge in the library, and are filterable via the new "Folder" option in the format dropdown. Audio uses the shared vgmstream/`convert_wem` pipeline, cached under `AUDIO_CACHE_DIR`. The chart `<offset>` from the first non-vocals XML is now propagated to the frontend via `song_info.offset` and applied in `highway.setTime()` so loose folders authored against non-silence-padded audio stay in sync. Pairs with the companion `feedBack-plugin-loosefolder` plugin which adds an in-player Fix Sync UI for nudging and saving offset corrections.
- Highway note-state hook (#254). New `highway.setNoteStateProvider(fn)` lets a scorer plugin publish a per-note judgment (`'hit'` / `'active'` for a sustain currently held correctly / `'miss'`, or `{ state, alpha, color }`) so the renderer lights up the **gem itself** on a correct hit and keeps a sustain trail glowing while it's still being played right — instead of a separate overlay ring near the note. The built-in 2D highway honors it in `drawNote` / `drawSustains` / the chord-frame path (bright string colour + additive halo on hits, bright vs dim sustain trail, faint red wash on misses); the bundled 3D highway reads the same data via `bundle.getNoteState` (bright string-tinted outline + bright body + glowing sustain on hit/active, red outline + suppressed body on miss). Custom renderers opt in by calling `bundle.getNoteState(note, chartTime)`. note_detect registers the provider (and still owns its HUD / diagnostic miss markers / "currently detected" indicator); renderers that ignore the hook simply don't light gems. On a confirmed hit/active the renderers add a contained "sparkle/sizzle" on the note — the 2D highway: additive throbbing halo + flickering hot core + crackling spark lines (+ an expanding shockwave ring on a fresh strike) on the gem and a glowing/jittery sustain trail; the 3D highway: a few twinkling bright dots and short crackling arc segments hugging the note's rectangle (no bloom past the note), drawn on its overlay and projected through the camera so they ride the note. Also adds `highway.isDefaultRenderer()` so overlays that position with the 2D-highway helpers (`project` / `fretX`) can skip rendering when a custom renderer is active — fixes note_detect's miss markers appearing in random places over the 3D highway. New 3D-highway setting **Show note preview on the fretboard** (on by default) toggles the board-projection ghost — the translucent preview of the upcoming note on the fretboard surface. (Note: the companion change in the **note_detect plugin repo** turns its full-screen green/red edge flash off by default and adds a toggle to re-enable it — ships separately with note_detect, not in this feedBack release.)
- Diagnostic bundle export (#166). New "Export Diagnostics" + "Preview Bundle" buttons in Settings produce a single redacted zip combining server logs (tail of `LOG_FILE`), system info (Python/OS/version), hardware probe (CPU model + cores + freq + RAM, GPU via `nvidia-smi`/`rocm-smi`/`system_profiler`, container/Electron/bare runtime detection), full plugin inventory with git SHA + remote URL (read directly from `.git/HEAD` so it works in minimal runtime images without `git` installed) + orphan/failed-to-load detection, the browser console transcript (all levels: log/info/warn/error/debug + window.onerror + unhandledrejection, 500-entry ring buffer), browser hardware (WebGL/WebGPU adapter info, navigator + userAgentData), filtered localStorage, and per-plugin contributed diagnostics. Top-level `manifest.json` lists every file with its versioned schema id (`system.hardware.v1`, `client.console.v1`, etc.) so AI agents can dispatch by schema. Redaction is on by default: DLC paths, song filenames (`<song:HASH8>` stable per-bundle), IPv4/IPv6 addresses, bearer tokens, and `key=`/`token=` query strings are replaced. Plugins opt their backend diagnostics in via a new `diagnostics` manifest field (`server_files` allowlist mirroring `settings.server_files` semantics, plus an optional `callable: "<module>:<function>"` resolved lazily via `load_sibling`). Frontend plugins push diagnostics via `window.feedBack.diagnostics.contribute(plugin_id, payload)`. Three new endpoints: `POST /api/diagnostics/export`, `GET /api/diagnostics/preview`, `GET /api/diagnostics/hardware`. Full bundle format spec in `docs/diagnostics-bundle-spec.md`.
- Structured logging bootstrap (phase 1 of #155). Three new environment variables control server log output: `LOG_LEVEL` (default `INFO`), `LOG_FORMAT` (`text` for coloured console, `json` for one-JSON-object-per-line suitable for Loki/ELK/Promtail), and `LOG_FILE` (optional path, rotated at 10 MB with 5 backups). HTTP responses now include a `X-Request-ID` correlation header (via `CorrelationIdMiddleware`); the same request ID appears as `request_id` in structured log lines emitted via the stdlib `logging` / `structlog` APIs during that request.
- Structured logging migration completed (phase 2 of #155, #159, #242). The 42 `print()` calls and 6 `traceback.print_exc()` calls across `server.py` and `lib/` have been migrated to levelled `feedBack.*` loggers. Silent `except: pass` blocks in `lib/sloppak.py` and `lib/sloppak_convert.py` now surface as `log.warning` / `log.debug` with the exception attached. WebSocket handlers (`highway_ws`, `ws_retune`) bind a `ws_conn_id` contextvar at accept time so every log line within a session carries a connection ID. A CI grep guard in `.github/workflows/tests.yml` fails the build if either `print(` or `traceback.print_exc(` reappears in `server.py` or `lib/`.
- **Lyrics Karaoke plugin** — end-to-end karaoke setup for Sloppak songs in one workflow. The setup screen shows a per-song checklist (vocals stem / synced lyrics / per-syllable pitch) and a single "Build Karaoke" button that runs whatever's missing: Whisper alignment of pasted lyric text against the vocals stem, then `librosa.pyin` pitch extraction. Both artifacts persist inside the Sloppak (`lyrics.json`, `vocal_pitch.json`). In the player, a "Karaoke" toggle swaps the text-lyrics overlay for a horizontal pitch ribbon (one bar per syllable, vertically positioned by pitch, sweeping playhead).
- Settings export/import (#113). Two buttons on the Settings page bundle server config, browser localStorage, and opted-in plugin server-side files into a single versioned JSON file for backup, migration, or sharing a calibrated setup. Server-side import is all-or-nothing for safety-critical failures: phase-1 validates the entire bundle (schema, path-traversal, encoding) before any disk writes; phase-2 commits each file via temp+rename. Plugin-state mismatches between export and import are handled leniently: files referenced for a plugin that isn't loaded are skipped with a warning, files referenced for a plugin whose manifest no longer declares them are skipped with a warning, and localStorage is merged (not cleared) so first-run defaults from plugins installed after the export are preserved. Path-traversal, absolute paths, schema mismatch, and decode failures remain hard refusals. Plugins opt their server-side files in by declaring `settings.server_files` in `plugin.json` (list of relpaths under `CONFIG_DIR`; trailing `/` denotes a directory).
- Library filtering by parts present or missing (#129, #69). New right-side Filters drawer (single button next to the format/sort row, with active-filter count badge and dismissible chips below) lets you require or exclude arrangements (Lead/Rhythm/Bass/Combo), specific stems on Sloppaks (drums/bass/vocals/piano/other), lyrics, and tuning. Multi-select within an axis is OR (Lead OR Rhythm); cross-axis is AND. State persists across reloads. New endpoint `GET /api/library/tuning-names` returns distinct tunings present in the library, ordered by musical distance.
- Sort library by year (#128). Two new options in the sort dropdown: "Year (newest)" and "Year (oldest)". Songs without a year are pushed to the bottom for both directions.
- **`highway.getLyrics()` accessor.** `createHighway()` now exposes the parsed timed lyric syllables (`[{t, d, w}]`) via `getLyrics()`, mirroring `getBeats()`/`getSections()`, so overlay plugins can render karaoke without opening a second highway WebSocket. Pure accessor; no behavior change.
### Changed
- **Perf (3D highway, feedBack#226)**: pre-warm `plugins/highway_3d/screen.js` object pools at board init. Previously the pool factory grew lazily on first `.get()` past the high-water mark, allocating a fresh `T.Mesh` mid-rAF on dense 7/8-string charts and stalling those frames; the meshes were then permanently added to `noteG` (the pool only hides on `reset()`, never removes), bloating the scene graph for the rest of the session. Pre-warming spends the cost up front. Fold the per-frame `updateStringHighlights()` per-string loop with the post-call `mGlow`/`mAccentCore` emissive writes — one walk over the per-string scratch arrays instead of two. Replace `longestConsecutiveRun`'s per-call array allocations with a `{start, len}` index pair (trades two per-call sub-array allocations for one small 2-key object — net reduction in per-visible-chord allocation churn). Opt-in perf bench harness via `?h3dbench=1` URL param: `console.log` p50/p95/max for six update() segments every 5 seconds; when the URL flag is absent the mark helpers are bound to empty functions at renderer-instance init (each `createHighway()` panel re-checks the flag), so the hot-path calls are no-ops with negligible overhead (typically JIT-inlined).
- **License**: Relicensed to AGPL-3.0-only. Prior versions claimed MIT in the README, but the bundled desktop build statically links JUCE 8 (AGPL-3.0), so AGPL terms have effectively governed the desktop distribution since JUCE was added. AGPL-3.0-only is now the canonical license for the project — see [LICENSE](LICENSE) and [CONTRIBUTING.md](CONTRIBUTING.md) (DCO sign-off + plugin licensing policy). Bundled and vendored third-party code keeps its original license.
- Tuning sort is now ordered by musical distance from E Standard (#22) instead of alphabetical: E Standard first, then Drop D / F Standard at distance 2, then Eb Standard / F# Standard at distance 6, etc. Within a magnitude tier, down-tuned variants come before up-tuned, then alphabetical.
- Settings page restructured into separate "FeedBack" (core) and "Plugins" sections, with each plugin's settings rendered as a collapsible panel (collapsed by default). "Plugin Updates" moved into the Plugins section.
- **Lyrics Sync** is now a redirect stub. Its alignment + save endpoints moved into the new Lyrics Karaoke plugin alongside the pitch extraction. Existing nav entries and bookmarks land on a "moved" page that auto-redirects to the merged plugin.
### Security
- **Path traversal in archive extractors and library path resolution.** `lib/sloppak.py::_unpack_zip` and `server.py::_resolve_dlc_path` previously concatenated attacker-controlled entry names or filenames directly onto the extraction or library directory, so a crafted sloppak zip member or library filename with `..` segments, an absolute path, or backslash separators could write or read outside the intended directory. Any code path that unpacks a user-supplied archive (library upload, click-to-play, retune) or resolves a library path was reachable. Both locations now delegate to a new `lib/safepath.py::safe_join` helper that resolves each destination once and rejects entries that don't fall under the target directory; rejected entries are logged and skipped, the rest of the archive still extracts. The stem-split paths in `lib/sloppak_convert.py::split_stems` and `scripts/split_stems.py` previously called `ZipFile.extractall()` directly on user-supplied sloppaks; both now delegate to the same hardened `lib/sloppak.py::_unpack_zip` so every sloppak-unzip site in the codebase shares one containment guarantee. Tests in `tests/test_archive_traversal.py` and `tests/test_safepath.py` pin the contract for `../`, deep traversal, absolute paths, mixed `subdir/../../` forms, Windows-style separators, NUL bytes, names that resolve to the unpack root, and symlinked roots.
### Fixed
- E Standard retune now stays metadata-consistent across a chart's arrangement files (feedBack-plugin-notedetect#50). Previously the retune path could shift the audio and update manifests while leaving some arrangement metadata untouched, so `load_song()` later exposed the *original* tuning at runtime. `lib/retune.py` now updates every arrangement's tuning metadata consistently before applying the E Standard tuning, and raises on a partial update instead of silently packing split tuning metadata. EStd files generated before this fix should be re-converted so their metadata is consistent.
- Keyboard shortcut help now opens from the Player/3D Highway context when Linux/Electron reports Shift+Slash as `key="/"`, including while player controls such as the visualization picker are focused (#598).
- 3D Highway left-handed mode now has regression coverage for fret-axis mirroring, board rebuilds on runtime lefty changes, and mirrored camera state including the lookahead target and shoulder offset; the maintainer guide no longer claims the renderer ignores `bundle.lefty` (#321).
- Chord-level `fretHandMute` is now parsed into each note's `fret_hand_mute` (wire `fhm`) instead of being folded into `mute` (`mt`), matching `_parse_note` and preserving wire-format fidelity for both the template-expanded (synthetic-note) and explicit-`chordNote` paths. The 3D highway renders the fret-hand-mute X for `mt` *or* `fhm` notes, so the muted-chord overlay still shows. Also fixes the per-note fret-connector label vanishing exactly at the hit line (the fade now holds full opacity through `dt = 0`).
- `gp2rs` now respects the time-signature denominator when emitting ebeat subdivisions, fixing misaligned beat grids in 6/8 and other non-quarter-note meters.
- Settings dropdowns (Default Arrangement, Platform Filter) now persist immediately when changed. Previously a dropdown selection was only written to `config.json` when an unrelated "Save" button (Library Folder or Demucs Server) was clicked, so picking a default arrangement and navigating away silently discarded it. Both `<select>` controls now POST the single changed field on `change` via the partial-merge `/api/settings` endpoint, matching the auto-save behaviour of the A/V Sync Offset and mastery sliders. The text inputs (Library Folder Path, Demucs Server URL) keep their explicit Save buttons. Autosaves are sent through a single client-side queue (one request in flight at a time, in selection order), and the `POST /api/settings` handler now serializes its read-modify-write of `config.json` under a lock so concurrent partial updates can no longer overwrite each other and drop a key. The config write is also atomic (temp + rename) and `/api/settings/import` shares the same lock, so readers never observe a half-written file and a settings import can't race a concurrent partial save.
- Demucs stem split failing on Windows desktop with `OSError: Could not load this library: libtorchcodec_core4.dll` or `ImportError: TorchCodec is required for save_with_torchcodec`. The demucs subprocess now bootstraps a `torchaudio.save` → `soundfile.write` shim before importing demucs, sidestepping the torchcodec dependency entirely. The override stays in place across torchaudio versions — soundfile's WAV writes are behaviorally equivalent for demucs's float32 outputs.
- Splitscreen pop-out windows briefly flashed the library/song grid before showing the popped panel. A popup loads the full app (whose default screen, `#home`, is the library) and only swaps to the player once the splitscreen plugin loads; app init now detects `?ssFollower=1` and switches to the player screen up front, so the popup shows player chrome the whole time.
- Sloppak assembly dropped all tone data — affected sloppaks showed no signal chain in the Tones plugin and no tone-change markers on the highway. The assembler (`lib/sloppak_convert.py`) now lifts each arrangement's tones from the source chart via the new `lib/tones.py` helper and embeds them inline in the arrangement JSON under a `tones` key (`base`, `changes`, `definitions` — see `docs/sloppak-spec.md` §3.9). The highway WebSocket reads `base`/`changes` for sloppaks, and the Tones plugin (≥ 1.1.0) reads `definitions` to render the gear chain. Sloppaks built before this release carry no tone data and must be rebuilt from their source chart to gain it.
- Tab View (feedBack-plugin-tabview ≥ 3.0.1): the bottom row of tablature was permanently hidden behind the player controls bar (#336). The overlay reserved 60px at the *top* (clearing the transparent HUD) and extended all the way to the bottom of `#player`, where the opaque `#player-controls` (z-index 10) drew over the last row. The overlay now measures `#player-hud` and `#player-controls` dynamically and insets both edges; a `ResizeObserver` on the controls bar re-runs the inset when it wraps to a second row on narrow viewports.
- Tab View (feedBack-plugin-tabview ≥ 3.0.1): the cursor highlight led playback by roughly one beat (#336). alphaTab snaps `tickPosition` to the start of the *next* beat, so the cursor would race ahead by 500ms+ at typical tempos. The plugin now sends `tickPosition` one beat earlier so the snap lands on the current beat, and the highlight overlay tracks the bar cursor (`.at-cursor-bar`) instead of the next-beat cursor (`.at-cursor-beat`).
### Migration notes
- **Constitution amended to 1.1.0 (Principle II — Vanilla Frontend).** Prebuilt Tailwind (`static/tailwind.min.css`) is now codified as non-negotiable: no Play CDN / runtime CSS JIT anywhere, core or plugin. Plugin authors: a plugin that uses Tailwind classes not guaranteed in core — especially arbitrary values like `w-[37px]` — MUST ship its own compiled stylesheet via the new `styles` manifest key, built with `corePlugins.preflight = false`. Plugins that use only core-guaranteed utilities, or that ship no Tailwind at all, need no change. Contributors: after adding any Tailwind class to core or a bundled plugin, run `bash scripts/build-tailwind.sh` and commit the regenerated CSS, or the `tailwind-fresh` CI job fails.
- The library filters depend on three new columns (`stem_ids`, `tuning_name`, `tuning_sort_key`) that are populated as songs are scanned. If filters look empty after upgrading, run **Settings → Full Rescan** to repopulate; alternatively the periodic background rescan picks them up over time.
## [0.2.4] - 2026-04-22
### Added
- Version badge in navbar (`/api/version` endpoint + `VERSION` file)
- `CHANGELOG.md` and semantic versioning
- Step Mode plugin
- `gp2midi` improvements and expanded test coverage
- Note Detection plugin factory-pattern refactor with multi-instance/splitscreen support
- Per-panel note detection in Split Screen plugin with M/L/R channel routing for multi-input interfaces
### Fixed
- `SLOPPAK_CACHE_DIR` moved to `CONFIG_DIR` for AppImage compatibility
- Improved error message when plugin requirements fail to install
**Frontend scripts** — `screen.js` runs in the global scope via a `<script>` tag. It can access `window.playSong`, `window.showScreen`, `window.createHighway`, the `<audio>` element, and the `window.feedBack` event emitter.
**ES-module plugins (`scriptType:"module"`)** — a plugin may instead ship a native ES-module graph with **no build step**: set `"scriptType": "module"` in `plugin.json`, make `screen.js` a one-line `import './src/main.js'`, and put the module tree under `src/` (served by the sandboxed `/api/plugins/<id>/src/{path}` route). The host injects it as `<script type="module">`, whose `onload` fires only after the whole static-import graph evaluates — so the loader's completion-by-`onload` + `_loadingPluginId` + `playSong` wrapper-chain ordering all hold. Resolve your own asset URLs (worklets, WASM) with `import.meta.url` — `document.currentScript` is `null` in a module. Module top-level code does **not** re-run when the user re-enters the screen at the same version (the host loads screen.js once and `showScreen` re-injects nothing), so keep per-visit re-init in a `screen:changed` handler, exactly as classic plugins do. Classic global-scope `screen.js` remains fully supported. See `docs/plugin-modules.md`.
**The playSong wrapper chain** — Plugins commonly wrap `window.playSong` to hook into song playback. Plugins load alphabetically, so the last-loaded (alphabetically later) wrapper runs first, while the alphabetically first plugin runs closest to the original. Be aware that `await` calls in inner wrappers yield to the event loop — WebSocket messages can arrive before outer wrappers finish setup.
- The built-in 2D highway consults it in `drawNote` / `drawSustains` / the chord-frame path: 'hit'/'active' → bright string colour + additive halo + a contained "sizzle" (crackling sparks, throbbing core, a shockwave ring on a fresh strike) on the gem and a bright (vs dim) sustain trail; 'miss' → faint red wash. The bundled **3D highway** reads the same data via `bundle.getNoteState` (bright string-tinted outline + bright body + glowing sustain + a contained sparkle hugging the note rect on hit/active; red outline + suppressed body on miss). Custom renderers that want it call `bundle.getNoteState(note, chartTime)` — it null-guards and returns the normalized `{ state, alpha, color }` (or null).
- This is orthogonal to the overlay contract: note_detect remains an overlay (HUD, diagnostic miss markers, the "currently detected" indicator) *and* a scorer that feeds this provider. A renderer that ignores `getNoteState` simply doesn't light gems — nothing breaks.
#### 4. Chart-transform provider — remap the chart before rendering AND scoring (feedBack#952)
The core-owned `chart-transform` provider coordinator applies synchronous chart substitutions after difficulty filtering. Register and select providers through the capability domain; it owns persistence, refresh, splitscreen propagation, failure attribution, and diagnostics.
Provider inputs and staged outputs are isolated copies. Async returns or provider errors fail back to the original chart and expose only a fixed public failure reason. `getSongInfo()` remains the original chart contract; transform-aware consumers use the renderer bundle or `getStringCount()`, `getTuning()`, `getCapo()`, and `getCentOffset()`.
See [docs/capability-recipes.md](docs/capability-recipes.md#chart-transform-provider) for the manifest and registration example.
### Audio mixer fader registration (feedBack#87)
Plugins that produce audio outside the song's `<audio>` element (NAM amp output, synth voices, etc.) can register a labeled fader so users can balance them against the song from one mixer popover in the player controls.
Loaded from `static/diagnostics.js` ASAP in `<head>` so the console-wrap is in place before any other script runs. Available on the `window.feedBack.diagnostics` namespace alongside `snapshotConsole()`, `snapshotHardware()`, `snapshotUa()`, `snapshotLocalStorage()`, `snapshotContributions()`. Keep your payload small (< 100 KB) and don't include secrets — bundles are shared with maintainers.
### Detachable panes — pop your panel out into its own window
If your plugin has a floating panel that sits over the player — a mixer, a camera rig, a settings board — you can let the user pop it out into its own OS window and leave it there: while they play, across song switches, on a second monitor, minimized to the tray. Two calls:
```js
feedBack.panes.register({
id: 'camera_director',
title: 'Camera Director',
icon: '🎥',
element: () => panelEl, // your existing panel, exactly as it is
**The host moves your real element.** Not a copy, not a re-render — the actual DOM node, adopted into the pop-out window, keeping its listeners and its closures. Your panel goes on running *your* code against *your* state. It looks and behaves like what was popped out because it **is** what was popped out. Nothing to mirror, nothing to keep in sync.
The rules below are all things that have already gone wrong. Full contract: **[docs/plugin-panes.md](docs/plugin-panes.md)**.
- **Your code still runs in the main window.** The element is *displayed* elsewhere; its closures, timers and `document` references still belong to the main realm. That is exactly why everything keeps working — and exactly why `document.body.appendChild(myPopover)` lands in the **main window, not the pane**. Anchor tooltips, popovers and menus to your panel, not to `document.body`. Measure with `el.ownerDocument.defaultView`, never a cached `window`.
- **Don't hide your panel yourself when it pops out.** Core hides it and leaves a "bring it back" stub. If you also hide it, you will hide the node that just moved — and blank the pane window.
- **Prefer `hidden` or a class over inline `display` for show/hide.** While popped out, core neutralises *placement* with `.fb-paned` (`position`, `inset`, `width`, `z-index`, `box-shadow`). An inline `display:none` on your panel reasserts itself the moment the pane docks back and the class is removed, so your panel returns invisible.
- **`element` is a function so it can be resolved late.** Return the *live* node. If you rebuild your panel (Camera Director rebuilds on every mode change), re-run `attachChip` — it returns a `detach()`; call it before re-attaching, and again in your teardown.
- **`isConnected` does not mean "docked".** A panel sitting in a pane window is very much connected — just not to *this* document. Test `el.ownerDocument === document`, or take the `onHost(hostId, el)` callback.
- **rAF is throttled while the main window is backgrounded** — and it will be, whenever the user is looking at your pane. Event-driven panels (sliders, buttons) are unaffected. A panel that *animates continuously* may run slowly while it is the only thing on screen.
- **Don't reach for BroadcastChannel, `postMessage`, or a second copy of your state.** There is one realm and one panel. If you find yourself synchronising, you have misunderstood the model.
- **Nothing is required.** No panes API on the host → skip both calls, and your panel behaves exactly as it does today.
### Keyboard Shortcuts
Plugins can register keyboard shortcuts via the global `window.registerShortcut()` function. Shortcuts appear in the `?` help panel.
@@ -540,6 +596,21 @@ tab, key/scale annotations, etc.). Published as **feedpak**; this codebase still
**sloppak** name internally — same on-disk format. [docs/sloppak-spec.md](docs/sloppak-spec.md) is
a local pointer + code map.
**The spec is sacrosanct — read it BEFORE changing how this app reads or writes packs.** The
spec repo defines the format; this app merely implements it ("a change is not part of the format
until it lands here" — feedpak-spec/GOVERNANCE.md). Any new manifest key, file, or directory the
app touches must land in the spec **first**, via the
- **Storage** — `localStorage` for all user preferences
- **Styling** — Tailwind CSS utility classes, dark theme (`bg-dark-600`, `text-gray-300`, accent `#4080e0`, gold `#e8c040`). Tailwind is served as a **prebuilt** stylesheet (`static/tailwind.min.css`, regenerated by `bash scripts/build-tailwind.sh`), **never** the runtime Play CDN — the CDN's on-the-fly JIT rescanned the DOM on the main thread and dropped ~26% of frames with the 3D highway (feedBack-desktop#110). The committed CSS only contains classes the build scanner saw, so CI (`tailwind-fresh`) rebuilds and diffs it; run the build script and commit when you add new classes. A plugin that uses classes not guaranteed in core (notably arbitrary values like `w-[37px]`) MUST ship its own compiled stylesheet via the `styles` manifest key, built with `corePlugins.preflight = false` (utilities only — core ships the one base reset). Plugins MUST NOT load the Tailwind Play CDN or any runtime CSS JIT. See constitution Principle II.
- **Naming** — camelCase for JS functions, kebab-case for CSS classes, snake_case for plugin IDs
- **Text selection (v3)** — the v3 UI defaults to `user-select: none` on `html` (in `static/v3/v3.css`) so accidental drag/double-click selection of chrome never looks broken. Form fields are always re-enabled, and a **plugin's mounted screen subtree (`.screen[id^="plugin-"]`) stays selectable by default**, so a plugin's copy-worthy text (lyrics, chord names, results, diagnostics) is unaffected — *unless your plugin renders copyable content OUTSIDE its `plugin-<id>` screen* (e.g. injected into the player chrome / a HUD overlay), which inherits the non-select default. Opt such content back in with the core-served **`.fb-selectable`** class (it sets `user-select: text` on the element + descendants; works for runtime-installed plugins since it's hand-authored in core CSS, not a scanned Tailwind utility). Never use a `* { user-select: none }` rule (breaks input carets/IME), and never use `user-select: none` to "lock" text — keep errors, IDs, paths, versions, and metadata selectable.
- **Player layout** — `#player` is `display:flex; flex-direction:column; position:fixed; inset:0`. `#highway` is `flex:1`. `#player-controls` sits at the bottom. Hiding the highway collapses the layout — use `margin-top: auto` on controls if you need to hide it.
| [Create from Tab](https://github.com/got-feedback/feedBack-plugin-ug) | Search Ultimate Guitar for GP tabs and convert to playable songs | `git clone ...feedBack-plugin-ug.git ultimate_guitar` |
| [Import Tab](https://github.com/got-feedback/feedBack-plugin-tabimport) | Drag and drop Guitar Pro files to create songs | `git clone ...feedBack-plugin-tabimport.git tab_import` |
| [Practice Journal](https://github.com/got-feedback/feedBack-plugin-practice) | Auto-track practice time, speed, loops. Dashboard with charts | `git clone ...feedBack-plugin-practice.git practice_journal` |
| [Metronome](https://github.com/got-feedback/feedBack-plugin-metronome) | Audible click and visual beat flash synced to song tempo | `git clone ...feedBack-plugin-metronome.git metronome` |
| [Tone Player](https://github.com/got-feedback/feedBack-plugin-tones) | View amp/pedal/cab signal chains with gear artwork | `git clone ...feedBack-plugin-tones.git tones` |
| [Fretboard View](https://github.com/got-feedback/feedBack-plugin-fretboard) | Live fretboard overlay showing active notes in real-time | `git clone ...feedBack-plugin-fretboard.git fretboard` |
| [MIDI Amp Control](https://github.com/got-feedback/feedBack-plugin-midi) | Auto-switch amp/modeler presets via MIDI on tone changes | `git clone ...feedBack-plugin-midi.git midi_amp` |
| [Section Map](https://github.com/got-feedback/feedBack-plugin-sectionmap) | Color-coded song structure minimap with clickable navigation | `git clone ...feedBack-plugin-sectionmap.git section_map` |
| [Arrangement Editor](https://github.com/got-feedback/feedBack-plugin-editor) | DAW-like visual editor for creating and editing song note charts | `git clone ...feedBack-plugin-editor.git editor` |
| [MIDI Capo](https://github.com/masc0t/slopsmith-plugin-midi-capo) | MIDI capo control for real-time transposition | `git clone ...slopsmith-plugin-midi-capo.git midi_capo` |
| [Note Detection](https://github.com/got-feedback/feedBack-plugin-notedetect) | Real-time pitch detection and scoring against highway notes | `git clone ...feedBack-plugin-notedetect.git note_detect` |
| [Find More](https://github.com/masc0t/slopsmith-plugin-find-more) | Search for more songs by the same artist | `git clone ...slopsmith-plugin-find-more.git find_more` |
| [Piano Highway](https://github.com/got-feedback/feedBack-plugin-piano) | Scrolling piano/keyboard view for Keys arrangements with MIDI input | `git clone ...feedBack-plugin-piano.git piano` |
| [Studio](https://github.com/got-feedback/feedBack-plugin-studio) | Collaborative band recording and multi-track mixing | `git clone ...feedBack-plugin-studio.git studio` |
| [Drum Highway](https://github.com/got-feedback/feedBack-plugin-drums) | Lane-based drum highway with MIDI drum pad input and built-in sounds | `git clone ...feedBack-plugin-drums.git drums` |
| [Invert Highway](https://github.com/masc0t/slopsmith-plugin-invert-highway) | Flip the highway note direction | `git clone ...slopsmith-plugin-invert-highway.git invert_highway` |
| [Jumping Tab](https://github.com/renanboni/slopsmith-plugin-jumpingtab) | Yousician-style 2D horizontal tab with trajectory arcs and hopping ball | `git clone ...slopsmith-plugin-jumpingtab.git jumpingtab` |
| [Step Mode](https://github.com/got-feedback/feedBack-plugin-stepmode) | Step-by-step practice mode — highway freezes at each note until played (via Note Detection) or Space | `git clone ...feedBack-plugin-stepmode.git step_mode` |
| [Lyrics Sync](https://github.com/got-feedback/feedBack-plugin-lyrics-sync) | Generate synced LRC lyrics from text + vocals stem via Whisper alignment | `git clone ...feedBack-plugin-lyrics-sync.git lyrics_sync` |
| [NAM Tone Engine](https://github.com/got-feedback/feedBack-plugin-nam-tone) | In-browser amp modeling with NAM WASM, cabinet IRs, tone auto-switching | `git clone ...feedBack-plugin-nam-tone.git nam_tone` |
| [Guitar Theory Lab](https://github.com/topkoa/slopsmith-plugin-guitar-theory) | Explore scales, chords, intervals, tunings, and voicings on a fully interactive fretboard | `git clone ...slopsmith-plugin-guitar-theory.git guitar-theory-lab` |
| [Themes](https://github.com/masc0t/slopsmith-plugin-themes) | Offers several basic recolorings of the interface | `git clone ...slopsmith-plugin-themes.git themes` |
| [Update Manager](https://github.com/masc0t/slopsmith-update-manager) | Installs, updates, and uninstalls other plugins and the feedBack core itself | `git clone ...slopsmith-update-manager.git update_manager` |
| [Simplify Chords](https://github.com/bkranendonk/slopsmith-plugin-simplify-chords) | Changes complex chords on the note highway to simpler ones. Inspired by Ultimate Guitar's Simplify button. | `git clone ...slopsmith-plugin-simplify-chords.git simplify-chords` |
| [Key Bindings](https://github.com/jackipicco/slopsmith-plugin-key-bindings) | Highway key bindings for keyboard and TV remote | `git clone ...slopsmith-plugin-key-bindings.git key_bindings` |
| [Virtuoso](https://github.com/got-feedback/feedBack-plugin-virtuoso) | Practice studio for guitar & bass — scale, technique, and rhythm drills, timed workouts, and jam backing that teach skills you take off the screen. | `git clone ...feedBack-plugin-virtuoso.git virtuoso` |
| [Audio Preview](https://github.com/saleemk/slopsmith-plugin-audio-preview) | Quick audio previews from library cards with configurable start time, volume, and duration | `git clone ...slopsmith-plugin-audio-preview.git audio_preview` |
| [Song Mastery](https://github.com/jamesgaiser/slopsmith-plugin-song-mastery) | Auto-adjusts difficulty based on your rolling note accuracy and saves the slider position per song | `git clone ...slopsmith-plugin-song-mastery.git song_mastery` |
| [Mobile Note Highway](https://github.com/saleemk/slopsmith-plugin-mobile-note-highway) | Touch-optimized player with collapsible controls, highway gestures, and device-adaptive layouts for phones and tablets | `git clone ...slopsmith-plugin-mobile-note-highway.git mobile_note_highway` |
Install any plugin by cloning it into your `plugins/` directory and restarting:
@@ -153,6 +153,14 @@ The legacy chart-coupled surface — `highway.setNoteStateProvider(fn)`, the sin
Diagnostics live under `feedBack.note_detection_capability.v1` and contain provider ids/labels/kinds, binding summaries (requester, provider, redacted context, target size), availability, and the last bounded outcome (event, binding, provider, MIDI number, hit flag) — never raw audio buffers, sample data, device labels, or song identity.
## Chart-Transform Domain
The chart-transform slice (#952) is a core-owned provider coordinator implemented by [static/capabilities/chart-transform.js](../static/capabilities/chart-transform.js). Its commands register, select, clear, and refresh providers; `chart.transform` is the provider operation. Selection persists by provider id and applies to the primary highway and announced splitscreen instances.
The synchronous `highway.setChartTransform` data-plane hook runs at chart ready, mastery changes, and refresh—not per frame. Transforms receive isolated chart data after difficulty filtering and may replace notes, chords, anchors, hand shapes, chord templates, string count, tuning, capo, and cent offset. Outputs are isolated and timeline arrays are time-sorted before the built-in renderer, renderer bundle, or public getters read them. Async returns and other provider failures clear the stage and retain the original chart.
`getSongInfo()` retains original metadata; effective values are exposed by the renderer bundle and dedicated highway getters. Diagnostics under `feedBack.chart_transform.diagnostics.v1` contain provider selection/install state and a fixed public failure reason, never chart data, song identity, or raw exceptions. The domain has no compatibility shim because no earlier chart-substitution surface exists.
## MIDI-Input Domain
The MIDI-input slice (spec 012, issues #873/#880) promotes `midi-input` as a **core-owned** provider-coordinator implemented by [static/capabilities/midi-input.js](../static/capabilities/midi-input.js) — the MIDI analog of `audio-input`. It is deliberately separate from `audio-input` (whose source/`open` contract is audio-frame-centric: channel shapes, sample buffers) because MIDI carries discrete messages, not audio; and it is **not** owned by any feature plugin, so the device-access boundary outlives the input-setup wizard (exactly as `audio-input` is `core.audio.session`-owned). Consumers — the `input_setup` onboarding wizard, the `piano`/keys and `drums` plugins, and (as a follow-up, #881) note-detection's Web-MIDI provider — converge here on ONE device-access boundary: one permission prompt, one source list, one redaction boundary, retiring private per-plugin `navigator.requestMIDIAccess()` calls.
@@ -192,7 +200,7 @@ Core domains include review metadata in diagnostics:
-`active`: wired to current FeedBack behavior and expected to work as an integration point.
PR1 includes only the delivered domains listed in [capability-roadmap.md](capability-roadmap.md): `pipeline`, `diagnostics`, and `library`. The follow-up audio graph/session slice promotes `audio-mix`, `audio-input`, `audio-monitoring`, and a coordinated `stems` surface. The playback slice promotes `playback` as an active transport control plane. The audio-effects slice promotes provider-selected effect-chain planning while leaving physical processor loading to compatible executors such as trusted Desktop native audio or a browser/WASM executor. The visualization slice promotes `visualization` as the highway renderer provider-coordinator, and the note-detection slice (spec 009) promotes `note-detection` as the detection-binding control plane. Backend routes, app UI, settings, and other hardware-facing domains remain documented in the roadmap and safety matrix until their own host workflow/provider slice exists.
PR1 includes only the delivered domains listed in [capability-roadmap.md](capability-roadmap.md): `pipeline`, `diagnostics`, and `library`. The follow-up audio graph/session slice promotes `audio-mix`, `audio-input`, `audio-monitoring`, and a coordinated `stems` surface. The playback slice promotes `playback` as an active transport control plane. The audio-effects slice promotes provider-selected effect-chain planning while leaving physical processor loading to compatible executors such as trusted Desktop native audio or a browser/WASM executor. The visualization slice promotes `visualization` as the highway renderer provider-coordinator, the note-detection slice (spec 009) promotes `note-detection` as the detection-binding control plane, and the chart-transform slice (#952) promotes `chart-transform` as the pre-render/pre-scoring chart substitution coordinator. Backend routes, app UI, settings, and other hardware-facing domains remain documented in the roadmap and safety matrix until their own host workflow/provider slice exists.
Capability metadata is versioned by the `capability-pipelines.v1` standard. Invalid roles, commands, operations, requests, observes, emits, events, owner kinds, compatibility modes, ownership policies, safety classes, or version fields are excluded from the capability graph and surfaced through `capability_validation_warnings`; legacy plugin fields continue to load through their existing app paths. Plugins that declare a future `capability-pipelines` version are reported through `capability_unsupported_versions` and their runtime handlers are marked incompatible.
@@ -248,7 +256,7 @@ UI placement and settings contributions are real FeedBack surfaces, but they are
The library provider workflow is the PR1 core adapter and is implemented natively as the `library` capability module. Provider refresh, selection, and sync run through `library` owner commands; backend provider registration remains the way providers enter the library registry, and the browser module turns that registry into provider participants. The app event bus continues to dispatch local `window.feedBack` events for legacy listeners; playback now mirrors song transport, route, seek, and loop lifecycle into `playback`, and visualization attributes renderer selection/failure, while navigation, note, and route-only surfaces remain outside capability domains until their own slices land.
The direct `window.highway` object remains the renderer data plane. Per-frame reads such as notes, chords, beats, and renderer hooks should not be moved behind asynchronous capability commands until there is a dedicated chart/render facade.
The direct `window.highway` object remains the renderer data plane. Per-frame reads such as notes, chords, beats, and renderer hooks should not be moved behind asynchronous capability commands until there is a dedicated chart/render facade. The `chart-transform` domain follows this doctrine: its substitution runs through the synchronous `highway.setChartTransform` hook (staged once per chart change), while the capability surface owns only registration, selection, and diagnostics.
## First-Party Management Plugins
@@ -295,8 +303,9 @@ From the `feedBack/` directory:
Plugins that transpose, simplify, annotate, or otherwise rewrite chart data register as `chart-transform` providers (#952). The effective chart reaches the built-in highway, custom renderers, and highway getters on primary and splitscreen instances.
`transform(input)` receives filtered `notes`, `chords`, `anchors`, and `handShapes`, plus full-difficulty `allNotes`/`allChords`, `chordTemplates`, `stringCount`, and `songInfo`. It may synchronously return any subset of those arrays plus `tuning`, `capo`, or `centOffset`; null leaves the chart unchanged. The host isolates provider inputs and outputs, time-sorts accepted timelines, and falls back to the original chart on failure.
Transforms run at chart ready, mastery recompute, and explicit `refresh`, never per frame. Selection persists by provider id. `getSongInfo()` retains original metadata; effective metadata is available through the renderer bundle and `getStringCount()`, `getTuning()`, `getCapo()`, and `getCentOffset()`.
## Future Expansion Domains
Some domain names are reserved for expected future contracts, but they are not registered in the runtime graph yet. For example, `ui.player-panels` is documented as a likely panel-host surface, but FeedBack does not currently expose a capability command for panel contributions. See [capability-roadmap.md](capability-roadmap.md) for the PR1 domain set and deferred-domain checklist.
@@ -60,6 +60,10 @@ The progression slice (spec 010) promotes `progression` as an active exclusive-o
Deferred follow-up slices: a `contributor` role so plugins ship their own challenge/quest content (drums challenges from a drums-scoring plugin, quest-pool entries from minigame plugins), and drums scoring wiring so `song_completed {instrument: "drums"}` goals become satisfiable.
## Chart-Transform Control Plane Slice
The chart-transform slice (#952) is an active provider-coordinator domain. It owns provider lifecycle, persisted selection, refresh, failure attribution, and redaction-safe diagnostics. Its synchronous highway hook applies isolated provider output after difficulty filtering to built-in, custom-renderer, and getter consumers across primary and splitscreen highways. No compatibility shim is needed; per-panel independent selection remains a follow-up.
## Recommended Next Slices
The plugin inventory suggests this migration order after the audio graph/session and playback slices:
@@ -20,7 +20,7 @@ Core domains also have a review scope. **Active contract** domains are wired to
| visualization | provider-coordinator | safe | inspect, list-providers, select-renderer, clear-renderer | renderer.create, renderer.destroy | Highway renderer provider registry, picker-delegated selection, auto-match attribution, and failure fallback. `renderer.create` maps to the legacy `window.feedBackViz_*` factory `init(canvas, ctx)` call; `renderer.destroy` maps to the factory `destroy()` teardown. Legacy `type: "visualization"` manifests and `window.feedBackViz_*` globals are accounted compatibility shims. Diagnostics carry provider ids/labels, selection source, last auto-match outcome, and last failure — no song filenames, titles, or arrangement names. |
| note-detection | provider-coordinator | sensitive | inspect, register-provider, unregister-provider, open-binding, close-binding, set-target, clear-target | pitch.estimate, verify.target | Detection-binding control plane (spec 009): providers (midi/engine/js) serve primitives; each requester binds its own redacted tuning context; consumers own judgment, hit/miss flow as observability events. Legacy `highway.setNoteStateProvider` is an accounted shim. Diagnostics carry provider/binding summaries and bounded outcomes — no raw audio, sample data, device labels, or song identity. |
| chart-transform | provider-coordinator | safe | inspect, list-providers, register-provider, unregister-provider, select-provider, clear-provider, refresh | chart.transform | Synchronous chart substitution after difficulty filtering (#952). Provider data is isolated, timelines are sorted, and failures retain the original chart with a fixed public reason. Diagnostics contain provider and selection state, never chart data, song identity, or raw exceptions. |
| midi-input | provider-coordinator | sensitive | inspect, list-sources, discover, select-source, open-source, close-source | source.enumerate, source.describe, source.open, source.close | Core-owned MIDI device control plane (spec 012), the MIDI analog of `audio-input`. Inspect/list/select are prompt-free; `discover` is the Web-MIDI permission boundary (`requestMIDIAccess()` gates the whole input list) and records denied/unavailable outcomes; `open-source` attaches a shared listener session and never re-prompts. Selection persists by redaction-safe `logicalSourceKey`. Diagnostics redact device labels and never include raw MIDI messages or live handles. |
Privileged commands are roadmap-only until they have: a visible user confirmation path, diagnostics redaction rules, failure recovery, and tests that prove disabled or incompatible participants cannot execute handlers.
`tools/check_spec_conformance.py`, run in CI as the `feedpak-spec` job.
## Why
feedpak is published as an **open format**: its own repo
([got-feedback/feedpak-spec](https://github.com/got-feedback/feedpak-spec)), a normative spec, JSON
Schemas, and a reference validator. That is a promise to everyone outside this codebase — third-party
packers, converters, and players build against the spec, and the spec is meant to be the complete and
authoritative description of a pack.
The moment core reads a manifest key the spec doesn't define, that promise breaks silently:
- A spec-compliant pack is no longer guaranteed to be a fully-working pack.
- The reference validator can't warn authors about a key it has never heard of — it will happily green-light
the key, and every misspelling of it.
- The format's real definition drifts into our source tree. In the case that motivated this gate
([#933](https://github.com/got-feedback/feedback/issues/933)), third-party tooling started emitting an
`original/` directory that no code anywhere requires — the convention was reverse-engineered from an
example in a *code comment*.
The rule this gate enforces: **any manifest key core reads _or writes_ must be in the spec before core
ships code that depends on it.** Spec first, implementation second. Writes are not exempt — a key core
writes lands in every pack we emit, so an undeclared one seeds the ecosystem with non-spec data.
Note that "get it into the spec" is not automatically the right fix for an existing violation — for
`original_audio` it isn't. The spec already carries the pre-separation mixdown as a stem
(`{id: full, file: stems/full.ogg}`), so that key added a *second, redundant* location for audio to a format
that already had one, and the resolution is to remove it rather than bless it. The gate takes no position on
which way a violation resolves; it only insists that one of the two happens deliberately, in the open,
before the code merges.
## What it checks
We can't mechanically prove core *interprets* a key the way the spec means. We can prove four surface
properties, and they cover the drift that actually occurs.
| Layer | Check | Catches |
|---|---|---|
| 1. key-coverage | Every manifest key core reads **or writes** is declared in the spec's `manifest.schema.json`. | Core growing a key the spec never defined — the #933 class. |
| 2. allowlist-closed | `feedpak-spec-exceptions.yml` has not **grown** relative to the base branch. | Someone routing around the FEP process by allowlisting their own new key. |
| 3. forward | Core's `load_song()` ingests every example pack the spec ships. | The spec adding or tightening something core ignores or breaks on. |
| 4. reverse | Every pack committed to this repo passes the spec's `tools/validate.py`. | Core (or a contributor) committing a pack the spec would reject. |
Layer 1 works by walking the AST of the modules listed in `READERS` and collecting every literal key touched
on a manifest dict (`manifest.get("x")`, `manifest["x"]`, and the wrapped
`(load_manifest(p) or {}).get("x")` form used in `lib/enrichment.py`).
**Reads and writes are both checked, and reported differently.** A key core *writes*
(`manifest["x"] = v`, as `lib/songmeta.py` does) is spec surface pointed outward: it puts a key into every
pack we emit, so an undeclared one seeds the ecosystem with non-spec data. Subscripts are classified by AST
context — `Store` is a write, `Load` is a read — so `manifest["year"] = ...` is not miscounted as a read.
## When it fails
You added a manifest key the spec doesn't define. **There is exactly one way forward, and it is not in this
repo.**
Land the key in the spec through the **feedpak Enhancement Proposal (FEP)** process
**Trigger:** a plugin UI feature accidentally "carved itself into a single theme."
## 1. Problem
A results-card feature in the `note_detect` plugin (a glow-ring hero button + a
gradient-filled accuracy number) was built and visually verified against **only the
default skin** ("neon"). On the other skins it broke: on "esports" — a deliberately
glow-less, near-monochrome design language — the glow ring and the colour gradient
simply **vanished**. The colours adapted (everything used CSS custom-property tokens),
but the **visual devices themselves did not port**, because nothing in the system says
"this theme does / doesn't do glow rings."
### Root cause (three findings)
1.**Themes are design *languages*, not palettes.** neon = glow + animation + gradients;
esports = no-glow, square, near-monochrome amber; metal = brushed steel + hard bevels +
drop-shadows. Tokens made *colour* portable; they never made a *device* portable.
2.**Tokens are named by *device*, not *intent*.** e.g. `--nd-glow-*` holds a glow in neon
but a **hard drop-shadow** in metal — the metal skin is already repurposing a
device-named slot to express a different language. The cure is to finish that move:
name slots by intent, with "off" (`none`) a legal value.
3.**No "text-legible-on-accent" role.** White-on-accent was hardcoded in several places;
on esports' amber accent that's a contrast failure. And `--nd-accent2` was
**double-booked** (gradient-end *and* S-grade colour), so the hero gradient resolved
amber→near-white and washed out.
A process gap compounds it: **verification covered one skin**, so the regression was
invisible until a user switched themes. And this recurs ecosystem-wide — other plugins
ship their own independent skin systems too.
## 2. Current state (two disconnected systems)
| System | What it is | Limits |
| --- | --- | --- |
| **Host themes** (`static/v3/theme-core.js`, `html[data-fb-theme]`) | Cosmetic "shop" themes that recolour `fb-*` Tailwind tokens (surfaces/text/borders). | Apply-only & recolour-only. `--fbv-*` vars exist **only while a theme is equipped** (nothing to read in the default state). No read API, no capability signal, no normalized `theme:changed` event. Comment explicitly says it *leaves decorative accents (rings/shadows) at defaults* → **devices are an ownerless gap.** |
| **Plugin skins** (e.g. `note_detect``data-nd-skin`) | Full per-plugin design languages (neon/esports/metal) as CSS-var blocks. | Each plugin reinvents the wheel; disconnected from host themes; a feature can't see both. |
## 3. Goals / non-goals
- **Goal:** a feature, authored once, renders correctly in **any** theme — including ones not
yet invented — and degrades **intentionally** (neon ring → esports border), never accidentally.
- **Goal:** the host owns a canonical contract so plugins consume instead of reinventing.
- **Non-goal:** forcing every plugin skin to become a host theme. Skins stay plugin-local but
**implement** the contract.
- **Non-goal:** backward-compat with pre-v3 hosts. Everything here is additive + feature-detected.
| **on-fill (new)** | `--fb-on-accent``--fb-on-good``--fb-on-warn``--fb-on-bad` | **Rule: every role used as a fill behind text gets a paired `--fb-on-*`** (fixes white-on-amber). Required + contrast-linted (§6). |
| **focus (new)** | `--fb-focus-ring` | focus indicator independent of `accent`, so focus stays visible when `accent ≈ surface` |
| core `static/highway.js` → residual `renderer-2d.js` (post-split) | ~2,400–2,900 est. | **3,000** | 60 fps hot path; no module boundary inside the per-frame loop | Byron | after the highway.js split |
| core `plugins/highway_3d/` → residual renderer | sized at split; likely **>3,000** | set at split, flagged now | same hot-path rule; the draw core can't be cut without behavior risk | Byron | after the highway_3d split |
| tutorials `builtin/reading-the-highway/generate.py` | 1,818 | 2,000 | offline content generator, never imported at runtime, deps not in runtime requirements | Byron | if a 3rd builtin pack appears |
| desktop `src/audio/NodeAddon.cpp` | 3,542 | as-is | C++, outside the ESM/routes playbooks; under active use-after-free crash work — do not churn | Byron | after crash-class work settles |
| desktop `src/audio/AudioEngine.cpp` | 2,977 | as-is | same | Byron | same |
| desktop `src/vst-host/main.cpp` | 1,928 | as-is | same | Byron | same |
| virtuoso `screen.js` (authored, own track) | 25,741 | as-is until its own split | authored plugin on a separate roadmap; migrates on its own schedule | Christian | virtuoso split kickoff |
## Split-when-touched (no scheduled train; row retires when split)
(2,413 — was 14,037; ratcheted by the R3 `MetadataDB` + `AudioEffectsMappingDB`
extractions and twenty-two `routers/` modules, plus lib/library_registry.py for the provider-registry classes (album-art in `lib/routers/art.py`, the settings + export/import bundle in `lib/routers/settings.py`); the ~930-line metadata-enrichment subsystem — MB/CAA/AcoustID transport, matcher, background worker — now lives in `lib/enrichment.py`) ·
`lib/metadata_db.py` (4,373 — new in R3; the `MetadataDB` class alone is 4,018 lines
and is a monolith in its own right, to be split per-table once the router train
"description":"Career mode — gig your way from a local bar to the arena, and build a passport wall of genre badges per instrument. Earn stars per song; the crowd reacts to how you play.",
<pclass="text-sm text-gray-400 mb-4">Earn stars by playing songs well — 60% accuracy is a star, 75% two, 85% three. Stars unlock bigger stages, and the crowd plays along with you.</p>
<pclass="text-sm text-gray-400 mb-4">Commit to an instrument, pick a genre, and stamp your way to its badge — five ★★ songs mint a Bronze. Your passport wall is who you are as a musician.</p>
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.