feat(editor): tracks become first-class — the persistent track-session tree - #289
Merged
Conversation
…n tree
A song's tracks (master recording, studio stems, transcription parts,
optional folders) are now one ordered, persistent tree — the
editor_track_session manifest extension key — instead of surfaces
re-deriving ad-hoc lists from whatever the song contains.
The tree LAYERS OVER canonical song data, never replacing it:
- rows reference sources ('master' / bare manifest stem ids) and chart
tracks (_partViewKeyPure keys — the same dialect editor_stem_links
speaks); unknown references drop on normalize, new parts append,
parent cycles repair;
- pairing is NOT stored: rows project S.stemLinks, so the two stores
can never disagree (a link to a tombstoned source projects unpaired);
- sources are derived on demand from (audioUrl, S.stems) — no parallel
copy to go stale under /stem-op renames;
- removing an audio track is non-destructive: a removedSourceIds
tombstone hides the row, the media stays in the pack, restore
re-appends it;
- the tempo-guide role (which source is the timing reference, locked or
not, audio vs metronome) persists on the tree; its behavior ships
separately;
- a fully-default tree saves as null and the backend removes the key,
so untouched packs stay byte-identical (absent ≠ null: an older
client's absence has no authority to erase a persisted tree).
Wired at every persistence boundary (save, save-as, build, load) and
through the create/import seam #286 reserved (installCreatedTrackSession
now real: seeds S.stems from the server's audio_sources — bare ids,
unconditional reset so a previous song's stems can't leak into a fresh
import). This slice is model + persistence; the unified Tracks surface
that renders the tree ships separately.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EBQCHCNA81E9tHmSDHSe2Q
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (2)
🚧 Files skipped from review as they are similar to previous changes (2)
📝 WalkthroughWalkthroughAdds a persistent editor track-session model with normalized track trees, folder and source operations, tempo-guide state, manifest load/save/build wiring, create/import initialization, and Python/JavaScript regression tests. ChangesTrack session persistence
Estimated code review effort: 4 (Complex) | ~45 minutes Sequence Diagram(s)sequenceDiagram
participant Editor
participant file_ops
participant routes
participant Sloppak
Editor->>file_ops: load track_session
file_ops->>routes: request persisted session
routes->>Sloppak: read editor_track_session
Sloppak-->>routes: return manifest payload
routes-->>file_ops: return sanitized track_session
file_ops-->>Editor: install normalized session
Editor->>file_ops: save track_session
file_ops->>routes: send serialized session
routes->>Sloppak: write editor_track_session
Possibly related issues
Possibly related PRs
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@routes.py`:
- Around line 484-491: Update the removedSourceIds handling near
_track_session_id to bound the number of processed entries and deduplicate IDs
with a set for linear-time membership checks, while preserving the existing
filtering and output order in removed_sources. Use the established request limit
or an appropriate fixed maximum, and stop processing once that bound is reached.
In `@src/track-session.js`:
- Around line 124-130: Update the persisted-row normalization before the
canonical loops in track-session handling so folder or mismatched leaf records
cannot occupy IDs returned by audioTrackId or transcriptionTrackId. Reserve
canonical source and target IDs while processing persisted rows, or assign
collision-free IDs to conflicting leaf rows, ensuring the loops over
visibleSources and targets always append every canonical track.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 63d75665-aeab-4c37-9adf-3481bf812ef4
📒 Files selected for processing (9)
CHANGELOG.mdroutes.pysrc/create.jssrc/file-ops.jssrc/main.jssrc/state.jssrc/track-session.jstests/test_track_session_persist.pytests/track_session.test.mjs
This was referenced Jul 15, 2026
ChrisBeWithYou
pushed a commit
that referenced
this pull request
Jul 16, 2026
A click/reference stem can be locked as the session's TEMPO GUIDE (the ♩ button in the Audio tracks manager; the role persists on the track-session tree). Assisted mapping (G) then analyzes THAT source — decoded on demand into an analysis-only cache, playback and the visible waveform untouched — with the click-track contract: consolidated transients are beat pulses walked by the chart's authored beats-per-measure, so tempo changes in the click are followed directly instead of rejected as drift. Pulse dropout continues on the recent median gap at low confidence, honestly flagged. Ordinary (non-guide) G fits now carry a low-confidence completion tail to the final authored barline (never auto-committed — proposals only), and a new Accept Whole Fit verb (tempo toolbar button + registry command) commits every proposal as ONE undoable TempoMapCmd. Accepting through the final downbeat rescales the open last measure's interior beats onto the accepted tempo (equal-length invariant preserved). Anchor semantics: the focused barline always anchors the fit — locked or not — and a stale multi-selection no longer resets analysis toward the beginning or caps the march (it is only an anchor fallback when nothing has focus). In the metronome engine a locked barline keeps its authored time without disturbing the pulse cursor, so one stale lock cannot phase-shift every later suggestion. Guide-routing discipline: G awaits the guide decode and revalidates its preconditions after the await (mode still on, guide unchanged); a new-song boundary resets the guide cache and orphans in-flight decodes. Design-doc and user-guide sections updated to match. Stacked on feat/editor-track-session (#289) — the guide fields ride the persistent track-session tree it introduced. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EBQCHCNA81E9tHmSDHSe2Q
ChrisBeWithYou
pushed a commit
that referenced
this pull request
Jul 16, 2026
A click/reference stem can be locked as the session's TEMPO GUIDE (the ♩ button in the Audio tracks manager; the role persists on the track-session tree). Assisted mapping (G) then analyzes THAT source — decoded on demand into an analysis-only cache, playback and the visible waveform untouched — with the click-track contract: consolidated transients are beat pulses walked by the chart's authored beats-per-measure, so tempo changes in the click are followed directly instead of rejected as drift. Pulse dropout continues on the recent median gap at low confidence, honestly flagged. Ordinary (non-guide) G fits now carry a low-confidence completion tail to the final authored barline (never auto-committed — proposals only), and a new Accept Whole Fit verb (tempo toolbar button + registry command) commits every proposal as ONE undoable TempoMapCmd. Accepting through the final downbeat rescales the open last measure's interior beats onto the accepted tempo (equal-length invariant preserved). Anchor semantics: the focused barline always anchors the fit — locked or not — and a stale multi-selection no longer resets analysis toward the beginning or caps the march (it is only an anchor fallback when nothing has focus). In the metronome engine a locked barline keeps its authored time without disturbing the pulse cursor, so one stale lock cannot phase-shift every later suggestion. Guide-routing discipline: G awaits the guide decode and revalidates its preconditions after the await (mode still on, guide unchanged); a new-song boundary resets the guide cache and orphans in-flight decodes. Design-doc and user-guide sections updated to match. Stacked on feat/editor-track-session (#289) — the guide fields ride the persistent track-session tree it introduced. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EBQCHCNA81E9tHmSDHSe2Q
byrongamatos
pushed a commit
that referenced
this pull request
Jul 16, 2026
A click/reference stem can be locked as the session's TEMPO GUIDE (the ♩ button in the Audio tracks manager; the role persists on the track-session tree). Assisted mapping (G) then analyzes THAT source — decoded on demand into an analysis-only cache, playback and the visible waveform untouched — with the click-track contract: consolidated transients are beat pulses walked by the chart's authored beats-per-measure, so tempo changes in the click are followed directly instead of rejected as drift. Pulse dropout continues on the recent median gap at low confidence, honestly flagged. Ordinary (non-guide) G fits now carry a low-confidence completion tail to the final authored barline (never auto-committed — proposals only), and a new Accept Whole Fit verb (tempo toolbar button + registry command) commits every proposal as ONE undoable TempoMapCmd. Accepting through the final downbeat rescales the open last measure's interior beats onto the accepted tempo (equal-length invariant preserved). Anchor semantics: the focused barline always anchors the fit — locked or not — and a stale multi-selection no longer resets analysis toward the beginning or caps the march (it is only an anchor fallback when nothing has focus). In the metronome engine a locked barline keeps its authored time without disturbing the pulse cursor, so one stale lock cannot phase-shift every later suggestion. Guide-routing discipline: G awaits the guide decode and revalidates its preconditions after the await (mode still on, guide unchanged); a new-song boundary resets the guide cache and orphans in-flight decodes. Design-doc and user-guide sections updated to match. Stacked on feat/editor-track-session (#289) — the guide fields ride the persistent track-session tree it introduced. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EBQCHCNA81E9tHmSDHSe2Q
byrongamatos
added a commit
that referenced
this pull request
Jul 16, 2026
…it) (#290) * feat(editor): metronome guide + whole-song tempo fits (Accept Whole Fit) A click/reference stem can be locked as the session's TEMPO GUIDE (the ♩ button in the Audio tracks manager; the role persists on the track-session tree). Assisted mapping (G) then analyzes THAT source — decoded on demand into an analysis-only cache, playback and the visible waveform untouched — with the click-track contract: consolidated transients are beat pulses walked by the chart's authored beats-per-measure, so tempo changes in the click are followed directly instead of rejected as drift. Pulse dropout continues on the recent median gap at low confidence, honestly flagged. Ordinary (non-guide) G fits now carry a low-confidence completion tail to the final authored barline (never auto-committed — proposals only), and a new Accept Whole Fit verb (tempo toolbar button + registry command) commits every proposal as ONE undoable TempoMapCmd. Accepting through the final downbeat rescales the open last measure's interior beats onto the accepted tempo (equal-length invariant preserved). Anchor semantics: the focused barline always anchors the fit — locked or not — and a stale multi-selection no longer resets analysis toward the beginning or caps the march (it is only an anchor fallback when nothing has focus). In the metronome engine a locked barline keeps its authored time without disturbing the pulse cursor, so one stale lock cannot phase-shift every later suggestion. Guide-routing discipline: G awaits the guide decode and revalidates its preconditions after the await (mode still on, guide unchanged); a new-song boundary resets the guide cache and orphans in-flight decodes. Design-doc and user-guide sections updated to match. Stacked on feat/editor-track-session (#289) — the guide fields ride the persistent track-session tree it introduced. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EBQCHCNA81E9tHmSDHSe2Q * Revalidate asynchronous tempo guide analysis * fix(editor): unlock the tempo guide when its stem is renamed/deleted Locking a stem as the metronome guide records its source id in the track session, but the tracks-manager rename/delete controls rewrite S.stems (via _adopt) without touching the guide role, so tempoGuideSourceId can dangle. A dangling LOCKED guide is not harmless: assisted mapping (G) can't find the source live, and the save-time normalize silently repoints the still-locked role onto the first surviving source (usually the master recording) — so a reopened song would analyze the wrong track as a click. Reconcile at the choke point every stem-list mutation flows through (_adopt): if the locked guide's source vanished, unlock back to the default instead of transferring the lock. Reorder keeps ids, so it is a no-op there. Regression test drives the delete/rename path and pins that a live guide is left untouched. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(editor): address CodeRabbit review on tempo guide (PR #290) Three still-valid findings from the CodeRabbit pass: - audio.js: coalesce concurrent guide analyses. Two rapid G presses for the same (sourceId, url) each bumped _guideGeneration, so the older request superseded itself and returned null, briefly overwriting the newer success with an analysis error. Reuse a keyed in-flight promise instead; different guides / song switches still supersede via the token. - tempo-suggest.js: stop proposal generation when an authoritative lock conflicts with the proposed chronology. The completion tail no longer demotes a lock behind the last emitted time into an unlocked inferred proposal at a fabricated time, and the metronome walk no longer emits an equal/decreasing locked downbeat time when the pulse walk has run ahead — both terminate. - track-session.js: unlock a stale locked tempo guide at load. A persisted tempoGuideSourceId that no longer exists was silently repointed onto master while staying locked (reconcile then bailed because the id resolved); the load seam now clears the lock/mode, mirroring the stem-op reconcile. Skipped: the "account for missing pulses" finding (heavy heuristic lift) — inserting inferred pulse slots on oversized gaps can't distinguish a missed transient from a genuine tempo change, which the metronome engine is designed to follow; it would fight the module's stated contract and needs tuning against real audio. Tracked as a follow-up, not a minimal root-cause fix. Regression tests added to tests/metronome_guide.test.mjs (all fail pre-fix). 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>
This was referenced Jul 17, 2026
Merged
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this is
The foundation slice of the DAW Tracks re-cut: a song's tracks — the master recording, studio stems, and every transcription part, optionally grouped into folders — become one ordered, persistent tree (
src/track-session.js, persisted as theeditor_track_sessionmanifest extension key), instead of each surface re-deriving its own ad-hoc list.This supersedes the data-model portion of #285, rebuilt on the current main so it speaks the landed vocabulary (
S.stems/S.stemLinksfrom #283,partMixkeys, theinstallCreatedTrackSessionseam #286 reserved). The rest of #285 (Metronome Guide, tempo-suggest continuation, marquee/bulk locks, the Tracks lanes UI) follows as separate PRs on top of this model.Design (the load-bearing choices)
'master'/ bare manifest stem ids) and chart tracks (_partViewKeyPurekeys — the SAME dialecteditor_stem_linksspeaks). Normalization drops unknown references, appends new song parts, and repairs parent cycles — a corrupted tree can never hide a branch or crash a load.editor_stem_links; tree rows project it (_trackSessionRowsPure), so the two stores cannot disagree. A link pointing at a tombstoned source projects as unpaired instead of resurrecting it.(audioUrl, S.stems)on demand, so a/stem-oprename/delete can't leave a stale parallel copy.removedSourceIds; the media stays in the pack and restore re-appends the row. The tempo-guide role repairs to the next visible source.nullwhen the tree carries nothing the song doesn't already express, and the backend removes the key — untouched packs stay byte-identical. Absent ≠ null: an older client's absence has no authority to erase a persisted tree (_FIELD_ABSENT, same contract aseditor_stem_links).tempoGuideSourceId/Locked/Mode) ship in the schema now, inert — so the Metronome Guide PR is frontend-only.installCreatedTrackSessionis now real (was the inert feat(editor): unify project creation around imported tracks #286 stub): create/import seedsS.stemsfrom the server'saudio_sources(bare ids — thestem:prefix is create-payload namespace, not persisted identity), with an unconditional reset so a previous song's stems can't leak into a fresh import (S.stemswas previously only written byloadCDLCand stem-op adoption).Wiring
Every persistence boundary:
/save(parse + manifest apply), save-as,/build(meta + apply),/load(manifest →data.track_session→installTrackSession, after arrangements/stems/drumTab so normalization sees the whole song, withdata.audio_urlpassed explicitly becauseS.audioUrlstill points at the previous song at that line).Tests
tests/track_session.test.mjs(13, real-import ESM): normalization/canonical order, key dialects, pairing projection incl. tombstone case, folder ops + descendant guard + cycle repair, non-destructive remove/restore round-trip, is-default matrix, the create-seam stems reset, and the save/build wire checks.tests/test_track_session_persist.py(7): absent/null/malformed authority contract, YAML round-trip, id bounds (names-with-spaces pass — chart keys are arrangement names today — control chars and unbounded ids don't), shape filtering, 300-track/120-char caps, and no-pairing-field-on-rows.Gates:
npm test181 suites green,python -m pytest291 passed / 2 skipped,npm run lint0 errors (3 pre-existing baseline warnings).What this deliberately does NOT do
No new UI surface (the unified Tracks lanes/header column is the follow-up), no engine changes, no stable arrangement ids (that lands with the instrument-identity track and slots in via
_partViewKeyPure's existingarr.idpreference), and no migration ofmanifest["stems"]order — the manifest list stays ingest-order-authoritative for the stem manager until the unified surface supersedes it.🤖 Generated with Claude Code
https://claude.ai/code/session_01EBQCHCNA81E9tHmSDHSe2Q
Summary by CodeRabbit