- ignore unconfigured backends when clearing completed downloads
- keep the post-download cleanup route best-effort after a successful wishlist run
- add regression coverage for the orchestrator clear step
First lift in the new PR5 discovery-workers series. Pulls the 448-line
playlist sync background worker out of `web_server.py` into its own
focused module under `core/discovery/`. Pure 1:1 lift — wrappers keep
the original entry-point name so the four callers
(`sync_executor.submit(_run_sync_task, ...)`) continue to work without
changes.
What the sync worker does:
1. Convert frontend JSON tracks → SpotifyTrack/SpotifyPlaylist objects.
2. Normalize artist/album shapes for downstream wishlist parity.
3. Wire a progress_callback that updates `sync_states` + automation card.
4. Patch sync_service for database-only fallback when no media server is
connected.
5. `run_async(sync_service.sync_playlist(...))` and capture the result.
6. Update sync_states to 'finished', push playlist poster image to
Plex / Jellyfin / Emby, record sync history (with re-sync vs new-sync
branching), emit `playlist_synced` event for automation engine, and
persist sync status with a tracks_hash for smart-skip on the next
scheduled sync.
7. On exception → mark error in sync_states + automation; finally clear
progress callback + drop `_original_tracks_map` from sync_service.
Dependencies injected via `SyncDeps` (11 fields) — config_manager,
sync_service, plex_client, jellyfin_client, automation_engine, run_async,
record_sync_history_start, update_automation_progress,
update_and_save_sync_status, sync_states dict, sync_lock. The only
structural drift from a pure paste is the top-of-function variable
binding: original used `global sync_states, sync_service`, lifted version
rebinds them as locals from deps (`sync_states = deps.sync_states` etc.)
since the names aren't module-level in the new file. Same behaviour
otherwise — diff against the original after `deps.X` → global X
normalization is **zero differences**.
Tests: 18 new under tests/discovery/test_discovery_sync.py covering
sync history recording (new + resync), setup error path (with and
without automation_id), missing sync_service handling, sync_playlist
exception handling, successful sync state transition, unmatched-tracks
summary, playlist image upload (plex + jellyfin + zero-synced gate),
automation engine emit, automation progress finished call, sync history
DB persistence (completion + match_details), tracks_hash persistence,
and finally-block cleanup (callback clear + map drop).
Full suite: 1068 passing (was 1050). Ruff clean.
Kicks off the PR5 series — 9 discovery workers totaling ~2,400 lines
across `_run_sync_task`, `_run_*_discovery_worker` family,
`_run_quality_scanner`, and `_process_watchlist_scan_automatically`.
Wishlist-related extractions deliberately skipped to avoid overlap with
kettui's planned `core/wishlist/` package.
- add module-level loggers for the wishlist package instead of threading the web server logger through runtime objects
- default wishlist helper runtimes and cleanup helpers to their package logger while still allowing test overrides
- keep web_server.py as a thin caller that no longer injects its logger into wishlist flows
- extract the remaining wishlist endpoint behavior from web_server.py into core/wishlist/routes.py
- keep web_server.py as a thin Flask adapter around the new route helpers
- add tests that cover wishlist counts, stats, track listing, clear/remove flows, cycle updates, and album-track adds
- add core/wishlist as the home for wishlist payload, resolution, state, processing, reporting, and selection helpers
- move wishlist-specific tests into tests/wishlist alongside the new package layout
- keep web_server.py and the import/search callers as thin adapters for now
Three drifts caught in line-by-line review against original — all reverted
so the lifted body is now byte-identical to the web_server.py original
(after deps.X → global X normalization):
1. `import traceback` was hoisted to module-level imports. Original imports
it inline inside the except block. Moved back to inline (same behavior,
but strict location parity).
2. Trailing whitespace on the blank line after analysis_results.append.
3. Trailing whitespace on the blank line inside the wishlist except handler.
Diff against original now produces ZERO differences. Tests still 21/21
passing, ruff clean.
Final extraction in the download orchestrator series. Lifts the 586-line
master worker that drives the entire missing-tracks pipeline from
`web_server.py` into `core/downloads/master.py`. Pure 1:1 lift — wrappers
keep the original entry-point name so the three callers
(`missing_download_executor.submit(_run_full_missing_tracks_process, ...)`)
continue to work without changes.
What the master worker does:
1. PHASE 1 ANALYSIS — per-track DB ownership check with album fast path
(lookup album by name+artist, match tracks within it) plus a
MusicBrainz release-cache preflight so per-track post-processing all
uses the same release MBID (prevents Navidrome album splits).
2. Wishlist removal for tracks already in the library.
3. Explicit-content filter.
4. PHASE 2 transition — if nothing missing, mark batch complete, update
per-source playlist phases, kick auto-wishlist completion handler.
5. Soulseek album pre-flight — search for a complete album folder before
falling back to track-by-track search, cache the source for reuse.
6. Wishlist album grouping — derive per-album disc counts and resolve ONE
artist context per album so collab albums don't fold-split.
7. Task creation with explicit album/artist context injection +
playlist-folder-mode flag propagation.
8. Hand off to download_monitor + start_next_batch_of_downloads.
9. Error handler — phase=error, reset YouTube playlist phase to
'discovered', reset auto-wishlist globals on auto-initiated batches.
Dependencies injected via `MasterDeps` (21 fields) — wide surface
covering config, MB caches/locks, soulseek client, source-page state
dicts, multiple callbacks (wishlist removal, explicit filter, executor
+ auto-completion fn, monitor, start_next_batch). The only behaviour
difference from a pure paste is `import traceback` hoisted to module
scope (was inline in the except block) — same behaviour. Trailing
whitespace on two blank lines also got normalized away by the editor;
neither has any runtime effect.
`reset_wishlist_auto_processing` callback wraps the
`global wishlist_auto_processing, wishlist_auto_processing_timestamp`
write + `wishlist_timer_lock` since `global` can't reach back into
web_server.py from a separate module.
Tests: 21 new under tests/downloads/test_downloads_master.py covering
analysis-phase state, force_download_all, found-track wishlist removal,
explicit filter, no-missing complete + per-source state updates, auto
wishlist completion submit, album fast path (direct + fallthrough),
MB preflight (caches both keys, no-mb-worker no-op), task creation
(queue + tasks dict, explicit context for albums, wishlist album
grouping consistency, playlist folder mode), monitor + next-batch
handoff, multi-disc total_discs computation, error handler (phase set,
youtube reset, auto wishlist reset), and batch-removed-mid-flight
defensive path.
Full suite: 1050 passing (was 1029). Ruff clean.
End of the PR4 series — `web_server.py` lost ~590 lines on this commit
alone; total trim across PR4a–PR4h is ~2900 lines of orchestrator code
moved into focused `core/downloads/*.py` modules.
Seventh sub-PR in the download orchestrator series. Strict 1:1 lift —
zero behavior change. ~570 lines moved out of web_server.py.
What moved (lifted as 3 tightly-coupled functions in one module):
- _start_next_batch_of_downloads → start_next_batch_of_downloads
- _on_download_completed → on_download_completed
- _check_batch_completion_v2 → check_batch_completion_v2
Dependencies bundled in `LifecycleDeps` (15+ refs):
- config_manager, automation_engine, download_monitor, repair_worker,
mb_worker (live globals)
- is_shutting_down (lambda over IS_SHUTTING_DOWN flag)
- get_batch_lock (web_server helper for batch_locks dict)
- submit_download_track_worker (lambda wrapping
missing_download_executor.submit + _download_track_worker)
- submit_failed_to_wishlist + submit_failed_to_wishlist_with_auto_completion
(async, used by on_download_completed) AND process_failed_to_wishlist
+ process_failed_to_wishlist_with_auto_completion (sync, used by
check_batch_completion_v2 — direct call matches original v2 behavior;
the non-v2 path always submitted to executor)
- ensure_spotify_track_format, get_track_artist_name,
check_and_remove_from_wishlist, regenerate_batch_m3u (web_server
helpers — large, will lift in follow-up PRs)
- youtube_playlist_states, tidal_discovery_states,
deezer_discovery_states, spotify_public_discovery_states
(per-source playlist state dicts — phase transitions on batch
completion)
Direct imports for already-lifted helpers:
- core.runtime_state.{download_tasks, download_batches, tasks_lock,
add_activity_item}
- core.downloads.history.record_sync_history_completion (PR4a)
- core.album_consistency.run_album_consistency
- core.metadata.common.get_file_lock
Behavior parity verified line-by-line:
- start_next_batch: same batch lock acquisition, same shutdown gate,
same V2-cancelled-task skip, same searching-status-set-before-submit,
same submit-fails-no-ghost-worker semantics
- on_download_completed: same duplicate-call detection (skip decrement
but still check completion), same failed-track tracking with
spotify_track formatting + activity items + automation event emission,
same wishlist removal on success, same active_count decrement, same
stuck-detection (searching > 10min → not_found, post_processing >
5min → completed), same M3U regeneration + repair worker hand-off
+ album consistency pass + wishlist failed-tracks submission
- check_batch_completion_v2: same finished-count tally, same stuck
detection, same already-complete short-circuit returning True, same
per-source playlist phase updates, same album consistency pass,
same DIRECT (sync) wishlist processing call (NOT submit-to-executor
— matches original v2 which called process_* functions directly)
CRITICAL drift caught + fixed during review:
- Initial lift had v2 routing wishlist calls through submit_* deps
(async). Original v2 called process_* directly (sync). Added separate
process_* deps to LifecycleDeps and routed v2 to them. Tests updated.
Two minor defensive additions documented:
- `is_auto_batch = False` initialized before conditional in v2 (Python
scope rules made this unnecessary in original, but explicit is safer)
- Variable rename inside the queue-completion-check loop in
on_download_completed: `task_id` → `queue_task_id` to avoid shadowing
the outer parameter. Log output preserves the same task ID.
Tests: 28 new under tests/downloads/test_downloads_lifecycle.py
covering start-next (early-returns, shutdown gate, max_concurrent,
cancelled-task skip, searching-status-set, submit-failure-no-ghost,
orphan task), on-complete (decrement, duplicate skip, failed/cancelled
tracking, automation emit, wishlist removal, batch completion + emit
+ source phase update, stuck detection, auto vs manual routing),
check-v2 (missing batch, not-complete, complete-marking, already-
complete, auto routing, exception handling).
Full suite: 1029 passing (was 1001). Ruff clean.
Fifth sub-PR in the download orchestrator series. Strict 1:1 lift —
zero behavior change.
What moved:
- _build_batch_status_data → build_batch_status_data
- get_batch_download_status route body → build_single_batch_status
- get_batched_download_statuses route body → build_batched_status
- get_all_downloads_unified route body → build_unified_downloads_response
- Status priority dict → module-level _STATUS_PRIORITY constant
Dependencies bundled in `StatusDeps` dataclass:
- config_manager, docker_resolve_path, find_completed_file,
make_context_key, submit_post_processing (lambda wrapping
missing_download_executor.submit + _run_post_processing_worker),
get_cached_transfer_data
Direct imports from core.runtime_state for download_tasks /
download_batches / tasks_lock (already lifted by kettui).
Behavior parity:
- Same response payload shape across all 3 endpoints
- Same safety-valve mutation: stuck downloading task with file recovered
→ status='post_processing' + submit worker; stuck searching → not_found;
stuck downloading no file → failed
- Same live transfer state mapping (Cancelled/Canceled, Failed/Errored/
Rejected/TimedOut, Completed/Succeeded with byte-mismatch verification,
InProgress, default queued)
- Same intermediate post_processing status promotion + single-shot worker
submission (only when status != 'post_processing')
- Same 'Errored' handling: keeps current status to let monitor retry
- Same 17-key item dict in unified response with same field order
- Same artist/album/artwork normalization (handles string, dict, list,
list-of-dicts, list-of-strings variants)
- Same sort: (priority asc, -timestamp desc)
- Same batch summary aggregation
- Same items[:limit] slicing
- Same logger messages text-for-text
- Same lock scope (single tasks_lock per call) — no new contention
Pre-existing bug preserved (will fix in follow-up PR):
- batched_status `debug_info` block iterates `response["batches"]` and
guards with `if "error" not in batch_status`. Every successful
payload includes `"error": batch.get('error')` (key always present,
value usually None) so the guard is always False and debug_info
never populates in production. Test documents the buggy behavior so
the next PR can flip the check to `batch_status.get('error') is None`.
Tests: 32 new under tests/downloads/test_downloads_status.py covering
phase routing (analysis vs downloading vs unknown), task formatting +
sort + V2 fields, every live transfer state mapping (Cancelled,
Succeeded with full + partial bytes, InProgress, Errored, terminal-
not-overridden), safety valve (stuck searching → not_found, stuck
downloading recovered → post_processing, stuck downloading no file →
failed), all 3 route helpers (single, batched, unified), unified
artist/album/artwork normalization, batch summary aggregation, limit
slicing, plus debug_info bug documentation.
Full suite: 982 passing (was 950). Ruff clean.
Fourth sub-PR in the download orchestrator series. Strict 1:1 lift —
zero behavior change. ~407 lines moved out of web_server.py.
What moved:
- _run_post_processing_worker → run_post_processing_worker
The lifted function is intentionally kept as one ~400-line block to
preserve byte-for-byte parity with the original. Refactoring it into
smaller helpers (context lookup, file search loop, transfer-folder
handler, downloads-folder handler) gets its own follow-up PR.
Dependencies: 9 callbacks bundled in `PostProcessDeps` dataclass.
- config_manager, soulseek_client, run_async (live refs)
- docker_resolve_path, extract_filename, make_context_key
(small utilities still in web_server.py — will lift in a future PR
alongside other shared utilities)
- find_completed_file (file search helper, still in web_server.py)
- enhance_file_metadata, wipe_source_tags (web_server wrappers around
core.metadata.enrichment)
- post_process_with_verification (web_server wrapper around
core.imports.pipeline)
- mark_task_completed (wraps runtime_state.mark_task_completed +
session counter)
- on_download_completed (deferred to PR4g batch lifecycle)
Direct imports for already-lifted helpers (no injection needed):
- core.imports.album_naming.resolve_album_group
- core.imports.context.{get_import_clean_title, get_import_clean_album,
get_import_original_search, get_import_context_artist,
get_import_context_album, normalize_import_context}
- core.imports.filename.extract_track_number_from_filename
- core.metadata.enrichment (re-exported as metadata_enrichment)
- core.runtime_state.{download_tasks, tasks_lock,
matched_downloads_context, matched_context_lock}
Behavior parity:
- Same control flow: missing-task short-circuit → cancelled/completed
short-circuit → missing-filename failure → docker path resolution →
context lookup with fuzzy fallback → expected filename generation →
YouTube special-case path resolution → 5-attempt search loop with
Strategy 1 (original filename in download+transfer) and Strategy 2
(expected final filename in transfer) → file-not-found failure →
transfer-folder handler with metadata enhancement → downloads-folder
handler with full post-process verification
- Same retry count (5), sleep duration (5s), per-attempt logging
- Same album_info dict construction with is_album=True for explicit
album downloads
- Same album grouping skip when context.is_album_download is True
- Same wipe_source_tags fallback when enhancement context missing
- Same matched_downloads_context cleanup on success
- Same exception swallowing at processing-error and critical-error
layers, both setting status='failed' + error_message + calling
on_download_completed(b, t, success=False)
- Every logger message text preserved verbatim (so log filters keep
working)
Tests: 16 new under tests/downloads/test_downloads_post_processing.py
covering missing task, cancelled, already-completed, stream_processed,
missing filename + username, file-not-found-after-retries with sleep
mocked, stream-processor-completes-mid-search, transfer-folder with
metadata enhanced + with no context (wipes tags), downloads-folder
with + without context, processing exception, critical outer
exception, YouTube special path, fuzzy context matching.
Full suite: 950 passing (was 934). Ruff clean.
Third sub-PR in the download orchestrator series. Strict 1:1 lift —
zero behavior change.
What moved:
- _automatic_wishlist_cleanup_after_db_update → cleanup_wishlist_after_db_update
The lifted fn takes config_manager as an arg (so core/downloads/cleanup.py
doesn't need to import web_server). Other deps (wishlist_service,
MusicDatabase, get_database) stay as in-function imports — matches the
original deferred-import pattern.
The single caller in web_server.py (missing_download_executor.submit at
L18028) keeps using the same wrapper name with no signature change.
Behavior parity:
- Same per-profile iteration via get_all_profiles()
- Same essential-field skip (no name / no artists / no spotify_track_id)
- Same artist normalization (string / dict / fallback to str())
- Same 0.7 confidence threshold for db match
- Same break-on-first-artist-match semantics
- Same album extraction (dict.name vs string passthrough)
- Same active_server pulled via config_manager.get_active_media_server()
- Same per-track exception swallowing inside the loops
- Same top-level exception swallow with traceback.print_exc()
- Same logger messages (exact text match for "[Auto Cleanup]" prefix)
Tests: 13 new under tests/downloads/test_downloads_cleanup.py covering
empty wishlist short-circuit, found-in-db removal, missed track stays,
low-confidence skip, missing-fields skip, dict + string artist formats,
break-on-first-match, multi-profile walk, album dict/string handling,
db check failure continuing to next artist, top-level exception swallow,
active server propagation.
Full suite: 934 passing (was 921). Ruff clean.
Second sub-PR in the download orchestrator series. Strict 1:1 lift —
zero behavior change.
What moved:
- cancel_download (single slskd cancel) → cancel_single_download
- cancel_all_downloads (cancel + clear + sweep) → cancel_all_active
- clear_finished_downloads (slskd clear + sweep) → clear_finished_active
- clear_completed_downloads (local task tracker prune) → clear_completed_local
Slskd-touching helpers take (soulseek_client, run_async, sweep_callback)
explicitly so the route layer wires the live client + the existing
_sweep_empty_download_directories helper. The local-state helper imports
download_tasks/download_batches/batch_locks/tasks_lock straight from
core.runtime_state since those are module-level shared globals.
Prep change: `batch_locks` dict moved from web_server.py global into
core/runtime_state.py alongside the other download globals. web_server.py
re-imports from runtime_state so the ~3 existing call sites in
web_server.py keep resolving without modification. Identity preserved
(same dict across all importers).
Out of scope (deferred to PR4g batch lifecycle):
- cancel_download_task (calls _on_download_completed)
- cancel_task_v2 + _atomic_cancel_task + _find_task_by_playlist_track
(manipulate batch active_count directly, deeply coupled to lifecycle)
Behavior parity:
- Same response shapes + status codes on each route
- Same call order (cancel_all → clear_all_completed → sweep)
- Same conditional sweep on clear_finished (skipped on failure)
- Same sweep ALWAYS runs after cancel_all even if clear_all returns False
(matches original — clear failure was non-fatal in cancel_all path)
- Same TERMINAL_STATUSES set: completed/failed/not_found/cancelled/skipped/
already_owned (lifted to module-level constant)
- Same empty-batch pruning + same batch_locks cleanup
- Same lock acquisition pattern (single tasks_lock)
Tests: 14 new under tests/downloads/test_downloads_cancel.py covering
single cancel, cancel-all happy + failure paths, clear-finished + sweep
gate, local task pruning across all 7 active/terminal states, batch
queue trimming, batch_locks cleanup.
Full suite: 921 passing (was 907). Ruff clean.
First sub-PR in the download orchestrator series. Strict 1:1 lift —
zero behavior change.
What moved:
- _record_sync_history_start → record_sync_history_start
- _record_sync_history_completion → record_sync_history_completion
- _detect_sync_source → detect_sync_source
- Source prefix map → module-level _SOURCE_PREFIX_MAP constant
What stayed:
- web_server.py keeps three thin wrappers (_detect_sync_source,
_record_sync_history_start, _record_sync_history_completion) that
delegate into core/downloads/history.py. ~60 callers of these names
in web_server.py keep resolving without touching every site.
Each lifted function takes `database` as an arg (was
`db = MusicDatabase()` inline). The wrappers construct
`MusicDatabase()` per call to mirror the exact original behavior —
each invocation got a fresh DB connection.
Behavior parity:
- Same SQL UPDATE statement (preserves the in-place update path when
a sync_history entry already exists for the playlist_id)
- Same JSON serialization with ensure_ascii=False
- Same thumb URL extraction order (album_context.images → image_url
→ first track album.images)
- Same per-track result shape (index, name, artist, album, image_url,
duration_ms, source_track_id, status, confidence, matched_track,
download_status)
- Same status mapping (found/not_found, completed/failed)
- Same best-effort exception swallowing (sync history failure must
never break the actual download)
- Reads `download_tasks` from core.runtime_state (already lifted by
kettui in PR378)
Tests: 34 new under tests/downloads/test_downloads_history.py
covering source detection (16 prefixes), start happy paths + thumb
extraction + duplicate-update + DB error swallowing, completion stats
+ per-track results JSON shape + edge cases.
Full suite: 907 passing (was 873). Ruff clean.
User report: SoulSync was only pulling MusicBrainz genres from the
recording (track-level) endpoint. Most MB recordings don't carry genres
at the track level — they live on the release (album) or artist. So
the MB tier was contributing nothing to the genre merge for the
overwhelming majority of tracks.
Fix:
- Added `'genres'` to the release-detail `includes` (was missing).
- After release-detail processing, if pp['mb_genres'] is still empty,
populate from release_detail['genres'] (sorted by count desc).
- If still empty AND artist_mbid is set, fetch artist with
`includes=['genres']` and use those.
No extra API call when the recording (or release) already had genres —
the artist fetch only fires when both upstream tiers came back empty.
The downstream genre merge in _embed_metadata_genres is unchanged; this
just makes the MB feed into it richer.
Tests: 4 new (recording present, recording empty → release, recording
+ release empty → artist, all empty → []). Full suite 873 passing.
Ruff clean.
Reported by @kcaoyef421 in Discord.
The endpoint was returning a 200-line literal dict inline. Moved the
three lists (TRIGGERS, ACTIONS, NOTIFICATIONS) to module-level constants
in core/automation/blocks.py. Route shrinks to 7 lines. Data is now
importable for tests + future docs.
Added 8 shape tests so a typo in the dict (missing 'type', wrong
field type, missing options on a select, etc.) gets caught by CI
instead of breaking the builder UI silently.
The `known_signals` field stays computed at request time via
_collect_known_signals(database) since it's dynamic.
No behavior change. Same response shape. 869 tests passing (was 861).
Ruff clean.
Routes moved to thin parse-args/jsonify handlers; logic now lives in
three focused modules under core/automation/. 436 lines deleted from
web_server.py; 53 added back as wrappers.
Module split:
- core/automation/api.py — CRUD + run + history helpers. Each function
takes (database, automation_engine, ...) explicitly and returns
(response_body, http_status). Includes signal cycle detection
preflight checks for create + update.
- core/automation/progress.py — owns the in-memory progress state dict
+ lock (mirroring the original web_server.py globals as module-level
shared state so all callers see one view), init/update/history
helpers, and the WebSocket emit loop.
- core/automation/signals.py — collect_known_signals for the builder
autocomplete.
Out of scope (deferred):
- _register_automation_handlers — the 23+ action handler closures stay
in web_server.py because each one is tightly coupled to feature-
specific implementations (wishlist, watchlist, library scan, etc.).
- Worker functions (_process_wishlist_automatically, etc.) — belong
with their feature lifts.
- _run_sync_task / _run_playlist_discovery_worker — sync + discovery
PRs.
Behavior preserved 1:1:
- Same route response shapes + status codes
- Same JSON field hydration (trigger_config, action_config,
notify_config, last_result, then_actions)
- Same backward-compat: empty then_actions + notify_type set →
synthesize then_actions from notify_type/notify_config
- Same signal cycle detection behavior on create + update
- Same system-automation protection on delete + duplicate
- Same reschedule/cancel logic on toggle + bulk-toggle + update
- Same progress state shape (status, progress, phase, current_item,
log capped at 50, started_at/finished_at, action_type)
- Same emit-on-finish socketio push from update_progress
- Same emit loop semantics (1s tick, snapshot active states, reap
finished after window)
Pre-existing bugs preserved (will fix in follow-up PRs):
- emit_progress_loop uses naive datetime.now() against tz-aware
started_at/finished_at, so the timeout-zombie check raises
TypeError → caught → never fires, and the cleanup-after-window
check raises → caught → state is reaped on FIRST tick regardless
of the window. Tests document this behavior so the next PR can
flip them to the corrected expectation.
Tests: 72 new under tests/automation/ (signals 10, progress 24,
api 38). Full suite: 861 passing (was 789). Ruff clean.
Three drifts caught in line-by-line review against the pre-lift
web_server.py. All addressed for strict 1:1 behavior parity.
1. /api/enhanced-search/source/<src> now returns plain JSON
`{"artists":[],"albums":[],"tracks":[],"available":false}` (or
`{"videos":[],"available":false}` for youtube_videos) when the
source's client isn't available, matching the original endpoint
contract. Previously streamed an NDJSON `{"type":"done"}` line
instead.
Restructured by splitting the orchestrator into resolve+stream
helpers:
- `resolve_client(source_name, deps)` — already existed, used
for /api/enhanced-search single-source mode
- `resolve_youtube_videos_client(deps)` — new, returns the
soulseek_client.youtube subclient or None
- `stream_metadata_source(source_name, query, client)` — pure
NDJSON generator, caller resolves client first
- `stream_youtube_videos(query, youtube_client, run_async)` —
same shape for the yt-dlp path
The route now decides plain-JSON-vs-stream based on resolution
result, mirroring the original control flow exactly.
2. core/search/library_check.py — reverted the defensive `(x or '')`
and `getattr(plex_client, 'server', None) is not None` patterns
to original byte-for-byte (`x.get('name', '')`,
`plex_client.server`, no try/except around `get_plex_config`).
Lift PR shouldn't change crash semantics; if the original raises
on malformed input, mine should too. Pre-existing edge cases get
their own follow-up PR.
3. core/search/stream.py — same revert: `soulseek_client.youtube`
instead of `getattr(..., 'youtube', None)` etc.
Also removed the module-level `EMPTY_SOURCE` from sources.py and
moved its (per-call) duplicate into _fan_out_response as a local —
the original used a per-request local dict and the identity-check
behavior depends on that. Module-level was a footgun for future
mutations.
789 tests still pass (95 search), ruff clean.
Line-by-line review of the search lift caught one drift: cache.get_cache_key
was coercing falsy provider returns ('', None, 0) to 'unknown' / False.
Original web_server.py only fell back to those sentinels on exception, not
on falsy success values.
Real-world impact: low — get_active_media_server() and get_primary_source()
return non-empty strings in practice. But cache keys are tuples with no
schema enforcement, so any drift here can silently fragment the cache.
Restored 1:1 parity with original semantics.
Added test covering the falsy-success path so this can't drift again.
789 tests pass, ruff clean.
Routes moved to thin parse-args/jsonify handlers; logic now lives in
six focused modules under core/search/. 720 lines deleted from
web_server.py; 109 added back as wrappers; ~700 lines of new core code
plus ~700 lines of tests.
Module split:
- core/search/cache.py — TTL+LRU cache for enhanced-search responses,
keyed by (query, active_server, fallback_source, hydrabase_active,
source_tag) so config changes don't poison stale entries.
- core/search/sources.py — per-kind metadata search (artists/albums/
tracks) and the multi-kind ThreadPoolExecutor that fans them out.
- core/search/library_check.py — library + wishlist presence check
with Plex thumb URL resolution; profile-aware wishlist with legacy
fallback for older DBs missing the profile_id column.
- core/search/stream.py — single-track preview search; effective stream
mode resolution, query-variant generation, retry walk, matching
engine integration.
- core/search/basic.py — flat Soulseek file search, quality-sorted.
- core/search/orchestrator.py — main enhanced-search dispatch
(short-query fast path, single-source bypass, hydrabase-primary fan
out, alternate source list builder), NDJSON streaming generator
for /source/<src>, and the SearchDeps dataclass that bundles the
cross-cutting deps.
Routes pass clients (spotify, hydrabase, hydrabase_worker, soulseek)
and helpers (config_manager, fix_artist_image_url,
_is_hydrabase_active, _get_metadata_fallback_*, _run_background_
comparison, run_async, dev_mode_enabled_provider) into core/search via
a SearchDeps bundle built per-request. fix_artist_image_url stays in
web_server.py because it touches 31 other call sites.
Behavior preserved 1:1:
- Same response shapes (db_artists, spotify_artists, spotify_albums,
spotify_tracks, primary_source, metadata_source, alternate_sources,
source_available)
- Same NDJSON line ordering (artists/albums/tracks as they finish, plus
done marker)
- Same per-kind exception swallowing
- Same hydrabase-worker mirror on dev mode
- Same cache key shape (5-tuple) and TTL/LRU semantics
- Same stream-track effective-mode resolution including the
Soulseek-coerce-to-YouTube edge case
- Same library-check Plex thumb URL rewriting and wishlist fallback
for older DBs
Tests: 94 new (cache TTL/LRU/key, sources happy/partial/all-fail,
library presence with library + wishlist + thumbs, stream effective
mode + query gen + retry, orchestrator client resolution + short
query + single source + fan-out alternates + hydrabase primary +
NDJSON drain). Full suite: 788 passing (was 694).
Ruff clean.
Stats route logic moves into core/stats/queries.py as pure-ish functions
that take dependencies (database, image-url fixer, listening worker) as
arguments. The 13 route handlers in web_server.py shrink to thin
parse-args / jsonify wrappers.
What moved to core/stats/queries.py:
- stats_cached: 3-key metadata cache lookup + image url fix-up
- stats_overview / timeline / genres / library_health / db_storage
- stats_top_artists / top_albums / top_tracks: top-N + DB enrichment
- stats_recent: listening_history readback
- stats_resolve_track: title+artist -> file_path lookup for playback
- listening_stats_sync: spawns daemon thread that runs worker._poll
- listening_stats_status: stats payload, with None-worker fallback shape
No behavior change. Same response shapes, same error handling, same
silent-except on per-row enrichment failure. fix_artist_image_url
stays in web_server.py and is passed through as a callback so we
don't have to lift its config_manager / media-server dependencies in
this PR.
Adds tests/stats/test_stats_queries.py — 27 tests covering happy
paths, edge cases, image-url plumbing, worker glue.
Ruff clean. 694 tests pass (was 667 + 27 new).
Lifted-then-not-deleted leftovers from the PR378 merge:
- web_server.py `_resolve_album_group` and `_build_final_path_for_track`
were already imported at module top from `core/imports/`. Removed the
shadowing local copies.
- Mutagen reimports (FLAC/MP4/OggVorbis) at L17736-17738 shadowed the
top-of-file imports. Picture/MP4Cover/MP4FreeForm were unused. Dropped
the whole block.
- core/imports/context.py: `getattr(artist, "name")` -> `artist.name`
(B009).
Ruff clean, 667 tests pass.
- Normalize album import track display handling so queue labels and match rows stay consistent
- Bound MusicBrainz caches and avoid caching transient lookup failures
- Stop swallowing programmer errors in source enrichment helpers
- Restore import config test seams without reintroducing lazy imports
- Guard task completion calls and fix the Windows path test expectation
- Keep file lock tracking from growing without bound
- Cover search_result fallback normalization and ambiguous album detection.
- Add staging metadata, multi-disc path, and MusicBrainz enrichment cases.
- Move the single-track context test next to the imports code it exercises.
- keep single-track import lookup in imports/resolution.py
- normalize simple-download search_result data before wishlist matching
- run wishlist cleanup for simple-download post-processing
- keep source-only artist detail on resolved names and MB short-circuit
- pass the selected manual match through singles import
- keep the import context source-aware so artist and album stay correct
- avoid treating non-Spotify IDs as wishlist Spotify IDs
- make wishlist logging and local variable names source-neutral
- Move the import pipeline runtime factory into core.imports.pipeline
- Move the metadata runtime factory into core.metadata.enrichment
- Keep the web server wiring thin and drop the shared glue module
- Add contract tests that keep the two runtime bundles separate
- Relocate the shared metadata helper module from core/metadata_common.py into core/metadata/common.py.
- Update the new metadata package, the import pipeline, and the web entrypoint to use the package-scoped helper.
- Keep the shared config, mutagen, file-lock, and tag-writing helpers centralized without touching unrelated files.
- Pass the live runtime bundle into the shared metadata facade so worker-backed source enrichment can actually run.
- Forward runtime from the import pipeline and web-server wrapper into embed_source_ids.
- Add a regression test that verifies the runtime object reaches the source-ID embedding path.
- Keep existing metadata_cache and metadata_service at the top level for now
- Move the new branch-local metadata helpers under core/metadata
- Share MusicBrainz release cache state from core.metadata.source and update import sites
- Move app-wide task and activity registries out of core/imports
- Share one runtime-state module across the web server, API, and import pipeline
- Keep import-specific helpers focused on context and post-processing
- Move import flow modules into a dedicated package
- Update app and test imports to the new namespace
- Group the import-focused tests under tests/imports
- remove stale wrapper helpers from web_server and metadata_common
- import provider helpers directly in metadata_source
- keep the metadata modules' public surface explicit
- remove runtime from metadata helper APIs where it only carried config, logger, mutagen, and database access
- keep runtime only for the source-ID enrichment path that still needs live worker handles
- add the new metadata helper modules and update the tests to match the slimmer interfaces
- Move filename and staging helpers into their canonical modules
- Extract album naming and grouping from path handling
- Update import and test call sites to the new layout
- Extract the import pipeline, album import, staging, path, file ops, guards, runtime state, side effects, and metadata enrichment out of .
- Canonicalize the refactored import path around and remove legacy , , , and request shapes from the import endpoints.
- Make album and track metadata lookups follow the configured provider priority instead of hard-coding Spotify, while still falling back when needed.
- Update the import routes and frontend payloads to use the new core helpers.
- Add coverage for the extracted helpers and the refactored import flows.
PS. apologies to anyone who might check this commit out - the intention was to start small, but things kinda snowballed out of control at some point since the logic just kept going on and on, and everything kinda had to be changed all at once for it all to make any sense
Six items from a Cin-style line-by-line pass on PR #383:
- resolve_cors_origins: list of non-string entries (`[None, 123]`) now
drops them instead of coercing to junk strings like `'None'`/`'123'`.
- will_reject: backwards-compat shim removed. Production callers always
pass `request.scheme` (Flask-guaranteed); the shim only existed for
tests/non-Flask callers and made the production code path branchier
than necessary. Tests now pass scheme explicitly.
- maybe_log: redundant `if not origin` early-return dropped. will_reject
handles missing origin (engineio's own behavior — server.py:207).
- RejectionLogger.__init__: `int(dedup_cap)` wrapped in try/except so
bad-type input falls back to DEFAULT_DEDUP_CAP instead of raising.
- web_server.py: docstring on the before_request hook explains why the
hook fires on every request (Flask doesn't scope before_request to a
path prefix; the early-return string compare is the cheapest option).
- settings.js: cors-origins URL regex tightened from `[^\s/]+` to
`[^\s/?#]+` so query/fragment chars don't pass validation. Engineio
would silently fail to match those anyway; better to flag at save.
Test changes:
- parametrize gained an explicit `scheme` column (12 cases updated).
- New explicit case: scheme-mismatch rejects (engineio compares full
`{scheme}://{host}` strings).
- `test_will_reject_falls_back_to_host_only_when_no_scheme_info`
deleted — the shim it tested is gone.
- `test_will_reject_honors_x_forwarded_host` now passes scheme info.
Net: -9 production lines, -3 test lines. Production code path is
straight-line. 603 tests pass.
Self-review pass on the security fix uncovered five issues, all fixed
here:
1. will_reject scheme handling. Engineio compares full {scheme}://{host}
strings, not just hostnames. A TLS-terminating proxy can leave the
backend seeing http while the browser's Origin is https — engineio
rejects, but the original predictor said "allow" → no helpful log
line. Added request_scheme + forwarded_proto params, build full
candidate strings to match engineio.
2. EITHER-forwarded-header rule. Engineio adds the forwarded candidate
when EITHER X-Forwarded-Proto OR X-Forwarded-Host is present (it
falls back to HTTP_HOST for the missing one). The original predictor
only added it when forwarded_host was set — false negative for
misconfigs sending only X-Forwarded-Proto. Now mirrors engineio.
3. will_reject incorrectly rejected missing-Origin requests. Engineio
(server.py:207: `if origin: validate`) skips CORS validation when
no Origin header is sent — non-browser clients (curl etc.) are
intentionally permitted. The original code rejected them. Test was
asserting the wrong behavior. Both fixed.
4. RejectionLogger had unbounded dedup set growth. A hostile actor
opening connections from many distinct fake origins would fill
memory unboundedly. Capped at 100 unique origins (configurable);
when cap hit, one overflow notice is emitted and further rejections
are silently dropped until restart.
5. Lock pattern: the overflow log path called logger.warning() while
holding the dedup lock, inconsistent with the normal path. Fixed
to pick the message under the lock and log after release. Critical
section is now minimal and uniform.
Plus polish:
- Stale module docstring fixed (said "empty list" instead of "None").
- settings.js validates each cors_origins line against a URL regex on
save; toasts a one-shot warning if entries are malformed (resolver
silently filters them, but user gets feedback now).
- web_server.py wiring passes request.scheme + X-Forwarded-Proto so
the predictor has full proxy info.
Tests:
- 51 unit tests in tests/test_socketio_cors.py (was 45). New cases:
* scheme comparison (5 cases including TLS-terminating proxies)
* forwarded_proto-alone misconfig
* missing-origin matches engineio (was asserting wrong behavior)
* dedup cap with overflow + reset
* default cap is reasonable (uses public DEFAULT_DEDUP_CAP constant)
Engineio behavior independently verified by reading engineio/server.py
and engineio/base_server.py source. Predictor mirrors both files.
604 tests pass.
Closes#366 (reported by JohnBaumb).
Socket.IO was initialized with `cors_allowed_origins='*'`, accepting
WebSocket connections from any origin. A malicious site could open a
WS to a user's local SoulSync instance and exfiltrate live progress /
toast / activity events.
This commit:
- Defaults to engineio's same-origin behavior (`cors_allowed_origins=None`),
which automatically honors X-Forwarded-Host so reverse proxies that
send that header (Caddy / Traefik by default, properly-configured
Nginx) work transparently.
- Adds a `security.cors_origins` config setting + Settings → Security
textarea where users behind unusual proxies / Electron wrappers /
cross-origin integrations can whitelist their origin. Accepts comma
or newline separated values; `*` on its own line opts back into the
legacy wildcard with a startup-warning log.
- Logs a clear warning the first time engineio rejects each unique
origin, naming the rejected Origin and request Host and pointing
users to the settings field. Without this, engineio silently 403s
the upgrade and the user just sees a half-broken UI with no clue
why. Threadsafe dedup so a hostile origin can't spam logs.
Logic lives in `core/socketio_cors.py` (resolver, rejection
predictor, dedup logger class, startup-status emitter) — pure
functions, no Flask dependency. `web_server.py` adds 23 lines of
wiring and imports.
Important catch during review: my first pass used `cors_allowed_origins=[]`
as the "secure default." Reading engineio's source revealed `[]` actually
means "DISABLE CORS HANDLING" (engineio/server.py:202: `if cors_allowed_origins != []:`)
— identical security to `'*'`. Fixed to use `None` (engineio's actual
same-origin sentinel) and pinned with a regression test that asserts
the resolver never returns `[]` for any input shape.
Tests:
- tests/test_socketio_cors.py — 45 unit tests covering 19 resolver shape
cases (None, empty, whitespace, comma, newline, garbage types, lists),
the `[]`-must-never-be-returned security regression, 12 rejection
prediction cases, X-Forwarded-Host handling, dedup logger behavior,
threadsafe race (8 threads × 50 hammers → exactly 1 warning), and
startup-status emitter outputs.
Frontend:
- Settings → Security gains an "Allowed WebSocket Origins" textarea
with help text explaining same-origin default + when to add a domain
+ the `*` opt-out.
- helper.js — new '2.4.1' WHATS_NEW block (hidden until version bump)
with a chill-voice entry describing the change.
Conftest.py left at `'*'` — test environment, no security concern.
598 tests pass.
Five issues kettui flagged on PR #377:
- Worker race (reorganize_queue.py): _next_queued() picked an item and
released the lock, then re-acquired to flip status='running'. A
cancel() landing in that window marked the item cancelled but the
worker still ran it. Replaced with _claim_next_or_wait() that picks
AND flips under one lock acquisition.
- Wakeup race (reorganize_queue.py): _wakeup.clear() after the empty
check could lose an enqueue's _wakeup.set(), parking a freshly-queued
album for up to 60 seconds. Replaced Lock + Event with a single
threading.Condition; cond.wait() releases and re-acquires atomically
on notify.
- Bulk dedupe (reorganize_queue.py:enqueue_many): looped single-item
enqueue, so a duplicate album_id later in the same batch could slip
through if the worker finished the first copy before the loop
reached the second. Now holds the lock for the whole batch and tracks
a per-batch seen set, so intra-batch duplicates dedupe against each
other and not just pre-existing items.
- Preview button stuck disabled (library.js:loadReorganizePreview):
early returns and thrown errors skipped the re-enable line. Moved
state into a canApply flag committed in finally, so any exit path
lands the button correctly.
- DB helpers swallowing failures (music_database.py): get_album_display_meta
and get_artist_albums_for_reorganize used to catch every Exception
and return None / [], so a real DB outage masqueraded as "album not
found" / "no albums". Now lets exceptions bubble; the route layer
already wraps them as 500.
Tests:
- test_cancel_and_run_are_mutually_exclusive — hammers enqueue+cancel
pairs and asserts the invariant that no successfully-cancelled item
ever ran (catches regressions to the atomic pick).
- test_enqueue_many_dedupes_batch_internal_duplicates — pins the
intra-batch dedupe.
- test_get_album_display_meta_propagates_db_errors and
test_get_artist_albums_for_reorganize_propagates_db_errors — pin
the bubble-up behavior.
Changelog updated in helper.js and version modal.
Replaces the single-slot "one reorganize at a time, return 409 on collision"
model with a per-user FIFO queue. Buttons stay clickable, "Reorganize All"
is one backend call instead of an N-call JS loop, and a status panel mounted
at the top of the artist actions bar shows live progress (active item,
queued count, recent completions) with per-item cancel buttons.
Backend
- core/reorganize_queue.py: singleton queue + worker thread, dedupe-on-
enqueue, cancel rules (queued cancellable, running not), enqueue_many
for bulk operations, progress fan-out via update_active_progress
- core/reorganize_runner.py: factory builds the worker's runner closure
with injected dependencies. Reads config per-call so changing the
download path in Settings takes effect on the next reorganize without
a server restart
- database/music_database.py: get_album_display_meta and
get_artist_albums_for_reorganize — moves the SQL out of route handlers
- web_server.py: thin enqueue/snapshot/cancel/clear endpoints, runner
registration at module load. Old _reorganize_state globals + status
endpoint deleted. Static-asset cache buster (?v=<server-start>)
added so JS/CSS updates ship live without users clearing cache
Frontend
- webui/static/library.js: status panel mount, polling (1.5s when
active, 8s when idle), expand/collapse, per-item cancel, debounced
enhanced-view reload (one reload per artist batch instead of N).
Per-album reorganize button paints with queued/running indicator
and short-circuits to a toast when the album is already in queue
- webui/static/style.css: panel + button styling matching the existing
glass-UI accents
- webui/static/helper.js + version modal: WHATS_NEW entry
Tests (22 new)
- tests/test_reorganize_queue.py (19 tests): FIFO order, dedupe,
per-item source, cancel rules, continue-on-failure, snapshot
shape, progress propagation, bulk enqueue
- tests/test_reorganize_runner.py (4 tests): per-call config reads,
setup-failure summary, dependency injection, progress fan-out
- tests/test_reorganize_db_methods.py (7 tests): SQL JOIN behavior,
ordering, fallback for blank strings, artist isolation
Full suite 549 passed in 27s.
Four changes addressing kettui's PR #377 review comments:
1. **`_finalize_track` no longer over-counts on DB failure (🔴 bug).**
The function previously bailed on DB-update failure but
`_process_one_track` still incremented `summary['moved']`
unconditionally — overstating how many tracks the UI knows are
at their new locations. Fixed by:
- `_finalize_track` now returns ``bool`` (True only when DB row
was updated AND original was dealt with)
- Caller checks the return; on False, records as a failed track
with a clear message ("Track landed at new location but DB
update failed — file is at both old and new paths until library
scan re-indexes")
- Existing `test_db_update_failure_leaves_original_in_place` now
also asserts `moved == 0`, `failed == 1`, and that the error
message names the cause
2. **`executeReorganize` toast no longer says "undefined tracks" (🐛
bug).** `/reorganize` doesn't return `result.total` anymore (the
track count is determined server-side after planning), so the
"Reorganizing undefined tracks..." string was meaningless. Now uses
`result.message` from the backend instead.
3. **`_pollReorganizeStatus` distinguishes completed from skipped
(🟡 risk).** Backend now propagates the orchestrator's status
(`completed` / `no_source_id` / `no_album` / `no_tracks` /
`setup_failed` / `error`) into `_reorganize_state['result_status']`
so the frontend can warn appropriately. Two new helpers:
- `_classifyReorganizeOutcome(state)` — returns 'success' only
when `result_status === 'completed'` AND `failed === 0`;
'warning' otherwise
- `_formatReorganizeResultMessage(state)` — returns a message
specific to the outcome ("Reorganize skipped — album has no
metadata source ID. Run enrichment first." for `no_source_id`,
etc.)
Zero-failure non-completed runs now show as warnings instead of
green checkmarks.
4. **Bulk mode no longer counts skipped albums as succeeded (🟡
risk).** `_executeReorganizeAll`'s loop was treating any HTTP
200 response as success, ignoring the orchestrator's actual
outcome for that album. Fixed by:
- `_waitForReorganizeComplete()` now resolves with the final
state object (was: void)
- Loop checks `finalState.result_status === 'completed'` AND
`finalState.failed === 0` before counting `succeeded++`;
otherwise increments `skipped` (with a per-album warning
toast) or `failed` accordingly
- Final summary toast now reads
"Reorganized N of M albums, K skipped, J failed" and only
shows green when nothing was skipped or failed
All four addressed in a single commit because they form one
coherent UX-correctness fix — the bug bug (#1) and the count-
overstatement bug (#4) both made the user see "everything succeeded"
when reality was different. Together they make the UI honestly
reflect what actually happened.
Files:
- core/library_reorganize.py — `_finalize_track` returns bool,
`_process_one_track` reads it
- web_server.py — `_reorganize_state['result_status']` populated
from orchestrator's summary on success and on exception
- webui/static/library.js — `_classifyReorganizeOutcome` /
`_formatReorganizeResultMessage` helpers, single-album +
bulk-mode flows both consume them
- tests/test_library_reorganize_orchestrator.py — strengthened
the existing DB-failure test to assert moved/failed counts
Credit: kettui — four PR #377 review comments named all of these
precisely with line numbers and severity.
Reported by kettui on PR #374 review: the inline filter that backed
`set_album_api_track_count` only counted rows where `type_ == 'track'`,
but `discogs_client.get_album_tracks` itself accepts both `'track'`
AND empty `type_` as real songs (line 660: `type_ in ('track', '')`).
Releases where Discogs returns some real tracks with an empty `type_`
field would be undercounted, which would silently disagree with the
repair job's fallback `_get_expected_total` path (which calls into
`get_album_tracks_for_source` and therefore uses the client's count).
Extracted the filter into `count_discogs_real_tracks(tracklist)` —
single source of truth for the rule, testable in isolation, and the
worker call site is now a one-liner that names what it's doing. Also
defensive about the input shape: `type_ == None`, missing field, and
empty/None tracklist all handled cleanly.
10 tests pin the behavior:
- empty/missing/None type_ all count as a real track (the kettui case)
- 'heading', 'index', 'sub_track' excluded
- unknown future type strings excluded conservatively
- realistic multi-disc tracklist with mixed shapes counts correctly
- empty/None input returns 0 without raising
Credit: kettui — the PR #374 review comment that flagged this.
Reported on Discord by winecountrygames. The library "Reorganize" tool
had several layered bugs that all traced to the same root cause: the
endpoint reinvented every wheel post-processing already turns — its own
template engine, its own disc-number resolution from file tags, its own
sidecar sweep, its own collision detection — and each had drifted from
the canonical path used by fresh downloads. Reported symptoms:
- 3-disc Aerosmith deluxe collapsed to a flat single-disc layout
- Half the tracks on other albums silently skipped, no error / no count
- Re-runs left empty leftover album folders cluttering the artist dir
Architecture: stop reinventing wheels. Route reorganize through exactly
the same pipeline downloads use. Per-album:
1. Fetch the canonical tracklist from a metadata source (Spotify /
iTunes / Deezer / Discogs / Hydrabase) using the album's stored
source IDs. New `core/library_reorganize.py::plan_album_reorganize`
does this — primary-source-first, fall through priority chain
unless the user picked a specific source in the modal (strict mode).
2. For each local track, find the matching API entry via a scored
candidate matcher. Score components: exact-title (100),
substring-with-length-ratio (40-90), track-number agreement (20).
Hard reject when the two titles have different version
differentiators (Remix vs no-remix means different recordings,
not annotation drift). Below threshold = unmatched, surfaced as
"not in source's tracklist, left in place" rather than silently
mis-routing.
3. Copy the file to a per-album staging directory, build the same
context dict the import flow builds (`spotify_album` /
`track_info` / etc. with `is_album_download=True` so the path
builder enters ALBUM mode, not SINGLE mode), call
`_post_process_matched_download(...)` — same function fresh
downloads use. Post-process handles tagging, multi-disc subfolder
decisions, sidecar regeneration, AcoustID verification.
4. Read `context['_final_processed_path']` to learn where it landed.
Update `tracks.file_path` in the DB BEFORE removing the original
(DB-update failure leaves the file at both locations, recoverable
via library scan; the reverse would orphan the row). Delete
per-track sidecars (post-process recreates them at the new
destination).
3 concurrent workers per album via ThreadPoolExecutor, matching the
download path's per-batch worker count. State mutations all guarded by
a single lock; staging filenames carry a UUID prefix so concurrent
copies of identically-named source files don't overwrite each other.
Source picker in the modal lets the user choose which source to read
the tracklist from. Two endpoints feed it:
- `/api/library/album/<id>/reorganize/sources` — sources for THIS
album that are both authed AND have a stored ID. For the per-
album modal.
- `/api/library/reorganize/sources` — all authed sources globally.
For the bulk "Reorganize All" modal where per-album ID coverage
varies.
When the user picks a specific source, the orchestrator runs in
`strict_source=True` mode (no fallback chain) — picking Spotify means
"use Spotify or fail", not "use Spotify and silently fall back."
Preview endpoint shares the same planning logic as apply via
`preview_album_reorganize` — the destination path comes from the same
`_build_final_path_for_track` post-process uses, so what you see in
the preview is exactly what you get on apply.
Empty destination folders (from earlier failed runs OR from the
current run when post-process creates a dir then fails AcoustID)
get cleaned up after each successful run: walk up to the artist
folder from any successful destination, prune empty album-sibling
folders one level deep. Bounded scope = won't touch unrelated user
dirs.
Web_server.py shrinks by ~450 net lines. The endpoint handler is now
a thin wrapper that builds injected callables (path resolver, post-
process function, DB updater, empty-dir cleaner), spawns a thread
that calls `reorganize_album()`, and returns. All actual logic lives
in `core/library_reorganize.py` where it's unit-testable without
spinning up Flask.
Frontend cleanup: the per-call template input in both reorganize
modals (per-album and bulk) was redundant — the backend always uses
the configured global download template. Removed the input and the
variables-grid reference UI it was for.
39 new unit tests pin every contract:
- source resolution (no_source_id when album has none, fallthrough
chain when primary returns nothing, strict mode bypasses fallback)
- matcher scoring (exact / substring / multi-disc disambiguation /
smart-quote tolerance / dash-vs-parens / bonus-track substring /
Remix-vs-original differentiator rejection / "Real" doesn't false-
match "Real Real Real" / track-number-only no longer fires)
- file safety (DB-update failure leaves original in place, post-
process failure leaves original in place, post-process exception
caught and original preserved, success removes original AND
updates DB in the right order)
- sidecar handling (per-track .lrc/.nfo deleted on success, kept on
failure; album-level cover.jpg/folder.jpg cleaned only when
directory has no remaining audio)
- staging cleanup (recreated between tracks because post-process
nukes it, dir cleaned up on success AND on failure)
- destination-dir prune (empty siblings removed, real album with
files preserved, no recursive sweep)
- source picker (only authed-with-stored-ID sources for per-album,
all authed sources for bulk; strict mode doesn't fall back)
- concurrency (3 workers in flight, state stays consistent under
races, stop_check cuts off pending tasks)
- preview parity (preview produces same destination as apply for
multi-disc; ALBUM mode not SINGLE mode; unmatched/no-path tracks
surfaced with reasons)
Limitations (deliberate punts, NOT in this PR):
- Renamed local titles on multi-disc albums where track_number
also disagrees: matcher returns nothing (track is "not in
source"). Fixable by using duration_ms as a tertiary signal.
- Per-track in-modal source switching with per-album track-count
hints (would need a second API call before opening the modal).
- UI status panel on the artist page during a run — currently
just toasts. Documented as a follow-up PR.
Files:
- core/library_reorganize.py — new module: plan_album_reorganize,
preview_album_reorganize, reorganize_album, available_sources_for_album,
authed_sources, _score_candidate, helpers for staging/post-
processing/finalizing, sidecar + dest-dir cleanup
- core/metadata_service.py — no changes; reused get_album_for_source,
get_album_tracks_for_source, get_source_priority,
get_client_for_source
- web_server.py — three endpoints (preview / apply / sources GETs)
are thin wrappers; -450 net lines
- tests/test_library_reorganize_orchestrator.py — 39 tests covering
every contract above
- webui/static/library.js — source picker UI in both modals; dead
template input + variables-grid removed
- webui/static/style.css — dropdown option styling fix (white-on-
white was unreadable)
Reported on Discord by winecountrygames — his bug report named the
trigger button (Enhanced view → Reorganize All) and both symptoms
(multi-disc collapse, half-album skip), which let the diagnosis go
straight to the architectural problem.
Reported on Discord by winecountrygames — Spotify auth granted, then
re-banned for 4 hours within ~30 seconds, repeatedly. Trace from his
captured log:
< 12:05 [pre-log] Spotify ban active when log starts
15:21:27 First ban EXPIRED → 5-minute post-ban cooldown begins
15:26:27 Cooldown ends, spotify_client.is_authenticated() probe
allowed again → client initialized
15:26:59 First Spotify API call after cooldown — get_artist_albums
for an artist whose discography a background worker was
enriching — gets 429 immediately with no Retry-After
header → new ban activated for 14400s (4 hours)
Root cause: `_POST_BAN_COOLDOWN = 300` (5 minutes) is shorter than
Spotify's actual server-side memory of the previous offense. The
cooldown exists specifically to prevent the "ban expires → we probe →
re-ban" cycle (`spotify_client.py:65-68` documents that intent
explicitly), but the value was wrong: Spotify's server still
considered this user banned 5 minutes after our local ban window
ended, so the very first call after cooldown got slapped.
The 4-hour re-ban itself is correct behavior — `_BASE_MAX_RETRIES_BAN`
fires when spotipy reports "max retries", which means the client
exhausted its internal retry budget on 429s before raising. That's a
severe-ban signal and a long default is the right response.
Fix: bump `_POST_BAN_COOLDOWN` to 1800 seconds (30 min). This is the
smallest change that addresses the immediate "re-probe → re-ban" loop
in the report. 30 minutes is an empirical floor — long enough for
Spotify to actually clear its server-side memory in the cases we've
observed, short enough not to keep functional users locked out beyond
necessary. Can be revisited if reports persist.
What this PR does NOT fix (important context for the same user):
This bump only helps the "ban expires → we re-probe → re-ban" loop.
It does NOT help winecountrygames's other symptom — Spotify being
banned within 30 seconds of his FIRST EVER authorization (no prior
ban). That's a separate failure mode: on first auth, enrichment
workers immediately fan out across the user's library (250 artists
in his case), hammering Spotify endpoints with bulk get_artist_albums
calls before any rate-limit feedback can land. Spotify's hidden
per-endpoint daily quotas — which BoulderBadgeDad has empirically
documented but the global rate limiter doesn't see — flag the burst
and impose a multi-hour cooldown that LOOKS like a bot-detection ban
to us. A proper fix needs a fresh-auth ramp-up: start with very low
Spotify QPS for the first N minutes, scale up only if no rate-limit
feedback arrives. That's a separate PR.
Documented as additional follow-ups (NOT in this change):
- Adaptive cooldown that scales with the size of the previous ban —
a 4-hour MAX_RETRIES ban probably warrants a 1-hour cooldown,
while a 60-second Retry-After-honored ban can resume in 5 minutes.
The system already distinguishes these in `_set_global_rate_limit`,
it just doesn't propagate the distinction to cooldown duration.
- Probe-with-light-call pattern — make the first post-cooldown call
a single inexpensive endpoint (`current_user`) rather than
allowing a background worker's heavy `get_artist_albums` to be
the canary. Failed probe extends cooldown silently instead of
triggering a fresh 4-hour ban.
- Fresh-auth ramp-up (per the limitation above).
Files:
- core/spotify_client.py — `_POST_BAN_COOLDOWN` 300 → 1800. Comment
expanded to cite the report so the value isn't bumped back without
context.
- webui/static/helper.js — WHATS_NEW entry under 2.40 explaining
the change for affected users.
No tests added — the cooldown logic itself is unchanged, only the
constant. Tests asserting on a constant value are theater.
Reported on Discord by winecountrygames — his captured log made the
"ban-expires-to-re-ban" timing chain unambiguous.
The "Link Tidal Account" device-flow UI displayed a verification URL
like `link.tidal.com/XBXYT` that, when clicked, navigated back to the
SoulSync origin (e.g. `http://localhost:8889/link.tidal.com/XBXYT`)
instead of to Tidal's activation page.
Root cause: tidalapi returns `login.verification_uri_complete` as a
schemeless string. settings.js drops it straight into `<a href>`, and
browsers treat schemeless hrefs as same-origin relative URLs.
Normalize the URI in `start_device_auth` — if it doesn't already
start with `http://` or `https://`, prepend `https://`. Same
treatment for the `link.tidal.com/{user_code}` fallback so the
defensive path stays well-formed too.
Reported on Discord by Netti93: with Tidal configured for "HiRes only"
and "Allow Quality Fallback" disabled, tracks were still downloading
successfully — as m4a 320kbps files. Some "successful" downloads were
less than half the file size of the same track pulled via Tidarr/tiddl
from the same Tidal account.
Root cause: Tidal's API silently degrades to the best quality your
account + the track + your region permits. Setting
`session.audio_quality = Quality.hi_res_lossless` and calling
`track.get_stream()` on a track that's only available in AAC returns
an AAC stream with no error. The downloader wrote the m4a file to
disk, the ~7MB size sailed past the 100KB stub threshold, and the
download reported success.
The pre-existing "verify quality wasn't silently downgraded" block
only LOGGED a warning when this happened; it did not fail the tier.
Two knock-on effects:
- Users with "HiRes only, no fallback" got m4a files anyway, which
defeats the setting entirely.
- The worker-level fallback chain (hires → lossless → high → low)
couldn't advance past the first tier, because every tier
"succeeded" at whatever Tidal happened to serve.
Fix: after `track.get_stream()`, compare `stream.audio_quality`
against the tier we asked for using a rank-based ordering:
LOW < HIGH < LOSSLESS < HI_RES < HI_RES_LOSSLESS
- Same tier or higher → accept (so the occasional Tidal upgrade
doesn't get rejected just because it's not an exact match).
- Lower tier → reject THIS tier. The loop `continue`s and the next
fallback tier is tried, or the whole download fails honestly
when the user has fallback disabled. The existing final-error
log already has a hint directing users to enable fallback if
they want automatic Lossless substitution.
- Unrecognized `audioQuality` value (e.g. a new Tidal tier we
haven't mapped) → reject conservatively, so the next fallback
tier gets a chance and the diagnostic log names the unknown
value.
Why the rank-based approach instead of strict equality:
Tidal's API doesn't technically promise an exact-tier match on
serving; on tracks that are flagged in its catalog as a higher
tier, it can serve higher than the session setting. Rejecting
higher-than-asked quality would be user-hostile. And the `HI_RES`
(legacy MQA) value — not in tidalapi's modern `Quality` enum but
possibly still present on old catalog entries — needs to rank
below `HI_RES_LOSSLESS`: users asking for true lossless HiRes
should reject MQA since MQA is a lossy format.
tidalapi's `Quality` enum is a `str` subclass whose VALUES (not
member names) match what the Tidal API returns in the
`audioQuality` field (e.g. `Quality.hi_res_lossless.value ==
'HI_RES_LOSSLESS'`, `Quality.low_320k.value == 'HIGH'`). Both
sides of the comparison are coerced to `str` before use, so the
check is robust to whichever tidalapi version exposes the served
quality as an enum or a plain string.
The check is extracted as `_verify_stream_tier(stream, q_info,
q_key) -> (ok, reason)` at module scope — a pure function with no
I/O, unit-tested independently. Ten tests: match, three upgrade
cases (LOSSLESS → HI_RES_LOSSLESS, LOSSLESS → HI_RES, LOW → any
higher), three downgrade cases (the reported HiRes → AAC, HiRes
Lossless → MQA HiRes, Lossless → AAC), one unrecognized-tier case,
and two defensive paths for older tidalapi builds without
`audio_quality` on the stream object and for QUALITY_MAP entries
that lack `tidal_quality` (e.g. tidalapi wasn't importable at
module load). Test stub updated to use uppercase `Quality` values
matching real tidalapi so case-sensitivity regressions get caught.
Also removed the old codec-string-based warning block — the new
tier check is strictly stronger, and keeping the warning around
would just be dead code waiting to drift out of sync.
Deliberately NOT tackling in this PR (documented as follow-ups):
- Bit-depth verification of HiRes FLAC files via mutagen. The
`stream.audio_quality` tier check catches the main "HiRes
requested, got AAC" case; bit-depth would only matter if Tidal
labeled a stream HI_RES_LOSSLESS but served a 16-bit FLAC
(`Stream.bit_depth` isn't reliable for this — tidalapi defaults
missing `bitDepth` fields to 16, so a trust-the-stream check
would spuriously reject valid HiRes whenever Tidal omits the
field). A proper fix runs mutagen post-download to inspect the
actual file, then decides whether to delete + retry the next
tier — a whole new failure mode with design trade-offs that
deserve their own PR. The support logs don't show this
happening.
- The "manual remap still says Not Found" symptom. Might be
downstream of this same bug (silent-AAC "success" hitting a
later rejection), might be a separate task-state issue. Not
guessing without logs from the retry path.
- Quality-aware stub threshold. 100KB is a reasonable floor for
real stub/preview detection and there's no evidence the
universal threshold is misfiring in the wild.
Field-verified status: desk-verified via unit tests and empirical
checks against a live tidalapi import (confirming the `Quality`
enum's str-subclass behavior). Not yet smoke-tested end-to-end
against a real Tidal account with a HiRes-only-no-fallback
setting — Netti93 or anyone else with that config should notice
either the fix working (non-HiRes tracks fail honestly with a
clear log line) or any regression before wider release.
Files:
- core/tidal_download_client.py — new `_verify_stream_tier` helper
and `_QUALITY_RANK` table at module scope, called in the
download loop after the stream is fetched and before any
bandwidth is spent. Removed the old inline codec-based warning
since the new check supersedes it.
- tests/test_tidal_stream_tier_verification.py — ten tests covering
match / upgrade / downgrade / unknown / defensive paths.
- tests/test_tidal_search_shortening.py — fake `Quality` values
brought in line with tidalapi's real values so both files share
a consistent stub regardless of pytest collection order.
- webui/static/helper.js — WHATS_NEW entry under 2.40 describing
the rank-based tier comparison.
Reported on Discord by Netti93 — the "same account works via
Tidarr" comparison narrowed the cause to SoulSync's download path
rather than an account/region issue.
Reported by sassmastawillis: the Album Completeness maintenance job
scans 3127 albums in 0.1 seconds and reports 0 findings — for every
user, regardless of whether their library is actually complete.
Restoring an older DB surfaced 7 correct findings, so the code logic
works; the DB state is what's making everything look complete.
Root cause: `albums.track_count` is only ever written by server-sync
paths — Plex's `leafCount`/`childCount` and SoulSync standalone's
`len(tracks)`. It's the OBSERVED count of tracks SoulSync has indexed,
which is always exactly what `COUNT(tracks)` returns for that album.
The completeness job treated it as the EXPECTED total and compared it
against the observed count. They're equal by construction, so
`actual >= expected` is always true: skip, 0.1s scan, 0 findings.
Fix: new `api_track_count INTEGER` column on `albums`, written only by
metadata-source code paths. Populated in two places so the scan is
fast and the fallback is robust.
1. Enrichment workers — shared helper `set_album_api_track_count`
in `core/worker_utils.py`. Called by each worker's existing
`_update_album` method alongside its other album-column UPDATEs:
- spotify_worker: `album_obj.total_tracks` from the Spotify Album
dataclass (already in hand, zero new API calls)
- itunes_worker: same, from the iTunes Album dataclass
- deezer_worker: `nb_tracks` from full_data, falling back to
search_data when the full lookup didn't run
- discogs_worker: count of tracklist rows where `type_=='track'`
(Discogs tracklists interleave heading and index rows that
shouldn't count as songs)
Helper skips the write on zero/None/negative/non-numeric inputs
so a source lacking track info can't clobber a good value a
different source already wrote. Caller owns the transaction —
helper just queues an UPDATE on the caller's cursor without
committing, so it batches cleanly with each worker's existing
multi-UPDATE pattern.
Hydrabase worker deliberately not touched — it's a P2P mirror
that doesn't write album metadata to the local DB. Hydrabase-
primary users hit the fallback path below.
2. Album Completeness repair job — new `al.api_track_count` column
in the SELECT, read first in the scan loop. On miss (album never
enriched, or enrichment workers haven't run yet on a fresh
install), falls through to the existing `_get_expected_total()`
API lookup and persists the result via the same shared helper
(wrapped in connection/commit management since the repair job
runs outside a worker's batched transaction).
Also removed `al.track_count` from the scan's SELECT — now unused
since the observed count was the whole source of this bug, and
leaving a dead SELECT would invite a future engineer to re-introduce
the same comparison.
Help text on the job card was reworded so it honestly describes
current behavior ("counts cached during normal enrichment are used
when available; otherwise the job queries a metadata source
directly") rather than the old "active provider first, then others
as fallback" phrasing, which doesn't match how the cache actually
fills — any enrichment worker that runs can populate it, and the
last writer wins. Document-only follow-up if this edge case ever
bites in practice: add a `api_track_count_source` column so the
scan can prefer the configured primary source's count over others
(e.g. deluxe vs. standard edition mismatches). Not worth the
complexity today.
For existing users, the first completeness scan after upgrade is
fast to the extent their library is already enriched: the workers
already ran and populated `api_track_count` on their normal schedule.
For brand-new installs, the scan's fallback path handles the cold
start — slower, but correct, and subsequent scans are fast.
Does NOT affect:
- Download / post-processing / wishlist / sync code paths — none
of them read `track_count` for completeness semantics.
- Plex / Jellyfin / Navidrome / standalone sync — still write
`track_count` exactly as before; `api_track_count` is a separate
column they never touch.
- Other repair jobs.
- Any UI path — same finding schema, just correct counts now.
Files:
- database/music_database.py — idempotent migration adding
`api_track_count INTEGER DEFAULT NULL` to the existing album-column
check block.
- core/worker_utils.py — new `set_album_api_track_count` helper with
the documented skip-on-bad-input contract.
- core/spotify_worker.py, itunes_worker.py, deezer_worker.py,
discogs_worker.py — one-liner call from each `_update_album`.
- core/repair_jobs/album_completeness.py — scan uses the cache;
fallback path persists API-lookup results via the shared helper;
help text updated to match actual behavior.
- tests/test_worker_utils_album_track_count.py — 9 tests covering
the helper's write/skip contract + no-commit invariant.
- tests/test_album_completeness_job.py — 2 tests for the repair
job's fallback-path wrapper.
- webui/static/helper.js — WHATS_NEW entry.
Credit: sassmastawillis spotted the bug; the "restored older DB
finds 7 albums" signal pinpointed DB state over code logic and
made the diagnosis tractable.
Typing "michael jackson" returned 7 identical-looking cards because
MusicBrainz has many different PEOPLE sharing a canonical name — the
King of Pop plus a NZ poet, a photographer, a mashup artist, a
didgeridoo player, and more, all scoring 80+ on exact-name match.
All 7 passed the score filter. All 7 rendered with the same
fallback image because iTunes/Deezer only know the famous one.
Fix dedupes by normalized (lowercase, whitespace-trimmed) name before
building Artist dataclasses. Keeps the highest-scoring entry per name,
so the King of Pop (score 100) wins over the others (all score 80-81).
Artists with genuinely different names stay separate — a search for
"the beatles" still surfaces tribute bands if they're above threshold.
Implementation note: fetch `max(limit*3, 10)` from MB instead of
`limit` directly, so the dedup pool is large enough to still return
`limit` distinct artists after collapsing duplicates. Previously the
raw fetch was capped at the caller's limit, which would have left
fewer-than-requested results after dedup for common names.
3 new tests (49 total):
- Dedupe collapses 5 same-named entries to 1 (keeps highest score).
- Dedup key is case-insensitive and whitespace-normalized.
- Dedup preserves distinct names ("The Beatles" vs "The Beatles Revival"
stay separate).
Live-verified: "michael jackson" now returns 1 card, "kendrick lamar"
returns 1 card.
Credit: kettui spotted duplicate Michael Jackson cards in the search UI.
Three bugs from kettui's follow-up review pass on the MusicBrainz
search PR, all fixed in one commit because they share UI context.
1. Missing artist images on MB artist results
MusicBrainz doesn't store artist images directly. My earlier commit
returned `image_url=None` on every artist result and trusted the
frontend's lazy-loader — but the lazy-loader's `/api/artist/<id>/image?
source=musicbrainz` endpoint had no handler for MusicBrainz, so it
silently returned None and the emoji placeholder stayed.
Fix plumbs the artist name through:
- `renderCompactSection` stashes `data-artist-name` on artist cards.
- `search.js` and `downloads.js` lazy-loaders pass `name=<artist>` as a
query param.
- `/api/artist/<id>/image` accepts an optional `name` param.
- `metadata_service.get_artist_image_url` has a new `musicbrainz`
branch: since MB has no artist art, it searches fallback sources
(iTunes/Deezer by configured priority) for the artist name and
returns the first image found.
Verified live — Metallica/Kendrick Lamar/Daft Punk all resolve to
Deezer artist images via the name lookup.
2. total_tracks off-by-one on tracks with a release
`_recording_to_track` initialized `total_tracks = 1` and then summed
media track-counts on top. For an 11-track album, it reported 12. An
adapter-level regression introduced when the recording-projection
helper was extracted during the main MB refactor.
Fix: initialize at 0, sum normally. Standalone recordings with no
release (can happen for uncredited remixes etc.) still report 1 via
an explicit fallback — so the existing "single track" case isn't
broken.
3. "Artist Album Title" queries buried specific albums in the
discography list
Bare-name queries like "The Beatles Abbey Road" used to resolve "The
Beatles" as the artist and then browse their full discography — Abbey
Road was buried alphabetically among 200+ releases instead of being
the top result.
Fix adds a title-hint extractor. When the query starts with the
resolved artist name followed by more words, the trailing portion is
treated as a title hint. Browse results are filtered to those whose
release-group title contains the hint. If the filter matches nothing,
falls back to text-search with the hint as the title (the "keep the
old split-by-whitespace fallback" path kettui called for). If text-
search also misses, shows the full discography rather than nothing.
10 new tests in tests/test_musicbrainz_search.py (46 total):
- Title-hint extractor: basic match, case-insensitive, whitespace
tolerance, bare-artist-no-hint, artist-not-prefix-no-hint, word-
boundary required (no false splits on "Metallicasomething").
- Browse filtering by title hint.
- Text-search fallback when the title hint matches nothing in browse.
- Bare-artist queries return the full discography unfiltered.
- total_tracks for single-release, multi-disc, and no-release cases.
Clicking a MusicBrainz album returned 404 because the browse-based
search path now stores release-GROUP MBIDs in Album.id, but `get_album`
still hit `/ws/2/release/<mbid>` directly. Release-group MBIDs don't
resolve as release MBIDs — MB 404s. User log:
GET /api/spotify/album/b88655ba...?source=musicbrainz → 404
Error fetching release b88655ba...: 404 Client Error
The fix requires a two-step resolution for the new browse path:
1. Look up the release-group with `inc=releases+artist-credits` to get
the list of releases inside (original + reissues + regional + promo
editions). MB release-groups routinely hold 5-20 releases.
2. Pick a representative release: prefer Official status over Promo,
prefer releases with a real tracklist over stubs, then earliest date.
3. Fetch that release's full tracklist via `get_release`.
Two extra seconds at the 1-rps rate limit, but it's on click, not on
search results rendering.
Structure:
- New `MusicBrainzClient.get_release_group(mbid, includes)` method.
- New `_pick_representative_release(releases)` helper encapsulates the
ranking logic.
- Tracklist projection extracted into `_render_release_as_album` so
both paths share the same shape construction.
- `get_album` tries release-group first; falls back to direct release
lookup when the MBID turns out to be a release from the text-search
fallback path.
- Canonical Album.id stays the release-group MBID so a re-fetch with
the same URL hits the same code path idempotently.
3 new tests (now 33 total):
- End-to-end release-group → release resolution with mocked client
- Fallback to direct release lookup when rg lookup misses
- Representative-release picker ranks correctly
Verified against live API with the exact MBID that 404'd for the user
(b88655ba... for DAMN. by Kendrick Lamar): now returns in 1.2s with
the full 14-track listing (BLOOD., DNA., YAH., ELEMENT., FEEL., ...).
Three related fixes to make album/track results look like a real
artist discography instead of a firehose of fan-compiled bootlegs.
1. Drop 'compilation' from the release-group browse primary-type filter.
MB's OR filter (`type=album|ep|single|compilation`) silently breaks
when 'compilation' is included — Metallica drops from 1076 matches
to 82 because `compilation` is a SECONDARY type on MB, not a primary
type. The invalid value corrupts the filter for all types, not just
itself. Now we request `type=album|ep|single` which returns the full
1076; actual compilations (primary=Album + secondary=[Compilation])
are filtered out by the studio-preference logic below.
2. Filter release-groups with non-studio secondary-types
(Live/Compilation/Soundtrack/Remix/Demo/Mixtape/Interview/Audiobook/
Audio drama). For Metallica, the first 100 browse results are 12
studio albums + 83 live bootlegs + 5 compilations — without this
filter the Albums section was dominated by 2019-2021 broadcast
recordings. Falls back to the unfiltered list if filtering leaves
the result set empty (covers live-only niche artists).
3. Sort chronologically ASC by first-release-date. Wikipedia-style
discography ordering — debut album on top, then chronological.
Previous DESC sort put the most recent release on top which, for
prolific artists, meant 2020s material before their classics.
Track side of the same fix:
- Re-orders each recording's `releases` array to put studio releases
first before `_recording_to_track` picks up the first release for
album context. Without this, MB's arbitrary release order often
buried the canonical studio album under random live bootlegs.
- Filters out recordings that only exist on live/compilation release-
groups (keeps the ones with at least one studio release). Falls
back to the full set if the artist has no studio recordings at all.
- Sorts recordings by earliest studio-release year ASC so classic
tracks surface first.
Smoke test against live MB API confirmed:
- Artists: [Metallica score=100]
- Albums: Kill 'Em All (1983) → Ride the Lightning → Master of Puppets
→ ...And Justice for All → Metallica (Black Album) → Load → Reload
→ St. Anger → Death Magnetic → Lulu (2011)
- Tracks: real Metallica recordings (Killing Time, Nothing Else
Matters, Creeping Death, etc.) — a few remastered demos still leak
in where MB metadata quality is thin, but the bulk is correct.
- Total latency: 3.5 seconds.
4 new tests covering the studio filter, live-only fallback, preferred
release ordering, and live-only recording exclusion.
Credit: kettui flagged the poor MB results during PR #371 review.
The previous commit's `browse_artist_recordings` call passed
`inc=releases+artist-credits` — but MusicBrainz's recording browse
endpoint rejects `inc=releases` with HTTP 400. The adapter's error
handler returned an empty list, so the Tracks section stayed empty
even though the fix was supposed to populate it.
Browse without release info is useless for our search UI (tracks
would render with no album), so swap to the fielded Lucene search
`arid:<mbid>` on the `/recording` endpoint. That's the canonical MB
pattern for "find recordings by this artist WITH release context":
- arid: search accepts the artist MBID and returns recordings with
`releases` (release-group, date, media) embedded in each result.
- One API call per lookup, same as browse would have been.
Renamed the method to `search_recordings_by_artist_mbid` so the name
matches its behaviour — it's a search, not a browse. Adapter updated
to call the new name; tests updated to match.
Verified against the live API: Metallica's MBID returns 5 recordings
in ~1.8 seconds (vs the previous 400 error).
Cover Art Archive URLs are deterministic from the MBID: a GET either
307-redirects to the image or returns 404. The previous adapter fired
`requests.head(timeout=3)` per search result to probe for the image
first. 10 results × 3s worst-case = up to 30s of blocking HEAD calls
before a search returned.
The probe was defensive overhead — the frontend already handles 404 via
`<img onerror>` fallback. Building the URL deterministically and letting
the browser load it lazily collapses the tail latency to the real MB API
calls (artist-search + browse = ~3s at the 1-rps rate limit).
Also prefer release-group scope over per-release scope when both are
available — release-group covers every edition of an album, so the hit
rate is noticeably higher than pinning to a specific regional release.
Removes now-unused `self._art_cache` and the `requests` import.
Bare name queries (typing 'metallica') now resolve to an artist MBID via
the fuzzy search added in the previous commit, then BROWSE that artist's
release-groups and recordings instead of text-searching release/recording
titles. That's the only way to fix the core garbage-results issue: MB
indexes release/recording titles, not artist names, so 'recording:metallica'
matches random tracks literally titled 'Metallica' (all scoring 100).
Structure:
- `_split_structured_query` — detects 'Artist - Title' / 'Artist – Title' /
'Artist — Title' shapes. When present, text-search is correct (user
gave an explicit title to match).
- `_resolve_top_artist` — memoized per-instance lookup for the top-scoring
artist MBID. Backend fires artists/albums/tracks searches in parallel
against one shared client instance, and albums+tracks both need the
same artist lookup. Cache + lock means one HTTP call instead of three.
- `_release_group_to_album` / `_recording_to_track` — shared projection
helpers between the browse and text paths so both paths return the
same dataclass shape.
Search flow per kind:
- `search_albums('metallica')` → resolve top artist → browse release-groups
with `type=album|ep|single|compilation` → sort by type priority then
release date desc → Album dataclasses for top N.
- `search_tracks('metallica')` → resolve top artist → browse recordings
with `inc=releases+artist-credits` → dedupe by normalized title (MB
has many live/compilation variants of the same song) → sort by release
date desc → Track dataclasses for top N.
- `search_albums('foo - bar')` → structured query → text-search path
(unchanged behavior, now score-filtered to 80+).
- `search_tracks('foo - bar')` → same.
- Both text-search paths also dedupe through `_search_albums_text` /
`_search_tracks_text` helpers, which apply the 80-score filter that
the artist-first path gets free from the resolver's threshold.
Also dedupes text-path tracks through the new `_recording_to_track`
helper, replacing ~60 lines of inline projection code. Net change is
more lines overall (browse + helpers) but the text paths shrank and
the garbage-results issue is fixed.
Credit: kettui flagged the missing Artists section + unusable track
results during PR #371 review.
`MusicBrainzSearchClient.search_artists` has been a `return []` stub
since the feature landed, with a comment claiming the MB tab 'doesn't
show artists.' That's why kettui saw a missing Artists section on the
search page — not a missing render, a hardcoded empty list.
Re-enable it properly:
- New `strict=False` parameter on `MusicBrainzClient.search_artist`
sends a bare Lucene query instead of `artist:"..."`. MusicBrainz
matches bare queries against alias+artist+sortname indexes together,
which is the right behavior for user-facing fuzzy search (finds
typos, aliases, sortname variants). `strict=True` remains the
default for enrichment/AcoustID callers that want exact matches.
- Adapter filters results to `score >= 80`. MB assigns a 0-100 Lucene
score on every hit; the true artist + close variants score 100,
tribute bands and lookalikes typically land in the 40-65 range.
The cutoff keeps "Metallica" (100) and drops "Black Metallica
Tribute Band" (60) without hand-curated lists.
- Results returned as the same `Artist` dataclass used elsewhere in
the search-tab adapter layer. `popularity` carries the MB score
(0-100) so the frontend can sort/highlight top matches if desired.
Add `browse_artist_release_groups(mbid)` and `browse_artist_recordings(mbid)`
to MusicBrainzClient. These hit `/ws/2/release-group?artist=<mbid>` and
`/ws/2/recording?artist=<mbid>` respectively — the correct MusicBrainz
pattern for "give me everything linked to this artist."
Why this matters: our current search adapter calls text-search
(`release?query=...` / `recording?query=...`) for albums and tracks,
which matches entity titles literally. Typing "metallica" hits unrelated
releases titled "Metallica" and recordings named "Metallica" by obscure
bands — every garbage match scores 100 because they're all exact title
matches on the wrong field.
Browse walks the artist→release-group and artist→recording links
directly. Once we know the artist's MBID (from `search_artist`), browse
returns their actual discography instead of title collisions.
No behavior change yet — search adapter still uses the old path. Follow-
up commit wires the new endpoints in.
Reference: https://musicbrainz.org/doc/MusicBrainz_API — "Browse queries
retrieve entities linked to a known entity" vs search.
MusicBrainz mandates a meaningful User-Agent with contact info, warning
that bare strings can trigger IP blocking under load. Our client was
sending `SoulSync/2.3` with no contact — and the search adapter passed
an app version hard-coded at "2.3" that's now stale (UI is at 2.40).
Fix: default contact to the project URL (`https://github.com/Nezreka/SoulSync`)
when no email is supplied, so every request lands as
`SoulSync/<version> ( https://github.com/Nezreka/SoulSync )`. Drop the
search-adapter version suffix to a generic "2" since the exact UI minor
version would add noise to every MB request without helping operators
track issues.
Reference: https://musicbrainz.org/doc/MusicBrainz_API — "it is
important that your application sets a proper User-Agent string."
JohnBaumb's review: "If we're going to refactor the web_server.py soon,
might as well start moving stuff away from web_server.py in our PRs.
_build_source_only_artist_detail, make it a module, it's perfect."
This continues the pattern the prior commit started with the source-ID
lookup helpers: move the pure data-building logic to a side-effect-free
core module, leave a thin wrapper in web_server.py that bridges the
Flask response and the module-global clients.
**core/artist_source_detail.py** — pure function that takes the artist id,
name, and source plus dependency-injected per-source clients (spotify,
deezer, itunes, discogs) and a Last.fm API key. Returns
(payload_dict, http_status) so it isn't coupled to Flask.
**web_server.py wrapper** — builds the client bag from the module globals
(checks Spotify auth, constructs the Discogs client from the configured
token, reads the Last.fm API key) and wraps the core return in jsonify.
147 lines of logic go away from web_server.py; the 24-line wrapper is
purely glue.
**tests/test_artist_source_detail.py** — 21 focused tests covering the
response envelope, the source-specific ID-field stamping for all six
supported sources, the dedup_variants=False contract (the behaviour
that originally motivated the split of MetadataLookupOptions), per-source
genre/follower extraction with safe handling of missing or throwing
clients, and the Last.fm enrichment branch including the no-key and
error-path cases. Runtime 0.26s.
Cin observed that database/api_call_history.json was occasionally
landing on disk truncated mid-write — `_load()` would log
`History file is not valid JSON, starting fresh` and 24h of metrics
would be lost.
Root cause: `save()` opened the file in 'w' mode (which truncates to
0 bytes immediately) and then streamed JSON via `json.dump`. Any
SIGINT/SIGTERM/crash between truncate and final write left the file
half-formed — exactly Cin's symptom of the JSON cutting off mid-array.
Switch to the standard atomic pattern: write to a sibling .tmp file,
flush + fsync, then `os.replace` (atomic on every platform we run on).
Failed writes also clean up the leftover .tmp file. The canonical
file is now either the previous good copy or the new good copy —
never a partial one.
Cin's review note: typing artist_name as plain `str` forced callers
that didn't have a name to pass `""` as a placeholder, which leaks the
parameter's emptiness contract into every call site and reads badly in
tests. Switching to `Optional[str] = None` lets callers omit it.
The function body's `if artist_name and active_server:` check already
handles None and "" identically, so no body changes were needed. Tests
that previously passed `artist_name=""` drop the argument; one new test
covers the omitted-arg path explicitly.
The web_server.py wrapper takes the same default for symmetry.
Cin pointed out that the prior version of test_artist_source_lookup.py
AST-parsed web_server.py to verify a constant and to string-match a
function's response keys. That was a workaround for the fact that
web_server.py can't be imported at test time (it boots Spotify,
Soulseek, Plex, etc.) — the right answer is to move the logic into a
side-effect-free module so it can be imported and tested directly.
This commit:
- adds core/artist_source_lookup.py containing the SOURCE_ID_FIELD
map, the SOURCE_ONLY_ARTIST_SOURCES set, and find_library_artist_for_source
- replaces the inline definitions in web_server.py with imports +
a thin wrapper that injects the active media server
- rewrites the tests to import from the core module directly:
* mapping correctness is now a plain equality assertion
* lookup behaviour is exercised against a real MusicDatabase
* the AST parse and the string-matching contract test class are
gone
- drops the _build_source_only_artist_detail contract test entirely
(the weakest of the four — it was just string-matching the function
body); when that function moves to core/ it can get a real
behavioural test alongside.
Test runtime drops from ~161s to ~5.8s. All 18 tests pass.
Source artists landing on /artist-detail were rendering an almost-blank
hero — image + name + a tiny Download button — because the backend
response only had {id, name, image_url, server_source: null, genres: []}.
The library.js renderers do their best with what they have, and that
wasn't much.
Backend changes (_build_source_only_artist_detail):
- Set the source-specific ID field (deezer_id / spotify_artist_id /
itunes_artist_id / discogs_id / soul_id / musicbrainz_id) on
artist_info so the corresponding service badge renders on the hero.
- Try the source's own get_artist_info / get_artist for genres +
followers (Spotify always; Deezer/iTunes/Discogs when available).
Spotify also fills image_url if metadata_service.get_artist_image_url
came up empty.
- Last.fm enrichment by artist name — bio + listeners + playcount +
lastfm_url. Mirrors what library artists get from the cached
enrichment workers but on demand for source artists.
- All enrichment lookups are wrapped in try/except so a 500 from any
one source doesn't break the whole response.
Frontend (library.js populateArtistDetailPage):
- Watchlist button now initialises for source artists too. Falls back
to artist.id + artist.name when there's no canonical Spotify
identity (which is the common case for non-library artists).
Discography dedup opt-out:
- Added dedup_variants flag to MetadataLookupOptions (default True so
library artists are unchanged). Source-only path now passes
dedup_variants=False so every "Deluxe Edition" / "Remastered" /
"Anniversary" variant the source returns is shown — matches the
inline /artists page behaviour the user was comparing against.
Result: source artists' hero now shows badges + bio + listeners +
playcount + watchlist button + genres in addition to image and name.
Discography lists every release the source returns, not the deduped
canonical view.
User reported searching "Maduk - Leave A Light On" on Tidal silently
downloaded Tom Walker's completely different song of the same name, then
embedded Maduk's metadata into Tom Walker's audio. Three layers of
defense all failed permissively. Two of them are fixed here; the third
(score formula weights) was left alone since these two together cover it.
Layer 1 fix — candidate artist gate (web_server.py:27782)
Old: `if _best_artist < 0.4 and confidence < 0.85: continue`
New: `if _best_artist < 0.5 and confidence < 0.85: continue`
SequenceMatcher returns exactly 0.400 for "maduk" vs "tom walker"
(5-char vs 10-char strings with coincidental char matches), which
slipped past the strict `< 0.4` check. The word-boundary containment
check earlier in the function already short-circuits legitimate
formatting variations to sim=1.0, so falling to SequenceMatcher means
strings are genuinely different. 0.5 closes the fencepost AND gives
a small safety buffer.
Layer 3 fix — AcoustID verification (acoustid_verification.py:316)
When title matches but artist doesn't AND expected artist isn't found
anywhere in AcoustID's returned recordings:
Old: always SKIP (let file through, assume cover/collab)
New: FAIL if artist_sim < 0.3 (clear mismatch)
SKIP if artist_sim >= 0.3 (ambiguous — cover/collab/formatting)
The 0.3 cutoff catches hard mismatches like Maduk/Tom Walker (sim ~0.2)
while preserving benefit-of-the-doubt for borderline artist formatting
differences. Legitimate covers and collabs where the expected artist
appears anywhere in AcoustID's recordings still PASS via the existing
secondary-match loop above.
Both fixes are defense-in-depth — either alone would have caught this
bug. Together they close the pre-download AND post-download gaps.
All 292 tests pass. Version bumped to 2.39 with changelog entries.
Tidal's search engine chokes on long queries with multiple qualifier
words (remix credits, edit labels, bonus-disc markers). User reported
case: "maduk transformations remixed fire away fred v remix" returns 0,
but shortening to "maduk transformations remixed fire away" works.
Behaviour change:
- On a 0-result search, retry with progressively-shortened variants
(capped at 5 total attempts, 100ms pause between).
- Variants (in priority order):
1. strip trailing "(...)" / "[...]"
2. strip all parentheticals/brackets
3-5. drop last 1 / 2 / 3 tokens
6. keep first half of tokens (rounded up)
- Dedupes so identical variants don't re-query.
Safety — qualifier-aware filter:
- Variant keywords (Live / Remix / Acoustic / Extended / Unplugged /
Instrumental / Karaoke / etc.) are extracted from the original query
using word-boundary match so "edit" doesn't match "edition" and
"mix" doesn't match "remixed".
- If the original query carries any qualifiers, fallback results MUST
contain those qualifiers in their track names — otherwise a shortened
query could silently downgrade "Song (Live)" to the studio "Song".
- Tracks that fail the filter are dropped. If no variant produces
qualifier-matching tracks, returns ([], []) — the same outcome as the
original code, so no regression.
Contract preservation:
- Never raises to caller (outer try/except catches orchestration errors).
- Returns ([], []) on any failure path, same as original.
- Original-query successes take the same code path as before — no
behavioural change for queries that already work.
- Defensive guards for None/empty/non-string query (early return).
Logging:
- Preserves original warning/error/info messages for back-compat log
scraping.
- Adds fallback-success INFO log ("Tidal fallback query succeeded: ...")
so successful retries are visible in production logs.
- Adds qualifier-filter INFO/DEBUG logs with kept/total counts.
- Per-attempt exception logs at DEBUG (not ERROR) to avoid noise when
retries succeed.
- Traceback preserved on final failure.
Tests (16 regression tests in tests/test_tidal_search_shortening.py):
- Skowl's reported query reaches his working variant within the cap.
- Paren/bracket stripping priority.
- Short queries produce no variants.
- All variants unique (dedup guard).
- Progressive token drops present for long queries.
- Qualifier extraction is word-bounded (no "edit" in "edition").
- Qualifier extraction is case-insensitive.
- Track name filter requires ALL qualifiers.
- Empty-qualifier list passes every track (original-query behaviour).
All 292 tests pass.
New smart template variable that emits "CD01" / "CD02" etc. in filenames
on multi-disc albums, and expands to empty string on single-disc albums
so mixed libraries don't end up with "CD01" on every single.
Template behaviour:
- total_discs > 1 -> "CD{disc:02d}" (zero-padded, CD prefix)
- total_discs <= 1 -> empty string
- Both $cdnum and ${cdnum} bracket form supported
- Empty value collapses cleanly via existing double-dash regex plus new
leading-dash cleanup pass
Wiring:
- _apply_path_template in web_server.py (download pipeline)
- _apply_path_template in core/repair_jobs/library_reorganize.py
(Reorganize repair job)
- total_discs added to every album-mode template context:
* download pipeline album branch (uses resolved total_discs even for
single-track downloads from search)
* per-album Reorganize preview + apply endpoints (pre-scan all track
tags once, take max disc_number)
* Library Reorganize repair job (already had album_total_discs map,
just added to context dict)
Leading-dash cleanup added to _get_file_path_from_template (web_server)
and _build_path_from_template (library_reorganize) so templates like
"$cdnum - $track - $title" don't leave "- 05 - Title" on single-disc
albums.
UI:
- Template hint in Settings -> File Organization documents $cdnum
- Template validation variable list includes $cdnum
- Reorganize modal variable reference shows $cdnum with example "CD01"
Verified:
- Multi-disc disc 1 -> "CD01 - 05 - Track"
- Multi-disc disc 2 -> "CD02 - 05 - Track"
- Single-disc -> "05 - Track" (no leading dash)
- Templates without $cdnum behave unchanged
- 276/276 tests pass
Two bugs reported in issue #320:
1. Auto-watchlist scan bypassed Global Override settings.
scan_watchlist_profile applied _apply_global_watchlist_overrides, but
the scheduled auto-scan called scan_watchlist_artists directly —
bypassing the override. Users who unchecked "Albums" or "Live" under
Watchlist → Global Override still saw full albums and live tracks
added during nightly scans (per-artist defaults, which include
everything, won).
Moved override application into scan_watchlist_artists itself so
every entry point respects it. scan_watchlist_profile now forwards
the apply_global_overrides flag through to avoid double-application.
2. is_live_version (watchlist + discography backfill) and
live_commentary_cleaner's content patterns used bare \blive\b, which
matched verb uses like "What We Live For" by American Authors,
"Live Forever" by Oasis, "Live and Let Die" by Wings.
Tightened the live patterns to require clear recording context:
(Live) / [Live Version] / - Live / Live at|from|in|on|version|
session|recording|performance|album|show|tour|concert|edit|cut|take
/ In Concert / On Stage / Unplugged / Concert.
Locked in 11 regression tests covering the reported false positives
(What We Live For, Live Forever, Living on a Prayer, Live and Let Die)
and the reported true positives (Dimension - Live at Big Day Out,
MTV Unplugged, etc.).
Version bumped to 2.37 with changelog entries.
Root-cause fix for "scanning 50 artists" then silence: when the master
repair worker was paused, force-run still kicked off _run_job but the
job's first wait_if_paused() blocked forever because is_paused was tied
to the master-enabled state. Force-run now bypasses master-pause —
scheduled runs still respect it.
Also fixes Fix All on discography findings doing nothing: the backend
bulk_fix_findings query had a fixable_types allowlist that excluded
missing_discography_track (and acoustid_mismatch). Added both.
Backfill job rebuild:
- auto_add_to_wishlist opt-in setting — creates findings AND pushes to
wishlist during the scan
- 3-option fix dialog (Add to Wishlist / Just Clear / Cancel) on single
Fix, Bulk Fix selection, and Fix All (page-level)
- Fix All "Just Clear" path uses the clear endpoint with job_id filter
instead of the generic "may delete files" bulk-fix warning
- Batched in-memory matching using get_candidate_albums_for_artist +
get_candidate_tracks_for_albums (same fast path the Library pages use)
- Rich album context per finding (id, name, album_type, release_date,
images, artists, total_tracks) — flows through the wishlist pipeline
so auto-processor classifies each track into the right cycle
(albums vs singles) and post-processing gets correct folder/tags/art
- Per-artist progress logs [N/50] Scanning ArtistName
- Default interval 24h (was 168h); all release types default on; settings
reordered with _section_* group headers (Core / Release Types /
Content Filters)
Repair settings UI:
- Generic _section_<name> key convention renders as an uppercase group
divider in the settings panel — any job can opt in
- .repair-setting-row gets a dashed bottom border so label↔toggle pairing
is visually clear
- _prettifyRepairSettingKey fixes acronym capitalization (EPs, not Eps)
Version bumped to 2.36 with changelog entries.
Two bugs kept this job from finding anything useful on a typical library.
1. Wrong Deezer column name. The artists table has a deezer_id column
(per music_database.py:1986), but the job looked for deezer_artist_id
in both _scan_artist (line 132) and _get_library_artists (line 345).
For Deezer-primary users, this meant the Deezer ID never made it into
the source_ids map, so get_artist_discography fell back to artist-
name-only search — slower and less accurate than an ID lookup.
2. Spotify-reported EPs were silently excluded. Spotify lumps EPs and
true singles under album_type='single'. The previous
_should_include_release short-circuited on album_type='single' and
returned the include_singles setting (default False), so 4-6 track
EPs on Spotify-primary libraries never survived the filter — even
though include_eps defaulted to True. Only 7+ track full albums
made it through. This is the main reason users felt the job did
nothing.
Fixes:
- Use the correct deezer_id column name in both reference sites.
- Restructure _should_include_release so only 'album', 'ep', and
'compilation' are trusted outright. Anything else (including
'single' and missing type) falls through to a track-count
disambiguation matching the download pipeline's _get_album_type_display:
1-3 tracks = true single, 4-6 = EP, 7+ = album. A Spotify-returned
'single' with 5 tracks now correctly counts as an EP.
Full suite stays at 263 passed. Ruff clean.
PR #340 added ruff to the build-and-test.yml CI gate, which surfaced
286 pre-existing lint errors. Left unfixed, every feature branch push
fails CI. This commit resolves all of them so CI goes green and
contributors can actually land work.
Auto-fixes (248 of 286): removed unused f-string prefixes (F541),
renamed unused loop control variables with underscore prefix (B007),
removed duplicate imports (F811).
Manually fixed 10 latent bugs ruff caught (all wrapped in try/except
today, silently failing):
- music_database.py: _add_discovery_tables() called undefined
conn.commit() — would have crashed the iTunes-support migration
for existing databases. Now uses cursor.connection.commit().
- web_server.py settings GET: referenced undefined download_orchestrator
when it should be soulseek_client. Feature (_source_status on the
settings payload) was silently missing for UI auto-disable logic.
- web_server.py _process_wishlist_automatically: active_server
undefined in track-ownership check. Auto-wishlist was falling
through to the error handler and re-downloading owned tracks.
- web_server.py start_wishlist_missing_downloads: same active_server
bug in the manual wishlist path.
- web_server.py _process_failed_tracks_to_wishlist_exact: emitted
wishlist_item_added automation event with undefined artist_name
and track. Automation event silently never fired correctly.
- web_server.py discovery metadata enrichment: referenced cache
without calling get_metadata_cache() first. Track enrichment from
cached API responses was silently skipped.
- web_server.py Beatport discovery worker: wing-it fallback branch
used undefined successful_discoveries variable. Wing-it counter
never incremented correctly. Now uses state['spotify_matches']
consistently with the rest of the function.
- web_server.py _run_full_missing_tracks_process: stale import json
mid-function shadowed the module-level import, making an earlier
json.dumps() call reference an unbound local (F823).
- web_server.py discovery loop: platform loop variable shadowed
the module-level platform import (F402).
- core/watchlist_scanner.py: 7 lambda captures of loop variables
(B023 classic Python closure-in-loop bug) now bind at creation.
No existing tests had to change. Full suite stays at 263 passed.
- Move /api/artist/<artist_id>/image resolution into core.metadata_service.
- Resolve artist artwork through source priority, with explicit source/plugin overrides preserved.
- Keep Spotify call tracking inside the client layer to avoid double counting.
- Update similar-artist lazy loading to pass source context and add service coverage.
- Relocate the streamed MusicMap similar-artist flow out of web_server.py and into core.metadata_service.
- Match similar artists through the configured source-priority chain instead of assuming Spotify first.
- Add iTunes artwork fallback so streamed artist payloads still carry image_url when search results are sparse.
- Cover the new service behavior with tests.
Artist detail pages ran check_album_exists_with_editions and check_track_exists
per discography item, each firing 5+ title variations times 3 artist variations
of fuzzy LIKE searches plus fallback broad-artist queries. For a 30-album artist
that was ~450 SQL round-trips just to answer "which of these do I own."
Hoist the artist's library albums and tracks into memory once per request via
two new helpers — get_candidate_albums_for_artist and get_candidate_tracks_for_albums —
and thread them through as optional candidate_albums / candidate_tracks params on
check_album_exists_with_editions, check_album_exists_with_completeness,
check_track_exists, check_album_completion, and check_single_completion.
Batched path scores the same _calculate_album_confidence / _calculate_track_confidence
against the in-memory list, preserving Smart Edition Matching and accuracy.
Title-only cross-artist fallback still fires for collaborative-album edge cases.
None on either param preserves legacy per-item SQL behavior for unaffected callers.
Applied to both /api/library/completion-stream (library artist detail page) and
iter_artist_discography_completion_events (Artists search page). Timing logs
added to confirm the pre-fetch cost and loop elapsed time.
On a Kendrick page load, per-album resolution drops from ~8 seconds to under
the 50ms streaming sleep floor. Observed ~100x SQL reduction on the happy path.
- collapse old multi-line debug bursts into single structured rows
- remove leftover DEBUG-style prefixes from message text
- keep the app log readable without losing useful trace detail
If the application was using a non-standard location for app.log, the other logs would still go to the default location. Now everything goes under the same, configured folder
When staging files are organized as Artist/Albums/AlbumFolder or
Artist/AlbumFolder, the auto-import now uses the parent folder name
as the artist instead of trusting embedded file tags.
Uses relative path from staging root to determine folder depth, so
albums directly in staging root don't accidentally pick up container
paths as artist names. Common category subfolder names (Albums,
Singles, EPs, Mixtapes, etc.) are recognized and skipped.
Fixes mixtapes and compilations where file tags have DJ names or
incorrect artists (e.g. files tagged as "Slim" in a 2Pac folder).
New repair job that scans each artist in the library, fetches their
full discography from metadata sources, and creates findings for any
tracks not already owned. Users review findings and click "Add to
Wishlist" to queue missing tracks for download.
Respects content filters (live/remix/acoustic/instrumental/compilation)
and release type filters (album/EP/single). Opt-in, disabled by default,
runs weekly, processes up to 50 artists per run with rate limiting.
Jobs with interval_hours set to 0 caused ZeroDivisionError in
_pick_next_job staleness calculation. Now skips jobs with invalid
(zero or negative) intervals.
The Subsonic getArtist endpoint doesn't support musicFolderId filtering,
so when an artist exists in multiple libraries, all their albums were
imported regardless of which music folder was selected in settings.
Now passes musicFolderId to getArtist (in case Navidrome supports it),
and as a fallback filters albums against a cached set of album IDs
built from getAlbumList2 (which reliably supports musicFolderId).
The set is built once per session and invalidated on folder change.
The high-confidence fingerprint skip (≥0.95) assumed title mismatches
were language/script differences and bypassed verification. But a high
fingerprint score just means AcoustID identified the audio confidently —
not that it matches the requested track. Now requires partial title
(≥0.55) or artist (≥0.60) similarity before skipping, so completely
wrong files (e.g. different song/artist from same remix producer) are
correctly rejected.
- Move album-track resolution into metadata_service
- Use the configured provider order instead of Spotify-first branching
- Switch the frontend to the unified /api/album/<id>/tracks endpoint
- Add tests for source-priority lookup, DB resolution, and formatting
iTunes API can return collection metadata without song tracks for
region-restricted albums. The _lookup fallback only checked if results
was empty, so a collection-only response was accepted and cached as
{'items': []}. All future lookups returned the cached empty result.
Three fixes:
- get_album_tracks now checks for actual song items and tries fallback
storefronts when only collection metadata is returned
- Skip cached results with empty items array (prevents stale cache hits)
- Backend returns descriptive 404 error, frontend surfaces it in toast
Fixed 5 critical gaps in the download orchestrator where lidarr was
missing from client loops: get_all_downloads, get_download_status,
cancel_download fallback, clear_all_completed_downloads, and
cancel_all_downloads. Without these, lidarr downloads were invisible
to the UI, couldn't be cancelled, and accumulated in memory.
Also: error messages now visible in download list (appended to
filename on error state), removed "(Development)" label from UI.
- pass provider-specific artist ids into the source-priority discography lookup
- stop relying on the local library artist id when querying external metadata
- add a regression test for source-specific artist id resolution
- Stop passing in spotify_id as the id in the UI, use the actual db id instead
- Fixes an issue where albums for another artist would end up being returned for the actual searched artist
- Remove the redundant artist_id filtering code
- Fixes an issue where not-currently-owned albums would be filtered out from the results, even if they were successfully fetched from the configured metadata provider
The retag fix for AcoustID mismatches was only updating the DB
record (title, artist_id) without writing corrected tags to the
actual audio file. Users would click Fix, the finding disappeared,
but the file on disk stayed unchanged. Now writes title and artist
tags to the file via Mutagen after the DB update.
Also fixed artist INSERT missing server_source when creating a new
artist during retag — now uses the active media server value.
The listening stats worker ran three N+1 query patterns on every
30-minute poll cycle:
1. _resolve_db_track_id was called once per history event (up to
500 events = 500 SELECTs).
2. _map_play_counts_to_db ran one SELECT per server track ID.
3. _enrich_stats_items ran one SELECT per top_artist, top_album,
and top_track (typically 60 extra queries per rebuild).
All three paths now use batched IN queries with 500-row chunks
(well under SQLite's default variable limit of 999). Case-insensitive
matching and LIMIT 1 semantics are preserved via setdefault() on the
Python-side result dict.
Track resolution uses SQLite row-value IN ((?,?), ...) on
(LOWER(title), LOWER(artist_name)), available in SQLite 3.15+
(bundled with Python 3.13).
Four enrichment workers (Last.fm, MusicBrainz, Tidal, Qobuz) had a
bug where every background loop re-processed the same rows because
the existing-ID short-circuit path never set match_status, and two
workers queried the wrong column when checking for an existing ID.
lastfm_worker._get_existing_id queried a non-existent lastfm_id
column; the real column is lastfm_url. The method now reads
lastfm_url for all three entity types.
musicbrainz_worker._get_existing_id queried musicbrainz_id for all
entity types, but albums use musicbrainz_release_id and tracks use
musicbrainz_recording_id. The method now uses a per-type column map.
All four workers (lastfm, musicbrainz, tidal, qobuz) now write
match_status='matched' when they short-circuit on an already-present
external ID, so these rows are no longer re-selected on the next
worker sweep.
A new migration (_backfill_match_status_for_existing_ids) runs once
on startup to retroactively set match_status='matched' for rows that
already have an external ID but NULL match_status. This covers legacy
data, manual matches, and rows populated from file tags outside the
worker.
MetadataCache.get_search_results previously looped over each cached
entity ID and issued one SELECT per ID, producing N extra queries per
cached search hit. It now resolves all entities in a single batched
IN query (chunked at 500 to stay under the SQLite variable limit),
then reconstructs the result list in the original result_ids order
using an in-memory dict lookup.
- broaden the artist-detail dedup helper to catch trailing parenthetical edition and remaster variants
- keep the legacy hyphenated suffix fallback for older metadata
- add regression coverage for language-specific Edition and remaster cases
- move artist-detail discography resolution onto the shared source-priority metadata service
- keep the variant dedup helper in the UI-facing adapter
- pass the chosen source through completion checks
- add coverage for the new adapter and dedup behavior
Move completion checks into metadata_service and make them follow the configured metadata source priority.
Drop the old test-mode path, remove the web_server wrapper indirection, and keep artist inference on explicit release metadata instead of guessing from a track search.
Add coverage for the source-priority completion behavior and the safer artist-name handling.
When no cached token exists, spotipy's auth probe starts an interactive
OAuth flow that binds 127.0.0.1:<redirect_port> inside the container.
This either steals Flask's port 8008 (crash loop) or binds loopback-only
on 8888 (unreachable from Docker host — 'connection reset by peer').
Now checks for a cached token before probing. If none exists, returns
False immediately so users authenticate via the SoulSync web UI instead.
No behavior change for already-authenticated users.
Fixes#269
New core/genre_filter.py with ~180 curated default genres. When strict
mode is enabled in Settings → Library Preferences → Genre Whitelist,
only whitelisted genres pass through during enrichment. Junk tags from
Last.fm (artist names, radio shows, playlist names) are silently dropped.
Applied at all 10 genre write points: Spotify, Last.fm, AudioDB, Deezer,
Discogs, iTunes, Qobuz enrichment workers + post-processing genre merge
+ initial download artist/album creation.
Strict mode is OFF by default — zero behavior change for existing users.
First enable auto-populates the whitelist with defaults. Users can add,
remove, search, and reset genres via the Settings UI.
The Duplicate Detector repair job had its own ignore_cross_album setting
that was independent of the global allow_duplicate_tracks setting. When
a user enabled 'Allow duplicate tracks across albums', the detector
still flagged same-titled tracks on different albums as duplicates.
Now respects the global setting — if duplicates are allowed, cross-album
matches are always skipped.
Users can now override which metadata provider (Spotify, Deezer, Apple Music,
Discogs) is used when scanning a specific watchlist artist for new releases.
The selector appears in the artist config modal and only shows sources the
artist has enrichment IDs for. Default behavior is unchanged — all artists
use the global metadata source unless explicitly overridden.
The redownload branch had `import json, uuid` locally inside the function,
which caused Python to treat `uuid` as a local variable for the entire
function scope. When the retag branch ran instead, `uuid` was unbound.
Both modules are already imported at the top of the file.
Move Hydrabase availability checks into metadata_service so source resolution owns the policy. Keep web_server delegating to the centralized helper and add tests for the enabled/disabled cases.
Move artist discography resolution into core metadata_service, introduce MetadataLookupOptions, and keep web_server focused on request handling. Add focused tests for the new service boundary and preserve current fallback behavior for now.
New MusicBrainz tab in Enhanced and Global search — finds tracks and
albums on MusicBrainz's community database with Cover Art Archive
images. Covers obscure tracks that Spotify/Deezer/iTunes miss.
- core/musicbrainz_search.py: search adapter with Track/Artist/Album
dataclasses, Cover Art Archive integration, smart query parsing
- Albums deduplicated (keeps best version with date and art)
- No artist results shown (MusicBrainz has no artist images)
- Album detail with full tracklist for download modal
- Smart word-boundary splitting for queries without separators
- Global search results container widened from 620px to 920px
- UI version bumped to 2.32
Files with embedded tags (artist+title from post-processing) were
failing import because the metadata search scored low (66%) and the
AcoustID result returned before the tag-preference code could run.
- Tag-based identification now returns 85% confidence when embedded
tags have an artist field, borrowing album art from weak metadata
- AcoustID search result only accepted at 80%+ confidence, otherwise
kept as fallback (doesn't short-circuit past tag preference)
- AcoustID None artist/title falls back to tag data via 'or' operator
- Stop retrying failed/unidentified items every scan cycle
Items with status needs_identification, failed, or rejected were not
in the skip list, causing them to be re-scanned and re-logged every
60 seconds indefinitely. Now skips all terminal statuses.
New 'soulsync' media server option manages the library directly from
the filesystem, bypassing Plex/Jellyfin/Navidrome entirely.
Two paths populate the library:
1. Downloads/imports write artist/album/track to DB immediately at
post-processing completion, with pre-populated enrichment IDs
(Spotify, Deezer, MusicBrainz) so workers skip re-discovery
2. soulsync_client.py scans Transfer folder for incremental/deep scan
via DatabaseUpdateWorker (same interface as server clients)
New files:
- core/soulsync_client.py: filesystem scanner implementing the same
interface as Plex/Jellyfin/Navidrome clients. Recursive folder scan,
Mutagen tag reading, artist/album/track grouping, hash-based stable
IDs, incremental scan by modification time.
Modified:
- web_server.py: _record_soulsync_library_entry() at post-processing
completion, client init, scan endpoint integration, status endpoint,
web_scan_manager media_clients dict, test-connection cache updates
- config/settings.py: accept 'soulsync' in set_active_media_server,
get_active_media_server_config, is_configured, validate_config
- core/web_scan_manager.py: add soulsync to server_client_map
Dedup: checks existing artist/album by name across ALL server sources
before inserting to avoid duplicates. Enrichment IDs only written when
the column is empty (won't overwrite existing data).
Race condition: scanner re-scanned folders while post-processing was
still moving files, causing partial matches and ghost failures. Now
tracks in-progress paths and skips them on subsequent scans.
Coverage penalty fix: individual tracks that match at 80%+ confidence
now auto-import even when overall album coverage is low (e.g. 2 of 18
tracks present). Previously low coverage killed the entire import.
Import page: stats bar, filter pills, Scan Now, Approve All, Clear
History (clears imported + failed), live scan progress.
- Track numbers defaulted to 1 instead of using metadata source values
- Release dates not captured, causing missing year in path templates
- Cover art missing for Deezer (direct image_url not checked)
- Track names in expanded view showed Unknown (wrong JSON field name)
- Read year/date from embedded file tags as fallback
- Add Deezer get_album_metadata/get_album_tracks fallbacks
- Handle Deezer tracks.data response format
Loose audio files in the staging root are now picked up alongside album
folders. Singles are identified via embedded tags, filename parsing
(Artist - Title.ext), or AcoustID fingerprinting, then matched against
the configured metadata source. Confidence-gated processing applies
the same way as album folders (90%+ auto, 70-90% review, <70% manual).
Soulseek results from "Various Artists", "VA", "Unknown Artist", and
"Unknown Album" folders are now rejected before scoring. These
compilation folders rarely contain properly tagged files for the target
artist.
Clearing the wishlist now also cancels any active wishlist download
batch and resets the auto-processing flag, so downloads don't keep
running after the source tracks are removed.
Priority 0 query (artist + album + title) was gated behind a download
mode check that excluded Soulseek, the source that benefits most from
it. Soulseek searches match against file paths where users organize as
Artist/Album/Track — without the album name, ambiguous artist names
could match wrong-artist results (e.g. "Bleakness" as an album folder
instead of an artist). Removed the mode gate so all sources get the
most specific query first.
Repair-worker album fills now generate explicit track IDs when copying rows, instead of relying on SQLite auto-assignment that no longer exists for TEXT primary keys. The unknown-artist fixer now does the same for new artists.
Also add a regression test for the album-fill copy branch and keep the AcoustID scanner resilient to legacy null-ID rows.
Full auto-import pipeline: background worker watches the staging folder,
identifies music using embedded tags → folder name parsing → AcoustID
fingerprinting, matches files to metadata source tracklists, and
processes high-confidence matches through the existing post-processing
pipeline automatically.
Worker: AutoImportWorker with start/stop/pause/resume, configurable
scan interval (default 60s), confidence threshold (default 90%), and
auto-process toggle. Processes one folder per cycle, alphabetical
order. Disc folder detection, stability checking, content hash dedup.
Confidence gate: 90%+ auto-processes silently, 70-90% queued as
pending review with approve/dismiss actions, <70% flagged for manual
identification. Track matching uses weighted algorithm (title 45%,
artist 15%, track number 30%, album tag 10%).
Database: auto_import_history table tracks every scan result with
folder hash, match data JSON, confidence, status, timestamps.
API: 7 endpoints — status, toggle, settings (GET/POST), results
(filtered/paginated), approve, reject.
UI: Auto tab on Import page with enable toggle, confidence slider,
scan interval selector. Live result cards with album art, confidence
bar (green/yellow/red), status badges, match stats. 5-second polling.
Switch similar-artist backfill to the shared provider-priority flow instead of assuming iTunes as the fallback.
Reuse the generic metadata search helpers, keep a compatibility alias for the old helper name, and update the scanner tests to cover the new path.
Add a regression test that verifies backfill walks each available fallback provider and persists the resolved IDs per source.
Shift similar-artist lookup to the shared metadata provider priority flow.
Use generic provider clients for search and metadata extraction instead of
branching on Spotify/iTunes-specific paths.
Add a regression test that verifies MusicMap matching queries the provider
priority list and preserves canonical metadata from the best match.
Make discovery pool population and curated playlists follow the configured metadata source order. Keep Spotify strict where fallback would corrupt source-specific IDs, and trim fan-out with smaller similar-artist samples and page caps. Leave the remaining incremental path for follow-up.
Reduce request volume in the discovery helpers while keeping the source-priority model intact.
- make cache_discovery_recent_albums source-priority aware
- cap Spotify artist-album pagination in the discovery and incremental paths
- reduce the similar-artist sample size for the cache-refresh helper
- keep Spotify strict where fallback would contaminate source-specific IDs
- add regression coverage for source order, strict Spotify lookups, and pagination caps
Watchlist scanner: empty discography (no new releases in lookback) was
treated as API failure, causing "Failed to get artist discography" for
artists like Kendrick Lamar who simply had no recent releases. Now
distinguishes None (API failure → try next source) from [] (success,
no new tracks). Spotify backfill now uses the authenticated client
instance instead of creating a fresh unauthenticated one.
Wishlist nebula: album remove now sends album_name (API updated to
accept album_name as fallback alongside album_id). Track remove
re-renders the nebula after deletion. Toned down processing pulse
animation.
Updated test to verify fallback triggers on API failure (None), not
on empty results.
Replaced track-count-only release selection with deterministic scoring
across 6 factors: track count match (40pts), release status (10pts),
country preference with US/worldwide bias (10pts), format preference
favoring Digital/CD over Vinyl/Cassette (10pts), barcode presence (3pts),
and date completeness (2pts). Same inputs always produce the same release.
Also fixed critical bug: _embed_source_ids was missing the context
parameter, silently skipping ALL source ID tag embedding since the
MusicBrainz consistency commit. Now passes context from the caller.
Make discovery pool population respect provider priority while keeping Spotify strict, and reduce unnecessary request volume in the hot discovery paths.
- keep discovery fan-out source-priority aware
- preserve cache use where freshness is not required
- cap Spotify artist-album pagination in discovery and cache refresh paths
- keep incremental release checks to a single page, since they only need the newest releases
- add regression coverage for provider order, strict Spotify handling, and pagination caps
Rewrote the AcoustID scanner job to scan all library tracks (via DB file
paths resolved to disk) instead of only the Transfer folder. Checkpoints
by track ID for robust resume across restarts. Defaults changed to
enabled, 24h interval, batch size 200.
Added _fix_acoustid_mismatch handler with three actions:
- retag: update DB title/artist to match actual audio content
- redownload: add expected track to wishlist and delete wrong file
- delete: remove wrong file and DB record
This catches cases like a file tagged as "Dinosaur Bones" that is
actually "Helicopters" — the scanner fingerprints the audio, detects
the mismatch, and the user can fix it from Library Maintenance findings.
Resolve Spotify artist matching through the exact Spotify client only, so watchlist ID backfill cannot drift to fallback-provider results. Remove the remaining preemptive provider availability check from the backfill loop.
Allow cached Spotify search results to return even when Spotify is rate-limited or temporarily unavailable, and remove redundant rate-limit gating after auth checks.
Drop the old active-provider artist lookup helpers from watchlist_scanner now that the web scan flow resolves sources through the shared metadata priority.
Keep the Spotify-specific feature toggles in place for discovery and sync paths that still use them.
Move the web watchlist scan core onto the shared metadata source priority so primary provider settings are respected during artist, album, and image resolution.
Add coverage for primary-source-first discography lookup and fallback to later providers when the primary source has no albums.
Bring placeholder tracklist skipping back into the shared watchlist scan path, and centralize the DB-only artist image backfill helper so both web scan entrypoints reuse the same logic.
