The MATERIALIZED-CTE scalar guard forced Postgres to materialize all object
rows with their full JSON (~100 GB) to temp before json_each — on NFS that's a
huge spill and pathologically slow (risks disk-full). Replace with an inline
CASE that feeds json_each an empty object for non-object rows: same scalar
guard, but a single streaming pass with no materialization.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Some image_record rows store tagger_predictions as a JSON scalar/null rather
than an object; json_each throws 'cannot deconstruct a scalar' on those,
rolling back the whole migration. Filter to json_typeof = 'object' in a
MATERIALIZED CTE so the guard runs before json_each ever evaluates a scalar.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The #764 in-place prune (rewrite tagger_predictions to >=0.70) is too slow on
100 GB of TOAST and fails at its soft limit (interrupts a query mid-flight ->
'another command is already in progress'). #768 supersedes it: extract only
the >=floor predictions into image_prediction via this set-based backfill,
then drop the column (step 3) — reading 100 GB once + writing ~840k small rows
beats rewriting 100 GB in place.
So this backfill no longer assumes the prune ran: it filters by
ml_settings.tagger_store_floor (default 0.70) itself, handling the full or
partially-pruned JSON identically.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Normalize tagger predictions out of the image_record.tagger_predictions JSON
blob into a queryable per-prediction table. Step 1 of the cutover (expand):
additive + low-risk — reads still use the JSON, this just adds the table and
keeps it populated.
- ImagePrediction(image_record_id, raw_name, category, score) — stores the
RAW tagger vocab name (not tag_id) so read-time alias→canonical resolution
is unchanged. Indexed for per-image reads + by (raw_name, score).
- Migration 0045: create table + set-based backfill from the JSON via
json_each (fast post-#764-prune). The old column stays (vestigial) and is
dropped in a later follow-up — DROP needs an ACCESS EXCLUSIVE lock on the
hot image_record table, so it waits for a quiesced-worker window.
- tag_and_embed dual-writes the rows (delete-then-insert, idempotent);
tagger_store_floor already applied in infer().
Next: switch suggestion + allowlist reads to the table, then drop the JSON
write. Plan-task #768.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Promotes the prediction store-floor from the TAGGER_STORE_FLOOR env (default
0.05) to a DB-backed, Settings-UI-tunable ml_settings column (default 0.70).
Storing every tag down to 0.05 from a ~10k-tag tagger is what grew
image_record's TOAST to ~100 GB; the suggestion path already filters at 0.70
and the centroid/learned path covers lower-confidence preferred tags, so the
sub-0.70 tail is redundant. Foundation for plan-task #764 (backfill + reclaim
land next; this only changes the write gate for NEW imports).
- ml_settings.tagger_store_floor (migration 0044, default 0.70)
- tagger.Tagger.infer(store_floor=...); ml task passes settings.tagger_store_floor
- ML admin GET/PATCH expose it; PATCH rejects a category suggestion threshold
below the floor (nothing below the floor is stored, so the gap surfaces
nothing) — server backstop for the UI slider clamp
- Settings → ML: store-floor slider + caption; category sliders min-bound to it
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The native Patreon backfill flooded the feed with bare 'Post <id>' shells
(1589 for Anduo). Root cause: PostAttachment.sha256 was GLOBALLY unique, so a
non-art file reused across posts only ever linked to the first one, and
_capture_attachment created the Post before that dedup check — leaving later
posts with no image and no attachment. Duplicate IMAGES had the mirror gap:
attach_in_place returned duplicate_hash/duplicate_phash before _apply_sidecar,
so the second post got no provenance row, and the feed only rendered via
primary_post_id (one post per image).
Operator requirement: a duplicate item must show on EVERY post it appears in.
Unify the fix as link-not-suppress:
- importer: on duplicate_hash / duplicate_phash(larger_exists), append an
image_provenance row for the new post (keep primary on the first). Both the
download path (attach_in_place) and the filesystem path (_import_media).
- post_feed_service: render thumbnails by image_provenance UNION primary_post_id,
so a cross-posted image shows on every post (and legacy primary-only images
still show).
- PostAttachment: per-post uniqueness — drop UNIQUE(sha256), add partial
UNIQUE(post_id, sha256) + partial UNIQUE(sha256) WHERE post_id IS NULL
(migration 0043); _capture_attachment dedups per-(post,sha) over the shared
sha-addressed blob, so no post is left bare.
- cleanup: new prune-bare-posts maintenance action (cleanup_service
_bare_post_conditions shared by preview/count/delete per preview/apply parity;
admin endpoint; PostMaintenanceCard). Deletes posts with zero image links
(primary or provenance) AND zero attachments. Run after the feed fix so a
hidden provenance link spares the post instead of deleting it.
Tests: dup image shows on both posts; dup attachment shows on both posts; feed
renders provenance-linked duplicates; prune-bare delete-path == preview.
Operator redeploys (migration 0043) then runs the prune to clear the shells.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Reverses the advisory-lock approach (7309d1d) — it treated a replica race that
wasn't the cause and added a new indefinite-hang mode (a sibling/stale migrator
holding the xact lock).
Real cause of the 0040 hang (operator-diagnosed 2026-06-07): web has always been
a single replica. The migration's ALTER series_page queued behind a concurrent
tag-merge that held a series_page lock for minutes — _do_merge repoints
series_page then runs _create_protective_aliases, an unindexed full scan of
image_record (JSON column, ~59k rows). Migrations ran with no lock_timeout, so
the DDL hung indefinitely and silently.
Fix: SET lock_timeout (default 30s, env-overridable) on the migration connection
before alembic's transaction. A blocked DDL now fails fast with 'canceling
statement due to lock timeout'; the entrypoint exits non-zero so the deploy
retries / surfaces loudly instead of wedging. General protection for every
future migration. (The slow _create_protective_aliases scan — the actual lock
holder — is the separate perf fix still under discussion.)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Operator feedback: thumbnails too small to judge order, no obvious way to mark
'this installment is Part 2', and the permanent two-pane picker was busy and
competed with the ordering work.
- Full-width parts, each a card with a big page grid (150px, contain so whole
pages are visible) and drag-to-reorder; positional page number as a badge.
- Editable Part # (hero field) backed by new series_chapter.stated_part —
separate from the auto-managed chapter_number, mirroring the page_number vs
stated_page split so reorder/delete renumbering can't wipe a hand-set part.
Missing-Part hints when consecutive parts' stated_part jump >1.
- Each part labels its source post (derived from pages' primary_post_id) and
shows the printed-page range with clear labels.
- Picker demoted to an on-demand right slide-over ('Add pages') with a target-
part selector; part actions (move/merge/delete) collapsed into an overflow ⋮.
alembic 0042 adds series_chapter.stated_part (nullable int).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Every web replica runs 'alembic upgrade head' in its entrypoint, so under
docker stack deploy two replicas can boot at once and race the same DDL —
0040 raced in prod (operator-flagged 2026-06-07): one backend wedged on the
series_page lock while a second tried to re-CREATE series_chapter, and the
loser died with AdminShutdown, crash-looping the web service.
Wrap run_migrations() in a transaction-scoped pg_advisory_xact_lock acquired
BEFORE the version table is read. The first replica to reach it migrates and
holds the lock for the whole upgrade; siblings block, then find the version
already at head and apply nothing. Works regardless of replica count and
needs no Swarm depends_on ordering (which stack deploy ignores anyway).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Confirm-only "this post may continue this series" matcher.
- series_suggestion table (post_id, series_tag_id, score, signals jsonb, status
pending|added|dismissed, UNIQUE(post,series)); migration 0041 + two settings
knobs (series_suggest_enabled, series_suggest_threshold).
- series_match_service: weighted additive score (title-stem / same-artist /
page-continuity / shared-distinctive-tags), no single signal gating. The title
"pattern" is derived on the fly from the post titles already in a series, so it
sharpens as more are confirmed (no persisted state to drift). Candidates are
bounded to the post's artist. match_post upserts pending suggestions (UNIQUE +
on-conflict, respecting prior added/dismissed decisions).
- accept reuses add_post_as_chapter then marks 'added'; dismiss marks 'dismissed'.
- rescan_series_suggestions_task: settings-gated, time-boxed + self-resuming from
a post-id cursor (maintenance_long lane), like normalize_tags_task.
- API: GET /series/suggestions, POST .../<id>/accept|dismiss, POST .../rescan.
- Settings: enabled + threshold exposed via /settings/import.
- Tests: pure scoring helpers + matcher/accept/dismiss/rescan lifecycle + UNIQUE
dedup.
Frontend (Suggestions tab + settings card) lands next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds an ordered chapter layer to series. Reading order becomes
(series_chapter.chapter_number, series_page.page_number); a chapter may be a
placeholder reserving a slot, and carries an optional parsed stated-page range
used to flag missing-page gaps. An image still lives in at most one series ⇒ one
chapter (image_id stays UNIQUE).
- models: series_chapter; series_page gains chapter_id (NOT NULL, cascade) +
stated_page. Migration 0040 backfills every existing series into one
auto-chapter holding its current flat pages — no data loss.
- SeriesService: chapter CRUD (create/update/reorder/delete/merge), page→chapter
assignment, reorder_pages, chapter-aware set_cover; list_pages now returns
chapters[] + gaps[] alongside a back-compat flat pages[]. Legacy series-wide
reorder operates on the single default chapter and rejects multi-chapter series.
- API: chapter endpoints under /api/series/<tag>/chapters; POST pages accepts an
optional chapter_id.
- TagService.merge now repoints series_chapter too, so a merged series' chapters
(and their pages) survive the source tag's deletion instead of cascading away.
- Tests: new chapter suite; updated the 4 direct SeriesPage(...) constructions to
supply chapter_id.
Frontend (chapter-aware manage view + reader) lands next; until then the
existing UI keeps working via the flat pages[] + single default chapter.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
scan_library_for_rule ran one 2-hour pass that timed out on large libraries and
held the concurrency-1 maintenance queue the whole time, starving vacuum/backup/
normalize (operator-flagged — it was the dominant entry in the 24h failures).
It now runs ~10-min chunks and re-enqueues itself until the library is
exhausted, matching the operator's preferred pattern (reasonable timeout → retry
queued → other things process between). New columns (alembic 0039):
resume_after_id persists the keyset cursor so a chunk continues where the last
left off; last_progress_at lets the recovery sweep tell a progressing multi-
chunk audit from a dead one (it now measures staleness from last_progress_at,
not started_at). Matches accumulate across chunks. soft/hard limits dropped
2h→15/16.7 min so the in-chunk budget fires first; a soft-limit backstop
re-enqueues to resume instead of erroring the whole run.
Tests: time-box → re-enqueue (status stays running); resume carries prior
matches and appends new ones. Existing full-scan tests unchanged (small sets
finish in one chunk).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A media that fails every walk (404'd CDN, deleted post, geo-blocked Mux,
persistently-corrupt bytes) used to re-error forever and re-burn chunks.
New `patreon_failed_media` table (alembic 0038, chains 0037) records
per-media attempts; once attempts reach DEAD_LETTER_THRESHOLD (3) the
ingester skips it on routine tick/backfill walks (tier-1.5, folded into the
seen/skip predicate). Recovery BYPASSES it (the operator's "try everything
again" re-attempts dead media). A clean download clears the row (recovered);
errors/quarantines upsert-increment it. Surfaced as
run_stats.dead_lettered_count.
- New PatreonFailedMedia model + migration; ingester _dead_keys /
_record_failures (on_conflict increment) / _clear_failures.
- skip = seen | dead (empty in recovery); failures recorded post-fetch on
short sessions (same pattern as the seen-ledger).
Tests: a media erroring 3× is dead-lettered + skipped (no download attempt);
recovery re-attempts a dead media and clears it on success; a clean download
clears a sub-threshold failure.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
patreon_seen_media(source_id, filehash, post_id, seen_at), UNIQUE(source_id,
filehash) — our own queryable replacement for gallery-dl's archive.sqlite3.
Routine walks skip seen media; recovery mode bypasses the ledger. filehash is
a 32-hex CDN MD5 or a video:<post>:<media> sentinel (String(128)). alembic
0037 (← 0036). Integration test covers dedup + savepoint recovery.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GalleryService.similar() ranks images by pgvector cosine distance to a source
image's precomputed SigLIP embedding — no query-time ML inference. Composes
with the Phase-1/2 scope filters (AND) but replaces the date sort (always
nearest-first, bounded top-N, no cursor). Returns None for a missing source
(→404), [] for a source with no embedding (video / pending ML); excludes self
and NULL-embedding rows. New GET /api/gallery/similar?similar_to=<id>&limit=N.
Image-detail payload gains has_embedding so the UI can hide the surface.
Alembic 0036 adds an HNSW vector_cosine_ops index on siglip_embedding (1152<2000
dims) so the search is sub-50ms ANN instead of a full scan; one-time ~30-60s
build over existing embeddings on deploy. Shared _gallery_images/_image_json
helpers de-dup the scroll/similar builders.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The gallery cursored on COALESCE(post.post_date, image_record.created_at)
across the Post outer join — an expression spanning two tables that no
index can serve, so every /scroll sorted a large slice of the library
(and the old frontend fired ten serially). Materialize it:
- image_record.effective_date column + ix_image_record_effective_date
(effective_date DESC, id DESC); alembic 0035 backfills
COALESCE(primary post's post_date, created_at) for existing rows.
- gallery_service._effective_date_col() now returns the column, so scroll
/ timeline / jump / neighbors all order off the index instead of
re-deriving the COALESCE. _neighbors reads record.effective_date
directly (drops an extra Post lookup).
- importer._apply_sidecar maintains it: when a primary post with a date is
linked, effective_date = post.post_date; plain inserts keep the
created_at-equivalent server default.
Tests: sidecar import asserts effective_date == post.post_date; gallery
ordering/timeline/jump test seeds set effective_date alongside created_at.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Per-artist "+N" accent pill on the artists directory and a "N new since
last visit" banner inside ArtistView. Counts new IMAGES (not posts) so
multi-image posts increment correctly.
- alembic 0034: artist_visit (artist_id PK, last_viewed_at NOT NULL).
Seeds every existing artist with last_viewed_at=NOW() so the badge
starts at 0 across the board — no noisy "5000 unseen images" on
first deploy.
- ArtistService.find_or_create autoseeds a visit row alongside new
artists, so freshly imported content doesn't read as unseen.
- ArtistService.overview reads pre-visit last_viewed_at, counts images
created since, then atomically UPSERTs last_viewed_at=NOW() via
postgres ON CONFLICT DO UPDATE (no SELECT-then-INSERT race per
reference_scalar_one_or_none_duplicates). Returns the pre-update
count as `unseen_count_at_visit` so the banner has data.
- ArtistDirectoryService.list_artists adds an `unseen_count` aggregate
to each card via LEFT JOIN artist_visit + conditional COUNT. NULL
last_viewed_at (artist created before this code shipped) defensively
counts as "never visited" → all images unseen.
- Frontend: ArtistCard renders an accent pill in the preview-strip
corner when unseen_count > 0 (capped at 99+); ArtistView shows a
closable v-alert banner on initial load when
unseen_count_at_visit > 0, re-arms on slug change.
Single-row-per-artist (no user_id) — rule #47 multi-user ACL is
aspirational; widens to (user_id, artist_id) PK when User lands, per
rule #22.
Scribe plan #597.
Operator-flagged 2026-06-02 — the 0.50 default (set on 2026-06-01)
surfaces too many low-confidence picks in the modal's Suggestions
rail. 0.70 keeps the rail signal-rich while still showing more than
the original 0.95 (which hid almost everything).
Alembic 0033 updates the singleton row conditionally — only rows
still at the old 0.50 default flip to 0.70. Operators who tuned to
some other value via Settings → ML keep their pick.
Settings UI already exposes both sliders (MLThresholdSliders.vue),
so further tuning continues to work without a deploy.
Alembic 0032 adds Source.error_type (varchar(32), indexed).
_update_source_health stamps it alongside last_error on status='error'
and clears it on 'ok'. SourceRecord/to_dict exposes it.
FailingSourcesCard renders a colored chip next to the consecutive-
failures count, with a tooltip explaining the suggested operator
action. Color reflects intent:
- warning (yellow) — operator action needed (auth_error)
- info (blue) — backend-paced (rate_limited / timeout /
network_error / partial / tier_limited)
- error (red) — likely terminal without intervention
(not_found / access_denied / validation_failed /
unsupported_url / http_error / unknown_error)
Audit 2026-06-02: the backend computed 13 ErrorType categories but
only the free-text last_error reached the operator. Bulk-triage by
class ("all auth_error → rotate cookies", "12 rate_limited → just
wait") required opening Logs per row.
Routine subscription polls walked the entire post history every tick
even when nothing had changed, because gallery-dl's default `skip: True`
continues iterating archived posts. A creator with ~550 archived posts
(Knuxy patreon) saturates the 870s wall-clock cap before completing,
even with zero downloads needed. Plus, a tier-limited run that
downloaded hundreds of files but ran out the clock should be a
warning, not an error.
Two coupled changes, both operator-flagged 2026-06-01:
* **Tick mode (default, cron polls).** New `TICK_SKIP_VALUE = "exit:20"`
asks gallery-dl to exit after 20 contiguous archived items. Fresh
subscriptions + new-content cases still walk normally; established
subscription with zero new content exits in ~30s of HEAD requests
instead of pegging the timeout. 20 (not 5) gives headroom against
paywall warnings interleaving with archived items.
* **Backfill mode (explicit, operator-triggered).** Sticky for N runs
via new `Source.backfill_runs_remaining` (alembic 0031). While > 0,
downloads use `skip: True` + 1800s timeout. Auto-decrements per run
with early-reset to 0 when a clean run finds zero files (queue
drained). N defaults to 3 — multiple runs give the system enough
budget to finish a deep walk across timeout boundaries. New
`POST /api/sources/{id}/backfill` arms the source; "Deep scan"
button on each SourceRow (chip shows remaining count) wires it.
Plus partial-success classifier: non-zero gallery-dl exit + ≥1 file
downloaded + no source-level error fires `ErrorType.PARTIAL`, which
download_service maps to `status=\"ok\"`. The run did real work; the
next tick continues via gallery-dl's archive. No more red events for
"timed out mid-walk after downloading 300 files."
Retires `SourceConfig.skip_existing` — skip value is now derived from
the source state and passed as a separate `skip_value` parameter
through download() / _build_config_for_source(). `GD_DEFAULTS` drops
the now-dead key (was inert data after this refactor).
Tests cover:
* tick + backfill skip-value emission in _build_config_for_source
* PARTIAL classifier branch + TIER_LIMITED-wins-over-PARTIAL ordering
* SourceService.set_backfill_runs validation + persistence
* /api/sources/{id}/backfill 200/400/404 paths
* download_service auto-decrement / auto-reset / tick-mode-no-touch
* PARTIAL → status=ok in the orchestrator (no consecutive_failures bump)
Operator-asked 2026-06-01 after the Dymkens orphan investigation
(Scribe plan #540). The pre-2030 sidecar-synthetic Source pattern
(`sidecar:<platform>:<slug>` enabled=false rows) existed solely to
satisfy `Post.source_id NOT NULL`, and leaked into the Subscriptions
UI as phantom subscriptions. Now the data model says what's true:
filesystem-imported content with no live subscription has NULL
source_id, full stop.
## Schema (alembic 0030)
- `post.artist_id` — NEW NOT NULL FK to artist (CASCADE). Backfilled
from source.artist_id in the migration. Indexed for the artist-filter
queries.
- `post.source_id` — NOT NULL → nullable; FK ondelete CASCADE → SET
NULL. Deleting a Source detaches its Posts instead of destroying
archived content (subscription ends, archive stays).
- `image_provenance.source_id` — same nullable + SET NULL.
- Partial unique index `uq_post_artist_external_id_null_source` on
(artist_id, external_post_id) WHERE source_id IS NULL — guards
filesystem-import dedup since the existing source-bound unique
ignores NULLs (Postgres treats NULL != NULL).
- Sidecar synthetic Sources deleted: NULL out FKs in post,
image_provenance first, then DELETE FROM source WHERE url LIKE
'sidecar:%'. The Dymkens cleanup.
## Model + service changes
- `Post.source_id` → `Mapped[int | None]`; new `Post.artist_id`
denormalized.
- `ImageProvenance.source_id` → `Mapped[int | None]`.
- Importer: `_source_for_sidecar` (synthetic-creating) →
`_lookup_source_for_sidecar` (returns None when no subscription).
`_find_or_create_post` takes required `artist_id`; matches on
(source_id, external_post_id) for source-bound posts or
(artist_id, external_post_id) for NULL-source posts.
- Service queries switched off the Source detour to use Post.artist_id
directly: post_feed_service.scroll/around/get_post (LEFT JOIN to
Source so NULL-source posts surface); artist_service date_row/
activity/post_count; provenance_service.for_image/for_post (LEFT
JOIN); gallery_service._provenance_exists_where_artist via
Post.artist_id instead of ImageProvenance.source_id → Source.
- `_to_dict` and provenance dict-builders emit `"source": null` for
NULL-source rows.
## Frontend
- `ProvenancePanel.vue` + `PostCard.vue`: render `e.source?.platform
?? 'filesystem import'` so NULL-source posts get a clear
"filesystem import" affordance instead of a NaN crash.
## Tests
- `test_importer_upsert_helpers`: removed the four synthetic-anchor
tests; added `_find_or_create_post_idempotent_with_null_source`
(dedup via the partial unique index) and
`_lookup_source_for_sidecar_returns_*` (existing-subscription +
none cases). The existing `_find_or_create_post_idempotent` now
also passes `artist_id` and asserts it.
- 8 other test files updated: every direct `Post(...)` construction
gains `artist_id=<artist>.id`. The `_seed_post` helper in
`test_post_feed_service` looks up artist_id from the source row so
callsites stay one-arg.
## Verification on deploy
After alembic 0030 runs:
- `SELECT COUNT(*) FROM source WHERE url LIKE 'sidecar:%'` → 0.
- `SELECT COUNT(*) FROM post WHERE source_id IS NULL` → count of
filesystem-imported posts (Dymkens + any other historical).
- Every `post.artist_id` non-null; consistent with source.artist_id
for source-bound rows.
- Subscriptions tab: no Dymkens phantom row.
- Artist detail → Posts/Gallery: Dymkens's content still reachable
via Post.artist_id.
- Provenance panel renders "filesystem import" chip for NULL-source
posts; PostCard same.
## Out of scope
- UI to manage/delete orphan NULL-source Posts. Data model is right;
UI follows if operator wants it.
Four coupled operator-asked changes to the view modal (Scribe plan #509):
1. **Autofocus tag entry on modal open** — TagAutocomplete grabs focus
in onMounted/nextTick so the caret is in the input the moment the
modal renders. No click needed to start typing.
2. **General suggestions expanded by default** — SuggestionsPanel's
general-category group now mounts with `:default-open="true"`.
Operator can collapse if too noisy, but the v1 frame shows them.
3. **Lower general threshold default 0.95 → 0.50** — MLSettings.
suggestion_threshold_general default matches character. Alembic
0029 also bumps the existing singleton row's value if it's still
at the old 0.95. Operator can re-tune from Settings → ML.
4. **Retire `copyright` + `artist` as ML suggestion categories** —
neither feeds a Tag.kind (`artist` retired in FC-2d-vii-c, never
really existed as a copyright tag-kind). They were surfaced in the
suggestions pipeline + threshold settings UI but had no follow-
through. Drop from SURFACED_CATEGORIES, suggestions._threshold_for,
ml_admin GET/PATCH allowlist, MLSettings columns (alembic 0029
drops the two columns), frontend CATEGORY_ORDER + CATEGORY_LABELS,
SuggestionsPanel.peopleCats, AliasPickerDialog kind-check, and
MLThresholdSliders rows.
Out of scope (intentional): `tag_kind` Postgres enum still includes
`artist` for historic Tag row queryability (per the model comment);
no operator pain reported, no enum-shrink needed.
Tests:
- test_surfaced_categories asserts {character, general}, excludes
artist + copyright.
- test_threshold_for_artist_is_unsurfaced extended to cover copyright.
- test_get_and_patch_settings asserts new 0.50 default and the absent
artist + copyright keys in the GET payload.
Two coupled bugs surfaced 2026-05-31 by the Subscriptions UI showing
"phantom" subscriptions like `sidecar:patreon:dpmaker`:
1. `SourceService.list()` returned every Source, no filter on URL.
alembic 0022 (2026-05-26) consolidated old per-post-URL Sources into
one canonical row per (artist, platform); when no real campaign URL
was salvageable it rewrote the canonical to `sidecar:<plat>:<slug>`
enabled=false as a disabled anchor. The UI then listed those
anchors as if they were polls — disabled, but visible. Fix: `list()`
excludes `url LIKE 'sidecar:%'` by default; `include_synthetic=True`
opts back in for admin tooling.
2. `importer._source_for_sidecar` picked the lowest-id Source for
(artist, platform). When alembic 0022 had rewritten a per-post row
into a synthetic anchor (lower id) AND the operator later added the
real subscription (higher id), every gallery-dl download silently
attached its Post to the SYNTHETIC instead of the real Source. Fix:
prefer a non-`sidecar:%` URL when one exists; fall back to the
synthetic; only create a new synthetic when nothing exists for
(artist, platform).
alembic 0028 is the data half: for every (artist, platform) with both
a synthetic AND a real Source, pre-merge Post+ImageProvenance
collisions on the canonical, bulk-repoint Posts/ImageProvenance/
DownloadEvent.source_id onto the real Source, and delete the
synthetic. Lone synthetics (no real twin) are left intact — they
anchor real imported content the operator may still want; the
list-filter hides them so they no longer surface as phantoms.
Layer 1 of the import-task resilience work (operator-requested
2026-05-28). The recover_interrupted_tasks sweep re-queues rows stuck
in 'processing' — correct for a worker crash, but without a cap a row
that RELIABLY hard-crashes the worker (OOM/segfault/SIGKILL on a
corrupt or oversized input) loops forever: re-queue → crash → re-queue,
burning a worker slot every 5 min. A caught exception flips to terminal
'failed' and never enters this loop; only process-killing inputs do.
- alembic 0026: import_task.recovery_count (int, default 0) +
import_task.refetched (bool, default false — backs Layer 2).
- recover_interrupted_tasks now runs a poison-pill UPDATE FIRST: stuck
rows whose recovery_count has already reached MAX_RECOVERY_ATTEMPTS-1
are marked 'failed' with a diagnostic ("crashed or stalled the worker
N times … likely a corrupt or oversized input … inspect/replace the
file, then retry via /api/import/retry-failed") instead of re-queued.
The re-queue pass then handles the remaining stuck rows and bumps
recovery_count. Shared stuck_predicate (and_/or_) keeps the
media-5min / archive-40min split.
- MAX_RECOVERY_ATTEMPTS=3 (two recoveries then give up).
The failed poison pill surfaces in the existing import-failures view
with its file path, directly answering "help me identify them."
Test test_recover_interrupted_poison_pill_caps_at_max pins both
branches: a row at the cap is failed (not re-enqueued, diagnostic
present), a row one short is re-queued + incremented.
Operator-flagged: the claim that 'a future deep-scan via the new parser
will fix HF and Discord post_url' was conditional on the operator
actually running a deep-scan, which they might not do for ages. Until
then HF posts stay at post_url=NULL (HF sidecars have no `url` field)
and Discord posts stay pointing at cdn.discordapp.com/attachments/...
(the file URL, not the message permalink).
The migration was already opening sidecar files for SubscribeStar.
Generalizing the loop to also handle HF and Discord is a tiny addition
that closes the gap without operator intervention.
Per-platform Part 1 logic now:
subscribestar — read sidecar.post_id, overwrite external_post_id +
post_url with the derived /posts/<post_id> permalink.
hentaifoundry — read sidecar.user + .index, overwrite post_url with
/pictures/user/<u>/<i>. external_post_id (= index) unchanged.
discord — read sidecar.server_id + .channel_id + .message_id,
overwrite post_url with the discord.com/channels/.../<m> triple.
external_post_id (= message_id) unchanged.
Part 2 (SubscribeStar fragment merge) and Part 3 (pure-SQL Pixiv
post_url backfill) unchanged.
Posts whose related ImageRecord paths don't resolve on disk (orphan
filesystem state) are reported per-platform in the migration output —
those still need a future deep-scan, but the in-DB-with-on-disk-files
common case is now fully covered by the migration alone.
Audit of one sample sidecar per platform on the operator's
/mnt/Data/Patreon/ archive surfaced three parser bugs that have been
silently corrupting non-Patreon Posts since FC-3 shipped:
1. SubscribeStar `id` vs `post_id` confusion. gallery-dl puts the
per-attachment id in `id` (e.g. 711509) and the actual post id in
`post_id` (e.g. 360360). FC's external_post_id chain had `id`
winning, so every multi-image SubscribeStar post was fragmented into
N Post rows in the database. Reorder the chain to
`("post_id", "id", "index", "message_id")` — Patreon/Pixiv (no
`post_id`), HF (uses `index`), Discord (uses `message_id`) all
unaffected.
2. Discord `message` field not captured. Discord posts put the body in
`message`, not `content`. Append it to the description fallback chain
`("content", "description", "caption", "message")`.
3. post_url is the file URL on SubscribeStar/Pixiv/HF/Discord. New
`_derive_post_url(platform, data)` helper synthesizes proper
permalinks from per-platform fields:
subscribestar → https://www.subscribestar.com/posts/<post_id>
pixiv → https://www.pixiv.net/artworks/<id>
hentaifoundry → https://www.hentai-foundry.com/pictures/user/<user>/<index>
discord → https://discord.com/channels/<server>/<channel>/<message>
Patreon's bare `url` IS a real permalink and is used as-is. For the
four file-URL platforms, the bare `url` is NEVER trusted: derive or
return None rather than persist a CDN URL.
Tests:
- `test_parse_core_fields_and_id_priority` flipped to assert post_id
wins over id.
- New `test_parse_id_used_when_no_post_id` covers the Patreon real
shape.
- New `test_parse_message_used_as_description_fallback` covers Discord
bodies.
- Five new tests cover per-platform post_url derivation
(SubscribeStar/Pixiv/HF/Discord/Patreon-untouched + missing-fields →
None).
Cleanup migration alembic 0025_fix_subscribestar_post_ids:
- For each SubscribeStar Post: find a related ImageRecord.path, walk to
its sidecar JSON, read `post_id`, overwrite Post.external_post_id +
post_url with the corrected values.
- After all updates, every group of Posts under one source sharing the
same NEW external_post_id is a fragment-set — merge to a canonical
row using the same ImageProvenance pre-delete + repoint dance as
alembic 0022 (banked pattern).
- Pure-SQL backfill of Pixiv post_url: replace any `i.pximg.net`-shape
url with the derived `/artworks/<id>` permalink.
- HF and Discord post_url backfills skipped — HF would need the `user`
field (not stored on Post), Discord needs server/channel triple.
Both will be corrected by a deep-scan re-applying sidecars through
the new parser.
Idempotent: re-running on already-corrected data is a no-op.
SubscribeStar gallery-dl always writes `title: ""` and embeds the leading
sentence inside `content` HTML. Confirmed against the operator's
/mnt/Data/Patreon/Cheunart/subscribestar/ dump 2026-05-27: every
post's JSON has `title: ""` and a content like
`<div>Lets say hello to you guys with my Belle <br><br><br></div>`.
FC's sidecar parser, treating empty strings as missing, had been leaving
post_title NULL on every subscribestar post since FC-3 shipped.
Fix at two layers:
1. `backend/app/utils/sidecar.py` — new `_first_line_text(body, limit)`
helper strips HTML tags, collapses whitespace, returns the first
non-empty line truncated to 120 chars with ellipsis. `parse_sidecar`
now falls back to this when `title` resolves to None and a
`content`/`description`/`caption` value is present. Patreon's
non-empty titles short-circuit the fallback so existing behavior is
unchanged. Four new tests in test_sidecar_util.py pin: derivation
from content, truncation at 120 chars, explicit-title precedence,
no-content no-fallback.
2. `alembic 0024_backfill_post_title_from_description` — backfills the
same logic across existing Post rows where `post_title IS NULL OR
post_title = ''` AND description is present. Idempotent (re-running
is a no-op once titles are populated). Downgrade is a no-op since
there's no safe way to tell derived rows from genuine ones.
After deploy + migration: subscribestar posts will surface a meaningful
title in PostCard, post feed search, etc.
extension/manifest.json: add content_security_policy.extension_pages = "script-src 'self'; object-src 'self';" — explicitly omits the upgrade-insecure-requests directive that MV3 inherits by default. Without this, every fetch(http://curator.../...) silently upgrades to https:// at the browser layer (Sec-Fetch-Site=same-origin, NS_ERROR_GENERATE_FAILURE), regardless of about:config. Bump XPI version 1.0.3 → 1.0.4 so a fresh signed build replaces the cached one. Operator-troubleshot 2026-05-26 via Inspect-the-extension dev tools showing the silent scheme upgrade.
alembic 0023: drop ck_tag_fandom_requires_character before the tag_kind type swap and recreate after. Postgres can't resolve `kind = 'character'` across the rename (column on tag_kind_old, literal binds to new tag_kind → "operator does not exist"). Same dance on downgrade. Banked under reference_tag_kind_enum_swap_check_drop.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Operator-flagged 2026-05-26: Atole artist detail page showed 406 Sources where 1 was right.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
bvandeusen