573228b9da2bb08584d6d1a9704ecc3d04c64a9d
26
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
  bvandeusenandClaude Opus 5
|
573228b9da |
db: finish reconciling the models with the deployed schema (#3275)
CI / lint (push) Failing after 3s
Build images / sign-extension (push) Successful in 4s
CI / extension-version (push) Successful in 3s
Build images / build-agent (push) Successful in 9s
CI / frontend-build (push) Successful in 34s
Build images / build-ml (push) Successful in 53s
Build images / build-web (push) Successful in 44s
CI / integration (push) Successful in 4m5s
CI / backend-lint-and-test (push) Successful in 1m6s
Closes the residue the first reconciliation pass left, and corrects a
factual error I put into the record.
sha256 was NOT missing a uniqueness guarantee. I read
`op.create_index("ix_image_record_sha256", ...)` at 0001 line 151 and
concluded duplicates were possible, without reading line 149 two lines
above it:
sa.UniqueConstraint("sha256", name="uq_image_record_sha256"),
Uniqueness has held since the initial schema. The database expresses it
as a CONSTRAINT plus a separate non-unique lookup index; the model said
`unique=True, index=True`, which is one UNIQUE index under a different
name. Same guarantee, different objects — which is exactly why the two
schemas did not line up. The model now declares both objects. No DDL.
0088's docstring, which repeated the claim, is corrected in place.
Two real divergences, both the MODEL over-claiming:
* source: uq_source_artist_platform_url (alembic 0010) was declared
nowhere in the models — source.py had no __table_args__ at all — so
autogenerate would have proposed DROPPING it.
* head_metrics_snapshot.tag_id: model said NOT NULL, 0060 created it
nullable. Left nullable; the FK already cascades.
Seven constraints renamed to what the chain actually created, rather than
what base.py's naming convention renders: uq_series_page_image,
uq_series_chapter_anchor_page, fk_series_chapter_anchor_page,
fk_image_record_artist_id, fk_image_provenance_from_attachment, and the
two hand-shortened fk_tsr_* names from 0003.
Float server_defaults now mirror their own migration, per column. The
chain is MIXED: a plain string renders DEFAULT '0.90'::double precision,
sa.text() renders DEFAULT 0.90, and the migrations used both. Seven
columns take text(); the rest stay strings. Two literals also disagreed
outright — process_{auto_apply,conflict}_threshold said 0.9/0.5 against
the migration's 0.90/0.50.
baseline.yml gains two things. A repair for a SECOND generator defect in
the same class as the missing pgvector import: base.py's ck convention
contains %(constraint_name)s, so it applies even to a NAMED
CheckConstraint — autogenerate writes the already-rendered name into the
migration and running it applies the convention again, yielding
ck_ml_settings_ck_ml_settings_singleton. That is round-tripping damage,
not a claim the models make, so it is undone rather than counted.
And the diff now runs twice. Column ORDER differs permanently between a
schema built by 87 ADD COLUMNs and one built in a single shot — the
operator's database keeps chain order forever, a fresh install gets model
order — so a check that failed on it could never pass. The second pass
SORTS column lines within each CREATE TABLE instead of deleting them,
which cannot hide a column present on one side only, or one whose type,
nullability or default differs. Ordered diff is reported as information;
the order-insensitive one is the verdict.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017QHszn9H8VBvx5Ke8x1hvw
|
||
|
bvandeusen
|
5e1996e77f |
db: reconcile the models with the deployed schema (#3275)
Build images / sign-extension (push) Successful in 4s
CI / lint (push) Failing after 2s
CI / extension-version (push) Successful in 2s
Build images / build-agent (push) Successful in 7s
CI / frontend-build (push) Successful in 27s
Build images / build-ml (push) Successful in 48s
CI / backend-lint-and-test (push) Successful in 1m7s
Build images / build-web (push) Successful in 40s
CI / integration (push) Successful in 4m1s
Milestone 328's acceptance test compared a database built by the real
0001..0087 chain against one built from the models, and found ~130
places where they disagree. This closes them.
Almost all were the MODEL being wrong, so almost all of this is model
edits with no DDL — the database already had these things, nothing in it
changes, and no deploy is needed for this part:
* 92 columns gained server_default. The models carried Python-side
`default=` only, so the ORM filled the value and the column had no
database default. Anything inserting outside the ORM behaved
differently from production.
* Eleven indexes that existed only in migrations are now declared:
the three backup_run reporting indexes, the two date-ordered
image_record browse indexes, import_task and presentation_review,
and the three task_run history indexes. All use text() for their DESC
ordering and postgresql_where for the partial one.
* Two UNIQUE indexes that autogenerate silently proposed DROPPING,
because neither is expressible as a UniqueConstraint:
uq_tag_name_kind_fandom — an EXPRESSION index over
(name, kind, COALESCE(fandom_id, 0))
uq_post_artist_external_id_null_source — PARTIAL, WHERE source_id
IS NULL
post.py already had a comment describing the second one. The comment
was right; nothing declared it.
* The two external_link enum CHECKs (host, status) — rule 36 territory,
and absent from the model entirely.
* Two indexes were named explicitly. A bare index=True generated
ix_tag_alias_canonical_tag_id where the database has
ix_tag_alias_canonical, so autogenerate proposed a drop+create of an
index that was already there under another name. Same for
tag_suggestion_rejection.
Only ONE thing needed DDL, as 0088: tag.fandom_id is declared
index=True but no migration ever created that index.
Deliberately NOT here: image_record.sha256. The model says unique=True;
0001 created a plain index. Duplicates are possible today and the ORM
believes otherwise. The fix depends on whether duplicates already exist
— if they do, that is a dedupe decision, not a constraint — so it waits
on an answer about live data.
The real severity of #3275 is not the squash. It is that --autogenerate
has been unsafe on this project: run against the old models it would
have proposed dropping eleven indexes and two uniqueness guarantees.
|
||
|
bvandeusenandClaude Opus 4.8
|
05df51b749 |
fix(ml): drop unnecessary quotes on MLSettings.load annotations (UP037)
Python 3.14 evaluates annotations lazily, so the self-referential return type needs no forward-ref quotes — matches ImportSettings.load. Fixes the ruff lint lane on the DRY-pass push. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NsmJSQxnNxGgtM5Yz4GAqi |
||
|
bvandeusenandClaude Opus 4.8
|
666b3a2ec8 |
refactor(ml): DRY pass — shared sweep helpers + table-driven settings (#161)
Consolidate duplication accrued across the ML tagging + settings backend, behavior-preserving (over-DRY guard applied — the three auto-apply sweep BODIES stay separate; only their shared inner helpers are extracted). - _sigmoid / _conflict_scores / _insert_presentation_review (heads.py): the score→prob transform (6 inlined sites), the presentation conflict signal (2 sites), and the ring-loud PresentationReview insert (2 sites, single- sourced so the mode column can't drift on the shared composite PK). - _applied_or_rejected (training_data.py): the per-tag "applied ∪ rejected" skip-set, byte-identical at 3 sweep sites (heads.py x2, tasks/ml.py ccip). - ccip sweep divergence fixes: import ccip._FIGURE_KINDS + training_data._l2norm instead of local copies that silently drift when the canonical changes. - MLSettings.load / .load_sync classmethods (mirror ImportSettings); route all 8 scalar_one singleton reads through them (the session.get None-path stays). - GET serializers for MLSettings + ImportSettings are now table-driven off the same _EDITABLE tuples PATCH writes, so a new field can't be silently absent from GET (the split that historically dropped fields). - AUTO_APPLY_THRESHOLD_MIN/MAX constant single-sources the [0.5,0.999] operating range across the service clamp + the 5 API validators. - test_ml_dry_helpers.py pins _applied_or_rejected + _sigmoid. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NsmJSQxnNxGgtM5Yz4GAqi |
||
|
bvandeusenandClaude Opus 4.8
|
ad2a5fc5fe |
feat(system-tags): process vs chrome groups + WIP provisional auto-apply (#1464)
Backend for the system-tag behavior refactor (milestone #157). editor screenshot moves from chrome (hidden) to the PROCESS group (shown, like wip); wip+editor gain provisional auto-apply so they stop needing endless manual identification — without a runaway loop. - tag.py: split PRESENTATION_SYSTEM_TAGS → CHROME_SYSTEM_TAGS (banner) + PROCESS_SYSTEM_TAGS (wip, editor screenshot). - heads.py: generalize presentation_auto_apply_sweep → system_tag_auto_apply_sweep (mode chrome|process). Same Guard 1 (skip human/confirmed) + Guard 2 (ring-loud conflict → PresentationReview). process mode uses source 'process_auto' and does NOT hide (hide is a gallery-query effect of group membership). - training_data._AUTO_SOURCES += 'process_auto' → the head never trains on its own auto-applied output; only wip_title/manual train it (the runaway break). - ml_settings: process_auto_apply_enabled (OFF, opt-in) + threshold + conflict threshold. presentation_review.mode ('chrome'|'process'). Migration 0086. - gallery_service: default-hide reads CHROME only (editor now shows); Explore neighbors exclude the whole PROCESS group. - tasks/ml + celery beat: scheduled_process_auto_apply (daily, opt-in); prune covers both modes. - api: ml_admin process_* CRUD+validation; hidden-review returns mode. - tests: rename chrome sweep calls; new test_process_auto_apply (apply, guards, mode flag, no-self-train); gallery test asserts editor now visible. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> |
||
|
bvandeusenandClaude Opus 4.8
|
ab63d94249 |
feat(ml): presentation auto-hide settings + review table (#141 step 3)
MLSettings gains presentation_auto_apply_enabled / _threshold (default 0.90) + presentation_conflict_threshold (default 0.50): banner/editor auto-hide with a FLAT threshold (decoupled from content-head graduation), plus the "also looks like content" conflict cut. New presentation_review table (image, presentation tag, conflict tag + score, created/resolved_at) records auto-hides flagged for review. Migration 0082 (columns + table), ml_admin API (editable + get_settings + _validate bounds), settings roundtrip/bounds test. The sweep that reads these knobs + the Settings UI land in step 4. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CDgx8bQS5YrGRK76v8HUnM |
||
|
bvandeusenandClaude Opus 4.8
|
7d3a3b4a83 |
revert(ml): keep head auto-apply precision at 0.97 (operator: general tuning was fine)
Milestone 139 raised head_auto_apply_precision 0.97→0.98; operator confirmed the general-tag confidence was already well tuned, so revert that. The support floor (min_positives 30→50) and CCIP match confidence (0.92→0.95) stay. Migration 0081 (not yet deployed) edited to drop the precision bump. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CDgx8bQS5YrGRK76v8HUnM |
||
|
bvandeusenandClaude Opus 4.8
|
cbc3e11a53 |
feat(ml): stricter auto-apply defaults to cut misfires (milestone 139)
head_auto_apply_precision 0.97→0.98, head_auto_apply_min_positives 30→50, ccip_auto_apply_threshold 0.92→0.95 (operator-asked). Model defaults change for fresh installs; migration 0081 bumps the existing singleton row IFF still at the old default (won't clobber a deliberate operator change). ml_admin bounds already permit these. Fixed a stale comment in the auto-apply test. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CDgx8bQS5YrGRK76v8HUnM |
||
|
bvandeusenandClaude Opus 4.8
|
f24dc81764 |
feat(ccip): schema for precomputed incremental character prototypes (#1317, m138 step 1)
Foundation for making CCIP character references a precomputed, INCREMENTAL artifact instead of a request-path rebuild (kills the per-accept ~4s suggestions stall; cost will scale with change, not library size): - character_prototype: a character's reference CCIP vectors, capped to MLSettings.ccip_prototype_cap so match cost doesn't grow with popularity. - ccip_prototype_state: per-character fingerprint (ref count + max region id) + updated_at → drives per-character incremental rebuilds and the matcher cache's reload-only-what-advanced. - MLSettings.ccip_ref_signature (cheap global change gate) + ccip_prototype_cap. Migration 0079. Schema + models only — the builder service, refresh task/beat, and matcher rewrite land in the following steps. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CDgx8bQS5YrGRK76v8HUnM |
||
|
bvandeusenandClaude Opus 4.8
|
62ec70b9e4 |
feat(ml): detector config in MLSettings with working defaults (#134 step 1)
Move the crop-proposer config (per-proposer enable + weights + conf, caps, dedupe IoU) into the DB so it's UI-tunable and can be announced to the GPU agent in the lease (like the embedder model) — no restart, agent env becomes bootstrap-only. Migration 0078 adds the columns with working server_defaults so existing rows + fresh installs crop out-of-the-box with all three proposers ON (operator: default-on): person=yolo11n.pt, anatomy=booru_yolo yolov11m_aa22 (URL, license unstated/private-homelab-OK), panel=mosesb best.pt. Plain columns, no CHECK enum. Steps 2 (lease announce + agent apply) and 3 (Settings UI) follow. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CDgx8bQS5YrGRK76v8HUnM |
||
|
bvandeusenandClaude Fable 5
|
19b962f1a7 |
feat(b3): ml-worker becomes optional — embed-only role, decoupled GPU coordination, cpu-embed switch
The ml-worker's ONLY processing role is now the CPU whole-image embed fallback (tag_and_embed renamed embed_image — Camie tagging was retired #1189 and the name kept implying otherwise; videos were already handled agent-style: frame sampling + mean-pool). Detection/cropping/CCIP stay GPU-agent-only, and their completion is judged per-pipeline: ccip by gpu_job rows, siglip by concept regions at the current model version — never by image_record.siglip_embedding. A CPU embed therefore can NEVER close crop work for the agent (regression test pins this; only the whole-image 'embed' job, the same artifact, is satisfied). Making removal actually safe (operator will drop the container): - GPU-queue coordination (enqueue_gpu_backfill, recover_orphaned_gpu_jobs, reprocess_gpu_jobs) moved verbatim to tasks/gpu_queue.py on the maintenance quick lane — it lived on the 'ml' queue only by module colocation, which made the ml-worker a hard dependency of the whole agent pipeline. - New ml_settings.cpu_embed_enabled (migration 0074, default ON so agent-less installs keep working): OFF stops the four import hooks queueing embed work nothing will consume and no-ops the manual backfill; switch lives on the renamed 'CPU embedding backfill' card. - NB heads training / auto-apply still run on the ml image (sklearn) — a stack that removes the container gives those up too. Deploy note: in-flight messages under the old task names are dropped by the new workers; the 60s orphan sweep + hourly backfill re-fire under the new names immediately. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CDgx8bQS5YrGRK76v8HUnM |
||
|
bvandeusenandClaude Opus 4.8
|
359bc5a283 |
feat(ml): default to SigLIP 2 (new installs) + model dropdown, no free-text (#1203)
- Migration 0069: new installs default to SigLIP 2 (so400m, 512px, 1152-d drop-in) — UPDATE applies ONLY where no image is embedded yet (fresh install), so an existing library is NOT silently invalidated; it switches deliberately via the dropdown → Re-embed → Retrain. Column server_defaults moved to SigLIP 2. - GET /api/ml/embedder-models: server-authoritative supported list (SigLIP 2 512 recommended / 384 faster / SigLIP 1 384 original) so the UI never free-types. - GpuAgentCard: the two name/version text fields → a single model dropdown; Save sets name+version from the picked option (the current model is always selectable even if off-list). - embedder.py DEFAULT_MODEL_NAME unchanged (stays the baked local-dir SigLIP 1) to avoid a local-dir/weights mismatch; SigLIP 2 loads by HF name, cached on the ml-worker's persistent HF_HOME. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ttrj5P7upUTueSfoJcxEqa |
||
|
bvandeusenandClaude Opus 4.8
|
bc6d43d3f2 |
refactor(ml): drop dead tagger/suggestion settings + columns (#1199)
Hygiene follow-up to the Camie retirement (#1189) — these were left inert to bound that change; nothing reads them now. Migration 0068 drops: - ml_settings: tagger_store_floor, tagger_model_version, suggestion_threshold_ character/general (already dead pre-retirement — scoring uses per-head thresholds), video_min_tag_frames (only the deleted video-prediction aggregator used it). - image_record: tagger_model_version (no writer), centroid_scores (dead JSON cache, no reader). Also: ml_admin _EDITABLE/GET/_validate pruned (dropped the store-floor invariant + video_min_tag_frames check); MLThresholdSliders trimmed to a video-embedding card (interval + max frames only); importer no longer resets the dropped cols; download_models drops the Camie fetch; stale CASCADE comments in cleanup_service no longer name the removed tables. Tests updated. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ttrj5P7upUTueSfoJcxEqa |
||
|
bvandeusenandClaude Opus 4.8
|
3d77a38a25 |
refactor(ml): remove the dead per-tag centroid subsystem (#1189)
The v2 pivot replaced per-tag SigLIP centroids with learned heads + CCIP. Centroids were still recomputed (on every tag merge + a daily beat) but NOTHING read them — suggestions come from heads+CCIP and apply_allowlist_tags applies via Camie predictions, not centroids. Pure dead wiring; remove it. Removed: CentroidService, recompute_centroid/recompute_centroids tasks, the daily beat, POST /api/ml/recompute-centroids, the recompute-on-merge trigger, the tag_reference_embedding table + model, the centroid_similarity_threshold + min_reference_images settings (migration 0066), the CentroidRecomputeCard + its store action + MaintenancePanel tile, and the centroid slider in MLThresholdSliders. _keep_as_alias drops its vestigial has-centroid branch (the allowlist branch already covers "could re-emit"); tag merge no longer clears a table that no longer exists. NOT touched (still live, parallel to heads): the Camie tagger, ImagePrediction, and the allowlist bulk-apply — accepting a suggestion still allowlists + applies it across the library. The tag-eval "centroid" baseline metric is unrelated (in-memory) and stays. (image_record.centroid_scores JSON column also remains — separate legacy field, its own micro-cleanup.) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ttrj5P7upUTueSfoJcxEqa |
||
|
bvandeusenandClaude Opus 4.8
|
4daa3f2790 |
feat(ml): operator model swap — GPU re-embed + embedder as a setting (#1190)
Make the SigLIP embedder an operator choice (drop-in to SigLIP 2:
google/siglip2-so400m-patch16-512 is a verified 1152-d model at 512px → no
schema change, better small-cue fidelity). A swap = set model + re-embed +
retrain, all operator-driven; the GPU agent does the re-embed so it's fast.
- settings: embedder_model_name is now a setting (migration 0065) alongside the
existing embedder_model_version; both editable + validated (non-empty) in the
ml admin API. The server embedder loads by HF name (AutoImageProcessor/Model,
model-agnostic), preferring the pre-downloaded local dir for the default so
existing deploys don't re-download; rebuilds on a name change.
- agent: new 'embed' job = whole-image SigLIP embedding (mean-pool video frames)
under the lease-announced model → POST /jobs/submit_embedding writes
image_record.siglip_embedding + siglip_model_version. The lease now announces
the model FROM THE SETTING (not a constant).
- re-embed routing: enqueue_gpu_backfill('embed') selects unembedded + stale-
version images; 'siglip' now re-embeds concept crops whose version != current
(so a swap re-triggers crops, not just the never-embedded back-catalogue). The
CPU ml-worker backfill no longer re-embeds on a version mismatch (it can't
churn the library at 512px) — the GPU agent owns version re-embeds. Daily
'embed' + 'siglip' beats self-heal.
- scoring: score_image only bags embeddings in the CURRENT model's space (whole-
image gated by siglip_model_version, concept regions by embedding_version) so a
mid-swap stale vector isn't scored by new-space heads; legacy NULL = current.
- UI: GpuAgentCard "Embedding model (advanced)" — edit name/version, Save, and
"Re-embed library (GPU)" (queues embed + siglip); points at SigLIP 2.
Tests: lease announces model + submit_embedding round-trip; enqueue 'embed'
selects stale/unembedded; stale-version excluded from scoring; embedder model
settable + empty rejected; siglip gate updated to current-version concept.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ttrj5P7upUTueSfoJcxEqa
|
||
|
bvandeusenandClaude Opus 4.8
|
b91a230f12 |
feat(ccip): automation + reference quality — keep identity flowing hands-free (#114)
Works through the optional CCIP ideas + the "keep moving even if I forget" ask:
AUTOMATION (no button needed):
- Hourly beat auto-enqueues CCIP backfill — new images get embedded (and errored
ones retried) on their own; the queue never goes idle waiting for a click.
- CCIP auto-apply: a daily sweep tags confident matches (source='ccip_auto') so
identity tags keep flowing. ON by default (opt-out, like head auto-apply);
ml_settings.ccip_auto_apply_enabled + _threshold (0.92, above the suggest cut),
migration 0064. Vectorized (one matmul + reduceat per image), reversible, skips
already-applied/rejected. Switch + threshold in the GPU agent card; GET/PATCH
/api/ml/settings; auto_applied count in /api/ccip/overview.
REFERENCE QUALITY (the over-fire root cause):
- character_references now draws ONLY from single-character images — on a
multi-character image the tag is image-level, so every figure would otherwise
pollute each character's prototypes (a 2-char image tagged 'Velma' made
Daphne's figure a Velma reference). This is the contamination behind residual
over-firing.
- Cached on a cheap signature (char-tag count + ccip-region count/max-id) so the
reference load isn't redone on every modal open.
Tests: multi-character image not used as a reference; auto-apply tags a confident
match as ccip_auto.
NEXT (not done, confirmed): comic-panel cropping + SigLIP concept crops ("spot
interesting content").
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ttrj5P7upUTueSfoJcxEqa
|
||
|
bvandeusenandClaude Opus 4.8
|
625336b6b4 |
feat(ccip): tunable match threshold, default 0.85 (#114)
Live data showed the v1 flat 0.75 cosine over-fired — ~64% of matched images got
3-10 character guesses dominated by the most-referenced characters (a 27-ref
character clears a low bar on many images). A sweep showed 0.85 collapses the
noise (noisy multi-matches 47→3) while keeping the confident single-character
matches.
- ml_settings.ccip_match_threshold (migration 0063, default 0.85); match_image
reads it (override still accepted). DEFAULT_SIM_THRESHOLD fallback 0.75→0.85.
- Exposed in GET/PATCH /api/ml/settings (validated 0.5–0.999).
- Slider in the GPU agent card ("Character-match strictness") — tune live, no
redeploy, same observe-and-tune loop as auto-apply.
Test: a ~0.9-cosine figure matches at 0.85, dropped at 0.95.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ttrj5P7upUTueSfoJcxEqa
|
||
|
bvandeusenandClaude Opus 4.8
|
48c8811d69 |
feat(heads): auto-apply observability + on by default (#114 auto-apply B)
Auto-apply is now ON by default (operator-asked: opt-OUT, not opt-in) — migration 0059 + model default flipped. The support (>=30) + measured-precision gates keep it safe and every auto-tag is reversible. Observability so the operator can tune from real data: - MISFIRE = an auto-applied (source='head_auto') tag the operator later removes. UNDER-FIRE = a tag with a head the operator adds by hand (the head missed it). Both captured at correction time in TagService.add_to_image/remove_from_image (source is lost on delete) into durable per-tag counters (head_metric), keyed by tag so they survive head retrain/prune. - Daily snapshot_head_metrics writes a per-concept time-series point (head_metrics_snapshot): auto-applied volume + cumulative misfires/under-fires + head quality; 180-day retention; daily beat. - GET /api/heads/metrics: per-concept current counts + realized misfire rate + head quality, plus the snapshot time-series — the report to tune the precision target + support floor. Migration 0060. Tests: misfire/under-fire counting (and the negatives — manual removal isn't a misfire, headless manual add isn't an under-fire), snapshot time-series, metrics API. What's the autofire threshold? There's no single number — each graduated head derives its OWN probability cutoff from its PR curve: the operating point that holds precision >= head_auto_apply_precision (0.97) at max recall. The global knobs are that target + the >=30 support floor. NEXT (slice 3): UI — enable toggle, dry-run preview, per-concept trends. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ttrj5P7upUTueSfoJcxEqa |
||
|
bvandeusenandClaude Opus 4.8
|
74fef908d2 |
feat(heads): earned auto-apply — sweep mechanism, off by default (#114 auto-apply A)
Graduated heads can now apply their tag without a human — gated so it's safe:
- FIRING GATE: a head fires only when the master switch (head_auto_apply_enabled,
default OFF) is on AND it has >= head_auto_apply_min_positives (default 30)
clean labels. A precise-looking but under-supported low-N head can't spray tags.
- auto_apply_sweep (heads.py): streams every embedded image in chunks, scores
against the eligible heads (numpy, no sklearn), applies each head's tag where
score >= its auto_apply_threshold and the tag isn't already applied/rejected,
with source='head_auto' (distinguishable + reversible). dry_run counts only.
- HeadAutoApplyRun (migration 0059) tracks each sweep / preview; apply_head_tags
task (ml queue) + scheduled_apply_head_tags daily beat (no-op unless enabled)
+ recovery sweep + retention(20).
- API: POST /api/heads/auto-apply {dry_run} (202 / 409 running / 400 disabled),
GET /api/heads/auto-apply (recent runs + per-concept report). Settings
head_auto_apply_enabled + min_positives via /api/ml/settings.
Tests: sweep applies above threshold, dry-run writes nothing, skips under-
supported + ungraduated heads; API disabled/dry-run/conflict guards.
NEXT (slice 2): the observability the operator asked for — per-concept misfire
(auto-applied-then-removed) + under-fire tracking, time-series snapshots, and a
reporting API to tune. Slice 3: the UI (enable, preview, trends).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ttrj5P7upUTueSfoJcxEqa
|
||
|
bvandeusenandClaude Opus 4.8
|
22c3b54746 |
feat(heads): production per-concept heads — train + score backend (#114 A)
The eval (#1130) proved the frozen-embedding + trained-head spine; this lands its production form (the first of three slices that make heads the suggestion source, replacing Camie + centroid). - tag_head: one logistic-regression head per general/character concept with enough labelled positives. Weights (pgvector), honest CV-derived suggest threshold + earned-auto-apply point, and per-concept quality metrics. - head_training_run: persisted batch lifecycle (mirrors tag_eval_run) so the admin card shows live + historical status across navigation. - services/ml/heads.py: TRAIN (sync, ml worker, reuses tag_eval's proven data loaders + metric math so production heads match measured eval numbers) and SCORE (async, API worker — numpy via pgvector, no scikit-learn): score one image's embedding against all heads → the rail's suggestions, cached on (count, max trained_at) so a retrain invalidates without per-request loads. - tasks.ml.train_heads (ml queue, commits per head so a kill leaves progress) + recover_stalled_head_training_runs sweep + retention(20) + 5-min beat (rule 89). - api/heads.py: POST /api/heads/train (one run at a time, 409 guard) + GET /api/heads (count, graduated, last-trained, running, per-concept table, recent runs). - ml_settings: head_min_positives + head_auto_apply_precision, tunable via /api/ml/settings. Scoring isn't wired into the rail yet (slice C) and the admin UI is slice B — this slice makes training + scoring exist and CI-verifiable. 'precision' column stored as precision_cv (SQL reserved word). Migration 0058. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ttrj5P7upUTueSfoJcxEqa |
||
|
bvandeusenandClaude Opus 4.8
|
369e3de684 |
feat(ml): cadence-based video frame sampling + min-frame tag aggregation (#747)
Video tag noise root cause: frames were a FIXED count (6) max-pooled — a tag firing on one frame survived at peak confidence, and a fixed count under-samples long multi-scene videos so real scene-local tags looked like noise. Redesign (operator-steered): - Sample at a fixed CADENCE — one frame every `video_frame_interval_seconds` (default 4) across the 5–95% window — so a tag's frame-presence reflects real screen time independent of video length. Capped at `video_max_frames` (default 64): a long video stretches the spacing instead of exploding into hundreds of inferences, bounding per-video cost on the single ml-worker (per-frame ffmpeg timeout also cut 60s→30s). - Aggregate with `_aggregate_video_predictions`: keep a tag only if it appears in >= `video_min_tag_frames` sampled frames (≈ that many × interval seconds on screen — duration-independent noise rejection), with confidence = MEAN over the frames it appears in (not max). Clamps the threshold to the sample count so a 1–2-frame short video still tags. - All three knobs are DB-backed ml_settings (migration 0053), patchable via /api/ml/settings + sliders in the ML settings card — replaces the VIDEO_ML_FRAMES env var (product-not-project). Tests: aggregation drops one-frame noise + means corroborated tags + clamps on short videos; settings round-trip + min>max validation. Replaced the _maxpool_predictions unit test. NOTE: this is the QUALITY half of #747. The perf half — the ml-worker runs CPU-only — is GPU enablement, tracked separately in #872. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> |
||
|
bvandeusenandClaude Opus 4.8
|
3f92669f12 |
feat(ml): DB-backed tagger_store_floor (default 0.70), the ingest confidence floor
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> |
||
|
bvandeusen
|
1fd594baaf |
chore(ml): suggestion_threshold default 0.50 → 0.70
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. |
||
|
bvandeusen
|
af7b5c95e9 |
feat(modal): autofocus tag input, expand general suggestions, retire copyright/artist categories
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. |
||
|
bvandeusenandClaude Opus 4.7
|
0a4eb0bdc0 |
fix(fc2b): bare CHECK constraint names (naming-convention double-prefix)
Base.metadata's convention applies ck_%(table_name)s_%(constraint_name)s.
ml_settings and tag_allowlist passed already-prefixed names
(ck_ml_settings_singleton / ck_tag_allowlist_confidence_range), so the
ORM-side names came out doubled (ck_ml_settings_ck_ml_settings_singleton
etc.) and the migration-0003 smoke tests failed.
Same class of bug fixed in FC-2a for ImportSettings — should have applied
that lesson here. Bare names ('singleton', 'confidence_range') let the
convention produce the final names that match migration 0003's literal
DDL. Migration unchanged; only the model __table_args__.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
||
|
bvandeusenandClaude Opus 4.7
|
906804140c |
feat(fc2b): schema migration 0003 — ML pipeline tables
Renames image_record.wd14_* -> tagger_* (we're on Camie now, not WD14). Adds tag_allowlist (auto-apply opt-in, per-tag confidence), tag_suggestion_rejection (per-image dismissals), tag_alias (composite (string, category) -> canonical tag, resolved at read time), tag_reference_embedding (per-tag SigLIP centroids), and the ml_settings singleton (per-category + centroid thresholds, model version pins). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> |