Last consumer of the on-disk thumbnail pipeline was the vision
worker reading /data/thumbs/{id}/medium.webp. Now it asks Nextcloud
for a 640px preview (the same edge size the old thumb used) and
decodes the bytes in-memory — no disk dependency.
- nextcloud_dav.get_preview_bytes: sync sibling of get_preview_async,
for the celery vision worker (which is sync).
- vision._load_thumb: tries NC preview first; transitional disk
fallback stays for rows still indexed during the rollout.
- thumbs.WORKER_THUMB_SIZES = set() — generate_thumbnails still runs
the decode + pHash side-effect (perceptual dedup is mule-only and
needs original-resolution pixels) but no longer writes thumbnail
files.
The HTTP thumbnail endpoint's disk fallback path stays in place
unchanged: for NC-404 cases (e.g. iPhone JPEGs mis-extensioned as
.DNG), inline Pillow regeneration still writes a tiny per-photo
file so subsequent requests are fast. That path is rare and the
files are small.
Disk impact: /data/thumbs currently has ~22k medium.webp totaling
~1 GB. They'll stop being read after the worker-vision container
restarts, but no automatic delete — purge with the same find
pattern used for small/large reclaim when ready:
find /data/thumbs -name "medium.webp" -delete
Drops face recognition, OCR, object detection, and semantic embeddings.
The sole remaining vision task is a CLIP-based binary classifier
(photography vs other); photos in "other" get needs_review=true so
screenshots, documents, memes and scans can be triaged from a new
filter pill in the UI.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds Redis-backed feature flags for vision stages with admin UI toggles
and manual backfill trigger, photo upload and download routers with
frontend upload modal, and rawpy-based RAW decoding with JPEG fallback
for misnamed DNGs. Fixes pgvector serialization, is_trashed filter, and
naive-datetime bind in incremental duplicate regrouping; bumps Celery
time limits on regroup tasks beyond the 5-minute default.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Sharing:
- New HeapShare and FolderShare models with read/write permissions
- Sharing API router (CRUD for heap and folder shares)
- Heap endpoints accept shared access (photo_ids, add/remove with write)
- Photo list drops user_id filter in shared context, adds owner_username
- Media serving (thumb/original/proxy) falls back to share check on 404
- ShareDialog component for managing shares from kebab menus
- HeapsPanel shows "Shared with me" section for shared heaps
- LeftSidebar shows "Shared with me" section for shared folders
- Owner badge on PhotoThumbnail for photos from other users
Auth:
- Access token default bumped to 1 year, refresh to 10 years
- Refresh token persisted in localStorage (survives page reload)
- Timer-based refresh replaced with 401 axios interceptor
Vision pipeline fixes:
- Bootstrap sets Redis ready key even on partial export failure
- Export functions run conditionally (only for actually missing models)
- _load_thumb handles multi-user path (/data/thumbs/{user_id}/{photo_id}/)
- can_access_photo_via_share uses single subquery instead of N+1 loop
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Addresses 16 robustness, transparency, and performance issues across
the Celery media processing pipeline:
Critical:
- Singleton DB engine in vision tasks (was leaking one per task call)
- acks_late + task_reject_on_worker_lost so crashed workers don't lose tasks
- Global soft/hard time limits (5/10 min) to prevent hung worker slots
- Thumbnail copy-before-resize (in-place mutation degraded larger sizes)
- backfill_vision now checks each task type independently (OCR, faces, etc.)
- Parameterized LIMIT in backfill_vision (was f-string SQL injection)
High:
- try/except + retry(max=3) on all vision inference tasks
- extract_metadata writes processing_error on exiftool failure
- PIL Image handles closed in _load_thumb/_load_original
- Scan progress Redis keys auto-expire after 1 hour
- Watcher lock renewal is wall-clock based (30s) not event-count based
- worker_process_init signal warms up vision models on startup
Medium:
- Explicit task_routes for every task name (wildcards never matched)
- app.services.metadata added to Celery include list
- POST /maintenance/recover-stuck endpoint for photos stuck in processing
- Docker healthchecks for worker-light, worker-vision, and Redis
- Task ID in vision log lines for distributed tracing
- Bare except:pass narrowed to specific exceptions
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- detect_objects, classify_content, recluster_faces now look up the
photo's user_id and set it on created Tag rows — fixes tags being
invisible to the owning user due to NULL user_id
- Initial admin setup creates source root at the mount root (/photos)
instead of a subdirectory, since the admin owns the entire library
- Revert to OpenCLIP ViT-B/32 (512-d) as default embedder — SigLIP
requires transformers version alignment not yet available in the
Docker image. SigLIP2 code remains for future enablement.
- Add transformers to requirements for future SigLIP support
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Scan now automatically queues backfill_vision (+90s) and recluster_faces
(+300s) after dispatching folder scans. Face extraction also schedules a
debounced recluster via Redis so incremental file-watcher imports get
clustered without manual intervention.
The ScanProgress widget now tracks worker queue activity beyond the scan
phase, showing a "Processing Photos" indicator with vision queue counts
while background tasks (embeddings, faces, tags, OCR) are running.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Three overlapping fixes so the ingestion pipeline actually runs and the
user can see what it's doing:
Pipeline recovery
- app/database.py: use NullPool when MULITA_CELERY_WORKER=1 so each
Celery task opens a fresh asyncpg connection on its own event loop.
Fixes "another operation in progress" and "Future attached to a
different loop" errors that were dropping ~every thumbnail +
extract_metadata task on the floor.
- app/tasks/thumbs.py: initialize photo=None before the try and rollback
on error so a transport failure in the initial SELECT doesn't raise
UnboundLocalError in the except block and leak rows stuck in 'pending'.
- app/services/vision/bootstrap_models.py: on missing model files,
invoke export_models automatically instead of just warning. First
boot of a fresh install now self-heals.
- app/services/vision/export_models.py: shutil.move instead of
Path.rename so the YOLO export survives the /app → /data/models
cross-volume hop.
- requirements.txt: add ultralytics so export works in a stock image.
Worker topology
- docker-compose.yml: replace the single worker with worker-light
(default/high/low queues, c=2, IO-bound) and worker-vision (vision
queue, c=5, OMP_NUM_THREADS=1 to avoid oversubscription on 6 cores).
Vision is pinned to ≤5 parallel inferences so ONNX doesn't each
spawn an all-cores intra-op pool.
- .env / .env.example: CELERYD_CONCURRENCY replaced with
CELERY_LIGHT_CONCURRENCY + CELERY_VISION_CONCURRENCY.
- Backfill queries in thumbs / scan / vision now ORDER BY taken_at
DESC NULLS LAST so newest photos finish first — the library fills
in top-down in the UI instead of arbitrary insertion order.
Settings visibility
- routers/library.py: new GET /maintenance/pipeline-stats returning
done/total per stage (thumbnails, exif, gps, phash, embeddings,
tags, ocr, faces, face clusters, duplicate groups). Worker-status
now also reports the `vision` queue depth, which was missing.
- services/api.ts: PipelineStats / PipelineStage / ScanStatus types
and the matching client call.
- components/dialogs/SettingsDialog.tsx:
- new Pipeline Progress card with one progress bar per stage
- inline scan banner (processed/total/current folder) inside the
Library section while a scan is running
- Tasks/min throughput computed by diffing worker processed counters
between polls
- Workers section calls out the vision queue and documents the
CELERY_LIGHT/VISION_CONCURRENCY + docker compose up -d scale path
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Cross-reference face detections with YOLO 'person' bounding boxes —
only keep faces that overlap >= 50% with a detected human body. This
eliminates false positives on dogs, paintings, and cartoons without
needing an aggressive score threshold.
Lower face detection threshold back to 0.6 since the person-overlap
check is now the primary precision filter.
Tested: 6 verified faces from 4 photos, zero false positives.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- recluster_faces now writes photo_tags rows for each face cluster so
the tag count and tag_ids filter work (previously count was always 0)
- Old cluster tags and photo_tags are cleaned up before re-clustering
- Raise face detection threshold from 0.4 to 0.7 to reduce false
positives (was detecting dog faces as people)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Rewrite faces.py to use cv2.FaceDetectorYN instead of raw ONNX
(handles multi-scale anchor decoding and NMS internally)
- Load original photo files at up to 4000px for face detection instead
of 240px thumbnails — faces were too small to detect at thumbnail res
- Falls back to thumbnail if original is unavailable
Tested: 33 faces extracted from 13 photos, clustered into 1 person.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Rewrite all vision tasks to use sync psycopg2 sessions instead of
asyncpg — fixes 'another operation in progress' and event loop errors
when Celery forks workers sharing the async connection pool
- Letterbox-pad images to exactly 640x640 for YuNet face detector
(was crashing on non-square thumbnails)
- Deduplicate object detections per label per photo — keep highest
confidence only to avoid photo_tags PK violation on multiple
detections of the same class
- Add all queues (-Q default,high,low,vision) to worker command
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Create face_embeddings table with pgvector Vector(128) + HNSW index
- Implement extract_faces task (YuNet detection + SFace recognition)
- Implement recluster_faces task (DBSCAN clustering → Tag(kind=face_cluster))
- Clusters are named "Person N" and get representative_photo_id
- cluster_id FK → tags.id, SET NULL on delete for merge/rename support
Migration 0005 creates the face_embeddings table.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Implement detect_objects Celery task:
- Runs YOLOv8n on 640px thumbnail via ONNX Runtime
- Creates Tag(kind=object) rows for each COCO class detected
- Writes photo_tags associations with confidence, bbox, and source
- Wipes previous detections per source model on re-run
No new tables/migrations — uses the unified Tag model from PR3.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Create ocr_text table for storing per-region OCR results
- Add tsvector search_vector column to photos with GIN index and
auto-update trigger on filename/user_title/user_notes
- Implement ocr_photo Celery task using rapidocr-onnxruntime
- Add FTS leg to hybrid search: queries photos.search_vector and
ocr_text via UNION, fused with semantic results via RRF (k=60)
Migration 0004 backfills search_vector for existing rows.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Wire the full embedding flow:
- Rewrite Embedding model to use pgvector Vector(512) with HNSW index
- Add embed_photo, vision_fanout, backfill_vision Celery tasks on
dedicated `vision` queue
- Hook vision_fanout into generate_thumbnails completion
- Add POST /api/v1/photos/search with hybrid RRF ranking (semantic-only
for now; FTS leg added in PR5)
- Stub ocr_photo, detect_objects, extract_faces tasks for later PRs
Migration 0003 drops/recreates the embeddings table (was never populated).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
