Thirdshop/Kazier
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(api): OCR réel — POST /ocr/jobs + GET /ocr/jobs/:id (tesseract syscall, jobs asynchrones)

Browse Source 
- ocr/tesseract.go : Engine → Tesseract subprocess (OCR_LANG fra+eng) ; PDF → calque texte via ledongthuc/pdf (go.mod : nouveau dep)
- repository/ocr_jobs.go : queued→processing→done/failed, scoping device, text/error NULLIFés
- service/ocr.go : Create valide le fichier (GetFile), queue + goroutine de traitement ; physique résolu par glob UPLOAD_DIR/<device>/<id>.* ; fail propre (fichier illisible, erreur moteur)
- handlers/ocr.go : réels (validate fileId, NOT_FOUND si job d'un autre device), fin 501 OCR
- tests end-to-end : cycle queued→done (stub moteur), NOT_FOUND fichier inconnu, scoping device
- smoke réel : PNG 'VAULTDROP' → job done text='VAULTDROP' (tesseract installé)
- docs/AGENTS : §5 + état des routes (tout V1 réel)
This commit is contained in:
m
2026-09-10 19:54:16 +02:00
parent 5d96814853
commit d5886c678b
12 changed files with 452 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
AGENTS.md
+2 -2
View File
@@ -34,8 +34,8 @@ cd mobile && npm run test:db
- `config/` — env (`godotenv`, optionnel) + defaults: `PORT`, `DATABASE_URL`, `UPLOAD_DIR`, `MAX_FILE_SIZE_MB`, `OCR_LANG`, secret paseto
- `models/` — domain entities (users, devices, documents/resources, clients)
- `service/` — business logic (permissions, upload, create folder, move, **sync outbox + snapshot**)
- `handlers/` — HTTP handlers (health, devices register + paseto, files CRUD/upload/search, folders, **sync/ops + sync/permissions** — réels ; OCR stubs 501)
- `repository/` — Postgres persistence réelle (`repository.Resources` : insert/list/get/soft-delete scoping `owner_id`, **search, move, rename, root-name unique index**, `repository.Devices.Upsert`, `repository.Operations` : trace outbox idempotente `(device_id, operation_id)`, `ListOwned` pour le snapshot) ; IDs sont TEXT 32-hex, `NewID()` = `crypto/rand` 16 octets hex (jamais UUID conversion, cf. `docs/api-v1.md`)
- `handlers/` — HTTP handlers (health, devices register + paseto, files CRUD/upload/search, folders, **sync/ops + sync/permissions, ocr/jobs** — réels)
- `repository/` — Postgres persistence réelle (`repository.Resources` : insert/list/get/soft-delete scoping `owner_id`, **search, move, rename, root-name unique index**, `repository.Devices.Upsert`, `repository.Operations` : trace outbox idempotente `(device_id, operation_id)`, `ListOwned` pour le snapshot, `repository.OcrJobs` : jobs queued→processing→done/failed) ; IDs sont TEXT 32-hex, `NewID()` = `crypto/rand` 16 octets hex (jamais UUID conversion, cf. `docs/api-v1.md`)
- `db/` — package migrations (`golang-migrate/v4`, embarquées via `embed` dans `db/migrations/*.sql`) : `db.MigrateDatabase(url)` au boot du serveur ; test harness `db/migrations_test.go` (up → assertions schéma → down, `TEST_DATABASE_URL`, skip si PG indisponible) ; `dbtest/` — helper cross-package pour les tests repo/handlers (crée la DB test si absente, reset schema, migrate ; skip si PG down)
- `ocr/` — OCR engine behind an interface (Tesseract system call, `OCR_LANG` défaut `fra+eng`)
- Response helpers: `pkg/api/response.go`