809823a8ae
- embeddings: corpus ca matrice numpy cu norme precalculate; suggest_nearest
= un matvec (~0.6ms/query fata de ~500ms cosine pur-Python la 17k vectori)
- enrich_suggestions: vot ponderat cu similaritatea pe top-5 vecini (NUL =
eticheta proprie); prag 0.5 -> 0.88, calibrat LOO pe corpusul SILVER
(tools/mapare-llm/knn_calibrate.py): precizie 90.5% -> 93.1%, cod gresit
preselectat 7.2% -> 4.7%; sub prag abtinere -> preselectie fuzzy
- UI: codul sugerat de sistem afisat explicit cu sursa si scorul, separat de
lista fuzzy ("potrivire text"); indicator de progres reparat pe upload
(display:inline anula .htmx-indicator) si adaugat pe pasii 2->3 si
"Salveaza maparile"
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
30 lines
1.2 KiB
Plaintext
30 lines
1.2 KiB
Plaintext
# Gateway RAR AUTOPASS — dependinte runtime
|
|
# Versiuni aliniate la ce e instalat in container (2026-06-15).
|
|
fastapi==0.115.0
|
|
uvicorn[standard]==0.30.0
|
|
httpx==0.27.*
|
|
jinja2==3.1.*
|
|
pydantic==2.8.2
|
|
pydantic-settings==2.*
|
|
python-multipart==0.0.*
|
|
# SessionMiddleware (app/main.py) semneaza cookie-urile de sesiune via itsdangerous.
|
|
# Necesar explicit: nu e tras tranzitiv in imaginea Docker, doar in dev local.
|
|
itsdangerous==2.2.0
|
|
# Fuzzy lookup pentru editorul de mapari operatii (app/mapping.py). Pur Python/C, fara build extern.
|
|
rapidfuzz==3.14.5
|
|
# Criptare creds RAR efemere in submissions (app/crypto.py, Fernet). Zero-storage at rest.
|
|
cryptography==46.0.5
|
|
|
|
# Parsare xlsx/xls pentru import fisiere (Treapta 2, Issue 4 — PINNED).
|
|
openpyxl==3.1.5
|
|
|
|
# Migrare DBF (tools/import_dbf.py). Necesar doar pentru import optional, nu pentru runtime.
|
|
dbfread==2.0.7
|
|
|
|
# Embeddings in-proces pentru sugestie cod RAR (L14-S4, PRD 5.14 Decision #16).
|
|
# Model multilingv (~230MB pe disc, ONNX quantizat), fastembed fara torch, lazy-load la runtime.
|
|
# Degradare gratioasa daca lipseste la runtime (is_available()=False, suggest_nearest=[]).
|
|
fastembed>=0.8.0
|
|
# Folosit direct la cautarea NN (matvec pe corpus); oricum tras de fastembed/onnxruntime.
|
|
numpy>=1.26
|