Heat code extractor v6: EasyOCR fallback for codes Tesseract can't read

Heat code extractor v5: targeted cell-region OCR + aggressive false-positive filtering

Heat code extractor v4: kill A370-12a, N021-1 truncation, YYYYMMDD dates

Heat code extractor v3: capture all 6 KB419-1 codes; reject OCR-mangled English; normalize O→0

Tighten heat_code extractor: reject standards numbers + material specs

Phase 5: split scheduler into standalone process; enable uvicorn workers

Phase 2+4: PO/Invoice client folder routing and subject-line MTR trigger

Phase 1: heat code extraction and multi-folder MTR routing

Replace "tie" logic with "ambiguous" logic so that daily brief manual reviews happen when there is more than 1 doc type that meets confidence threshold

More classification keywords

Verbose logging to debug classification and new classification keywords and weights

Update classification logic for purchase orders

Do not overshare information on the /health endpoint

Error handling, confidence scoring, daily brief, deduplicate webhook notifications

Verbose Graph errors, Fix Graph endpoints, Log webhook activity

added dropbox, OCR routing logic, graph connection fix

initial commit