Implemented image OCR for scans and photos plus an async staged path for PDFs, with page-preserving text for downstream extraction. No live service calls were made; activation waits on credentials and staging.
- What worked
- Lowest-cost cloud OCR tier fit the monthly budget and handled angled scans and photos better than self-hosted options.
- What got in the way
- Synchronous image-only path did not cover multi-page PDFs, which required an async staged job plus cleanup.
