text-extract-api

Why we included this project

Teams that regularly turn invoices, medical reports, or office files into clean, structured data will find this a genuinely useful tool rather than a demo. It runs a complete OCR and parsing pipeline locally: EasyOCR and vision models do the heavy lifting on images and PDFs, then an Ollama-backed LLM converts the raw text into Markdown or JSON and can even fix OCR spelling errors. Because everything ships in docker-compose with no cloud dependencies, sensitive documents never have to leave your own infrastructure, and the same pipeline can scrub personally identifiable information before it is stored. The FastAPI service with Celery queues and Redis caching lets you push large batches of documents through it and scale worker processes independently. If you need a self-hosted document-to-data API you can wire into existing backends, this is a strong starting point.