docu-ocr-engine 0.1.0

Structured OCR extraction engine

docu-ocr-engine (Self-hosted OCR Engine)

A backend-agnostic, self-hosted OCR engine packaged in Python and shipped as a Docker container. It exposes:

• HTTP API (FastAPI) for Laravel/Node/any backend
• CLI for batch/offline OCR
• Pluggable OCR engines (Tesseract by default) with an interface for future engines

This repo is production-oriented but intentionally minimal so you can extend it.

Features

• Document processing endpoint (/v1/process) accepting:
  • local file upload (multipart)
  • or remote URL fetch
• Status tracking (/v1/jobs/{job_id}) with in-memory store (swap to Redis/Postgres easily)
• Retry-friendly design: idempotency key support
• OCR engine abstraction (BaseOcrEngine)
• Output as structured JSON (text + basic metadata)
• Dockerized for easy self-hosting

Quickstart (Docker)

docker build -t docu-ocr-engine:local -f docker/Dockerfile .
docker run --rm -p 8000:8000 docu-ocr-engine:local

Open:

• http://localhost:8000/health
• http://localhost:8000/docs

Quickstart (Local dev)

python -m venv .venv
# Windows: .venv\Scripts\activate
# Linux/macOS: source .venv/bin/activate
pip install -r requirements.txt
uvicorn docu_ocr.api.http:app --host 0.0.0.0 --port 8000 --reload

CLI

python -m docu_ocr.cli.main process --file samples/sample.png --out /tmp/result.json

API usage examples

cURL (upload)

curl -X POST "http://localhost:8000/v1/process" \
  -H "X-Idempotency-Key: demo-123" \
  -F "file=@samples/sample.png" \
  -F "language=eng"

cURL (status)

curl "http://localhost:8000/v1/jobs/<job_id>"

Notes

• Default OCR engine is Tesseract. The Docker image includes tesseract-ocr.
• For PDFs, the service uses pdftoppm (poppler-utils) to render pages to images.

Extending

• Add new engines in docu_ocr/ocr/ and wire via DOCU_OCR_ENGINE=<name>
• Replace InMemoryJobStore with Redis/Postgres
• Add webhooks from core/pipeline.py after completion

License

MIT

ocr-medz