OCR-D extension to recognize text using open models via OpenAI interface

ocrd_llm is an OCR‑D module that tries to recognize text using open models via OpenAI interface. Currently tested with Academic Cloud.

Disclaimer: Work in progeress (maybe implement layout recognition in the future, too)

Installation

pip install .

Or install via Docker:

- docker compose build
- docker-compose run ocrd-llm

Quick Start

ocrd-llm -I OCR-D-SEG-REGION-LINES -O OCR-D-SEG-REGION-LINES-RECOGNIZED-LLM --overwrite -p '{"model_id":"qwen3.5-35b-a3b", "api_endpoint": "", "api_key": ""}'