Support RapidOCR with custom OCR models

[maksimkurb]

There is a page in docling documentation that describes how to configure RapidOCR to use custom OCR models.

I have tried to figure out how to configure it in docling-serve specifically because I want to use it with multilanguage model as API for document OCR pre-processing, but haven't succeeded.

Is there any way to pass custom model paths for RapidOCROptions in docling-serve without forking this project?

[dosubot]

Hi @maksimkurb! I'm Dosu and I’m helping the docling-serve team.

You can use custom RapidOCR models in docling-serve without forking the project by mounting a directory containing your models and setting the environment variable DOCLING_SERVE_ARTIFACTS_PATH to point to that directory. This makes your custom models available to the API for OCR pre-processing, including multilanguage models. For example, if you are running docling-serve in Docker:

docker run -p 5001:5001 \
  -v /path/to/your/models:/opt/app-root/src/models \
  -e DOCLING_SERVE_ARTIFACTS_PATH="/opt/app-root/src/models" \
  quay.io/docling-project/docling-serve

This approach works for both local and production/cluster deployments (e.g., with a PersistentVolumeClaim in Kubernetes) and does not require any code changes or forking. Just ensure your models are in the correct directory structure expected by RapidOCR and docling-serve. If a required model is missing, docling-serve will raise a runtime error, so double-check the path and contents [source] [source].

Currently, docling-serve does not support passing custom model paths for RapidOCROptions per request or per pipeline via the API; the model directory is set globally via the environment variable. For more advanced customization (like plugging in a new backend), you can use the plugin factory system described in the Docling plugin documentation [source].

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}