Headless version of RapidOCR for containerized environments

[jfloe]

first of all, thank you for your great work on RapidOCR — it is a very powerful and useful project.

We are currently integrating RapidOCR into a containerized production pipeline. In this environment, there is no need for any GUI-related functionality, and we also do not have access to system-level graphical libraries.

We have run into an issue where one of RapidOCR’s dependencies ultimately pulls in opencv-python, which requires graphical system libraries. This causes installation and runtime complications in minimal container images.

While there is already a well-known alternative (opencv-python-headless) that works perfectly in such environments, we noticed that the dependency chain currently enforces opencv-python as a hard requirement via a subdependency (in our case through Rapidover).

Because package resolution tools like uv strictly resolve and lock dependencies, it is not possible for us to override this cleanly without hacks or forking dependencies.

Request

It would be extremely helpful if a headless variant of RapidOCR (or its relevant dependency chain) could be provided, where:

opencv-python-headless is used instead of opencv-python
No GUI/system display dependencies are required
The package remains fully compatible with containerized and server-side environments

This would make adoption in production systems, CI pipelines, and cloud-native deployments significantly easier.

We believe many users running OCR workloads in Docker/Kubernetes or other headless environments would benefit from this greatly.

Thank you very much for considering this request, and for your continued work on RapidOCR!

Best regards