Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
71 lines (45 loc) · 2.42 KB

CONTRIBUTING.md

File metadata and controls

71 lines (45 loc) · 2.42 KB
Raw

Contributing to OpenCR

Thanks for considering a contribution.

OpenCR exists to make Turkish-language and archival OCR easy to run, share, and improve — every PR, issue, and dataset published with OpenCR helps that goal.

Ways to help

  • File issues. Found a bug, a layout that OpenCR struggles with, or a confusing piece of docs? Open an issue with a short PDF (or page screenshot) we can reproduce against.

  • Add Turkish-language test fixtures. A small public-domain PDF + the expected text is one of the highest-leverage contributions.

  • Benchmarks. Compare OpenCR against Tesseract / Surya / PaddleOCR / Marker on a Turkish corpus and post the table — even informal numbers are useful.

  • Deployment recipes. vLLM, hosted GPU endpoints, and reproducible benchmark environments that improve throughput or quality.

  • Translations. README and dataset cards in additional languages.

Setup

git clone https://github.com/cdliai/opencr.git
cd opencr
make install
make test

make run starts a local dev server on http://localhost:39672 and points it at MODEL_SERVER_URL (default: http://localhost:39671). Start the bundled GPU stack with docker compose up -d, or provide another OpenAI-compatible GPU endpoint.

Code style

  • Python: keep it boring and explicit. Type hints on public functions. No new dependencies without a brief rationale in the PR.

  • Frontend: stays Alpine + plain CSS until the state model genuinely outgrows it. No build step, no framework rewrite.

  • Tests: every new code path should have a unit or integration test. We use pytest and pytest-asyncio.

Pull request flow

  1. Open an issue first for non-trivial changes — a 5-line discussion saves a 500-line rewrite.

  2. Branch from main, name it feat/... or fix/....

  3. Run make lint test before pushing.

  4. PR description: what changed, what it fixes, how to verify locally.

Reporting OCR-quality regressions

If a particular PDF regresses after a change, please attach (or link to a public copy of) the PDF, the page number, what OpenCR produced, and what was expected. Quality bugs without a reproducer are very hard to act on.

Code of conduct

Be respectful. We're a small project trying to do useful work for Turkish-language NLP — no room for harassment or discrimination here.

License

By submitting a PR, you agree your contribution is licensed under the project's Apache 2.0 License.