CONTRIBUTING.md

Contributing to Safetensors Studio & Bench

Thank you for considering contributing! This project is open-source (MIT) and welcomes developers and ML engineers.

Code of Conduct

Please read and follow our CODE_OF_CONDUCT.md.

How to Contribute

1. Setup Development Environment

git clone https://github.com/your-org/safetensors-studio.git
cd safetensors-studio
# Backend
cd backend && pip install -r requirements.txt && uvicorn app.main:app --reload
# Frontend
cd ../frontend && npm install && npm run dev

2. Branching

• main is protected. Create a feature branch: git checkout -b feat/your-feature
• Follow conventional commits: feat:, fix:, docs:, perf:

3. Areas to Contribute

• New Benchmark Suites: Add tasks to backend/app/benchmark.py:14 (keep id, name, category, prompt, expected, expected_regex, max_tokens)
• Languages: Extend SUPPORTED_LANGUAGES.md and dataset_parser.py:detect_language()
• Quant Backends: model_manager.py:26 (e.g., AWQ, GPTQ)
• Frontend: frontend/src/components/** (Tailwind + Shadcn)
• Docs: docs/ in English (required)

4. Pull Request Checklist

• npm run build passes
• python -m py_compile backend/app/**/*.py passes
• Added/updated docs in English
• Tested POST /api/model/validate and POST /api/benchmark/run-stream
• No secrets committed (check .env.local not included)
• Updated README.md if adding user-facing feature

5. Reporting Issues

Use GitHub Issues with template:

• Bug: steps, expected vs actual, uvicorn.log, http://localhost:8000/docs screenshot
• Feature: use case, proposed API, UI mock

6. Hugging Face Collaboration

• For model/dataset cards, see docs/HUGGINGFACE.md
• Tag PRs with hf if related to HF Hub integration

7. Review Process

• Maintainers review within 3 days
• CI must pass (build + lint)
• Squash-merge to main

Developer Certificate of Origin

By contributing, you agree that your contributions will be licensed under MIT.

Questions?

Open a Discussion or contact maintainers via GitHub.

Happy building!