Contributing to LiteResearcher

Thanks for your interest in improving LiteResearcher. This guide covers how to report issues, propose changes, and submit pull requests.

Repository layout

Directory	Purpose
inference/	Inference & evaluation harness
training/	RL training stack (GRPO + difficulty-aware curriculum), built on verl
datagen/	Training-data and corpus synthesis pipeline
environment/	Local search/browse environment (Milvus + BGE-M3, PostgreSQL)
docs/	Project page and trajectory case viewer

Each component has its own README.md and dependency file; start there for component-specific setup.

Reporting issues

• Search existing issues before opening a new one.
• Include the component (inference / training / datagen / environment), your environment (OS, Python, GPU/driver, key package versions), the exact command you ran, and the full error output.
• For reproducibility problems, note the model checkpoint and dataset revision you used.

Development setup

git clone https://github.com/simplexai-labs/LiteResearcher.git
cd LiteResearcher
# Install the component you are working on, e.g.:
cd inference && pip install -r requirements.txt

Pull requests

1. Fork the repo and create a topic branch from main (e.g. fix/eval-timeout, feat/corpus-dedup).
2. Keep each PR focused on a single concern. Unrelated cleanups belong in separate PRs.
3. Match the existing code style of the component you touch. The training/ stack inherits verl's pre-commit config — run pre-commit run --all-files there before pushing.
4. Update the relevant README.md when you change behavior, flags, or paths.
5. Write a clear PR description: what changed, why, and how you verified it.

Commit messages

Use concise, conventional-style prefixes where they fit (feat:, fix:, docs:, refactor:, test:) and explain the why in the body when it is not obvious.

License

By contributing, you agree that your contributions are licensed under the Apache License 2.0.