Skip to content

Contributing to BioML-bench

We welcome contributions to BioML-bench! This guide covers how to contribute effectively.

Development Setup

  1. Fork and clone the repository
  2. Install development dependencies: 
    uv sync --extra dev
  3. Build the base environment: 
    ./scripts/build_base_env.sh

Types of Contributions

  • New biomedical tasks - Add tasks from your domain expertise
  • Agent implementations - Contribute new AI agents
  • Data source integrations - Support new biomedical benchmark databases
  • Documentation improvements - Help improve this documentation
  • Bug fixes and optimizations - Code quality improvements

Contribution Process

  1. Create a feature branch
  2. Make your changes
  3. Test thoroughly with existing tasks
  4. Submit a pull request
  5. Address review feedback

Code Standards

  • Follow PEP 8 style guidelines
  • Add comprehensive docstrings
  • Include type hints
  • Update documentation for user-facing changes

Documentation

Building Documentation

BioML-bench uses MkDocs with auto-generated API documentation. When contributing:

Build and serve docs locally: 

# Build static docs
./scripts/build_docs.sh

# Build and serve docs locally
./scripts/serve_docs.sh

Documentation is automatically generated from:

  • Docstrings - All API documentation comes from code docstrings
  • CLI help - CLI documentation is extracted from argparse definitions
  • Manual content - User guides, tutorials, and conceptual documentation

When to update documentation:

  • Add docstrings to new functions/classes
  • Update docstrings when changing function signatures
  • Add examples to docstrings for complex functionality
  • Manual docs are rarely needed for API changes

Testing

Test your changes with: 

# Run basic tests
./scripts/test_environment.sh

# Test specific agent
biomlbench run-agent --agent dummy --task-id polarishub/tdcommons-caco2-wang

# Validate new tasks
biomlbench prepare -t my-source/my-new-task