Contributing Guidelines

The authoritative guidelines live in the repository, in .github/CONTRIBUTING.md. Read them before opening a pull request; this page only summarizes what is in them so you know whether you need to.

What to work on:

Representation, quantification, and calibration methods, downstream tasks, datasets and dataloaders, and ports of existing features to PyTorch, HuggingFace, and scikit-learn. For anything outside that scope, open an issue first so we can discuss it.

Workflow:

Fork main, clone the fork, commit, push, and open a pull request against main.

Development environment:

Python 3.13+ with uv, then uv sync --dev. The library and its dev and docs dependency groups require Python 3.13+. .python-version pins the development environment to this minimum.

Code style:

Ruff for linting and formatting, configured in pyproject.toml and applied by the pre-commit hook.

Documentation:

Public features are documented in their docstring, following the Google style guide. If the feature comes from a paper, cite it.

Credit:

If you use code from another source, check its license and credit the original authors.

See also

Adding a New Method for the file layout, registration mechanics, and quality checks that a new uncertainty method needs.