Contributing

Contributions should improve the library's correctness, scope, usability, documentation, or interoperability.

Useful contributions include focused bug fixes, tests, documentation corrections, detector compatibility notes, examples, and new methods or approaches for conformal anomaly detection, conformal inference, FDR control, covariate-shift workflows, or related detector interfaces.

Setup

git clone https://github.com/OliverHennhoefer/nonconform.git
cd nonconform
uv sync --group dev
uv pip install -e ".[all]"

Optional pre-commit hooks:

uv run pre-commit install

Checks

Run the checks that match the change:

uv run ruff format .
uv run ruff check . --fix
uv run pytest

For documentation changes:

uv run mkdocs build -f docs/mkdocs.yml

Pull Requests

Keep pull requests focused and include:

• what changed and why
• exact validation commands run
• public API impact, if any
• statistical-core impact, if any
• relevant papers, detector libraries, or implementation references when useful

Changes to p-values, FDR control, weighting, calibration, aggregation, metrics, or validity claims need tests and an explicit before/after rationale.

API Stability

nonconform is a v1 project. Public APIs should remain stable unless a breaking change is intentional and documented.

Public contracts include root exports, public module __all__ exports, documented constructor arguments, methods, properties, enum values, and dataclass fields. nonconform._internal is private.

Issues

Bug reports should include a minimal reproducible example, expected and actual behavior, environment details, and tracebacks or warnings when relevant.

Feature requests should describe the workflow, proposed method or approach, statistical or API impact, and relevant prior work when applicable.