I welcome focused bug fixes, tests, reproducibility improvements, and clearly scoped research extensions. For a large change, please open an issue first so the design and evidence requirements can be agreed before implementation.
CliffQuant requires Python 3.11 or newer.
python -m venv .venvActivate the environment with .venv\Scripts\Activate.ps1 in PowerShell or
source .venv/bin/activate on macOS and Linux, then install the project:
python -m pip install --upgrade pip
python -m pip install -e ".[dev,figures]"Run the same checks used in continuous integration:
python -m ruff check .
python -m ruff format --check .
python -m pytest --basetemp .pytest-local-contrib
python -m buildUse python -m ruff format . to apply the formatter.
- Add a regression test for behavior changes.
- Keep calibration, held-out evaluation, and solver certification separate.
- Support quantitative statements with a reproducible script, protocol, and raw checksummed result where applicable.
- Report negative or incomplete results plainly; do not broaden a result beyond the metric and model that were tested.
- Do not commit secrets, private data, local caches, environments, or build outputs.
Keep each pull request narrow enough to review and explain any intentional change to deterministic behavior or evidence provenance.