Installation¶

Install Conformity when you already have a scikit-learn model and need held-out uncertainty calibration:

uv add conformity-calib
# Or:
pip install conformity-calib

The distribution is named conformity-calib; import it as conformity. Python 3.9 or newer is supported. Runtime requirements are NumPy >=2.0.2, scikit-learn >=1.6.1 and typing-extensions >=4.15.0. The lock resolves compatible dependencies for each Python version; newer dependency releases may require newer Python.

Plotting is optional and uses Matplotlib:

uv add matplotlib
# Or:
pip install matplotlib

from conformity import ConformalClassifier, ConformalRegressor, __version__

assert isinstance(__version__, str)
assert callable(ConformalRegressor) and callable(ConformalClassifier)
print(__version__)

For a checkout, use the locked development instructions. Continue with getting started to fit your first calibrated model.