git_bayesect is a Python tool that extends git bisect with Bayesian inference to identify commits that changed the likelihood of flaky or probabilistic test failures. Unlike standard git bisect which requires deterministic pass/fail outcomes, it handles non-deterministic scenarios by using a Beta-Bernoulli conjugacy model and greedy entropy minimization to select the most informative commit to test next. It supports priors based on filenames, commit messages, or diffs, and can run automatically with a command. Installable via pip or uv.
Sort: