FlakyTestHunter
CI suites accumulate tests that fail on their own schedule, not the code's — nobody scores them, so engineers just learn to re-run the red build and hope. FlakyTestHunter ranks every test by how inconsistent it's actually been, and shows what quarantining the worst offenders does to your real pass rate.
Which tests can you actually trust?
Real flakiness scoring
A 0–100 score per test computed from the Shannon entropy of its pass/fail distribution blended with its run-to-run transition rate, dampened by a confidence factor for low run counts — not a lookup table.
Ranked history, one click deep
Every test shows a pass/fail sparkline sorted worst-to-best. Click a row to expand the full run grid with flip points (state changes) highlighted.
Quarantine, with a number that moves
Flip a toggle to exclude flagged-flaky tests from the pass rate and watch the suite's real, trustworthy pass rate animate upward live.
Not deployed yet — try it from source.
This demo runs entirely client-side (synthetic CI data, scored and rendered in the browser). Clone the repo and start the bundled static server:
git clone https://github.com/sridharvanka/flakytesthunter.git cd flakytesthunter node server.mjs # open http://localhost:3020