Hygiene of the case base
The “Hygiene” tab of a project finds what every case base gets overgrown with:
- likely duplicates — by text similarity (word bigrams, Jaccard index; ё = е, case and punctuation do not
matter). Cases can be compared side by side and one kept: the other becomes deprecated with a
duplicate-of-…tag; - not run for long — active cases without results for N days (new cases are not included);
- never failing — not a single failure in N+ runs: candidates for automation or less frequent runs;
- flaky — the result “flips” between passed and failed (≥ 30% changes in the latest results).
Thresholds are set on the page. A finding can be hidden (“all good”); hidden ones come back with one button.
API: GET /api/projects/:project/hygiene, POST|DELETE /api/projects/:project/hygiene/ignores.