Run the regression suite and mark known-good
- Open Admin → Operations.
- Click Run regression.
- Read the JSON. If ElevenLabs is not configured, scenarios report
skippedand only middleware eval ran. - Compare
diff.changesto the last known-good run. - If the run is the new baseline, click Mark known-good on that run id.
CI already runs apps/api/src/services/scenarios.test.ts (gates, vague-answer rejection, P2 queue, stale flag, apply-time gate) without voice. Use that as the merge bar. Treat the in-app simulation button as optional extra.