Chapter D · 4 of 4~10 min

Production monitoring lab

Watch **latency, quality, and cost** after deploy — alert, rollback, or investigate with traces.

Reviewed July 2026 · interactive lab

Paths & resources for this lesson

1Try it yourself

Playground

Production alert desk

Monitor quality, latency, and cost — alert vs rollback vs watch.

p95 latency 3× baseline

2Read & reflect

Recap

#What you just did

You chose alert vs rollback vs watch for live signals.

Read

#Minimum dashboard

| Metric | Action | |--------|--------| | [p95 latency](/reference/p95-latency) | Page if > 2× baseline | | Citation / groundedness rate | Rollback if −30% day-over-day | | Error rate on /api/chat | Alert on 5xx spike | | Token spend | Budget alert |

3Spark check

Continue learning · glossary & guides