Reference · Glossary
Runbook
A **step-by-step playbook** for common incidents — who to page, what to rollback, where logs live.
When to use
Repeatable ops for AI APIs, RAG apps, and agent services.
When not
As untested docs — run game-day drills on the runbook.
Example
“Citation rate drop → check index version → rollback if >30% → notify #ai-ops.”