Definition
Observability
Observability is the ability to explain why an agent did what it did, not just that a metric moved. Agent monitoring has a unique problem: HTTP 200 OK with a wrong answer. Traditional APM tools see a healthy system. Traces, spans, structured logs, and eval scores let you follow a hallucinated answer back to the outdated document that caused it. The book calls it non-negotiable and, at multi-agent scale, existential.
Explained in
Chapter 13: Production AI: Deployment, Monitoring, and Evaluation
The reality of running AI agents. It's nothing like the demo.
Related terms