medium10 pts · first answer countsobservabilityagents

Debugging a bad agent answer

A user reports a wrong answer from your multi-step agent. Your logs only store the final answer. What should you add so you can debug cases like this?

A.A daily dashboard of total requests and average response timeB.The user question stored next to the final answer in one rowC.A per-request trace with a span for every model, retrieval and tool stepD.Debug-level logging turned on for the web server and load balancer

Discussion 0

Join free to post in the discussion.

Debugging a bad agent answer

Discussion 0

Related questions