Trace and Investigate Agent Behavior
Now that traces are flowing, let’s use them. In this chapter you’ll pivot into Splunk Agent Observability and investigate exactly what your agent did on each request: the reasoning, the tool calls, the retrieved context, the tokens, and the latency, all in one place.
Persona
As Careful Health Provider’s AI engineer, you finally have complete visibility into the
assistant. When a patient reports a bad answer, you no longer guess; you open the trace and
see the exact path the agent took to get there.
Instant visibility. Splunk Agent Observability shows the root cause of errors
across complex, multi-step agent workflows. Instead of stitching together logs, you see the full request as a tree of spans you can expand and inspect.
The traces you’ll explore here come from the instrumented app in
~/workshop/healthcare-assistant/2-app-with-instrumentation (the traffic you generated in the
previous chapter).
Continue to explore and investigate a trace.