Trace and Investigate Agent Behavior

Explore Other Views

5 minutes

Let’s explore other views in Splunk Agent Observability using the traffic you just generated.

Exercise Investigate agent behavior
1

Open your project and agent stream

Return to your project and the default agent stream in Splunk Agent Observability.
2

View the Agent Graph

Click on the Agent graph tab, which aggregates agent behavior across all traces in the agent stream. It shows the most common execution paths, along with latency and frequency patterns, so you can quickly see how the agent is behaving with actual users.

Agent Graph

3

View Trends

We can also switch to the Trends tab to understand how the system is performing over time. For example, we can see how the overall latency and token usage trends across requests.

Trends

From investigation to automation

Reading individual traces is powerful for a single incident, but you can’t manually inspect millions of them. In the next chapter you’ll enable evaluators to score every trace automatically, so problems like ungrounded medical advice surface on their own.