AI-103: Developing AI Apps and Agents on Azure

By Microsoft · Question Mode

QuestionQ36

VerifiedImplement generative AI and agentic solutions

Your Microsoft Foundry project includes a customer support agent that calls an internal knowledge API tool before it generates responses.

Users are reporting the following issues:

  • Certain requests take longer than 15 seconds to complete.
  • Certain responses are incorrect, even though the knowledge API returns the expected data.

You need to inspect individual agent runs to view the ordered sequence of large language model (LLM) calls, tool invocations, and timing information.

Which observability capability should you use?

Explanation

Tracing in Microsoft Foundry captures, for each individual agent run, an ordered sequence of spans covering LLM calls, tool invocations, and their durations, letting you follow exactly what happened inside a single run to diagnose both the latency spikes and the cases where a correct tool result still produced a wrong response. Token usage, monitoring dashboards, and safety metrics are aggregate signals collected across many runs and do not expose the ordered, per-run call sequence needed for this kind of root-cause inspection.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!