Support engineers debugging a multi-agent failure in Microsoft Foundry can see that a user request failed, but they cannot reconstruct which agent ran, which tools were called, how many tokens were consumed, or how child spans related to the parent request. Which observability implementation should the team prioritize?
Select an answer to reveal the explanation.
Short Explanation
You need end-to-end tracing in Foundry that ties parent requests to agent runs and tool spans with correlation IDs, token and execution tracking, and alerting. Local-only prints without correlation don’t help production, screenshots aren’t traces, and turning logs off is sabotage.
Full Explanation
The correct implementation is tracing in Foundry that captures tokens, prompts subject to privacy policy, correlation IDs, alerting, and execution tracking across agents and tools. Observability for multi-agent solutions requires structured traces so failures can be reconstructed across agents and tools. The free-text debug strings only in local developer consoles without correlation are incorrect because uncorrelated local prints do not provide production cross-agent forensics. Relying exclusively on end-user screenshots of the chat window as the system of record for traces is incorrect because user screenshots lack structured spans, tokens, and tool payloads needed for root-cause analysis. Disabling all logging to ensure failures leave no forensic trail is incorrect because disabling logging prevents reliability monitoring and remediation. Exam focus: correlation IDs plus structured execution traces are essential multi-agent observability.