Question 34
Domain 3: Implement Generative AI SolutionsA Prompt Flow has been deployed as an endpoint. The team wants to automatically track which documents were retrieved, which chunks were used, and what the model's intermediate reasoning was for each request. What should you enable?
Correct answer: B
Explanation
Tracing in Azure AI Foundry records the execution path of a Prompt Flow endpoint, including retrieved documents, used chunks, and intermediate model steps. This matches the need to automatically track “which documents were retrieved, which chunks were used, and what the model's intermediate reasoning was” for each request.
Why each option is right or wrong
A. Azure Monitor custom metrics
B. Tracing in Azure AI Foundry
Azure AI Foundry Tracing is the feature that captures end-to-end execution details for a deployed Prompt Flow endpoint, including retrieval events, chunk selection, and intermediate node/model outputs for each invocation. That makes it the only option that can automatically record the request-level provenance the team is asking for, rather than just aggregate metrics or logs.
C. Content filter logging
D. Azure Application Insights only