Question 32
Domain 6: Evaluation and MonitoringWhich source is best for analyzing the live request and response payloads of a deployed AI agent endpoint over time?
Correct answer: A
Explanation
Inference tables store endpoint inference data, including request and response payloads, so they are the best source for analyzing live traffic over time. They let you review how a deployed AI agent endpoint is being used and inspect payloads after deployment.
Why each option is right or wrong
A. Inference tables
Inference tables are the Databricks SQL table-backed record of endpoint inference activity, and they capture the request and response payloads for each call made to a deployed model or AI agent endpoint. Because they persist this traffic over time, they are the only source here that supports retrospective analysis of live endpoint behavior rather than a one-off snapshot or deployment metadata.
B. The model version alias
A model version alias points to a deployed version; it does not store request/response traffic history.
C. The Prompt Registry commit history
Prompt Registry commit history tracks prompt changes and versions, not live endpoint payload observations.
D. A DBFS mount
A DBFS mount is a storage access location, not a purpose-built record of endpoint inference events.