QuestionQ132

Implement generative AI quality assurance and observability

An organization operates a customer-facing generative AI application built using Microsoft Foundry. The application uses multiple prompts connected to multiple workflows to generate production responses.

The application occasionally produces incomplete responses. Although the model call succeeds, the final message sometimes ends early.

The issue cannot be reproduced consistently in development.

You need to determine where and why response generation is ending early in production.

Which approach should you use?

  • A Run a pre-release evaluation workflow to score groundedness and relevance on a test dataset.
  • B Enable tracing and logging so that each workflow can be inspected.
  • C Increase max_tokens and temperature to reduce the chance of early termination.
  • D Replace the deployed model with a smaller model to reduce variability across responses.
Explanation

Microsoft Foundry tracing and logging capture request-level telemetry across agent and workflow execution, including model inputs and outputs, intermediate operations, timing, token usage, and errors. Inspecting these production traces identifies the workflow or operation at which the final response is truncated and supplies the context needed to determine why it terminated. Set up tracing in Microsoft Foundry

Learn more

Community Discussion

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