LLM Generated Feedback

After a session completes, an LLM reviews it and writes objective feedback. You get actionable suggestions without manual triage.

Feedback AI

What gets generated

  • Summary of the session goal and outcome
  • Strengths and issues observed in answers
  • Suggestions for prompts, tools, or content

Where it goes

  • Stored alongside the session in DynamoDB
  • Indexed in OpenSearch for exploration in the Admin Site

Benefits

  • Close the loop on quality with minimal effort
  • Spot recurring gaps and prioritize fixes
Non-blocking

Feedback is asynchronous and never slows down end‑user chats.

Configuration Required

The LLM Generated Feedback (Verify Response) feature must be enabled in your site-wide pika-config.ts file before chat apps can use it. Without this enablement, feedback generation will not be available regardless of individual chat app settings.

Suggest changes to this page
Last update at: 2025/09/17 14:37:11