RAG Service
API Docs Center
← Back to Documentation Center

Feedback Endpoint

Centralized API for collecting, storing, and analyzing user and system feedback on RAG and chat responses. Feedback is used to monitor quality, guide improvements, and support A/B testing and analytics.

Business & Functional Overview

The feedback system enables: - User satisfaction tracking: Collects ratings and comments on answers, surfacing issues and strengths. - A/B testing support: Associates feedback with experiments/variants for model evaluation. - Quality improvement: Provides a feedback loop for model and system tuning. - Analytics: Enables reporting on thumbs up/down rates, completion, and user trends.

Feedback is stored in Azure Table Storage or Blob Storage (with local fallback), deduplicated by response, and can be retrieved for analytics or audit.

Routes

  • POST /feedback
  • GET /feedback/{id}
  • GET /feedback?user_id=...
  • GET /feedback/summary — summary stats (thumbs up/down, sample)
  • GET /feedback/recent — recent feedback list
  • GET /feedback/ui — modern feedback dashboard UI

Description

Allows users and systems to submit feedback (ratings, comments, corrections) and retrieve feedback history. Feedback can be submitted programmatically or via the UI. Each feedback record is associated with a user, session, response, and (optionally) experiment/variant.

Request Body (POST)

See FeedbackRequest schema for canonical details. Example:

{
    "user_id": "user-123",
    "session_id": "sess-456",
    "experiment_name": "abtest1",
    "variant_name": "A",
    "response_id": "resp-789",
    "rating": 1,
    "feedback_text": "Great answer!",
    "response_time": 1.23,
    "task_completed": true,
    "response_payload": {"question": "...", "answer": "..."}
}

Response Example

See FeedbackResponse schema for canonical details. Example:

{
    "status": "success",
    "message": "Feedback recorded"
}

Parameters

  • user_id (string): User identifier
  • session_id (string): Session identifier
  • experiment_name (string, optional): A/B test experiment name
  • variant_name (string, optional): A/B test variant name
  • response_id (string): ID of the response being rated
  • rating (int): -1 (down), 0 (neutral/autosave), 1 (up)
  • feedback_text (string, optional): Free text
  • response_time (float, optional): Time taken to respond
  • task_completed (bool, optional): Whether the user completed the task
  • response_payload (object, optional): Context for training/analytics

Authentication

Requires valid API key or Azure AD token.

Errors

  • 400 Bad Request: Invalid input
  • 401 Unauthorized: Missing/invalid credentials
  • 404 Not Found: Feedback not found
  • 500 Internal Server Error: Unexpected failure

UI & Analytics

  • Dashboard: /api/v2/feedback/ui provides a modern, filterable dashboard for browsing, searching, and analyzing feedback records. Features include:
    • Filter by user, rating, text, time
    • View question, answer, sources, and metadata
    • Copy/collapse details, show raw JSON
    • Summary stats (total, thumbs up/down, up-rate)
  • Summary: /api/v2/feedback/summary returns aggregate stats and a sample of recent feedback for analytics.
  • Recent: /api/v2/feedback/recent returns a list of recent feedback records (with all metadata).

Usage Tips

  • Use feedback to improve model quality and monitor user satisfaction.
  • Retrieve feedback by user, session, or response for analytics.
  • Use the UI for quick triage and exploration; use the API for programmatic analysis.
  • Feedback is deduplicated by response and can be associated with A/B test variants.

See Also

Usage Tips

  • Use feedback to improve model quality and monitor user satisfaction.
  • Retrieve feedback by user or feedback ID for analytics.