Back to docs overview

XAI Feedback Operations

How explainability feedback from chat is reviewed in admin and converted into quality improvements.

What It Captures

XAI Feedback collects feedback from the explainability panel in chat. It focuses on explainability, source quality, and missing context instead of generic like/dislike signals.

  • Issue types: too_general, wrong_source, missing_context, other
  • Admin route: /xai-feedback
  • Feedback origin: chat explainability panel

Operational Flow

The admin page combines KPI cards, issue breakdowns, top agents, top sources, and a searchable table with row details. This helps separate recurring patterns from one-off incidents quickly.

Recurring findings should always turn into concrete actions: prompt tuning, retrieval fixes, source hygiene, guardrails, memory tuning, or governance escalation.

API Surface

The data is provided by the API Gateway route `chat_explain.py`. Feedback is tenant-scoped and can be reviewed across tenants by superadmins.

  • POST /api/v1/chat/explain/feedback
  • GET /api/v1/chat/explain/feedback
  • GET /api/v1/chat/explain/feedback/stats