Assistant quality gate
Answer quality, not just retrieval quality
This page reads the latest persisted assistant RAG reports, sanitized job-fixture checks, and pending improvement proposals created from insufficient-answer feedback.
No report yet
Generate the report locally, then refresh this page.
python scripts/evaluate_assistant_rag_answer_quality.pyImprovement Proposals
Insufficient-answer feedback creates pending proposals. They are not applied until reviewed, gated, and approved.
0 pending/recent
Idle
Pending: 0
Approved: 0
Rejected: 0
Staged cases: 0
Staged review file:
/app/reports/assistant_rag_feedback/staged_approved_eval_cases.json
No feedback-driven proposals have been recorded yet.