📊

Agentic-RAG Observability has moved to Grafana

RAG trace volume, grader/checker pass rate, citations, retries, latency and token use are live panels on the Founder / Platform Ops dashboard — never stale.

📊 Founder Ops dashboard →

Grafana login: admin · password in infra/mcp/.env.mcp

Agentic RAG Observability is hive-scoped.

You need to join or create a hive first to see your agentic-RAG traces.

Go to Hive board →
Computing observability verdict…
How this dashboard works (open the engineering panel)

All numbers are computed from the per-hive agentic_rag_traces table via a head-only count query in the selected window. Hide-zero filter dims any row where calls = 0 so a quiet route reads as calm, not broken.

Route breakdown

Volume + retry rate + checker pass rate per route. Source: agentic_rag_traces grouped by route in the selected window.
RouteCallsAvg retriesGrader %Checker %Avg tokensp95 ms

Top 10 most token-heavy queries

Highest total_tokens in the selected window. Spot prompt drift or runaway retrievers early.
QuestionRouteRetriesCheckerTokensLatency msWhen

Recent traces

Most recent 20 agentic-rag-loop runs in the selected window.
WhenRouteQuestionRetriesGraderCheckerCitationsLatency ms