📊

LLM Observability has moved to Grafana

Calls, cache-hit rate, fallback & error rate, p50 latency, token burn, per-hive burn and cost-by-provider are all live panels on the Founder / Platform Ops dashboard — never stale.

📊 Founder Ops dashboard →

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

LLM Observability

Loading ai_cost_log + ai_cache rollups...

Total calls

-

↑ window-over-window

Cache hit rate

-

vs LLM-served calls

Fallback rate

-

non-primary provider

Tokens used

-

in + out (estimated)

p50 latency

-

ms (LLM-served only)

Error rate

-

fallback or schema fail

Per-task routing

Per-task LLM routing: calls, cache hit rate, p50 latency, provider mix
Task / Fn	Calls	Cache %	p50 ms	Provider mix
Loading...

Per-hive budget burn

Per-hive LLM budget burn: calls in window vs limit, last activity
Hive	Calls (window)	Limit/hr	Budget burn	Last call
Loading...