Observability & Prometheus Metrics

Gain deep visibility into latency, token usage, error rates, and provider performance.

Metrics Export (Prometheus) #

Naagmani exposes standard Prometheus metrics at :8081/metrics:

# TYPE naagmani_http_requests_total counter
naagmani_http_requests_total{status="200",provider="openai",model="gpt-4o"} 14209
naagmani_http_requests_total{status="429",provider="openai",model="gpt-4o"} 12

# TYPE naagmani_ttft_seconds histogram
naagmani_ttft_seconds_bucket{le="0.25",provider="anthropic"} 8400
naagmani_ttft_seconds_bucket{le="0.5",provider="anthropic"} 12100

Grafana Dashboards #

Pre-built dashboards are provided in the official repository under /deploy/grafana/ for instant monitoring of:

• Gateway Throughput & P99 Latency
• Provider Cascade Failures
• Real-Time Budget Burn Rate

Next Steps #

• Scaling & Concurrency
• Marketplace Overview