Monitoring - UnitBuilds-CC/V.E.L.O.C.I.T.Y.-MCP GitHub Wiki

Monitoring

Production-ready monitoring with Prometheus, Grafana, and OpenTelemetry.

Prometheus Metrics

Available at /metrics in HTTP mode:

curl http://localhost:3000/metrics

Available Metrics (20+)

Metric	Description
`velocity_requests_total`	Total requests by method and status
`velocity_request_duration_seconds`	Request latency histogram
`velocity_active_connections`	Current active connections
`velocity_active_sessions`	Current active sessions
`velocity_tool_calls_total`	Tool calls by name and outcome
`velocity_tool_call_duration_seconds`	Tool execution latency
`velocity_sandbox_violations_total`	Sandbox violations by category
`velocity_audit_log_entries`	Audit log entry count
`velocity_rate_limiter_rejections_total`	Rate-limited requests
`velocity_nda_operations_total`	NDA compile/read/execute ops
`velocity_errors_total`	Errors by type
`velocity_cache_hits_total`	Cache hit/miss counts

Alerting Rules

Pre-configured rules in monitoring/prometheus/alerts.yml:

High error rate (>5% over 5 min)
High latency (p95 > 1s over 5 min)
Sandbox violations detected
Rate limiter exhaustion
Session limit approaching capacity

Grafana Dashboard

Pre-built dashboard in monitoring/grafana/dashboard.json:

Request rate and latency
Tool call breakdown
Error rates and types
Sandbox violations
Cache hit rates
Session/connection counts

OpenTelemetry

cargo build --release --features observability
export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317
velocity_mcp --mode http

All tracing spans automatically exported to your OTLP collector.

Health Endpoints

# HTTP mode
curl http://localhost:3000/health
curl http://localhost:3000/performance

# Any mode (JSON-RPC)
echo '{"jsonrpc":"2.0","method":"health/check","id":1}' | velocity_mcp --mode stdio