Skip to main content

​
Real-time Metrics Export

The MetricsExporter provides per-agent metrics in Prometheus, JSON, and real-time streaming formats.

​
Setup

import { Agent, openai } from "@radaros/core";
import { MetricsExporter } from "@radaros/observability";
import { createAgentRouter } from "@radaros/transport";

const exporter = new MetricsExporter();
const agent = new Agent({ name: "bot", model: openai("gpt-4o") });
exporter.attach(agent.eventBus);

const router = createAgentRouter({
  metricsExporter: exporter,
  cors: true,
});

​
REST Endpoints

EndpointFormatDescription
GET /metricsPrometheus textScrape-ready for Prometheus/Grafana
GET /metrics/jsonJSONDashboard-friendly JSON
GET /metrics/json?agent=botJSONPer-agent metrics
GET /metrics/streamSSEReal-time event stream

​
Available Metrics

  • runs — total agent runs
  • errors — total errors
  • avgDurationMs — average run duration
  • p95DurationMs — 95th percentile latency
  • totalCost — accumulated USD cost
  • totalTokens — total tokens consumed
  • toolCallCount — total tool invocations
  • toolUsageFrequency — per-tool call counts
  • errorRate — error ratio
  • tokensPerRun — average tokens per successful run

​
Prometheus Format

radaros_agent_runs_total{agent="bot"} 42
radaros_agent_errors_total{agent="bot"} 1
radaros_agent_duration_ms_avg{agent="bot"} 1250
radaros_agent_duration_ms_p95{agent="bot"} 3200
radaros_agent_tokens_total{agent="bot"} 125000
radaros_agent_cost_usd_total{agent="bot"} 0.85
radaros_agent_tool_calls_total{agent="bot"} 156