observability.md

Metrics and logging

Enable Prometheus metrics with the built-in collector and configure log verbosity with CONDUCTOR_LOG_LEVEL:

import { createMetricsCollector, MetricsServer, TaskHandler } from "@io-orkes/conductor-javascript";

const metrics = createMetricsCollector();
await new MetricsServer(metrics, 9090).start();

const handler = new TaskHandler({ client, eventListeners: [metrics], scanForDecorated: true });
await handler.startWorkers();
// GET http://localhost:9090/metrics — Prometheus text format

Two metric surfaces exist: legacy (default) and canonical (opt-in via WORKER_CANONICAL_METRICS=true, unprefixed Histogram-based names, bounded uri labels). See METRICS.md for the full catalog and migration guide.

For agent executions, inspect the shared workflow record (agents/reference/client.md) for inputs, outputs, tool calls, retries, and status. Avoid logging credentials or unredacted sensitive data.