MicrometerAgentListener currently emits counters (executions, retries, errors). For production observability we want per-node duration histograms so dashboards can show p50 / p95 / p99 latency per agent in a graph.
Acceptance criteria
Suggested metric naming
agentflow.node.duration (timer, new)
agentflow.node.executions (counter, already exists)
MicrometerAgentListenercurrently emits counters (executions, retries, errors). For production observability we want per-node duration histograms so dashboards can show p50 / p95 / p99 latency per agent in a graph.Acceptance criteria
MicrometerAgentListenerrecords aTimerper node, tagged withnode,graph,outcome(success / failed / interrupted)docs/observability.md(new or existing) shows a Grafana panel exampleSuggested metric naming
agentflow.node.duration(timer, new)agentflow.node.executions(counter, already exists)