Monitoring

Every workflow execution is fully recorded. The platform provides built-in observability through the WorkflowRun and TaskRun ledger, with Sentry for errors and Logfire for distributed tracing.

Workflow observability

Every workflow execution produces a WorkflowRun record with:

run identifier
trigger source (user / schedule / API)
execution start and end time
final status

Each step produces a TaskRun record with:

step name and verb
execution status
error details (if failed)
execution duration

These records are viewable under Workflows → Runs in the platform.

Run states

State	Meaning
`PENDING`	Created, awaiting worker
`RUNNING`	Currently executing
`SUCCEEDED`	All steps completed
`FAILED`	A step encountered an error
`SUSPENDED`	Awaiting governance approval
`CANCELLED`	Cancelled by user

Integration monitoring

Adapters communicate with external systems. Monitor for:

API authentication failures (credential expiry)
Network connectivity errors
Unexpected schema changes in adapter responses
Rate limit responses from external APIs

Repeated adapter failures in a short window are a strong signal of credential expiry or a breaking API change upstream.

Queue monitoring

The ARQ worker processes jobs from Redis. Monitor:

queue depth
worker throughput
processing latency (time from enqueue to execution start)

High queue latency without a corresponding spike in job volume usually means a worker instance is down or under memory pressure.

External observability

Tool	Purpose
Sentry	Error tracking and alerting
Logfire	Distributed tracing across API + worker

For production deployments, configure Sentry alerts on new error groups and Logfire dashboards for workflow execution duration and adapter error rates.

Alerting recommendations

Alert on:

any WorkflowRun in FAILED state older than 5 minutes
adapter error rate above baseline for a sustained window
queue backlog growing without worker consumption
SUSPENDED workflows older than your SLA for approvals