Documentation

Orchestrate runs with clarity.

Workflows are stored in Redis and executed as runs. Each run emits a timeline of step states, approvals, and outcomes, providing a complete audit trail.

Capabilities

DAG execution with depends_on
Approval, delay, notify steps
for_each fan-out with aggregation
Retries with exponential backoff
Rerun from step & dry-run mode
Schema validation (Input/Output)

Step Types

job

approval

delay

notify

condition

for_each

Steps dispatch jobs as sys.job.submit events on the bus.

Workflow Definition Example

workflow.yaml

id: incident.triage
input_schema: IncidentContext
steps:
  - id: enrich
    uses: worker.incident.enricher
    
  - id: approval
    type: approval
    reason: "Prod write detected"
    
  - id: patch
    uses: worker.patch.apply
    requires: ["prod", "write"]
    retry:
      max_attempts: 3
      backoff: "10s"