Promotion intervention loop
Routes demand signals through planners, recommendation logic, and release gates before activation.
schedulemanualsystem_signal
3 nodes · owner usr-001
Workflow orchestration
Operate agent, rule, tool, API, and review logic through one orchestration layer.
Workflow Orchestration keeps node-edge execution visible with run history, trigger posture, health overlays, and human-in-the-loop gates before release.
3Workflows
1Active or waiting review
2Healthy workflows
9Modeled nodes
Claims intake and routing
Moves incoming packets through document extraction, severity scoring, and reviewer escalation.
api_eventwebhook
3 nodes · owner usr-003
Platform governance monitor
Scans agent fleet, integrations, and release posture for control-plane exceptions.
schedulesystem_signal
3 nodes · owner usr-004
State machine and retries
Pending, running, waiting, failed, and completed states stay explicit under orchestration control.
max retries 2
if confidence < 0.82 -> human_review
Parallel branches enabled
max retries 1
if missing_docs -> human_review
Sequential execution only
max retries 3
if integration_missing -> incident_flow
Parallel branches enabled
Environment controls
Workflow execution policy changes as promotion moves from dev to protected environments.
Relaxed execution for rapid config iteration.
Regression, evaluation, and approval checks active.
Usage observation and escalation controls active.
Human review, policy enforcement, and immutable audit logs required.
Run history
Status overlays and trigger context stay visible before production decisions.
Started 2026-03-16T08:00:00Z
Duration 192,000 ms · ended 2026-03-16T08:03:12Z
Started 2026-03-16T10:18:00Z
Duration 48,000 ms · ended 2026-03-16T10:18:48Z
Started 2026-03-16T11:40:00Z
Duration 26,000 ms · ended 2026-03-16T11:40:26Z
Execution trace
Workflow execution trace with tool dependency failure.
Trace `/trace-003` exposes orchestration steps, tool calls, decision points, latency, token cost, and suggested remediation without opening backend internals.
Fleet health monitorfailedIncident bridge trigger failed
Observability Bus · Alerting
Integration missing in prod environment
180 in · 46 out · $0.30
model 90 ms · tool 170 ms · orchestration 110 ms
Decision path stayed inside registered tools, applied runtime guardrails, and routed to review whenever confidence or policy state required intervention.
Alerting channel was not enabled in production environment. Suggested fix: Promote alerting integration with protected secret scope and rerun health check.