Operations begins from persisted truth. A model response, log line, SQS delivery, or external API response is not by itself proof that a job advanced or an effect succeeded.
Operating model
Understand workers, outboxes, retries, and ownership.
Operational control plane
Use the needs-you inbox, resumable job shells, durable controls, and batch execution.
Recover failures
Classify failures and resume safely.
Audit authority
Trace digest-bound human decisions.
Inspect telemetry
Correlate jobs, operations, traces, and usage.
Verify a release
Create exact-commit evidence and scan every deployable image.
Respond to alarms
Contain incidents and reconcile effect-bearing work.
Restore a backup
Exercise recovery into an isolated table.
Rotate service tokens
Replace generated internal credentials safely.
Run local load
Measure a built loopback service without hosted traffic.
Bounded autonomy and operational KPIs
Where Harmonia reasons, where deterministic code controls execution, where humans retain authority, and how the demo is measured.
Operational control plane
Operator attention, resumable live state, durable job controls, connector capabilities, and large-batch execution.
Failure and recovery reference
Typed failure categories, retry limits, escalation, replay, and fail-closed recovery behavior.
State ownership and retention
The authoritative owner, scope, lifetime, writers, readers, and retention boundary for every Harmonia state store.
Authenticated record and replay
Capture, sanitize, review, and replay authenticated runs without fabricating evidence or mutating durable state.
How to respond to an operational alarm
Triage Harmonia alarms from persisted state, contain effects, and recover without replaying uncertain work.
Last modified on September 14, 2026
