Skip to content

System Alerts & Self-Healing

Keel System Alerts & Self-Healing settings — alert destination, self-healing mode, and metrics endpoint

Location in Keel: /settings/system-alerts

System Alerts controls where the appliance sends alerts about its own health — and whether it may try to fix problems automatically. This is separate from the detections and incidents raised from your log data.

Configure a destination first

With no email recipient or notification channel set, system alerts are dropped. Add a destination so health alerts can actually be delivered.

Alert destination

Field Purpose
Enabled Turn self-health alerting on
Email recipients Comma-separated addresses (e.g. [email protected], [email protected]); uses your SMTP settings
Minimum severity Warning and above or Critical only
Send recovery notices Also notify when a problem clears
Also send to notification channels Route alerts to Slack/Teams/webhook as well

Use Send test alert to confirm delivery.

Self-healing

When a problem is detected, Keel can attempt a bounded, non-destructive fix — restart a service, reconcile config, re-warm models. It never deletes data, and the AI may only choose from a fixed list of safe actions.

Mode Behavior
Off Detect and alert only — never take action
Recommend only Propose a fix; an admin applies it from the activity log
Auto (safe) Run deterministic fixes automatically; AI-suggested fixes remain recommend-only
Auto (full) Run both deterministic and AI-selected non-destructive fixes automatically

Recent self-healing activity lists remediations taken or proposed.

Prometheus metrics endpoint

Enable the metrics endpoint to expose the appliance's own health at /api/system/metrics for your monitoring to scrape. It is disabled by default and authenticated — use an API key / bearer token.