Skip to content

Alerts

Alerts authors alert rules: a metric, a comparison, a threshold, a time window, a cooldown, notification emails, and an enabled state. Each rule is a versioned config object, so it lives with the rest of the authoring surfaces in System even though its console route stays under Monitoring for deep-link stability. Below the rule list, Fire history shows recent threshold breaches read from the platform.

The Alerts panel in the admin console, showing the alert-rule table with metric, condition, and status columns, an alert-rule form, and a fire-history table below.
Monitoring → Alerts (config authored from System)Citadel synthetic data on a non-production Arbitex demo instance. Never client data.
  1. Open Alerts (console URL /monitoring/alerts — the route was kept there for deep-link stability, though the surface is a System-style authoring page) and click New alert.

  2. Enter a Name, then choose a Metric from the fixed list: Cost (USD), Request count, Throttled count, Error rate, Latency p95 (ms), DLP events, or Quality anomaly score.

  3. Choose a Comparison (greater than, less than, at least, at most, or equals) and a Threshold value.

  4. Set the Window (minutes) the metric is evaluated over. A cooldown prevents the same rule from re-firing immediately after it fires.

  5. Add one or more Notify emails as chips, toggle Enabled, then Save alert. Use Edit or Delete on any existing rule.

  6. Review Fire history below for a timeline of past breaches — when each fired, its type, score, observed value versus threshold, and model.

List rules, author a rule, and read fire history:

Terminal window
curl -X POST "https://admin.arbitex.ai/api/v1/admin/monitoring/alerts/list" \
-H "Authorization: Bearer $ADMIN_TOKEN"
curl -X POST "https://admin.arbitex.ai/api/v1/admin/monitoring/alerts/set" \
-H "Authorization: Bearer $ADMIN_TOKEN" \
-H "Content-Type: application/json" \
-d '{"name": "High DLP volume", "metric": "dlp_events", "comparison": "gt", "threshold": 100, "window_minutes": 60, "cooldown_minutes": 60, "enabled": true, "notify_emails": ["[email protected]"]}'
# → { "rule": { "id": "...", "name": "High DLP volume", ... }, "revision": 7 }
curl -X POST "https://admin.arbitex.ai/api/v1/admin/monitoring/alerts/delete" \
-H "Authorization: Bearer $ADMIN_TOKEN" \
-H "Content-Type: application/json" \
-d '{"id": "$RULE_ID"}'
curl -X POST "https://admin.arbitex.ai/api/v1/admin/monitoring/alerts/history" \
-H "Authorization: Bearer $ADMIN_TOKEN"

Alert rule authoring is available on every plan. The number of active rules is not gated by plan tier today.

On a Hybrid Outpost, alert rules evaluate against that deployment’s own telemetry, so a rule authored there only fires on that Outpost’s traffic. In an air-gapped deployment, rule evaluation and fire history both work against locally collected telemetry, with no external dependency; email notification delivery depends on your local mail relay being reachable.

  • Budget — a related, but separate, spend-visibility surface (not an alert rule).
  • Monitoring overview — where usage, cost, and DLP telemetry that feeds these metrics is reviewed in depth.