Skip to main content
Alerts (/analytics/alerts) is the workspace alert configuration. Where the Observe surface gives passive read-only telemetry, alert rules turn those metrics into active page-me predicates.
AlertsAlerts

Alerts — Forge Console at /analytics/alerts

What you can do

  • Author an alert rule of one of five types: cost_daily, cost_per_run, step_failure_rate, approval_age and workflow_duration. The evaluator also runs three trend types (budget_burn_rate, cost_anomaly, failure_rate_trend), eight types in all, but this page cannot create the trend types.
  • Set the threshold, the comparison (gt, gte, lt, lte, eq) and an optional dimension.
  • Edit, enable, disable or delete a rule.
Rules are evaluated every 30 minutes. A rule that crosses its threshold fires through the notification bus, which delivers to every channel the deployment’s notification settings enable (in-app bell, Slack, email); a rule has no per-rule routing target.

Key state to watch

  • Last triggered — when the rule last fired.
  • Data status — for step_failure_rate, whether the daily rollup it reads is ready; the rule does not fire on incomplete data.
Alert rules are Console-only: a machine credential cannot read or author them. Machines subscribe to webhooks for events instead.
  • Observe alerts: event-count rules that notify the bell or a Slack channel.
  • Notifications: which channels deliver a fired alert to you.
  • Cost: the spend that the cost rule types measure.