Alerts not firing

Symptom

A metric crossed what you thought was the threshold, but no email or Slack arrived. The usual causes are timing (the monitor only evaluates on its schedule), the monitor being muted, the threshold direction or value being set differently than you remember, or the notification channel itself failing. Check the schedule and mute state first — those explain most “missing” alerts.

• A metric clearly breached the limit but no notification came.
• Alerts used to arrive and stopped.
• The alert log shows nothing for the period you expected.

---

Quick checks

• One alert_frequency cycle has elapsed since the breach (minimum 5, default 60 minutes).
• The monitor is not muted (is_mute).
• threshold_operator and the critical value match the direction of the breach.
• notification_emails and/or slack_webhook_url are set and valid.

Causes and fixes

Cause	What you see	Fix
Frequency hasn’t elapsed	A brief breach between evaluation runs left no alert	A monitor evaluates on alert_frequency (minimum 5, default 60 minutes). Lower the frequency if you need faster detection.
Monitor is muted	The monitor keeps evaluating but no notification arrives	is_mute stops notifications while evaluation continues; unmute it.
Threshold direction or value	A spike didn’t fire a “less than” monitor (or vice versa)	threshold_operator (Greater than / Less than) and the critical value must match the breach you expect.
Percentage-change baseline	A new project never alerts on a percentage-change monitor	A percentage-change monitor needs enough history in its auto_threshold_time_window to compute a baseline.
Notification channel	The alert log shows a fire, but no email/Slack arrives	Verify notification_emails (up to 5) and/or slack_webhook_url; a bad webhook silently drops the message.

Diagnostic checks

Open the monitor and read its frequency, mute state, operator, threshold value, and notification channels, then check the alert log:

• A log entry with no email/Slack points at the notification channel (notification_emails / slack_webhook_url).
• No log entry at all points at timing (alert_frequency), mute (is_mute), or the threshold direction.

Minimal smoke test

Set a deliberately easy threshold, wait one alert_frequency cycle, and confirm an alert log entry plus the email/Slack message arrive. Then restore the real threshold.

Escalate

If the monitor still won’t fire on a confirmed breach, contact support@futureagi.com with the monitor name, its config, and the breach timestamp.

Prevent recurrence

• Match alert_frequency to how fast you need to know — don’t leave it at 60 if minutes matter.
• Test each notification channel once when you create the monitor.

Next steps