Alerts#

Alerts is the home for conditions the platform would otherwise swallow silently — the things an operator would want to know about, and would complain about not knowing.

What surfaces here#

• Suppressed re-triggers — a deal re-entered a trigger stage while an approval was already in flight, so no new card was created. The suppression is recorded instead of being a quiet thread note.
• Stale waiting approvals — an approval that's been sitting undecided past a threshold.
• Dispatch failures — a write that was dead-lettered.
• Open DLQ entries — dead-letters that haven't been resolved.
• Connector auth-expiry — an OAuth refresh failed; the connection needs to be reconnected ("reconnect required").

How it works#

Alerts come from two sources, unified into one list:

• Raised alerts are recorded the moment a point-in-time condition occurs (a suppression, a dispatch failure). They have a lifecycle — open → acknowledged → resolved — and repeats of the same condition collapse onto one entry with a count.
• Derived alerts are computed live from the platform's own state (stale approvals, open dead-letters). They clear themselves the moment the underlying condition clears — there's nothing to acknowledge; you fix the condition.

In the console#

The Alerts item in the platform nav carries an open-count badge. The screen sorts the worst, freshest thing first, links each alert back to the workflow, connection, or dead-letter it's about, and lets you acknowledge or resolve raised alerts.