AI Observability: Incident Manager Integration
Three module jobs register incidents related to Claude subscription limits in Incident Manager. An incident card shows the member's name and seat type, the state of the limit window, and the steps to take. It also links to the AI Observability: Claude Code - Member dashboard for that member over the last seven days.
| Job | When an incident is registered |
|---|---|
ai_observability_alert_limit_exhausted | the member has used the full limit of the five-hour or weekly subscription window |
ai_observability_alert_limit_weekly_80 | the member has used at least 80% of the weekly subscription window limit |
ai_observability_alert_limit_collector_silent | the member's limit snapshots were previously received but have not arrived for more than a day |
The jobs use snapshots from the subscription-limit collector and obtain the member's name and seat type from the claude_code_members lookup. Another incident for the same member is registered no more than once every six hours.
Incident Examples
No Subscription Limit Data
The ai_observability_alert_limit_collector_silent job registers an incident when a member's limit snapshots have not arrived for more than a day. The card shows the member's name, seat type, the number of hours without snapshots, and recommendations to check the status-line configuration and OTLP collector availability on the member's machine.
There are no incident-card examples for ai_observability_alert_limit_exhausted or ai_observability_alert_limit_weekly_80: their trigger conditions did not occur during module testing.