-
Notifications
You must be signed in to change notification settings - Fork 44
Open
Description
Context
When an AI agent is investigating a data quality issue, knowing that the upstream service had an incident last night is essential context. Incident history connects operational events to data assets.
Scope
New extractors for incident management platforms:
PagerDuty
- Extract incidents: title, status, severity, service, timeline
- Extract services and their escalation policies
- Extract on-call schedules
- Emit relationships linking incidents to affected services and data assets
OpsGenie
- Extract alerts and incidents: title, status, priority, tags
- Extract teams and on-call schedules
- Emit relationships where inferrable from tags or service mappings
Design Considerations
- Time-windowed extraction — only pull recent incidents (configurable lookback)
- Relationship inference: match incident service names to known assets
- Incident metadata should include resolution status and duration
Why
Operational context is critical for AI reasoning about data trust. "This table's freshness dropped" is more actionable when paired with "the producing service had a P1 incident 6 hours ago."
References
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
No labels