Magic observability: Sentry, Datadog, Linear, Jira, GitHub
Canon: kb/167_unified_telemetry_observability.md
Configuration, not one-off glue
Observability profiles treat sinks as org-level configuration: vault-backed credentials, webhook_subscriptions rows, and a test ping so you know delivery works before production traffic.
What you can wire
| Class | Examples |
|---|---|
| APM / logs | Datadog, Sentry-style HTTPS |
| Issue trackers | Linear, Jira, GitHub Issues (connector + wf-2; fingerprint dedupe on incident links) |
| Generic | Signed or HMAC external_webhook |
Tenant vs operator
Tenant-safe subscriptions fan out customer-relevant patterns. platform_internal subscriptions (strict ACL) carry telemetry.platform.* and capacity signals — tenants never list those rows.
Progressive disclosure
Start with telemetry.run.failed, incident, and escalation signals — not the full firehose — then widen event_pattern when needed.
Series
- Previous: Unified telemetry overview
- Next: Partner telemetry