Operations

The Operations page is home base for keeping the pipeline healthy.

Connector health

Every connector shows status, last sync, last error (code + message + time) and its circuit-breaker state. A breaker opens after repeated failures and half-opens to probe recovery — so one flapping source can’t take down ingestion.

Dead-letter queue (DLQ)

Deliveries that fail to normalize land in the DLQ with a failure class and reason. Operators can:

Replay an item — reprocess it (idempotent; dedup keys prevent duplicates).
Discard a poison item that will never succeed — marks it resolved with an audit entry.

GET  /api/v1/admin/dlq
POST /api/v1/admin/dlq/{dlqId}/replay
POST /api/v1/admin/dlq/{dlqId}/discard

Backfill & replay

Re-pull a time window for any syncable connector:

POST /api/v1/admin/replay
{ "integrationId": "int_…", "since": "2026-01-01T00:00:00Z" }

Omit since for an incremental pull from the connector’s checkpoint. The scheduler also polls due connectors automatically on their configured cadence.

Environment drift

The drift card flags services whose latest artifact differs across environments — a signal that a rollout is stale or inconsistent.

Linking hygiene

A project-wide health report grades your evidence linking:

GET /api/v1/projects/{projectId}/linking-hygiene

It returns orphan rate, uncertain rate, heuristic reliance, unlinked-PR count and a recommendation — so you know where to tighten repo/branch mappings or curate.

Tenant usage & health

Org admins get usage counters and a health roll-up (GET /api/v1/orgs/{id}/usage and /health) plus RED metrics at /api/v1/admin/metrics (rate, errors, duration, outbox lag, DLQ depth).

Integrations guide · Notifications guide · Admin guide