Connector sync
Connectors are the intake valve of the flywheel. Beakr manages OAuth credentials and sync state -- and every sync ends with an ingestion log entry that ties new knowledge base pages back to the exact external items that produced them.
Beakr manages both OAuth credentials and sync state. The pipeline is five hops from trigger to knowledge base page.
Key models
Sync cadence and triggers
Connectors do not listen for changes from external providers. Syncs are triggered explicitly by users or by the system when a connector is first configured. Syncs can run when a connector is first configured, when a user explicitly starts a sync, on scheduled cadences such as nightly refreshes, and through provider webhook triggers where supported. Scope controls still determine what data can enter the system, so teams can balance freshness, cost, and permissions.
Each connector has a scope mode that determines what gets synced:
- All -- the connector syncs everything accessible via the OAuth grant. Used for broad-coverage providers like Slack where channel-level filtering happens inside the pipeline.
- Restricted -- the user specifies exactly which folders, channels, or projects to sync. Used for Drive, Confluence, and other providers where the total volume would be prohibitive.
Auth health monitoring
| Mechanism | Trigger | What it does |
|---|---|---|
| OAuth provider webhooks | Real-time | Webhook events fire when a connection's OAuth token is refreshed, revoked, or encounters an error. Beakr updates the connector's health status immediately. |
| Scheduled token refresh | Every 2 hours | Proactively refreshes tokens before they expire. Catches cases where provider webhooks are delayed or missed. |
Change detection and memory
The infrastructure for delta sync exists in the data model: each source record tracks a content fingerprint and last-modified timestamp. Delta tracking is not only a sync optimization; it is also part of the memory layer. The system can reason about what changed, when it changed, and which downstream pages, citations, graph edges, or agent answers may be affected.
- Each sync re-enumerates items in scope and compares source metadata against the previous source record.
- Deduplication at the source-record level prevents duplicate pages and keeps source-to-page lineage stable across syncs.
- Changed items can trigger targeted recompilation, provenance updates, and graph health checks instead of treating every sync as a fresh import.
Change-aware sync roadmap
Content-change webhooks
Automatic detection of file edits, new messages, and ticket updates from connected providers -- triggering re-sync without manual intervention.
Scheduled re-sync
Automatic periodic re-sync on a configurable cadence, keeping your knowledge base fresh without manual triggers.
Delta sync
Intelligent change detection that skips unchanged items, making re-syncs faster and more cost-efficient.