Skip to main content
Workspaces connect to external systems through connections. A connection holds a provider, its credentials, and — for storage providers — a sync schedule. There are two kinds:

Object Storage

s3, azure, gcs — sync files in and out of a workspace

LLM Inference

openai, anthropic, ollama — back AI-driven detection

Creating a Connection

The body carries a displayName, a typed config, and optionally isActive and a sync schedule. The provider is inferred from the config.
ollama requires baseUrl rather than credentials — it is the self-hosted option, so there is no key to supply. openai and anthropic accept an optional baseUrl for gateways and compatible endpoints.

Storage Configuration

LLM Configuration

Verifying a Connection

Check reachability before relying on a connection:
The response contains reachable and, on failure, an error.
Verify after any credential rotation. A connection that stops being reachable will surface as failed syncs rather than an obvious error.

Syncing Files

Storage connections can sync on a schedule or on demand.

Schedule

deletionPolicy: "delete" removes workspace files when their remote counterpart is gone. Use ignore unless the remote is authoritative.

Sync Status

A sync record reports status, triggerType, recordsSynced, attempt, and errorMessage on failure. Triggers are manual, scheduled, or webhook. Cancel a running sync with:

Webhooks

Register endpoints to receive events as runs and syncs progress:
Use the test endpoint to confirm delivery before depending on it.

Managing Connections

List endpoints accept a provider filter, and workspace-wide syncs also accept status. Set isActive: false to pause a connection without deleting it — scheduled syncs stop, and the configuration is retained.

Next Steps

API Reference

Connection and sync endpoints

TypeScript SDK

connections, syncs, and webhooks