Orchestration for Forge connectors: schedule and fan out recurring ingestion tasks

Forge connectors built on the @forge/teamwork-graph SDK can now use Orchestration - a platform-managed scheduling and execution layer that runs background ingestion work without an external scheduler or cron.

More details

Instead of processing an entire data source in a single web request, you describe the work as a tree of tasks and hand them to Teamwork Graph. The platform then schedules your taskRunner function on a configurable recurring cadence (minutes, hours, or days), fans out work via child tasks, retries based on structured failure reasons.

Key capabilities:

• graph.scheduleOrUpdateTask- Schedule a recurring root task per connection, or update an existing schedule in place.
• graph.scheduleChildTask - Fan out from a running task into smaller per-item child tasks with persistent metadata.
• graph.updateTaskStatus - Report success or failure with structured failureReason values (UNAUTHORIZED, RATE_LIMITED, AUTH_TOKEN_EXPIRED, etc.) so the platform can respond intelligently.
• Long-running execution - Opt into a 15-minute execution budget via timeoutSeconds: 900 in manifest.yml.

For a step-by-step walkthrough with a Google Drive worked example, see https://developer.atlassian.com/platform/teamwork-graph/orchestration-for-forge-connectors/. For the underlying concepts (IDs, idempotency, lifecycle), see https://developer.atlassian.com/platform/teamwork-graph/orchestration-concepts/. For the SDK method reference, see https://developer.atlassian.com/platform/teamwork-graph/connector-reference/task-operations/overview/.