Factory
What a Factory is, the states it moves through, and how it relates to a project.
A Factory is a project's deployable agent configuration. It packages an agent manifest into immutable revisions and runs them for the project.
Factory and project
A Factory belongs to exactly one project. The project's status owns the
archived state, so a non-active project reports its Factory as archived.
Revisions
Each deployment creates an immutable revision with a revision number. The active revision is the one the Factory runs.
Lifecycle
A Factory has four states:
draft- deployed but not activated.active- an active revision is admitted and running.paused- admission is paused while the revision is retained.archived- the owning project is not active; history stays readable.
The available commands are deploy, activate, pause, resume, and
archive:
deploycreates a revision and leaves a new Factory indraft; deploying another revision preserves an existing active revision and its state.activateactivates the deployed revision. When the Factory ispaused, activation retains that state; useresumeto admit the new revision.pausemoves anactiveFactory topaused.resumemoves apausedFactory back toactive.archivearchives the owning project. This makes the Factoryarchivedand terminal; its history remains readable.
Runtime bindings
channelsbinds the Factory to its channel selections.connectionsbinds the Factory to its agent connections.schedulesbinds scheduled work to the Factory.workflowoptionally binds the Factory to a Workflow.
These runtime bindings are set when a revision is activated.
Endpoints
The Factory endpoint family is
/v1/teams/:teamId/projects/:projectId/factory/*. Reads require
automation:read and the factory.read action grant, while mutations require
automation:manage plus the matching Factory action grant. See the API
overview for the action grants.
Next steps
- Deploy and activate a Factory - take a local directory to an active revision
- Workflow and Run lifecycle
- Core concepts
- API overview