Concepts

The objects you work with as a developer are organized around the workspace. Before you call anything, understand the object hierarchy and a few easily confused concept pairs.

Workspaces and the object hierarchy

A workspace is the main governance boundary for resources; it holds members, roles, data, workflows, and compute resources. You usually obtain and use object IDs along this hierarchy:

Workspace
├─ Members (linked to user accounts) and roles
├─ Catalog → database → table or volume → file
├─ Connector → import task → import run
├─ Workflow → workflow job → artifacts and data lineage
├─ Knowledge base → knowledge sources and semantic configuration
├─ Agent → session or automation task → agent artifacts
└─ Compute instances, data publications, and data subscriptions

Easily confused concepts

Direct model calls and model-service configuration. When an application needs model output, use the Genesis API or a compatible SDK. When a workflow, knowledge base, or agent needs a model, configure LLM model services in AI Studio. This configuration lets AI Studio select model services for internal use; it does not create a model-calling address that can replace the Genesis service address. To change the selection rules, see Configure model routing.

Definitions and run instances. A workflow definition, an import task, and an automation task are long-lived objects; a workflow job, an import run, and an agent task are run instances produced on each trigger. Saving or deploying only updates the definition — it does not mean anything has run. A successful submission does not mean execution finished: save the run instance ID and query its status and result. Deleting a definition or task does not roll back data already written.

User accounts and workspace members. A user account is the platform-level login identity; a workspace member is that account’s membership in a workspace, with a role and tags (see User Management). Removing a workspace member only ends that workspace relationship — it does not delete the platform account.

Data publications and data subscriptions. With Data sharing, one workspace publishes a database for others; another workspace then sets up access to it (see Data subscriptions). Deleting a subscription does not delete the source data; before deleting a publication, check existing subscriptions and downstream dependencies.

Next steps

Last updated on