Skip to main content

Concepts

AI Cloud uses the same vocabulary in the UI, CLI, SDKs, and APIs.

Hierarchy

ConceptWhat it means
TenantA customer or organizational workspace.
ProjectA scoped workspace for workloads, access, and usage attribution.
RuntimeA launched compute node, app, notebook, scheduler, or endpoint.
AppA packaged runtime experience that can be launched repeatedly.
SKU or profileA capacity shape such as GPU type, slice, image, region, or runtime family.
MemberA user with tenant or project access.
Service accountAn automation identity scoped to a project or tenant purpose.
Entitlement or quotaA policy that limits what can be launched or consumed.
Correlation IDA trace handle used by support when something fails.

For ownership, roles, billing groups, service accounts, and access decisions, use Resource Hierarchy And Roles.

Tenant And Project

Most actions happen in a selected tenant and project. Before launching work or changing access, confirm the top bar shows the tenant and project you intend to use.

Project context matters because it controls:

  • who can see or manage the runtime;
  • where usage is attributed;
  • which quota and entitlement apply;
  • which service accounts and storage references are available;
  • which support owner should respond to a blocked state.

Runtime Lifecycle

The product should always show the current lifecycle state and the next safe action. If a state is blocked, capture the correlation ID before escalating.

Access Model

Access combines identity, role, project membership, account security, and runtime-specific connect permissions.

Access areaExamples
Account securityMFA, sessions, SSH keys, API keys
Project membershipmember, viewer, admin, owner
Automationservice accounts and scoped credentials
Runtime connectionSSH, terminal, notebook, route, endpoint

If an access path is blocked, use Troubleshooting Playbook to decide whether the owner is the user, tenant admin, app owner, or platform support.

Product Boundary

Public user and builder docs explain how to use AI Cloud. Protected engineering docs cover implementation, deployment, security evidence, provider operations, and internal runbooks.