Skip to content

Services

A service is a concrete instance inside a domain. Initializing a service records the service, resolves features, and can start assignment-linked inference.

/v2/clients/{client_id}/domains/{domain_id}/service/{service_id}

Service initialization accepts caller-provided features, validates them against the domain feature definitions, applies defaults, stores the resolved feature set, and returns identifiers for follow-up work.

sequenceDiagram
    participant Caller
    participant API
    participant DomainAssignment
    participant Job

    Caller->>API: POST /service/init
    API->>API: Validate and resolve features
    API->>DomainAssignment: Run initialization assignment
    DomainAssignment->>Job: Create inference job
    API-->>Caller: service_id and job_id
  • Provided features are exactly what the caller supplied.
  • Resolved features are the values after defaults from JSON Schema are applied.

Use resolved features when reasoning about what the model actually saw.

Service assignments expose stable keys for work scoped to one service. The key is stable; the active model version behind it can change over time. Running a job through the service assignment preserves the assignment context on the durable job record.

Service status tells callers whether the service is still initializing, available, failed, or not found. Workflows that need model output should poll the returned job rather than infer completion from service status alone.