Get startedDevelopers · Automation
Build with the SDK and REST API
Automate the same data-to-production lifecycle used in the Score Studio interface.
Source reviewed 2026-09-16. Availability depends on your installation, permissions, and compatible runtimes. A supported path is not a guarantee of model quality or production readiness.
Outcome
Create reproducible scripts, CI jobs, and external applications against workspace-scoped contracts.
Start here
What you need to know first
Check the inputs below before starting. If you are new, begin with the first-project guide. A dataset holds media and labels; a model produces results; a deployment makes a selected model version callable. Creating one does not create the others.
Bring these inputs
- — Workspace base URL and scoped bearer token
- — Validated request bodies with version IDs and idempotency strategy
Expected output and limits
- — Typed resources and durable task IDs
- — Structured errors suitable for retry and observability
Core workflow
Create an API key
Open a deployment → API quickstart → Create endpoint key to issue an inference-only credential restricted to that endpoint. Use a signed-in session for management routes and separate scoped keys for inference and MCP. Keys are not interchangeable with sessions and their secret is shown once.
Configure the client
Set the base URL and bearer token, then verify API status.
Call lifecycle primitives
Create datasets and versions, launch runs, evaluate artifacts, deploy, and query tasks.
Handle durable work
Poll or subscribe to task state. For inference, preserve request_id and model_version_id, retry only explicit transient statuses with a bound, and use the request ID to correlate failures with deployment logs.
What this surface supports
Python client with JSON responses
OpenAPI REST contract
Route-specific session or API-key authentication
Durable task endpoints
Expert section
Contracts, signals, and failure modes
Use this section when you are defining acceptance criteria, automating the surface, or reviewing whether its output is safe to promote downstream.
Quality and operating signals
- — HTTP status and error code
- — Rate-limit headers
- — Task state and request correlation ID
Common failure modes
- — Leaking tokens
- — Polling without backoff
- — Retrying mutations without idempotency
- — Using names where immutable IDs are required
Expert release checklist
□ Inputs and dependencies are pinned to immutable versions.
□ Acceptance metrics include critical classes and operating slices.
□ Failure, retry, cost, and rollback behavior are understood.
□ The resulting artifact has an owner and a downstream review path.
Engineering safeguards
- — Use least-privilege keys
- — Never embed secrets in clients
- — Treat asynchronous IDs as durable references