Responses API reference
The Starfire-native Responses resource is designed for workloads that need richer Starfire capabilities than the compatibility chat endpoint can express.Resource
The Alpha v1 architecture includes:Response inputs
The native request model can carry concepts such as:- model or routing preference
- input/messages
- reasoning configuration
- tools
- structured-output requirements
- project or organization context
- streaming preference
Synchronous vs run-backed work
A normal response can complete within one request/stream. Work that can take materially longer—such as deep research or FORGE—should use a durable run/build resource instead of holding one request open indefinitely.Tools
Tool use is constrained by model capability, developer scope, plan, organization policy, and tool configuration. A request cannot grant itself a tool simply by naming it in the payload if the credential/context is not eligible.Usage
The response can contribute to the same Starfire metering foundation as web usage while remaining visible as Developer Platform usage.Request ID
Capture the request identifier returned by Starfire for support and log correlation.Structured output
When the active model/route supports schema-constrained output, validate the resulting object in your application before acting on it.Use the native Responses resource when you want Starfire’s platform semantics. Use compatibility endpoints when interoperability with an existing client contract is the primary goal.