Responses

Workflow reference | Model configuration

Kind: Responses
Configuration object: model

Use this step for a Responses-compatible provider endpoint, including supported provider-native tools and response events.

Inputs

The Model input contract applies. Responses input can be a string or an array of input items. Put {{input}} in the request to map input explicitly. When there are no request placeholders, a nonempty rendered input is appended using the Responses input format.

Configuration

All Model configuration properties apply. This kind forces the effective protocol to Responses at invocation. Keep protocol set to Responses in the saved document and use a compatible endpoint.

Configure the request with input. Supplying both input and messages is rejected. Capture provider-native tool and structured-output settings from the supported playground. Runtime continuation belongs to the conversation runtime and is excluded from saved editor state.

Outputs

Returns the complete extracted response. Supported response deltas, usage, and native agent activity appear in Preview separately. Valid JSON response text is available to downstream expressions as typed JSON.

A compatible OpenAI Responses model with multi-agent mode enabled can also be referenced by the root workflow's specialized orchestration configuration. Start from a native orchestration template for that input and presentation contract.

Example

{
  "id": "respond",
  "name": "Answer with Responses",
  "kind": "Responses",
  "model": {
    "playground": "openai",
    "model": "my-chat-deployment",
    "protocol": "Responses",
    "endpoint": "https://example.openai.azure.com/openai/v1/responses",
    "credentialName": "MODEL_API_KEY",
    "authenticationHeader": "api-key",
    "requestBody": "{\"model\":\"my-chat-deployment\",\"input\":\"{{input}}\",\"stream\":true}"
  }
}

Use Multi-turn with a Responses child for conversation feedback and model-callable MCP or JIT handoffs.