Endpoints
Responses
Create OpenAI Responses-compatible output with Capriole AI Responses models.
POST
Use this endpoint when your application, SDK, or agent expects the OpenAI Responses request and response format.
Send a standard Responses payload with
model and input. Capriole AI resolves latest aliases to the selected upstream model ID before upstream dispatch, forwards compatible fields such as instructions, tools, reasoning, previous_response_id, and stream, and returns the upstream JSON or SSE stream unchanged.
POST /v1/responses accepts openai-latest, openai/gpt-5.6-terra, openai/gpt-5.6-luna, openai/gpt-5.5, and openai/gpt-5.4-mini. Use openai-latest to let Capriole AI choose our recommended flagship OpenAI Responses model.
Claude and Google latest aliases are not supported on this endpoint. Grok 4.5 and GLM 5.2 use Chat Completions and are also not supported here. Use POST /v1/chat/completions when you want one OpenAI-compatible endpoint that can call every public model.
For simple Capriole-native text generation, use POST /v1/chat. Use POST /v1/responses when OpenAI Responses wire behavior is required.Authorizations
Use an API key created in the Capriole AI page. Send it as Authorization: Bearer sk-....
Body
application/json
Public Responses-compatible model identifier or OpenAI latest alias
Available options:
openai-latest, openai/gpt-5.6-terra, openai/gpt-5.6-luna, openai/gpt-5.5, openai/gpt-5.4-mini OpenAI Responses input string or item list.
Optional system-level instructions.
OpenAI Responses-compatible tool definitions.
OpenAI Responses-compatible reasoning options.
Stream the response as OpenAI Responses-compatible server-sent events.
Response
OpenAI Responses-compatible JSON response or SSE stream
OpenAI Responses-compatible response body.