Responses подходит для клиентов и coding agents, которые используют протокол OpenAI Responses.
https://www.routeapi.ai/v1/responses
curl https://www.routeapi.ai/v1/responses \
-H "Authorization: Bearer $ROUTEAPI_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"gpt-4o","input":"Explain RouteAPI"}'
| Field | Type | Обяз. | Описание |
|---|
model | string | Да | Model ID |
input | string/array | Да | Input content |
instructions | string | Нет | System instruction |
stream | boolean | Нет | Streaming output |
tools | array | Нет | Tool definitions |
tool_choice | string/object | Нет | Tool selection policy |
temperature | number | Нет | Sampling temperature |
max_output_tokens | number | Нет | Maximum output tokens |
| Item | Chat Completions | Responses |
|---|
| Input | messages | input + instructions |
| Clients | Traditional SDKs | New protocol clients |
| Tool calls | Supported, depends on model | Supported, usually more expressive |
| Scenario | General chat | Coding agents and complex tasks |
- Use
/v1/responses when the client requires Responses.
- Use
/v1/chat/completions for traditional chat clients.
- Support depends on model capability.