Responses

Send input and instructions. Receive a response containing typed output items.

POST/openai/v1/responses

Create a response

Use your OpenAI key and model gpt-5.6-luna. For a simple request, input can be a string.

curl · whole response

curl "https://proxy.litechat.ai/openai/v1/responses" \
  -H "Authorization: Bearer $BUILD_OPENAI_KEY" \
  -H "Content-Type: application/json" \
  --data '{
    "model": "gpt-5.6-luna",
    "input": "Say hello in one sentence.",
    "max_output_tokens": 128,
    "reasoning": {"effort": "none"},
    "store": false
  }'

Read output items

Check status, then inspect output. A message item can contain an output_text part. Other items can represent reasoning or tool calls. In the Python SDK, response.output_text is a convenience accessor, not a top-level REST field.

Illustrative whole response · selected fields

{
  "id": "resp_example",
  "object": "response",
  "model": "gpt-5.6-luna",
  "status": "completed",
  "output": [{
    "id": "msg_example",
    "type": "message",
    "role": "assistant",
    "status": "completed",
    "content": [{"type": "output_text", "text": "Hello! How can I help?", "annotations": []}]
  }],
  "usage": {"input_tokens": 12, "output_tokens": 8, "total_tokens": 20}
}

IDs, text, and usage are illustrative. incomplete and failed are distinct from completed. A completed response can still require a tool action.

Common parameters

input and instructions

Use a string or input items, with optional instructions. Send the required conversation and tool history in subsequent requests. Stored response retrieval and previous_response_id are not supported.

max_output_tokens and reasoning

Set an output limit and an effort level. Use {"effort":"none"} to disable reasoning, as in the example.

text.format

Select text, JSON object, or JSON Schema output. Generated schema output is validated. Use local schema references.

tools, store, and background

Your backend executes function tools and returns results with the same call IDs. Hosted search and code execution are unavailable. Use false for storage and background options; stored and background responses are unsupported.

Streaming

Add "stream": true to the request body above and pass --no-buffer to curl. SSE events have a type; text increments use response.output_text.delta. Check the terminal response.completed, response.incomplete, or response.failed event and handle errors or interruptions. There is no [DONE] sentinel.

See the OpenAI Responses streaming reference.