Responses

Install the SDK with pip install openai or npm install openai. The system prompt goes in instructions.

curl https://raon-api.krafton.ai/v1/responses \
  -H "Authorization: Bearer $RAON_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "raon-1-mini-preview", "instructions": "You are a helpful assistant.", "input": "Hello!"}'

The model's reasoning comes as a reasoning item in output, before the message item. When the answer reaches max_output_tokens, status is incomplete and incomplete_details.reason is max_output_tokens.

No stored responsesCopy Anchor Link

On OpenAI, store keeps a response on the server so a later request can continue from it with previous_response_id. Raon does not offer this: store is always false, and requests that refer to stored state are refused. Send the whole conversation in input on every turn, previous answers included.

See What is refused for the full list.