Responses

Updated: 2026-09-18

Overview

A compatible implementation of the OpenAI Responses API. Compared with Chat Completions it collapses multi-turn conversation, tool calls and reasoning into a single input/output shape, and lets the server carry the context for you via previous_response_id.

POST/responses

Request Parameters

Parameter	Type	Required	Description
model	string	Required	Model ID
input	string \| array	Required	The input. A string is a single turn; an array can mix messages, tool results and other items
instructions	string	Optional	System instructions, equivalent to the system message in Chat Completions
previous_response_id	string	Optional	The id of the previous response. Pass it instead of rebuilding the history yourself
stream	boolean	Optional	Stream the response. Defaults to false
max_output_tokens	integer	Optional	Output token limit
tools	array	Optional	Tool definitions, same shape as the Chat Completions tools field

Example

curl -X POST "https://www.starunion.net/v1/responses" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o-mini",
    "input": "Explain quantum entanglement in one sentence"
  }'

Differences from OpenAI

Only POST /responses (and /responses/compact) is implemented. The retrieval sub-routes GET /responses/{id}, DELETE /responses/{id}, POST /responses/{id}/cancel and GET /responses/{id}/input_items are not available yet.

previous_response_id still works (the upstream keeps that state), but you cannot fetch an earlier turn back by id afterwards. If you need the transcript, store each response on your side.

Didn't find what you were looking for?Contact us →