Responses API | LLMoxy Docs

LLMoxy provides an OpenAI Responses-compatible route:

POST /v1/responses

If your application already uses the Responses API structure, or needs fields such as input, instructions, reasoning, and previous_response_id, you can use this endpoint.

Basic Request

curl https://llmoxy.com/v1/responses \
  -H "Authorization: Bearer <LLMOXY_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o-mini",
    "input": "Explain gateway routing in one sentence.",
    "instructions": "Be direct."
  }'

Common Fields

Field	Description
`model`	Required. Model ID.
`input`	String or message-style array.
`instructions`	System-level instruction text.
`max_output_tokens`	Output token limit.
`temperature`, `top_p`	Sampling control parameters.
`stream`	Enable streaming when supported by upstream.
`tools`, `tool_choice`	Tool calling parameters.
`reasoning.effort`	Supported reasoning models can use `low`, `medium`, `high`.
`reasoning.summary`	Request reasoning summary when supported by upstream.
`previous_response_id`	Continue the previous response when supported by upstream.
`truncation`	`auto` or `disabled`.

This endpoint is for compatibility. Actual behavior depends on the selected model.

Conversation Compaction

LLMoxy also provides:

POST /v1/responses/compact

Supported models can compress long conversation history through this endpoint. Authentication is the same as other /v1 requests.