Create a stateless OpenAI Responses-compatible response

POST /v1/responses

Create a stateless OpenAI Responses-compatible response

cURL

curl --request POST \
  --url https://api.abliteration.ai/v1/responses \
  --header 'Content-Type: application/json' \
  --data '
{
  "model": "<string>",
  "input": "<string>",
  "instructions": "<string>",
  "stream": false,
  "store": true,
  "background": true,
  "previous_response_id": "<string>",
  "max_output_tokens": 499995,
  "max_tool_calls": 2,
  "temperature": 123,
  "top_p": 123,
  "presence_penalty": 0,
  "frequency_penalty": 0,
  "tools": [
    {}
  ],
  "parallel_tool_calls": true,
  "reasoning": {
    "enabled": true,
    "max_tokens": 499995,
    "exclude": true
  },
  "include_reasoning": true,
  "text": {
    "format": {
      "type": "text"
    }
  },
  "include": [],
  "metadata": {},
  "client_metadata": {},
  "prompt_cache_key": "<string>",
  "user": "<string>",
  "safety_identifier": "<string>",
  "policy_ids": [
    "<string>"
  ]
}
'

200

{
  "id": "<string>",
  "status": "completed",
  "model": "<string>",
  "output": [
    {}
  ],
  "object": "response",
  "usage": {
    "input_tokens": 1,
    "output_tokens": 1,
    "total_tokens": 1,
    "input_tokens_details": {},
    "output_tokens_details": {}
  }
}

Headers

Authorization · string | null

Bearer API key.

x-api-key · string | null

API key alternative for Anthropic-compatible clients. Authorization takes precedence when both headers are present.

Body

application/json

OpenAI Responses-compatible request.

model · string · required

Public model ID.

Minimum string length: 1

input · string Input · object[] · required

instructions · string | null

stream · boolean · default: false

store · boolean | null

background · boolean | null

previous_response_id · string | null

max_output_tokens · integer | null

Required range: 1 <= x <= 999990

max_tool_calls · integer | null

Required range: x >= 1

temperature · number | null

top_p · number | null

presence_penalty · number | null

Required range: -2 <= x <= 2

frequency_penalty · number | null

Required range: -2 <= x <= 2

tools · Tools · object[]

tool_choice · enum<string> Tool Choice · object

Available options: none, auto, required

parallel_tool_calls · boolean | null

reasoning · ResponsesReasoning · object | null

Native reasoning.effort plus the reviewed compatibility aliases.

The accepted profile decision requires Responses to take enablement, budget, and visibility aliases, not effort alone, so an OpenRouter- or Codex-shaped client resolves the same intent here that it already resolves on Chat Completions and Messages. max_tokens is the budget alias; the total output cap remains max_output_tokens.

Show child attributes

include_reasoning · boolean | null

text · ResponsesTextConfig · object | null

Show child attributes

include · enum<string>[] | null

Available options: message.input_image.image_url, message.output_text.logprobs, reasoning.encrypted_content, code_interpreter_call.outputs, computer_call_output.output.image_url, file_search_call.results

truncation · enum<string> | null

Available options: auto, disabled

metadata · Metadata · object | null

Show child attributes

client_metadata · Client Metadata · object | null

prompt_cache_key · string | null

service_tier · enum<string> | null

Available options: auto, default, flex, scale, priority

user · string | null

Maximum string length: 512

safety_identifier · string | null

Maximum string length: 64

policy_ids · string[]

Required string length: 1 - 120

Response

200

application/json

Successful Response

Native response object with required billing/public identity fields.

id · string · required

Pattern: ^resp_

status · enum<string> · required

Available options: completed, failed, in_progress, incomplete, queued

model · string · required

output · Output · object[] · required

object · string · default: response

Allowed value: "response"

usage · ResponseUsage · object | null

Show child attributes