Responses

OpenAI-compatible This endpoint follows the OpenAI Responses API shape.

Request Parameters

model

string

required

Target model ID.

input

string | array

required

Text input, or an array of structured input items (official Responses API style).Supported forms:

string: plain text input
array: list of input items (supported type: message, function_call, function_call_output, reasoning)

Show Input item: message (type: "message")

Shape:

{ "type":"message", "role":"user|assistant|system|developer", "content":"string|array", "name":"optional" }

type (required): "message"
role (required): "user" | "assistant" | "system" | "developer"
name (optional): author name
content (optional): string or an array of typed parts

Show message.content[] parts (discriminator: type)

Show input_text

{ "type": "input_text", "text": "..." }

text (required): string

Show output_text

{ "type": "output_text", "text": "...", "annotations": [] }

text (required): string
annotations (optional): array (provider-specific)

Show input_image

{ "type": "input_image", "image_url": "https://..." }

image_url can be:

a URL string, or

an object:

{ "url": "https://...", "detail": "auto|low|high" }

detail defaults to "auto".

Show input_audio

{ "type": "input_audio", "input_audio": { "data": "<base64>", "format": "wav|mp3" } }

input_audio.data (required): raw base64 string (not a Data URI)
input_audio.format (required): "wav" or "mp3"

Show input_file (inline payloads only; gateway subset)

Inline:

{ "type":"input_file", "filename":"x.pdf", "file_data":"data:application/pdf;base64,..." }

Nested variant:

{ "type":"input_file", "input_file": { "filename":"x.pdf", "file_data":"https://..." } }

filename (optional): string
file_data (optional): Data URI (data:...;base64,...) or URL
input_file (optional): nested object with the same filename / file_data

Notes:

This gateway intentionally does not support file references like file_id for /v1/responses.

Show Input item: function_call (type: "function_call")

{ "type":"function_call", "call_id":"call_...", "name":"my_function", "arguments":"{\"k\":\"v\"}" }

call_id (required): string
name (required): string
arguments (optional): JSON string (defaults to "")

Show Input item: function_call_output (type: "function_call_output")

{ "type":"function_call_output", "call_id":"call_...", "output":"{\"result\":123}" }

call_id (required): string (must match the corresponding function_call.call_id)
output (optional): string (often JSON; defaults to "")

Show Input item: reasoning (type: "reasoning")

{ "type":"reasoning", "content": [] }

content (optional): array (provider-specific structured reasoning items)

instructions

string

A system (or developer) instruction inserted into the model’s context.

max_output_tokens

integer

Maximum number of output tokens to generate (minimum: 1).

temperature

number

Sampling temperature (0..2). Higher values make output more random.

top_p

number

Nucleus sampling probability (0..1). Alternative to temperature.

stop

string | string[]

Stop sequence(s). The model stops generating when any stop string is encountered.

logit_bias

object

Map of token IDs (as strings) to bias values. Provider/model dependent.

tools

array

Tool definitions (Responses API format). Tool availability depends on the model/provider.

Show Tool object types (discriminator: type)

Show function

{ "type": "function", "name": "my_tool", "description": "optional", "parameters": {}, "strict": true }

name (optional): string
description (optional): string
parameters (optional): object (JSON Schema for arguments)
strict (optional): boolean
function (optional): legacy nested function object (compat)

Show web_search

{ "type": "web_search", "max_results": 5 }

max_results (optional): integer

Show file_search

{ "type": "file_search" }

Show computer_use_preview

{ "type": "computer_use_preview" }

Show code_interpreter

{ "type": "code_interpreter" }

tool_choice

string | object

Tool choice strategy.

Show Supported shapes

"auto" | "none" | "required"

Force a function call:

{ "type": "function", "name": "get_weather" }

Legacy function forcing (Chat Completions-style):

{ "type": "function", "function": { "name": "get_weather" } }

Allowed tools (official Responses API shape):

{ "type": "allowed_tools", "mode": "auto|required", "tools": [ { "...": "..." } ] }

parallel_tool_calls

boolean

Enable parallel tool calls (if supported by the model/provider).

reasoning

object

Reasoning configuration (gateway subset that maps into Chat Completions).

Show properties

effort (optional): "minimal" | "low" | "medium" | "high"
summary (optional): "auto" | "concise" | "detailed"

text

object

Output formatting configuration.

Show properties

format (optional): object
- type (required): "text" | "json_object" | "json_schema"
- For type: "json_schema":
  - name (optional): string
  - description (optional): string
  - schema (optional): object (JSON Schema)
  - strict (optional): boolean

stream

boolean

default:"false"

Whether to stream the response.

Usage examples

from openai import OpenAI

client = OpenAI(
    base_url="https://api.naga.ac/v1",
    api_key="YOUR_API_KEY",
)

resp = client.responses.create(
    model="gpt-4o-mini",
    input="Hello!",
)

print(resp)

Response

Returns an OpenAI-compatible Responses API object.

API Guides

API Reference

Request Parameters

Usage examples

Response

API Guides

API Reference

​Request Parameters

​Usage examples

​Response

Request Parameters

Usage examples

Response