Create a ChatCompletion

post

https://api.vcita.biz/v3/ai/chat_completions

Overview

Create chat completions using multiple LLM providers (OpenAI, Anthropic, Google). Supports text and multimodal input (images, audio, video, files), streaming responses via SSE, async mode for long-running thinking models, tool calling (function calling), and structured output (JSON schema).

Supported Models

Provider	Models
OpenAI	`openai/gpt-5`, `openai/gpt-4o`, `openai/gpt-4o-mini`, `openai/o3-mini`, `openai/o1`
Anthropic	`anthropic/claude-sonnet-4-5-20250929`, `anthropic/claude-haiku-4-5-20251001`, `anthropic/claude-3-5-sonnet-latest`, `anthropic/claude-3-5-haiku-latest`
Google	`google/gemini-2.5-pro`, `google/gemini-2.0-flash`, `google/gemini-2.0-pro`

Streaming

By default, responses are streamed as Server-Sent Events (SSE). Set stream: false for a single JSON response.

Async Mode

Set async: true to queue the request and receive a run UID. Poll GET /v3/ai/chat_completion_runs/{uid} for the result. Recommended for thinking models.

Available for Staff tokens

Recent Requests

Time	Status	User Agent
Retrieving recent requests…

Loading…

Body Params

model

string

required

Model identifier in provider/model format. Supported providers: openai, anthropic, google (e.g., "openai/gpt-4o", "anthropic/claude-sonnet-4-5-20250929", "google/gemini-2.5-pro").

messages

array of objects

required

Array of messages representing the conversation history. Must include at least one message with role 'user'.

messages*

stream

boolean

Defaults to true

Whether to stream the response as Server-Sent Events (SSE). When true, the response is delivered incrementally as text/event-stream. Defaults to true.

async

boolean

Defaults to false

Process the request asynchronously. Returns a run UID immediately that can be polled via GET /v3/ai/chat_completion_runs/{uid}. Recommended for thinking models (e.g., openai/o1, openai/o3-mini) that may take longer to respond.

temperature

number

0 to 2

Sampling temperature between 0 and 2. Higher values (e.g., 0.8) make the output more random, while lower values (e.g., 0.2) make it more focused and deterministic.

max_tokens

integer

≥ 1

Maximum number of tokens to generate in the completion. The total length of input tokens and generated tokens is limited by the model's context length.

top_p

number

0 to 1

Top-p (nucleus) sampling parameter between 0 and 1. The model considers the results of the tokens with top_p probability mass. For example, 0.1 means only tokens comprising the top 10% probability mass are considered.

tools

array of objects

Array of tool definitions available for the model to call during the completion. The model may choose to call zero or more of these tools based on the conversation context.

tools

response_format

object

Response format specification for structured output. Use this to constrain the model's output to a specific format such as JSON. When not specified, the model returns free-form text.

stop

Up to 4 sequences where the model will stop generating further tokens. The returned text will not contain the stop sequence.

Headers

string

enum

Defaults to application/json

Generated from available response content types

Allowed:

Responses

200Successful non-streaming completion.

202Async mode: job queued successfully. Poll the returned run UID for the result.

400Bad Request - The request body is malformed or contains invalid syntax.

401Unauthorized - The bearer token is missing, expired, or invalid.

422Unprocessable Entity - Validation error with valid syntax.

429Too Many Requests - Rate limit exceeded (10 requests per minute).

502Bad Gateway - The upstream LLM provider returned an error.