OpenAI Compatible API

Use this route when you want the fastest path to production or when your application already speaks the OpenAI Chat Completions format.

Base URL

https://gateway.api.anyint.ai/openai/v1

Published route

POST /chat/completions

Full URL:

https://gateway.api.anyint.ai/openai/v1/chat/completions

Authentication

Authorization: Bearer <ANYINT_API_KEY>
Content-Type: application/json

When to use this route

You already use the OpenAI Python or JavaScript SDK
You want one stable chat entrypoint for multiple model families
You want SSE streaming with minimal integration work

If you need Gemini-native request bodies or Anthropic-specific guides such as token counting, use those provider-compatible pages directly.

If a client specifically requires OpenAI Responses API wiring, use OpenAI Responses API. For embeddings, image generation, audio, files, batches, assistants, threads, and videos, see OpenAI Extended Endpoints.

Core request shape

Field	Type	Required	Meaning	Example
`model`	string	Yes	Model ID to use for the request. Fetch this from Models API instead of guessing.	`claude-sonnet-4-6`
`messages`	array	Yes	Ordered conversation turns in OpenAI Chat Completions format.	`[{"role":"user","content":"Hello"}]`
`stream`	boolean	No	Set `true` to receive SSE chunks; omit or set `false` for a single JSON response.	`true`
`max_tokens`	integer	No	Maximum number of output tokens the model should generate.	`512`
`temperature`	number	No	Controls randomness. Lower values are more deterministic; higher values are more varied.	`0.7`
`top_p`	number	No	Nucleus sampling control. Use either `temperature` or `top_p` tuning first; avoid changing both without a reason.	`1`
`stop`	string or array	No	Stop sequence or sequences that end generation early.	`["\nUser:"]`
`tools`	array	No	OpenAI-compatible tool definitions for model tool calling when supported by the selected model.	`[{"type":"function",...}]`
`tool_choice`	string or object	No	Controls whether the model may call tools, must call a specific tool, or should not call tools.	`"auto"`
`presence_penalty`	number	No	Penalizes new tokens based on whether they already appeared in the text.	`0`
`frequency_penalty`	number	No	Penalizes repeated tokens based on frequency.	`0`
`user`	string	No	End-user identifier for your own abuse monitoring or request tracing. Do not send private personal data unless your policy allows it.	`user_123`

The current published schema documents the minimum request body needed to get a completion working. If your client already uses other OpenAI-compatible fields, validate them against the target model before relying on them in production.

Message fields

Field	Type	Required	Meaning	Example
`role`	string	Yes	Message author. Common values are `system`, `user`, `assistant`, and `tool`.	`user`
`content`	string or array	Usually	Text content or multimodal content blocks when supported by the selected model.	`Explain AnyInt in one sentence.`
`name`	string	No	Optional participant name used by some OpenAI-compatible clients.	`planner`
`tool_call_id`	string	Tool responses only	Associates a tool response with a prior tool call.	`call_123`

Response fields

Field	Meaning
`id`	Response identifier returned by the compatible API path
`object`	Response object type, such as `chat.completion` or `chat.completion.chunk`
`created`	Unix timestamp for response creation
`model`	Model that handled the request
`choices[]`	Generated message or stream delta choices
`choices[].message.content`	Generated text for non-streaming responses
`choices[].delta.content`	Incremental text for streaming responses
`usage.prompt_tokens`	Input tokens counted by the provider when available
`usage.completion_tokens`	Output tokens counted by the provider when available
`usage.total_tokens`	Total tokens counted by the provider when available

cURL example

curl https://gateway.api.anyint.ai/openai/v1/chat/completions \
  -H "Authorization: Bearer $ANYINT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "stream": true,
    "messages": [
      {"role": "system", "content": "You are a concise assistant."},
      {"role": "user", "content": "Explain AnyInt in one sentence."}
    ]
  }'

Python example

from openai import OpenAI

client = OpenAI(
    base_url="https://gateway.api.anyint.ai/openai/v1",
    api_key="your-anyint-api-key",
)

response = client.chat.completions.create(
    model="claude-sonnet-4-6",
    stream=True,
    messages=[
        {"role": "system", "content": "You are a concise assistant."},
        {"role": "user", "content": "Explain AnyInt in one sentence."},
    ],
)

for chunk in response:
    delta = chunk.choices[0].delta.content
    if delta:
        print(delta, end="")

JavaScript example

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://gateway.api.anyint.ai/openai/v1",
  apiKey: process.env.ANYINT_API_KEY,
});

const stream = await client.chat.completions.create({
  model: "claude-sonnet-4-6",
  stream: true,
  messages: [
    { role: "system", content: "You are a concise assistant." },
    { role: "user", content: "Explain AnyInt in one sentence." },
  ],
});

for await (const chunk of stream) {
  const delta = chunk.choices?.[0]?.delta?.content;
  if (delta) process.stdout.write(delta);
}

Response behavior

With stream: true, the route returns SSE chunks
The published schema uses chat.completion.chunk style objects
Each chunk can contain choices[0].delta.content
The stream ends with [DONE]
Final usage data can arrive near the end of the stream

Common mistakes

Using x-api-key instead of Authorization: Bearer
Hardcoding a model ID before checking Models API
Treating a partial stream chunk as the final answer
Mixing provider-native request fields into the OpenAI payload without validation

OpenAI Compatible API

On this page