API reference

WorldRouter exposes an OpenAI-compatible /chat/completions endpoint. Any SDK or HTTP client that accepts a custom base URL can call it with no code changes. For image and video generation, see Media Generation.

Chat completions

Endpoint

POST https://inference-api-pre-d80ca3.worldrouter.ai/v1/chat/completions

Authentication

Pass your API key as a Bearer token in the Authorization header:

Authorization: Bearer your_api_key

Keys are scoped to your account. Create and rotate them in the API Keys dashboard.

Request body

Any other field in the OpenAI /chat/completions schema (top_p, stop, seed, response_format, …) is accepted unchanged.

Web search

WorldRouter can ground any model’s answer in live web results. There is no account-level toggle: search runs only when a request asks for it, and a request that doesn’t ask never incurs a search fee.

The simplest way to enable it is the :online model suffix:

{
  "model": "qwen3.5-flash:online",
  "messages": [
    { "role": "user", "content": "What changed in the latest release?" }
  ]
}

For control over how searching behaves, add the tool explicitly instead:

{
  "model": "qwen3.5-flash",
  "messages": [
    { "role": "user", "content": "What changed in the latest release?" }
  ],
  "tools": [
    {
      "type": "worldrouter:web_search",
      "parameters": {
        "max_uses": 3,
        "max_results": 5,
        "allowed_domains": ["worldrouter.ai"],
        "excluded_domains": ["example.com"]
      }
    }
  ]
}

The model decides when and how often to search: it may search, read the results, refine the query and search again, or answer without searching at all. Sources are cited as [1], [2], and so on.

Pricing

WorldRouter searches cost $0.007 each, covering the first 10 results; results beyond 10 add $0.001 apiece. Fees appear in your usage history alongside the model call that used them — a multi-hop answer shows one row per search round, each labelled with its round number. That is itemization, not duplicate billing. When engine resolves to a provider’s native search, the provider’s own search pricing applies instead.

If your balance cannot cover the request’s worst case (max_uses × the per-search fee), the request is rejected with 402 before anything runs.

Citations

When WorldRouter’s search ran, the response carries the sources it used — annotations on the message, in the same url_citation shape OpenAI uses, and a search counter under usage:

{
"choices": [
  {
    "message": {
      "content": "The latest release added ... [1]",
      "annotations": [
        {
          "type": "url_citation",
          "url_citation": {
            "url": "https://worldrouter.ai/changelog",
            "title": "Changelog",
            "content": "…excerpt from the page…",
            "start_index": 30,
            "end_index": 33
          }
        }
      ]
    }
  }
],
"usage": {
  "server_tool_use": { "web_search_requests": 1 }
}
}

Response

A non-streaming response matches the OpenAI shape:

{
"id": "chatcmpl-...",
"object": "chat.completion",
"created": 1738960610,
"model": "gpt-5.4",
"choices": [
  {
    "index": 0,
    "message": { "role": "assistant", "content": "Hello! How can I help you today?" },
    "finish_reason": "stop"
  }
],
"usage": {
  "prompt_tokens": 13,
  "completion_tokens": 9,
  "total_tokens": 22
}
}

Examples

curl https://inference-api-pre-d80ca3.worldrouter.ai/v1/chat/completions \
  -H "Authorization: Bearer your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4",
    "messages": [{ "role": "user", "content": "Hello" }]
  }'

from openai import OpenAI

client = OpenAI(
    api_key="your_api_key",
    base_url="https://inference-api-pre-d80ca3.worldrouter.ai/v1",
)

response = client.chat.completions.create(
    model="gpt-5.4",
    messages=[{"role": "user", "content": "Hello"}],
)

print(response.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "your_api_key",
  baseURL: "https://inference-api-pre-d80ca3.worldrouter.ai/v1",
});

const response = await client.chat.completions.create({
  model: "gpt-5.4",
  messages: [{ role: "user", content: "Hello" }],
});

console.log(response.choices[0].message.content);

Streaming

Set stream: true in the request body. The response becomes a Server-Sent Events stream. Each chunk follows the OpenAI chat.completion.chunk shape, and the stream terminates with a data: [DONE] line:

curl https://inference-api-pre-d80ca3.worldrouter.ai/v1/chat/completions \
  -H "Authorization: Bearer your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4",
    "stream": true,
    "messages": [{ "role": "user", "content": "Hello" }]
  }'

from openai import OpenAI

client = OpenAI(
    api_key="your_api_key",
    base_url="https://inference-api-pre-d80ca3.worldrouter.ai/v1",
)

stream = client.chat.completions.create(
    model="gpt-5.4",
    messages=[{"role": "user", "content": "Hello"}],
    stream=True,
)

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

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "your_api_key",
  baseURL: "https://inference-api-pre-d80ca3.worldrouter.ai/v1",
});

const stream = await client.chat.completions.create({
  model: "gpt-5.4",
  messages: [{ role: "user", content: "Hello" }],
  stream: true,
});

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

Error codes

See also

• OpenAI Chat Completions reference : WorldRouter mirrors this schema.
• Media Generation: async image tasks, plus Seedance, Kling, and PixVerse video.
• Async image tasks: create image-generation jobs, poll status, and receive stored image URLs.
• Seedance video guide: async video generation with /api/v3/contents/generations/tasks.
• PixVerse video guide: async video generation with /openapi/v2/video/....
• Models: full catalog with live pricing.
• Quickstart: API key, base URL, and first call.