Call XiuRouter with an SDK

Configure the OpenAI Python or JavaScript SDK and understand Chat Completions parameters and compatibility limits.

Use XiuRouter with an OpenAI-compatible SDK by setting its Base URL to https://router-api.xiu.ai/v1, providing a dedicated XiuRouter API key, and copying an exact model ID from the current catalog.

Before you start

Create a dedicated key using the quickstart.
Copy a model ID from Models and pricing.
Install the OpenAI SDK used by your project.

Keep the key out of source files, lockfiles, frontend bundles, and Git. These examples read environment variables:

export XIUROUTER_API_KEY="YOUR_XIUROUTER_API_KEY"
export XIUROUTER_MODEL="YOUR_MODEL_ID"

Python

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["XIUROUTER_API_KEY"],
    base_url="https://router-api.xiu.ai/v1",
)

response = client.chat.completions.create(
    model=os.environ["XIUROUTER_MODEL"],
    messages=[
        {"role": "user", "content": "Reply only: Python SDK connected"}
    ],
)

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

JavaScript

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.XIUROUTER_API_KEY,
  baseURL: "https://router-api.xiu.ai/v1",
});

const response = await client.chat.completions.create({
  model: process.env.XIUROUTER_MODEL,
  messages: [
    { role: "user", content: "Reply only: JavaScript SDK connected" },
  ],
});

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

Common parameters

Parameter	Purpose	Limits
`model`	Select the model	Required; copy the exact ID from the catalog or model list
`messages`	Supply conversation messages	Used by Chat Completions; roles and multimodal content depend on the model
`max_completion_tokens`	Limit output tokens	Preferred for newer models; some upstreams accept only `max_tokens`
`temperature`, `top_p`	Control sampling	Reasoning models may ignore or reject some sampling options
`stop`	Set stop sequences	Depends on model support
`tools`, `tool_choice`	Declare tools	Both the model and upstream must support tool calling
`response_format`	Request JSON or JSON Schema	Requires structured-output support
`stream`	Enable streaming	Use the direct API domain and parse SSE
`stream_options.include_usage`	Include usage near the end of a stream	Only compatible upstreams return it
`reasoning_effort`	Set reasoning effort	Only for reasoning models that accept the field

The gateway can parse these fields without every model accepting them. Keep the original error and request ID, remove optional parameters, then add them back one at a time.

Verify the result

Confirm that the program receives complete text, rather than only passing authentication.
In XiuRouter usage logs, check the model, service group, tokens, cost, and status.
Record the x-oneapi-request-id response header for troubleshooting.

Troubleshoot

Symptom	Next step
Requests still go to the official provider	Check Python’s `base_url` or JavaScript’s `baseURL`
`401 Invalid token`	Confirm that the running process can read `XIUROUTER_API_KEY`
Model unavailable	Fetch `GET /v1/models` again and check the key’s group and model scope
Parameter rejected	Keep only `model` and minimal `messages`, then retry
Text works but tools or JSON fail	Test tools, structured output, and streaming separately

Sources and review date

The Chat Completions route, request fields, and OpenAI Python/JavaScript custom Base URL settings were reviewed on 2026-08-19. This translation preserves that review date. Verify parameter support with the actual target model.