OpenAI SDK with ModelPilot

Use your existing OpenAI code with ModelPilot for intelligent routing and cost optimization.

Why Use OpenAI SDK?

Zero Code Changes

Your existing OpenAI code works as-is

Gradual Migration

Test ModelPilot without rewriting

Framework Support

Works with LangChain, LlamaIndex, etc.

Intelligent Routing

Automatic model selection and cost optimization

Installation

bash

npm install openai

Configuration

Basic Setup

javascript

const { OpenAI } = require('openai');

const client = new OpenAI({
  apiKey: process.env.MODELPILOT_API_KEY,    // Use mp_xxx key
  baseURL: `https://modelpilot.co/api/router/${process.env.MODELPILOT_ROUTER_ID}`
});

Critical Points

Use your ModelPilot API key (starts with mp_), NOT your OpenAI key
The baseURL must include your router ID in the path
The SDK automatically appends /chat/completions to the URL
Omit the model parameter - your router automatically selects the best model

Migration Guide

Before (Pure OpenAI)

javascript

const { OpenAI } = require('openai');

const client = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY
});

const completion = await client
  .chat.completions.create({
    model: 'gpt-5',
    messages: [
      { role: 'user', content: 'Hello!' }
    ],
    temperature: 0.7
  });

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

After (ModelPilot)

javascript

const { OpenAI } = require('openai');

const client = new OpenAI({
  apiKey: process.env.MODELPILOT_API_KEY, // ← Changed
  baseURL: `https://modelpilot.co/api/router/${process.env.MODELPILOT_ROUTER_ID}` // ← Added
});

const completion = await client
  .chat.completions.create({
    // NO model parameter - router selects!
    messages: [
      { role: 'user', content: 'Hello!' }
    ],
    temperature: 0.7
  });
); // ← New!

Migration Checklist

Create a router in the dashboard with your preferred models

Replace OPENAI_API_KEY with MODELPILOT_API_KEY

Add baseURL with your router ID

Remove model parameter - let router select automatically

Test with a simple request

Monitor model used and costs via _meta field

Basic Usage

Simple Completion

javascript

const completion = await client.chat.completions.create({
  messages: [
    { role: 'user', content: 'What is the capital of France?' }
  ]
});

console.log(completion.choices[0].message.content);
// Output: "The capital of France is Paris."

Streaming

javascript

const stream = await client.chat.completions.create({
  messages: [{ role: 'user', content: 'Tell me a story' }],
  stream: true
});

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

Framework Integration

LangChain

javascript

const { ChatOpenAI } = require('langchain/chat_models/openai');

const chat = new ChatOpenAI({
  openAIApiKey: process.env.MODELPILOT_API_KEY,
  configuration: {
    baseURL: `https://modelpilot.co/api/router/${process.env.MODELPILOT_ROUTER_ID}`
  },
  temperature: 0.7
});

const response = await chat.call([
  { role: 'user', content: 'Hello!' }
]);

Vercel AI SDK

javascript

import { OpenAIStream, StreamingTextResponse } from 'ai';
import { Configuration, OpenAIApi } from 'openai-edge';
import { CodeBlock } from '@/components/code-block';

const config = new Configuration({
  apiKey: process.env.MODELPILOT_API_KEY,
  basePath: `https://modelpilot.co/api/router/${process.env.MODELPILOT_ROUTER_ID}`
});

const openai = new OpenAIApi(config);

export async function POST(req) {
  const { messages } = await req.json();
  
  const response = await openai.createChatCompletion({
    messages,
    stream: true
  });
  
  const stream = OpenAIStream(response);
  return new StreamingTextResponse(stream);
}

Differences from Pure OpenAI

What's the Same ✓

• All request parameters
• Response format
• Streaming behavior
• Function/tool calling
• Error structure
• SDK methods

What's Different +

• Additional _meta field with cost, latency
• Router may select different model
• Environmental impact tracking
• Automatic fallback handling
• Multi-provider access

Next Steps