How to Build a Multi-Model AI App in 10 Minutes (With Code)

#opensource

Have you ever wanted to use different AI models for different tasks — fast for simple queries, reasoning for complex problems, vision for images — without managing multiple API keys?

Here's how to build a multi-model app in 10 minutes.

The Concept

Instead of picking one model, you route each request to the model that fits best:

Simple chat → Fast model ($0.30/M)
Code review → Code model ($0.80/M)
Image analysis → Vision model ($5.00/M)
Complex logic → Reasoning model ($6.00/M)

Step 1: Install the SDK

npm install openai

Step 2: Set Up the Client

import OpenAI from 'openai';

const dubhe = new OpenAI({
  baseURL: 'https://dubhehub.com/v1',
  apiKey: process.env.DUBHE_API_KEY
});

Step 3: Route Requests by Type

async function routeRequest(task, content) {
  const modelMap = {
    'chat': 'dubhe-fast',
    'code': 'dubhe-code',
    'reason': 'dubhe-reasoning',
    'vision': 'dubhe-vision'
  };

  return await dubhe.chat.completions.create({
    model: modelMap[task] || 'dubhe-fast',
    messages: [{ role: 'user', content }]
  });
}

Step 4: Use It

// Simple chat — costs ~$0.00022
const chat = await routeRequest('chat', 'Explain what Docker is');

// Code review — costs ~$0.0028  
const review = await routeRequest('code', 'Review this function for bugs: ...');

// Deep reasoning — costs ~$0.009
const analysis = await routeRequest('reason', 'Analyze this business case...');