Connect Shopify to Self-Hosted AI on GPU

What You’ll Connect

After this guide, your Shopify store will have AI features powered by your own GPU server — no API costs, no rate limits. A custom Shopify app calls your vLLM or Ollama endpoint on dedicated GPU hardware to generate product descriptions, handle customer queries via an embedded chatbot, and analyse reviews automatically.

The integration uses Shopify’s App Bridge and Admin API alongside a lightweight backend that proxies AI requests to your self-hosted model. Merchants get AI capabilities within their Shopify admin without subscribing to expensive third-party AI tools.

Custom App UI –> App Backend (Node.js) –> GPU Server (vLLM) (App Bridge) (React embed) proxies AI requests LLM inference | | on dedicated GPU Merchant clicks Calls /v1/chat/ | “Generate Description” completions | | | | UI displays <-- App backend <-- JSON response <-- Model completion AI content returns text from GPU server returned -->

Prerequisites

• A GigaGPU server with a running LLM behind an OpenAI-compatible API (self-host guide)
• A Shopify Partner account and a development store for testing
• Node.js 18+ for the app backend (using Shopify CLI)
• HTTPS endpoint for your GPU server (Nginx proxy guide)
• Basic familiarity with Shopify app development and Remix or Express

Integration Steps

Scaffold a new Shopify app using the Shopify CLI: shopify app init. This creates a Remix-based app with authentication already configured. Your app runs as a separate server that Shopify loads within an iframe in the admin panel.

Add a route in your app backend that accepts a product ID, fetches product details via the Shopify Admin API, sends the product title and attributes to your GPU-hosted inference API as a structured prompt, and returns the generated description. The front end calls this route when a merchant clicks “Generate with AI” in the product editor.

For the storefront chatbot, create a Shopify ScriptTag or App Embed that injects a chat widget onto the store. The widget sends visitor messages to an app endpoint that proxies to your GPU model and streams responses back. This turns your store into an AI-powered customer support channel.

Code Example

This Express route in your Shopify app backend calls the GPU model to generate a product description using the OpenAI-compatible API:

import express from 'express';
import OpenAI from 'openai';
import { shopifyApi } from '@shopify/shopify-api';

const router = express.Router();
const llm = new OpenAI({
  baseURL: 'https://your-gpu-server.gigagpu.com/v1',
  apiKey: process.env.GPU_API_KEY,
});

router.post('/api/generate-description', async (req, res) => {
  const { productTitle, attributes, tone } = req.body;

  const prompt = `Write a compelling product description for an e-commerce store.
Product: ${productTitle}
Attributes: ${attributes}
Tone: ${tone || 'professional and persuasive'}
Keep it under 150 words.`;

  const completion = await llm.chat.completions.create({
    model: 'meta-llama/Llama-3-70b-chat-hf',
    messages: [
      { role: 'system', content: 'You are an expert e-commerce copywriter.' },
      { role: 'user', content: prompt }
    ],
    max_tokens: 400,
    temperature: 0.7,
  });

  res.json({ description: completion.choices[0].message.content });
});

export default router;

Testing Your Integration

Install the app on your development store using Shopify CLI’s shopify app dev command, which handles tunnelling and OAuth. Navigate to a product in the admin, trigger the AI generation, and verify a relevant product description appears. Compare the output quality across different product types — electronics, clothing, food — to ensure the system prompt handles diverse categories.

Test the storefront chatbot by visiting your dev store as a customer. Send a query and confirm the response returns within a reasonable time. Check your GPU server logs for both admin and storefront requests.

Production Tips

Shopify stores often need AI for repetitive tasks like writing descriptions for hundreds of products. Build a bulk generation feature that iterates through products tagged “needs-description” and processes them sequentially. Queue the requests to avoid overwhelming your GPU during catalogue imports.

For the customer-facing chatbot, implement response caching for frequently asked questions (shipping times, return policies). This reduces GPU load for common queries while keeping the AI available for unique questions. Rate-limit storefront requests per session to prevent abuse.

Merchants benefit from keeping product data and customer conversations private. Self-hosted AI on open-source models ensures no store data flows to external AI providers. Secure the connection with our API security guide, explore more tutorials, or get started with GigaGPU to add AI to your Shopify store.