Build a 30‑Line Node.js WhatsApp Chatbot with GPT‑4.1 and Heltar in 15 Minutes

Turn WhatsApp into a button‑driven AI in less time than a coffee run. This post walks you through wiring GPT‑4.1 to Heltar’s WhatsApp API using just JavaScript and ~30 lines of Express code. No bloated frameworks, no endless setup—just a solid API, a few lines of code, and a working feature by the time your coffee cools.

This guide will walk you through building a fully functional chatbot that:

• Gets messages from WhatsApp via Heltar
• Uses GPT-4.1 to decide what button to show next
• Sends a reply straight to the user

It’s fast, clean, and surprisingly powerful. Let’s get into it.

What You’ll Need?

Here’s the quick stack:

​If you don't have an account on Heltar you can set up an account for free by following this tutorial: 

1. Package Installation & Setup

Install everything in one go:

npm i openai express axios dotenv

Set up your .env file:

OPENAI_API_KEY=sk-...
HELTAR_TOKEN=ht_...

2. Tell GPT‑4.1 What a “Button” Means

To keep replies predictable, we’ll use structured outputs (OpenAI function calling). Here’s a simple schema GPT‑4.1 will follow:

const quickReply = {
  name: 'quick_reply',
  description: 'Return exactly one of the provided buttons',
  parameters: {
    type: 'object',
    properties: {
      choice: { type: 'string', enum: ['Approve', 'Reject'] }
    },
    required: ['choice']
  }
};

This ensures GPT doesn’t hallucinate—it sticks to our defined buttons.

3. The 30‑Line Express Server

Here’s the core of your bot:

// index.js
import express from 'express';
import axios from 'axios';
import 'dotenv/config';
import OpenAI from 'openai';
const openai = new OpenAI();
const app = express();
app.use(express.json());
const HISTORY = [];
app.post('/whatsapp', async (req, res) => {
  const { text: userMsg, chat_id } = req.body;
  HISTORY.push({ role: 'user', content: userMsg });
  const chat = await openai.chat.completions.create({
    model: 'gpt-4.1-chat-bison',
    messages: HISTORY,
    functions: [quickReply],
    temperature: 0.3
  });
  const reply = chat.choices[0].message;
  let payload;
  if (reply.function_call) {
    const labels = quickReply.parameters.properties.choice.enum;
    payload = {
      chat_id,
      text: 'Pick one:',
      buttons: labels.map(l => ({ type: 'reply', label: l, payload: l }))
    };
  } else {
    payload = { chat_id, text: reply.content };
  }
  HISTORY.push({ role: reply.role, ...reply });
  await axios.post('https://api.heltar.io/v1/messages', payload, {
    headers: { Authorization: `Bearer ${process.env.HELTAR_TOKEN}` }
  });
  res.send('ok');
});
app.listen(3000, () => console.log('Server running on port 3000'));

This is your bot’s brain and voice, all in one file.

4. Test It Out

Start the server:

node index.js

Then expose it with Ngrok:

ngrok http 3000

Point your Heltar webhook to https://<your-ngrok-id>.ngrok.io/whatsapp, message your sandbox number, and you should see buttons pop up immediately.

Why This Works So Well?

Structured outputs (function calling) solve a ton of problems:

• Reliable: GPT sticks to the schema you give it—no surprises.
  

• Clean: No need to parse text or use fragile regex.
  

• Flexible: Easy to chain into multi-step flows or menus.
  

Next Steps You Can Build Right Now

• Add more button values to build nested menus.
  

• Include image URLs—WhatsApp previews them automatically.

Got Stuck?

Ping us anytime—our engineers at Heltar love debugging and will get you unblocked quickly.​

You’re Done!

In under 15 minutes, you’ve wired GPT‑4.1 into WhatsApp using Heltar and built a working chatbot that replies with dynamic buttons—all in one file.

Clone it. Tweak it. Make it yours.

For more smart WhatsApp Solutions, check out heltar.com/blogs.