5 Problems I Hit Running a WhatsApp AI Bot 24/7 on Windows - And How I Fixed Them

A developer documented five reliability problems encountered while running a WhatsApp AI bot 24/7 on a Windows PC, including process persistence, repeated QR-code authentication, local LLM inference speed, and unhandled errors. The fixes rely on PM2 for process management, whatsapp-web.js LocalAuth for persistent sessions, smaller Ollama models such as llama3.2:3b for faster replies, and try/catch error handling around every external call. The resulting architecture runs entirely on owned hardware with no VPS or paid AI API.

Running a WhatsApp AI bot locally sounds simple: WhatsApp → Node.js → Ollama → done. In practice, keeping it running reliably 24/7 on a Windows PC taught me a few things the hard way. Here are 5 problems I ran into — and how I fixed them. During development, running node index.js worked perfectly. But the moment the terminal was closed, the bot disappeared with it. Use a process manager such as PM2: npm install -g pm2 Then: pm2 start index.js --name whatsapp-ai-bot And save the process list: pm2 save This makes the bot much easier to manage and restart. Scanning a QR code every time the bot starts isn't practical for a machine that's supposed to run 24/7. With whatsapp-web.js , use persistent local authentication: const { Client, LocalAuth } = require 'whatsapp-web.js' ; const client = new Client { authStrategy: new LocalAuth } ; The authentication session is stored locally, so normal restarts don't require pairing the account again. A local LLM means no paid AI API, but your hardware now does the inference. Running a model that's too large can make a WhatsApp bot feel unusable. Start with a smaller model. For example: ollama run llama3.2:3b Then test response time before moving to something larger. For a chat bot, a fast smaller model can be more useful than a smarter model that takes forever to answer. Network requests fail. Ollama may temporarily be unavailable. Messages may contain unexpected content. Without error handling, one bad request can take down a bot that's supposed to run unattended. Treat every external operation as something that can fail. try { const response = await askOllama message.body ; await message.reply response ; } catch error { console.error 'Bot error:', error ; } Also log failures instead of silently ignoring them. This was an important distinction. The setup can avoid: But the Windows machine still needs to remain powered on and connected to the internet. For me, that's the useful part of this architecture: reuse hardware you already own instead of paying for another server every month. The architecture ended up being surprisingly small: WhatsApp → whatsapp-web.js → Node.js → Ollama → Local LLM No VPS. No paid AI API. No monthly hosting subscription. If you want the complete Windows setup, source code, configuration, and step-by-step instructions, I put everything together here: 👉 Complete WhatsApp AI Bot Setup Guide + Code If you're building something similar, I'd also be interested to hear what problems you ran into.