💬 chat.aibiz · install guide

Up and running in about 5 minutes

One download, one double-click. The launcher installs everything else for you — Node.js and the free local AI — no terminal, no coding.

Download host (.zip) ↓

Windows 10/11 · macOS · Linux — free, ~27 KB zip (the AI model downloads separately, once)

1Download & unzip

Grab chataibiz-host.zip and extract it anywhere — your Desktop is fine. You'll see start.bat, start.sh, server.js, and a README.

2Run the launcher

Windows: double-click start.bat.  Mac/Linux: open Terminal in the folder and run sh start.sh.

  • ✓ Missing Node.js? It installs it for you (winget on Windows, Homebrew on Mac).
  • ✓ Missing Ollama (the free local AI)? Installed for you too.
  • ✓ First run downloads the AI model — one time, about 2 GB. Grab a coffee.

If the launcher installs something, it will ask you to close the window and double-click it once more — that's normal (Windows needs a fresh window to see new programs).

3Open your two links

  • 🛍️ Customer chat — http://localhost:8801 — what your visitors see.
  • 🛠️ Your admin inbox — http://localhost:8802 — watch every conversation, click take over to reply as a human. The console prints your sign-in token on first run (saved in chat-host.config.json).

4Go public (optional)

Customers on the internet can't reach localhost. Two options: a free ngrok tunnel (ngrok http 8801), or a managed you.chat.aibiz tunnel on the Pro plan — your own name, no third-party URL.

If something gets in the way

Windows says "Windows protected your PC" — that's SmartScreen being cautious about a new file. Click More info → Run anyway.
Mac says the file "cannot be opened" — run it from Terminal instead: cd into the folder, then sh start.sh.
"winget is not recognized" (older Windows 10) — install App Installer from the Microsoft Store, or install Node.js and Ollama by hand, then double-click start.bat again.
The AI replies slowly or not at all — the model needs ~8 GB of RAM to be comfortable. Chat still works without the AI: messages simply wait in your admin inbox for a human reply (nothing is lost).
Model download is slow — it's ~2 GB, one time only. Leave the window open; it resumes where it left off if interrupted.
Something already uses port 8801/8802 — edit chat-host.config.json and change the ports, then run the launcher again.

← back to chat.aibiz · see the live demo