I built a free AI desktop pet for Windows powered by Llama 3.3 70B

I built BEBO the PET, a free open-source AI desktop companion for Windows.

The idea is simple: instead of opening a browser every time you need quick AI help, BEBO lives on your desktop as a small animated assistant. Click it, and a compact panel opens with focused writing tools: summarize, draft email, fix grammar, simplify text, humanize AI-written text, and ask general questions.

Website: https://bhavik8025.github.io/BEBO-the-PET/

GitHub: https://github.com/bhavik8025/BEBO-the-PET

Download: BEBO the PET v1.0.0 for Windows

Why I built it

I kept running into the same tiny productivity friction: I wanted AI help, but I did not always want to open a browser, switch tabs, sign in somewhere, paste text into a large chat window, and then switch back to what I was doing.

Most AI tools are powerful, but they still feel like a separate destination. I wanted something closer to a desktop utility: always nearby, fast to open, and focused on the small writing tasks people repeat every day.

That became BEBO.

What BEBO does

BEBO currently has six focused actions:

• 
  Summarize: turn long text into a clear short summary


• 
  Draft Email: turn notes or intent into a professional email


• 
  Fix Grammar: clean grammar, spelling, and punctuation


• 
  Simplify: rewrite complex text in plain language


• 
  Humanize: make robotic AI text sound more natural


• 
  Ask AI: use a general prompt when the fixed tools are not enough

It also supports keyboard shortcuts, a system tray menu, draggable positioning, and a resizable assistant panel.

The stack

BEBO is built with:

• 
  Electron for the Windows desktop app


• 
  Node.js for the main process and app logic


• 
  Vanilla HTML, CSS, and JavaScript for the pet and assistant panel


• 
  Groq API for fast inference


• 
  Llama 3.3 70B Versatile as the model behind the AI actions

I kept the frontend simple on purpose. For this kind of app, speed and reliability matter more than framework complexity.

How the app is structured

The app uses two main windows:

1. A transparent pet window that stays on the desktop


2. An assistant panel window that opens near the pet when needed

The pet communicates with the main Electron process through IPC. When the user clicks BEBO or uses a shortcut, the main process toggles the assistant panel. When the user runs an AI task, the panel sends the input to the main process, which calls the Groq API and returns clean plain text back to the UI.

That structure made the app feel more native and less like a webpage wrapped in a desktop shell.

A small UX decision I cared about

I did not want BEBO to become another giant chatbot window. The panel is intentionally focused around tasks people actually repeat:

• "summarize this"


• "make this email sound professional"


• "fix this paragraph"


• "explain this more simply"

That constraint helped keep the app lightweight.

What I learned

The biggest lesson was that desktop AI tools need a different rhythm from web AI tools. On the desktop, the assistant should feel fast, nearby, and low-friction. If opening the tool feels heavy, people will just go back to the browser.

I also learned that small product details matter a lot in Electron apps: panel positioning, always-on-top behavior, tray actions, keyboard shortcuts, and clean IPC boundaries all affect whether the app feels polished.

What I would like feedback on

I would love feedback from developers and Windows users on three things:

• Is the desktop-pet interface useful, or should the app become more minimal?


• Which AI actions should be added next?


• Should BEBO stay a lightweight utility, or grow into a larger productivity assistant?

The project is free and open-source, so issues, ideas, and pull requests are welcome.

Website: BEBO the PET

GitHub: bhavik8025/BEBO-the-PET