Squobert

Squobert is a conversational AI agent built with Pipecat and running on Pipecat Cloud. He was featured at the 2025 AI Engineer World's Fair in San Francisco.

Squobert is powered by a Raspberry Pi and a Jabra 410 USB speaker/microphone. His face is a React app that runs in a fullscreen browser.

Running the bot

cd bot
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python bot.py -t daily

To deploy it to Pipecat Cloud, edit bot/build.sh and bot/pcc-deploy.toml and run:

cd bot
./build.sh
pcc deploy

There's also a GitHub Action in this repo to automatically build and deploy the Pipecat Cloud bot on pushes to main.

Running the client UI

cd client
npm i
npm run dev

Then add a Pipecat Cloud public key to client/src/providers/RTVIProvider.tsx.

The client code is a bit of a mess to make the touch interactions work the way I wanted them to, but you should be able to understand it well enough to do something useful with it if you want. Claude wrote most of it, after all. :)

Building your own

It's not for the faint of heart, but you totally can! Here's a parts list:

• Squishmallows Eden The Purple Bigfoot, 10 inch
• Raspberry Pi, or similar small-ish PC
• Jabra 410 Speakerphone. You're going to want one that has good built-in echo cancellation.
• Logitech C925 Webcam (or similar)
• A 7-inch HDMI touch screen like this one

You'll also need a decent 3d printer for some of the parts. All of the parts are available in this OnShape document. You might need to adapt the designs a bit to work with different hardware.

And be prepared to do some terrible things to a stuffed animal.