The Ultimate Thinking Cap

This project will show you how to code and build a hat with the power of locally run AI that helps you enhance your ideas. Not only is it usefule, it's fashionable (if you like cubes)! We will use Ollama to pull the AI model tinyllama, then use vosk and a python script to take voice input and feed it to the AI. Finally, pyttsx3 will speak out the AI's result. The original code I found was created by iotstation on github, but I have modified the code a heavy amount because the original code wasn't working for me.

Electronics

1x Raspberry Pi and Raspberry Pi power cord

1x. MicroSD Card (16 GB and above)

1x Tiny USB Microphone (found here)

1x USB Speaker (found here)

3D Printed Parts (Designed in Tinkercad)

1x Cover.stl

1x Speaker Holder.stl

1x Bottom.stl

You will also need a string of some kind, preferably an elastic band.

Download the Raspberry Pi Imager and open it. You will need to flash your Raspberry Pi with Raspberry Pi OS 64-bit. Put your microSD card in your computer then select it from the choose storage menu. Make sure to apply OS customization settings and edit them to include a password and your wifi network information. Click Next and then Write. Once it's done, put the microSD card in your raspberry pi.

Once your RPI is plugged in, SSH into it using your terminal. Once you are in, run the command:

git clone https://github.com/jgame123/aionahat

Now we have to make a virtual environment. To do this, run the commands:

cd aionahat

python -m venv ai

source ai/bin/activate

Then create a folder called models using the command

mkdir models

For our AI to understand our voice, we need to download a vosk model. To do that, cd into the models directory then run the following commands:

wget https://alphacephei.com/vosk/models/vosk-model-small-en-us-0.15.zip

mv vosk-model-small-en-us-0.15.zip vosk-model-small-en-us.zip

unzip vosk-model-small-en-us-0.15.zip

rm vosk-model-small-en-us-0.15.zip

This will rename the vosk model then unzip the vosk zip and then delete the no longer needed zip. To install the rest of the prerequisites, cd back into the aionahat directory then run

pip install -r requirements.txt

sudo apt install espeak-ng libespeak1

That's all you need (except for the AI itself).

For our AI assistant we will be using Ollama, a tool that makes it unbelievably easy to run a number of AI models. First, install Ollama using the following command

curl -fsSL https://ollama.com/install.sh | sh

We will be using one of the smallest AI models because of our hardware limitations on the Rpi. One of these small models is called tinyllama. To install it run

ollama pull tinyllama

This will take a little bit, but once its done, the AI is up and ready to go. If you want to test it out, just run

ollama run tinyllama

This will open a chatbot in your terminal!

We want our AI assistant to be ready as soon as we plug our hat in. To do this, we need to make a bash script then run it at system start. Make sure you are in the home directory, then start by making the script:

nano myscript.sh

Then you want to copy this text into the terminal:

#!/bin/bash

cd aionahat

source ai/bin/activate

python main.py

Then press control and x and the same time, then the y button, then enter. This will save it, but to make it executable run this:

chmod +x myscript.sh

Now lets make this script run at system start up. To do this, type

sudo nano /home/pi/.bashrc

Next, use the down arrow key to scroll all the way to the bottom of the file. Press enter a few times then type

/home/pi/myscript.sh

Finally, press control and x together, then y, then enter to save it. Now let's begin building the hat! Run the following command to shut down the Pi, then unplug it.

sudo shutdown now

Begin by plugging in the speaker and microphone, then placing the Raspberry Pi into the holder. The USB ports will line up with the front, and the power port will line up with the hole on the side. Apply glue to the sides of the cover, then place it over the Raspberry Pi. It will fit snuggly, and once the glue dries it will be pretty stable. Take the speaker holder and glue it onto the top of the cover. Make sure that the slit for the cord is facing the USB ports. Once the glue is dry, take the Raspberry Pi power cord (make sure it's unplugged) and put it through the hole in the cover and into the Raspberry Pi. Next, place the speaker onto the box, letting the cord rest on the slit. You might need to wind the cord around the box a couple times for a tight fit. Finally, take your string and feed it through the two slots on either side of the Raspberry Pi holder and tie them in place.

Put on the hat and plug it in! The hat will ask you for a prompt, then you can ask it whatever you want. Let your imagination loose, and you ideas flow with the help of the ultimate thinking cap!