PotatOS

Network Setup

The PotatOS system consists of components running on different devices:

1. Ollama Server (On your powerful computer)
2. RAG Server (On Jetson device)
3. Main Coordinator (On Jetson device)

Quick Setup

1. Update IP addresses in config.json file
2. On your computer: Run start_ollama_server.bat (Windows) or start_ollama_server.sh (Linux/Mac)
3. On Jetson: Start the RAG server
4. On Jetson: Run the main coordinator

For detailed setup instructions: NETWORK_SETUP.md

---

Setup

git clone --recursive git@github.com:davesarmoury/PotatOS.git
cd PotatOS

pip3 install -r requirements.txt
sudo groupadd -f -r gpio
sudo usermod -a -G $USER

sudo cp lib/python/Jetson/GPIO/99-gpio.rules /etc/udev/rules.d/
#sudo reboot

Jetson Containers

bash jetson-containers/install.sh

Patch platform detect (optional)

cd patch/
./patch_platformdetect.sh
cd ..

LLM (Network Setup)

With the new network configuration, Ollama runs on your powerful computer instead of Jetson:

On your computer:

• Run start_ollama_server.bat (Windows) or start_ollama_server.sh (Linux/Mac)
• This will start Ollama with network access

Previous Jetson setup (deprecated): jetson-containers run --name ollama $(autotag ollama) ollama run llama3.2:3b --keepalive 60m

RAG Server

cd wiki_rag/
pip3 install -r requirements.txt

# Choose one of the following RAG implementations:

# Option 1: Simple Ollama-based RAG (Recommended)
python3 wiki_ollama.py

# Option 2: LlamaIndex-based RAG (More advanced)
#python3 wiki_llamaindex_preprocess.py
#python3 wiki_llamaindex.py

Both implementations now connect to the network Ollama server configured in config.json

Audio

See https://developer.nvidia.com/embedded/learn/tutorials/connecting-bluetooth-audio, then pair a bluetooth speaker with bluetoothctl. Then set the volume

sudo nano /lib/systemd/system/bluetooth.service.d/nv-bluetooth-service.conf


amixer -D pulse sset Master 50%

VOSK

pip3 install -r vosk-server/requirements.txt 
pip3 install sounddevice
sudo apt install libportaudio2

cd vosk-server/websocket
wget https://alphacephei.com/vosk/models/vosk-model-small-en-us-0.15.zip
unzip vosk-model-small-en-us-0.15
rm vosk-model-small-en-us-0.15.zip
mv vosk-model-small-en-us-0.15 model

python3 asr_server.py

Run Piper Server

Train or download the GLaDOS Piper voice. See train_piper for more details. If you don't want to train you own, you can download pre-trained models using

cd jetson-containers/data/models/piper/
wget https://huggingface.co/DavesArmoury/GLaDOS_TTS/resolve/main/glados_piper_medium.onnx.json
wget https://huggingface.co/DavesArmoury/GLaDOS_TTS/resolve/main/glados_piper_medium.onnx

On the jetson, put the onnx and onnx.json files into jetson-containers/data/models/piper, then run the command below

docker pull dustynv/piper-tts:r36.2.0
jetson-containers run $(autotag piper-tts) python3 -m piper.http_server --port 5001 -m /data/models/piper/glados_piper_medium.onnx

Run a quick test with

curl -G --data-urlencode 'text=I like big butts, I cannot lie.' --output - 'localhost:5001' | aplay

Coordinator

python3 potatos.py -r 44100 -d 0

PulseAudio Quirk

PulseAudio will automatically disconnect when no one is logged into the machine. That means if you are trying to run it headless, your audio will break. This is gross, but just ssh the terminal into itself so the machine sees a "remote" connection

ssh localhost