pi-voice-assistant

Voice Assistant For Raspberry Pi 5

Stack:

• ollama local models for LLM inference
• fast whisper for speech to text
• open wake word for waking up the program

Setup

Ollama

Keep the model loaded in memory indefinitely to avoid cold start latency after idle periods:

sudo systemctl edit ollama.service

Add the following under [Service]:

[Service]
Environment="OLLAMA_KEEP_ALIVE=-1"

Then reload and restart:

sudo systemctl daemon-reload && sudo systemctl restart ollama

Running the project

1. ssh into your raspberry pi
2. clone the repo
3. run the run.sh shell script

Running unit tests

uv run python -m pytest