Local LLM Chat Bot

A local voice-enabled chat bot running on DietPi (Raspberry Pi).

Prerequisites

• OS: DietPi is recommended. See device-stuff for configuration files (diet-pi_config.txt, rpi_config.txt).

Installation

1. Run Setup Script This script installs system dependencies (apt), Ollama, Piper TTS, and downloads necessary models.

   ./scripts/setup.sh
   

2. Create Python Environment It is recommended to use --system-site-packages to allow access to system-level libraries installed via apt (like python3-lgpio).

   python3 -m venv --system-site-packages .venv
   source .venv/bin/activate
   pip install -r requirements.txt  # If you have one, or install dependencies manually
   # Dependencies: ollama, fast-whisper, sounddevice, soundfile, numpy, gpiozero, piper-tts, fastapi, uvicorn, python-multipart
   # Note: local_chat.py dependencies are currently managed manually or via setup.sh hints.
   

   Note

   : If you run into issues with lgpio or audio, ensuring --system-site-packages is used is often the fix.

Running the Bot

To run the chat bot manually:

source .venv/bin/activate
python3 src/local_chat/local_chat.py

Systemd Services (Autostart)

To configure the bot to start automatically on boot:

1. Install Service Files Run the startup setup script to copy systemd files to /etc/systemd/system/.

   sudo ./scripts/setup_startup.sh
   

2. Enable and Start Services

   # Reload systemd
   sudo systemctl daemon-reload
   
   # Enable Ollama (Dependecy)
   sudo systemctl enable --now ollama
   
   # Enable Chat Bot
   sudo systemctl enable --now local-chat-bot
   

3. Optional: Faster Whisper Server If you are using the separate Whisper server:

   sudo systemctl enable --now faster-whisper-server
   

Project Structure

• src/local_chat/: Python source code (local_chat.py, assistant.py, etc.)
• scripts/: Shell scripts for setup and maintenance.
• device-stuff/: Hardware/OS configuration files.
• assets/: Models and sounds.