AI-Computer (Local Offline Voice Assistant)

This project implements a privacy-focused, local voice assistant for Windows. It listens for the wake-word "Computer", responds using a synthetic Text-to-Speech (TTS) voice, fetches weather data via an API, and controls Nanoleaf lights within your local network. Complex open-ended questions are processed entirely locally using Ollama. It's main use is in German, but you can change it to your language in the code.

Features

• 100% Local Speech-to-Text (STT): Powered by Vosk for offline speech recognition.
• Local AI Model: Handles general knowledge questions locally via Ollama (gemma4:e2b).
• Weather Integration: Accurate current forecasts and rain reports via OpenWeatherMap.
• Smart Home Control: Automatic discovery and toggling of Nanoleaf devices (e.g., Shapes).
• Text-to-Speech (TTS): Responds directly in German using the system's native voice.

---

Prerequisites & Installation

1. Windows Dependencies (PyAudio)

Since installing PyAudio directly on Windows can sometimes fail, it is highly recommended to install it using pipwin:

pip install pipwin
pipwin install pyaudio

2. Install Python Libraries

Install all other required packages with the following command:

pip install vosk pyttsx3 requests langchain-core langchain-ollama nanoleafapi zeroconf

3. Download the Vosk Speech Model

For speech recognition to function, you simply need to download the model and place it into your project. No training or extra installation is required:

1. Download the vosk-model-small-de-0.15 model from Vosk Models (https://alphacephei.com/vosk/models).
2. Extract the downloaded folder into your main project directory.
3. Rename the extracted folder exactly to model_small.

4. Set Up Ollama

1. Install Ollama for Windows (https://ollama.com/).
2. Open your terminal and download the model specified in the code:

ollama pull gemma4:e2b

(If you prefer to use a different model like llama3, simply change the model name inside the Python script where OllamaLLM(model="...") is defined).

---

Configuration

Open the main file of your script and adjust the configuration block at the top to match your environment:

# ==========================================
# CONFIGURATION & SECRETS (CUSTOMIZE HERE)
# ==========================================
OPENWEATHER_API_KEY = "YOUR_API_KEY_HERE"  # Create for free on openweathermap.org
LATITUDE = "50.0"                          # Your latitude coordinate
LONGITUDE = "0.0"                          # Your longitude coordinate

# Absolute path to ollama.exe so the script can launch Ollama if it isn't running
OLLAMA_EXECUTABLE_PATH = r"C:\Path\To\Your\Ollama\ollama.exe"

# Nanoleaf name as it appears in your network
NANOLEAF_NAME = "Shapes 9121"

# Index of your microphone input device
INPUT_DEVICE_INDEX = 3
# ==========================================

How to find your INPUT_DEVICE_INDEX

To find out which ID (index) belongs to your microphone, you can run this short helper script in a separate Python file:

import pyaudio

p = pyaudio.PyAudio()
print("Available audio input devices:")
for i in range(p.get_device_count()):
    dev = p.get_device_info_by_index(i)
    if dev.get('maxInputChannels') > 0:
        print(f"Index {i}: {dev.get('name')}")
p.terminate()

Run the script, check the terminal output for your desired microphone, and enter its corresponding index number into the INPUT_DEVICE_INDEX variable above.

Important for Nanoleaf: During the very first launch, the program will look for your Nanoleaf panel. Press and hold the power button on your Nanoleaf controller for about 5-7 seconds (until the LEDs start flashing) to enable pairing mode. The script will then automatically generate a permanent authentication token.

---

Usage

Start the script via your terminal:

python main.py

The script will check in the background if Ollama is running (and launch it if necessary), connect to your microphone using the configured index, and start listening for your commands.

Voice Commands (Always start with "Computer"):

• "Computer, wie wird das Wetter heute?" (Computer, how is the weather today?)
• "Computer, wird es morgen regnen?" (Computer, will it rain tomorrow?)
• "Computer, Licht an!" / "Computer, Licht aus!" (Computer, turn the light on! / turn the light off!)
• "Computer, warum ist der Himmel blau?" (Computer, why is the sky blue? -> forwards the question to your local AI model)