🎙️ Mysaa Web Wake Word Engine

A real-time, browser-based wake word detection application built with FastAPI and openWakeWord.

---

🌟 Overview

This is the web-based inference engine for the Mysaa Wake Word Assistant. It streams live audio from the browser's microphone over WebSockets to a FastAPI server, which processes the audio using the highly optimized openWakeWord library running ONNX models.

When the target wake word (e.g., "Hey Jarvis") is detected, the server instantly sends a WebSocket event back to the frontend, which displays a beautiful glassmorphic UI alert.

✨ Key Features

• Real-Time Streaming: Seamlessly streams 16kHz audio from the browser to the backend using WebSockets.
• Low Latency: Uses asyncio.to_thread for non-blocking ONNX model inference.
• Modern Interface: A sleek, responsive, glassmorphic UI equipped with an animated listening orb.
• Plug-and-Play Models: Easily swap between different openWakeWord models.

📂 Structure

webapp/
├── server.py             # FastAPI WebSocket server and ONNX inference logic
├── hey_jarvis_v0.1.onnx  # Pre-trained wake word ONNX model
├── requirements.txt      # Python dependencies
└── static/               
    ├── index.html        # Main glassmorphic frontend
    ├── styles.css        # Animations and CSS styling
    └── app.js            # Audio capture and WebSocket logic

🚀 Getting Started

1. Requirements

Ensure you have Python 3.8+ installed. Install the dependencies:

pip install -r requirements.txt

(Dependencies generally include fastapi, uvicorn, openwakeword, and numpy)

2. Running the Server

Start the FastAPI application via Uvicorn:

python server.py

3. Usage

1. Open your browser and navigate to http://localhost:8000/static/index.html
2. Click Enable Microphone to allow the browser to capture audio.
3. Speak the wake word (e.g., "Hey Jarvis").
4. The frontend orb will react, and if the word is detected above the confidence threshold, a glassmorphic popup modal will appear!

🔧 Customization

• Change the Wake Word: To use a custom word like "Mysaa", generate a new ONNX model using openWakeWord's text-to-wakeword generator. Swap the .onnx file and update server.py and index.html.
• Threshold Adjustment: If you're experiencing false positives, you can adjust the detection threshold in server.py (currently set to > 0.5).