Audio-to-Audio Architecture

This folder contains a complete bidirectional audio-to-audio communication system using Google's Gemini Live API. The system enables real-time voice conversations with AI through (smart glasses or) any web client.

Architecture Overview

• Backend: Python WebSocket server (server.py) that connects to Gemini Live API
• Frontend: JavaScript audio client (audio-client.js) with HTML interface
• Communication: Real-time bidirectional audio streaming via WebSocket

Files

Core Files (focus on server.py)

• server.py - Main WebSocket server for Scholar AI assistant

• audio-client.js - Frontend audio client class for recording/playback

• index_for_server.html - Web interface for testing

• common.py - Shared utilities and base classes

• requirements.txt - Python dependencies (minimal, server.py focused)

• archive/ - Contains alternative implementations and unused files

Features

• Real-time audio recording and streaming
• Bidirectional communication (speak and listen)
• WebSocket-based architecture
• Browser-based client interface

Setup

Prerequisites

• Python 3.8+
• Google API Key for Gemini Live API

Installation

1. Clone and navigate to project:

   git clone <repository-url>
   cd audio-to-audio-architecture

2. Create virtual environment (recommended):

   python3 -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

   Note: Virtual environments (venv/, .venv/) are excluded from git via .gitignore as they're platform-specific and recreatable.

3. Install dependencies:

   pip install -r requirements.txt

4. Set API key:

   export GOOGLE_API_KEY="your_api_key_here"

5. Run server:

   python server.py

6. Open frontend: Open index_for_server.html in your browser and start recording!

   google-chrome index_for_server.html 

Usage

1. Set your GOOGLE_API_KEY environment variable
2. Install dependencies: pip install -r requirements.txt
3. Run server: python server.py
4. Open index_for_server.html in browser
5. Start recording and speak with Scholar