🐾 EchoPaw: AI-Powered Therapeutic Companion for Dementia Care

EchoPaw is an AI companion system designed to help elderly dementia patients through natural conversation. Built for privacy and empathy, it runs entirely on your local computer without sending any data to the cloud.

🎯 What Does EchoPaw Do?

• Natural Conversations: Talk to EchoPaw like you would with a friend
• Remembers Everything: Stores personal memories and stories locally on your device
• Completely Private: No data ever leaves your computer - perfect for healthcare settings
• Voice-Enabled: Speaks back to you with realistic AI-generated voice
• Therapeutic Focus: Designed specifically for dementia care and cognitive support

🏥 The Problem We're Solving

In NHS hospitals, dementia patients occupy 25% of acute beds and stay 30-60 days (vs 16 days in Japan). EchoPaw aims to prevent cognitive decline during hospital stays, potentially reducing costs and improving patient outcomes.

🛠️ How It Works

EchoPaw combines four AI technologies:

1. Speech-to-Text (Whisper) - Understands what you say
2. Memory System (RAG) - Remembers your conversations and personal details
3. Language Model (Llama 3.2) - Generates caring, therapeutic responses
4. Text-to-Speech (Voice Cloning) - Speaks back with a warm, human-like voice

Everything runs locally on your computer for complete privacy.

🚀 Quick Start

Prerequisites

• Python 3.13+
• 32GB RAM recommended (minimum 16GB)
• GPU optional but recommended (NVIDIA, Apple Silicon, or CPU fallback)
• Microphone for voice input

Installation

1. Clone the repository

   git clone git@github.com:Shreyas8612/EchoPaw.git

2. Install dependencies

   # Using uv (recommended)
   uv sync
   
   # Or using pip
   pip install -r requirements.txt

3. Test the system

   python EchoPaw.py

Running the Web Interface

For the best experience, use the web interface:

python Web.py

Then open your browser and go to: http://localhost:5000

The web interface provides:

• 🎤 Easy voice recording
• 💬 Clean chat interface
• 📊 Memory statistics
• 🔊 Audio playback

💻 Usage Options

Option 1: Web Interface (Recommended)

python Web.py

• Open http://localhost:5000 in your browser
• Click "Speak Now" to record audio
• EchoPaw will respond with both text and voice

Option 2: Command Line

python EchoPaw.py

• Press Enter to record audio
• Type 'good-bye' to exit
• Say 'demo' to see memory system in action

Option 3: Memory Demo

python RAG_demo.py

• See how EchoPaw stores and recalls memories
• Perfect for understanding the system capabilities

🧠 Memory System

EchoPaw automatically remembers:

• Personal details (family, pets, hobbies)
• Important events and stories
• Preferences and interests
• Medical information (if shared)

All memories are stored locally in the memory/ folder and never leave your computer.

🎛️ Configuration

Performance Optimization

For NVIDIA GPUs:

• CUDA acceleration enabled automatically
• Faster processing for real-time conversations

For Apple Silicon (M1/M2/M3):

• MPS acceleration for optimal performance
• Native ARM64 support

For CPU-only systems:

• Optimized for Intel/AMD processors
• Slightly slower but fully functional

Voice Customization

The system uses Naomi Scott's voice by default. To use a different voice:

1. Replace Naomi Scott.mp3 with your audio file
2. Update the text in TTS.py to match your audio content
3. Restart the system

📁 Project Structure

echopaw/
├── EchoPaw.py          # Main command-line interface
├── Web.py              # Web server and interface
├── Web.html            # Web UI (served automatically)
├── STT.py              # Speech-to-text (Whisper)
├── LLM.py              # Language model (Llama 3.2)
├── RAG.py              # Memory system
├── TTS.py              # Text-to-speech with voice cloning
├── RAG_demo.py         # Memory system demonstration
├── memory/             # Local memory storage (created automatically)
├── requirements.txt    # Python dependencies
└── README.md           # This file