<<<<<<< HEAD

AI-Showcase

=======

🤖 Local AI Demo Stack

Complete AI application with Voice, Text, and Vision - 100% Local & Free

✨ Features

• 🎤 Voice Chat: Speak → AI transcribes → Generates response → Speaks back
• 💬 Text Chat: Type messages with optional voice output
• 👁️ Vision Analysis: Upload images for AI description and analysis
• 🔒 100% Local: No external APIs, complete privacy
• 💰 Free: No usage costs, open-source models only

🚀 Quick Start

1. Install Dependencies

# Clone or download this project
cd local-ai-demo

# Install Python packages
pip install -r requirements.txt

# Or run setup script
python setup.py

2. Run Application

python app.py

3. Open Interface

Open your browser to: http://localhost:7860

🛠️ System Requirements

Minimum

• RAM: 8GB
• VRAM: 4GB (GPU recommended)
• Storage: 10GB for models
• Python: 3.8+

Recommended

• RAM: 16GB+
• VRAM: 8GB+ (NVIDIA GPU)
• Storage: 20GB
• OS: Linux/Windows/macOS

🎯 How It Works

Voice Chat Flow

🎙️ Record Audio → 🗣️ Whisper STT → 🧠 Local LLM → 🔊 TTS → 🎵 Audio Response

Text Chat Flow

⌨️ Type Message → 🧠 Local LLM → 📝 Text Response (+ Optional 🔊 TTS)

Vision Analysis Flow

📷 Upload Image → 👁️ BLIP-2 VLM → 📋 Description & Analysis

🤖 AI Models Used

Component	Model	Size	Purpose
Speech-to-Text	OpenAI Whisper	~1GB	Voice transcription
Language Model	DialoGPT-medium	~1.5GB	Text generation
Text-to-Speech	pyttsx3	<1MB	Voice synthesis
Vision	BLIP-2-2.7B	~3GB	Image analysis

Alternative: Ollama Integration

For better LLM performance, install Ollama:

# Install Ollama
curl -fsSL https://ollama.ai/install.sh | sh

# Pull a model
ollama pull llama2:7b-chat

# App will automatically use Ollama if available

📱 Interface Overview

🎤 Voice Chat Tab

• Record audio with microphone
• Real-time speech-to-text transcription
• AI generates contextual responses
• Text-to-speech audio output
• Conversation history display

💬 Text Chat Tab

• Standard chat interface
• Optional voice output toggle
• Conversation memory
• Clear chat functionality

👁️ Vision Analysis Tab

• Image upload or camera capture
• Custom question input
• AI-powered image description
• Detailed visual analysis

⚙️ Settings Tab

• System information display
• Model configuration details
• Performance tips
• Privacy information

🔒 Privacy & Security

• ✅ 100% Local Processing: All AI runs on your machine
• ✅ No External APIs: No data sent to cloud services
• ✅ Offline Capable: Works without internet connection
• ✅ No Data Collection: Conversations stay private
• ✅ Open Source: Fully transparent implementation

🎛️ Configuration

Edit config.py to customize:

# Model Selection
LLM_MODEL = "microsoft/DialoGPT-medium"
WHISPER_MODEL = "base"  # tiny, base, small, medium
VLM_MODEL = "Salesforce/blip2-opt-2.7b"

# Performance Settings
DEVICE = "cuda"  # or "cpu"
MAX_TOKENS = 512
TEMPERATURE = 0.7

# UI Settings
GRADIO_PORT = 7860

🚨 Troubleshooting

Common Issues

Models downloading slowly?

• Models download automatically on first use
• Subsequent runs are much faster
• Check internet connection for initial setup

Out of memory errors?

• Reduce model sizes in config.py
• Use CPU instead of GPU: DEVICE = "cpu"
• Close other applications

Audio not working?

• Check microphone permissions
• Install audio drivers
• Try different audio input device

Ollama connection failed?

• Install Ollama: curl -fsSL https://ollama.ai/install.sh | sh
• Start Ollama service: ollama serve
• Pull model: ollama pull llama2:7b-chat

Performance Tips

1. Use GPU: Significantly faster inference
2. Smaller Models: Trade quality for speed
3. Close Apps: Free up RAM and VRAM
4. SSD Storage: Faster model loading

🛣️ Roadmap

• Real-time streaming responses
• Multi-language support
• Custom model fine-tuning
• Mobile app version
• Advanced conversation memory
• Plugin system for extensions

🤝 Contributing

1. Fork the repository
2. Create feature branch
3. Make improvements
4. Test thoroughly
5. Submit pull request

📄 License

MIT License - Feel free to use and modify!

🙏 Acknowledgments

• OpenAI Whisper: Speech recognition
• Hugging Face: Model hosting and transformers
• Gradio: Web interface framework
• Ollama: Local LLM serving
• Open Source Community: Making AI accessible

---

🚀 Built with ❤️ for local AI enthusiasts

No cloud required • No API costs • Complete control