README.md

metadata

title: Vish AI
sdk: gradio
emoji: 🔥
colorFrom: green
colorTo: blue
pinned: false

🌟 Vish AI - Virtual Intelligent System Hub

Powerful AI assistant powered by Microsoft Phi-3 Mini (3.8B parameters) - Optimized for Hugging Face Spaces

🚀 Features

Core Capabilities

• 💬 Chat Assistant: Intelligent conversation with 4K context window (Phi-3)
• 📝 Text Summarization: Advanced text condensing with AI understanding (Phi-3)
• 😊 Sentiment Analysis: Accurate emotion detection (Phi-3)
• 🔐 Supabase Authentication: Optional secure user management
• 📊 Usage Logging: Track interactions in Supabase database

Performance Specs

• Model: Microsoft Phi-3 Mini 4K Instruct (3.8B parameters)
• Model Size: ~7.4GB (unified model for all tasks)
• Response Time: 2-5 seconds on CPU (faster on GPU)
• Memory Usage: ~8GB RAM recommended
• CPU Optimized: Works on free tier, better on GPU

🎯 Use Cases

1. Advanced Chatbot: High-quality conversational AI
2. Content Analysis: Professional-grade summarization and sentiment detection
3. Educational Tool: Intelligent learning assistant
4. Research Assistant: Context-aware information processing
5. VIJ Project Integration: Powerful AI backend

📦 Quick Deploy to Hugging Face Spaces

Method 1: Direct Upload

1. Create a new Space on Hugging Face

   • Go to: https://huggingface.co/new-space
   • Select: Gradio SDK
   • Python: 3.10 or 3.11 (recommended)
   • Hardware: CPU basic (free, slower) or T4 GPU (faster)

2. Upload these files:

   • app.py (main application)
   • requirements.txt (dependencies)
   • README.md (this file)

3. Wait for build (15-20 minutes on first run):

   • Installing dependencies (~3 minutes)

   • Downloading Phi-3 model (~10-15 minutes, 7GB)

   • Building app (~2 minutes)

   • Add these secrets:

     NEXT_PUBLIC_SUPABASE_URL=https://lyebtceryednzafhyunq.supabase.co
     NEXT_PUBLIC_SUPABASE_ANON_KEY=your_anon_key_here
     

4. Deploy: The space will automatically build and deploy!

Local Development

# Clone the repository
git clone https://github.com/vishwas896/Vish_AI.git
cd Vish_AI

# Install dependencies
pip install -r requirements.txt

# Set up environment variables
cp .env.example .env
# Edit .env with your Supabase credentials

# Run the application
python app.py

🗄️ Supabase Setup

Create Logs Table

Run this SQL in your Supabase SQL Editor:

-- Create table for logging Vish AI interactions
CREATE TABLE IF NOT EXISTS vish_ai_logs (
    id BIGSERIAL PRIMARY KEY,
    user_email TEXT,
    prompt TEXT,
    response TEXT,
    model_type TEXT,
    timestamp TIMESTAMPTZ DEFAULT NOW()
);

-- Create index for faster queries
CREATE INDEX idx_vish_ai_logs_user ON vish_ai_logs(user_email);
CREATE INDEX idx_vish_ai_logs_timestamp ON vish_ai_logs(timestamp DESC);

-- Enable Row Level Security (RLS)
ALTER TABLE vish_ai_logs ENABLE ROW LEVEL SECURITY;

-- Policy: Users can view their own logs
CREATE POLICY "Users can view own logs"
    ON vish_ai_logs FOR SELECT
    USING (auth.jwt() ->> 'email' = user_email);

-- Policy: Service role can insert logs
CREATE POLICY "Service role can insert logs"
    ON vish_ai_logs FOR INSERT
    WITH CHECK (true);

🔧 Configuration

Model Selection

The AI uses these lightweight models:

Model	Size	Speed	Purpose
DistilGPT2	82MB	~0.5-2s	Chat conversations
DistilBART-CNN-6-6	300MB	~1-3s	Text summarization
DistilBERT-SST2	255MB	~0.3-1s	Sentiment analysis

Why These Models?

✅ Optimized for CPU - No GPU required
✅ Fast inference - Sub-3 second responses
✅ Low memory - Runs on 2GB RAM
✅ Good accuracy - Distilled from larger models
✅ Free tier friendly - Fits Hugging Face limits

🌐 VIJ Project Integration

Connect from v0.dev/Next.js

// In your VIJ project (Next.js/React)
const callVishAI = async (message: string, userToken: string) => {
  const response = await fetch('https://vishwas896-vish-ai.hf.space/api/predict', {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
    },
    body: JSON.stringify({
      data: [message, [], userToken]
    })
  });
  
  const result = await response.json();
  return result.data[0];
};

// Usage with Supabase auth
const { data: { session } } = await supabase.auth.getSession();
const aiResponse = await callVishAI(
  "Hello Vish AI!",
  session?.access_token || ""
);

API Endpoints

Once deployed, your space will have these endpoints:

• Chat: POST /api/predict (function_index: 0)
• Summarize: POST /api/predict (function_index: 1)
• Sentiment: POST /api/predict (function_index: 2)

📊 Performance Benchmarks

Tested on Hugging Face CPU basic (free tier):

Task	Avg Response Time	Memory Usage
Chat (50 words)	1.2s	~800MB
Summarization (500 words)	2.4s	~1.2GB
Sentiment Analysis	0.6s	~600MB

🔒 Security

• Environment Variables: Sensitive keys stored in HF Secrets
• Supabase RLS: Row-level security on logs table
• JWT Validation: Optional user authentication
• Anonymous Mode: Works without authentication

🚦 Usage Limits (Free Tier)

• CPU Time: Reasonable for personal projects
• Memory: 2GB RAM limit (well within our ~1.5GB usage)
• Storage: 50GB (models cache ~2GB)
• Sleeps after 48h inactivity: First request wakes it up

🛠️ Troubleshooting

Models Loading Slowly

• Normal on first run (downloads ~650MB)
• Cached after first load
• Takes 30-60 seconds initially

Out of Memory Error

• Reduce max_length in text generation
• Use smaller batch sizes
• Consider upgrading to CPU upgrade tier ($0)

Supabase Connection Issues

• Verify environment variables are set
• Check Supabase project is active
• Ensure RLS policies are correct

📈 Roadmap

• Add image analysis (CLIP model)
• Voice input/output
• Multi-language support
• Custom model fine-tuning
• Advanced analytics dashboard
• WebSocket for real-time chat

🤝 Contributing

Contributions welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Submit a pull request

📝 License

MIT License - feel free to use in your projects!

🙏 Acknowledgments

• Hugging Face: For free hosting and amazing models
• Supabase: For backend infrastructure
• v0.dev: For VIJ project development
• Gradio: For beautiful UI framework

📧 Contact

Vishwas

• Hugging Face: @Vishwas896
• GitHub: @vishwas896

---

Built with ❤️ for the VIJ Project