AskAce: D'RAG 🎯

Ultra-fast RAG chatbot for document Q&A

✨ Features

🔒 Complete Privacy - All processing happens locally
⚡ Lightning Fast - 2-second startup, 1-3 second responses
📚 Multi-Format - PDF, DOCX, TXT, MD support
🎯 Smart Retrieval - FAISS vector search with citations
🚀 Easy Deployment - One-click cloud deployment
💰 Zero API Costs - Uses local Ollama LLMs
🌐 Web Version - Static HTML/CSS/JS for GitHub Pages

🚀 Quick Start

1. Install Dependencies

pip install -r requirements.txt

2. Setup Ollama

# Install from https://ollama.com
ollama pull llama3.2:1b  # Fast model
# OR
ollama pull llama3.2:3b  # Better quality

4. Add Your Documents

• Drop files into data/ folder
• Click "🚀 Build Index"
• Start asking questions!

⚡ Performance Optimizations

• Smart Caching - Index and models cached in memory
• Lazy Loading - Components load only when needed
• Batch Processing - Embeddings generated in efficient batches
• Optimized Chunking - 500-char chunks with minimal overlap
• Fast Models - Prioritized smaller, faster LLMs

🏗️ Architecture

Documents → Chunking → Embeddings → FAISS Index
     ↓
User Query → Embedding → Similarity Search → Context → LLM → Answer

📊 Benchmarks

Metric	Performance
Cold Start	~2 seconds
Query Response	1-3 seconds
Index Building	~100 chunks/second
Memory Usage	200-500 MB
Supported Docs	Unlimited

🛠️ Tech Stack

• 🎨 Frontend: Streamlit + HTML/CSS/JS
• 🔤 Embeddings: SentenceTransformers (local)
• 🗂️ Vector DB: FAISS
• 🤖 LLM: Ollama (local) / OpenAI (cloud)
• 📄 Documents: PyPDF, python-docx

🔧 Configuration

Models

• Fast: llama3.2:1b (~1.5GB)
• Balanced: phi3:mini (~2.2GB)
• Quality: llama3.2:3b (~2.0GB)

Embeddings

• Default: all-MiniLM-L6-v2 (22MB, 384 dims)
• Alternative: paraphrase-MiniLM-L6-v2 (22MB, 384 dims)

📁 Project Structure

askace-drag/
├── index.html             # 🌐 Static web version
├── app.py                 # 🐍 Main Streamlit app
├── app_cloud.py           # ☁️ Cloud version
├── start.py               # 🚀 Optimized launcher
├── rag/
│   ├── llm_client.py     # 🤖 Optimized LLM & embeddings
│   ├── rag_core.py       # 🧠 RAG pipeline with caching  
│   └── ingest.py         # 📄 Fast document processing
├── data/                  # 📂 Your documents
├── storage/              # 💾 Generated indices
├── requirements.txt      # 📦 Dependencies
├── WEB_VERSION.md        # 🌐 Web deployment guide
└── deployment guides/    # 📚 Hosting instructions

🚀 Getting Started

Local Development

git clone https://github.com/areekaraza/askace-d-rag.git
cd askace-d-rag
pip install -r requirements.txt
ollama pull llama3.2:1b
python start.py

🤝 Contributing

1. Fork the repository
2. Create feature branch (git checkout -b feature/amazing-feature)
3. Commit changes (git commit -m 'Add amazing feature')
4. Push to branch (git push origin feature/amazing-feature)
5. Open Pull Request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• Ollama - Local LLM serving
• FAISS - Efficient similarity search
• SentenceTransformers - Semantic embeddings
• Streamlit - Rapid web app development

💡 Support

• Documentation: Check the deployment guides in this repo
• Issues: GitHub Issues
• Discussions: GitHub Discussions

---

🌟 Star this repo if it helped you build an awesome RAG chatbot!