Offline RAG (Retrieval Augmented Generation) Project

Project Overview

First step towards GenAI and LLMs. My first experiment with RAG based chatbot, with an attempt to include features to make it faster, accurate and more robust. Main feature : this RAG based document retrieval and question-answering system can operate fully offline, without needing any api key(but after downloading/pulling the LLM and embedding models)

Older obselete versions are also updated to keep track of the progress with time.

Still few bugs exists, which needs to be fixed.

1. instead of web scrapping, it needs to be changed to web search with Tavily integration
2. Web layout needs to be updated.
3. More models to be downloaded and tested.

Versions

• Version 1: Basic RAG implementation with FastAPI (Desktop version also available)
• Version 2: Enhanced chatbot with FAISS (Desktop version also available)
• Version 3: Advanced features and optimizations(v0, v1 , v2) [Streamlit versions]

Version 3

Advanced Features

Document Processing

• Multi-format document support (PDF, Text, Code)
• OCR capabilities with PyMuPDF and Tesseract
• Intelligent PDF parsing with fallback mechanisms
• Code syntax highlighting and language detection

Vector Search & Embeddings

• FAISS vector store integration
• Multiple embedding options:
  • Ollama embeddings
  • HuggingFace embeddings
  • Configurable model selection

Performance

• Asynchronous document processing
• Advanced caching system with:
  • LRU cache
  • Disk cache
  • Document fingerprinting
• Multi-threaded operations

Intelligence

• Multi-language support
• Context-aware responses
• Source attribution
• Response optimization
• Conversation state management

Monitoring & Logging

• Progress tracking
• Detailed logging
• Cache statistics
• Performance metrics

🌟 Key Features of the Project

• Multi-Version Support: Three distinct versions with incremental improvements
• Document Processing: Handle PDFs, TXTs, and other text-based formats
• Vector Storage: FAISS-based efficient similarity search
• Multiple Interfaces: FastAPI and Streamlit implementations
• Async Processing: Enhanced performance with asynchronous operations
• Caching System: Optimized response times
• Testing Suite: Comprehensive unit and integration tests

📁 Project Structure

RAG_new/
├── src/
│   ├── v1/         # Base implementation
│   ├── v2/         # Enhanced features
│   └── v3/         # Latest upgrades
├── data/           # Document and vector stores
├── tests/          # Testing suite
├── config/         # Configuration files
├── utils/          # Helper utilities
└── web/           # Web interface assets

🚀 Quick Start

# Clone repository
git clone https://github.com/yourusername/RAG_new.git

# Install dependencies
pip install -r requirements.txt

# Run tests
python run_tests.py

# Start web interface
python src/v1/RAG_Search_web.py

🔧 Configuration

1. Set up Ollama LLM
2. Configure vector store path in config/settings.py
3. Add documents to data/documents/

💻 Usage Examples

from src.v1.RAG_Search_new import RAG_search

# Simple query
response = RAG_search("How does RAG work?")

# Document ingestion
from src.v1.RAG_Search_new import create_vector_store_from_pdfs
create_vector_store_from_pdfs("path/to/docs")

🧪 Testing

# Run all tests
pytest tests -v

# Run specific test category
pytest tests/unit -v
pytest tests/integration -v

📚 Version Details

• V1: Basic RAG implementation with FastAPI
• V2: Enhanced chatbot with FAISS
• V3: Advanced features and optimizations

📋 Requirements

• check requirements file

📜 License

MIT License

👥 Contributing

See CONTRIBUTING.md for guidelines.