GenRAG

GenRAG is an AI-powered document intelligence system that implements Retrieval-Augmented Generation (RAG) from scratch. Built without high-level frameworks like LangChain or vector databases, it features a modern GUI, semantic search through 1,400+ text chunks, and AI-powered responses. The system processes PDFs, generates embeddings, and provides instant answers to questions about document content with sub-millisecond search speeds.

Features

• 🎨 Modern GUI: GitHub-inspired dark theme interface
• 📄 PDF Processing: Automatic text extraction and chunking
• 🧠 Smart Embeddings: Sentence-BERT powered semantic search
• ⚡ Fast Search: Sub-millisecond query processing
• 🤖 AI Responses: Google Gemini API integration
• 💾 Lightweight: CSV-based storage (no vector database needed)
• 🔧 Python 3.8+ Compatible: Works on older Python versions

Installation

1. Clone the repository:

   git clone https://github.com/AkprasadoP/GenRAG.git
   cd GenRAG

2. Create and activate a virtual environment:

   python -m venv env
   call env\Scripts\activate  # On Windows
   # source env/bin/activate  # On Linux/Mac

3. Install compatible dependencies:

   pip install -r requirements_minimal.txt

Setup

1. Generate Embeddings (The PDF is already included):

   python create_embeddings_auto.py

   This will process "The Intelligent Investor" PDF and create embeddings.

2. Configure Gemini API (Optional for enhanced responses):

   • Edit .env file and add your API key:

     GEMINI_API_KEY=your_actual_api_key_here
     

Usage

GUI Application (Recommended)

python app.py

Or simply double-click run.bat

Terminal Version

python main.py

Test Multiple Queries

python test_rag.py

LLM Response

You can use both a local LLM or an LLM from an API like Gemini for generating responses.

• Local LLM: If you have the capability to run a local LLM, you can use it for generating responses. Cause mine is too slow :(

• LLM from API: If your system is not powerful enough for local inference, you can use an API like Gemini. To do this, create a .env file and pass the Gemini API key.

Using Gemini API

1. Get your API key from Google AI Studio
2. Edit the existing .env file:

   GEMINI_API_KEY=your_actual_api_key_here

3. Restart the application to use enhanced AI responses

Note: The system works perfectly with fallback responses even without the API key.

---

Author

Ashish Prasad - @AkprasadoP

This project is developed and maintained by Ashish Prasad.

Credits

Special thanks to the following YouTube channels and research papers for their invaluable resources and insights:

YouTube Channels

• AI Anytime
• Daniel Brouke
• Krish Naik
• Codebasics
• CampusX

Research Papers

• Patrick Lewis ., "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks" arXiv:2005.11401
• Vaswani et al., "Attention is All You Need" arXiv:1706.03762
• Reimers et al., "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks" arXiv:1908.10084

License

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