PDF RAG Assistant 📚

A powerful Streamlit application that enables you to upload PDF documents and ask questions about their content using AI-powered Retrieval-Augmented Generation (RAG) technology.

🚀 Live Demo

Try it now: https://pdfanalyzerwithrag-ezbzxnaza2ng9lpd9fgf5x.streamlit.app/

Note: You'll need your own Gemini API key to use the live demo.

Features ✨

• PDF Upload & Processing: Upload PDF documents and extract text content
• Intelligent Text Chunking: Automatically splits documents into semantic chunks with overlap
• Vector Embeddings: Creates semantic embeddings using sentence transformers
• Semantic Search: Uses FAISS for efficient similarity search
• AI-Powered Q&A: Leverages Google Gemini models for generating contextual answers
• Multiple Model Support: Choose between different Gemini models (2.0 Flash, 1.5 Pro, 1.5 Flash)
• Interactive UI: Clean, user-friendly interface with suggested questions
• Source Context: View the document chunks used to generate answers
• Session Management: All processing happens in-memory with session state management

Outputs 📸

The outputs/ folder contains 4 demonstration screenshots showing the complete workflow:

1. Main Interface

Initial application interface with configuration panel, API key setup, and upload section

2. PDF Processing

PDF processing in progress showing file upload confirmation and loading states

3. Question Interface

Successfully processed PDF with active Q&A interface and suggested questions

4. Generated Answer

AI-generated response with source context and session information

Installation 🚀

1. Clone the repository

   git clone <your-repo-url>
   cd pdf-rag-assistant

2. Create a virtual environment

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies

   pip install -r requirements.txt

4. Get a Gemini API key

   • Visit Google AI Studio
   • Create a new API key
   • Keep it secure for use in the application

Usage 💡

1. Start the application

   streamlit run app.py

2. Configure the application

   • Enter your Gemini API key in the sidebar
   • Select your preferred Gemini model

3. Upload and process PDF

   • Upload a PDF file using the file uploader
   • Click "Process PDF" to extract text and create embeddings

4. Ask questions

   • Enable question answering with the checkbox
   • Type your question or use suggested questions
   • View AI-generated answers with source context

Technical Architecture 🔧

Core Components

• Text Extraction: PyPDF2 for extracting text from PDF files
• Text Chunking: Smart chunking with sentence boundary detection and overlap
• Embeddings: all-MiniLM-L6-v2 sentence transformer model
• Vector Store: FAISS for efficient similarity search
• AI Model: Google Gemini for generating contextual answers
• Frontend: Streamlit for the web interface

Data Flow

1. PDF Upload → Text extraction using PyPDF2
2. Text Processing → Chunking with overlap and boundary detection
3. Embedding Generation → Convert chunks to vector embeddings
4. Vector Storage → Store embeddings in FAISS index
5. Query Processing → Convert user question to embedding
6. Similarity Search → Find relevant document chunks
7. Answer Generation → Use Gemini to generate contextual answers

File Structure 📁

pdf-rag-assistant/
├── app.py                 # Main Streamlit application
├── requirements.txt       # Python dependencies
├── README.md             # This file
└── outputs/              # Output screenshots (4 files)
    ├── 1.jpg  # Main interface
    ├── 2.jpg  # PDF processing
    ├── 3.jpg  # Q&A interface
    └── 4.jpg  # Generated answer

Configuration Options ⚙️

Gemini Models

• gemini-2.0-flash-exp: Fastest response, good for quick queries
• gemini-1.5-pro: Most capable, best for complex questions
• gemini-1.5-flash: Balanced speed and capability

Processing Parameters

• Chunk Size: 1000 characters (configurable in code)
• Chunk Overlap: 200 characters (configurable in code)
• Similar Chunks Retrieved: 5 (configurable in code)
• Embedding Model: all-MiniLM-L6-v2 (cached for performance)

Troubleshooting 🔍

Common Issues

1. PDF Processing Fails

   • Ensure the PDF contains readable text (not just images)
   • Try with a different PDF file
   • Check that the file is not corrupted

2. Slow Embedding Generation

   • First-time model loading takes a few minutes
   • Subsequent uses are faster due to caching
   • Consider using smaller PDF files for testing

3. API Key Issues

   • Verify your Gemini API key is valid
   • Check that you have remaining quota
   • Ensure the key has proper permissions

4. Memory Issues

   • For large PDFs, consider splitting them into smaller sections
   • The application stores all data in memory during the session
   • Restart the application to clear memory if needed

Performance Tips

• The embeddings model is cached for better performance
• Use the "Clear Session" button to reset if needed
• Smaller, focused questions often yield better results
• Try rephrasing questions if answers aren't satisfactory

Dependencies 📦

• streamlit: Web application framework
• PyPDF2: PDF text extraction
• google-generativeai: Google Gemini API client
• sentence-transformers: Text embedding generation
• faiss-cpu: Efficient vector similarity search
• numpy: Numerical computations
• torch: PyTorch for transformer models

Contributing 🤝

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

License 📄

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

Acknowledgments 🙏

• Google AI for the Gemini API
• Hugging Face for the sentence transformers
• Facebook Research for FAISS
• Streamlit team for the amazing framework

Future Enhancements 🚀

• Support for multiple file formats (DOCX, TXT, etc.)
• Batch processing of multiple PDFs
• Conversation history and chat interface
• Custom embedding model selection
• Document comparison features
• Export functionality for Q&A sessions
• Advanced filtering and search options

---

Happy Document Analysis! 🎉

For questions or support, please open an issue in the repository.