PDFfusion 📄✨

"Don't just read — ask."

PDFfusion is a modern, AI-powered web application that allows users to upload PDF documents and ask questions about their content . Built with Flask and LangChain, it provides an intuitive chat interface for document analysis.

🌟 Features

• Smart PDF Analysis: Upload any PDF and get intelligent answers to your questions
• Wikipedia Integration: Access external knowledge through Wikipedia API for comprehensive answers
• Dual Processing Modes: Choose between simple QA chain or advanced agent-based processing
• Modern UI: Beautiful, responsive interface with animated particles and glassmorphism design
• Drag & Drop Upload: Easy file upload with visual feedback
• Real-time Chat: Interactive chat interface for seamless Q&A experience
• Local AI Processing: Uses Ollama for privacy-focused, local AI processing
• Vector Search: Efficient document retrieval using ChromaDB for accurate context
• Agentic RAG: Advanced Retrieval-Augmented Generation with intelligent agent decision-making
• Simple QA Chain: Efficient question-answering using retrieval-augmented generation

🛠️ Technologies Used

• Backend: Flask (Python)
• AI/ML: LangChain, Ollama (Llama 3.1)
• RAG Architecture: Agentic RAG with tool integration and simple rag also (options)
• External Knowledge: Wikipedia API integration
• Vector Database: ChromaDB
• PDF Processing: PyPDF2
• Frontend: HTML5, CSS3, JavaScript
• Styling: Custom CSS with animations and modern design

📋 Prerequisites

Before running PDFfusion, make sure you have:

• Python 3.8 or higher
• Ollama installed and running
• Llama 3.1 model downloaded via Ollama

Installing Ollama and Llama 3.1

# Install Ollama (macOS/Linux)
curl -fsSL https://ollama.ai/install.sh | sh

# Pull Llama 3.1 model
ollama pull llama3.1

# Start Ollama service
ollama serve

🔧 Installation

1. Clone the repository

   git clone https://github.com/yourusername/pdffusion.git
   cd pdffusion

2. Create virtual environment

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

3. Install dependencies

   pip install -r requirements.txt

4. Create required directories

   mkdir uploads templates

5. Move HTML template

   # Make sure index.html is in the templates/ directory
   mv index.html templates/

📦 Dependencies

Create a requirements.txt file with:

Flask==2.3.3
langchain==0.1.0
langchain-community==0.0.10
langchain-ollama==0.1.0
langchain-text-splitters==0.0.1
PyPDF2==3.0.1
chromadb==0.4.18
requests==2.31.0
wikipedia==1.4.0

🚀 Usage

1. Start the application

   python app.py

2. Open your browser Navigate to http://localhost:5000

3. Upload a PDF

   • Click the upload area or drag & drop a PDF file
   • Supported format: PDF only

4. Ask questions

   • Type your question in the chat input
   • Get AI-powered answers based on the PDF content

📁 Project Structure

pdffusion/
├── app.py              # Main Flask application
├── lang.py             # LangChain processing logic
│                       # - Simple QA Chain
│                       # - Agentic RAG with Wikipedia
│                       # - PDF processing utilities
├── templates/
│   └── index.html      # Frontend template
├── uploads/            # Temporary file storage
├── requirements.txt    # Python dependencies
└── README.md          # This file

⚙️ Configuration

The application uses the following default settings:

• Upload folder: uploads/
• Ollama model: llama3.1
• Processing modes:
  • Simple QA Chain (default in current setup)
  • Agentic RAG with Wikipedia integration (available in code)
• Chunk size: 1000 characters
• Chunk overlap: 200 characters
• Vector search results: Top 3 relevant chunks
• Wikipedia search: Top 1 result, max 200 characters

Switching Between Processing Modes

The application includes two processing approaches:

1. Simple QA Chain (process_pdf_question_simple):

   • Fast, direct PDF-only processing
   • Currently used in app.py

2. Agentic RAG (process_pdf_question_with_agent):

   • Advanced agent with PDF retriever + Wikipedia tools
   • Makes intelligent decisions about which sources to use
   • To enable: Change app.py import to use process_pdf_question_with_agent

🔄 Future Enhancements

• Support for multiple PDF uploads
• Additional external knowledge sources (beyond Wikipedia)
• Document comparison features
• Export conversation history
• Advanced search filters
• Support for other document formats (DOCX, TXT)
• User authentication and session management
• Cloud deployment options
• Custom agent tools and workflows

📄 License

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

📧 Contact

Author - Khushi Kumari

⭐ If you found this project helpful, please give it a star!