pdfPulse

Overview

RAG PDF Chatbot is an advanced document analysis application that enables users to upload PDF documents and interact with their content through an intelligent, context-aware question-answering system. By leveraging cutting-edge AI technologies, this application transforms static PDF documents into interactive knowledge bases.

Key Features

• PDF Text Extraction: Seamlessly extract text from uploaded PDF documents
• Semantic Search: Utilize advanced embedding techniques to find contextually relevant information
• AI-Powered Responses: Generate precise, context-aware answers using Google's Gemini AI
• Vector Database Integration: Implement efficient information retrieval with Pinecone vector database

Technology Stack

• Streamlit
• Google Gemini AI
• Pinecone Vector Database
• PyPDF2
• Python 3.8+

Preview

Prerequisites

• Python 3.8 or higher
• Pinecone API Key
• Google Gemini API Key

Installation

1. Clone the repository:

git clone https://github.com/leeh-nix/pdfPulse.git
cd pdfPulse

2. Install required dependencies:

pip install -r requirements.txt

Configuration

Create a .secret file in the .streamlit folder (optional, for local development):

GEMINI_API_KEY=your_gemini_api_key
PINECONE_API_KEY=your_pinecone_api_key

Running the Application

streamlit run app.py

Usage Instructions

1. Navigate to the Streamlit application
2. Enter your Gemini and Pinecone API keys in the sidebar
3. Upload a PDF document
4. Ask questions about the document's content
5. Receive contextually accurate responses

How It Works

The application implements a Retrieval-Augmented Generation (RAG) workflow:

1. Extract text from uploaded PDF
2. Split text into semantic chunks
3. Generate vector embeddings for chunks
4. Store embeddings in Pinecone vector database
5. When a query is made, retrieve the most relevant chunks
6. Generate a response using retrieved context

Security Note

• API keys are used locally and not stored
• Sensitive information remains in memory during the session

Limitations

• Supports PDF documents only
• Requires active internet connection
• Response accuracy depends on document complexity and AI model capabilities

Contributing

Contributions are welcome! Please submit pull requests or open issues on the GitHub repository.

Acknowledgments

• Streamlit
• Gemini AI
• Pinecone