PDF Chat Assistant 📚

A Streamlit-based application that allows users to upload PDF documents and interact with their content using natural language queries. The app uses OpenAI's GPT-4 model and FAISS vector storage for efficient document retrieval and question answering.

Features

• PDF document upload and processing
• Natural language querying of PDF content
• Conversation memory for context-aware responses
• Response caching for improved performance
• Cost estimation for embeddings and queries
• Clear chat history functionality
• Vector store persistence for faster subsequent loads

Prerequisites

• Python 3.8 or higher
• OpenAI API key

Installation

1. Clone the repository:

git clone https://github.com/rajat343/multiple_pdf_qa.git
cd multiple_pdf_qa

2. Create and activate a virtual environment (recommended):

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

3. Install the required packages:

pip install -r requirements.txt

4. Create a .env file in the project root directory and add your OpenAI API key:

OPENAI_API_KEY=your_api_key_here

Usage

1. Start the Streamlit app:

streamlit run app.py

2. Open your web browser and navigate to the URL shown in the terminal (typically http://localhost:8501).

3. Upload one or more PDF documents using the file uploader.

4. Start asking questions about the content of your PDFs in the chat interface.

Project Structure

pdf-chat-assistant/
├── app.py              # Main application file
├── requirements.txt    # Python dependencies
├── .env               # Environment variables
├── uploaded_pdfs/     # Directory for stored PDFs
└── embeddings/        # Directory for stored vector embeddings

Features in Detail

Document Processing

• PDFs are processed once and stored locally
• Document embeddings are cached for faster subsequent loads
• Multiple PDFs can be processed simultaneously

Conversation Management

• Maintains conversation history for context-aware responses
• Allows clearing of chat history
• Caches question-answer pairs for improved performance

Cost Management

• Displays estimated costs for embeddings generation
• Shows per-query costs for API usage
• Uses efficient retrieval methods to minimize API calls

Technical Implementation

The application uses:

• langchain for document processing and chat chain management
• FAISS for efficient vector similarity search
• OpenAI's GPT-4 for generating responses
• Streamlit for the web interface
• Document hashing for efficient storage and retrieval

Limitations

• PDF processing may take longer for large documents
• API costs can accumulate with heavy usage
• Requires stable internet connection for API calls
• Maximum token limit applies based on GPT-4 model constraints

Cost Considerations

The application uses OpenAI's API which has associated costs:

• Embedding generation: $0.0001 per 1K tokens
• Query processing:
  • Input: $0.0015 per 1K tokens
  • Output: $0.002 per 1K tokens

Contributing

Feel free to submit issues, fork the repository, and create pull requests for any improvements.

Acknowledgments

• Built with Streamlit
• Uses LangChain for document processing
• Powered by OpenAI GPT-4