Chat with your PDF Document!

This project put together by Manogya allows users to upload PDF documents, extract text from them, and interact with the content using a conversational AI model.

Features

• Upload a single or multiple PDF files.
• Extract text from the uploaded PDFs.
• Split the extracted text into manageable chunks.
• Store text embeddings locally using FAISS.
• Ask questions related to the uploaded PDF content and receive detailed answers.

Requirements

• Python 3.x
• Required libraries:
  • os
  • dotenv
  • pickle
  • google.generativeai
  • streamlit
  • PyPDF2
  • pathlib
  • langchain_google_genai
  • langchain
  • langchain_community

Installation

1. Clone the repository:

   https://github.com/manogyaguragai/Talk-to-PDF.git

2. Install the required libraries:

   pip install -r requirements.txt

3. Create a .env file in the root directory and add your Google API key:

   GOOGLE_API_KEY=your_google_api_key

Usage

1. Run the Streamlit application:

   streamlit run talk_to_pdf.py

2. In the Streamlit interface:

   • Upload your PDF files.
   • Click the "Submit" button to process the PDFs.
   • Ask questions related to the content of the uploaded PDFs in the provided input box.

Code Overview

• get_pdf_text(pdf_docs): Extracts text from all uploaded PDF files.
• get_text_chunks(text): Splits the extracted text into chunks.
• get_vector_store(text_chunks): Creates a local vector store for text embeddings.
• get_conversation_chain(retr): Creates a conversational chain with a predefined prompt and selected model.
• user_input(user_question): Handles user input, performs similarity search, and retrieves answers.
• main(): Main function to run the Streamlit application.

Points to Note

• It is recommended to fact-check the answers provided by this model.
• The information may sometimes be incomplete due to limited output size.