Chat With PDF

Overview

This project, Chat With PDF, is a Streamlit-based application that enables users to upload a PDF file and interact with its content using a chatbot powered by large language models (LLMs). The application leverages embeddings, vector stores, and a question-answering pipeline to extract and analyze text from PDFs, enabling users to ask meaningful questions about the uploaded content.

Features

• PDF Upload and Parsing: Users can upload a PDF file, which is processed to extract text from all pages.
• Text Chunking: Text from the PDF is split into manageable chunks using LangChain’s RecursiveCharacterTextSplitter for efficient processing.
• Embeddings: The application uses HuggingFace embeddings (via the sentence-transformers/all-MiniLM-L6-v2 model) to create vector representations of text chunks.
• Vector Storage: The project uses FAISS for similarity search to retrieve relevant chunks based on user queries.
• Question Answering: A transformer-based pipeline (from HuggingFace’s deepset/roberta-base-squad2 model) answers user questions by analyzing the relevant text chunks.
• Streamlit Interface: A simple and interactive UI for uploading PDFs, asking questions, and viewing results.

Technologies Used

Libraries and Frameworks

• Streamlit: Frontend framework for building interactive web applications.
• LangChain: For text processing and chunking.
• HuggingFace: Provides pre-trained models for embeddings and question answering.
• FAISS: Vector storage and similarity search.
• PyPDF2: For reading and extracting text from PDF files.
• dotenv: For environment variable management.

File Details

1. app.py:

   • The main script that implements the application’s logic.
   • Defines the UI, PDF processing pipeline, embeddings generation, and QA pipeline.

2. requirements.txt:

   • Contains the list of dependencies required to run the application.

Installation

Follow these steps to set up and run the project locally:

1. Clone the repository:

   git clone https://github.com/mahesh-diwan/chat-with-pdf.git
   cd chat-with-pdf

2. Create a virtual environment:

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

3. Install the required packages:

   pip install -r requirements.txt

4. Run the application:

   streamlit run app.py

5. Open the app in your browser at http://localhost:8501.

Usage

1. Upload a PDF File:
   • Use the "Upload your PDF file" button to upload a document.
2. Ask Questions:
   • Enter your question in the text box, and the app will retrieve relevant text chunks and provide an answer.
3. View Results:
   • The answer will be displayed along with any supporting context.

Example Use Case

• Upload a research paper and ask specific questions about methodologies or results.
• Analyze long reports by querying particular sections without manually searching through the document.

Future Improvements

• Add support for multiple file formats (e.g., Word, Excel).
• Optimize embeddings for faster performance with larger documents.
• Enhance UI/UX with additional features like visualization of search results.

Developer Information

Author: Mahesh
Contact: Feel free to reach out for any queries or suggestions.

User Interface

Chat WIth PDF