Talk2PDF - Chat with your Documents

Talk2PDF is an intelligent web application that allows you to have interactive conversations with your PDF documents. Upload a PDF, and the app will enable you to ask questions about its content, leveraging powerful AI models to provide accurate answers from the text and even explain images within the document.

This application is built with Flask for the backend, LangChain for orchestrating language model interactions, and Google's Gemini models for state-of-the-art text and vision understanding.

---

Features

• Interactive Chat Interface: Ask questions in natural language and get responses in real-time.
• PDF Content Analysis: Processes the entire text content of your uploaded PDF to provide comprehensive answers.
• Image Understanding (Optional): Extracts images from the PDF, analyzes them, and can answer questions about their content.
• Deduplication: Intelligently detects and skips duplicate images to save processing time.
• Easy Setup: Run the application with just a few simple commands.
• Secure API Key Handling: Prompts for your API key if not found in the environment, avoiding hard-coding.

---

How It Works

The application follows a multi-step workflow to process your documents and answer questions intelligently.

User Flow

1. Upload a PDF file via the web interface.
2. Choose between Text Only or Text & Image analysis.
3. Ask questions through the chat interface.
4. Receive AI-generated answers based on the PDF's content.

Backend Processing (LangChain Flow)

1. Upload PDF: The user uploads a PDF file through the web interface.
2. Store File: The file is temporarily stored on the server.
3. Setup QA System: The core question-answering system is initialized.
4. Text Splitting: The document’s text is split into smaller chunks.
5. Generate Embeddings: Text chunks are converted into embeddings using Google’s AI.
6. Create RetrievalQA Chain: A LangChain-based chain is created to match questions with relevant content.
7. Ask Question: The user enters a question.
8. Display Answer: The system retrieves the relevant chunks and generates a human-like answer.

---

Getting Started

Follow these steps to set up and run the application locally.

Prerequisites

• Python 3.8 or higher
• pip (Python package manager)

Installation

1. Clone the repository:

   git clone https://github.com/nishanthnaa52/Talk2pdf
   cd Talk2pdf

2. Install dependencies:

   pip install -r requirements.txt

---

Configuration

This app uses the Google Gemini API, which requires an API key.

Step 1: Get an API Key

Visit Google AI Studio to generate your API key.

Step 2: Provide the API Key

Choose one of the two methods:

• Recommended (via .env file):
  Create a .env file in the project root and add:

  GOOGLE_API_KEY="your_gemini_api_key_here"
  

• Manual (on runtime):
  If no .env file is found, the app will prompt for your key in the terminal.

---

Usage

1. Start the Flask server:

   python app.py

2. Open your browser at:

   http://127.0.0.1:5000
   

3. Upload your PDF:

   • Click “Choose File” and select a PDF document.
   • Choose between:
     • Text Only: For text content analysis.
     • Text & Image: To analyze both text and images.

4. Start Chatting:

   • After the PDF is processed, the chat interface appears.
   • Ask any question related to the document and receive answers instantly.

---

Technologies Used

• Backend: Flask
• AI Orchestration: LangChain
• Language & Vision Models: Google Gemini (langchain-google-genai)
• Vector Store: FAISS (Facebook AI Similarity Search)
• PDF Processing: PyMuPDF (fitz)
• Image Processing: Pillow, imagehash
• Frontend: HTML, CSS, JavaScript

---

Contributing

Pull requests are welcome! For major changes, please open an issue first to discuss what you'd like to change.

---

Contact

Feel free to connect on GitHub for questions or suggestions.