🎥 Click here to watch the demo

Adv_Rag

Adv_Rag is an advanced document processing and question-answering system built with various cutting-edge technologies, enabling users to upload PDF or image documents, process them, and ask questions based on the extracted content. The system uses a custom AI-powered backend, powered by Google's Gemini API, and integrates with popular Python libraries and cloud services.

Features

• PDF and Image Upload: Upload PDF or image files for text extraction using the Vision API and pdfplumber.
• Question Answering: Use an advanced AI model (Gemini 2.0 Flash) to answer questions based on the content of the uploaded documents.
• JWT Authentication: Secure the backend routes with JWT tokens, ensuring only authenticated users can access the system.
• Vector Storage: Vectorize document text using FAISS for efficient search and retrieval of relevant document chunks.
• Smart Confidence Scoring: Provide confidence scores along with the answers to indicate the model's certainty.

Technologies Used

• Backend: Python, Flask
• Frontend: ReactJS, Tailwind CSS (via npm run dev)
• Text Extraction: pdfplumber for PDF text extraction, Google Cloud Vision API for image-to-text extraction.
• Vector Database: FAISS for efficient document search and retrieval.
• AI Model: Google Gemini 2.0 Flash for natural language processing and question answering.
• Authentication: JWT-based token system for user authentication.

Installation Guide

Prerequisites

• Python 3.10 or +
• Node.js and npm for running the frontend
• Google Cloud Account for API keys and credentials
• Gemini API Key for integrating the AI model

Setup

1. Clone the repository:

git clone https://github.com/yourusername/Adv_Rag.git
cd RAG

2. Install the backend dependencies:

pip install -r requirements.txt

3. Setup your environment variables:

Create a .env file in the root of the project and add the following:

GEMINI_API_KEY=your_api_key_here

4. Set up the Google Cloud Vision credentials:

• Download the service account credentials from your Google Cloud console.
• Set the GOOGLE_APPLICATION_CREDENTIALS environment variable to point to the path of your credentials file:

export GOOGLE_APPLICATION_CREDENTIALS="path_to_your_credentials.json"

5. Install the frontend dependencies:

cd Bot
npm install

6. Run the backend server:

In the backend directory, run:

python server.py

7. Run the frontend:

In the frontend directory, run:

npm run dev

The frontend will be available at http://localhost:3000 by default, and the backend will be accessible at http://localhost:5000.

API Endpoints

1. /login [POST]

Description: Authenticate a user and return a JWT token.

Request:

{
  "email": "email",
  "password": "password"
}

Use the following credentials to log in:

• Email: [email protected]
• Password: deepak

Response:

{
  "token": "your_jwt_token_here"
}

2. /upload [POST]

Description: Upload a PDF or image file for text extraction and processing.

Request (Form Data):

• file: The PDF or image file to be uploaded.

Response:

{
  "message": "PDF processed and vector DB updated.",
  "success": true
}

3. / [POST]

Description: Ask a question based on the uploaded document.

Request:

{
  "question": "What is the main idea of the document?"
}

Response:

{
  "answer": "The main idea of the document is...",
  "confidence": 80
}

How It Works

1. Upload PDF/Image: Users can upload PDF or image files, and the system will extract text from them using either pdfplumber (for PDFs) or Google Vision API (for images).

2. Text Vectorization: The extracted text is split into chunks and vectorized using FAISS for efficient retrieval.

3. Question Answering: After processing, users can ask questions, and the system will generate relevant answers based on the extracted text using Google’s Gemini 2.0 Flash model.

4. Confidence Scoring: The system returns answers with a confidence score, helping users gauge the reliability of the response.

Additional Notes

• Google Cloud Vision API: This is required for image-to-text extraction. Be sure to configure your Google Cloud credentials correctly.
• JWT Authentication: Secure your system with JWT tokens to ensure only authorized users can access protected routes.
• Error Handling: The system includes comprehensive error handling to ensure robustness, with clear messages in case of failures.

Contributing

If you'd like to contribute, please fork the repository, make your changes, and submit a pull request. Ensure to follow the existing code style and provide detailed commit messages.

License

This project is licensed under the MIT License - see the LICENSE file for details.