Medical Chatbot

A Flask-based application that uses RAG (Retrieval Augmented Generation) to answer medical questions by retrieving relevant information from medical PDFs.

Overview

This project implements a conversational AI chatbot specialized in medical information using:

• LangChain framework for RAG implementation
• Flask for the web interface
• Pinecone as the vector database
• OpenAI for generating responses
• HuggingFace embeddings for text processing

The chatbot retrieves information from medical documents that have been processed, chunked, and stored in a Pinecone vector database to provide accurate and concise medical information.

Features

• User-friendly chat interface
• Retrieval of relevant medical information from indexed PDFs
• Concise responses (limited to three sentences)
• Similarity-based document retrieval

Prerequisites

• Python 3.7+
• Pinecone API key
• OpenAI API key

Installation

1. Clone the repository:

   git clone https://github.com/yasithS/Healthie-AI
   cd medical-chatbot
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Create a .env file in the root directory with the following variables:

   PINECONE_API_KEY=your_pinecone_api_key
   OPENAI_API_KEY=your_openai_api_key
   

Usage

Data Preparation and Indexing

1. Place your medical PDF documents in a Data/ directory in the project root.

2. Run the indexing script to process and store the documents in Pinecone:

   python store_index.py
   

Running the Application

1. Start the Flask application:

   python app.py
   

2. Open a web browser and navigate to:

   http://localhost:8080
   

3. Interact with the chatbot by typing medical questions in the chat interface.

Project Structure

medical-chatbot/
├── app.py                  # Main Flask application
├── Data/                   # Directory for medical PDF documents
├── Research/               # Directory contains notebooks               
├── setup.py                # Package setup file
├── requirements.txt        # Project dependencies
├── store_index.py          # Script to process and index documents
├── src/
│   ├── __init__.py
│   ├── helper.py           # Utilities for document loading and processing
│   └── prompt.py           # System prompt for the chatbot
├── static/
│   └── style.css           # CSS styles for the chat interface
└── templates/
    └── chat.html           # HTML template for the chat interface

How It Works

1. Document Processing:

   • Medical PDFs are loaded and split into smaller chunks
   • Each chunk is embedded using HuggingFace's MiniLM model

2. Information Retrieval:

   • User queries are processed to find the most relevant document chunks
   • The system retrieves the top 3 most similar chunks using Pinecone

3. Response Generation:

   • Retrieved contexts are provided to OpenAI's model
   • Responses are generated following the system prompt guidelines (concise, three sentences maximum)

Customization

• Modify the system prompt in src/prompt.py to change the chatbot's behavior
• Adjust the chunk_size and chunk_overlap in src/helper.py to optimize retrieval performance
• Change the temperature and max_tokens in app.py to control response length and creativity

License

MIT License

Acknowledgements

• Created by Yasith Gunawardhana
• [email protected]