MedPrompt AI Chatbot: Advanced Medical Assistant 🩺

MedPrompt AI is an innovative medical assistant chatbot engineered to deliver accurate, empathetic, and comprehensive information based on a robust medical knowledge base. It harnesses the power of cutting-edge large language models (LLMs) and an advanced Retrieval-Augmented Generation (RAG) system, designed to be your intelligent companion for medical inquiries.

---

🚀 Features

• Dual LLM Integration: Seamlessly integrates with both Google's Gemini-1.5-Flash and Mistral AI models, ensuring diverse perspectives and robust medical insights.
• Retrieval-Augmented Generation (RAG): Built upon a sophisticated RAG architecture that queries a FAISS vector index of pre-processed medical text chunks, providing context-aware and highly accurate responses grounded in factual data.
• Smart Conversation History: Efficiently manages and intelligently summarizes long chat histories to maintain crucial conversational context throughout your interactions.
• Streamed Responses: Delivers AI-generated content in a real-time streaming fashion, enhancing user engagement and providing a dynamic chat experience.
• Intuitive User Interface: Features a responsive and modern web interface, developed with Flask, HTML, CSS (Tailwind CSS), and JavaScript, ensuring seamless and visually appealing interactions.
• Voice Capabilities (STT & TTS): Enhances accessibility with both speech-to-text (STT) for microphone input and optional text-to-speech (TTS) for voice output of AI responses.
• Crystal Clear Formatting: AI responses are meticulously formatted using markdown, incorporating bold text, clear headings (##), and organized lists (*) for superior readability and information retention.
• Quick Action Buttons: Offers pre-defined quick action buttons for common medical queries, simplifying the process of initiating conversations and exploring topics.

---

🛠️ Technologies Used

• Backend Framework: Flask (Python)
• Large Language Models:
  • Google Gemini (gemini-1.5-flash)
  • Mistral AI (mistral-tiny)
• Embeddings: sentence-transformers (model: BAAI/bge-base-en-v1.5)
• Vector Database: FAISS (faiss-cpu)
• Natural Language Toolkit (NLTK): For text processing utilities like sentence tokenization.
• Data Handling: numpy, pickle, requests.
• Frontend: HTML, CSS (Tailwind CSS), JavaScript, Font Awesome.
• Environment Variables: python-dotenv.

---

⚙️ Setup and Installation

Follow these steps to get the MedPrompt AI Chatbot up and running on your local machine.

Prerequisites

• Python 3.9+
• pip (Python package installer)

1. Obtain API Keys 🔑

The application requires API keys for Google Gemini and Mistral AI to function.

• Gemini API Key: Obtain yours from the Google AI Studio.
• Mistral API Key: Get your key from the Mistral AI Platform.

Once you have your keys, create a .env file in the root directory of your project (e.g., MedPrompt AI Chatbot/.env) with the following content, replacing the placeholder values with your actual API keys:

GEMINI_API_KEY=YOUR_GEMINI_API_KEY
MISTRAL_API_KEY=YOUR_MISTRAL_API_KEY

Security Note: Remember to add .env to your .gitignore file to prevent your API keys from being accidentally committed to version control.

2. Dataset Information 📚

The MedPrompt AI Chatbot's RAG system is built upon a knowledge base derived from medical textbooks. While this project uses a custom dataset, the MedQA-USMLE Dataset on Kaggle is a relevant public medical dataset that can be referenced or used for similar applications.

3. Set Up Pre-processed Data 🗃️

The chatbot relies on pre-processed medical text chunks and a FAISS index for its RAG capabilities. You have two convenient options to obtain these:

Option A: Download Pre-built Data (Recommended) 👇

You can directly download the medical_faiss_index.bin and medical_text_chunks.pkl files from the following Google Drive link. After downloading, create a new directory named medprompt_ai_data in the root of your project and place these files inside it:

MedPrompt AI Data on Google Drive

Option B: Generate Data Using model.ipynb 🧑‍💻

If you prefer to generate the data yourself, you can run the model.ipynb notebook on Google Colab.

• Recommendation: Utilize a T4 GPU runtime in Colab for significantly faster processing.
• This notebook will guide you through:
  1. Installing necessary libraries.
  2. Downloading NLTK data (handled automatically by the notebook).
  3. Mounting your Google Drive to access your medical textbooks (you'll need to provide your own .txt medical textbook files in a folder like Medical_Textbooks on your Google Drive).
  4. Processing the textbooks, generating embeddings, and building the FAISS index.
  5. Saving medical_faiss_index.bin and medical_text_chunks.pkl to your specified SAVE_DIR (e.g., /content/drive/MyDrive/medprompt_ai_data).

Ensure these generated files are then accessible to your app.py (e.g., by placing them in a medprompt_ai_data folder in your project directory).

4. Install Dependencies 📦

You can install all the required Python packages using pip.

Method 1: Using requirements.txt (Standard)

If you have a requirements.txt file (which lists all project dependencies), simply run:

pip install -r requirements.txt

Method 2: Installing Directly via pip (Alternate)

Alternatively, you can install each core dependency by listing them in a single pip install command:

pip install flask python-dotenv google-generativeai sentence-transformers faiss-cpu nltk numpy requests scipy

Method 3: Using a Virtual Environment (Recommended Best Practice) 🐍

It's highly recommended to install dependencies within a virtual environment. This isolates your project's dependencies, preventing conflicts with other Python projects on your system.

1. Create a virtual environment:

   python -m venv venv

2. Activate the virtual environment:
   • On macOS/Linux:

     source venv/bin/activate

   • On Windows:

     .\venv\Scripts\activate

3. Install the dependencies within the activated environment:

   pip install flask python-dotenv google-generativeai sentence-transformers faiss-cpu nltk numpy requests scipy

   Once installed, you can run your application. To exit the virtual environment, simply type deactivate.

5. Run the Application ▶️

Once you have set up your API keys and the data files, you can launch the Flask application:

python app.py

The application will start, and you can usually access it by opening your web browser and navigating to http://127.0.0.1:5000/.

Note: For production deployment, it is recommended to use a production-ready WSGI server like Gunicorn instead of Flask's built-in development server.

---

💬 Usage

1. Access the Chatbot: Open your web browser and go to the address where the Flask app is running (e.g., http://127.0.0.1:5000/).
2. Ask a Question: Type your medical question into the input box at the bottom of the chat interface.
3. Send Message: Press Enter or click the send button (the paper plane icon) to submit your question.
4. Receive Response: The AI will swiftly process your question, retrieve relevant context, and synthesize a comprehensive answer, which will be streamed to the chat interface.
5. Voice Input/Output:
   • Click the microphone icon to speak your question directly.
   • Click the volume icon to toggle voice output for the AI's responses on or off.
6. New Chat: Click the "New Chat" button to start a fresh conversation at any time.

---

DISCLAIMER: This AI tool is for informational purposes only and is not a substitute for professional medical advice. Always consult a qualified healthcare provider.