Medical Chatbot

This is a Flask-based medical chatbot application that leverages Hugging Face embeddings and Pinecone vector stores and Llama-2 7b model for question answering on medical topics. The chatbot allows users to interact in real-time and stores chat history in a PostgreSQL database.

Features

• Real-time chat interface.
• Medical question answering leveraging Hugging Face embeddings and Pinecone vector stores.
• Persistent storage of chat history in a PostgreSQL database.

Requirements

• Python 3.10.9
• Flask
• psycopg2
• Pinecone
• Hugging Face transformers
• dotenv

Installation

1. Clone the repository:

   git clone https://github.com/NaveenPradhaph/D-Nexus
   cd D-Nexus

2. Install dependencies:

   pip install -r requirements.txt

3. Create a .env file

4. Set up environment variables:

   PINECONE_API_KEY = "xxxxx-xxxx-xxxx"
   PINECONE_API_ENV = "xxxxx"
   
   NEON_USERNAME = "xxxx"
   NEON_PASSWORD = "xxxx"
   NEON_HOST = "xx-xxx-xxxxxx"
   NEON_PORT = "xxxx"
   NEON_PROJECT = "xxxx"

5. Download the model from huggingface:

   Select the model according to the system requirements if the RAM is less than 16GB it is advisable to select the lesser bit model such as 2-bit or 4-bit model

   You can use this TheBloke/Llama-2-7B-GGML

   save the model in the folder named model

6. Data for the model:

   The data for the model are multiple pdf's that are store in the folder named data

   Important note the pdf's must be editable not scanned pdf

   run the following code only once to store the vectors in the Pinecone vector stores

   python store_index.py

7. Run the application:

   python app.py

Usage

• Open your web browser and go to http://localhost:8080.
• Type your medical-related questions in the chatbox and hit Enter.
• Receive real-time responses from the chatbot.
• Chat history is stored in the PostgreSQL database and can be viewed on the web interface.

API Endpoints

• /get: Retrieve chat history from the database.
• /chat: Send messages to the chatbot.