Medical Chatbot Llama2

This repository contains the code for the Medical Chatbot Llama2, a sophisticated AI-powered chatbot designed to provide medical information and assistance. The project leverages advanced models and frameworks to ensure accurate and efficient performance.

Screenshots

Table of Contents

• Features
• Installation
• Usage
• Models and Technologies
• Lessons Learned
• License

Features

• Utilizes the all-MiniLM-L6-v2 model from Hugging Face for embeddings.
• Employs Pinecone Vector Database with cosine similarity for indexing and searching.
• Integrates the llama-2-7b-chat.ggmlv3.q4_0.bin model for chatbot responses.
• Offers an experimental option to use PineconeHybridSearchRetriever with dotproduct metric, available in hybridSearch.ipynb.
• Beautifully designed and implemented with Flask.

Installation

Prerequisites

• Python 3.8 or higher
• Flask
• Hugging Face Transformers
• Pinecone
• Llama 2 model files

Steps

1. Clone the repository:

   git clone https://github.com/bhaveshk22/Medical_Chatbot_Llama2.git
   cd medical-chatbot-llama2

2. Install the required Python packages:

   pip install -r requirements.txt

3. Set up Pinecone:

   • Create a Pinecone account and get your API key.
   • Set up your Pinecone index with cosine similarity.

4. Download the llama-2-7b-chat.ggmlv3.q4_0.bin model file and place it in the model directory.

Usage

Storing the Index

• Run the store_index.py:

  python store_index.py

Running the Chatbot

1. Start the Flask server:

   python app.py

2. Open your web browser and go to http://127.0.0.1:8080/ to interact with the chatbot.

Using PineconeHybridSearchRetriever (Experimental)

1. Open hybridSearch.ipynb in your preferred Jupyter Notebook environment.
2. Follow the instructions to set up and test the PineconeHybridSearchRetriever with dotproduct metric.

Models and Technologies

• Hugging Face Embeddings Model: all-MiniLM-L6-v2
  • Used for generating embeddings for text.
• Pinecone Vector Database:
  • Stores the embeddings and performs similarity search with cosine metric.
• Chatbot Model: llama-2-7b-chat.ggmlv3.q4_0.bin
  • Generates responses based on user input.
• Flask:
  • Provides a web interface for interacting with the chatbot.

Lessons Learned

Throughout the development of the Medical Chatbot Llama2, several key lessons were learned:

1. Model Selection: Choosing the right model for embeddings and chatbot responses significantly impacts performance and accuracy. The all-MiniLM-L6-v2 model from Hugging Face and the llama-2-7b-chat.ggmlv3.q4_0.bin model provided a good balance of speed and accuracy.

2. Vector Database Efficiency: Pinecone's vector database with cosine similarity proved to be efficient for storing and retrieving embeddings. Experimenting with the PineconeHybridSearchRetriever and different metrics like dotproduct can further optimize performance.

3. Flask Integration: Integrating advanced AI models with Flask for the web interface required careful consideration of API design and asynchronous processing to handle user interactions smoothly.

4. Scalability Considerations: As the complexity of the chatbot increases, considerations for scaling the application, such as load balancing and distributed computing, become critical to maintain performance.

5. User Experience: A well-designed user interface significantly enhances user interaction and satisfaction. Iterative testing and feedback helped refine the Flask-based web interface to be more intuitive and responsive.

License

This project is licensed under the MIT License.