MediChat
MediChat is an advanced medical chatbot designed to assist with clinical queries and provide information based on medical literature. It leverages state-of-the-art models and embeddings to deliver accurate and reliable responses.
Demo for MediChat
Features
Medical Expertise:
Provides information based on extensive training on medical literature.
Advanced Model:
Utilizes the Llama-2-7B-Chat model from Hugging Face.
Vector Database:
Incorporates Chroma DB for efficient data retrieval.
High-Quality Embeddings:
Uses sentence-transformers/all-MiniLM-L6-v2 embeddings from Hugging Face.
Model Details
Llama-2-7B-Chat Model:
- Source: Link
- Description: An open-source large language model optimized for chat-based interactions, capable of understanding and generating human-like text.
Database and Embeddings
Chroma DB:
A vector database that allows for efficient storage and retrieval of embeddings.
Embeddings Model:
- Source: Link
- Description: A small, fast, and high-quality embedding model that provides dense vector representations of text.
Training Data
MediChat is trained on embeddings derived from the following medical books:
- Clinical Emergency Medicine (PDFDrive.com)
- Current Essentials of Medicine
- Gale Encyclopedia of Medicine Vol. 4 (N-S)
Installation
To set up MediChat, follow these steps:
Clone the Repository:
- git clone https://github.com/Gyanbardhan/MediChat.git
- cd MediChat
Install Dependencies:
- pip install -r requirements.txt
Download the Model and Embeddings:
- Llama-2-7B-Chat model: Download from Hugging Face and place it in the models directory.
- Sentence-Transformers embeddings: Download from Hugging Face and place them in the embeddings directory.
- Set Up Chroma DB
Run the MediChat application with the following command:
Interact with MediChat via the provided interface, asking medical questions and receiving expert responses based on the embedded medical literature.
