Med Gemma - Medical Multimodal Chatbot (Radiology)

Med Gemma is a medical chatbot based on the edwardlo12/medgemma-4b-it-Q4_K_M language model, accessible via a Chainlit Chat interface. Note: The used model is image-text-to-text. On M-Series Mac or eventually PC with GPU or suffisant ram and vram, I recommend to install LM Studio and download a multimodal model in GGUF-Format or mlx 4 bit vision for Mac for a quicker start. Medgemma from google is the first multimodal model with its minimalistic size of 4B parameters. It might be only capable of performing easy radiology vision tasks I guess. Check ollama medgemma vision. Because in the meanwhile new vision models of medgemma are available on ollama model search simililar to the one chosen. You might want to use one of those instead.

Prerequisites

Before you begin, ensure that Ollama is installed and running.

1. Install Ollama: Visit ollama.com and download the version for your operating system.

2. Download the MedGemma Model: Open your terminal and run the following command to download the specific model:

   ollama pull edwardlo12/medgemma-4b-it-Q4_K_M

   Ensure the Ollama service is running in the background before starting the application.

Installation

You can install the project dependencies using either Conda or pip.

Option 1: Using Conda

1. Create a new Conda environment (recommended):

   • download & install miniconda

   conda create -n medgemma python=3.11  # You can choose a different Python version if desired
   conda activate medgemma

2. Install dependencies: Create a requirements.txt file (if not already present) with the following content:

   // filepath: requirements.txt
   chainlit
   langchain
   langchain-community
   ollama

   Then install the packages:

   pip install -r requirements.txt

   (Note: Although we are using Conda, pip within a Conda environment is often the easiest way to install Python packages that are not directly available as Conda packages or to use specific versions from requirements.txt.)

Option 2: Using pip

1. Create a virtual environment (recommended):

   python -m venv venv
   source venv/bin/activate  # On macOS/Linux
   # venv\Scripts\activate  # On Windows

2. Install dependencies: Create a requirements.txt file (if not already present) with the following content:

   // filepath: requirements.txt
   chainlit
   langchain
   langchain-community
   ollama

   Then install the packages:

   pip install -r requirements.txt

Running the Application

1. Ensure your (Conda or virtual) environment is activated.
2. Ensure the Ollama service is running and the edwardlo12/medgemma-4b-it-Q4_K_M model is available.
3. Navigate to the project directory in your terminal.
4. Start the Chainlit application:

   chainlit run main.py -w

   The -w flag enables automatic reloading on code changes.

Accessing the Chat in the Browser

After starting the application, you will see output in the terminal similar to this:

Your app is available at http://localhost:8000

Open your web browser and navigate to the displayed address (defaults to http://localhost:8000, but may vary if the port is already in use). You should now see the Med Gemma chat interface.