MedicAI: Your Private AI Health Assistant

MedicAI is a private AI solution designed to assist users with health-related queries in a secure and personalized manner. It is built to be accessible even to non-technical users, providing a user-friendly interface for interacting with a powerful language model.

⚠️ MedicAI is not a replacement for qualified medical advice and should only be used as a helpful supplement. Always consult a healthcare professional for medical concerns. MedicAI should also tell you this as you use it.

---

Table of Contents

1. Core Features
2. How It Works
3. Prerequisites
4. Usage
   • Optional Personal Settings
   • Advanced Users
5. License

---

Core Features

• Private AI Solution: All interactions are processed locally, ensuring your data remains private.
• Personalized Responses: Tailored responses based on user-configured personal health information.
• Accessible to All: Simple setup and usage, designed for non-technical users.

---

How It Works

MedicAI operates as a two-service system with custom "system" prompts for the language model (LLM):

1. Services:

   • App Service: Hosts the user interface and handles chat interactions.
   • Ollama Service: Runs the LLM and processes requests.

2. System Prompts:

   • A core system prompt is provided to the LLM on first launch to define its behavior as medicAI.
   • If the user configures any Personal Settings, all subsequent chat interactions will embed the settings so that responses are tailored to the user.

---

Prerequisites

• Docker: Install Docker from docker.com. For most users, Docker Desktop is what you should install.

---

Usage

Starting the Application

To start MedicAI, clone or download this repository and run the following command in the project directory:

docker compose up --build

NOTE: This can take a few minutes to start up the first time you run it. This is because Ollama will need to download the LLM (~2GB), which may take some time depending on your internet speed. If you attempt to send chat messages before Ollama is ready, you will see an Error in the UI. Just be patient and wait for Ollama to retrieve the model components!

Once the services are running, open your browser and navigate to http://localhost:7860 to access the app.

Stopping the Application

Press Ctrl+C in your terminal to stop it.

To also destroy the container resources that were created:

docker compose down

Optional Personal Settings

For improved response quality, configure the Personal Settings section. This ensures that all bot responses are tailored to your specific needs.

Advanced Users

Changing Models

To switch the default model, update the entrypoint in the docker-compose.yaml file under the ollama service. For example:

entrypoint: /root/health-models/entrypoint.sh llama3.2

Enabling GPU Support

• Windows/Linux: Follow the Docker GPU support guide to configure GPU usage.
• macOS: Run the services locally instead of using containers (see below).

Running without Docker

You can also run this application directly on your hardware. You may want to do this if you have a GPU that is currently unavailable to Docker (e.g. Apple Silicon). In this case, you should install and run ollama directly (not in a container), and you may also need to run the python app directly.

# Download and install Ollama. Mac: https://ollama.com/download/mac

# Start ollama (skip this if you are running Ollama app with the UI)
ollama serve

# -- In a separate shell --

# Create the custom "health" model from the Modelfile
# This will download the base model (llama3.2 by default)
ollama create health -f ./health-models/Modelfile

# Once the custom model finishes creating, you can start the UI
# Install gradio requirement
pip install -r requirements.txt

# Start the UI
python app.py

License

This project is licensed under the MIT License. See the LICENSE file for details.