🪸 Coral Bleach Predictor

This application predicts coral bleaching risk using two models:

• Coral Bleaching Predictor – our custom-trained ML model.
• LLaMA 3.1 – a large language model accessed via Ollama.

---

🔧 Setup Instructions

1. Start Ollama with LLaMA 3.1

Make sure Ollama is installed and running.

Start the LLaMA 3.1 model using:

ollama run llama3.1

By default, Ollama runs on http://localhost:11434.

---

2. Generate the Custom ML Model Files

Due to their size, the trained model files are not included in this repository.
You must generate them before running the app:

• src/models/scaler.pkl
• src/models/coral_bleaching_predictor.pkl

To generate these files, run the notebook:

notebooks/noaa_evaluation.ipynb

Once generated, place the files in the following directory:

app/backend/models/
├── coral_bleaching_predictor.pkl
└── scaler.pkl

---

3. Install Python Dependencies

Navigate to the backend directory and install required packages:

cd app/backend
pip install -r requirements.txt

---

4. Run the Backend API

Start the backend server:

cd api
python app.py

This will launch the backend on http://localhost:8000.

---

5. Test the Application in Your Browser

Open your browser and visit:

http://localhost:8000

You can now test coral bleaching predictions using:

• Coral Bleaching Predictor (our custom ML model)
• LLaMA 3.1 (via Ollama)

and use interactive chat powered by LLaMA 3.1, capable of answering questions related to coral reefs, ocean data, and bleaching risk.

---

🛠 Notes

• The backend handles predictions from both models.

• Make sure Ollama is running before using the Coral Reef AI Assistant.

• If any issues occur:

  • Confirm ports 8000 and 11434 are open.
  • Ensure LLaMA 3.1 is fully loaded before sending requests.

• Notebook Notes Before Running:

  • Please make sure to create a virtual env and attach in the notebook kernel
  • Run pip install -e . from the root of the repo

---