🐳 Flask ML API + Kubernetes Deployment

This project demonstrates how to containerize a simple Flask-based machine learning API, deploy it to a Kubernetes cluster with Minikube, and serve predictions through a /predict endpoint.

🚀 Features

• Flask API for serving ML model predictions
• Docker containerization
• Kubernetes deployment via Minikube
• RESTful /predict endpoint
• Local testable setup with curl/Postman
• SSH-authenticated GitHub repository

📁 Project Structure

.
├── app.py                # Flask API app
├── model.pkl             # Serialized ML model (generated)
├── generate_model.py     # Script to create model.pkl
├── requirements.txt      # Python dependencies
├── Dockerfile            # Image definition
├── deployment.yaml       # Kubernetes deployment config
├── service.yaml          # Kubernetes service config
└── .gitignore

---

🧰 Prerequisites

• Python 3.9+
• Docker Desktop
• Minikube
• kubectl
• GitHub SSH key setup (for pushing)

---

🔧 Setup Instructions

1. Clone the repo

git clone git@github.com:kevinmastascusa/flask-k8s-deploy.git
cd flask-k8s-deploy

2. Generate the model file

pip install -r requirements.txt
python generate_model.py

3. Build and push Docker image

docker build -t kzmastascusa/flask-model-api:latest .
docker push kzmastascusa/flask-model-api:latest

4. Start Minikube and deploy

minikube start
kubectl apply -f deployment.yaml
kubectl apply -f service.yaml

5. Access the service

minikube service flask-api-service

You’ll get a URL like http://127.0.0.1:64893

---

📡 Test the API

POST /predict

Request

curl -X POST http://127.0.0.1:64893/predict   -H "Content-Type: application/json"   -d '{"features": [1, 2]}'

Response

{
  "prediction": 0
}

---

🧠 Author

Kevin Mastascusa
🧑‍💻 GitHub: @kevinmastascusa

---

📄 License

MIT License – free to use, share, or build on.