Clarion 🚀

A lightweight FastAPI service for serving machine learning classification models — containerized and ready for local development with S3-compatible storage (MinIO).

---

📌 Features

• ✅ FastAPI web server with /predict endpoint
• ✅ Loads a serialized ML model (.pkl) for live predictions
• ✅ Integrates with S3-compatible storage (MinIO)
• ✅ Local Docker Compose setup for full-stack dev
• ✅ Clean Python package structure (src/clarion/)

---

🚀 Getting Started

1️⃣ Clone the repo

git clone https://github.com/chishxd/clarion.git
cd clarion

---

2️⃣ Local development

# Create and activate virtual environment
python -m venv .venv
source .venv/bin/activate

# Install in editable mode
pip install -e .

# Run the API server with hot reload
uvicorn clarion.app:app --reload

---

🐳 Docker Compose (Recommended)

Run the API server + MinIO together:

# Make sure you have a .env file with your MinIO credentials:
echo "AWS_ACCESS_KEY_ID=admin" > .env
echo "AWS_SECRET_ACCESS_KEY=root@123" >> .env

# Then build and run containers
docker compose up --build -d

• Clarion API: http://localhost:80
• MinIO Console: http://localhost:9001

---

🔬 Example API Request

curl -X POST "http://localhost:80/predict" \
  -H "Content-Type: application/json" \
  -d '{"PassengerId": 123, "Pclass": 1, ...}'

Example Response

{
  "Survived": 1
}

---

✅ Tests

Run unit tests with pytest:

pytest

---

⚙️ Roadmap

• Local FastAPI prediction endpoint
• Model versioning & download from MinIO
• Docker Compose setup
• Production cloud deployment (On hold — budget-dependent)

---

📜 License

Distributed under the MIT License.

---

✨ Credits

Clarion is built as part of the Cloud Computing for Data Science curriculum — and follows up on AutoCleanSE.

---

⚠️ Notes

• Keep .env in .gitignore — don’t commit credentials.
• For production, use real secrets + a secure MinIO setup.
• Current version is for local dev & student projects — not hardened for prod yet.

---