april8-examples

Example repository demonstrating how to deploy machine learning models with April8 — a GitOps model serving platform built on KServe.

How it works

1. Add an april8.yaml to your repo declaring your model deployments
2. Push to main
3. April8's webhook picks it up, downloads your model files, uploads to S3, and creates a KServe InferenceService
4. Your model is live at a predictable HTTPS endpoint

Examples in this repo

Model	Framework	Task	Notebook
anomaly-detector	sklearn	Network anomaly detection	notebooks/01-sklearn-anomaly-detector.ipynb
churn-predictor	xgboost	Customer churn prediction	notebooks/02-xgboost-churn-predictor.ipynb
sentiment-classifier	huggingface	Sentiment analysis	notebooks/03-huggingface-sentiment.ipynb

Quickstart

python -m venv .venv
source .venv/bin/activate

pip install -r requirements.txt

# Generate model files by running the notebooks
jupyter notebook notebooks/

After running the notebooks, model files appear in models/. Commit and push to deploy:

git add models/
git commit -m "feat: train and add model files"
git push

april8.yaml reference

See CLAUDE.md for the full schema reference, framework requirements, tier defaults, and endpoint URL patterns.

version: "1"
project: <team>/<project>

deployments:
  my-model:
    model: ./models/my-model/       # path to model directory
    framework: sklearn               # sklearn | xgboost | pytorch | tensorflow | onnx | huggingface
    tier: staging                    # dev | staging | production
    mlflow_model: my-model          # optional: register in MLflow
    resources:
      cpu: "1"
      memory: "2Gi"
    scaling:
      min_replicas: 0
      max_replicas: 5

Supported frameworks

Framework	Model file	ServingRuntime
sklearn	model.joblib	kserve-sklearnserver
xgboost	model.bst	kserve-xgboostserver
huggingface	standard HF directory	kserve-huggingfaceserver
pytorch	model.mar	kserve-torchserve
tensorflow	SavedModel 1/ directory	kserve-tensorflow
onnx	model.onnx	kserve-tritonserver