Yatai is a Kubernetes-native deployment operator that enables scalable, GitOps-friendly deployment and management of BentoML machine learning services.
Model Deployment at Scale on Kubernetes 🦄️
Yatai is designed for DevOps and ML engineering teams who want to integrate BentoML model deployment into Kubernetes-based CI/CD and GitOps workflows. It simplifies deploying, scaling, and managing ML services on any Kubernetes cluster, making it ideal for production-grade ML service orchestration.
Yatai for BentoML 1.2 is currently under construction; users should refer to the Yatai 2.0 proposal for upcoming changes. The provided installation script is intended for development and testing only; production deployments require following the official installation guide. On macOS, using the hyperkit driver with minikube is recommended to avoid Docker Desktop network issues. Maintaining the kubectl port-forward session is necessary to access the Yatai web UI during setup and usage.
Install latest minikube from https://minikube.sigs.k8s.io/docs/start/
Install latest Helm from https://helm.sh/docs/intro/install/
Start a minikube Kubernetes cluster with `minikube start --cpus 4 --memory 4096`
On macOS, use the hyperkit driver to avoid Docker Desktop network limitations
Verify minikube cluster status with `minikube status`
Ensure kubectl is configured with the minikube context using `kubectl config current-context`
Enable ingress controller with `minikube addons enable ingress`
Run the quick install script: `bash <(curl -s "https://raw.githubusercontent.com/bentoml/yatai/main/scripts/quick-install-yatai.sh")` to install Yatai and dependencies
For production deployment, follow the official Installation Guide at https://docs.yatai.io/en/latest/installation/index.html
Port-forward Yatai service with `kubectl --namespace yatai-system port-forward svc/yatai 8080:80`kubectl --namespace yatai-system port-forward svc/yatai 8080:80
Port-forwards the Yatai service to localhost for web UI access
YATAI_INITIALIZATION_TOKEN=$(kubectl get secret yatai-env --namespace yatai-system -o jsonpath="{.data.YATAI_INITIALIZATION_TOKEN}" | base64 --decode) echo "Open in browser: http://127.0.0.1:8080/setup?token=$YATAI_INITIALIZATION_TOKEN"
Retrieves the Yatai initialization token and prints the URL to complete admin setup
bentoml yatai login --api-token {YOUR_TOKEN} --endpoint http://127.0.0.1:8080
Logs into Yatai CLI using an API token generated from the Yatai web UI
bash <(curl -s "https://raw.githubusercontent.com/bentoml/yatai/main/scripts/quick-install-yatai.sh")
Installs Yatai along with PostgreSQL and MinIO on a minikube cluster for development/testing