Documentation

Install Sceptre with Helm on local Kubernetes

Run the complete tabular AutoML workspace on Docker Desktop Kubernetes, k3d, or another conformant local cluster.

You do not need local Python, Node.js, PostgreSQL, SeaweedFS, MLflow, a private registry, or a source checkout. Kubernetes pulls the public versioned images and Helm installs the stack.

What you need

An x86-64/AMD64 workstation with a supported local Kubernetes cluster
Helm 3 or 4 and a working kubectl context
A default dynamic StorageClass
At least 4 logical CPU cores; 6 or more is the practical recommendation
16 GiB host memory, with 10–12 GiB available to the cluster
At least 50 GiB free disk space for volumes, images, and training cache

The chart accepts Kubernetes 1.27 or newer for API compatibility. Use an actively supported upstream minor for a new installation.

Select the Sceptre release

CHART_VERSION="<version shown in Artifact Hub>"

helm show chart oci://registry-1.docker.io/maponyacharles/sceptre \
  --version "$CHART_VERSION"

Find the current immutable chart version on the Sceptre Artifact Hub listing.

Generate local-only secrets

umask 077
JWT_SECRET="$(openssl rand -hex 32)"
DATABASE_PASSWORD="$(openssl rand -hex 32)"
OBJECT_STORE_PASSWORD="$(openssl rand -hex 32)"

cat > .env.sceptre-local.yaml <<EOF
auth:
  jwtSecret: "${JWT_SECRET}"
postgresql:
  auth:
    password: "${DATABASE_PASSWORD}"
seaweedfs:
  auth:
    secretKey: "${OBJECT_STORE_PASSWORD}"
EOF

unset JWT_SECRET DATABASE_PASSWORD OBJECT_STORE_PASSWORD

Keep this file for upgrades, do not share it, and never commit it. The complete repository guide includes an equivalent PowerShell path for Windows.

Install the public Helm chart

helm upgrade --install sceptre oci://registry-1.docker.io/maponyacharles/sceptre \
  --version "$CHART_VERSION" \
  --namespace sceptre \
  --create-namespace \
  --set environment=local \
  --values .env.sceptre-local.yaml \
  --wait --wait-for-jobs --timeout 30m

Verify and open Sceptre

helm status sceptre --namespace sceptre
kubectl --namespace sceptre get pods,jobs,pvc
helm test sceptre -n sceptre
kubectl -n sceptre port-forward service/sceptre-ui 8080:80

Open http://127.0.0.1:8080. Long-running pods should be ready, migration Jobs complete, and persistent volume claims bound.

What Helm creates

Application

FastAPI, the React UI, training configuration, namespace-scoped permissions, and inference support.

Data services

PostgreSQL, SeaweedFS, MLflow, durable volumes, database bootstrap, and migrations.

NextReview production readiness

For Windows and Linux setup, secrets, first use, upgrades, and troubleshooting, use the complete repository installation guide.