Local Quickstart
Install the complete evaluation stack, open the dashboard, and create a real preview. The quickstart chart includes Envy, Istio, PostgreSQL, and a tea-shop application. No setup script, database configuration, or Envy CLI is needed.
Prerequisites
Section titled “Prerequisites”- A dedicated local Kubernetes cluster, such as Kind or Docker Desktop
Kubernetes, with
kubectlpointing at it. Use a cluster without an existing Istio installation. See Kind’s quick start if you need one. - Cluster-admin installation permissions: the bundled mesh creates cluster-wide CRDs and RBAC. Do not use a shared cluster for this quickstart.
- Helm:
brew install helmon macOS installs the standard Homebrew formula. - A default storage class for PostgreSQL’s 1 GiB volume. Kind and Docker Desktop normally provide one.
- Network access to download the chart and images from Docker Hub. Start with 4 CPUs and 8 GiB of Docker memory; initial image downloads can take several minutes.
No mesh or database needs to be installed beforehand. For a cluster with existing infrastructure, use Install a Release instead.
1. Install everything
Section titled “1. Install everything”helm upgrade --install envy oci://registry-1.docker.io/davey/envy-quickstart \ --version 0.9.0 --namespace envy-quickstart --create-namespace \ --wait --wait-for-jobs --timeout 10mUse the release name and namespace shown. This intentionally fixed local setup creates the mesh, database, Envy, and sample application, then registers the Tea shop / staging baseline. Helm waits for the setup Job to finish. Database credentials are generated automatically.
2. Open the dashboard
Section titled “2. Open the dashboard”kubectl -n envy-quickstart port-forward service/envy-ingress 8080:80Keep that terminal running and open http://127.0.0.1:8080. Sign in with admin / admin. Leave Demo simulation off: this installation has a real sample application.
The same port-forward handles the dashboard and preview traffic.
The baseline is at http://shop.envy.localhost:8080/products;
it shows tea priced at 1200 minor units (£12.00).
3. Create a working preview
Section titled “3. Create a working preview”- Open Previews, choose Create preview, and select Tea shop / staging.
- Select the pricing component to override.
- Choose Direct image and enter
davey/envy-demo:0.9.0-v2. - Give the preview a name, review the changes, and create it.
- Wait for readiness, then open the preview URL. Its root redirects to
/products.
The preview shows a price of 990 (£9.90). Reload the baseline: it still shows 1200 (£12.00). The storefront is shared; only the preview’s pricing service uses the new image.
When finished, destroy the preview in the dashboard. See the web interface walkthrough for more controls.
Local addresses
Section titled “Local addresses”Use 127.0.0.1:8080 for dashboard login. localhost is a different
authentication origin. CLI users can optionally install the CLI
and set ENVY_API_URL=http://127.0.0.1:8080.
Most browsers resolve *.localhost to loopback. If your system resolver does
not, use curl’s explicit mapping for the baseline or returned preview hostname:
curl --resolve shop.envy.localhost:8080:127.0.0.1 \ http://shop.envy.localhost:8080/productsPort 8080 must be available on your laptop. Stop another local service using it before opening the port-forward. No public DNS or certificates are needed.
Stop or remove the evaluation
Section titled “Stop or remove the evaluation”Stopping the port-forward closes local access; it does not delete workloads. Destroy all previews in the dashboard before uninstalling:
helm uninstall envy --namespace envy-quickstartThe PostgreSQL volume and credential Secret are retained so a reinstall can use the same database. Istio CRDs also remain. For a complete reset, delete the dedicated local cluster through Kind or Docker Desktop; this deletes its database contents too. Do not delete a cluster that contains other work.
Move to your own installation
Section titled “Move to your own installation”This chart is for local evaluation, with a default password, plain HTTP, and a single PostgreSQL pod. When you are ready to configure a shared installation, follow Install a Release with your managed database, mesh, authentication, and HTTPS settings.