These serve as an example. You will probably want to use kustomize or some other configuration/deployment tool.
We recommend using a StatefulSet instead of deployment to mitigate high churn in long term metrics - statefulset' stable pod name format (which is propagated to Prometheus' instance label) makes it easier for Prometheus to calculate the SLO over long time periods.
We recommend to either building own docker image based on the upstream one with the configuration baked in or including configuration as a versioned configmap(s), in order to simplify rollbacks.
In this example we use it without the versioning just for simplification.