The ideal flow, in my opinion, would be we can have a separate repo that is responsible for talking to the cluster. If you want, you can have a flag on the open dataset generation. There could be a button here that says talk to the cluster if you want to expand.