How can I (ideally per code location or even more fine grained) control which agent is running an job/op/asset materialization? I see that for accelerator hardware i.e. GPUs https://dagster.slack.com/archives/C02LJ7G0LAZ/p1677528206949639?thread_ts=1677527782.079859&cid=C02LJ7G0LAZ kubernetes tags for specific jobs might be sufficient. But all jobs are still using the same dagster-cloud agent to launch. However, my problem is a bit different: I am in a hybrid cloud setup where some jobs need to launch on-premise and others in the cloud. Currently, I think it would be easiest to deploy separate agents and tag these (if possible) and somehow apply the tags on jobs/ops/assets so the right agent will be chosen. Is this a possibility? If not, how does the way forward look like?

Currently the solution is to create a second deployment for on prem jobs and have the on prem agent point at that.

Are there any roadmap plans to simplify this as indicated above?

It’s not on the roadmap currently