So had a quick play with this, and appreciate it’s early days, but I’m a huge fan of having a Prometheus Endpoint built into the node, as it’s saves a whole heap of re-work running a separate node-exporter linking into an exposed API on the node 
Things of note:
-
Would be nice to specify the TCP port the Endpoint runs on, and if not present then do the whole random port thing to keep it easy to deploy for a single instance.
-
The docker local hosts hack doesn’t scale, and is a complete nightmare
across Win$ows / Linux / Docker / Podman / Swarm etc as they all do / might / don’t support this - can see you’ve done it for security, but maybe an allowed Prometheus Server IP list, although that shouldn’t really be down to the node to manage hmm 
-
Metric names are nice and verbose, and meaningful which is really helpful. Was able to add a few custom ones into Grafana and the name alone was enough to work out what data was exposed.
Looking forward to seeing how this progresses, my slightly changed dashboard below - keep up the great work ! 
