It's not really document-centric, but rather expecting to capture metrics about running Python code (decorators to measure execution time, etc.). There are ways to generate the documents it expects, though.
Pull appears to be the preferred method for prometheus, but push makes more sense based on the fact that archstats is already the puller (however, push is still allowed)
Could maybe avoid the asyncio clients by using the pull model and asynchronously updating the statistics, but I don't like it much. Will hold off for now.
Prometheus, as suggested by Murali?
Grafana supports these out of the box: