Similar to the other comments, would be nice to be able to add some `fetcher` plugins, that know how to pull data from external resources (Graphite, Google Analytics etc), preprocess it, and then push it to the dashboard. If you go down that route however, then the question is why run a server in the first place?
It kinda goes against the essence of a command-line tool, which should be lightweight, quick and can run (almost) anywhere. I'm not trying to discourage you, but am thinking whether instead of server-based architecture, this might be better off as a client with different 'plugins' for different types of data (e.g. a sysinfo plugin that gets memory and cpu, a graphite plugin to fetch data from your graphite server, an SQL plugin etc)
(disclaimer: I wrote Giraffe, a graphite dashboard, so perhaps I'm suffering from tunnel-vision based on the graphite architecture).