That's the hard part. The rest of this can probably be handled by a combination of:
Rendering the dot graphs on the client side. (Probably the best bet for filtering out the standard library inclusions - can just remove every node which has a label that doesn't start with the project root path.)
A 'bound' parameter to limit how deep of a graph gets generated in the first place. It's unimportant if this happens client-side or server-side for this command, but it will matter much more for the call hierarchy graph, so I'm inclined to see what happens there before implementing this.
The following features would make the inclusions commands more useful:
inclusion-hierarchy
.inclusion-hierarchy
, note which identifiers are referenced from each inclusion. This will make it easy to sanity-check your inclusions.