I believe we agreed that it can be done either way. The client side does not have to know where the meow originates from as there are different possible architectures:
full medication overview prebuilt in a dedicated repository;
core medication overview prebuilt in a dedicated repository, other related data (eP, eD) requested from other sources and added where relevant;
no overview prebuilt, everything constructed on the fly from whichever sources needed. There is possibly a need to have some kind of medication overview repository anyway, because there is likely some information that would not come from anywhere else. However, medication overview should be technically constructable based on ePrescriptions and eDispensations only (not advised from functional perspective).