VE's API request for retrieving files from Commons is already with the right set of parameters that will invoke media search once it becomes a default profile.
We could slightly enhance it (if desirable) by passing along a language, which would help us inform which entities to include (or exclude, because they're likely irrelevant/undesirable in that language). This could be either:
- the wiki's content language
- the user's interface language
I suggest the interface language, because:
- that's the principle we use on Commons as well
- that can be changed by the user should they need to tweak things to find a better image (they can't change the default wiki language)