should the data source operate by implementing interfaces defined in the repository

imho, you won't gain much with interfaces there. Just inject it directly into Repository implementation.

alright, thanks.. will explore this option

IMHO it depends. Sometimes it's handy to make cache in memory implementation and start to use it, and later to replace it with sqlite implementation, this is a good fit interface approach. But if you know that you won't have another implementations, for backend for instance, then no need for excessive interfaces.