In addition to the thin client there is a client which can eagerly load data into your application called DaVinci. You can think of Davinci as a means to do local ephemeral or persistent cacheing within your application. This is ideal for usecases which have a not super huge dataset but want very low latency. Larger datasets can be accomodated with partitioning in davinci (basically, user app managed partition subscription), but can get complicated. So not always strictly preferred.