Java design pattern for Mongo DB

What are the good design principles for java as a client with MongoDB as no-SQL db. Should java execute simple queries and pull all data on its end and process it on server ( Jboss etc) OR the queries should be executed on MOngoDB and less processing on Java end.

Please see that the main requirement on my end is queries on MongoDB should be configuration based (less coding in java). i.e. Java Client request--> configuration -->java request to MongoDB using config.

This way less development cycles should be spend for new requirement or queries. Any pointers to good design principles will be appreciated.