Cloud search syncs your data from mongoDB by tailing the database oplog. All updates to your mongoDB database are automatically reflected in your search index in a few milliseconds.
The search engine only uses the in-memory skiplist indexes to process queries. For regex queries for example, the index keys are scanned and the regex operation applied on each key and then a set union operation is performed on the set associated with each key to complete the query. The Textx and text operator uses an inverted index. Geo spatial queries leverages google's S2 library. All other attributes perform a btree search to find the skip list associated with each key. Logical operators execute using the resulting skip list indexes from btree scans.
Because cloud search is mongoDB compatible and supports almost all CRUD operations except mongodb map-reduce, it is possible to use cloud search as your only database. The search engine supports CRUD operations the same way mongoDB does. You could also keep only your search documents in the cloud search engine and your other documents in mongoDB.