I have several chorma DB collections, and I want to route queries to the most relevant collection based on a similarity score. Ideally, the document with the highest similarity score within a collection would be retrieved. Can LlamaIndex leverage its retrieval capabilities to achieve this?