Context

Our existing retriever in langchain does not use the new $rankFusion or $scoreFusion built in operators.
See docs: https://www.mongodb.com/docs/atlas/atlas-vector-search/hybrid-search/
There is also a blog post and such.

Definition of done

What must be done to consider the task complete?

• Ask gaurab.aryal7@gmail.com for a short Product Description.
• Create a Scope / Design Doc. We now have 4 alternative ways to do Hybrid Search. (1) $rankFusion operator (2) $scoreFusion operator (3) The original pipeline (3b) The pipeline with additional weights that came from a pull-request of a community contributor.
• Given our work in pymongo-search-utils, and the importance of hybrid search generally, I propose that we consider creating a separate Retrievers, but that's what the design doc is for.
• Implementation
• Testing. I would specifically create test that compares the results with weightings that make them equivalent.