Hybrid semantic/text search

[mbertrand]

Description/Context

Experiment with various approaches to running semantic search queries with opensearch, and combining the results from standard text-based search.

Plan/Design

Some relevant links:
https://opensearch.org/docs/latest/search-plugins/hybrid-search/
https://blog.reactivesearch.io/knn-search-with-opensearch-and-openai-embeddings-an-in-depth-guide
https://opster.com/guides/opensearch/opensearch-machine-learning/how-to-set-up-vector-search-in-opensearch/
https://opster.com/guides/opensearch/opensearch-machine-learning/opensearch-hybrid-search/

[mbertrand]

Related doc: https://docs.google.com/document/d/1rERx31uYHDDMDQXgMfaaEI02ucQ_Qz09133m1GddKU4/edit?usp=sharing

Brief summary: Opensearch does not support paging, aggregations, or nested embeddings when performing hybrid searches (yet), but elasticsearch does. Performing separate semantic and text searches, then combining results programatically, is more or less feasible but aggregation counts would not be accurate and paging/scoring would require some custom implementation.