Summary
In this chapter, you explored a variety of concepts related to vector search. The chapter delved into how high-dimensional vectors produced from embedding models can be useful measures of semantic similarity among the unstructured data passed into those models. It examined the HNSW index and how it can be used to accelerate vector similarity comparisons between a query vector and many indexed vectors.
The chapter then illustrated how this type of index can be applied in various real-world contexts by large organizations, including such architecture patterns as RAG, semantic search, and RPA. Finally, the chapter reviewed some of the best practices for building vector search systems within MongoDB Atlas, ranging from ingestion time considerations, such as metadata extraction, to deployment model considerations, such as dedicated search nodes.
In the next chapter, you will discover the crucial aspects of designing AI/ML applications. You will learn how to effectively manage...
