In this guest post, Frank Liu, Head of ML&AI at Zilliz, explores the intricate realm of vector search, comparing vector databases, vector search plugins, and vector search libraries.
This article is too surface-level to influence any decisions. Simply using the analogy of adding vector search functionality to a preexisting database is akin to electrifying a ICE-powered vehicle (which often IS actually a great idea!) isn't nearly enough. There are far more tradeoffs that must be considered, especially if you're integrating into a preexisting system. I'm sure Milvus is a fine technology, but it doesn't have nearly the tenure that Postgres or Elasticsearch do. While those might not be "optimal" in retrieval times, they're battle-tested tools with far more capabilities than Milvus.
This article is too surface-level to influence any decisions. Simply using the analogy of adding vector search functionality to a preexisting database is akin to electrifying a ICE-powered vehicle (which often IS actually a great idea!) isn't nearly enough. There are far more tradeoffs that must be considered, especially if you're integrating into a preexisting system. I'm sure Milvus is a fine technology, but it doesn't have nearly the tenure that Postgres or Elasticsearch do. While those might not be "optimal" in retrieval times, they're battle-tested tools with far more capabilities than Milvus.
Nice post but I miss old TheSequence with vendor-neutral content.
Please single-space coding.