As an open source Postgres extension, pgvector lets you store and query vector embeddings alongside your relational data, using the same tables, transactions, and tooling you already run. Last October, Alex Jacobs published a post called “The Case Against pgvector” that’s made the rounds in engineering circles. His argument is that the wave of blog posts evangelizing pgvector as a drop-in replacement for dedicated vector databases glosses over operational realities of running it at scale (I won’t go into those here, but the blog is…
