Postgres is all you need, even for vectors

And sometimes numpy is all you need, especially for POCs when the embeddings fit in memory.