Building a Vector Database That Never Decompresses Your Vectors
go
dev.to
Preamble: What vector search is all about If you've spent any time near an LLM in the last couple of years, you've heard the term "embeddings." The idea is simple: take some text (or an image, or whatever), feed it through a neural network, and out comes a list of numbers, a vector. Texts that are semantically similar end up with vectors that point in roughly the same direction. The immediate question is: given a query vector, how do you find the most similar ones in your database?