Skip to main content

Embeddings & Vector Search

Intermediate

An embedding turns a piece of text into a list of numbers (a vector) that captures its meaning. Texts with similar meaning get vectors that are close together — even if they share no words. That's the trick behind semantic search and RAG.

The intuition​

Imagine every sentence placed as a point in a huge multi-dimensional space, arranged so that similar meanings sit near each other. "How do I reset my password?" lands near "I forgot my login," far from "best pizza in Rome."

  • Keyword search matches literal words ("password" finds "password").
  • Semantic search matches meaning — "I can't sign in" finds the password-reset doc even without the word "password."

Best results often combine both (hybrid search).

How a vector search works​

  1. Embed your documents (usually split into chunks) and store the vectors in a vector database.
  2. At query time, embed the query.
  3. Find the nearest vectors (by cosine similarity / distance).
  4. Return those chunks — typically to feed into RAG.

Practical notes​

  • Chunking matters. Too big = noisy matches; too small = lost context. Tune it.
  • Use one embedding model consistently — vectors from different models aren't comparable.
  • Metadata + filters (date, source, type) make retrieval far more precise.
  • A vector DB isn't always needed — for small corpora, a simple in-memory search is fine.

Next​