Step 8 of 15
TL;DR: A vector database is a special filing cabinet for embeddings. It stores all those meaning-numbers and finds the closest matches super fast, even across millions of items.
Remember the map where similar words sit close together? A vector database is the giant, super-fast map keeper. You say "find things near this spot," and it instantly hands you the closest neighbors.
It's a database built to store embeddings (those lists of meaning-numbers) and to answer one question really fast: which stored items are most similar to this one? Regular databases are bad at that; vector databases are made for it.
Normal databases find exact matches — like "show rows where name = John." Vector databases find close matches by meaning, even across millions of records in a blink. That's a totally different kind of search.
Here's the combo: you turn your documents into embeddings, store them in a vector database, and later ask "what's most related to this question?" The database returns the best-matching chunks, which you then hand to the LLM. That handoff is RAG — the next step.
Tools like Pinecone, Weaviate, Chroma, and pgvector (an add-on for Postgres) are common vector databases. As a builder you don't need all of them — just one you're comfortable with.
Not always. For a few documents you can skip it. But once you have lots of content to search by meaning, a vector database keeps it fast and organized.
Not anymore. Many are cloud services you can start using in minutes, and some run right inside databases you may already know, like Postgres.
Old search matches keywords. Vector search matches meaning using embeddings, so it finds the right answer even when the exact words don't match.
Want to know exactly where you stand on the road from using AI to building it? Take the free 3-minute AI Builder Scorecard and get your single biggest gap.