Vector search
On this pageWhat it is
What it is
Vector search finds items that are similar in meaning. A piece of text, such as a help article, is turned into a long list of numbers called an embedding, and the database stores it. A search query is converted the same way, and the database returns the closest matches. A search for "cancel my plan" can then find an article titled "How to end your subscription". Supabase offers this inside its database.
Why it matters
Keyword search misses anything that uses different words. Vector search is the basis for AI features such as support assistants that answer from your own documents, recommendations and duplicate detection. Keeping the vectors in the same database as the rest of your data means one system to run.
It matters if you are building AI features on your own data. It adds little for ordinary filtering and sorting, and it is a technical feature, so expect developer involvement.
What to check
- How many vectors and how large a dimension size are supported?
- Can similarity search be combined with ordinary filters, such as a customer or date?
- How does the speed hold up on your data volume?
- Which embedding models can you use, and can you change them later?
Tools with Vector search
Backend platform built on managed Postgres with auth, APIs, realtime subscriptions, and file storage.