LEVEL 02 · LESSON 16
Vector databases
A vector database helps find similar representations; the surrounding system makes them useful.
What it is
A vector database stores numerical representations and supports searching for nearby vectors. Many also store text and metadata or combine vector search with other search methods.
Why it matters
It makes similarity search practical over larger collections, but it is only one component of a knowledge system.
A business example
A company indexes thousands of support articles. A query vector helps find semantically similar articles, while metadata restricts results to the customer’s product version.
How it actually works
An embedding model produces vectors that are stored in an index. At search time, the system compares a query vector with indexed vectors. Approximate search can trade some retrieval accuracy for speed. Content updates and deletions must be reflected in the index. Changing embedding models requires compatible representations, often meaning re-embedding stored content. Evaluate retrieval quality, filtering, access controls, and operational cost together.
A common misconception
A vector database does not understand whether a document is true or whether its contents should be shown to a particular user.
ONE SENTENCE TO REMEMBER
A vector database helps find similar representations; the surrounding system makes them useful.
CHECK YOUR UNDERSTANDING
A vendor says its vector database will make your company assistant accurate and secure. What is the most realistic view?
Show the answer
Answer: D.
A vector database makes similarity search practical over large collections, and many also store text and metadata. It does not know whether a document is true or who should see it. Retrieval quality, filtering, updates, and access controls must be evaluated together.
Take it into your next meeting
What happens to the search index when a source policy is withdrawn?
One more concept, understood.
Mark this lesson complete to keep your place.