What Is a Vector Database? Storage Built for Similarity Search
DICTIONARY · AI

What Is a Vector Database?

A vector database stores embeddings and finds the closest matches to a query quickly, at scale.

In plain English

Comparing a query against a few thousand vectors is easy. Comparing it against millions, fast enough for a live application, is the problem these databases solve.

They are infrastructure for retrieval. If your corpus is small, you do not need one; several general databases now support vector search adequately.

What to know

Stores vectors
Embeddings plus the metadata needed to return the source.
Approximate search
Trades a little accuracy for a great deal of speed.
Filtering
Combining similarity with ordinary filters like date or category.
Optional at small scale
A few thousand documents rarely justify dedicated infrastructure.

Why it matters

For anyone building retrieval over a substantial document set, this is the component that makes it practical. For everyone else it is a decision worth deferring until the corpus is genuinely large.

Common mistakes

×Adopting one before there is enough content to need it.
×Storing vectors without the metadata needed to cite the source.
×Never re-indexing after changing the embedding model.
×Assuming the database is at fault when the chunking was the problem.

FAQs

Do I need a dedicated vector database?

Only at scale. Many standard databases now handle vector search well enough.

What matters most in a RAG system?

Chunking and retrieval quality, well ahead of the database choice.

WRITTEN BY TARIQ SALLAM
Marketing Consultant. Entrepreneur. Content Creator.

I'm a marketing consultant, entrepreneur and content creator. I help businesses grow through practical marketing, websites, SEO, content and AI.

More About Tariq →