Explainer2 min read

What Is a Vector Database (and Do You Need One)?

Vector databases went from obscure to mandatory-sounding in two years. You may need one. You may also already have one hiding inside the database you run today.

TrueCodeAI Engineering
Agents, Voice & ML practice
Published
RAGExplainerInfrastructure
Network of connected glowing cubes

What it stores

A vector database stores embeddings — lists of numbers that represent the meaning of a piece of text or an image — and finds the stored items closest to a query vector. "Closest" means "most similar in meaning", which is what makes semantic search work.

Your options

Vector storage options
OptionGood forTrade-off
Vector extension on PostgresUp to millions of vectors, apps already on PostgresOne less system; tuning needed at scale
Managed vector databaseLarge scale, many tenants, high query volumeAnother vendor and bill
Search engine with vector supportHybrid keyword + semantic searchHeavier to operate
In-memory indexPrototypes, small static setsNot durable

Do you need a dedicated one?

For most business assistants — tens of thousands to a few million chunks — a vector extension on the Postgres you already run is enough and keeps permissions, backups and joins in one place. A dedicated vector database earns its place at very large scale, very high query rates, or when you need features like multi-tenant isolation out of the box.

What matters more than the database

  • Chunking quality — bad chunks cannot be rescued by a fast index.
  • Hybrid search — combining keyword and vector search fixes exact-term misses.
  • Reranking — a second pass that reorders the top results improves answer quality.
  • Permission filtering — every query must respect who can see what.

Frequently asked questions

Is a vector database a replacement for my main database?

No. It is a search index. Your source of truth stays where it is.

How much does one cost?

Postgres extensions add little cost. Managed services charge by storage and queries; at business-assistant scale it is usually modest.

Can I switch later?

Yes, if embeddings and chunking live in your own pipeline. We keep them portable by design.

Tell us what you want to exist.

We reply within 24 hours at hello@truecodeai.com with how we would build it.

Get a fixed price WhatsApp