Vector Database
Last reviewed 3 Nov 2026
A vector database stores text, images or other content as lists of numbers called embeddings, so software can find items with similar meaning rather than matching exact words. It is the search engine behind most AI assistants that answer from company documents.
Key Facts
| Used for | Semantic search, AI chatbots over company documents, recommendations, duplicate detection |
|---|---|
| Common options | pgvector inside PostgreSQL, plus dedicated services such as Pinecone, Weaviate and Qdrant |
| For most SMEs | pgvector in an existing PostgreSQL database is usually enough and keeps data in one place |
| Cost drivers | Number of documents, how often content changes, and query volume |
How it fits into an AI assistant
- Documents are split into passages.
- Each passage is converted into an embedding.
- A question is converted the same way.
- The vector database returns the passages closest in meaning.
- The LLM answers using those passages.
This is the retrieval step in RAG.
Do you need a separate product
Often not. If your application already uses PostgreSQL, the pgvector extension handles hundreds of thousands of passages comfortably. Dedicated services make sense at very large scale or for special features.
Frequently Asked Questions
Is my data safe in a vector database?
It needs the same access control and encryption as any database. Embeddings can reveal content, so protect them accordingly.
Does it replace our normal database?
No. It sits alongside it, often inside the same PostgreSQL instance.
Related Glossary
Need help implementing this in your business?
Turbo Bytes Consulting helps businesses streamline operations and build custom software architectures that scale without chaos.