All terms
Glossary
Vector database
A database that stores vectors and finds similar entries quickly, the basis for search by meaning.
The difference to a classic database is the query: not equal or unequal, but similar. That needs special index methods trading accuracy against speed.
There are dedicated products and extensions for existing databases, for example for PostgreSQL.
How you notice it
- There are hundreds of thousands of text chunks to search.
- Search should capture meaning, not just words.
- Filters by tenant or area are required.
Frequently asked
Do I need a dedicated database for this?
Below roughly a million chunks, usually not. PostgreSQL with pgvector is enough and saves another system in operations. A specialized database pays off with very large corpora or very high query rates.