Vector database
Vektör veritabanı
D3 · Prompts, context, RAG, tools and agents
A vector database stores vector representations and supports retrieving records by vector similarity.
Review status: 2026-12-05
Technical explanation
Vector search can be provided by a dedicated database or an extension to an existing database.
Conceptual boundaries
Storing and searching vectors is different from generating embeddings or generating a final answer.
Provider-neutral example
Illustrative: a document service retrieves records with embeddings close to a query vector for later review.
Limitations
Approximate search can trade recall for speed; vector proximity alone does not establish factual correctness.
Related concepts
Atomic claims and evidence
1.1Weaviate Database
- Source
- Weaviate Database
- Source role
- Authoritative source
- Exact locator
- What is Weaviate?: objects, embeddings and semantic/hybrid search
- Supported claim
- Weaviate stores and indexes objects with vector embeddings; pgvector adds vector similarity search to Postgres.
- Last verification
- Review due
- Scope limitation
- These are two implementations, not an exhaustive taxonomy of databases.
1.2pgvector: vector similarity search for Postgres
- Source
- pgvector: vector similarity search for Postgres
- Source role
- Authoritative source
- Exact locator
- README: Storing; Querying; Indexing — exact and approximate nearest neighbor search
- Supported claim
- Weaviate stores and indexes objects with vector embeddings; pgvector adds vector similarity search to Postgres.
- Last verification
- Review due
- Scope limitation
- These are two implementations, not an exhaustive taxonomy of databases.
2.1pgvector: vector similarity search for Postgres
- Source
- pgvector: vector similarity search for Postgres
- Source role
- Authoritative source
- Exact locator
- README: Storing; Querying; Indexing — exact and approximate nearest neighbor search
- Supported claim
- pgvector documents exact search and approximate indexes that trade recall for speed.
- Last verification
- Review due
- Scope limitation
- This describes pgvector’s search options; no universal ranking of index methods is implied.