What is it?
A vector database is a data system designed to store, index, and search vectors.
A typical stored item includes a vector, a reference to the original content, and metadata such as a document ID or source.
08 · Embeddings, Search & RAG · Topic 08
What is a Vector Database?
A vector database stores vector representations and provides ways to search them efficiently.
Think of it like...
Think of it like a library where books are arranged by mathematical similarity rather than only alphabetically.
A new question can be matched with nearby books whose meaning is useful, even when the words are different.
How it works
1 · ContentStart with source materialDocuments, product records, images, or other content enter the system.
2 · Embedding modelConvert meaning to numbersAn embedding model represents each item as a vector.
3 · VectorsKeep useful representationsThe vector is stored with its original-content reference and metadata.
4 · Vector databaseIndex the vectorsAn index helps the system search many vectors efficiently.
5 · Query vectorRepresent the questionA user query is converted into a vector in the same space.
6 · Similarity searchFind matching recordsThe system returns nearby vectors and their linked content.
Content ↓ Embedding Model ↓ Vectors ↓ Vector Database ↓ Query Vector ↓ Similarity Search ↓ Matching Records
Real-world examples
Everyday example · Photo search
Input: “Find my beach photos”
System: Compares the query vector with stored photo vectors
Output: Similar photos, even if filenames do not contain “beach”
AI product example · Knowledge assistant
Input: An employee asks about leave policy
System: Finds nearby policy chunks and returns their source references
Output: Relevant context for an answer
What it is NOT
Vector Database ≠ Embedding
Vector database: Stores and indexes vectors.
Embedding: A numerical representation created from content.
Vector Database ≠ Vector Search
Vector database: The data system and its indexes.
Vector search: The search method that compares vectors.
Vector Database ≠ RAG
Vector database: Can provide searchable records.
RAG: A broader pattern that retrieves context before generation.
Related concepts
Content → Embeddings → Vector Database → Vector Search → Retrieval → RAG
A vector database can support vector search and retrieval, which may become part of a RAG system.
Explore next
Remember this
A vector database stores and indexes vectors so applications can search them efficiently.
Independent explainer
Video
A video explainer for this topic is not available yet.