Vector Store
A specialized database designed to store, index, and search high-dimensional vector embeddings for AI applications.
Also known as: Vector Database, Vector DB
Category: AI
Tags: ai, databases, embeddings, infrastructure, machine-learning, search
Explanation
A Vector Store (or Vector Database) is a specialized database designed to store, index, and search high-dimensional vector embeddings. Unlike traditional databases that store text and numbers in rows and columns, vector stores work with mathematical representations where similar items cluster together in multidimensional space.
When machine learning models process text, images, or audio, they convert these into dense numerical vectors called embeddings. Each dimension captures different features and attributes. Similar meanings result in nearby vectors, while different meanings produce distant vectors. Typical embeddings have 384 to 1536 dimensions.
Key operations in vector stores include:
- **Indexing**: Storing vectors with associated metadata
- **Similarity search**: Finding nearest neighbors using distance metrics like cosine similarity, euclidean distance, or dot product
- **Filtering**: Combining vector search with metadata filters
- **Hybrid search**: Mixing semantic and keyword search for better results
Popular vector databases include Pinecone (managed, easy RAG setup), Milvus (billion-scale, enterprise), Weaviate (model integrations), Chroma (simple, developer-friendly), Qdrant (open source, high performance), pgvector (PostgreSQL extension), and Redis (sub-millisecond latency).
Common use cases include: Retrieval-Augmented Generation (RAG) for LLMs, semantic search across documents, AI chatbots with contextual memory, image and audio similarity search, recommendation systems, and duplicate detection.
Vector stores are essential infrastructure for modern AI applications, enabling semantic understanding and intelligent retrieval at scale.
Related Concepts
← Back to all concepts