Vector database
A digital library that organizes information by what it actually means rather than just matching specific words, allowing AI to quickly find relevant facts to answer your questions.
A specialized storage system that indexes data as mathematical vectors, enabling fast similarity searches that allow AI models to retrieve contextually relevant information for RAG workflows.
A datastore that indexes embeddings and retrieves by semantic (nearest-neighbour) similarity, underpinning retrieval-augmented generation and semantic search.
evolution
- 2013 · historyWord2Vec Introduction
Google researchers introduced Word2Vec, enabling the representation of words as dense vectors in a continuous vector space.
- 2017 · historyTransformer Architecture
The 'Attention Is All You Need' paper established the foundation for high-dimensional embeddings that require specialized storage and retrieval.
- 2019 · historyFAISS Release
Meta (Facebook) open-sourced FAISS, a library for efficient similarity search and clustering of dense vectors, setting the standard for vector indexing.
- 2021 · historyMilvus and Pinecone Launch
Dedicated managed and open-source vector databases emerged to provide scalable, production-ready infrastructure for vector search.
- 2023 · historyRAG Mainstreaming
The explosion of LLMs made Retrieval-Augmented Generation (RAG) a standard pattern, cementing vector databases as essential AI infrastructure.