Qdrant - Vector Database

Vector Database

A vector database is a type of database that emerged with the rise of deep learning and is now widely used in AI applications, notably to enable semantic search over large datasets (e.g. in RAG pipelines with LLMs).

In a relational database, it is complicated to handle unstructured data (such as videos, images, or audio) in a meaningful way. You would have to manually assign tags to describe and categorise them.

That is why we use vector databases ; they index and store vector embeddings to enable fast retrieval and similarity search. Unstructured data is processed by an embedding model, which converts it into a vector embedding, which is essentially a list of numbers (e.g. apple = [2.5, -3, ...]). Then, we use indexes, which are optimized data structures to quickly find similar vectors (otherwise the database must search through all vectors, which becomes slow).

Usecases of using vector database is for long-term memory for LLMs, similarity search for text, images, audio or video data for instance or recommendation engine (for online retailers, to suggest items).