What is a vector embedding?

A vector embedding is a list of floating-point numbers that represents a piece of data — text, an image, audio — in a way that captures its meaning or content, positioned in a high-dimensional space such that similar items end up close together and dissimilar items end up far apart.

Embeddings are produced by machine learning models trained specifically for this purpose — a text embedding model like OpenAI's or a sentence-transformers model converts a sentence into, say, a 1536-dimension vector, while a model like CLIP can embed both images and text into a shared vector space so the two can be compared directly.

In LanceDB, embeddings are typically stored as a fixed-size vector column alongside the original data they were generated from, which is what makes semantic search possible: instead of matching exact keywords, a query is itself converted into a vector, and the database finds the stored vectors closest to it in that high-dimensional space.