An embedding is a vector, a long list of numbers, produced by a model from some text. Texts with similar meaning get vectors that sit near each other, so "reset my password" and "I can't log in" land close together even though they share few words.
That closeness can be measured, which makes embeddings the basis of semantic search. You embed your content once, store the vectors in a vector database, embed each incoming query the same way and return the nearest matches. This is the retrieval step in Retrieval-Augmented generation.
Two practical points. Use the same embedding model for content and queries, since vectors from different models are not comparable. And embed sensible pieces: the size of each document chunking unit affects how well matches work.
Related: Knowledge Base, AI Agent Memory.