Definition
A learned numeric vector that represents an item so that useful semantic or structural relationships can be measured geometrically.
Distinguish it from nearby terms
An embedding is the representation; a vector database stores and searches vectors; semantic search is an application of them.
Check your understanding
Nearby vectors indicate similarity under the embedding model, not factual equivalence or truth.