Embedding
Also known as: embeddings, vector embedding, text embedding, semantic embedding, word vector
An embedding is what you get when you run text through an embedding model: the text goes in, a vector (a long list of floating-point numbers, often hundreds or thousands of numbers long) comes out. These numbers encode the semantic content of the text in a way that similar texts have vectors that are close to each other in mathematical space. 'The dog ran fast' and 'the puppy sprinted quickly' end up with similar vectors even though they share no words.
Embeddings are the foundation of semantic search and RAG systems. You embed all your documents, store those vectors in a vector database (a database optimized for similarity lookup), and then when a user asks a question, you embed the question too and find the stored vectors closest to it. That finds semantically similar content even without keyword matching.
For builders, embeddings are often an invisible layer that runs behind retrieval systems. You choose an embedding model (OpenAI, Cohere, and open models like Nomic all offer them), run your documents through it at ingestion time, and query the resulting vector database at search time. The quality of your embedding model affects the quality of retrieval, and different models work better or worse for different domains.