Embedding

A fixed-length list of numbers (a vector) that encodes the meaning of a piece of text so that semantically similar texts end up with numerically similar vectors.

📌 Example: 'dog' and 'puppy' will have embeddings that are closer together than 'dog' and 'invoice'.

🛠 In practice: Embeddings are produced by a separate model (e.g. text-embedding-3-small). Choosing the right embedding model and chunk size affects retrieval quality significantly.

• depends on Vector DatabaseA vector database stores and searches embeddings — you need embeddings to populate and query one.