Glossary term

Embedding

A numeric representation of text or other content used for search, clustering, and retrieval workflows.

An embedding is a vector representation that lets a retrieval system compare pieces of content by meaning rather than only by exact words. Category: Models Related topics: 0 Related resources: 0

Why this term matters

Private RAG and document search depend on embeddings before the chat model ever writes an answer. A weak embedding choice can make a good local model look unreliable.

Examples

A notes workflow may embed each chunk of a document, then retrieve the closest chunks before the model answers.
Changing the embedding model can change which passages are found, even when the chat model stays the same.
Embeddings are a separate model decision from the LLM used for final generation.