For a RAG pipeline, which Cosmos DB capability lets you keep the vector and its source document together to avoid a second lookup?
- A Storing the embedding as a property on the same document as the text (correct answer)
- B Using a separate vector-only container
- C Storing vectors in analytical store
- D Using the change feed
Shortcut: Co-locating the embedding array and the source text in one document means the nearest-neighbour query returns the content directly, removing a round trip and its RU cost.