Question

A financial services company is building a RAG system so their assistant can answer questions using thousands of internal compliance documents. The team has already selected a foundation model and stored document embeddings in a vector database. During a design review, a stakeholder asks what specifically happens at query time to make the model's answer grounded in the correct documents. Which description best explains the retrieval step in a RAG pipeline?

Accepted Answer

The user's query is converted into an embedding and used to search the vector database for the most semantically relevant document chunks, which are then added to the prompt as context.

Answer

The foundation model's weights are updated with the compliance documents so the answer reflects the latest information.

Answer

The entire set of compliance documents is inserted into every prompt so the model always has full access to all information.

Answer

The model generates an answer first, then the vector database checks the answer against documents and rewrites any incorrect claims.