Question

A company is building a customer-support assistant on Amazon Bedrock using a Retrieval Augmented Generation (RAG) pattern. A developer wants to understand what happens after the system retrieves the most relevant documents from the vector store. In a standard RAG workflow, how are these retrieved documents used to produce the final answer?

Accepted Answer

The retrieved documents are added to the prompt as context and sent to the foundation model along with the user's question so the model generates a grounded response.

Answer

The retrieved documents are used to retrain the foundation model's weights before every response is generated.

Answer

The retrieved documents replace the foundation model entirely, and the answer is returned directly from the vector store without any generation step.

Answer

The retrieved documents are converted into embeddings a second time and returned to the user as the final answer.