Definition
A linear architectural pattern that augments Large Language Model responses by retrieving contextually relevant passages from an external vector database based on embedding similarity. While simple to implement, it relies on a single-pass retrieval step which can suffer from low precision if the initial query is ambiguous or the retrieval noise is high.
Often called 'Naive RAG' to distinguish it from iterative, agentic, or multi-stage retrieval workflows.
"An open-book exam where a student finds one specific chapter in a textbook to reference before writing their final answer."
- Vector Database(Prerequisite)
- Embeddings(Component)
- Top-k Retrieval(Component)
- Context Window(Prerequisite)
Conceptual Overview
A linear architectural pattern that augments Large Language Model responses by retrieving contextually relevant passages from an external vector database based on embedding similarity. While simple to implement, it relies on a single-pass retrieval step which can suffer from low precision if the initial query is ambiguous or the retrieval noise is high.
Disambiguation
Often called 'Naive RAG' to distinguish it from iterative, agentic, or multi-stage retrieval workflows.
Visual Analog
An open-book exam where a student finds one specific chapter in a textbook to reference before writing their final answer.