Retrieval-Augmented Generation

RAG

Kind: pattern

Layer: Correctness Core

Aliases: RAG

Record: architecture:retrieval-augmented-generation

Severity: contextual

Scope: language-model system, knowledge retrieval

Canonical: Ontology

A design pattern that retrieves relevant documents for a question and passes them to the model, so the answer can cite its sources.

Listed in Architecture principles, after Model Inference and before Vector Search.

Requires

• Retriever
• Document Store
• Grounding Strategy

Reinforces

• Explainability
• Knowledge Freshness

Enables

• Contextual Generation

Conflicts with

• Ungrounded Generation

In tension with

• Retrieval Quality/Latency

Tensions

• Retrieval-Augmented Generation (RAG) / Retrieval Quality/Latency

Violated by

• Ungrounded Generation

Refactored by

• Retriever
• Evidence Citation

Severity

• contextual

Category

• Model Architecture

Refactors

• Ungrounded Content

Reinforced by

• Vector Search

Linked from

• Systems built around a model
• Anti-patterns
• Model Architecture
• Model Architecture
• Severity levels
• The resolutions