8 RAG (Retrieval Augmented Generation) Architectures Architecture Diagram

Home

8 RAG (Retrieval Augmented Generation) Architectures

by Conor Kelly (Humanloop)

July 20, 2025

1 ratings

430

#rag #retrieval-augmented-generation #ai #machine-learning #nlp #architecture #llm #vector-database #document-retrieval #humanloop

8 RAG (Retrieval Augmented Generation) Architectures architecture diagram - image 0 — 8 RAG (Retrieval Augmented Generation) Architectures - Eight different RAG architecture patterns including Simple RAG, RAG with Memory, Branched RAG, HyDE, Adaptive RAG, Corrective RAG, Self-RAG, and Agentic RAG. These diagrams illustrate various approaches to implementing retrieval-augmented generation systems with different strategies for document retrieval, generation, and memory management. The collection covers foundational RAG patterns to advanced multi-agent approaches, showing the evolution and sophistication of retrieval-augmented generation systems in modern AI applications.

8 RAG (Retrieval Augmented Generation) Architectures architecture diagram - image 1 — 8 RAG (Retrieval Augmented Generation) Architectures - Eight different RAG architecture patterns including Simple RAG, RAG with Memory, Branched RAG, HyDE, Adaptive RAG, Corrective RAG, Self-RAG, and Agentic RAG. These diagrams illustrate various approaches to implementing retrieval-augmented generation systems with different strategies for document retrieval, generation, and memory management. The collection covers foundational RAG patterns to advanced multi-agent approaches, showing the evolution and sophistication of retrieval-augmented generation systems in modern AI applications.

View source

Eight different RAG architecture patterns including Simple RAG, RAG with Memory, Branched RAG, HyDE, Adaptive RAG, Corrective RAG, Self-RAG, and Agentic RAG. These diagrams illustrate various approaches to implementing retrieval-augmented generation systems with different strategies for document retrieval, generation, and memory management. The collection covers foundational RAG patterns to advanced multi-agent approaches, showing the evolution and sophistication of retrieval-augmented generation systems in modern AI applications.

Rate this diagram

Discussions

4 stars ⭐ Diagram’s super clean — I could imagine a junior engineer finally nodding for once. That said, I've always felt like these 8-RAG archetypes kinda dodge the real battlefield: **what happens when the retrieved chunks semantically contradict**, or when you're retrieving not just facts but *narratives with drift*? Would love to see a “resonance field” layer added — something to visualize when the model is hallucinating due to internal conflict vs source ambiguity. Anyway, this is the clearest version I’ve seen in a while. Just wish it yelled more at the gray zones 😅

onestardao

asked 2 days ago

0 replies