RAG vs. GraphRAG Architecture Comparison¶
Scaffold
This MicroSim has been scaffolded from its specification. The interactive implementation has not been built yet.
Learning Objective¶
TBD
- Bloom Level: TBD
- Bloom Verb: TBD
- Library: vis-network
Preview¶
Specification¶
The full specification below is extracted from Chapter 17: AI and Machine Learning System Architecture.
Type: Interactive architecture comparison
**sim-id:** rag-architecture-explorer<br/>
**Library:** vis-network<br/>
**Status:** Specified
**Purpose:** Side-by-side animated comparison of flat RAG and GraphRAG architectures showing the complete request flow from user query through retrieval to LLM generation.
**Left panel (RAG):** User query → Embedding model → Vector DB (similarity search) → Top-K documents → LLM prompt assembly → LLM → Response
**Right panel (GraphRAG):** User query → Entity extraction → Knowledge Graph (graph traversal) → Related entities + Vector DB (semantic search) → Combined context → LLM prompt → LLM → Response
**Interactions:**
- Click each component to see: purpose, latency contribution, quality risks
- "Show Latency Budget" mode: animate request flow with per-step timing
- Toggle "Complex reasoning query" vs. "Simple factual query" to see when GraphRAG advantage is most pronounced