Agentic GraphRAG with MAGMA Manifold Agency with Graph Models and Augmentation
summary
Claims that treating components as a magma — composable units — makes agents and graphs simple building blocks, and that composing agents generates emergent intelligent behavior rather than needing monolithic AGI. Demonstrates Graph RAG: vector search plus graph expansion using Cypher and Neo4j, shows 10 chunks -> ~100 pairwise combinations and orders-of-magnitude more answerable queries at scale, cites Microsoft’s precompute-and-summarize pattern, links to graphrag.com and describes agents-with-memory and graph-as-memory prototypes that reportedly work.
AI-written summary · about
chapters (7)
- 0:00MAGMA introduction
- 4:59Composition in software
- 10:55Graphs and databases
- 14:23GraphRAG for answers
- 19:18GraphRAG pattern basics
- 21:05Agents with graphs
- 22:49GraphRAG resources
related talks
24mAgentic GraphRAG: AI’s Logical Edge
36mAgentic GraphRAG: AI’s Logical Edge
30mAgentic RAG with Knowledge Graphs Explained
27mWhen Vectors Aren’t Enough: Unlocking AI’s Hidden Connections with Graphs
01h 26mThinking in Graphs: The Cognitive Foundations of Agentic AI
53mPractical GraphRAG - Making LLMs smarter with Knowledge Graphs
12mAgentic GraphRAG: Autonomous Knowledge Graph Construction and Adaptive Retrieval
26mPractical GraphRAG Making LLMs smarter with Knowledge Graphs