Connecting the Dots With Context Graphs
summary
Argues that persistent, queryable context beats one-off LLM calls and demonstrates a working pipeline: Neo4j + OpenAI embeddings + LangChain for a GraphRAG-style retrieval that stores short/long/reasoning memory in an open source package called Neo4j agent memory. Shows a podcast demo that indexes episodes with a PAULO entity model and a financial demo (Next.js UI, 10 MCP connectors) that recommends rejecting a $25k credit increase because it finds four identical rejections, unresolved fraud alerts and 14 risky transactions.
AI-written summary · about
chapters (12)
- 0:00Context graphs and LLMs
- 1:45Blue pill versus red pill in data
- 2:47Context graphs on Gartner’s hype cycle
- 3:28Knowledge graph basics
- 5:35Graph RAG vs vector RAG
- 7:20Three AI memory types
- 10:28Graph databases for context graphs
- 11:55Neo4j agent memory toolkit
- 13:18Demo 1: Podcast knowledge repo
- 15:43Demo 2: Credit approval system
- 19:55Bonus demo and use cases
- 21:07Resources and closing
related talks
56mConnecting the Dots with Context Graphs
18mConnecting the Dots with Context Graphs
29mThe AI Agent Memory Landscape
31mBuild Intelligent AI Agents With Context Engineering
49mConnecting the Dots with Context Graphs
27mGraphReader Agentic RAG: Rethinking Long-Context Retrieval Systems
48mOpening Keynote: Exploring Context Graphs: From Data to Decisions - Extended Version
01h 26mThinking in Graphs: The Cognitive Foundations of Agentic AI