Why your agents need decision traces, not just documents
summary
Argues that a context graph adds decision traces to RAG by combining short-term chat history, long-term extracted entities, and reasoning traces so past verdicts surface alongside semantic matches; it embeds graph structure and text into vectors for graph embeddings and standard vector search, uses a spaCy → GLiNER → LLM pipeline with deduplication in neo4j-agent-memory, and ships runnable scaffolding via uvx create-context-graph that seeds demo data, supports 22 built-in domains, and runs end-to-end.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:23Context graphs and why they matter
- 2:49Decision traces vs documents
- 4:19Demo: financial analyst agent
- 7:42Graph embeddings for retrieval
- 8:47One-line setup for context graphs
- 12:59Neo4j agent memory deep dive
- 15:06Resources and closing
- 16:13Q&A
related talks
29mHow to Structure Context to Make Your Agents Smarter
29mThe AI Agent Memory Landscape
56mConnecting the Dots with Context Graphs
23mConnecting the Dots With Context Graphs
31mBuild Intelligent AI Agents With Context Engineering
26mOpening Keynote: Exploring Context Graphs: From Data to Decisions
14mUsing Graphs to Fine-Tune Agents
28mData-Empowered AI Assistants