context-graphs

Neo4j's Will Lyon presents a recorded session on designing stateful AI agents with memory and context graphs, covering pitfalls and integrations with Salesforce Agentforce and Databricks.

This paper proposes Context Graphs, a live relational data structure for enterprise entities that enables proactive agents to surface relevant information before users query, formalizing components for delta detection, proactivity scoring, and LLM-powered surfacing.

Experiment shows AST-backed context graphs for coding agents can reduce token usage by 90% compared to broad snippets, while maintaining grounding, with a hybrid approach recommended to handle narrow retrieval cases.

A tweet asking if anyone has seen an elegant primitive for implementing stateful agents, decision traces, and context graphs.