Neo4j research engineer Zach Blumenfeld describes context graphs as a way to connect entities and events with past decisions, precedents and reasoning. He contrasts this with retrieving documents to answer questions.
Zach Blumenfeld walks through a demonstration combining semantic and graph-based retrieval of decision traces, then shows a tool for scaffolding a context graph application with example data. Parts of the recorded demonstration have caption gaps.
The talk discusses short-term memory, long-term entities and reasoning tracesAI agent memory is stored information that an agent can retrieve and use across steps, sessions, or changing contexts., as well as entity extraction and deduplication. Audience questions cover timestamps, schema preparation and the still-developing ability to write new traces.
Watch on YouTube




