How do you build a context graph?

What this post added

This post details Glean's approach to building context graphs, which are essential for AI agents to understand and automate work processes. It explains that context graphs model the 'how' of work by transforming actions into first-class entities with temporal traces and causal/correlational edges. The construction involves deep connectors and observability to capture change events, building a unified knowledge graph to infer entities and relationships, and creating a personal graph to understand individual tasks. Aggregating personal graphs and normalizing them into abstracted traces forms the context graph, which represents a probabilistic view of work flows. The post also highlights the hybrid storage model used for event data and the importance of closing the loop with agent execution, where agent traces are incorporated into the context graph for continuous learning and refinement. The context graph is presented as a joint model of human and agent behavior, owned by both the data and orchestration layers.