Reactive Agents with Drasi & Dapr

Agentic AI is commonly used to predict and resolve problems before they occur, requiring access to data in near real-time. However, traditional patterns for reacting to data changes come with major disadvantages, exacerbating the challenge of building resilient, near real-time distributed systems. These patterns include polling, which is simple to implement but has a fundamental tradeoff between minimizing load on data sources and minimizing data staleness. Change data capture (CDC) avoids the aforementioned issues, but requires teams to build and maintain custom infrastructure. Using Drasi’s ability to track derived entities (via Continuous Queries) and Dapr Agents’ out-of-the-box durability for agents and agentic workflows, this proposed integration would help developers build scalable, event-driven AI systems in products and internal platforms. To accomplish this, Drasi would need a new ”proxy” service that routes events to multiple publish-subscribe (pub/sub) topics. Dapr workflows and agents would consume from these topics via workflow decorator functions and agent middleware, supporting Drasi and providing the groundwork to integrate other backends.