Conceptual
Browse Instruct Curate

LLM Multi-Agent Architecture for Complex Event Processing on Multimedia Streams

An architecture that embeds a Large Language Model based multi-agent system (built with the Autogen orchestration framework) into a Kafka publish/subscribe pipeline so that complex event processing over multimedia video streams becomes dynamic and adaptable. The design lets LLM agents interpret and route diverse, changing data without hand-coded domain rules, and characterizes how agent count and video complexity trade off against end-to-end latency while preserving narrative coherence.