arXiv cs.AIThursday · May 28, 2026FREE

Discovery Agents for Real-Time Analytics: Toward Proactive Insight Systems

A new paper on arXiv (2605.27571) introduces a multi-agent architecture for proactive insight discovery in real-time streaming environments. The system uses a continuous discovery loop where agents generate hypotheses, compile them into executable analytics, validate artifacts, and produce visualizations and deployable applications. It leverages Apache Kafka for event-driven coordination, Apache Flink for stream processing, and large language models (LLMs) to implement specialized agents. A key design feature is contract-driven development using typed intermediate artifacts, ensuring modularity, observability, lineage, and safer execution of dynamically generated analytics. The architecture is evaluated through use cases in retail, finance, and public data, demonstrating a shift from query-driven to discovery-driven analytics. The paper is published on May 28, 2026, and is available on arXiv under cs.AI.