Multi-Agent Orchestration for Quality Control Optimization in Industrial Information Systems

Abstract

This paper presents a multi-agent orchestration architecture for automatic query optimization in PostgreSQL-based quality control systems. The system is implemented using the LangGraph framework and includes an orchestration graph with parsing, rewriting, indexing, prediction, and execution nodes, connected by conditional edges and retry loops. Agents execute in parallel up to four instances. Fault tolerance mechanisms include heartbeat notifications, node timeouts, circuit breaker, and automatic failover to backup models. Experimental evaluation on a dataset of ten thousand parts and five thousand orders demonstrates average query execution speedup of four point seven five times compared to baseline configuration. Orchestration overhead is zero point zero two eight milliseconds per query. Administrator intervention is reduced from eight hours to thirty minutes per week.