Author(s): Wen Du

Linked Author(s):

Keywords: Model Context Protocol; Multi-target routing engine; Retrieval-Augmented Generation; Water conservancy large model; Xinjiang Uygur Autonomous Region; Yellow River Basin

Abstract: To address core operational requirements in the Yellow River Basin and Xinjiang Uygur Autonomous Region, such as water resource management, flood and drought disaster prevention, engineering safety operation, government affairs and office administration, a collaborative large model service platform has been established at both basin and provincial levels. The platform aims to enhance collaborative intelligence and advance the modernization of water resources governance. In response to the issues such as the high coupling degree between large models and tools, the inability to directly access real-time data, the complexity of multi-target semantic processing for multiple tasks, and the bottleneck of water conservancy knowledge, a unified resource service large model was developed based on MCP. Through the multi-target routing engine technology, the large model could automatically decompose and execute tasks when performing multi-target complex semantic tasks. The knowledge base of the service was constructed by applying RAG technology. In future, the capacity to support water conservancy operations for target basins and provinces will be strengthened through the development of a Graph-RAG engine and the exploration of shared invocation mechanisms for large models.

DOI:

Year: 2026