基于LangChain+Deepseek的智慧燃气GlS领域问答系统研究与实现

打开文本图片集

中图分类号：TP18；P209 文献标识码：A 文章编号：2096-4706（2025）22-0012-05

Research and Implementation of an Intelligent Q&A System for the Smart Gas GlS Domain Based on LangChain+DeepSeek

SHI Shanzhong （Shanghai City GISDevelopingCo.，Ltd.，Shanghai 20oo32，China）

Abstract： To address the common issue of low accuracy in vertical applications of Large Language Models （LLMs）， this paper adopts LangChain as the core development framework and integrates DeepSeek's advanced LLMtechnology. By comparing and implementing cuting-edge model deployment solutions such as Ollama and vLLM，the scheme ultimately adopts vLLMtoachieve high-performance inference in production environmentsand efectively resolves Ollama'sperformance botlenecks inconcurrent scenarios.Through a modular architecture design，this paper developsa comprehensive technical framewok thatiegatesatetaalsisgine，ntellgntmemorysem，ultgetollboratiopututputpoig chain，andcustomizabletoolintegration.Toaddressthecomplexbusinescenaiosofsartas GSsstems，thisfrorkis utilized to create a high-performance intelligent dialogue system with deep industry understanding.

Keywords：LangChain; DeepSeek; Ollama; dialogue system

0 引言

目前，虽然人工智能与GIS的结合是学界和业界的热点，但将大语言模型技术深度融合于智慧燃气这类专业GIS平台，构建能够处理复杂业务逻辑的智能问答系统的研究与实践尚处于探索阶段。（剩余6789字）