In Japan, past disaster response cases and disaster countermeasure plans are not fully utilized in disaster response. In this study, we propose a disaster response knowledge-sharing system for local governments that utilizes Retrieval-Augmented Generation (RAG) technology to improve disaster response. In this system, local disaster prevention plans and past disaster response records created by local governments are databased, and necessary information is extracted and acquired through vector search using ChromaDB. The acquired information is natural language generated (NLG) using a large-scale language model (LLM), and necessary information is instantly searched and presented when a disaster occurs. This system consists of two types of chatbots: one dedicated to each uploaded disaster response plan and response record, and one that integrates all data. The dedicated chatbots propose countermeasures based on individual plans, and the integrated chatbot proposes comprehensive countermeasures across multiple data.