Since flora knowledge is primarily unstructured, traditional flora retrieval methods were low in efficiency and limited in functions. In order to enhance the flexibility, accuracy, and efficiency of flora knowledge retrieval, the large language models (LLMs), knowledge graphs, and retrieval-augmented generation (RAG) were used to investigate to construct a flora question-answering (QA) system by integrating relational databases, graph databases, and knowledge bases to address the above issues. LLMs were used for structured extraction of flora data to establish a relational database characterized by finer granularity and support for numerical attribute retrieval. The LLMs were used to construct the graph database suitable for diverse attribute storage and complex queries for the comprehension ability of entity relationships. Additionally, the vectorized knowledge base emphasizing semantic relevance and reasoning was built with the aid of semantic retrieval technologies. During the QA process, the mature language comprehension, programming, and text generation capabilities of LLMs were fully leveraged, and the corresponding issue results were retrieved from the aforementioned external databases and knowledge bases for data fusion to provide more accurate reference information in vertical field of flora for the LLMs and underpin the QA generation, thus constructing the flora QA system. The results show that such QA system can perform collaborative retrieval and information fusion from multi-source repositories to generate more accurate answers, is more flexible and efficient compared with the traditional flora retrieval method, and can effectively mitigate the “hallucination” problem of general-purpose LLMs. It is concluded that this study provides a new approach to achieve convenient and precise retrieval of flora knowledge, and its construction method could also offer a reference to the research and development of flora QA system in other regions, contributing to advancing the intelligent development of botanical knowledge services.