The article discusses the architecture and operating principles of EDRO Expert software for advanced vector database search and work with large language models capable of generating information based on additional knowledge of the system. The RAG (Retrieval Augmented Generation) method is a new method of expanding the knowledge of large language models without retraining the model's weight coefficients, which allows you to add new knowledge to the model in a short period of time. This approach has both advantages and disadvantages in the form of finding suitable data in a vector database at the user's request, an approach to storing new knowledge, as well as methods for splitting documents into chunks. The “EDRO Expert” software product partially solves a number of problems, which allows users without deep knowledge in the field of AI design to achieve good results in generating responses to user requests. The software package is open source and is well suited for scientific research.