Application of modular Retrieval-Augmented Generation system to support agricultural decision-making

Vladislav Viktorovich Skripnik; Irina Aleksandrovna Veselkova; Valentina Yurevna Kuznetsova

Vladislav Viktorovich Skripnik FSBEI HE «ASU named after V.N. Tatishchev»
Irina Aleksandrovna Veselkova ФГБОУ ВО "АГУ им. В.Н. Татищева"
Valentina Yurevna Kuznetsova FSBEI HE «ASТU»

Keywords: Retrieval-Augmented Generation (RAG), Modular architecture, Hybrid search, Vector embeddings, Intellectual agrarian assistant

Abstract

The paper examines the application of the Retrieval-Augmented Generation (RAG) architecture for intelligent decision support in agriculture. A modular RAG system is proposed that takes into account the complex structure of agricultural knowledge and provides higher relevance of the extracted information compared to naive approaches. The system architecture, methods for improving search quality, as well as practical scenarios are described.