Інтеграція Retrieval Augmented Generation у програмну систему SoilScout для генерації персоналізованих агрономічних рекомендацій

Abstract

This study explores enhancing the SoilScout IoT monitoring system by integrating artificial intelligence via the Retrieval-Augmented Generation approach. By linking large language models with 12 specialized datasets from Kaggle and Mendeley Data, the system provides personalized recommendations for crop selection and fertilization. The proposed architecture bridges the gap between real-time sensor data (pH, moisture, temperature) and expert agronomic knowledge. Implementing this AI module enables farmers to receive context aware advice, significantly improving decision-making efficiency and crop yields. Future work focuses on predictive analytics and time-series data mining to prevent soil depletion and forestall critical changes in the soil.