Llm-Based Farmer Advisory System Using Retrieval and Tool-Augmented Reasoning

Abstract

Timely, accurate, and context-aware advisory support is essential for improving agricultural productivity and decision-making. Traditional agricultural advisory systems often provide generalized recommendations and lack real-time adaptability to individual farmer requirements. To address these limitations, this paper presents an intelligent LLM-based Farmer Advisory System that integrates conversational artificial intelligence with Retrieval-Augmented Generation (RAG), tool-based reasoning, and multimodal analysis. The proposed system utilizes a LangGraph-based orchestration framework in which the Large Language Model dynamically determines whether to retrieve information from a FAISS-based knowledge repository, invoke external tools such as weather APIs, or generate responses directly based on user queries. The system further incorporates farmer-specific contextual information, including crop type, geographical location, and soil conditions, to provide personalized and relevant recommendations. In addition, a vision-based crop disease detection module is integrated to analyze crop images and assist farmers in identifying possible diseases. The system is implemented as a web-based application using Flask, enabling seamless interaction through a user-friendly interface. The proposed hybrid framework improves the reliability, adaptability, and usability of agricultural advisory services compared to conventional approaches. Experimental observations indicate that combining LLM reasoning with retrieval mechanisms and external tools significantly enhances the effectiveness of AI-driven agricultural assistance systems in real-world farming environments.