AI-enabled agriculture finds itself at a critical crossroads where smallholder farmers still do not have convenient access to relevant personalized information for crop, pest, and farm management. Although agricultural extension services exist, they have left much to be desired, particularly in their customization to the needs of smallholder farmers. Subsequently, farmers remain trapped by worsening climate conditions and low productivity. This research enhances the capability of existing solutions around AI-enabled agricultural assistance developed in previous stages by creating an AI-based agricultural chatbot powered with large language models and transformer technology. Starting with previous AI-powered agricultural advisory systems that utilized classical machine learning algorithms, rule-based systems, and primitive NLP techniques such as TF-IDF, Word2Vec, and LSTM, this research improves existing solutions by incorporating AI agricultural chatbot built on advanced LLMs and transformer models. Unlike previous research that examined a multitude of NLP approaches and techniques, we build upon such efforts by constructing and training agricultural specific domain models along with fine-tuning them. In the process, we evaluated several models and chose T5 and DistilBERT because their speed, accuracy, contextual understanding, and response accuracy was unmatched. This chatbot is designed to assist farmers with tailored, adaptive guidance while at the same time bridging the agricultural information gap. This innovation integrates different datasets such as climate data, soil health, and pest control; it ensures accurate and context-aware recommendations in real time. This approach enhances the accessibility and efficiency of agricultural assistance, helping smallholder farmers make informed decisions and better manage their farming practices.