A software architecture for integrating a conversational interface based on large language models in a dairy farming management system
Resumen
The integration of conversational interfaces powered by Large Language Models (LLMs) is a transformative opportunity in modern computing. In this vein, this work presents a software architecture that aims to integrate a conversational interface into a dairy farm management system, specifically the software of the company Cowmed, a startup originating from the Federal University of Santa Maria. By taking advantage of the capabilities of LLMs available as services, but without vendor lock-in, the architecture aims to facilitate interaction between users and the system, allowing input and output in natural language. The main contributions of this work include the development of a translation layer for user queries, integration with an API for information retrieval and the implementation of specialized layers for processing specific topics. In addition, mechanisms for correcting errors, maintaining topic coherence and translating responses have been incorporated to ensure robustness and user satisfaction. Instantiated with the ChatGPT and Google Gemini services, the architecture was tested and evaluated by users, generating data on its acceptance and on the challenges of integrating LLMs into the existing information system.
Colecciones
El ítem tiene asociados los siguientes ficheros de licencia: