Session Spotlight

Sensors, AI, and Cartoon Roosters

ChatGPT is fantastic at generating and analyzing text, but it's not really "doing stuff". Utilizing the OpenAI API, Azure Hosted LLMs, and Function (Tools) calling, I will demonstrate real world practical examples of how AI and a couple of real-time data sources, can create intelligent agents capable of bringing in difficult to engage user bases (by using Discord as the front-end). I will walk through the (mostly) painless experience of learning how to integrate Large Language Models (LLMs) into traditional software architectures to provide an engaging and intelligent assistant for hobby farm work management and monitoring. Lessons learned (as in this didn't work) will be covered around conversation management, function design, and general oddities in getting LLMs to behave well enough to use written code. In addition, we will cover how to use wifi cameras and GPT-4o's image capabilities to do things like collecting information on temperature, animal location/identification, and gate status, and then passing that information to give the system an overview of the current environment to be able to make informed decisions and communicate those to the user base. There will also be plenty of pictures of chickens and goats!