I built an offline AI that knows your last frost date, no internet, no API

Developer Sarvar Nadaf has introduced 'FrostWise,' an innovative, fully offline tool designed to predict the last spring frost date for gardeners. Unlike traditional weather services that rely on cloud APIs and internet connectivity, FrostWise runs entirely on a local machine. The project utilizes TabPFN-v2 for tabular forecasting and a local Gemma 3 model via Ollama to generate personalized planting advice. By processing climate data—such as latitude, elevation, and historical temperature trends—locally, the tool ensures user privacy and reliability in remote areas with poor signal. The project emphasizes the power of open-weight models, demonstrating that complex predictive tasks can be performed without external dependencies or subscription costs. FrostWise serves as a practical example of how local AI can solve real-world problems, providing gardeners with actionable, data-driven insights even when disconnected from the internet.