Yeah, features like spaghetti detection use more traditional ML, computer vision, image analysis types of AI algorithms. There are libraries to perform these sorts of detection, like OpenCV. Cloud services are used in the recent AI boom due to LLMs, which are inference models using significantly larger-scale compute resources, such as data centers with lots of AI-purposed hardware cards. Mosta those are for NLP-related tasks (translation, text document reading and analysis/summarization, query response, coding), little-to-none of which is used in 3d printing.
The only reasons one would even want/need cloud is for any sort of remote starting, managing, viewing, verification of one’s print jobs.