Need early AI integration advice for prototype

Seeking guidance - made a local python application that tracks user behavior- input/data is written into a local DB (sql lite). Goal is to have a an early prototype that uses AI to learn a user's baseline behavior and alert to deviations or abnormalities. Problem is, limited on both budget (free/open source preferred but not opposed to paying) and AI dev experience. Current main focus is completing a prototype, whether using local AI (like llama3) or cloud AI. Really looking for suggestions or recommendations on easiest and most efficient AI model and how to integrate for this early prototype. Any guidance or advice is greatly appreciated!!!