A fine-tuned Gemma 4 model for local, privacy-preserving AI safety coaching for minors. It classifies AI-chat prompts into coaching areas such as Privacy, Safety, Wellbeing, and Learning, then returns JSON for a local report without storing raw prompts.