Analyze whether a given text contains toxic, insulting, or harmful language using a fine-tuned Gemma3 model.