Defend is an open-source Python package that adds a security layer to any LLM-powered app. One line of code wraps your model with configurable input and output guards.