As applications move from development to production, ensuring their safety and predictable behavior becomes a primary objective. LLM-powered systems introduce specific challenges due to the variability of their outputs.

In this chapter, you will learn how to implement practical safeguards for your applications. We will cover how to use the safety module to build guardrails, perform content moderation, and detect and mask personally identifiable information (PII) to maintain user privacy. We will then address application reliability. You will learn strategies for testing systems that are inherently non-deterministic, including how to use the testing module to mock LLM calls for creating fast and repeatable unit tests.