Get reliable JSON or typed output from LLMs. Covers JSON schema, validation, retry-on-malformed, function-calling, and graceful degradation. Use whenever an LLM call needs to produce data your code consumes (not text for humans).