Guardrails Architecture

A production LLM deployment has three filtering stages: 1. Input ______ : validate/sanitize user input 2. The _____ : the actual language model 3. Output ______ : check/filter the response Describe this three-stage architecture. Flag format: CONGRESS{stage1_stage2_stage3_pipeline}