가드레일

AI 시스템의 발언과 행동을 제한하는 안전 통제로, 유해 콘텐츠 필터링·도구 권한 제한·정책 집행을 포함합니다.

Essential for agents that act on real systems: cap loop counts, require approval for irreversible actions, and log every tool call. Managed platforms like Bedrock ship guardrails as a built-in feature.

관련 리소스

A managed AWS service offering foundation models from multiple providers behind one API, with agents, guardrails, and knowledge bases.

A low-level orchestration framework from the LangChain team for building stateful, multi-step agent workflows with fine-grained control.

Dify LLM App Development Platform

An open-source LLM application development platform offering visual orchestration, a RAG engine, and Agent capabilities for building AI apps quickly.