Components 개념 MOC

Components 개념 MOC

AI Agent를 구성하는 재사용 가능한 컴포넌트를 역할별로 묶은 인덱스이다. 특정 프로젝트의 기능명보다, 다른 Agent에도 옮겨 쓸 수 있는 설계 원칙을 먼저 본다.

Agent Design

• Intent Classification
• Planning
• Capability Catalog
• Prompt-to-Action Planning
• Specialist Agent Pattern
• Reflection

Context

• Context Engineering
• Prompt Engineering
• System Prompt
• Data-aware Prompting
• Prompt Caching

Contracts and Safety

• Structured Output
• Block Contract
• Guardrails
• Contract Guardrail Pipeline
• Tool Execution Policy
• Human-in-the-loop

Runtime

• LLM Provider 추상화
• LLM Routing
• Async Resource Lifecycle
• Streaming
• SSE 기반 Agent Streaming

Memory

• Memory
• Session Context Store
• User Preference Locking
• Negative Feedback Memory

Interoperability

• MCP(Model Context Protocol)
• MCP Security
• A2A 프로토콜
• SKILL

Observability and Quality

• Observability
• Tool Observability Wrapping
• Trajectory
• Evaluation
• Grounded Generation
• LLM-as-Judge

Data and Multimodal

• Dataset Profiling
• OCR
• VLM
• SQLite 데이터 소스