Glossary

This glossary defines core terminology used throughout the manual. Feature-specific terms are revisited in their respective chapters.

Solution Basics

Term	Korean	Definition
Xgen	Xgen	Product name — an Agentic AI Platform that scales an enterprise's AI execution capability through natural-language task automation and integration with diverse systems
Agent Workspace	Agent 작업실	User workspace for creating and operating agents and agentflows
Admin Center	관리 설정	Administrative area accessible to users with SuperUser permission — or to non-SuperUsers granted specific menu-level permissions on the Admin Center pages
Dashboard	대시보드	Landing screen showing status and statistics at a glance

Core Objects

Term	Korean	Definition
Agent	에이전트	An AI work unit defined to automate a specific task
Agentflow	에이전트플로우	A defined AI workflow connecting multiple nodes — the unit of operational deployment
Node	노드	A single work unit within an agentflow (LLM call, tool execution, branching, etc.)
Canvas	캔버스	The area for drag-and-drop and visual editing of an agentflow
Tool	도구	An external function or API the agent can invoke
MCP	Model Context Protocol	The standard protocol connecting AI models with external tools
Prompt	프롬프트	An instruction passed to an AI model (System / User prompts)
Authentication Profile	인증 프로필	A bundle of credentials (keys/tokens) used for external API integrations

Agent vs Agentflow

• Agent: A single AI work unit performing one task
• Agentflow: A workflow weaving multiple agents and nodes — the deployment unit

Knowledge Management

Term	Korean	Definition
Collection	컬렉션	A grouping unit for documents in the knowledge store
Document	문서	The original asset (text, PDF, image) uploaded to a collection
Chunk	청크	A small text segment that a document is split into for embedding
Embedding	임베딩	The vector representation of a chunk that enables similarity search
Vector Database	벡터 데이터베이스	The DB that stores and searches embedding vectors (e.g., Qdrant)
Embedding Model	임베딩 모델	The model that converts text into vectors
Reranker	리랭커	The model that re-orders the initial vector search results
Ontology	온톨로지	Structured metadata defining concepts and relationships in documents

Execution and Chat

Term	Korean	Definition
Chat	채팅	A conversational session between a user and an agent
Message	메시지	A single utterance in a chat (user or agent)
Execution	실행	One run of an agentflow that produced results
Tool Call	도구 호출	An event where the agent invoked a tool or function
Citations	인용	The source documents or assets referenced by an AI response
Logs	로그	Step-by-step detailed records of an execution

Deployment and Operations

Term	Korean	Definition
Deploy	배포	Publishing an agentflow so users or external systems can invoke it
Embed	임베드	Code snippet that embeds an agentflow interface into an external page
Version	버전	The change-history unit for agentflows, prompts, and documents
Schedule	스케줄	Automated execution timing (daily, weekly, Cron, etc.)
Share	공유	Granting another user access to an item
Deployment Status	배포 상태	Lifecycle stage of an agentflow — Draft / Active / Inactive / Archived

Users and Permissions

Term	Korean	Definition
Standard User	일반 사용자	Base user permission tier — encompasses two Agent Workspace user types (Standard User / Agent Developer). Standard Users use Agent Chat / Technical Support / Dashboard, while Agent Developers receive additional menu permissions for Agent Creation / Tool Integration / Knowledge Management. A SuperUser may also grant additional individual admin-menu permissions per operational policy; the top-left Admin Center mode-switch becomes accessible when any such menu-level permission is granted
SuperUser	슈퍼유저	Administrative tier — accesses the top-left Admin Center mode and performs all administrative tasks (user/role management, system settings, AI governance)
Role	역할	A bundle of permissions (e.g., "System Administrator", "Governance Officer", "Operator", "Analyst")
Permission	권한	An access grant for a specific feature or resource

Tier (two levels) vs Role (many) vs Permission (ABAC keys)

The permission system is split into three independent layers:

• Tier: two levels — Standard User and SuperUser.
• Role: free-form labels defined by your organization — System Administrator, Governance Officer, Analyst, etc. Within the same SuperUser tier, roles control which sidebar menus you see.
• Permission: ABAC keys such as admin.user:read. They are bundled into roles or granted directly to a user, gating sections / tabs / buttons.
  • ABAC = Attribute-Based Access Control. A model that decides access from the user's attributes (roles, permissions, etc.); here those attributes take the form of resource:action strings.

The full picture and the order in which gates evaluate is documented in Role / Permission Management · Permission Model.

Governance and Security

Term	Korean	Definition
PII	Personally Identifiable Information	Information identifying an individual (national ID, phone, email, etc.)
Masking	마스킹	Hiding PII with characters such as asterisks (*)
Risk Level	위험 등급	Risk classification of an AI response or request
Audit Log	감사 로그	Retained records of user activity and system events

Notation Conventions

This manual follows the rules below.

• Korean primary, English in parentheses: First mention uses the form 에이전트플로우(Agentflow). Subsequent mentions in the same chapter use Korean only.
• Unified terms: Where the source code mixes terms (e.g., Canvas / 캔버스), this manual standardizes on the Korean form.
• Code and identifiers: Use monospace — e.g., customer.id, xgen.example.com.
• Language switch: This manual defaults to Korean. Use the language toggle (🌐) at the top right of the page to switch to English.