Glossary

All concepts that appear in the help documentation, arranged in alphabetical order. Click a "See" link to jump to the corresponding document.

A

Term	One-Sentence Description	See
Agent	The umbrella term for AI applications built on the platform, divided into Chatflow (conversational) and Workflow (batch-processing) types	Platform Overview
Agent Marketplace	Where you browse, use, and fork Agents published by others; marketplace content is for learning and exchange only and does not constitute a recommendation	Region-Specific Service Notes
Agent node	An autonomous execution node on the canvas: the LLM loops through tool calls on its own until the task is complete (Function Calling mode)	Agent Node
Agentic Chat	Lightweight creation mode: configure a prompt and attach tools to run, no canvas orchestration needed — a lightweight version of Chatflow	Tip 1
Answer node	The node in a Chatflow that outputs the reply to the user	Answer Node

B

Term	One-Sentence Description	See
Benefit Card	A time-limited higher-tier experience card: it does not change your subscription, and during its validity it upgrades feature entitlements and stacks extra usage	HKSG Benefit Card
Branch Aggregator	A node that merges same-type variables from multiple branches into a single variable (e.g. array[string])	Branch Aggregator Node

C

Term	One-Sentence Description	See
Charged By	The attribution rule that determines which Workspace a run's usage is billed to	Who Pays for Usage
Chatflow	Canvas-orchestrated multi-turn conversational Agents (e.g. intelligent customer service)	Platform Overview
Code node (Code Execution)	A node that runs Python3 / JavaScript code in a sandbox; the code must implement a main function and return a dictionary	Code Node
Context compression	Automatic compression of history in long conversations to avoid exceeding the context limit	Context and Memory Management

D

Term	One-Sentence Description	See
DSL	The import/export file (YAML) for application orchestration; model availability is validated automatically on import	Debugging and Observability

E

Term	One-Sentence Description	See
Editor	A Workspace member role: can create and edit resources within the space; the number is limited by plan	Member Roles
Embedding	Vectorization of knowledge base documents; one of the billing units	What Consumes Usage
End node	The end node of a Workflow, defining the final output	End Node
Error handling / retry on failure	Node-level fault-tolerance configuration: automatic retry on failure, handling exceptions by policy	Tool Node

F

Term	One-Sentence Description	See
Function Calling	The Agent node's tool-calling mode: the model uses tools via function calls, looping until the task is complete	Agent Node

G

Term	One-Sentence Description	See
Guardrail	Two-way compliance interception on input/output: PII masking, blocking of harmful content and financial violations	Guardrail

H

Term	One-Sentence Description	See
Http Request node	A node that calls external HTTP APIs to fetch data; supports importing cURL	Http Request Node
Human in the Loop (HITL)	A mode that proactively pauses at key nodes during execution and waits for human confirmation	Planning and Execution

I

Term	One-Sentence Description	See
IF Else node (Conditional Branch)	A routing node with IF / ELIF / ELSE multi-condition branching	IF Else Node
item variable	The variable that nodes inside an iteration use to reference "the element being processed in the current round"	Iteration Node
Iteration node	A container node that processes a list item by item, with parallel mode and error handling support	Iteration Node

J

Term	One-Sentence Description	See
Jinja2	The template language used by the Variables Transformer node (variables, filters, control structures)	Variables Transformer Node

K

Term	One-Sentence Description	See
Knowledge Base	Upload private documents for the Agent to retrieve and answer from; capacity depends on plan	Knowledge Base

L

Term	One-Sentence Description	See
Langfuse	The third-party tracing platform supported by Trace App Performance, used for debugging and performance analysis	Debugging and Observability
LLM node	The core node that calls a large language model (prompt, model, memory, structured output)	LLM Node
Logs (Run Logs)	View each run's inputs and outputs, execution path, and failure reasons	Debugging and Observability
LongbridgeAI Chatbot (chat client)	The official conversational product, which shares the same usage quota as the Agent platform	Chat Experience
Long-term / Short-term memory	Cross-session / in-session memory capabilities, configured via switches in nodes such as LLM	Context and Memory Management
Loop node	A container node that repeatedly executes the same logic until an exit condition or maximum count is reached	Loop Node

M

Term	One-Sentence Description	See
MCP (third-party MCP Server)	An external tool integration protocol, available on the Premium plan	Plan Benefits Quick Reference
Memory window	The number of context turns retained in a multi-turn conversation; setting the smallest sufficient value helps control usage	Context and Memory Management
Model tiers (Mainstream / SOTA)	Models are divided into two consumption tiers, with SOTA consuming more per unit; node-level model configuration is supported	Model Tiers and Usage

O

Term	One-Sentence Description	See
OpenAPI tools	External API tools configured by users themselves; calls are not billed	Tool Node
Owner	The single highest-privilege member of a Workspace, the only one who can perform commercial operations such as subscription purchases/upgrades	Member Roles

P

Term	One-Sentence Description	See
Parameter Extractor	A node that uses an LLM to extract structured parameters from natural language, commonly used to feed tool inputs	Parameter Extractor Node
Plan	The commercial object that determines a Workspace's feature scope and usage quota	Plan
Pro / Premium	Paid plan tiers (auto-renewing monthly subscription)	Plan
Prompt	The instruction text that directs model behavior, split into System / User layers	Prompt Writing Tips

Q

Term	One-Sentence Description	See
Question Classifier	A routing node that classifies user questions by intent and routes them to different branches accordingly	Question Classifier Node

R

Term	One-Sentence Description	See
RAG (Retrieval-Augmented Generation)	A Q&A paradigm that first retrieves from a knowledge base / search, then answers based on the retrieved results	RAG Retrieval Patterns
Rerank	Reordering of retrieval results; one of the billing units	What Consumes Usage

S

Term	One-Sentence Description	See
Security review	The platform's review of an Agent's name, avatar, description, and Prompt before publishing / marketplace listing	Security Review
Skill	An on-demand capability/methodology package that can be attached to an Agent; the maximum number depends on plan	Skill System
Start node	The entry node of a flow, defining the input fields	Start Node
Starter	The free plan, activated automatically upon registration	Plan
Statistics	The Agent analytics dashboard: trends in conversation count, active users, latency, satisfaction, etc.	Debugging and Observability
Structured output	The ability to make an LLM output reliably in a defined structure (e.g. JSON)	LLM Node
System Prompt	The stable part of a prompt: role definition, behavioral rules, output format, compliance red lines	Prompt Writing Tips

T

Term	One-Sentence Description	See
Test run (trial run)	Run a single node or the entire flow to validate results without publishing the application	Debugging and Observability
Tool node	A node that calls the platform's built-in tools or OpenAPI tools to fetch data or perform actions	Tool Node
Trace App Performance	Sends the application's execution context to a third-party tracing platform (Langfuse) for debugging and analysis	Debugging and Observability

U

Term	One-Sentence Description	See
Usage	The metering system for AI consumption: LLM calls, Embedding, sandbox runs, etc. are all converted into USD	Usage

V

Term	One-Sentence Description	See
Variable	The carrier for passing data between nodes, including input and output variables	Variables and Data Flow
Variables Transformer	A node that formats/transforms variables using Jinja2 templates	Variables Transformer Node
Viewer	A Workspace member role: read-only viewing and use of resources; the number is limited by plan	Member Roles

W

Term	One-Sentence Description	See
Weekly rolling release	The monthly quota is released on a rolling 7-day basis; weekly availability ≈ monthly total ÷ 4	How Quota Is Released
Workflow	Automated batch-processing Agents (one input, one output — e.g. research report generation)	Platform Overview
Workflow triggers	Scheduled / Webhook triggers that run Workflows automatically, unattended	Workflow Triggers
Workspace	The basic unit that hosts Agents, knowledge bases, and usage — the boundary within which a plan takes effect	Workspace and Plans