Glossary

The same word can mean different things in different products. This page fixes the meaning of each term as MAIA Plus uses it, with a link to the page that explains it in context.

Platform objects

Term	Meaning	Where
Agent	Persona plus policy: instructions, tools, knowledge scope, autonomy, limits. Not a model.	Agents
Model profile	One model inside a provider: identifier, capability, cost rates, parameters	AI providers
Provider	Connection definition: vendor kind, endpoint, authentication	AI providers
Workload route	Which profile serves Agent, Completion and Embedding work	AI providers
Tool	One callable capability with a key, schema and effect class	Tools and MCP
MCP server	External publisher of tools that MAIA Plus consumes	Tools and MCP
Procedure	Instructions that unlock and constrain tools for an agent	Procedures
Automation	Webhook endpoint that runs an agent or a legacy flow	Automations
Workflow	Saved directed graph of nodes that executes on a trigger	Workflow overview
Node	One step in a workflow graph, of a registered type	Node reference
Trigger	The first node; decides what starts a workflow	Triggers
Run	One execution, with steps, tool calls and a business result	Enterprise Audit
Revision	Monotonic version of a saved definition; a run records the one it used	Automations

Knowledge and retrieval

Term	Meaning	Where
Document	One retrievable unit: title, body, source tag, portal, visibility	Knowledge Vault
Chunk	A passage of a document; retrieval returns chunks	Indexing and search
Embedding	Fixed-length vector produced by an embedding model	AI providers
Source	Rule that keeps external content in the vault (sdpDB or file)	Knowledge Sources
Visibility	Ceiling for who may see a document (public, internal)	Roles, portals and scope
Knowledge space	Scope that caps an agent’s retrieval	Agents
RAG	Retrieval-augmented generation: retrieve chunks, then answer with them	Indexing and search

Data and prediction

Term	Meaning	Where
Assignment model	ML model that scores technician or group candidates	Assignment models
Prediction threshold	Minimum confidence for a prediction to be acted on	Assignment models
Shadow model	Trained version that predicts without acting until promoted	Assignment models
Business result	What a run reports it achieved: action executed, no action, or decision unavailable	Enterprise Audit
Ticket index	Searchable representation of tickets, lexical plus vector	Ticket Search
Incident	Cluster of similar tickets treated as one problem	Incident Radar
Data table	Typed, portal-scoped row storage owned by MAIA Plus	Data Tables

Security and operations

Term	Meaning	Where
Webhook key	Credential bound to exactly one target endpoint	Webhook keys
MAIA token	Token issued for a technician or requester session	Authentication
Portal	ServiceDesk Plus helpdesk; the usual scope boundary	Roles, portals and scope
Approval	Human decision required before a write takes effect	Approvals and autonomy
Autonomy mode	Assistant, Semi-auto, Auto — how much an agent may do alone	Approvals and autonomy
Baseline binary	Build for CPUs without AVX2	Installation
Payload	The installed code tree (current/), distinct from .env and ssl/	Upgrade
Support bundle	Diagnostic archive exported from Diagnostics	Diagnostics

Ambiguous words, resolved

• “Profile” alone is ambiguous: a model profile belongs to a provider, a run profile (automation) belongs to an endpoint. This documentation always qualifies it.
• “Agent” in a workflow means the AI Agent node; a registered agent is the reusable configuration.
• “Trigger” in ServiceDesk Plus means the SDP custom trigger; in MAIA Plus it means the workflow entry node.
• “Test” rarely means side-effect free: a provider test sends a request, an agent test can call tools.
• “Active” describes a definition or a connection, never a successful execution.

Related topics

Roles, portals and scope · Approvals and autonomy · Cost, tokens and time saved · Installation