Workflow
Process, description, execution
An arrangement of activities and control rules for carrying out a process. Distinguish the process, its description, and the system that executes it. [3] [1] Editorial interpretation
Open AI Glossary
A shared reference for research and engineering: compare meanings, inspect their boundaries, and follow the evidence.
Definitions with boundaries, evidence and an explicit repository proposal.
Browse definitions
Harness
The execution layer around a model
In agentic AI, a harness is software around a model that connects it to context and tools and controls execution. Sources differ on whether this means the control loop or the wider non-model runtime. [4] [5] Editorial interpretation
Model
Representation and inference component
Model has several scopes: a representation of a subject; an executable representation; or, in AI, the inference component transforming inputs into outputs. State the scope before using the term. [7] [8] Editorial interpretation
LLM
Large language model
A large language model is a deep learning language model trained on very large textual datasets. It is a kind of Model; its complete application or agent runtime is a separate system. [8] [14] Editorial interpretation
Agent
Goal-directed action with feedback
In general computing, an agent is a situated system with some control over its actions, responsiveness to its environment and goal-directed behavior. LLM-based agents are one implementation, not the whole category. [10] [14] Editorial interpretation
Schema
Data schema
A schema describes how data is organized and what its fields mean, with structural or value constraints where specified. State the data model and dialect: the exact obligations depend on the schema language. [16] [18] [19] Editorial interpretation
About this glossary
Read in the order Workflow, Harness, Model, LLM, Agent, Schema. The sequence is a reading path, not a claim that all six form a single subtype hierarchy.
External evidence and repository certification are separate. Each working definition remains a proposal until a documented repository consensus decision is recorded.
How we choose meanings
Research applicable formal institutional definitions first, then peer-reviewed papers, then industry standards, first-party documentation, preprints and academic materials, and finally Repository internal proposals.
Verify the original text, edition, publication status and context. A corporate engineering blog belongs to tier 3. Hosting a paper at a university does not change its publication type.
Prefer relevance over an unrelated authority. Preserve attributed disagreements, record source gaps, and do not turn a single paper into a consensus statement.
Citations accompany definitions, boundaries and typed relations. An editorial inference is marked explicitly; a source citation identifies its basis, not an endorsement of our inference.
Repository certification
Repository · Proposed / not certified
Yellow means Repository · Proposed / not certified. Green means Repository · Certified by consensus and requires an explicit unanimous repository decision, date, accessible decision link and evidence record. A merge, publication or authoritative external source does not establish repository certification.