name: ZenML Vocabulary description: >- Domain vocabulary for the ZenML MLOps and LLMOps platform, covering pipelines, stacks, components, runs, artifacts, models, deployments, schedules, secrets, projects, service connectors, and the broader ZenML governance and integration concepts. modified: '2026-05-03' terms: - term: Pipeline definition: A directed acyclic graph (DAG) of steps that processes data and produces artifacts and models. related: [Step, PipelineRun, Stack] - term: Step definition: A single, decorated Python function that becomes a node in a pipeline DAG. related: [Pipeline, Artifact, Materializer] - term: PipelineRun definition: An execution instance of a pipeline against a specific stack at a specific time. related: [Pipeline, Stack, Status] - term: Stack definition: A configurable collection of stack components used to execute a pipeline against target infrastructure. related: [StackComponent, Orchestrator, ArtifactStore] - term: StackComponent definition: An individual capability registered against a stack, such as an orchestrator, artifact store, or experiment tracker. related: [Stack, Flavor] - term: Orchestrator definition: A stack component that schedules and runs the steps of a pipeline (Kubernetes, Airflow, SageMaker, Vertex, etc.). related: [StackComponent] - term: ArtifactStore definition: A stack component that persists artifacts produced by pipeline steps (S3, GCS, Azure Blob, local). related: [StackComponent, Artifact] - term: ContainerRegistry definition: A stack component that hosts container images used by the orchestrator and step operators. related: [StackComponent] - term: ExperimentTracker definition: A stack component that records run metadata, metrics, and parameters (MLflow, W&B, Neptune, Comet). related: [StackComponent, PipelineRun] - term: ModelDeployer definition: A stack component responsible for serving registered models behind real-time or batch endpoints. related: [StackComponent, Model] - term: Artifact definition: A typed, materialized output of a step, persisted to the configured artifact store. related: [Materializer, ArtifactStore] - term: Materializer definition: A serialization/deserialization adapter that defines how an artifact type is read and written. related: [Artifact] - term: Model definition: A registered ML or LLM model in the ZenML Model Control Plane, with model card metadata and a chain of versions. related: [ModelVersion] - term: ModelVersion definition: A specific iteration of a model, tagged with a stage (none, staging, production, archived). related: [Model, Stage] - term: Stage definition: The lifecycle stage of a model version (none, staging, production, archived). related: [ModelVersion] - term: Deployment definition: A serving deployment that exposes a pipeline or a model behind an endpoint. related: [ModelDeployer, Pipeline] - term: Schedule definition: A cron- or interval-based trigger for repeatedly running a pipeline. related: [Pipeline, PipelineRun] - term: Secret definition: An encrypted credential or configuration value stored in the ZenML server. related: [ServiceConnector, Stack] - term: ServiceConnector definition: A reusable, scoped credential and configuration bundle for connecting to external infrastructure (AWS, GCP, Azure, Kubernetes). related: [Secret, Stack] - term: Project definition: A workspace boundary within a ZenML server for organizing pipelines, models, and runs. related: [Pipeline, Model] - term: Trigger definition: An external event source that initiates a pipeline run (webhook, schedule, manual). related: [Schedule, PipelineRun] - term: ControlPlane definition: The ZenML server component that hosts metadata, governance, and orchestration coordination. related: [Stack, PipelineRun] - term: Flavor definition: A specific implementation of a stack component type (e.g. kubeflow, sagemaker, mlflow). related: [StackComponent] - term: ZenMLPro definition: The managed ZenML control plane offering with multi-tenant governance, RBAC, and audit logs. related: [ControlPlane]