Glossary | OvalBee

Term	Meaning
Annotation	The labels of one image or video in a dataset: its objects, their geometry and class, and its tags. Each image has one annotation per dataset. See Assets and annotations.
Annotation tool	The labeling interface that labelers and reviewers open from a Labeling or Review stage. See Labeling.
Asset	One stored item: an image, video, annotation, model file or report.
Badge	A small label on a node showing a result of its last run, a schedule or a missing input.
Catalog	The pretrained models that come with OvalBee’s training and prediction nodes. Add one to your registry to use it. See Models and deployments.
Dataset	A named, versioned set of assets. See Datasets and versions.
Decision button	A button in the annotation tool, paired with the dataset it sends the image to. Set in the Decision buttons input of Labeling and Review. See Labeling.
Deployment	A model running as a long-lived service with an HTTP endpoint. See Models and deployments.
Edge	An arrow between two nodes. A flow edge runs the target after the source succeeds, a visual edge is drawing only. Edges never carry data. See Edges.
Flow	A graph of nodes that describes and runs a process. See Flows.
Group	A named, colored frame around related nodes on the canvas. Visual only.
Instance administrator	A person who configures the whole installation: sign-in, email, storage, compute, user accounts. See Administration.
Isolated run	A run of one node that doesn’t start the nodes after it. See Runs and automation.
Latest / pinned	A dataset or model input either follows the newest version (@latest) or stays on one exact version.
Live example	A real flow published read-only with its data and reports, opened by its public link. See Public examples.
Machine class	The kind of machine a node or deployment runs on, picked from the run button’s menu or in the Deploy dialog, or left on the installation’s default. Administrators define the classes and map them to real hardware. See Compute.
MCP	Model Context Protocol, the standard OvalBee uses to let external AI agents build and edit flows. See Build with an AI agent.
Node	One step of a flow, placed from a template in the Node Library. See Nodes.
Node template	The definition of a node in the Node Library: what it does, and its inputs and outputs.
Ontology	A versioned vocabulary of classes and tags that datasets and models are pinned to. See Ontologies.
Pool	A dataset holding work that waits for a labeling or review stage.
Reference	An input bound to another node’s output, resolved to that output’s latest value when the node runs. The app calls it a variable binding. See Edges and references.
Registry	Your workspace’s models and their versions, with metrics, classes and lineage.
Report	A document a node builds from its results, with charts, tables and image galleries. See Reports.
Run	One execution of one node, with its inputs, outputs, logs and artifacts. See Runs and automation.
Secret	A credential stored encrypted in a workspace for nodes to use. See Secrets and API tokens.
Stage	A Labeling or Review node that hands work from a pool to the people assigned to it.
Template	A reusable flow, or part of one, without any workspace data. See Templates and sharing.
Template setup	The step-by-step setup some templates open when added to a flow. It asks only what the template can’t decide, then Start creates its datasets and runs it.
Widget	The preview a node shows on the canvas: images, a report button, the annotation tool button, a file drop zone.
Workspace	Where flows, data, models, secrets and members live. Called a space in the sidebar. See Members and roles.