One canonical data model — expressed as a living knowledge graph — so every term, relationship, and measurement means the same thing everywhere.
Discovery data is only as good as the vocabulary underneath it. Halffield treats its data model as a first-class, browsable, editable knowledge graph: entities are nodes, their fields and values are properties, and the relationships between them are typed, cardinality-aware edges. Explore it as a tree-grid, visualize it as a graph, query paths between concepts, filter by DMTA stage — then export to OWL/RDF or Cytoscape.
Underneath sits a centralized vocabulary registry. Every controlled vocabulary opens one rich drawer showing its values, its source of truth, everything that uses it, and full editing — and can connect to an external ontology so your terms stay in sync with the world’s.
The result: no drift. One source of truth per concept, surfaced consistently across the designer, the campaign matrix, and the AI agents.
An ARIA tree-grid over categories → entities → fields → values, with a grain selector, facet filters, inline editing, and a knowledge-graph relationship layer.
A visual projection of the same node+edge model, with path, neighborhood, and DMTA-stage lenses that highlight the matching sub-graph in place.
Values + source + used-by + CRUD + parent→child cascades (species → strain), for any vocabulary.
Connect a vocabulary to Open Targets, EFO/MONDO, ChEMBL, or a custom source; pull values with provenance.
OWL/Turtle (classes, properties, cardinality) or Cytoscape property-graph, straight from the UI.
A CI gate blocks any change that would create a parallel catalog or an undocumented model — so the single source of truth stays single.