Console: Graph & Ontology
The Graph & Ontology pillar is where you understand and query your knowledge graph. It has four views: Explorer, Ontology, Data, and Query.
Explorer
Section titled “Explorer”A schema-first, interactive graph explorer (rendered with a force-directed canvas). You start at the schema map — the ontology’s entity types and how they connect — and drill down into instances, rather than being dumped into a hairball of every node at once.
Capabilities:
- Group by type to collapse instances into their entity types, then expand a type to see its members.
- Fit to view, re-layout, and a minimap for navigating large graphs.
- Sample projection to render a bounded subset when a graph is too large to draw whole — supernodes are handled deliberately, not blindly expanded.
- Click a node to open its detail panel: attributes, incoming/outgoing edges, and history, with paged reads for high-degree nodes.
This mirrors the SDK reads (entities.list, traverse, entity detail) but as a
visual, clickable surface.
Ontology
Section titled “Ontology”Inspect the graph’s ontology: entity types, relation types (with source/target domains and reducers), concepts, terms, and property definitions. The view can render the ontology as Turtle/RDF and show derived SHACL shapes (cardinality and allowed-value constraints derived from the ontology).
The ontology is fixed at graph creation, but you can additively evolve it —
widen a relation’s domain/range or add entity types by name — which bumps the
ontology version while preserving every existing record. To author a brand-new
ontology, create a new graph (the SDK/CLI define_ontology / graph create --ontology path, or MCP lbb_configure).
Browse and manage the actual facts: list entities with their flat attributes,
filter by type, and inspect assertions and their evidence. This is the
tabular counterpart to Explorer — useful when you want to read exact values
rather than navigate visually. It also surfaces per-entity indexing status
(bm25_indexed_commit_seq / ann_indexed_commit_seq) so you can see what’s
covered by the current index runs.
Run the three query surfaces against the selected graph and see results inline:
- Structured — build a SPARQL-subset SELECT/aggregate with typed filters,
GROUP BY(scalar and date-bucket keys), and aggregates. - SPARQL text — run conformant SPARQL 1.1 SELECT/ASK; results render as a table.
- SHACL — select matching focus nodes or validate and read a conformance report.
Results paginate with cursors and pin to a snapshot, matching the API behavior. This is the interactive way to develop a query before moving it into SDK code.