Knowledge Graph
Visual map of your ontology's object types and the relationships between them — explore your data model at a glance.
The Knowledge Graph page (/graph in Analytics) gives you a live, explorable map of your ontology. Each node represents an object type; each edge represents a declared link type between two types.
Reading the map
Every node (circle) on the canvas corresponds to one of your object types. Node size is proportional to the number of instances currently bound: a large circle means many rows are projecting through that type; a small one means few (or none yet).
Each edge connecting two types shows the link type's display name and a count:
| Edge appearance | Meaning |
|---|---|
| Solid line with a count | Edges have been inferred — the count is the number of linked pairs. |
| Dashed red with 0 | The link type matched nothing. The most common cause is a FK column that isn't bound, or values that differ in case or formatting between the source and target tables. |
| Dashed with ? | The count couldn't be computed yet — usually because one of the bound tables hasn't been indexed, or a required column mapping is missing. |
A 0 edge is not an error in the ontology definition — it means the link type is declared correctly but the data doesn't satisfy it yet. The Bindings tab is the right place to check whether the column map is pointing at the right columns.
Expanding a type
Double-click any node to load a ring of its instances around it (up to 50). Each instance appears as a smaller circle; its label comes from the type's identity rule.
If more than 50 instances exist, a +N more chip appears at the bottom of the ring. Clicking it opens the search panel pre-filtered to that type so you can find the record you're looking for.
Double-click an edge to load the actual linked pairs — the source and target instances that satisfy that relationship. This is useful for confirming that a link type is resolving the way you expect.
Searching
The search pill at the top of the canvas searches across all bound text values. As you type, matching instances are suggested with their type label. Selecting a result pins that record on the canvas and draws its immediate neighbors — all the instances it is linked to via any declared link type.
Legend and inspector
The legend (bottom-left) lists every object type in the current schema with its color swatch. Click a type name to toggle it on or off; hiding a type hides its nodes and any edges that connect only to hidden types.
The inspector panel (right side) opens when you select a record — either from the search results or by clicking an expanded instance node. It shows:
- The record's bound property values.
- Its connections, grouped by relationship, once neighbors have been loaded.
- A button to load the record's one-hop neighbors directly on the canvas (for a type node, the button expands the type's instances instead).
Diagnostics chip
The N issues chip in the page header appears when one or more link types have zero or unknown edge counts, or when bindings are not ready. Click it to see the full list with the reason for each:
- Link type, 0 edges — inferred no pairs; check that the FK column is in the column map and that values match.
- Link type, ? count — count couldn't be computed; check that the binding is ready for both sides of the relationship.
- Binding not ready — the table isn't picked yet or a required column is missing; fix it in the Bindings tab.
Selecting an ontology and branch
The header dropdowns let you switch between ontologies (org-wide and workspace-level) and between branches. The graph always reflects the schema on the selected branch, including any unpublished proposal branches — useful for previewing what a schema change will look like before publishing it.
Related
- Object types — define the nodes on the graph.
- Link types — define the edges.
- Bindings — connect nodes to real data and determine instance counts.
- Branches & proposals — the branch shown in the graph picker.