Graph explorer — click a label, expand neighbors, three lines of Cypher

The previous two lessons defined the semantic layer. This lesson confirms how the instances landed on top of that layer look. The Graph explorer is the second surface modelers live in most — it's the fastest place to verify that a new entity or relation works as intended.

Entry

Enter from the left sidebar via ONTOLOGY → Graph Explorer. Coming in via the collection context auto-applies that collection's graph scope.

The screen splits into three areas.

1. Left — metadata panel — Defined labels, relation types, and attribute keys, with a count next to each.
2. Center — visualization area — The canvas where nodes and edges arrange themselves in a force-directed layout.
3. Bottom — query editor — For writing and running Cypher directly.

The lesson assumes the previous lesson's retail 4 entities and 3 relations are loaded.

A single click — querying by label

In the left metadata panel's Labels (Nodes) area, the number next to RI_Product is the node count. Click a label once and the Cypher query for all nodes of that label runs automatically.

• Nodes spread across the visualization area in a force-directed layout.
• Node colors are auto-assigned per label — when multiple labels mix in one view, the colors distinguish them.
• Click one node and the right panel (or modal) shows every attribute of that instance. The Display Column is the primary title, with system keys below.

For datasets with too many rows (e.g. tens of thousands), LIMIT is applied automatically — to keep the screen from stalling.

Double-click — expand neighbors

Double-click one node and its connected neighbors are queried and added.

In the retail example, follow this once:

1. Click the RI_Region label once — a few region nodes appear.
2. Double-click the Capital region node — branches connected via RI_LOCATED_IN appear alongside.
3. Double-click one branch — products connected to that branch via RI_STOCKED_AT come up alongside.

In three clicks, region → branch → product N-hop neighbors are gathered into a single view. This pattern is the fastest exploration line in the Graph explorer.

The three-line Cypher pattern — MATCH · RETURN · LIMIT

Label clicks and double-clicks handle 80% of querying. The remaining 20% is written in Cypher at the bottom of the query editor. This lesson covers just the three-line pattern — anything beyond is in the Cypher guide manual.

Pattern 1 — all nodes of one label

MATCH (p:RI_Product)
RETURN p
LIMIT 100

This is exactly the form auto-run by a label click. Typing it once yourself makes the next patterns feel natural.

Pattern 2 — pairs connected by a single relation

MATCH (p:RI_Product)-[:RI_STOCKED_AT]->(b:RI_Branch)
RETURN p, b
LIMIT 50

The arrow direction (->) matches the source → target direction set in lesson 04. Reverse it and the result is empty.

Pattern 3 — filtering by condition

MATCH (p:RI_Product)-[r:RI_STOCKED_AT]->(b:RI_Branch)
WHERE r.stockout_risk_flag_synth = 1
RETURN p.name, b.name, r.inventory_qty
LIMIT 20

Filtering by a relation attribute (r.stockout_risk_flag_synth) brings up only the product-branch pairs at stockout risk. The relation attributes defined in lesson 04 deliver value directly here.

Three result-view modes

You read query results in one of three modes.

• Graph view — Nodes and relations visualized. For pattern and structure recognition.
• Table view — Tabular form. Sort, filter, and export.
• Text view — Raw JSON. For debugging.

Alternating between modes on the same result naturally builds the modeler's standard line: look graphically, verify tabularly.

URL deep links — shareable exploration links

Specific node selections or query states can be prefilled via URL query parameters so they're shareable. Useful for governance audits, investigation requests, and handoffs — this link reappears in lesson 06's handoff step.

Entity Map — when coordinate attributes exist

If an entity carries latitude and longitude attributes, the Entity Map tab can visualize markers.

• See spatial patterns like store distribution, IoT sensor locations, and logistics hubs.
• The coordinate system is WGS84 only. Korean KATEC / TM / UTM-K coordinates need to be converted up front.
• Marker clusters auto-cluster at density.

In the retail example, RI_Branch carries latitude and longitude attributes, so it can be put on the map as-is.

Self-check

• The label counts in the metadata panel are not 0 (if they are, go back to the lesson 02–03 mapping / sink steps).
• A single label click auto-expands nodes onto the visualization area.
• A node double-click brings in neighbor nodes alongside.
• You've typed at least two of the three-line Cypher patterns (MATCH · RETURN · LIMIT) by hand.
• You've viewed the same query in both Graph view and Table view.

Next lesson

The final lesson closes the Path — how to handle your graph at the maintenance stage: how model changes propagate, how to add a mapping when a new dataset arrives, and how to hand off to analysts and engineers, one paragraph each.