Dashboards and Neo4j exploration

Dashboard panels query one of two native datasources: the Neo4j read model (ArchiMate content, exported from PostgreSQL) or ArchiSpark's own PostgreSQL application database directly — both ArchiMate content (elements, relationships, views, element_properties) and organization-scoped business tables such as fournisseurs. Every built-in system dashboard queries PostgreSQL directly; Neo4j remains available for /explore and for any workspace-owned dashboard you create. PostgreSQL remains authoritative for ArchiMate content: export the workspace again whenever it changes if you build a Neo4j-backed panel.

Prepare the graph

Configure NEO4J_URI, NEO4J_USER, and NEO4J_PASSWORD, then call POST /api/export/neo4j for the active workspace. For batch imports, use pnpm import:workspace -- <workspace-uuid> or pnpm import:workspaces.

Every exported node and relationship carries organizationId. Re-exporting a workspace replaces only that workspace's graph.

Explore (/explore)

The explorer accepts read-only Cypher and JSON parameters. It provides presets, keeps the latest 20 executions in browser storage, renders compatible results as a graph or table, and exports rows to JSON or CSV.

$organizationId is injected by the server. Rows containing a node or relationship from another organization are removed from the result.

MATCH (element:Element {organizationId: $organizationId})
RETURN element.type AS type, count(element) AS count
ORDER BY count DESC

Dashboards (/dashboards)

• All organization members can list dashboards and execute panels; a panel returns graph, table, or metrics.
• Owners and admins can create and edit dashboards — saving an edit creates an immutable revision, deletion is soft.
• Every workspace sees the same fixed set of system dashboards (marked "System" in the admin list), shared globally rather than copied per workspace. They can be neither edited nor deleted.
• Graph panels provide controls for filtering element and relationship types, choosing the edge style and layout direction, and entering fullscreen (Escape or the minimize control to exit).
• The form uses structured metadata fields and validated JSON for parameters, panels, layouts, and tab groups; identifiers use lowercase kebab-case.
• Every panel query must target a native datasource (architecture-neo4j Cypher, or postgres-app-db read-only SQL), reference $organizationId, and select a compatible visualization: core/graph, core/table, or core/metric. Both datasources support graph panels — a SQL graph panel must return a nodeIds column (elements.uuid values); the server resolves matching node/edge metadata from elements/relationships.
• The dashboard and admin lists show each dashboard's datasource type(s) in parentheses next to its title — e.g. (postgres), (neo4j, postgres).

Current limits

• Neo4j is not updated automatically after model edits.
• Dashboard content uses a JSON editor, not a visual panel builder.
• Declared client-side transformations are accepted but not yet applied.
• Explorer queries are read-only and results may be truncated.

See Architecture for persistence and tenant-isolation details.