Skip to main content
Version: 3.2

2.7 Build

Build is the second step of the Model session. It takes the Target Entities you curated in the Model and processes them into the knowledge graph, where their nodes and Relationships become queryable. A build runs as a queued job, so you can start it and follow its progress in the Queue.

Data moves through two layers:

  • Staging layer: each entity's Data Import Configuration loads source data into the entity's staging tables, where it is held before it reaches the graph.
  • Graph layer: the staged data is processed into graph nodes and relationships.

A build works through these in a fixed order: it clears staging for any entity you flagged, imports each configuration into staging (one after another), rebuilds each entity's relationship content in the staging layer, and finally processes the staged data into the graph. Once the build finishes, you can query the result in Exploration. To run builds automatically or on a schedule, use Workflows.

Open the Build step​

In the Model, select the Target Entities you want on the canvas and choose Build Selected. The entities you selected carry over into the build queue and the session moves to the Build step.

You can adjust the queue on the Build screen:

  • The Available list on the left holds every Target Entity. Search it, then use the checkboxes (or Select all) to add more entities to the queue; Clear selection empties it.
  • Each queued entity appears as a configuration card on the right. The header counts how many entities are in build queue, and Back to Model returns you to the canvas.

The Build step: the Available entity list on the left and one configuration card per queued entity on the right, each with Staging and Graph options

Per-entity parameters​

Every card sets how one entity is processed, split into a Staging side and a Graph side.

Staging

  • Import mode: how imported rows are matched against the entity's existing staging data on the business key: Insert, Update or PreservingMerge. See Import options for what each mode does. Choosing a mode enables Select configuration.
  • Select configuration: which of the entity's Data Import Configurations to import from. You can pick more than one; they run in order.
  • Clear: clear the entity's existing staging data before importing.
  • Distinct: drop duplicate rows during import. Applies to Container sources only.

Graph

  • Relationship scope: which of the entity's relationships to (re)create in the graph: All, None, Incoming, Outgoing or Selected. The N rel badge shows how many relationships touch the entity; with Selected, choose them in Select relationships.
  • Clear: clear the entity's existing graph data before processing. This removes the entity's nodes from the graph and drops its search index.
  • Process: process the staged data into the graph. This is on by default; turn it off to import into staging only, without building the graph.
  • Date/Time: process only the data imported after this point in time, instead of everything in staging.

Add related adds the entities connected to this one by a relationship to the build queue, so a related set can be built together. Remove a card from the queue with its ✕.

tip

Match the import mode to what you want to happen: Insert only adds business keys that are not in staging yet, Update only changes rows that already exist, and PreservingMerge does both (add new rows and update existing ones) in one pass.

Bulk set parameters​

When several entities should share the same settings, choose Bulk set parameters. Pick the entities on the left, set any combination of Staging and Graph parameters on the right, and choose Set to apply them to all selected cards at once.

The Bulk set parameters dialog: select entities on the left, then set shared Staging and Graph parameters to apply to all of them

Run the build​

Choose Run build to queue the job. DCH confirms with "Import job has been queued." and a View job action that opens the job in the Queue. A build has no separate confirmation step, so review the cards first.

Follow the job in Queue Jobs, where each step reports its progress. The job first prepares the entities, then imports each configuration into staging, then rebuilds the relationships in the staging layer, and finally processes everything into the graph. A running job can be cancelled.

note

A build does only what the cards ask for. It does not clear staging or graph data unless you set the matching Clear option, and it does not touch the graph at all unless Process is on. Importing into staging with Process off is a valid way to stage data without building it.

Clear the entire graph​

Clear entire graph removes all data from the knowledge graph, not just the queued entities. DCH asks you to confirm before it runs.

caution

Clearing the entire graph deletes every node and relationship in it. Use it to start the graph over, not as part of a routine build.