Skip to main content

​
AI Node

  • Model — the models available to your tenant; the hint links to Settings → AI where a manager can bring your own Anthropic key.
  • System Prompt and Prompt — use {{variable}} to interpolate case data.
  • Output Format — Text or Structured JSON; with JSON, declare Output Fields.
  • Temperature, Max Retries, and Attach documents to pass images (for example vars.kyc.document_images).
Results land under Output variable and surface to analysts in the Customer context card.

​
Where the model runs

By default AI nodes run on Frayme’s account. A tenant admin can supply an Anthropic key in Settings → AI; the node then bills to that key. Case analysis in the risk engine always runs on Frayme’s account.

​
Sub-Flow

Runs a flow marked Callable as subflow as one or more child cases and waits for them.
  • Callable workflow — which flow to run.
  • Fan out over — an array expression (for example input.ubos) to create one child per element; Child ID identifies each.
  • Child role in parent — Ultimate Beneficial Owner, Owner, Legal Representative, none, or an expression. Analysts see it on the child.
  • Pass data to the child — maps the child’s inputs.
  • Output variable — where the children’s outputs are collected.

​
Fan out and child subjects

Under Advanced, choose how each child’s subject is built (from mapped fields or the related party), its Kind (Person or Business) and display fields.

​
Advanced

Timeout (seconds) with On timeout (Partial folds what finished and marks the rest unresolved; Fail errors the parent). Re-entry controls decide what happens when the parent runs the Subflow again: On each pass, When an element disappears (cancel or keep its child case) and Max cases per element.

​
Sub-Flow (start) and (wait)

On save the builder compiles a Subflow into a start and a wait step. If a flow is loaded with them expanded, their panels are read-only: edit through the Subflow node.