CLOUDCRUISE UNITED

LLM_VISION only

ExtractDatamodel — Cross-Origin Iframe

This page simulates a common healthcare portal pattern: benefits data displayed inside a cross-origin iframe. The contentDocument property returns null for cross-origin frames, so STATIC extraction via evaluateXPathOnIframes fails silently. Only LLM_VISION (screenshot) can extract this data.

Simulated scenario: The iframe below would normally point to benefits.insurance-partner.com (different origin).

In this demo, we render the benefits data as a same-origin iframe to show the visual layout. In production, the cross-origin iframe blocks contentDocument access entirely.

Cross-origin frame — contentDocument is null

Ground truth (what the LLM should extract from the screenshot)

Plan Code	Deductible	Copay	OOP Max	Status
PPO-500	$500	$25	$6,000	Active
HMO-250	$250	$15	$4,500	Active
HDHP-1500	$1,500	$0	$7,000	Inactive

Why only LLM_VISION works

Cross-origin iframes block contentDocument access (same-origin policy)
evaluateXPathOnIframes skips frames where contentDocument is null
LLM_DOM also fails — the serialized DOM cannot include cross-origin frame content
LLM_VISION works because the browser composites the cross-origin content into the screenshot
Real-world examples: insurance eligibility portals, embedded payment forms, third-party lab results viewers

Workflow node config

{
  "action": "EXTRACT_DATAMODEL",
  "parameters": {
    "execution": "LLM_VISION",
    "extract_data_model": {
      "type": "object",
      "properties": {
        "benefits": {
          "type": "array",
          "items": {
            "type": "object",
            "properties": {
              "plan_code":  { "type": "string" },
              "deductible": { "type": "string" },
              "copay":      { "type": "string" },
              "oop_max":    { "type": "string" },
              "status":     { "type": "string" }
            }
          }
        }
      }
    }
  }
}