Patient data lives inside an iframe that is itself inside another iframe. STATIC extraction only traverses one iframe level via getNonProtectedIframes, so XPath evaluation never reaches the inner content. LLM_DOM receives the fully serialized DOM (including nested iframe content) and extracts correctly.
evaluateXPathOnIframes only checks top-level iframes via getNonProtectedIframes(document)contentDocument is checked, but the inner iframe inside it is never traversedLLM_DOM works because getFullDocumentWithIframes recursively serializes nested iframes into the DOM string sent to the backend{
"action": "EXTRACT_DATAMODEL",
"parameters": {
"execution": "LLM_DOM",
"extract_data_model": {
"type": "object",
"properties": {
"patients": {
"type": "array",
"items": {
"type": "object",
"properties": {
"mrn": { "type": "string" },
"name": { "type": "string" },
"dob": { "type": "string" },
"diagnosis": { "type": "string" }
}
}
}
}
}
}
}