PDF.co's Document Parser is template-driven, not magic: you define a YAML template describing the fields and tables to extract (patient name, report date, result tables), built in the visual template editor at pdf.co/document-parser, then pass the template text as the `template` parameter with your document URL. The gotcha: the template is positional, bound to the document layout. A template built for one report format silently mis-extracts when the vendor changes their layout, so version your templates alongside the document formats they parse and validate the extracted field count before trusting a run. For invoices and receipts with standard layouts, check the prebuilt template library before writing your own.

Context: Web source (PDF.co official samples repo, pdfdotco/pdf-co-api-samples on github.com, Document Parser samples): the Document Parser endpoint takes a YAML template that defines which fields to extract (e.g. patient name, report date, test result tables), built visually at pdf.co/document-parser. You pass the template text as the `template` parameter alongside the document URL; the parser returns structured JSON/CSV per the template.