Data Sources

EN 16931 — the European Norm for electronic invoicing. Our canonical internal representation. Every endpoint accepts or produces EN 16931 JSON.

Peppol BIS 3.0 — Pan-European Public Procurement Online business interoperability specification. Universal Business Language (UBL) 2.1 profile.

XRechnung — German mandatory B2G format based on CII D16B (Cross-Industry Invoice). Two flavours: xrechnung_ubl and xrechnung_cii.

Factur-X / ZUGFeRD — hybrid PDF/A-3 with embedded CII D16B XML. Factur-X is the French variant, ZUGFeRD the German — they share the same technical base.

UBL 2.1 — Universal Business Language. Raw UBL invoice without a specific jurisdiction profile.

The /v1/intelligent/parse endpoint uses AI (Anthropic Claude) to extract structured invoice data from free-form text. Before any text leaves our EU infrastructure, it passes through Microsoft Presidio with a multilingual spaCy NER pipeline that replaces PII (names, VAT IDs, IBANs, emails, phones, addresses) with format-preserving placeholders.

After the AI response comes back, placeholders are reversed in-process. The AI never sees real PII, and strict-mode post-checks refuse the request (422 UNANONYMIZED_PII_DETECTED) if any residual PII slips through.

Legal basis for cross-border AI processing: EU-US Data Privacy Framework (DPF).

Schema validation uses the official XSD files published by OpenPEPPOL, XRechnung (KoSIT), and UN/CEFACT.

Business-rule validation uses Schematron rules sourced from the same upstream authorities. Rule sets are updated on each upstream release.