"""Checks for the deterministic invoice result v1 boundary.""" from __future__ import annotations import json import sys import unittest from pathlib import Path from unittest.mock import patch from jsonschema import validate PROJECT_ROOT = Path(__file__).resolve().parents[3] sys.path.insert(0, str(PROJECT_ROOT)) loaded_platform = sys.modules.get("platform") if loaded_platform is not None and not hasattr(loaded_platform, "__path__"): del sys.modules["platform"] from apps.wp_invoices.services.canonical_result import build_canonical_result, _normalize_tax_summary from apps.wp_invoices.services.extractor import MODEL_MAP from apps.wp_invoices.services.pipeline import extract_canonical_invoice APP_ROOT = Path(__file__).resolve().parents[1] SCHEMA = json.loads( (APP_ROOT / "contracts" / "invoice_result.v1.schema.json").read_text(encoding="utf-8") ) EXTRACTED = { "invoice_type": {"verbatim": "Corporate", "confidence": 0.9}, "header": { "supplier": { "name": {"verbatim": "Example Supplier", "confidence": 0.95}, "abn": {"verbatim": "51 824 753 556", "confidence": 0.9}, }, "invoice": { "number": {"verbatim": "INV-42", "confidence": 0.92}, "date": { "issue_date": {"verbatim": "2026-07-23", "confidence": 0.91}, }, }, }, "items": [ { "description": {"verbatim": "Service", "confidence": 0.93}, "qty": {"verbatim": "2", "confidence": 0.9}, "unit_price": {"verbatim": "$50.00", "confidence": 0.9}, "gst_line": {"computed": "10.00", "confidence": 0.8}, "line_total": {"verbatim": "100.00", "confidence": 0.94}, } ], "totals": { "subtotal": {"verbatim": "100.00", "confidence": 0.95}, "gst": {"verbatim": "10.00", "confidence": 0.95}, "grand_total": {"verbatim": "AUD 110.00", "confidence": 0.96}, }, "steps": "Private model reasoning must not cross the public boundary.", "server_path": "/private/invoice.pdf", } class CanonicalResultChecks(unittest.TestCase): def test_product_code_is_optional_and_derived_from_leading_source_token(self): from apps.wp_invoices.services.canonical_result import _normalize_item_v2 coded = _normalize_item_v2({"description": {"verbatim": "Tuna", "computed": None, "confidence": .9}, "source_text": "59496 Tuna Flav 95g 1.09 A", "source_line": 3}) uncoded = _normalize_item_v2({"description": {"verbatim": "Service", "computed": None, "confidence": .9}, "source_text": "Service fee 1.00", "source_line": 4}) self.assertEqual("59496", coded["product_code"]) self.assertEqual("observed_unverified", coded["product_code_evidence"]["evidence_state"]) self.assertIsNone(uncoded["product_code"]) def test_tax_rates_are_decimal_and_total_comes_only_from_rows(self): summary = _normalize_tax_summary({ "tax_inclusive": True, "tax_total": {"verbatim": "99.99", "computed": 99.99, "confidence": 1}, "lines": [ {"tax_code": "A", "rate": 0, "net_amount": 68.12, "tax_amount": 0}, {"tax_code": "B", "rate": 10, "net_amount": 19.80, "tax_amount": 1.98}, ], }, 99.99) self.assertEqual([0.0, 0.1], [line["rate"] for line in summary["lines"]]) self.assertEqual(1.98, summary["tax_total"]) self.assertEqual("derived_from_tax_summary_rows", summary["tax_total_origin"]) def test_mini_profile_uses_available_vision_model(self): self.assertEqual("gpt-4o-mini", MODEL_MAP["mini"]) def _result(self, extracted=None, filename="invoice.pdf"): return build_canonical_result( extracted=EXTRACTED if extracted is None else extracted, invoice_id="inv-1", filename=filename, content_type="application/pdf", size_bytes=123, artifact_id="artifact-1", engine="mini", ) def test_result_matches_canonical_json_schema(self): result = self._result() validate(instance=result, schema=SCHEMA) self.assertEqual("completed", result["status"]) self.assertEqual(110.0, result["total"]) self.assertTrue(result["validations"]["mathematically_consistent"]) def test_public_result_omits_reasoning_evidence_and_internal_paths(self): result = self._result(filename="../../private/invoice.pdf") serialized = json.dumps(result) self.assertEqual("invoice.pdf", result["document"]["filename"]) self.assertNotIn("steps", serialized) self.assertNotIn("server_path", serialized) self.assertNotIn("/private/", serialized) def test_computed_value_takes_precedence_and_currency_defaults_to_aud(self): extracted = json.loads(json.dumps(EXTRACTED)) extracted["totals"]["grand_total"]["computed"] = "111.00" result = self._result(extracted) self.assertEqual(111.0, result["total"]) self.assertEqual("AUD", result["currency"]) self.assertFalse(result["validations"]["subtotal_plus_tax_matches_total"]) self.assertIn("subtotal_plus_tax_does_not_match_total", result["warnings"]) def test_missing_fields_are_null_with_stable_warning_codes(self): result = self._result({}) validate(instance=result, schema=SCHEMA) self.assertIsNone(result["supplier"]["name"]) self.assertIsNone(result["total"]) self.assertEqual( [ "supplier_name_missing", "invoice_number_missing", "invoice_date_missing", "line_items_missing", "total_missing", ], result["warnings"], ) def test_math_is_recomputed_instead_of_trusting_model_flags(self): extracted = json.loads(json.dumps(EXTRACTED)) extracted["totals"]["validations"] = { "items_sum_matches_subtotal": False, "subtotal_plus_gst_plus_extras_equals_grand_total": False, } result = self._result(extracted) self.assertTrue(result["validations"]["items_sum_matches_subtotal"]) self.assertTrue(result["validations"]["subtotal_plus_tax_matches_total"]) def test_pipeline_boundary_preserves_context_for_ai_usage_accounting(self): context = {"tenant": {"tenant_id": "tenant-a"}, "request_id": "request-1"} with patch( "apps.wp_invoices.services.pipeline.process_invoice_bytes", return_value={"extracted": EXTRACTED, "checks": {}}, ) as process: result = extract_canonical_invoice( file_bytes=b"document", filename="invoice.pdf", content_type="application/pdf", invoice_id="inv-1", artifact_id="artifact-1", context=context, ) process.assert_called_once_with( file_bytes=b"document", filename="invoice.pdf", content_type="application/pdf", engine="mini", context=context, ) self.assertEqual(8, result["document"]["size_bytes"]) if __name__ == "__main__": unittest.main(verbosity=2)