Denied Party Screening Engine

{
  "code": "Both files created:\n\n## rosetta_core/denied_party.py (~300 lines)\n- SanctionsEntry dataclass: name, list_name, entity_type, country, aliases, identifiers, program\n- 12 embedded sample entries across 7 lists: SDN (4), BIS Entity List (2), UFLPA (3), EU Sanctions (1), UK OFSI (1), BIS Denied Persons (1)\n- _normalize(): lowercase, strip punctuation, collapse spaces\n- _levenshtein(): pure Python Levenshtein distance\n- _similarity(): combined containment + Levenshtein scoring (0-1)\n- ScreeningMatch: per-list match with confidence, match_type (exact/alias/fuzzy)\n- ScreeningResult: query_name, hit, match_count, matches, recommended_action, risk_level, provenance\n- screen_party(): Main function — checks all lists, country filter, entity_type filter, configurable threshold\n- Risk levels: none/low/medium/high/critical → clear/review/block\n- screen_shipment_parties(): Batch screening for all shipment parties with overall action\n- Confidence thresholds: exact=1.0, high=0.85, medium=0.70\n\n## tests/test_denied_party.py (~200 lines, 35+ tests)\n- TestNormalization: 3 tests\n- TestLevenshtein: 4 tests\n- TestSimilarity: 4 tests\n- TestScreenPartyExact: 9 tests (Huawei SDN, ZTE, SMIC Entity List, Hikvision, XPCC UFLPA, Hoshine, Wagner EU, Rosneft UK, Sberbank)\n- TestScreenPartyClear: 3 tests (Microsoft, Apple, Toyota)\n- TestScreenPartyEdgeCases: 6 tests (empty, whitespace, country filter, entity type filter, threshold, Karl Lee person)\n- TestRiskLevels: 2 tests\n- TestScreenShipmentParties: 3 tests (clean, flagged, mixed)\n- TestSerialization: 3 tests\n- TestProvenance: 1 test",
  "content": "Both files created:\n\n## rosetta_core/denied_party.py (~300 lines)\n- SanctionsEntry dataclass: name, list_name, entity_type, country, aliases, identifiers, program\n- 12 embedded sample entries across 7 lists: SDN (4), BIS Entity List (2), UFLPA (3), EU Sanctions (1), UK OFSI (1), BIS Denied Persons (1)\n- _normalize(): lowercase, strip punctuation, collapse spaces\n- _levenshtein(): pure Python Levenshtein distance\n- _similarity(): combined containment + Levenshtein scoring (0-1)\n- ScreeningMatch: per-list match with confidence, match_type (exact/alias/fuzzy)\n- ScreeningResult: query_name, hit, match_count, matches, recommended_action, risk_level, provenance\n- screen_party(): Main function — checks all lists, country filter, entity_type filter, configurable threshold\n- Risk levels: none/low/medium/high/critical → clear/review/block\n- screen_shipment_parties(): Batch screening for all shipment parties with overall action\n- Confidence thresholds: exact=1.0, high=0.85, medium=0.70\n\n## tests/test_denied_party.py (~200 lines, 35+ tests)\n- TestNormalization: 3 tests\n- TestLevenshtein: 4 tests\n- TestSimilarity: 4 tests\n- TestScreenPartyExact: 9 tests (Huawei SDN, ZTE, SMIC Entity List, Hikvision, XPCC UFLPA, Hoshine, Wagner EU, Rosneft UK, Sberbank)\n- TestScreenPartyClear: 3 tests (Microsoft, Apple, Toyota)\n- TestScreenPartyEdgeCases: 6 tests (empty, whitespace, country filter, entity type filter, threshold, Karl Lee person)\n- TestRiskLevels: 2 tests\n- TestScreenShipmentParties: 3 tests (clean, flagged, mixed)\n- TestSerialization: 3 tests\n- TestProvenance: 1 test"
}