feat: coverage V2 - historical evidence miner, KB evidence model + promotion audit + revalidation, Kodebar/OSCBR limits, KB/provider panel, seed import; 1.3.2 (Hub only)

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

central/app/enrichment/textnorm.py

@@ -7,6 +7,7 @@ UNIT_MAP = {"LT": "L", "LTS": "L", "LITRO": "L", "LITROS": "L", "L": "L", "ML":
 ABBREVIATIONS = {"REFRIG": "REFRIGERANTE", "REFRI": "REFRIGERANTE", "CHOC": "CHOCOLATE", "BISC": "BISCOITO", "SAB": "SABONETE", "DET": "DETERGENTE", "SH": "SHAMPOO", "COND": "CONDICIONADOR",
                  "MACARR": "MACARRAO", "MAC": "MACARRAO", "AMAC": "AMACIANTE", "DESOD": "DESODORANTE", "PAP": "PAPEL", "HIG": "HIGIENICO", "ORIG": "ORIGINAL", "TRAD": "TRADICIONAL"}
 STOP = {"DE", "DA", "DO", "DAS", "DOS", "E", "COM", "C", "PARA", "P", "EM", "A", "O"}
+MARKETING = {"ORIGINAL", "TRADICIONAL", "NOVO", "NOVA", "PROMO", "PROMOCAO", "OFERTA"}      # ignored ONLY by the fingerprint; a different pack/flavour still differs
 _QTY = re.compile(r"^(\d+(?:[.,]\d+)?)(ML|L|KG|G|MG|UN|PCT|CX|LT|LTS|GR)$")
 
 
@@ -35,7 +36,7 @@ def normalized(description) -> str:
 
 
 def fingerprint(description) -> str | None:
-    toks = tokens(description)
+    toks = [t for t in tokens(description) if t not in MARKETING]
     if not toks:
         return None
     return hashlib.sha256(" ".join(sorted(toks)).encode("utf-8")).hexdigest()