AgainFixedUserAndSupplierProblem

extractors/supplier_extractor.py

@@ -296,6 +296,8 @@ class ExpenseSupplierExtractor:
         date_phrase: str | None = None,
         excluded_phrases: list[str] | None = None,
         debug: bool = False,
+        score_threshold: float = 0.50,
+        combined_threshold: float = 0.48,
     ) -> dict[str, Any]:
         """
         Извлекает поставщика из текста.
@@ -305,11 +307,12 @@ class ExpenseSupplierExtractor:
             date_phrase: Фраза даты для исключения
             excluded_phrases: Дополнительные фразы для исключения
             debug: Включить отладочную информацию
+            score_threshold: Минимальный raw-score для принятия совпадения
+            combined_threshold: Минимальный combined-score для принятия совпадения
 
         Returns:
             Словарь с supplier, supplier_score, matched_supplier_phrase
         """
-        threshold = 0.50
         excluded_tokens: set[str] = set()
         if date_phrase:
             excluded_tokens.update(normalize_text(date_phrase).split())
@@ -333,7 +336,7 @@ class ExpenseSupplierExtractor:
             if len(token) > 1:
                 tokens.append(token)
 
-        tokens = [t for t in tokens if len(t) > 1 and t not in excluded_tokens]
+        tokens = [t for t in tokens if (len(t) > 1 or t.isdigit()) and t not in excluded_tokens]
 
         phrases: list[str] = []
         seen: set[str] = set()
@@ -367,7 +370,7 @@ class ExpenseSupplierExtractor:
                 })
 
             enriched = {**row, "combined": combined}
-            passes = score >= threshold or combined >= 0.48
+            passes = score >= score_threshold or combined >= combined_threshold
             if passes and (supplier not in best_by_supplier or combined > float(best_by_supplier[supplier].get("combined", -1.0))):
                 best_by_supplier[supplier] = enriched
 
@@ -405,6 +408,8 @@ class ExpenseSupplierExtractor:
                 "tokens": tokens,
                 "phrases_count": len(phrases),
                 "excluded_tokens": sorted(excluded_tokens)[:80],
+                "score_threshold": score_threshold,
+                "combined_threshold": combined_threshold,
                 "top_candidates": top_candidates,
             }
 

extractors/user_extractor.py

@@ -50,13 +50,14 @@ class ExpenseUserExtractor:
             date_phrase=date_phrase,
             excluded_phrases=[supplier_phrase] if supplier_phrase else None,
             debug=debug,
+            score_threshold=self.threshold,
+            combined_threshold=self.threshold,
         )
 
-        score = match.get("supplier_score")
         payload = {
-            "user": match.get("supplier") if score is not None and score >= self.threshold else None,
-            "user_score": score if score is not None and score >= self.threshold else None,
-            "matched_user_phrase": match.get("matched_supplier_phrase") if score is not None and score >= self.threshold else None,
+            "user": match.get("supplier"),
+            "user_score": match.get("supplier_score"),
+            "matched_user_phrase": match.get("matched_supplier_phrase"),
         }
 
         if debug: