Update app.py

app.py

@@ -487,7 +487,7 @@ class TextToSQLSystem:
     def _validate_and_fix_sql(self, question: str, raw_response: str) -> Tuple[Optional[str], str]:
         """
         (V23 / 统一实体识别版)
-        一個全面、多層次的 SQL 驗證與生成引擎。
+        一个全面、多层次的 SQL 验证与生成引擎。
         引入了全新的、统一的实体识别引擎，能够准确解析 "买家 Gap", "c0761n", 
         "买家ID c0761n" 等多种复杂的实体提问模式。
         """
@@ -503,19 +503,19 @@ class TextToSQLSystem:
         # 定义多种识别模式，【优先级从高到低】
         entity_patterns = [
             # 模式1: 匹配 "类型 + ID" (e.g., "买家ID C0761N") - 最高优先级
-            {'pattern': r"(买家|buyer)\s*(?:id|代號|代码)\s*'\"?\b([A-Z]\d{4}[A-Z])\b'\"?", 'column': 'sd.BuyerID', 'type': '買家ID'},
-            {'pattern': r"(申请方|申请厂商|applicant)\s*(?:id|代號|代码)\s*'\"?\b([A-Z]\d{4}[A-Z])\b'\"?", 'column': 'sd.ApplicantID', 'type': '申請方ID'},
+            {'pattern': r"(买家|buyer)\s*(?:id|代號|代码)\s*'\"?\b([A-Z]\d{4}[A-Z])\b'\"?", 'column': 'sd.BuyerID', 'type': '买家ID'},
+            {'pattern': r"(申请方|申请厂商|applicant)\s*(?:id|代號|代码)\s*'\"?\b([A-Z]\d{4}[A-Z])\b'\"?", 'column': 'sd.ApplicantID', 'type': '申请方ID'},
             {'pattern': r"(付款方|付款厂商|invoiceto)\s*(?:id|代號|代码)\s*'\"?\b([A-Z]\d{4}[A-Z])\b'\"?", 'column': 'sd.InvoiceToID', 'type': '付款方ID'},
             {'pattern': r"(代理商|agent)\s*(?:id|代號|代码)\s*'\"?\b([A-Z]\d{4}[A-Z])\b'\"?", 'column': 'sd.AgentID', 'type': '代理商ID'},
             
             # 模式2: 匹配 "类型 + 名称" (e.g., "买家 Gap")
-            {'pattern': r"(买家|buyer|客戶)\s*'\"?([a-zA-Z0-9&.\s-]+?)(?:\s*的|\s+|$|有)", 'column': 'sd.BuyerName', 'type': '買家'},
-            {'pattern': r"(申请方|申请厂商|applicant)\s*'\"?([a-zA-Z0-9&.\s-]+?)(?:\s*的|\s+|$|有)", 'column': 'sd.ApplicantName', 'type': '申請方'},
+            {'pattern': r"(买家|buyer|客戶)\s*'\"?([a-zA-Z0-9&.\s-]+?)(?:\s*的|\s+|$|有)", 'column': 'sd.BuyerName', 'type': '买家'},
+            {'pattern': r"(申请方|申请厂商|applicant)\s*'\"?([a-zA-Z0-9&.\s-]+?)(?:\s*的|\s+|$|有)", 'column': 'sd.ApplicantName', 'type': '申请方'},
             {'pattern': r"(付款方|付款厂商|invoiceto)\s*'\"?([a-zA-Z0-9&.\s-]+?)(?:\s*的|\s+|$|有)", 'column': 'sd.InvoiceToName', 'type': '付款方'},
             {'pattern': r"(代理商|agent)\s*'\"?([a-zA-Z0-9&.\s-]+?)(?:\s*的|\s+|$|有)", 'column': 'sd.AgentName', 'type': '代理商'},
             
             # 模式3: 单独匹配一个 ID (e.g., "c0761n") - 较低优先级
-            {'pattern': r"\b([A-Z]\d{4}[A-Z])\b", 'column': 'sd.ApplicantID', 'type': 'ID'} # 默认为 ApplicantID，可以根据业务调整
+            {'pattern': r"\b([A-Z]\d{4}[A-Z])\b", 'column': 'sd.ApplicantID', 'type': 'ID'}
         ]
         
         for p in entity_patterns:
@@ -528,19 +528,12 @@ class TextToSQLSystem:
                     "column": p['column']
                 }
                 break
-
+        
         # --- 预先检测其他意图 ---
         job_no_match = re.search(r"(?:工單|jobno)\s*'\"?([A-Z]{2,3}\d+)'\"?", question, re.IGNORECASE)
-        lab_group_match_data = None
-        LAB_GROUP_MAP = {'A':'TA','B':'TB','C':'TC','D':'TD','E':'TE','Y':'TY','TA':'TA','TB':'TB','TC':'TC','TD':'TD','TE':'TE','TY':'TY','WC':'WC','EO':'EO','GCI':'GCI','GCO':'GCO','MI':'MI'}
-        lab_group_match = re.findall(r"([A-Z]+)\s*組", question, re.IGNORECASE)
-        if lab_group_match:
-            codes = [LAB_GROUP_MAP.get(g.upper()) for g in lab_group_match if LAB_GROUP_MAP.get(g.upper())]
-            if codes: lab_group_match_data = {"codes": codes, "identifiers": lab_group_match}
-
+        
         # --- 判断逻辑: 依优先级进入对应的模板 ---
-
-        if any(kw in q_lower for kw in ['報告號碼', '報告清單', '列出報告', 'report number', 'list of reports']):
+        if any(kw in q_lower for kw in ['報告號碼', '報告清單', '列出報告']):
             year_match = re.search(r'(\d{4})\s*年?', question)
             month_match = re.search(r'(\d{1,2})\s*月', question)
             from_clause = "FROM JobTimeline AS jt"
@@ -590,12 +583,6 @@ class TextToSQLSystem:
         fixed_sql = " " + parsed_sql.strip() + " "
         fixes_applied_fallback = []
         
-        dialect_corrections = {r'YEAR\s*\(([^)]+)\)': r"strftime('%Y', \1)"}
-        for pattern, replacement in dialect_corrections.items():
-            if re.search(pattern, fixed_sql, re.IGNORECASE):
-                fixed_sql = re.sub(pattern, replacement, fixed_sql, flags=re.IGNORECASE)
-                fixes_applied_fallback.append(f"修正方言: {pattern}")
-        
         schema_corrections = {'TSR53Report':'TSR53SampleDescription', 'TSR53InvoiceReportNo':'JobNo', 'TSR53ReportNo':'JobNo', 'TSR53InvoiceNo':'JobNo', 'TSR53InvoiceCreditNoteNo':'InvoiceCreditNoteNo', 'TSR53InvoiceLocalAmount':'LocalAmount', 'Status':'OverallRating', 'ReportStatus':'OverallRating'}
         for wrong, correct in schema_corrections.items():
             pattern = r'\b' + re.escape(wrong) + r'\b'