feat:提示词约束2以及bug修复

template.yaml

@@ -78,12 +78,13 @@ template:
           <rule-detail>使用了函数（如 COUNT(), CAST(), SUM()）的字段，必须为其指定一个英文别名。</rule-detail>
           <rule-detail>计算占比或百分比时，结果保留两位小数，并以 '%' 符号结尾。示例：ROUND(COUNT(*) * 100.0 / (SELECT COUNT(*) FROM table), 2) || '%' (PostgreSQL语法)</rule-detail>
           <rule-detail>若查询结果包含枚举字段（如 gender=1,2），必须使用 CASE WHEN 语句将其转换为可读的标签。示例: SELECT CASE WHEN "gender" = '1' THEN '男' WHEN "gender" = '2' THEN '女' END AS "gender"</rule-detail>
-         <rule-detail>重点!重点!重点！涉及查询orgs表时，部门存在多层级，必须递归查询。使用 start .. with 语法。递归语法示例:
+         <rule-detail>重点!重点!重点！涉及查询orgs表时，部门存在多层级，大部分需要递归查询。使用 start .. with 语法。递归语法示例:
                   SELECT "id"
                   FROM "IUAP_APDOC_BASEDOC"."org_orgs" START
                   WITH "name"||"shortname" LIKE '%xx中心%' AND "dr"=0 AND "enable"=1 AND "code" LIKE '%CYJ%'
                   CONNECT BY PRIOR "id" = "parentid"
-        </rule-detail>
+         </rule-detail>
+       <rule-detail>当用户问题涉及查询是否时候。结果需要返回：是/否。而不是返回查询记录</rule-detail>
         </rule>
         <rule>
           <rule-title>关联与限制</rule-title>

util/load_ddl_doc.py

@@ -28,9 +28,10 @@ def add_documentation(vn: CustomVanna):
     random.shuffle(dd)
     # for d in dd:
     #     vn.add_question_sql(question=d['question'], sql=d['answer'])
-    for d in q_and_a_test1.question_and_answer:
+    ## tag  待定，容易过拟合
+    for d in dd:
         vn.add_question_sql(
-            question=f"{d['question']} [分类:{d['category']}] [标签:{','.join(d['tags'])}]",
+            question=f"{d['question']}",
             sql=d['answer']
         )