Merge branch 'dev_graph' of http://106.13.42.156:33077/lei_y601/sqlbot_agent into dev_graph

graph_chat/gen_data_report_agent.py

@@ -45,18 +45,15 @@ def _run_sql(state: DateReportAgentState) -> dict:
         update_conversation(id=state.get('id'), sql=sql)
         df = vn.run_sql_2(sql)
         result = df.to_dict(orient='records')
-        logger.info(f"run sql result {result}")
+        logger.info(f"run sql result => {result}")
         retry = state.get("retry_count", 0) + 1
         logger.info(f"old_sql:{sql}, retry sql {retry_sql}")
         dd= {'data': result, 'retry_count': retry, 'retry_sql': retry_sql,'run_sql_error':''}
-        #不需要模型反馈，运行成功后则成功
-        if not NEED_MODEL_FEEDBACK_QA:
-            dd['sql_correct']=True
         return dd
     except Exception as e:
         retry = state.get("retry_count", 0) + 1
         logger.error(f"Error running sql: {sql}, error: {e}")
-        return {'retry_count': retry, 'run_sql_error': '运行sql语句失败，请检查语法或联系管理员。'}
+        return {'retry_count': retry, 'run_sql_error': f'运行sql语句失败，{e}请检查语法或联系管理员。'}
 
 
 def _feedback_qa(state: DateReportAgentState) -> dict:
@@ -72,8 +69,10 @@ def _feedback_qa(state: DateReportAgentState) -> dict:
         ddl_list = vn.get_related_ddl(user_question)
         qa_list = vn.get_similar_question_sql(user_question)
         history = state.get('history', [])
+        document = vn.get_related_documentation(user_question)
+        sql_error = state.get('run_sql_error', '')
         sys_promot = feedback_temp['system'].format(question=user_question, sql=sql, sql_result=sql_result,history=history,
-                                                    current_time=datetime.now(),ddl_list=ddl_list,qa_list=qa_list)
+                                                    current_time=datetime.now(),ddl_list=ddl_list,qa_list=qa_list,document=document,sql_error=sql_error)
         logger.info(f"system_temp is {sys_promot}")
         result = gen_history_llm.invoke(sys_promot).text()
         logger.info(f"feedback result: {result}")
@@ -111,12 +110,14 @@ def handle_with_feedback(state: DateReportAgentState) -> str:
     logger.info(f"user:{state.get('user_id', '1')}  ---------------进入 handle_with_feedback 节点 ---------------")
     sql_error = state.get('run_sql_error', '')
     sql_correct = state.get('sql_correct', False)
-    force_check = config('FORCE_CHECK', False)
+    force_check = config('FORCE_CHECK', False,cast=bool)
 
     sql_retry_count = state.get('retry_count', 0)
     logger.info(f"handle with feedback  sql_retry_count is {sql_retry_count} "
                 f"sql error is {sql_error} sql correct is {sql_correct}")
+    logger.info(f"force check is {force_check} type")
     if force_check:
+        logger.info(f"进入 force check")
         if sql_retry_count < 3:
             if sql_correct:
                 return '_gen_report'
@@ -130,6 +131,7 @@ def handle_with_feedback(state: DateReportAgentState) -> str:
     else:
         if sql_retry_count < 3:
             if sql_error and len(sql_error) > 0:
+                logger.info(f"sql_error is {sql_error}")
                 return '_feedback_qa'
             else:
                 return '_gen_report'

template.yaml

@@ -98,7 +98,7 @@ template:
                   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>重点：当用户问题涉及查询:是否,有没有的时候。通过case/when语法结果需要返回：是/否。而不是返回查询记录</rule-detail>
+         <rule-detail>重点：当用户问题涉及查询:是否,有没有的时候。通过case/when语法结果需要返回：是/否。而不是返回查询记录</rule-detail>
         </rule>
         <rule>
           <rule-title>关联与限制</rule-title>
@@ -665,46 +665,79 @@ template:
   result_feedback:
     system: |
       # 角色 
-      你是一位经验丰富的数据分析师和SQL专家。你的核心任务是对SQL查询和其执行结果进行批判性审查。
+      你是一位注重实效的SQL审查专家。你的任务是**识别明显的技术错误**，而不是过度解读业务逻辑。
       # 你的任务
       我（用户）向你提供了一系列信息，请你进行分析和反思。
       # 输入信息
       [原始问题]: <question>{question}</question>
       [生成的SQL]: <sql>{sql}</sql>。
       [执行结果]: <sql_result>{sql_result}</sql_result>。
+      [执行sql报错]:<exec_sql_error>{sql_error}</exec_sql_error>
       [当前时间]: <current_time>{current_time}</current_time>
       [历史信息]: <history>{history}</history>
       [表结构信息]:<schema>{ddl_list}</schema>
       [问答参考]:<question_answer>{qa_list}</question_answer>
+      [注意事项文档]:<document>{document}</document>
+      # 核心审查原则
+        ## 重点关注（必须纠正的问题）
+        1. **语法硬伤**：明显的SQL语法错误、关键字冲突
+        2. **逻辑矛盾**：WHERE条件相互冲突、JOIN条件错误导致笛卡尔积
+        3. **数据安全**：敏感信息泄露、缺少必要的权限限制
+        4. **性能灾难**：缺失关键索引、全表扫描的写法
+        5. **结果明显异常**：数量级错误、数据类型明显不匹配
+        6. **结果错误**：执行sql有报错
+        
+        ## 适度关注（给出建议但不必判定为错误）
+        1. **业务逻辑合理性**：不同的业务理解可能有不同的实现方式
+        2. **模糊匹配精度**：LIKE查询的宽泛程度属于业务选择
+        3. **技术实现选择**：达梦与Oracle的语法兼容性等已验证的技术细节
+        4. **编码规范**：命名风格、格式问题等不影响结果正确性的问题
+        
+        ## 尊重上下文
+        - 信任SQL生成器对业务逻辑的理解
+        - 认可合理的模糊匹配和近似查询
+        - 理解技术选型的合理性
       # 核心规则
       请根据以上信息，进行全面、细致的反思，并最终判断这个结果是否能正确回答原始问题。你的反思需要包含以下几个步骤：
       1.  **核对问题理解**：
           回顾“原始问题”，确认其核心意图、时间范围、筛选条件和所需字段。
           我生成的SQL是否准确捕捉了问题的所有关键要素？有没有遗漏或误解？
           当sql和问题吻合，且能查询出结果时，一般判定为正确
-      2.  **通用业务定义**
-          上周:指完整的**上周一到上周日**
-      3.  **审查SQL逻辑**：
+      2.  **审查SQL逻辑**：
           逐行分析“生成的SQL”，检查其语法结构是否正确。
           关键子句（如 `WHERE`, `GROUP BY`, `HAVING`, `JOIN`）是否准确地反映了查询意图？
           聚合函数（如 `SUM`, `COUNT`, `AVG`）的使用是否恰当？分组维度是否正确？
-      4.  **评估结果合理性**：
+          例如：
+              SQL必须避免与数据库关键字冲突。
+              注意列名定义和使用的先后顺序，例如：SELECT阶段定义了列名，如果GROUP BY阶段先与SELECT阶段执行时，是不许在GROUP BY阶段引用列名的。
+              在ORDER BY、GROUP BY、WHERE子句中不要使用SELECT中定义的别名。
+              注意！当SELECT列中同时包含聚合列（COUNT(), SUM(), AVG()）和非聚合列时，必须要在GROUP BY子句中指定所有非聚合列
+              递归 WITH 子句必须具有列别名列表。
+              CONNECT BY子查询是独立的，无法访问外部查询的表别名,因此涉及CONNECT BY子查询时，里面禁止使用表别名。
+              使用了聚合函数（如 COUNT(), SUM(), AVG()）的SQL，必须配置相应的 GROUP BY 子句。
+      3.  **评估结果合理性**：
           观察“执行结果”，思考这个结果是否符合业务常识或数据的基本特征（例如，数量级、正负值、范围是否合理）？
           结果的列名和内容是否与问题期望的输出一致？
-      5.  **最终判断和建议**：
+      4.  **错误等级划分**
+            - **严重错误**：导致结果完全错误 → 必须纠正
+            - **一般问题**：可能影响部分结果 → 建议优化  
+            - **细节问题**：不影响结果正确性 → 可选优化
+      4.  **最终判断和建议**：
             **结论** : 用一句话明确指出：“结果正确”、“结果可能不正确”或“无法完全确定”。
             **原因** : 简练地解释你做出此判断的核心理由。
             **改进建议** (如果结果不正确):
                          *   具体指出SQL中可能存在的逻辑错误。
                          *   给出一个或多个修改后的SQL版本。
                          *   解释你为什么这样修改。
+    
       # 输出格式
       请严格按照以下JSON格式输出你的分析结果：
       ```json
       {{
         "conclusion": "结果正确 OR 结果可能不正确 OR 无法完全确定",
+        "confidence_level": "高 | 中 | 低",
         "reasoning": "对问题理解、SQL逻辑和结果合理性的综合分析说明。",
-        "is_result_correct": true OR false,
+        "is_result_correct": true | false,
         "suggested_sql": "如果结论为不正确，请在此处提供修改后的SQL。如果正确，则为null。"
       }}
 Resources:

util/q_and_a_dict.py

@@ -513,6 +513,82 @@ question_and_answer = [
                   GROUP BY p."code", p."name"
                   ORDER BY "在林芝工作天数" DESC LIMIT 1000
                   '''
-    }
-
+    },
+    #-----------------------------------------------按月查在藏--最早在藏---------------------------------------------------
+    {
+        "question": "张三最早在藏是哪一天",
+        "answer": '''
+				  SELECT date_value AS earliest_date
+                  FROM YJOA_APPSERVICE_DB."t_yj_person_status" a
+                  LEFT JOIN "YJOA_APPSERVICE_DB"."t_pr3rl2oj_yj_person_database" p 
+                  	   ON p."code" = a."person_id" 
+                  WHERE p."name" = '张三'
+                    AND a."dr" = 0
+                    AND p."dr" =0
+                    AND a."is_in_tibet" = 1
+                  ORDER BY a."date_value"  ASC LIMIT 1
+                  ''',
+        "tags": ["员工", "个人", "考勤", "工作地", "区域", "最早在藏时间"],
+        "category": "工作地考勤统计分析"
+    },
+    {
+        "question": "张三从5月到10月每个月分别在藏多长时间",
+        "answer": '''
+              SELECT TO_CHAR(a."date_value" , 'yyyy-MM') AS month, COUNT(DISTINCT TO_CHAR(a."date_value" , 'yyyy-MM-dd')) AS tibet_days
+              FROM YJOA_APPSERVICE_DB."t_yj_person_status" a
+              LEFT JOIN "YJOA_APPSERVICE_DB"."t_pr3rl2oj_yj_person_database" p 
+              	   ON p."code" = a."person_id" 
+              WHERE p."name" = '张三'
+                AND a."dr" = 0
+                AND p."dr" = 0
+                AND a."date_value" >= '2025-05-01'
+                AND a."date_value" < '2025-11-01'
+                AND a."is_in_tibet" = 1
+              GROUP BY TO_CHAR(a."date_value", 'yyyy-MM')
+              ORDER BY TO_CHAR(a."date_value", 'yyyy-MM') ASC
+              LIMIT 1000
+              ''',
+        "tags": ["员工", "个人", "考勤", "工作地", "区域", "每月在藏天数"],
+        "category": "工作地考勤统计分析"
+    },
+    {
+        "question": "XX中心在藏最长时间的人是谁",
+        "answer": '''
+          SELECT p."name" AS "姓名", p."code" AS "工号", COUNT(ps."id") AS "在藏天数"
+          FROM YJOA_APPSERVICE_DB."t_yj_person_status" ps
+                   JOIN YJOA_APPSERVICE_DB."t_pr3rl2oj_yj_person_database" p ON ps."person_id" = p."code"
+          WHERE ps."is_in_tibet" = 1
+            AND ps."dr" = 0
+            AND p."dr" = 0
+            and p.internal_dept in (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"
+              )
+          GROUP BY p."name", p."code"
+          ORDER BY COUNT (ps."id") DESC LIMIT 1
+          ''',
+        "tags": ["员工", "个人", "考勤", "工作地", "区域", "累计在藏统计"],
+        "category": "工作地考勤统计分析"
+    },
+    {
+        "question": "XX中心的张三在藏多少天了",
+        "answer": '''
+                  SELECT COUNT(*) AS days_in_tibet
+                  FROM YJOA_APPSERVICE_DB."t_yj_person_status" a
+				  LEFT JOIN "YJOA_APPSERVICE_DB"."t_pr3rl2oj_yj_person_database" p 
+				  ON p."code" = a."person_id"
+                  WHERE p."name"  = '张三'
+                  		AND p."internal_dept" IN (
+          						SELECT "id" FROM "IUAP_APDOC_BASEDOC"."org_orgs" 
+          						START WITH ("name" LIKE '%XX中心%' OR "shortname" LIKE '%XX中心%') 
+          						AND "dr" = 0 AND "enable" = 1 AND "code" LIKE '%CYJ%' CONNECT BY PRIOR "id" = "parentid"
+          						) 
+                    	AND a.dr = 0
+                    	AND a."is_in_tibet" = 1
+				 LIMIT 1000
+                  ''',
+        "tags": ["员工", "个人", "考勤", "工作地", "区域", "工作天数"],
+        "category": "工作地考勤统计分析"
+    },
 ]

util/q_and_a_test1.py

@@ -926,82 +926,6 @@ question_and_answer = [
         "tags": ["员工", "个人", "考勤", "工作地", "区域", "工作天数"],
         "category": "工作地考勤统计分析"
     },
-    {
-        "question": "XX中心的张三在藏多少天了",
-        "answer": '''
-                  SELECT COUNT(*) AS days_in_tibet
-                  FROM YJOA_APPSERVICE_DB."t_yj_person_status" a
-				  LEFT JOIN "YJOA_APPSERVICE_DB"."t_pr3rl2oj_yj_person_database" p 
-				  ON p."code" = a."person_id"
-                  WHERE p."name"  = '张三'
-                  		AND p."internal_dept" IN (
-          						SELECT "id" FROM "IUAP_APDOC_BASEDOC"."org_orgs" 
-          						START WITH ("name" LIKE '%XX中心%' OR "shortname" LIKE '%XX中心%') 
-          						AND "dr" = 0 AND "enable" = 1 AND "code" LIKE '%CYJ%' CONNECT BY PRIOR "id" = "parentid"
-          						) 
-                    	AND a.dr = 0
-                    	AND a."is_in_tibet" = 1
-				 LIMIT 1000
-                  ''',
-        "tags": ["员工", "个人", "考勤", "工作地", "区域", "工作天数"],
-        "category": "工作地考勤统计分析"
-    },
-    {
-        "question": "张三最早在藏是哪一天",
-        "answer": '''
-				  SELECT date_value AS earliest_date
-                  FROM YJOA_APPSERVICE_DB."t_yj_person_status" a
-                  LEFT JOIN "YJOA_APPSERVICE_DB"."t_pr3rl2oj_yj_person_database" p 
-                  	   ON p."code" = a."person_id" 
-                  WHERE p."name" = '张三'
-                    AND a."dr" = 0
-                    AND p."dr" =0
-                    AND a."is_in_tibet" = 1
-                  ORDER BY a."date_value"  ASC LIMIT 1
-                  ''',
-        "tags": ["员工", "个人", "考勤", "工作地", "区域", "最早在藏时间"],
-        "category": "工作地考勤统计分析"
-    },
-    {
-        "question": "XX中心在藏最长时间的人是谁",
-        "answer": '''
-              SELECT p."name" AS "姓名", p."code" AS "工号", COUNT(ps."id") AS "在藏天数"
-              FROM YJOA_APPSERVICE_DB."t_yj_person_status" ps
-                       JOIN YJOA_APPSERVICE_DB."t_pr3rl2oj_yj_person_database" p ON ps."person_id" = p."code"
-              WHERE ps."is_in_tibet" = 1
-                AND ps."dr" = 0
-                AND p."dr" = 0
-                and p.internal_dept in (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"
-                  )
-              GROUP BY p."name", p."code"
-              ORDER BY COUNT (ps."id") DESC LIMIT 1
-              ''',
-        "tags": ["员工", "个人", "考勤", "工作地", "区域", "累计在藏统计"],
-        "category": "工作地考勤统计分析"
-    },
-    {
-        "question": "张三从5月到10月每个月分别在藏多长时间",
-        "answer": '''
-                  SELECT TO_CHAR(a."date_value" , 'yyyy-MM') AS month, COUNT(DISTINCT TO_CHAR(a."date_value" , 'yyyy-MM-dd')) AS tibet_days
-                  FROM YJOA_APPSERVICE_DB."t_yj_person_status" a
-                  LEFT JOIN "YJOA_APPSERVICE_DB"."t_pr3rl2oj_yj_person_database" p 
-                  	   ON p."code" = a."person_id" 
-                  WHERE p."name" = '张三'
-                    AND a."dr" = 0
-                    AND p."dr" = 0
-                    AND a."date_value" >= '2025-05-01'
-                    AND a."date_value" < '2025-11-01'
-                    AND a."is_in_tibet" = 1
-                  GROUP BY TO_CHAR(a."date_value", 'yyyy-MM')
-                  ORDER BY TO_CHAR(a."date_value", 'yyyy-MM') ASC
-                  LIMIT 1000
-                  ''',
-        "tags": ["员工", "个人", "考勤", "工作地", "区域", "每月在藏天数"],
-        "category": "工作地考勤统计分析"
-    },
     {
         "question": "XX中心在林芝最长时间的是谁，待了多少天",
         "answer": '''
@@ -1218,6 +1142,181 @@ question_and_answer = [
                   ''',
         "tags": ["员工", "个人", "考勤", "工作地", "区域", "工作天数"],
         "category": "工作地考勤统计分析"
-    }
+    },
+    {
+        "question": "XX中心在藏最长时间的人是谁",
+        "answer": '''
+                SELECT  p."code" AS "人员编号",SUM(pt."count_in_tibat") AS "累计在藏天数", p."name" AS "姓名", o."name" AS "部门"
+                FROM "YJOA_APPSERVICE_DB"."t_pr3rl2oj_yj_person_database" p
+                INNER JOIN  "YJOA_APPSERVICE_DB"."t_yj_person_in_tibat" pt
+                ON p."code" = pt."person_id"  
+                INNER JOIN "IUAP_APDOC_BASEDOC"."org_orgs" o
+                ON p."internal_dept" = o."id"
+                WHERE p."dr" = 0
+                      AND pt."dr" = 0
+                      AND p."internal_dept" IN (
+                                 SELECT id
+                                 FROM "IUAP_APDOC_BASEDOC"."org_orgs" 
+                                 START WITH ("name" LIKE '%XX中心%' OR "shortname" LIKE '%XX中心%')
+                                       AND "dr" = 0 AND "enable" = 1 AND "code" LIKE '%CYJ%'
+                                 CONNECT BY PRIOR "id" = "parentid" 
+                      )
+                GROUP BY p."code", p."name", o."name"
+                ORDER BY SUM(pt."count_in_tibat") DESC
+                LIMIT 1
+      ''',
+        "tags": ["员工", "考勤", "工作地", "区域", "累计在藏统计","最值"],
+        "category": "工作地考勤统计分析"
+    },
+    {
+        "question": "XX中心的张三在藏多少天了",
+        "answer": '''
+            SELECT  p."code" AS "人员编号",SUM(pt."count_in_tibat") AS "累计在藏天数", p."name" AS "姓名", o."name" AS "部门"
+            FROM "YJOA_APPSERVICE_DB"."t_pr3rl2oj_yj_person_database" p
+            INNER JOIN  "YJOA_APPSERVICE_DB"."t_yj_person_in_tibat" pt
+            ON p."code" = pt."person_id"  
+            INNER JOIN "IUAP_APDOC_BASEDOC"."org_orgs" o
+            ON p."internal_dept" = o."id"
+            WHERE p."dr" = 0
+                  AND pt."dr" = 0
+                  AND p."name" = '张三'
+                  AND p."internal_dept" IN (
+                             SELECT id
+                             FROM "IUAP_APDOC_BASEDOC"."org_orgs" 
+                             START WITH ("name" LIKE '%XX中心%' OR "shortname" LIKE '%XX中心%')
+                                   AND "dr" = 0 AND "enable" = 1 AND "code" LIKE '%CYJ%'
+                             CONNECT BY PRIOR "id" = "parentid" 
+                  )
+            GROUP BY p."code", p."name", o."name"
+            ORDER BY SUM(pt."count_in_tibat") DESC
+              ''',
+        "tags": ["员工", "个人", "考勤", "工作地", "区域", "累计在藏天数"],
+        "category": "工作地考勤统计分析"
+    },
+    {
+        "question": "XX中心的张三2025年在藏多少天了",
+        "answer": '''
+                SELECT pt."year" AS "年份", p."code" AS "人员编号",pt."count_in_tibat" AS "2025年累计在藏天数", p."name" AS "姓名", o."name" AS "部门"
+                FROM "YJOA_APPSERVICE_DB"."t_pr3rl2oj_yj_person_database" p
+                INNER JOIN  "YJOA_APPSERVICE_DB"."t_yj_person_in_tibat" pt
+                ON p."code" = pt."person_id"  
+                INNER JOIN "IUAP_APDOC_BASEDOC"."org_orgs" o
+                ON p."internal_dept" = o."id"
+                WHERE p."dr" = 0
+                      AND pt."dr" = 0
+                      AND p."name" = '张三'
+                      AND pt."year" = '2025'
+                      AND p."internal_dept" IN (
+                                 SELECT id
+                                 FROM "IUAP_APDOC_BASEDOC"."org_orgs" 
+                                 START WITH ("name" LIKE '%XX中心%' OR "shortname" LIKE '%XX中心%')
+                                       AND "dr" = 0 AND "enable" = 1 AND "code" LIKE '%CYJ%'
+                                 CONNECT BY PRIOR "id" = "parentid" 
+                      )
+          ''',
+        "tags": ["员工", "个人", "考勤", "工作地", "区域", "累计在藏天数"],
+        "category": "工作地考勤统计分析"
+    },
+    {
+        "question": "XX中心的张三上一次累计在藏多少天",
+        "answer": '''
+                    SELECT  pt."year" AS "年份",p."code" AS "人员编号",pt."last_count_in_tibat" AS "上一次累计在藏天数", p."name" AS "姓名", o."name" AS "部门"
+                    FROM "YJOA_APPSERVICE_DB"."t_pr3rl2oj_yj_person_database" p
+                    INNER JOIN  "YJOA_APPSERVICE_DB"."t_yj_person_in_tibat" pt
+                    ON p."code" = pt."person_id"  
+                    INNER JOIN "IUAP_APDOC_BASEDOC"."org_orgs" o
+                    ON p."internal_dept" = o."id"
+                    WHERE p."dr" = 0
+                          AND pt."dr" = 0
+                          AND p."name" = '张三'
+                          AND p."internal_dept" IN (
+                                     SELECT id
+                                     FROM "IUAP_APDOC_BASEDOC"."org_orgs" 
+                                     START WITH ("name" LIKE '%XX中心%' OR "shortname" LIKE '%XX中心%')
+                                           AND "dr" = 0 AND "enable" = 1 AND "code" LIKE '%CYJ%'
+                                     CONNECT BY PRIOR "id" = "parentid" 
+                          )
+                    ORDER BY pt."year" DESC
+                    LIMIT 1
+          ''',
+        "tags": ["员工", "个人", "考勤", "工作地", "区域", "上一次累计在藏天数"],
+        "category": "工作地考勤统计分析"
+    },
+    {
+        "question": "XX中心连续在藏最长时间的人是谁",
+        "answer": '''
+                SELECT  pt."year" AS "年份",p."code" AS "人员编号",pt."continuous_in_tibet_days"  AS "连续在藏天数", p."name" AS "姓名", o."name" AS "部门"
+                FROM "YJOA_APPSERVICE_DB"."t_pr3rl2oj_yj_person_database" p
+                INNER JOIN  "YJOA_APPSERVICE_DB"."t_yj_person_in_tibat" pt
+                ON p."code" = pt."person_id"  
+                INNER JOIN "IUAP_APDOC_BASEDOC"."org_orgs" o
+                ON p."internal_dept" = o."id"
+                WHERE p."dr" = 0
+                      AND pt."dr" = 0
+                      AND p."internal_dept" IN (
+                                 SELECT id
+                                 FROM "IUAP_APDOC_BASEDOC"."org_orgs" 
+                                 START WITH ("name" LIKE '%XX中心%' OR "shortname" LIKE '%XX中心%')
+                                       AND "dr" = 0 AND "enable" = 1 AND "code" LIKE '%CYJ%'
+                                 CONNECT BY PRIOR "id" = "parentid" 
+                      )
+                ORDER BY pt."year",pt."continuous_in_tibet_days" DESC
+                LIMIT 1
+        ''',
+        "tags": ["员工", "考勤", "工作地", "区域", "连续在藏统计", "最值"],
+        "category": "工作地考勤统计分析"
+    },
+    {
+        "question": "XX中心2025年连续在藏最长时间的人是谁",
+        "answer": '''
+            SELECT  pt."year" AS "年份",p."code" AS "人员编号",pt."continuous_in_tibet_days"  AS "连续在藏天数", p."name" AS "姓名", o."name" AS "部门"
+            FROM "YJOA_APPSERVICE_DB"."t_pr3rl2oj_yj_person_database" p
+            INNER JOIN  "YJOA_APPSERVICE_DB"."t_yj_person_in_tibat" pt
+            ON p."code" = pt."person_id"  
+            INNER JOIN "IUAP_APDOC_BASEDOC"."org_orgs" o
+            ON p."internal_dept" = o."id"
+            WHERE p."dr" = 0
+                  AND pt."dr" = 0
+                  AND pt."year" = '2025'
+                  AND p."internal_dept" IN (
+                             SELECT id
+                             FROM "IUAP_APDOC_BASEDOC"."org_orgs" 
+                             START WITH ("name" LIKE '%XX中心%' OR "shortname" LIKE '%XX中心%')
+                                   AND "dr" = 0 AND "enable" = 1 AND "code" LIKE '%CYJ%'
+                             CONNECT BY PRIOR "id" = "parentid" 
+                  )
+            ORDER BY pt."continuous_in_tibet_days" DESC
+            LIMIT 1
+    ''',
+        "tags": ["员工", "考勤", "工作地", "区域", "连续在藏统计", "最值"],
+        "category": "工作地考勤统计分析"
+    },
+
+    {
+        "question": "XX中心的张三连续在藏多少天了",
+        "answer": '''
+                    SELECT  pt."year" AS "年份",p."code" AS "人员编号",pt."continuous_in_tibet_days"  AS "连续在藏天数", p."name" AS "姓名", o."name" AS "部门"
+                    FROM "YJOA_APPSERVICE_DB"."t_pr3rl2oj_yj_person_database" p
+                    INNER JOIN  "YJOA_APPSERVICE_DB"."t_yj_person_in_tibat" pt
+                    	  ON p."code" = pt."person_id"  
+                    INNER JOIN "IUAP_APDOC_BASEDOC"."org_orgs" o
+                    	  ON p."internal_dept" = o."id"
+                    WHERE p."dr" = 0
+                          AND pt."dr" = 0
+                          AND p."name" = '张三'
+                          AND p."internal_dept" IN (
+                                     SELECT id
+                                     FROM "IUAP_APDOC_BASEDOC"."org_orgs" 
+                                     START WITH ("name" LIKE '%XX中心%' OR "shortname" LIKE '%XX中心%')
+                                           AND "dr" = 0 AND "enable" = 1 AND "code" LIKE '%CYJ%'
+                                     CONNECT BY PRIOR "id" = "parentid" 
+                          )
+                    ORDER BY pt."year",pt."continuous_in_tibet_days" DESC
+                    LIMIT 1
+              ''',
+        "tags": ["员工", "个人", "考勤", "工作地", "区域", "连续在藏天数"],
+        "category": "工作地考勤统计分析"
+    },
+
 ]
 

util/train_ddl.py

@@ -12,6 +12,7 @@ train_document='''
                 查询部门信息时尽量使用internal_dept而非internal_unit<UNK>
                 数信部不是数信中心，两者不能等价
                 数信中心就叫数信中心，没有数字信息中心这个部门，请勿胡乱替换
+                部门表里code字段中中带CYJ的才是正式加入使用的部门，废弃或测试部门的部门编号则不带CYJ
         '''
 
 person_database_ddl = """
@@ -571,13 +572,7 @@ person_status_ddl='''
       "role": "dimension",
       "tags": ["休假信息", "请假类型", "枚举", "考勤管理"]
     },
-    {
-      "name": "is_in_tibet",
-      "type": "INT",
-      "comment": "是否在西藏地区",
-      "role": "dimension",
-      "tags": ["地区标识", "地理位置"]
-    },
+
     {
       "name": "business_trip_destination",
       "type": "VARCHAR(50)",
@@ -897,57 +892,57 @@ person_ac_area = '''{
 person_in_tibat = '''
 {
   "db_name": "YJOA_APPSERVICE_DB",
-  "table_name": "person_in_tibat",
-  "table_comment": "人员在藏情况表",
+  "table_name": "t_yj_person_in_tibat",
+  "table_comment": "人员累计在藏天数、连续在藏天数，情况表",
   "columns": [
     {
       "name": "id",
       "type": "VARCHAR(50)",
       "comment": "主键ID",
       "role": "dimension",
-      "tags": ["唯一标识"]
+      "tags": ["唯一标识", "主键"]
     },
     {
       "name": "person_id",
       "type": "VARCHAR(50)",
-      "comment": "人员ID",
+      "comment": "人员唯一标识",
       "role": "dimension",
-      "tags": ["人员标识"]
+      "tags": ["人员标识", "外键"]
     },
     {
       "name": "year",
       "type": "VARCHAR(20)",
-      "comment": "年份",
+      "comment": "统计年份",
       "role": "dimension",
-      "tags": ["时间维度"]
+      "tags": ["时间维度", "统计周期"]
     },
     {
-      "name": "create_time",
-      "type": "DATETIME",
-      "comment": "创建时间",
-      "role": "dimension",
-      "tags": ["时间信息"]
-    },
-    {
-      "name": "update_time",
-      "type": "DATETIME",
-      "comment": "更新时间",
-      "role": "dimension",
-      "tags": ["时间信息"]
-    },
-    {
-      "name": "dr",
+      "name": "count_in_tibat",
       "type": "INT",
-      "comment": "删除标记",
-      "role": "dimension",
-      "tags": ["数据状态"]
+      "comment": "该年累计在藏天数",
+      "role": "metric",
+      "tags": ["在藏统计", "累计天数","当年累计天数"]
+    },
+    {
+      "name": "last_count_in_tibat",
+      "type": "INT",
+      "comment": "上一年度累计在藏天数",
+      "role": "metric", 
+      "tags": ["历史统计", "同比数据","上次累计在藏"]
     },
     {
       "name": "continuous_in_tibet_days",
       "type": "INT",
       "comment": "连续在藏天数",
       "role": "metric",
-      "tags": ["连续在藏天数"]
+      "tags": ["连续统计", "持续天数"]
+    },
+    {
+      "name": "dr",
+      "type": "INT",
+      "comment": "软删除标记(0:正常, 1:删除)",
+      "role": "dimension",
+      "tags": ["数据状态", "系统字段"]
     }
   ],
     "relationships": [