auto-sync: vala-component-oops-stat 2026-04-13_14:34

SKILL.md

@@ -0,0 +1,45 @@
+---
+name: vala-component-oops-stat
+description: 瓦拉英语组件练习Oops率统计工具，支持按日期统计、自动剔除测试账号、分L1/L2等级、自定义样本量规则、导出Excel报表。使用场景：(1) 统计组件练习错误率/Oops率 (2) 按难度等级拆分L1(A1)/L2(A2)统计结果 (3) 练习次数不足10次自动补充历史记录 (4) 导出多sheet Excel报表 (5) 按Oops率降序排序展示高错误率组件
+---
+
+# 瓦拉英语组件Oops率统计技能
+
+## 功能说明
+用于统计飞书多维表格中组件练习的Oops（错误）率，支持灵活配置统计规则，自动处理数据口径问题，输出标准化统计结果。
+
+## 核心特性
+✅ 自动剔除测试账号：仅保留`bi_vala_app_account.status = 1`的正常用户练习记录
+✅ 等级拆分：自动按`level`字段拆分L1(A1)/L2(A2)两个难度等级分别统计
+✅ Oops判定规则：
+  - 练习结果为Oops/Opps → 记为Oops
+  - 练习结果为pass且c_type包含`core_`或`scence_` → 记为Oops
+  - 其余pass结果 → 记为Good，不计入Oops
+✅ 样本量规则：昨日练习≥10次用昨日全量数据，<10次自动补充历史记录至10次，历史不足10次取全部记录
+✅ 正确排序：按Oops率数值从高到低排序，避免字符串排序错误
+✅ 自动导出：生成包含两个sheet的Excel报表，直接发送给用户
+
+## 使用参数
+| 参数 | 说明 | 默认值 |
+|------|------|--------|
+| 统计日期 | 要统计的日期（格式YYYY-MM-DD） | 昨日（当前日期-1天） |
+| 最小样本量 | 组件最少统计样本量 | 10 |
+| 是否剔测试账号 | 是否排除测试账号练习记录 | 是 |
+| 是否分等级 | 是否拆分L1/L2分别统计 | 是 |
+| 导出格式 | 输出格式（CSV/Excel） | Excel |
+
+## 操作步骤
+1. 确认用户统计需求：统计日期、样本量规则、是否分等级等
+2. 执行对应等级的统计SQL脚本（`scripts/stat_l1.sql`、`scripts/stat_l2.sql`）
+3. 运行`scripts/generate_excel.py`生成Excel报表
+4. 将报表通过飞书发送给用户
+
+## 脚本说明
+### scripts/stat_l1.sql
+统计L1(A1)等级组件Oops率的SQL脚本，可修改日期参数调整统计时间
+### scripts/stat_l2.sql
+统计L2(A2)等级组件Oops率的SQL脚本，可修改日期参数调整统计时间
+### scripts/generate_excel.py
+将CSV统计结果合并生成带多sheet的Excel报表
+### references/table_schema.md
+相关数据表结构说明和字段含义参考

references/table_schema.md

@@ -0,0 +1,26 @@
+# 相关数据表结构说明
+
+## bi_user_component_play_record_* 分表（组件练习记录表）
+| 字段名 | 类型 | 说明 |
+|-------|------|------|
+| user_id | bigint | 角色ID，关联bi_vala_app_character.id |
+| component_unique_code | varchar | 组件唯一编码，业务系统中组件的唯一标识 |
+| play_result | varchar | 练习结果：Perfect/Good/Oops/Opps/pass |
+| c_type | varchar | 组件类型，包含core_前缀为核心题型，scence_前缀为场景题型 |
+| created_at | timestamp | 练习时间 |
+| level | varchar | 难度等级：A1(L1)/A2(L2) |
+| deleted_at | timestamp | 删除时间，为空表示记录有效 |
+
+## bi_vala_app_character（角色表）
+| 字段名 | 类型 | 说明 |
+|-------|------|------|
+| id | bigint | 角色ID |
+| account_id | bigint | 账号ID，关联bi_vala_app_account.id |
+| deleted_at | timestamp | 删除时间，为空表示记录有效 |
+
+## bi_vala_app_account（账号表）
+| 字段名 | 类型 | 说明 |
+|-------|------|------|
+| id | bigint | 账号ID |
+| status | int | 账号状态：1=正常用户，其他=测试账号/禁用账号 |
+| deleted_at | timestamp | 删除时间，为空表示记录有效 |

scripts/generate_excel.py

@@ -0,0 +1,34 @@
+import pandas as pd
+import numpy as np
+import sys
+
+# 兼容numpy版本
+try:
+    np._get_promotion_state = lambda *args, **kwargs: 0
+except:
+    pass
+
+def generate_excel(l1_csv_path, l2_csv_path, output_path):
+    """
+    生成包含L1和L2两个sheet的Excel报表
+    :param l1_csv_path: L1等级统计结果CSV路径
+    :param l2_csv_path: L2等级统计结果CSV路径
+    :param output_path: 输出Excel文件路径
+    """
+    # 读取CSV文件
+    df_l1 = pd.read_csv(l1_csv_path)
+    df_l2 = pd.read_csv(l2_csv_path)
+
+    # 创建Excel文件
+    with pd.ExcelWriter(output_path) as writer:
+        df_l1.to_excel(writer, sheet_name='L1等级组件', index=False)
+        df_l2.to_excel(writer, sheet_name='L2等级组件', index=False)
+
+    print(f"Excel报表生成成功：{output_path}")
+
+if __name__ == "__main__":
+    if len(sys.argv) != 4:
+        print("用法：python generate_excel.py <l1_csv路径> <l2_csv路径> <输出Excel路径>")
+        sys.exit(1)
+    
+    generate_excel(sys.argv[1], sys.argv[2], sys.argv[3])

scripts/stat_l1.sql

@@ -0,0 +1,90 @@
+-- 统计L1(A1)等级组件Oops率
+-- 修改统计日期时替换下面的'2026-04-12'为目标日期
+WITH all_component_records AS (
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_0 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_1 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_2 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_3 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_4 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_5 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_6 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_7 WHERE deleted_at IS NULL
+),
+valid_records AS (
+    SELECT 
+        cr.component_unique_code,
+        cr.play_result,
+        cr.created_at,
+        cr.c_type,
+        CASE WHEN DATE(cr.created_at) = '2026-04-12' THEN 1 ELSE 0 END AS is_yesterday
+    FROM all_component_records cr
+    JOIN bi_vala_app_character c ON cr.user_id = c.id AND c.deleted_at IS NULL
+    JOIN bi_vala_app_account a ON c.account_id = a.id AND a.status = 1 AND a.deleted_at IS NULL
+    WHERE cr.play_result IS NOT NULL AND cr.component_unique_code IS NOT NULL AND cr.level = 'A1'
+),
+-- 统计每个组件昨日练习次数
+yesterday_counts AS (
+    SELECT 
+        component_unique_code,
+        COUNT(*) AS yesterday_cnt
+    FROM valid_records
+    WHERE is_yesterday = 1
+    GROUP BY component_unique_code
+),
+-- 给记录排序：昨日记录排最前，历史记录按时间倒序
+ranked_records AS (
+    SELECT 
+        vr.component_unique_code,
+        vr.play_result,
+        vr.c_type,
+        vr.is_yesterday,
+        ROW_NUMBER() OVER (
+            PARTITION BY vr.component_unique_code 
+            ORDER BY vr.is_yesterday DESC, vr.created_at DESC
+        ) AS rn
+    FROM valid_records vr
+    JOIN yesterday_counts yc ON vr.component_unique_code = yc.component_unique_code
+),
+-- 筛选统计样本：
+-- 昨日练习≥10次：取全部昨日记录
+-- 昨日练习<10次：取全部昨日记录 + 最近历史记录补到10次
+filtered_samples AS (
+    SELECT 
+        r.component_unique_code,
+        r.play_result,
+        r.c_type
+    FROM ranked_records r
+    JOIN yesterday_counts yc ON r.component_unique_code = yc.component_unique_code
+    WHERE 
+        (yc.yesterday_cnt >= 10 AND r.is_yesterday = 1)
+        OR 
+        (yc.yesterday_cnt < 10 AND r.rn <= 10)
+),
+-- 统计结果，先按数值排序
+component_stats AS (
+    SELECT 
+        component_unique_code AS 组件唯一编码,
+        COUNT(*) AS 总练习次数,
+        SUM(CASE 
+            WHEN play_result IN ('Oops', 'Opps') THEN 1
+            WHEN play_result = 'pass' AND (c_type LIKE '%core_%' OR c_type LIKE '%scence_%') THEN 1
+            ELSE 0 
+        END) AS Oops次数,
+        ROUND(CASE WHEN COUNT(*) = 0 THEN 0 ELSE 
+            SUM(CASE 
+                WHEN play_result IN ('Oops', 'Opps') THEN 1
+                WHEN play_result = 'pass' AND (c_type LIKE '%core_%' OR c_type LIKE '%scence_%') THEN 1
+                ELSE 0 
+            END)::DECIMAL / COUNT(*) * 100 END, 2) AS Oops率数值
+    FROM filtered_samples
+    GROUP BY component_unique_code
+    ORDER BY Oops率数值 DESC
+)
+SELECT 组件唯一编码, 总练习次数, Oops次数, Oops率数值 || '%' AS Oops率 FROM component_stats;

scripts/stat_l2.sql

@@ -0,0 +1,90 @@
+-- 统计L2(A2)等级组件Oops率
+-- 修改统计日期时替换下面的'2026-04-12'为目标日期
+WITH all_component_records AS (
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_0 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_1 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_2 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_3 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_4 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_5 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_6 WHERE deleted_at IS NULL
+    UNION ALL
+    SELECT user_id, component_unique_code, play_result, created_at, level, c_type FROM bi_user_component_play_record_7 WHERE deleted_at IS NULL
+),
+valid_records AS (
+    SELECT 
+        cr.component_unique_code,
+        cr.play_result,
+        cr.created_at,
+        cr.c_type,
+        CASE WHEN DATE(cr.created_at) = '2026-04-12' THEN 1 ELSE 0 END AS is_yesterday
+    FROM all_component_records cr
+    JOIN bi_vala_app_character c ON cr.user_id = c.id AND c.deleted_at IS NULL
+    JOIN bi_vala_app_account a ON c.account_id = a.id AND a.status = 1 AND a.deleted_at IS NULL
+    WHERE cr.play_result IS NOT NULL AND cr.component_unique_code IS NOT NULL AND cr.level = 'A2'
+),
+-- 统计每个组件昨日练习次数
+yesterday_counts AS (
+    SELECT 
+        component_unique_code,
+        COUNT(*) AS yesterday_cnt
+    FROM valid_records
+    WHERE is_yesterday = 1
+    GROUP BY component_unique_code
+),
+-- 给记录排序：昨日记录排最前，历史记录按时间倒序
+ranked_records AS (
+    SELECT 
+        vr.component_unique_code,
+        vr.play_result,
+        vr.c_type,
+        vr.is_yesterday,
+        ROW_NUMBER() OVER (
+            PARTITION BY vr.component_unique_code 
+            ORDER BY vr.is_yesterday DESC, vr.created_at DESC
+        ) AS rn
+    FROM valid_records vr
+    JOIN yesterday_counts yc ON vr.component_unique_code = yc.component_unique_code
+),
+-- 筛选统计样本：
+-- 昨日练习≥10次：取全部昨日记录
+-- 昨日练习<10次：取全部昨日记录 + 最近历史记录补到10次
+filtered_samples AS (
+    SELECT 
+        r.component_unique_code,
+        r.play_result,
+        r.c_type
+    FROM ranked_records r
+    JOIN yesterday_counts yc ON r.component_unique_code = yc.component_unique_code
+    WHERE 
+        (yc.yesterday_cnt >= 10 AND r.is_yesterday = 1)
+        OR 
+        (yc.yesterday_cnt < 10 AND r.rn <= 10)
+),
+-- 统计结果，先按数值排序
+component_stats AS (
+    SELECT 
+        component_unique_code AS 组件唯一编码,
+        COUNT(*) AS 总练习次数,
+        SUM(CASE 
+            WHEN play_result IN ('Oops', 'Opps') THEN 1
+            WHEN play_result = 'pass' AND (c_type LIKE '%core_%' OR c_type LIKE '%scence_%') THEN 1
+            ELSE 0 
+        END) AS Oops次数,
+        ROUND(CASE WHEN COUNT(*) = 0 THEN 0 ELSE 
+            SUM(CASE 
+                WHEN play_result IN ('Oops', 'Opps') THEN 1
+                WHEN play_result = 'pass' AND (c_type LIKE '%core_%' OR c_type LIKE '%scence_%') THEN 1
+                ELSE 0 
+            END)::DECIMAL / COUNT(*) * 100 END, 2) AS Oops率数值
+    FROM filtered_samples
+    GROUP BY component_unique_code
+    ORDER BY Oops率数值 DESC
+)
+SELECT 组件唯一编码, 总练习次数, Oops次数, Oops率数值 || '%' AS Oops率 FROM component_stats;