@clickzetta/cz-cli-darwin-x64 0.3.81 → 0.3.83

package/bin/skills/clickzetta-data-science/SKILL.md

@@ -0,0 +1,125 @@
+---
+name: clickzetta-data-science
+description: |
+  数据科学家使用 ClickZetta Lakehouse 的端到端工作流指南。按工作阶段组织：
+  开发环境准备（Python 3.10+ 检查/搭建）、Jupyter Notebook 配置与使用、
+  项目结构规范（Cookiecutter DS 标准）、数据发现、数据质量评估、
+  数据清洗与整合、数据集构建、EDA 探索分析、
+  特征工程（SQL + ZettaPark）、模型推理上线（BITMAP 用户画像/UDF 批量推理/向量检索）。
+  当用户说"数据科学"、"机器学习"、"特征工程"、"EDA"、"数据探索"、
+  "ZettaPark 机器学习"、"Jupyter 连接 Lakehouse"、"notebook"、"ipynb"、
+  "jupyter kernel"、"%%sql"、"magic command"、"pandas 读取数据"、
+  "数据质量检查"、"数据采样"、"TABLESAMPLE"、"approx_percentile"、
+  "BITMAP 用户画像"、"人群圈选"、"批量推理"、"Python 3.10"、
+  "scikit-learn"、"项目目录结构"、"config.json"、".env"时触发。
+  Keywords: data science, Jupyter, EDA, feature engineering, ML, pandas, notebook
+---
+
+# ClickZetta Lakehouse 数据科学工作流
+
+## 工作流全景
+
+```
+环境准备 → Jupyter 配置 → 项目结构 → 数据发现 → 数据质量评估 → 数据清洗整合
+                                                                        ↓
+                                      模型推理上线 ← 特征工程 ← EDA ← 数据集构建
+```
+
+---
+
+## 硬性前提条件
+
+**Python 3.10+**（ZettaPark 硬性要求）。用户环境是 3.9 或更低时，先给升级方案再继续：
+
+```bash
+brew install pyenv && pyenv install 3.12.9 && pyenv local 3.12.9
+python -m venv .venv && source .venv/bin/activate
+```
+
+详细搭建步骤见 [references/setup.md](references/setup.md)。
+
+---
+
+## 项目结构
+
+```
+my-ds-project/
+├── notebooks/          # 00-env-check.ipynb 必须是第一个
+│   ├── 00-env-check.ipynb
+│   ├── 01-data-discovery.ipynb
+│   ├── 02-data-quality.ipynb
+│   ├── 03-eda.ipynb
+│   ├── 04-feature-engineering.ipynb
+│   └── 05-modeling.ipynb
+├── src/
+│   ├── config.py       # 连接配置，见 references/setup.md
+│   ├── data/
+│   └── features/
+├── sql/
+├── data/               # 全部加入 .gitignore
+├── models/             # 全部加入 .gitignore
+├── .env                # 绝不入 git
+└── .env.example        # 入 git
+```
+
+环境变量命名规范：`CLICKZETTA_SERVICE` / `CLICKZETTA_INSTANCE` / `CLICKZETTA_WORKSPACE` / `CLICKZETTA_USERNAME` / `CLICKZETTA_PASSWORD` / `CLICKZETTA_VCLUSTER` / `CLICKZETTA_SCHEMA`。
+
+---
+
+## 数据写入规则（禁止事项）
+
+| 方式 | 结论 |
+|------|------|
+| `session.create_dataframe(df).write.save_as_table()` | ✅ 推荐 |
+| `cursor` 批量 INSERT（每批 500 行） | ✅ Python 3.9 / ZettaPark 不可用时的 fallback |
+| `df.to_sql(conn, ...)` | ❌ 禁止，报 `'list' object has no attribute 'keys'` |
+| SQLAlchemy `clickzetta://...` | ❌ 禁止，dialect 不可靠 |
+
+代码模板见 [references/write-and-infer.md](references/write-and-infer.md)。
+
+---
+
+## 数据查看规则
+
+- 快速查看用 `.show()`，不需要 pandas 时不要 `.to_pandas()`
+- 大表操作默认加 `TABLESAMPLE ROW(10)` 采样，避免 OOM
+
+---
+
+## 数据验证规则
+
+导入数据后，**立即用已知基准值验证统计结果**，再进行后续分析。
+
+常见陷阱：运动员/用户级别的原始数据，团体项目每个参与者各有一条记录，直接 SUM 会重复计算。正确做法：先 `SELECT DISTINCT event, medal, ...` 去重，再聚合。
+
+---
+
+## ClickZetta SQL 不支持的语法
+
+| 不支持 | 替代方案 |
+|--------|---------|
+| `CREATE OR REPLACE TABLE` | `CREATE TABLE IF NOT EXISTS`（普通表不支持 OR REPLACE） |
+| `ARRAY_AGG(col IGNORE NULLS)` | `MAX(col)` 或 `COALESCE()` |
+| `QUALIFY` 子句 | 子查询 + `WHERE rn = 1` |
+| `UNION` / `INTERSECT` / `EXCEPT` | JOIN + 应用层合并 |
+| `BEGIN; COMMIT; ROLLBACK;` | 用 MERGE 实现原子操作 |
+| `NOW()` | `CURRENT_TIMESTAMP()` |
+
+遇到其他语法报错，加载 `clickzetta-sql-syntax-guide` skill。
+
+---
+
+## Schema 上下文
+
+Python 代码中 SQL 语句始终使用完整表名 `schema.table`，不依赖当前 schema 上下文。
+
+---
+
+## 参考文档
+
+- [环境搭建与项目配置](references/setup.md) — 环境搭建、config.py 模板、Jupyter 配置
+- [数据发现/质量/清洗/EDA 示例](references/data-patterns.md)
+- [数据写入/特征工程/模型推理示例](references/write-and-infer.md)
+- [ZettaPark API](references/zettapark-api.md)
+- [统计分析函数](references/stats-functions.md)
+- [BITMAP 用户画像](references/bitmap-profile.md)

package/bin/skills/clickzetta-data-science/eval_cases.jsonl

@@ -0,0 +1,12 @@
+{"case_id":"001","type":"should_call","user_input":"怎么用 Jupyter Notebook 连接 Lakehouse","expected_skill":"clickzetta-data-science","expected_output_contains":["Jupyter"]}
+{"case_id":"002","type":"should_call","user_input":"数据科学项目目录结构怎么组织","expected_skill":"clickzetta-data-science","expected_output_contains":["项目结构"]}
+{"case_id":"003","type":"should_call","user_input":"怎么做 EDA 探索性数据分析","expected_skill":"clickzetta-data-science","expected_output_contains":["EDA"]}
+{"case_id":"004","type":"should_call","user_input":"TABLESAMPLE 怎么做数据采样","expected_skill":"clickzetta-data-science","expected_output_contains":["TABLESAMPLE"]}
+{"case_id":"005","type":"should_call","user_input":"怎么用 ZettaPark 做特征工程","expected_skill":"clickzetta-data-science","expected_output_contains":["特征工程"]}
+{"case_id":"006","type":"should_call","user_input":"BITMAP 用户画像怎么做人群圈选","expected_skill":"clickzetta-data-science","expected_output_contains":["BITMAP"]}
+{"case_id":"007","type":"should_call","user_input":"%%sql magic command 怎么在 notebook 里用","expected_skill":"clickzetta-data-science","expected_output_contains":["%%sql"]}
+{"case_id":"008","type":"should_not_call","user_input":"TensorFlow 怎么训练模型","forbidden_skill":"clickzetta-data-science"}
+{"case_id":"009","type":"should_not_call","user_input":"帮我写一个 ETL 管道","forbidden_skill":"clickzetta-data-science"}
+{"case_id":"010","type":"should_not_call","user_input":"怎么创建 VCluster","forbidden_skill":"clickzetta-data-science"}
+{"case_id":"011","type":"should_not_call","user_input":"怎么做数据分享","forbidden_skill":"clickzetta-data-science"}
+{"case_id":"012","type":"should_not_call","user_input":"scikit-learn 怎么安装","forbidden_skill":"clickzetta-data-science"}

package/bin/skills/clickzetta-data-science/references/bitmap-profile.md

@@ -0,0 +1,146 @@
+# BITMAP 用户画像参考
+
+> 来源：https://www.yunqi.tech/documents/bitmap-type
+
+BITMAP 是 ClickZetta 中用于高效存储和处理整数集合的数据类型，基于 Roaring Bitmap 压缩算法，特别适合用户画像、人群圈选、UV 统计等数据科学场景。
+
+---
+
+## 核心限制
+
+- 支持 **64 位无符号整数**（0 到 2^64-1）
+- **不支持**比较操作（<、>、=）
+- **不支持** ORDER BY、GROUP BY、DISTINCT
+- **不能**作为 PRIMARY KEY、PARTITION KEY、CLUSTER KEY
+
+---
+
+## 构建用户标签 BITMAP
+
+```sql
+-- 方式 1：从行数据聚合构建（最常用）
+CREATE TABLE ds_workspace.user_tags AS
+SELECT
+    tag_name,
+    group_bitmap_state(user_id) AS user_bitmap
+FROM (
+    -- 高消费用户
+    SELECT 'high_value' AS tag_name, user_id
+    FROM my_schema.orders
+    WHERE total_amount_30d > 1000
+    UNION ALL
+    -- 近30天活跃用户
+    SELECT 'active_30d' AS tag_name, user_id
+    FROM my_schema.events
+    WHERE event_date >= CURRENT_DATE - INTERVAL 30 DAY
+    UNION ALL
+    -- 已流失用户（90天未活跃）
+    SELECT 'churned' AS tag_name, user_id
+    FROM my_schema.users
+    WHERE last_active_date < CURRENT_DATE - INTERVAL 90 DAY
+) t
+GROUP BY tag_name;
+
+-- 方式 2：从数组构建
+INSERT INTO ds_workspace.user_tags VALUES
+    ('vip', bitmap_build(ARRAY(1001, 1002, 1003, 1004)));
+```
+
+---
+
+## 人群圈选操作
+
+```sql
+-- 交集：同时满足多个标签（AND）
+SELECT bitmap_count(
+    bitmap_and(
+        (SELECT user_bitmap FROM ds_workspace.user_tags WHERE tag_name = 'high_value'),
+        (SELECT user_bitmap FROM ds_workspace.user_tags WHERE tag_name = 'active_30d')
+    )
+) AS target_count;
+
+-- 并集：满足任一标签（OR）
+SELECT bitmap_count(
+    bitmap_or(
+        (SELECT user_bitmap FROM ds_workspace.user_tags WHERE tag_name = 'high_value'),
+        (SELECT user_bitmap FROM ds_workspace.user_tags WHERE tag_name = 'active_30d')
+    )
+) AS reach_count;
+
+-- 差集：排除某类用户（ANDNOT）
+SELECT bitmap_count(
+    bitmap_andnot(
+        (SELECT user_bitmap FROM ds_workspace.user_tags WHERE tag_name = 'high_value'),
+        (SELECT user_bitmap FROM ds_workspace.user_tags WHERE tag_name = 'churned')
+    )
+) AS targetable_count;
+
+-- 获取目标用户 ID 列表
+SELECT bitmap_to_array(
+    bitmap_andnot(
+        (SELECT user_bitmap FROM ds_workspace.user_tags WHERE tag_name = 'high_value'),
+        (SELECT user_bitmap FROM ds_workspace.user_tags WHERE tag_name = 'churned')
+    )
+) AS target_user_ids;
+```
+
+---
+
+## UV 统计（去重计数）
+
+```sql
+-- 日活跃用户数（DAU）
+SELECT
+    event_date,
+    bitmap_count(group_bitmap_state(user_id)) AS dau
+FROM my_schema.events
+GROUP BY event_date
+ORDER BY event_date;
+
+-- 周活跃用户数（WAU）—— 跨天去重
+SELECT
+    DATE_TRUNC('week', event_date) AS week_start,
+    bitmap_count(
+        bitmap_or_agg(daily_bitmap)  -- 合并多天 bitmap
+    ) AS wau
+FROM (
+    SELECT event_date,
+           group_bitmap_state(user_id) AS daily_bitmap
+    FROM my_schema.events
+    GROUP BY event_date
+) t
+GROUP BY 1;
+
+-- 用户留存分析（新用户 vs 回访用户）
+SELECT
+    bitmap_count(
+        bitmap_and(new_users.user_bitmap, return_users.user_bitmap)
+    ) AS retained_users,
+    bitmap_count(
+        bitmap_andnot(new_users.user_bitmap, return_users.user_bitmap)
+    ) AS lost_users
+FROM
+    (SELECT group_bitmap_state(user_id) AS user_bitmap
+     FROM my_schema.events WHERE event_date = '2024-01-01') AS new_users,
+    (SELECT group_bitmap_state(user_id) AS user_bitmap
+     FROM my_schema.events WHERE event_date = '2024-01-08') AS return_users;
+```
+
+---
+
+## 常用 BITMAP 函数速查
+
+| 函数 | 说明 | 示例 |
+|---|---|---|
+| `group_bitmap_state(col)` | 聚合构建 BITMAP | `GROUP BY tag` |
+| `bitmap_count(bm)` | 计算元素个数（UV） | `bitmap_count(user_bm)` |
+| `bitmap_and(a, b)` | 交集 | 同时满足 A 和 B |
+| `bitmap_or(a, b)` | 并集 | 满足 A 或 B |
+| `bitmap_andnot(a, b)` | 差集 | 在 A 中但不在 B 中 |
+| `bitmap_xor(a, b)` | 异或（只在一个中） | A、B 各自独有的 |
+| `bitmap_to_array(bm)` | 转为整数数组 | 获取用户 ID 列表 |
+| `bitmap_build(arr)` | 从数组构建 | `bitmap_build(ARRAY(1,2,3))` |
+| `bitmap_contains(bm, val)` | 检查是否包含某值 | `bitmap_contains(bm, user_id)` |
+| `bitmap_min(bm)` | 最小元素 | — |
+| `bitmap_max(bm)` | 最大元素 | — |
+| `to_bitmap(val)` | 单值转 BITMAP | `to_bitmap(user_id)` |

package/bin/skills/clickzetta-data-science/references/data-patterns.md

@@ -0,0 +1,110 @@
+# 数据发现、质量评估、清洗、EDA 示例
+
+## 数据发现
+
+```python
+from src.config import get_session
+session = get_session()
+
+session.sql("SHOW SCHEMAS").show()
+session.sql("SHOW TABLES IN my_schema").show()
+session.sql("DESC EXTENDED my_schema.orders").show()
+session.sql("""
+    SELECT table_name, row_count,
+           ROUND(bytes/1024.0/1024/1024, 2) AS size_gb
+    FROM information_schema.tables
+    WHERE table_schema = 'my_schema'
+    ORDER BY bytes DESC
+""").show()
+```
+
+---
+
+## 数据质量评估
+
+```sql
+-- 基础统计
+SELECT
+    COUNT(*)                                                          AS total_rows,
+    COUNT(DISTINCT user_id)                                           AS unique_users,
+    MIN(event_time) AS earliest, MAX(event_time) AS latest,
+    ROUND(100.0 * SUM(CASE WHEN user_id IS NULL THEN 1 ELSE 0 END) / COUNT(*), 2) AS user_id_null_pct,
+    ROUND(100.0 * SUM(CASE WHEN amount  IS NULL THEN 1 ELSE 0 END) / COUNT(*), 2) AS amount_null_pct
+FROM my_schema.orders;
+
+-- 主键重复检查
+SELECT order_id, COUNT(*) AS cnt
+FROM my_schema.orders GROUP BY order_id HAVING cnt > 1 LIMIT 10;
+
+-- 数值分布（大表高效）
+SELECT
+    approx_percentile(amount, 0.25) AS p25,
+    approx_percentile(amount, 0.50) AS median,
+    approx_percentile(amount, 0.75) AS p75,
+    approx_percentile(amount, 0.99) AS p99,
+    MIN(amount) AS min_val, MAX(amount) AS max_val
+FROM my_schema.orders;
+
+-- 高频值 TOP-K
+SELECT approx_top_k(status, 10) AS top_statuses FROM my_schema.orders;
+
+-- 近似 UV
+SELECT approx_count_distinct(user_id) AS approx_uv FROM my_schema.events;
+```
+
+---
+
+## 数据清洗
+
+```sql
+-- 去重（保留最新一条）
+SELECT * FROM (
+    SELECT *, ROW_NUMBER() OVER (PARTITION BY order_id ORDER BY update_time DESC) AS rn
+    FROM my_schema.orders_raw
+) WHERE rn = 1;
+
+-- 缺失值处理 + 类型转换
+SELECT
+    order_id, user_id,
+    COALESCE(amount, 0.0)       AS amount,
+    COALESCE(status, 'UNKNOWN') AS status,
+    CAST(order_date AS DATE)    AS order_date
+FROM my_schema.orders_raw
+WHERE user_id IS NOT NULL;
+
+-- 多表整合
+SELECT o.order_id, o.user_id, o.amount, o.order_date,
+       u.age_group, u.city, p.category, p.brand
+FROM my_schema.orders o
+LEFT JOIN my_schema.users    u ON o.user_id    = u.user_id
+LEFT JOIN my_schema.products p ON o.product_id = p.product_id;
+```
+
+---
+
+## EDA
+
+```python
+# 采样策略
+df_quick = session.sql("""
+    SELECT * FROM my_schema.events TABLESAMPLE SYSTEM (0.1) LIMIT 50000
+""").to_pandas()  # SYSTEM：文件级，极快，适合 >100万行预览
+
+df_ml = session.sql("""
+    SELECT * FROM my_schema.events TABLESAMPLE ROW (10)
+""").to_pandas()  # ROW：行级精确，适合 ML 训练集
+
+# 时序分析
+session.sql("""
+    SELECT
+        DATE_TRUNC('day', order_time)  AS dt,
+        COUNT(*)                       AS daily_orders,
+        SUM(amount)                    AS daily_revenue,
+        AVG(SUM(amount)) OVER (
+            ORDER BY DATE_TRUNC('day', order_time)
+            ROWS BETWEEN 6 PRECEDING AND CURRENT ROW
+        )                              AS revenue_7d_ma
+    FROM my_schema.orders
+    GROUP BY 1 ORDER BY 1
+""").to_pandas().plot(x='dt', y=['daily_revenue', 'revenue_7d_ma'])
+```

package/bin/skills/clickzetta-data-science/references/setup.md

@@ -0,0 +1,160 @@
+# 环境搭建与项目配置
+
+## 环境搭建
+
+```bash
+# 方式 1：venv（推荐）
+python3.12 -m venv .venv
+source .venv/bin/activate          # macOS/Linux
+pip install clickzetta_zettapark_python clickzetta-connector-python \
+    python-dotenv pandas numpy scikit-learn pyarrow jupyterlab matplotlib seaborn \
+    -i https://pypi.tuna.tsinghua.edu.cn/simple
+
+# 方式 2：pyenv（需要切换 Python 版本时）
+pyenv install 3.12.9 && pyenv local 3.12.9
+python -m venv .venv && source .venv/bin/activate
+pip install clickzetta_zettapark_python clickzetta-connector-python \
+    python-dotenv pandas numpy scikit-learn pyarrow jupyterlab matplotlib seaborn \
+    -i https://pypi.tuna.tsinghua.edu.cn/simple
+
+# 方式 3：conda
+conda create -n lakehouse-ds python=3.12 -y && conda activate lakehouse-ds
+pip install clickzetta_zettapark_python clickzetta-connector-python \
+    python-dotenv pandas numpy scikit-learn pyarrow jupyterlab matplotlib seaborn \
+    -i https://pypi.tuna.tsinghua.edu.cn/simple
+```
+
+| 问题 | 修复 |
+|------|------|
+| Python 3.8/3.9 | `pyenv install 3.12.9` 或 `python3.12 -m venv .venv` |
+| `pyarrow` 版本冲突 | `pip install pyarrow==14.0.0` |
+| M1/M2 Mac 报错 | `pip install --no-binary :all:` 或改用 conda |
+| 连接超时 | VCluster 未启动，在 Studio 中手动启动 |
+
+---
+
+## Jupyter Kernel 配置
+
+```bash
+# 注册 venv 为 Jupyter kernel（关键步骤，否则 notebook 用系统 Python）
+source .venv/bin/activate
+pip install ipykernel jupyterlab
+python -m ipykernel install --user --name lakehouse-ds --display-name "Python (lakehouse-ds)"
+
+# 启动 JupyterLab
+jupyter lab --port=8888
+```
+
+VS Code / Cursor：打开 `.ipynb` → 右上角 "Select Kernel" → 选 "Python (lakehouse-ds)"
+
+| 问题 | 修复 |
+|------|------|
+| `ModuleNotFoundError: clickzetta` | kernel 未选对，切换到注册的 venv kernel |
+| `.env` 读不到 | `load_dotenv(dotenv_path='../.env')` 指定路径 |
+| `to_pandas()` OOM | 加 `TABLESAMPLE ROW(1)` 或 `LIMIT` |
+| 图表不显示 | notebook 开头加 `%matplotlib inline` |
+
+---
+
+## src/config.py 模板
+
+```python
+import os, sys
+from pathlib import Path
+from dotenv import load_dotenv
+from clickzetta.zettapark.session import Session
+import clickzetta
+
+# 多位置查找 .env
+for _p in [
+    Path(__file__).parent.parent / ".env",
+    Path.home() / ".config" / "kilo" / ".env",
+    Path.home() / ".czcode" / ".env",
+    Path.home() / ".env",
+]:
+    if _p.exists():
+        load_dotenv(dotenv_path=_p)
+        break
+
+def check_environment():
+    """在 00-env-check.ipynb 里调用，打印环境诊断。"""
+    ver = sys.version_info
+    if ver < (3, 10):
+        raise RuntimeError(
+            f"Python {ver.major}.{ver.minor} 不满足要求。ZettaPark 需要 Python 3.10+。\n"
+            "升级：brew install pyenv && pyenv install 3.12.9 && pyenv local 3.12.9"
+        )
+    print(f"✅ Python {ver.major}.{ver.minor}.{ver.micro}")
+    for pkg, mod in [
+        ("clickzetta_zettapark_python", "clickzetta.zettapark"),
+        ("clickzetta-connector-python", "clickzetta"),
+        ("pandas", "pandas"), ("python-dotenv", "dotenv"),
+    ]:
+        try:
+            m = __import__(mod.split(".")[0])
+            print(f"✅ {pkg}: {getattr(m, '__version__', 'ok')}")
+        except ImportError:
+            print(f"❌ {pkg}: 未安装 → pip install {pkg}")
+    try:
+        s = get_session()
+        print(f"✅ Lakehouse: {s.sql('SELECT current_workspace(), current_user()').collect()}")
+    except Exception as e:
+        print(f"❌ Lakehouse 连接失败: {e}")
+
+def get_session() -> Session:
+    return Session.builder.configs({
+        "service":   os.environ["CLICKZETTA_SERVICE"],
+        "instance":  os.environ["CLICKZETTA_INSTANCE"],
+        "workspace": os.environ["CLICKZETTA_WORKSPACE"],
+        "username":  os.environ["CLICKZETTA_USERNAME"],
+        "password":  os.environ["CLICKZETTA_PASSWORD"],
+        "vcluster":  os.environ.get("CLICKZETTA_VCLUSTER", "default_ap"),
+        "schema":    os.environ.get("CLICKZETTA_SCHEMA", "public"),
+    }).create()
+
+def get_connector_connection():
+    """仅用于 pd.read_sql。禁止用于 df.to_sql()。"""
+    return clickzetta.connect(
+        service=os.environ["CLICKZETTA_SERVICE"],
+        instance=os.environ["CLICKZETTA_INSTANCE"],
+        workspace=os.environ["CLICKZETTA_WORKSPACE"],
+        username=os.environ["CLICKZETTA_USERNAME"],
+        password=os.environ["CLICKZETTA_PASSWORD"],
+        vcluster=os.environ.get("CLICKZETTA_VCLUSTER", "default_ap"),
+        schema=os.environ.get("CLICKZETTA_SCHEMA", "public"),
+    )
+```
+
+---
+
+## .env 模板
+
+```bash
+CLICKZETTA_SERVICE=cn-shanghai-alicloud.api.clickzetta.com
+CLICKZETTA_INSTANCE=<instance-id>
+CLICKZETTA_WORKSPACE=<workspace>
+CLICKZETTA_USERNAME=<username>
+CLICKZETTA_PASSWORD=<password>
+CLICKZETTA_VCLUSTER=default_ap
+CLICKZETTA_SCHEMA=ds_workspace
+```
+
+## pyproject.toml
+
+```toml
+[project]
+name = "my-lakehouse-ds-project"
+requires-python = ">=3.10"
+dependencies = [
+    "clickzetta_zettapark_python>=0.1.2",
+    "clickzetta-connector-python>=1.0.0",
+    "python-dotenv>=1.0.0",
+    "pandas>=2.0.0",
+    "numpy>=1.24.0",
+    "scikit-learn>=1.3.0",
+    "pyarrow>=14.0.0",
+    "jupyterlab>=4.0.0",
+    "matplotlib>=3.7.0",
+    "seaborn>=0.12.0",
+]
+```