feat: add click gap pipeline and dashboard

.gitignore

@@ -0,0 +1,12 @@
+.DS_Store
+.env
+.env.*
+*.db
+*.sqlite
+*.sqlite3
+*.log
+__pycache__/
+*.py[cod]
+.cache/
+.cookies/
+browser-profile/

README.md

@@ -0,0 +1,62 @@
+# AppsFlyer Click Gap
+
+运行目标是 **mac2**。流水线复用 `appsflyer/af_click_gap.py` 的 Superset 查询、AppsFlyer payload、特殊包名映射和差值公式，但新入口增加 timeout/重试、严格错误边界、AppsFlyer 重复键按 `dt+pkg_name` 求和，并将日志写到仓库外 `~/Library/Logs/af-click-gap/pipeline.log`。固定数据库：`/Users/ads-center/Documents/database/af_click_gap.db`；不复制或提交数据库、cookie、token、`.env`、日志和缓存。
+
+## 文件和字段
+
+- `scripts/af_click_gap_pipeline.py`：提取、校验、发布和 Git 操作。
+- `data/YYYYMMDD.json`：**直接 JSON 数组**；无数据日期也写 `[]`。每行字段为 `dt`（`YYYYMMDD` 字符串）、`attr_platform`、`pkg_name`（非空字符串）、`shareit_clicks`、`af_clicks`、`installs_ua`、`clicks_diff`（有限 number）、`clicks_diff_rate`（有限 number 或 `null`）。业务键为 `pid + dt + pkg_name`。
+- `data/index.json`：版本、生成时间及 `data/YYYYMMDD.json` 清单。
+- `index.html`：无构建依赖的静态看板。
+
+默认窗口是 mac2 本地运行日及前 6 天，首尾包含，共 7 天；这是对旧脚本实际可能覆盖 8 个日期的修正。空 Superset 结果、无达标包或 AppsFlyer 合法空数组都是成功空结果（退出码 0），仍生成日期数组、同步清空 SQLite 窗口并可提交；HTTP、认证、响应格式、SQLite schema 或文件校验错误退出码 1，不覆盖正式文件。
+
+SQLite 必须预先存在 `click_gap` 表，列类型为 `pid TEXT NOT NULL, dt TEXT NOT NULL, attr_platform TEXT NOT NULL, pkg_name TEXT NOT NULL, shareit_clicks REAL NOT NULL, af_clicks REAL NOT NULL, installs_ua REAL NOT NULL, clicks_diff REAL NOT NULL, clicks_diff_rate REAL, updated_at TEXT NOT NULL`，主键严格为 `(pid, dt, pkg_name)`。脚本不建表、不迁移；每次成功运行先删除窗口内 `shareit` 旧行再插入完整新结果，因此消失日期/业务键也会清零。
+
+## 手工运行
+
+在 mac2 先确认依赖和登录环境（`pandas`、`requests`、`requests_toolbelt`、`af_browser`、`superset_login`）：
+
+```sh
+cd /Users/shareit./Library/CloudStorage/OneDrive-个人/ubuntu/Gitea/af-click-gap
+python3 -m py_compile scripts/af_click_gap_pipeline.py
+python3 scripts/af_click_gap_pipeline.py --help
+python3 scripts/af_click_gap_pipeline.py --dry-run
+# 正式写入并提交、push（默认行为）
+python3 scripts/af_click_gap_pipeline.py --push
+# 只写 data 和 SQLite，不操作 Git
+python3 scripts/af_click_gap_pipeline.py --no-push
+# 手工补数，日期首尾包含且最多 7 天
+python3 scripts/af_click_gap_pipeline.py --start-date 2026-10-10 --end-date 2026-10-16 --no-push
+```
+
+`--dry-run` 只生成仓库外临时 JSON 并校验，不写 data、数据库或执行 Git。正式文件先原子替换，再提交 SQLite，最后才 `git add/commit/push`；数据库提交失败时不执行 Git。commit/push 失败会保留本地文件和数据库，下一次可重跑或手工 `git add data index.html ... && git commit && git push`；不会伪造认证或 force push。工作区在 push 模式有非本任务改动时拒绝继续。
+
+## 看板
+
+直接打开 `index.html` 或运行 `python3 -m http.server` 后访问。页面读取并校验 `data/index.json` 及清单路径，只执行安全 JSON 解析；缺失/损坏文件显示“数据不完整”并跳过该文件，清单本身失败则不展示数据。默认最近 30 天，可选有效开始/结束日期和 `pkg_name`；默认 `dt` 降序、同日期 `shareit_clicks` 降序，点击表头切换排序。
+
+## mac2 launchd
+
+`launchd/com.roasbloom.af-click-gap.plist.template` 明确为 `TEMPLATE_ONLY`。当前开发机不能证明自己是 mac2，因此没有伪造最终 plist、解释器或安装结果。必须在 mac2 执行并记录：
+
+```sh
+hostname
+command -v python3
+python3 --version
+python3 -c 'import pandas,requests,requests_toolbelt,af_browser,superset_login; print("dependencies ok")'
+cd /Users/shareit./Library/CloudStorage/OneDrive-个人/ubuntu/Gitea/af-click-gap
+mkdir -p "$HOME/Library/Logs/af-click-gap"
+PY=$(command -v python3); sed -e "s#__MAC2_PYTHON3__#$PY#" -e "s#__PROJECT_DIR__#$PWD#g" -e "s#__LOG_DIR__#$HOME/Library/Logs/af-click-gap#g" launchd/com.roasbloom.af-click-gap.plist.template > /tmp/com.roasbloom.af-click-gap.plist
+plutil -lint /tmp/com.roasbloom.af-click-gap.plist
+cp /tmp/com.roasbloom.af-click-gap.plist "$HOME/Library/LaunchAgents/"
+launchctl bootstrap "gui/$(id -u)" "$HOME/Library/LaunchAgents/com.roasbloom.af-click-gap.plist"
+launchctl print "gui/$(id -u)/com.roasbloom.af-click-gap"
+# 卸载：launchctl bootout "gui/$(id -u)/com.roasbloom.af-click-gap"; rm "$HOME/Library/LaunchAgents/com.roasbloom.af-click-gap.plist"
+```
+
+定时任务失败不会触发成功 push；stdout/stderr 位于 mac2 用户日志目录。Git remote、SSH/HTTPS credential helper 或 Gitea token 必须由用户在 mac2 预先配置。若 `git remote -v` 为空，请按组织 Git 文档创建仓库并绑定 remote 后再运行；本项目不会猜测仓库名称或修改 git config。
+
+## 安全和验证
+
+允许提交的路径仅为脚本、README、看板、`data/*.json`、`data/index.json`、模板和 `.gitignore`。提交前禁止 `.env*`、SQLite、日志、浏览器状态，脚本还扫描 `cookie`、`authorization`、`password`、`api-key`、`token` 文本。建议验证：`python3 -m py_compile ...`、`--help`、mac2 上 `--dry-run`、`python3 -m json.tool data/index.json`，以及浏览器/静态 HTTP 打开页面。当前环境未验证 mac2 依赖、固定数据库、外部登录、launchd 安装和远程 push；这些不能以本机结果冒充。不要新增测试套件。

index.html

@@ -0,0 +1,20 @@
+<!doctype html>
+<html lang="zh-CN">
+<head>
+<meta charset="utf-8"><meta name="viewport" content="width=device-width, initial-scale=1"><title>AppsFlyer Click Gap</title>
+<style>body{font:14px system-ui,sans-serif;margin:2rem;color:#223}label{margin-right:1rem}input,select,button{padding:.35rem}table{border-collapse:collapse;width:100%;margin-top:1rem}th,td{border:1px solid #ccd;padding:.45rem;text-align:right}th{cursor:pointer;background:#eef}th:first-child,td:first-child,th:nth-child(2),td:nth-child(2),th:nth-child(3),td:nth-child(3){text-align:left}.status{padding:.7rem;background:#f5f5f5}.error{color:#a00}.sort{font-size:.8em}</style>
+</head>
+<body><h1>AppsFlyer Click Gap</h1><div><label>开始 <input id="start" type="date"></label><label>结束 <input id="end" type="date"></label><label>pkg_name <select id="pkg"><option value="">全部</option></select></label><button id="apply">查询</button><button id="reset">重置</button></div><p id="status" class="status">加载中…</p><table><thead><tr id="head"></tr></thead><tbody id="body"></tbody></table>
+<script>
+(() => {"use strict";
+const columns=["dt","attr_platform","pkg_name","shareit_clicks","af_clicks","installs_ua","clicks_diff","clicks_diff_rate"];let rows=[], failed=[];let sort={key:"dt",direction:-1};
+const $=id=>document.getElementById(id), status=$('status');
+function text(value){return value===null||value===undefined?"—":String(value)}
+function validDate(v){return /^\d{4}-\d{2}-\d{2}$/.test(v)&&!Number.isNaN(Date.parse(v))}
+function date8(v){return v.replaceAll('-','')}
+function show(){const start=$('start').value,end=$('end').value,pkg=$('pkg').value;if(!validDate(start)||!validDate(end)||start>end){status.textContent='请输入有效日期范围，且开始日期不得晚于结束日期。';status.className='status error';return}const filtered=rows.filter(r=>r.dt>=date8(start)&&r.dt<=date8(end)&&(!pkg||r.pkg_name===pkg));filtered.sort((a,b)=>{const av=a[sort.key],bv=b[sort.key];if(av===bv)return 0;if(av===null||av===undefined)return 1;if(bv===null||bv===undefined)return -1;const n=typeof av==='number'&&typeof bv==='number'?av-bv:String(av).localeCompare(String(bv));return n*sort.direction});$('body').replaceChildren();filtered.forEach(r=>{const tr=document.createElement('tr');columns.forEach(c=>{const td=document.createElement('td');td.textContent=text(r[c]);tr.append(td)});$('body').append(tr)});status.textContent=(failed.length?`数据不完整：跳过 ${failed.length} 个文件。`:'')+`显示 ${filtered.length} 行。`;status.className='status'+(failed.length?' error':'')}
+function setupHeaders(){const h=$('head');columns.forEach(c=>{const th=document.createElement('th');th.dataset.key=c;th.textContent=c;th.addEventListener('click',()=>{if(sort.key===c)sort.direction*=-1;else{sort.key=c;sort.direction=-1}setupHeaders();show()});h.append(th)});}
+async function load(){setupHeaders();try{const response=await fetch('data/index.json',{cache:'no-store'});if(!response.ok)throw Error('index.json 加载失败');const index=await response.json();if(!Array.isArray(index.files))throw Error('数据清单格式错误');const safe=index.files.filter(f=>/^\d{8}$/.test(f.date)&&f.path===`data/${f.date}.json`);const bad=index.files.length-safe.length;failed=bad?['清单路径']:[];const loaded=await Promise.all(safe.map(async f=>{try{const r=await fetch(f.path,{cache:'no-store'});if(!r.ok)throw Error('HTTP');const data=await r.json();if(!Array.isArray(data)||data.some(x=>!x||typeof x!=='object'||x.dt!==f.date))throw Error('格式');return data}catch(e){failed.push(f.date);return []}}));rows=loaded.flat();const dates=rows.map(r=>r.dt).sort();if(!dates.length){$('start').value=$('end').value='';status.textContent='暂无数据。';return}const end=dates.at(-1), endDate=new Date(`${end.slice(0,4)}-${end.slice(4,6)}-${end.slice(6)}T00:00:00`);endDate.setDate(endDate.getDate()-29);$('end').value=`${end.slice(0,4)}-${end.slice(4,6)}-${end.slice(6)}`;$('start').value=endDate.toISOString().slice(0,10);[...new Set(rows.map(r=>r.pkg_name).filter(Boolean))].sort().forEach(p=>{const o=document.createElement('option');o.value=p;o.textContent=p;$('pkg').append(o)});show()}catch(e){status.textContent=`无法加载数据：${e.message}`;status.className='status error'}}
+$('apply').addEventListener('click',show);$('reset').addEventListener('click',()=>{if(rows.length){$('pkg').value='';const d=rows.map(r=>r.dt).sort().at(-1);$('end').value=`${d.slice(0,4)}-${d.slice(4,6)}-${d.slice(6)}`;const x=new Date($('end').value);x.setDate(x.getDate()-29);$('start').value=x.toISOString().slice(0,10);show()}});load();
+})();
+</script></body></html>

launchd/com.roasbloom.af-click-gap.plist.template

@@ -0,0 +1,12 @@
+<!-- TEMPLATE_ONLY: render and validate on mac2 before installing. -->
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+<plist version="1.0"><dict>
+<key>Label</key><string>com.roasbloom.af-click-gap</string>
+<key>ProgramArguments</key><array><string>__MAC2_PYTHON3__</string><string>__PROJECT_DIR__/scripts/af_click_gap_pipeline.py</string><string>--push</string></array>
+<key>WorkingDirectory</key><string>__PROJECT_DIR__</string>
+<key>EnvironmentVariables</key><dict><key>DB_PATH</key><string>/Users/ads-center/Documents/database/af_click_gap.db</string></dict>
+<key>StartCalendarInterval</key><dict><key>Hour</key><integer>2</integer><key>Minute</key><integer>0</integer></dict>
+<key>StandardOutPath</key><string>__LOG_DIR__/pipeline.stdout.log</string><key>StandardErrorPath</key><string>__LOG_DIR__/pipeline.stderr.log</string>
+<key>RunAtLoad</key><false/>
+</dict></plist>

scripts/af_click_gap_pipeline.py

@@ -0,0 +1,366 @@
+#!/usr/bin/env python3
+"""Extract AppsFlyer click-gap reports and publish validated static JSON."""
+from __future__ import annotations
+
+import argparse
+import datetime as dt
+import json
+import logging
+import math
+import os
+import re
+import shutil
+import sqlite3
+import subprocess
+import sys
+import tempfile
+import time
+from pathlib import Path
+from typing import Any
+
+DB_PATH = "/Users/ads-center/Documents/database/af_click_gap.db"
+PID = "shareit"
+CLICK_NUM = 300
+REPORT_COLUMNS = ["dt", "attr_platform", "pkg_name", "shareit_clicks", "af_clicks", "installs_ua", "clicks_diff", "clicks_diff_rate"]
+SPECIAL_PKGS = {
+    "za.co.freeurl.easybetapp.fwdb": "za.co.freeurl.easybetapp.fwdb-Shareit",
+    "org.melbet.client": "org.melbet.client-Custom", "org.xbet.client1": "org.xbet.client1-Custom",
+    "org.betwinner_ng.client": "org.betwinner_ng.client-Custom", "org.betwinner_zm.client": "org.betwinner_zm.client-Custom",
+    "org.betwinner_cm.client": "org.betwinner_cm.client-Custom", "org.betwinner_cd.client": "org.betwinner_cd.client-Custom",
+}
+EXPECTED_TYPES = {"pid": "TEXT", "dt": "TEXT", "attr_platform": "TEXT", "pkg_name": "TEXT", "shareit_clicks": "REAL", "af_clicks": "REAL", "installs_ua": "REAL", "clicks_diff": "REAL", "clicks_diff_rate": "REAL", "updated_at": "TEXT"}
+LOG_PATH = Path(os.path.expanduser("~/Library/Logs/af-click-gap/pipeline.log"))
+
+
+def logger_setup() -> logging.Logger:
+    LOG_PATH.parent.mkdir(parents=True, exist_ok=True)
+    log = logging.getLogger("af_click_gap")
+    if not log.handlers:
+        log.setLevel(logging.INFO)
+        handler = logging.FileHandler(LOG_PATH, encoding="utf-8")
+        handler.setFormatter(logging.Formatter("%(asctime)s %(levelname)s %(message)s"))
+        log.addHandler(handler)
+        log.addHandler(logging.StreamHandler())
+    return log
+
+
+LOG = logger_setup()
+
+
+def date8(value: str) -> str:
+    try:
+        parsed = dt.datetime.strptime(value, "%Y%m%d")
+    except ValueError as exc:
+        raise ValueError(f"非法日期: {value}") from exc
+    return parsed.strftime("%Y%m%d")
+
+
+def date8_from_iso(value: str) -> str:
+    try:
+        return dt.datetime.strptime(value, "%Y-%m-%d").strftime("%Y%m%d")
+    except ValueError as exc:
+        raise ValueError(f"日期必须为 YYYY-MM-DD: {value}") from exc
+
+
+def window(start: str | None, end: str | None) -> tuple[str, str]:
+    today = dt.date.today()
+    if start is None and end is None:
+        first, last = today - dt.timedelta(days=6), today
+    else:
+        if not start or not end:
+            raise ValueError("--start-date 与 --end-date 必须同时提供")
+        first = dt.datetime.strptime(start, "%Y-%m-%d").date()
+        last = dt.datetime.strptime(end, "%Y-%m-%d").date()
+    if first > last or (last - first).days > 6:
+        raise ValueError("日期区间必须为包含首尾且不超过 7 天")
+    return first.strftime("%Y%m%d"), last.strftime("%Y%m%d")
+
+
+def finite_number(value: Any, field: str, allow_null: bool = False) -> int | float | None:
+    if value is None or (isinstance(value, float) and math.isnan(value)):
+        if allow_null:
+            return None
+        raise ValueError(f"{field} 不能为空")
+    try:
+        number = float(value)
+    except (TypeError, ValueError) as exc:
+        raise ValueError(f"{field} 必须为数值") from exc
+    if not math.isfinite(number):
+        raise ValueError(f"{field} 必须为有限数值")
+    return int(number) if number.is_integer() else number
+
+
+def validate_existing_schema(path: str) -> None:
+    db = Path(path)
+    if not db.is_file():
+        raise RuntimeError(f"SQLite 文件不存在: {path}")
+    try:
+        with sqlite3.connect(f"file:{db}?mode=ro", uri=True) as conn:
+            columns = conn.execute("PRAGMA table_info(click_gap)").fetchall()
+    except sqlite3.Error as exc:
+        raise RuntimeError(f"SQLite 不可读: {exc}") from exc
+    if not columns:
+        raise RuntimeError("click_gap 表不存在")
+    by_name = {row[1]: row for row in columns}
+    missing = set(EXPECTED_TYPES) - set(by_name)
+    if missing:
+        raise RuntimeError(f"click_gap 缺少列: {sorted(missing)}")
+    for name, expected in EXPECTED_TYPES.items():
+        if str(by_name[name][2]).upper() != expected:
+            raise RuntimeError(f"click_gap 列类型不兼容: {name}")
+    if [(row[1], row[5]) for row in columns if row[5]] != [("pid", 1), ("dt", 2), ("pkg_name", 3)]:
+        raise RuntimeError("click_gap 主键必须为 (pid, dt, pkg_name)")
+    nullable_required = {"pid", "dt", "pkg_name", "shareit_clicks", "af_clicks", "installs_ua", "clicks_diff", "updated_at"}
+    if any(by_name[name][3] != 1 for name in nullable_required):
+        raise RuntimeError("click_gap 必填列不得允许 NULL")
+
+
+def read_existing_rows(path: str, start: str, end: str, pid: str = PID) -> list[dict[str, Any]]:
+    validate_existing_schema(path)
+    with sqlite3.connect(path) as conn:
+        conn.row_factory = sqlite3.Row
+        return [dict(row) for row in conn.execute("SELECT * FROM click_gap WHERE pid=? AND dt BETWEEN ? AND ?", (pid, start, end))]
+
+
+def upsert_report(conn: sqlite3.Connection, rows: list[dict[str, Any]], start: str, end: str) -> int:
+    now = dt.datetime.now().astimezone().isoformat(timespec="seconds")
+    conn.execute("DELETE FROM click_gap WHERE pid=? AND dt BETWEEN ? AND ?", (PID, start, end))
+    values = []
+    for row in rows:
+        values.append((PID, row["dt"], row["attr_platform"], row["pkg_name"], row["shareit_clicks"], row["af_clicks"], row["installs_ua"], row["clicks_diff"], row["clicks_diff_rate"], now))
+    conn.executemany("INSERT INTO click_gap (pid,dt,attr_platform,pkg_name,shareit_clicks,af_clicks,installs_ua,clicks_diff,clicks_diff_rate,updated_at) VALUES (?,?,?,?,?,?,?,?,?,?)", values)
+    return len(values)
+
+
+def superset_rows(start: str, end: str) -> list[dict[str, Any]]:
+    from superset_login import get_cookies
+    import requests
+    from requests_toolbelt import MultipartEncoder
+    session, csrf = get_cookies()
+    nextday = (dt.datetime.strptime(end, "%Y%m%d").date() + dt.timedelta(days=1)).strftime("%Y-%m-%d")
+    start_iso = dt.datetime.strptime(start, "%Y%m%d").strftime("%Y-%m-%d")
+    query = {"datasource": "114__table", "viz_type": "table", "slice_id": 1344, "url_params": {}, "time_range_endpoints": ["inclusive", "exclusive"], "granularity_sqla": "dt_as_time", "time_grain_sqla": "P1D", "time_range": f"{start_iso}T00:00:00 : {nextday}T00:00:00", "query_mode": "aggregate", "groupby": ["dt", "attr_platform", "pkg_name"], "metrics": ["clicks", "pb_install"], "all_columns": [], "array_join": [], "percent_metrics": [], "timeseries_limit_metric": ["dt_sort"], "order_by_cols": [], "row_limit": None, "order_desc": True, "adhoc_filters": [{"clause": "WHERE", "expressionType": "SIMPLE", "comparator": ["appsflyer"], "operator": "IN", "subject": "attr_platform"}, {"clause": "WHERE", "expressionType": "SIMPLE", "comparator": ["0"], "operator": "IN", "subject": "is_internal"}], "table_timestamp_format": "smart_date", "color_pn": True, "show_cell_bars": True}
+    encoder = MultipartEncoder(fields={"form_data": json.dumps(query)})
+    response = session.post('https://ads-bi.ushareit.org/superset/explore_json/?form_data={"slice_id":1344}&force=true', headers={"x-csrftoken": csrf, "content-type": encoder.content_type}, data=encoder, timeout=(10, 60))
+    if response.status_code != 200:
+        raise RuntimeError(f"Superset 请求失败: HTTP {response.status_code}")
+    payload = response.json()
+    data = payload.get("data") if isinstance(payload, dict) else None
+    if not isinstance(data, dict) or not isinstance(data.get("records"), list) or not isinstance(data.get("columns"), list):
+        raise RuntimeError("Superset 响应格式错误")
+    return [dict(zip(data["columns"], row)) if isinstance(row, (list, tuple)) else row for row in data["records"]]
+
+
+def pkg_list(rows: list[dict[str, Any]]) -> list[str]:
+    totals: dict[str, float] = {}
+    for row in rows:
+        pkg = str(row.get("pkg_name") or "").strip()
+        if pkg:
+            totals[pkg] = totals.get(pkg, 0) + float(row.get("clicks") or 0)
+    return sorted(pkg for pkg, clicks in totals.items() if clicks > CLICK_NUM)
+
+
+def appsflyer_rows(packages: list[str], start: str, end: str, browser: Any) -> list[dict[str, Any]]:
+    import requests
+    cookie = browser.get_cookie_str()
+    if not cookie:
+        raise RuntimeError("AppsFlyer cookie 为空")
+    payload = {"dates": {"start": dt.datetime.strptime(start, "%Y%m%d").strftime("%Y-%m-%d"), "end": dt.datetime.strptime(end, "%Y%m%d").strftime("%Y-%m-%d")}, "filters": {"app-id": []}, "view-type": "unified", "localization": {"timezone": "UTC", "currency": "USD"}, "groupings": [{"dimension": "date", "limit": 200}], "summations": ["totals", "others"], "metrics": [{"metric-id": "clicks"}, {"metric-id": "installs-ua"}], "format": "json", "granularity": "days", "implicit-sorting": True}
+    result: list[dict[str, Any]] = []
+    url = "https://hq1.appsflyer.com/platform/dashboard?widget=platform-table__performance:3"
+    for package in packages:
+        app_id = SPECIAL_PKGS.get(package, package)
+        payload["filters"]["app-id"] = [app_id]
+        last_error = None
+        for attempt in range(4):
+            try:
+                response = requests.post(url, data=json.dumps(payload), headers={"content-type": "application/json;charset=UTF-8", "cookie": cookie}, timeout=(10, 60))
+                if response.status_code == 429 or response.status_code >= 500:
+                    raise RuntimeError(f"HTTP {response.status_code}")
+                if response.status_code != 200:
+                    raise RuntimeError(f"HTTP {response.status_code}")
+                data = response.json().get("data")
+                if not isinstance(data, list) or any(not isinstance(item, dict) for item in data):
+                    raise RuntimeError("响应格式错误")
+                for item in data:
+                    item["app_id"] = package
+                result.extend(data)
+                break
+            except Exception as exc:  # do not leak request contents
+                last_error = exc
+                if attempt < 3:
+                    time.sleep(2 ** (attempt + 1))
+        else:
+            raise RuntimeError(f"AppsFlyer 查询失败 pkg={package}: {last_error}")
+    return result
+
+
+def build_report(shareit: list[dict[str, Any]], af: list[dict[str, Any]], packages: list[str]) -> list[dict[str, Any]]:
+    shared: dict[tuple[str, str, str], float] = {}
+    for row in shareit:
+        if str(row.get("attr_platform") or "") != "appsflyer":
+            continue
+        pkg = str(row.get("pkg_name") or "").strip()
+        if not pkg or pkg not in packages:
+            continue
+        raw = str(row.get("dt") or "")[:10].replace("-", "")
+        date = date8(raw)
+        shared[(date, "appsflyer", pkg)] = shared.get((date, "appsflyer", pkg), 0) + float(row.get("clicks") or 0)
+    af_totals: dict[tuple[str, str], list[float]] = {}
+    for row in af:
+        raw = str(row.get("date") or "")[:10].replace("-", "")
+        date = date8(raw)
+        pkg = str(row.get("app_id") or "").strip()
+        key = (date, pkg)
+        values = af_totals.setdefault(key, [0.0, 0.0])
+        values[0] += float(row.get("filtersGranularityMetricIdClicksPeriod", row.get("clicks", 0)) or 0)
+        values[1] += float(row.get("attributionSourceAppsflyerFiltersGranularityMetricIdInstallsUaPeriod", row.get("installs_ua", 0)) or 0)
+    result = []
+    for (date, platform, pkg), share_clicks in shared.items():
+        af_clicks, installs = af_totals.get((date, pkg), [0.0, 0.0])
+        diff = share_clicks - af_clicks
+        result.append({"dt": date, "attr_platform": platform, "pkg_name": pkg, "shareit_clicks": finite_number(share_clicks, "shareit_clicks"), "af_clicks": finite_number(af_clicks, "af_clicks"), "installs_ua": finite_number(installs, "installs_ua"), "clicks_diff": finite_number(diff, "clicks_diff"), "clicks_diff_rate": None if share_clicks == 0 else finite_number(round(diff / share_clicks, 3), "clicks_diff_rate", True)})
+    result.sort(key=lambda row: (row["dt"], row["shareit_clicks"]), reverse=True)
+    if len({(r["dt"], r["pkg_name"]) for r in result}) != len(result):
+        raise RuntimeError("报告存在重复业务键")
+    return result
+
+
+def validate_rows(rows: list[dict[str, Any]], start: str, end: str) -> None:
+    for row in rows:
+        if list(row) != REPORT_COLUMNS or date8(row["dt"]) < start or date8(row["dt"]) > end or not str(row["pkg_name"]).strip():
+            raise ValueError("报告字段、日期或包名不合法")
+        for field in REPORT_COLUMNS[3:]:
+            finite_number(row[field], field, field == "clicks_diff_rate")
+
+
+def make_files(rows: list[dict[str, Any]], dates: list[str], out: Path) -> None:
+    out.mkdir(parents=True, exist_ok=True)
+    for date in dates:
+        content = [row for row in rows if row["dt"] == date]
+        target = out / f"{date}.json"
+        target.write_text(json.dumps(content, ensure_ascii=False, allow_nan=False, separators=(",", ":")) + "\n", encoding="utf-8")
+    index = {"schema_version": 1, "generated_at": dt.datetime.now().astimezone().isoformat(timespec="seconds"), "files": [{"date": date, "path": f"data/{date}.json"} for date in sorted(dates, reverse=True)]}
+    (out / "index.json").write_text(json.dumps(index, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
+
+
+def sensitive_check(paths: list[Path]) -> None:
+    pattern = re.compile(r"cookie|authorization|password|api[_-]?key|token", re.I)
+    for path in paths:
+        if path.name.startswith(".env") or path.suffix in {".db", ".log"}:
+            raise RuntimeError(f"禁止发布敏感文件: {path.name}")
+        if path.suffix in {".json", ".html", ".md"} and pattern.search(path.read_text(encoding="utf-8")):
+            raise RuntimeError(f"文件疑似包含敏感字段: {path.name}")
+
+
+def publish_files(repo: Path, generated: Path, dates: list[str]) -> None:
+    data = repo / "data"
+    data.mkdir(exist_ok=True)
+    for date in dates:
+        source = generated / f"{date}.json"
+        tmp = data / f".{date}.tmp"
+        shutil.copyfile(source, tmp)
+        os.replace(tmp, data / f"{date}.json")
+    os.replace(generated / "index.json", data / "index.json")
+
+
+def git_run(repo: Path, args: list[str]) -> subprocess.CompletedProcess[str]:
+    return subprocess.run(["git", *args], cwd=repo, text=True, capture_output=True, timeout=120, check=False)
+
+
+def git_publish(repo: Path, start: str, end: str, no_push: bool) -> str:
+    if not (repo / ".git").exists():
+        raise RuntimeError("仓库尚未初始化；请先执行 git init 并配置 remote")
+    status = git_run(repo, ["status", "--porcelain"]).stdout.strip().splitlines()
+    allowed = {"README.md", ".gitignore", "index.html", "scripts/af_click_gap_pipeline.py", "launchd/com.roasbloom.af-click-gap.plist.template"}
+    allowed.update({f"data/{date}.json" for date in (repo / "data").glob("*.json")})
+    unexpected = [line for line in status if line[3:] not in allowed]
+    if unexpected:
+        raise RuntimeError("工作区已有未预期改动: " + ", ".join(unexpected))
+    paths = ["README.md", ".gitignore", "index.html", "scripts/af_click_gap_pipeline.py", "launchd/com.roasbloom.af-click-gap.plist.template", "data"]
+    add = git_run(repo, ["add", *paths])
+    if add.returncode:
+        raise RuntimeError("git add 失败: " + add.stderr.strip())
+    staged = git_run(repo, ["diff", "--cached", "--name-only"])
+    if staged.returncode or not staged.stdout.strip():
+        raise RuntimeError("没有可提交的文件")
+    commit = git_run(repo, ["commit", "-m", f"feat: publish click gap {start}-{end}"])
+    if commit.returncode:
+        raise RuntimeError("git commit 失败: " + commit.stderr.strip())
+    commit_hash = git_run(repo, ["rev-parse", "HEAD"]).stdout.strip()
+    if no_push:
+        return commit_hash + " (未 push)"
+    remote = git_run(repo, ["remote", "-v"])
+    if remote.returncode or not remote.stdout.strip():
+        raise RuntimeError("未配置 Git remote；本地 commit 已保留")
+    push = git_run(repo, ["push"])
+    if push.returncode:
+        raise RuntimeError("git push 失败（本地 commit 已保留）: " + push.stderr.strip())
+    return commit_hash + " (已 push)"
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description="Extract and publish AppsFlyer click-gap JSON")
+    parser.add_argument("--start-date", help="补数开始日期 YYYY-MM-DD")
+    parser.add_argument("--end-date", help="补数结束日期 YYYY-MM-DD")
+    parser.add_argument("--dry-run", action="store_true", help="只提取并校验，不写正式文件、数据库或 Git")
+    group = parser.add_mutually_exclusive_group()
+    group.add_argument("--no-push", action="store_true", help="写文件和数据库但不 push")
+    group.add_argument("--push", action="store_true", help="写文件、数据库并 commit/push（默认）")
+    parser.add_argument("--db-path", default=DB_PATH, help=argparse.SUPPRESS)
+    return parser.parse_args()
+
+
+def main() -> int:
+    args = parse_args()
+    try:
+        if args.dry_run and (args.no_push or args.push):
+            raise ValueError("--dry-run 不可与 --push/--no-push 同时使用")
+        start, end = window(args.start_date, args.end_date)
+        validate_existing_schema(args.db_path)
+        shareit = superset_rows(start, end)
+        packages = pkg_list(shareit)
+        af: list[dict[str, Any]] = []
+        browser = None
+        if packages:
+            from af_browser import AFBrowser
+            browser = AFBrowser()
+            try:
+                af = appsflyer_rows(packages, start, end, browser)
+            finally:
+                try:
+                    browser.close()
+                except Exception:
+                    LOG.warning("关闭浏览器失败")
+        rows = build_report(shareit, af, packages)
+        validate_rows(rows, start, end)
+        dates = [(dt.datetime.strptime(start, "%Y%m%d").date() + dt.timedelta(days=i)).strftime("%Y%m%d") for i in range((dt.datetime.strptime(end, "%Y%m%d").date() - dt.datetime.strptime(start, "%Y%m%d").date()).days + 1)]
+        repo = Path(__file__).resolve().parents[1]
+        with tempfile.TemporaryDirectory(prefix="af-click-gap-") as temp:
+            generated = Path(temp) / "data"
+            make_files(rows, dates, generated)
+            sensitive_check(list(generated.glob("*")))
+            if args.dry_run:
+                LOG.info("dry-run 成功: %s-%s, %d 行", start, end, len(rows))
+                return 0
+            publish_files(repo, generated, dates)
+        with sqlite3.connect(args.db_path) as conn:
+            conn.execute("BEGIN")
+            upsert_report(conn, rows, start, end)
+            conn.commit()
+        if args.no_push:
+            LOG.info("已发布本地文件和 SQLite，未执行 Git")
+        else:
+            result = git_publish(repo, start, end, False)
+            LOG.info("Git 结果: %s", result)
+        return 0
+    except ValueError as exc:
+        LOG.error("输入失败: %s", exc)
+        return 2
+    except Exception as exc:
+        LOG.error("任务失败: %s", exc)
+        return 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())