butubb
16e3e162b2
用户反馈：一条任务覆盖 13 台设备，每台成功都推一条，群里被刷屏；批次汇总里
「任务：抖音养号」与标题重复、「样本：抖音养号」毫无信息量、「stopped」还是英文。
- task_manager：
· **多设备批次不发 `task.device.success`**（批次汇总里已有成功台数；单台任务照发）；
失败仍逐台发——那是少数，且要知道是哪台；
· `_BatchTracker` 收集失败设备（`名字（型号）：原因`，最多 5 条）并加进批次汇总；
· 设备级事件带 `model`：型号取自**设备池快照**（批次开始时一次查好带下去），
不是 worker 每次连接现采的（那次 `d.info()` 可能超时 → 空），也不查库；
- notifier：
· 聚合窗口口径修正：事件声明 `agg_window=0`（批次结束/服务启停/备份恢复等低频
高危事件）**不再被 hook 的窗口拖住**——原先事件级设置完全失效；
· 「样本」行只在真的合并了多条(>1)时出现，且按**设备**维度写（合并多台时写
任务名每条都一样）；与标题重复的字段行（「任务：x」）不再重复渲染；
· 设备名与地址同时存在时**只显示名字**（IP 是给日志看的）；
· 补 `stopped`/`failed_devices` 中文标签（原先直接显示英文 key）；
- notify_events：批次事件补 `skipped`/`failed_devices` 字段，设备事件补 `model`。
实测（dev 真机 + 假接收端）：
批次 2 台 → 只收到 batch.started/finished，无 device.success，汇总 总数2/成功1/跳过1；
单台任务 → 收到 device.success，显示「设备：cs1 / 型号：M2010J19SC」不含 IP。
文档：doc/NOTIFY.md §3（多设备批次怎么发 + 批次消息样例）、§6（窗口口径与注意事项）。
209 lines
11 KiB
Python
209 lines
11 KiB
Python
"""通知事件目录（**唯一真相**）。
|
||
|
||
约定：
|
||
- **key 点分层**：`模块.对象.动作`（如 `task.device.failed`），支持通配订阅
|
||
（`task.*`、`device.*`、`*`），匹配规则见 `match()`。
|
||
- **登记 ≠ 会发**：所有事件 `default` 一律 False —— 用户在「系统 → 通知」里
|
||
按 webhook 勾选才会推送。`recommend` 只用于界面高亮"建议开启"，不改默认值。
|
||
- **聚合**：`agg_window` 秒内的同 `(hook, event, agg_key 取值)` 事件合并成一条
|
||
（保留前几个样本）；`agg_window=0` = 不聚合，立即发（低频高危事件用它）。
|
||
- **fields**：该事件保证携带的上下文字段，供前端画字段表、也是模板占位符白名单。
|
||
|
||
新增事件：在 `EVENTS` 里加一条 + 在触发点调 `notifier.notify(key, **fields)`，
|
||
并同步 `doc/NOTIFY.md` 的事件表（见 doc/README.md 的文档同步红线）。
|
||
"""
|
||
from collections import namedtuple
|
||
|
||
# key, label, category, default, recommend, agg_window, agg_key, fields, desc
|
||
EventDef = namedtuple(
|
||
"EventDef",
|
||
"key label category default recommend agg_window agg_key fields desc")
|
||
|
||
|
||
def _e(key, label, category, fields, desc="", agg_window=30, agg_key=None,
|
||
recommend=False):
|
||
"""构造一条事件定义（default 恒为 False：登记不等于推送）。"""
|
||
return EventDef(key, label, category, False, recommend, agg_window,
|
||
agg_key, tuple(fields), desc)
|
||
|
||
|
||
# 聚合键：按"任务批次"合并（同一次任务的多台设备结果合成一条）
|
||
_BY_JOB = ("job_id",)
|
||
# 按"设备"合并（一台设备的重复抖动合成一条）
|
||
_BY_SERIAL = ("serial",)
|
||
|
||
EVENTS = [
|
||
# ---------------- 任务批次 ----------------
|
||
_e("task.batch.started", "任务批次开始", "任务批次",
|
||
["job_id", "job_name", "task_type", "device_count", "serials_preview"],
|
||
"一次任务开始铺开到 N 台设备", agg_window=0, recommend=True),
|
||
_e("task.batch.finished", "任务批次结束", "任务批次",
|
||
["job_id", "job_name", "total", "success", "failed", "stopped", "skipped",
|
||
"failed_devices", "duration_s"],
|
||
"整批跑完（成功/失败/停止台数 + 耗时；有失败时列出失败设备与型号）",
|
||
agg_window=0, recommend=True),
|
||
_e("task.batch.no_device", "任务无可用设备", "任务批次",
|
||
["job_id", "job_name", "target"],
|
||
"触发时一台可用设备都没有（任务空跑）", agg_window=0, recommend=True),
|
||
_e("task.batch.unknown_type", "任务类型不存在", "任务批次",
|
||
["job_id", "job_name", "task_type"],
|
||
"任务类型已从代码里删掉，任务永远不会执行", agg_window=0, recommend=True),
|
||
_e("task.cron.stopped", "定时停止任务", "任务批次",
|
||
["job_id", "job_name", "stopped_count", "serials"],
|
||
"cron_stop 到点，停掉了正在跑的设备", agg_window=0),
|
||
|
||
# ---------------- 任务 · 单设备（权威结论点） ----------------
|
||
_e("task.device.success", "设备任务成功", "任务·单设备",
|
||
["serial", "device_name", "model", "job_id", "job_name", "attempt", "duration_s"],
|
||
"某台设备上的任务最终成功。**只在目标设备只有 1 台时发送**——多设备批次看"
|
||
"「任务批次结束」就够了，13 台设备就是 13 条刷屏", agg_key=_BY_JOB),
|
||
|
||
_e("task.device.failed", "设备任务失败", "任务·单设备",
|
||
["serial", "device_name", "model", "job_id", "job_name", "attempts",
|
||
"cause", "msg"],
|
||
"某台设备上的任务最终失败（带真实原因与型号）", agg_key=_BY_JOB, recommend=True),
|
||
_e("task.device.offline", "设备离线放弃", "任务·单设备",
|
||
["serial", "device_name", "model", "job_id", "job_name", "error"],
|
||
"设备离线，不重试直接失败", agg_key=_BY_JOB, recommend=True),
|
||
_e("task.device.error", "单次尝试异常", "任务·单设备",
|
||
["serial", "device_name", "model", "job_name", "attempt", "error"],
|
||
"某次尝试抛异常（后面还会重试，噪音较大）", agg_key=_BY_JOB),
|
||
_e("task.device.retry", "设备任务重试", "任务·单设备",
|
||
["serial", "device_name", "job_name", "attempt", "next_attempt", "delay_s", "reason"],
|
||
"即将重试", agg_key=_BY_JOB),
|
||
_e("task.device.stopped", "设备任务被停止", "任务·单设备",
|
||
["serial", "device_name", "model", "job_name", "attempt", "phase"],
|
||
"用户手动停止 / cron 停止", agg_key=_BY_JOB),
|
||
_e("task.device.preempted", "设备被抢占", "任务·单设备",
|
||
["serial", "device_name", "job_name", "preempted_job_id", "preempted_job_name"],
|
||
"本任务抢占了该设备上正在跑的其他任务", agg_key=_BY_JOB, recommend=True),
|
||
_e("task.device.preempt_timeout", "抢占超时跳过", "任务·单设备",
|
||
["serial", "device_name", "job_name", "preempted_job_id"],
|
||
"等被抢占任务退出超时，本设备放弃执行", agg_window=0, recommend=True),
|
||
_e("task.device.released", "抢占结束归还", "任务·单设备",
|
||
["serial", "device_name", "job_name", "preempted_job_id", "returned", "reason"],
|
||
"抢占结束时重新拉起被抢占的任务", agg_key=_BY_JOB),
|
||
|
||
# ---------------- Worker 状态机（单次 attempt 级，默认关） ----------------
|
||
_e("worker.connected", "设备已连接", "Worker",
|
||
["serial", "model", "remote_adb_url"],
|
||
"worker 连上设备（单次尝试级，噪音大）"),
|
||
_e("worker.attempt.done", "单次执行完成", "Worker",
|
||
["serial", "task_type", "duration_s", "max_duration_hit"],
|
||
"单次 attempt 正常跑完（不代表任务最终成功）"),
|
||
_e("worker.attempt.error", "单次执行出错", "Worker",
|
||
["serial", "task_type", "error", "transient"],
|
||
"单次 attempt 抛异常"),
|
||
|
||
# ---------------- 业务（任务内容） ----------------
|
||
_e("task.selector.invalid", "选择器连续失效", "业务",
|
||
["serial", "device_name", "selector", "miss_count"],
|
||
"某选择器连续 10 次未命中——任务可能显示成功但什么都没做", agg_window=0,
|
||
recommend=True),
|
||
|
||
# ---------------- 设备 ----------------
|
||
_e("device.online", "设备恢复在线", "设备",
|
||
["serials", "count"],
|
||
"原本断联的设备又能连上了", agg_window=0, recommend=True),
|
||
_e("device.offline", "设备断联", "设备",
|
||
["serials", "count", "devices"],
|
||
"设备池里的设备连不上了", agg_window=0, recommend=True),
|
||
_e("device.discovered", "发现新设备", "设备",
|
||
["serials", "count"],
|
||
"扫描到尚未在池内的设备（待认领）", agg_window=0),
|
||
_e("device.claimed", "设备自动认领", "设备",
|
||
["pairs", "count"],
|
||
"指纹匹配成功，自动把旧记录迁到新地址", agg_window=0),
|
||
_e("device.heartbeat_timeout", "心跳超时", "设备",
|
||
["serial", "device_name", "timeout_s", "task_job", "model"],
|
||
"设备卡死（长时间没心跳），任务可能已中断", agg_window=0, recommend=True),
|
||
|
||
# ---------------- 应用安装 ----------------
|
||
_e("apk.install.started", "应用安装开始", "安装",
|
||
["apk_id", "apk_name", "package_name", "total"],
|
||
"开始往 N 台设备推装", agg_window=0),
|
||
_e("apk.install.finished", "应用安装完成", "安装",
|
||
["apk_name", "success", "failed", "skipped", "total", "failed_items"],
|
||
"批量安装结束（含失败台数与原因）", agg_window=0, recommend=True),
|
||
|
||
# ---------------- 系统 ----------------
|
||
_e("system.backup.exported", "备份已导出", "系统",
|
||
["filename", "size", "tables", "include_apk", "user"],
|
||
"有人导出了整库备份", agg_window=0, recommend=True),
|
||
_e("system.backup.imported", "备份导入已挂起", "系统",
|
||
["token_prefix", "env_label", "force"],
|
||
"上传了备份并确认导入（重启后生效）", agg_window=0),
|
||
_e("system.backup.restored", "备份已恢复", "系统",
|
||
["applied_rows", "schema_version"],
|
||
"启动时应用了待恢复的备份（所以这条是重启后才发）", agg_window=0,
|
||
recommend=True),
|
||
_e("system.backup.restore_failed", "备份恢复失败", "系统",
|
||
["error", "fail_dir"],
|
||
"恢复校验没过，已搁置（数据未被改动）", agg_window=0, recommend=True),
|
||
_e("service.started", "服务已启动", "系统",
|
||
["version", "env", "db_target", "device_count", "job_count"],
|
||
"平台进程起来了", agg_window=0, recommend=True),
|
||
_e("service.stopping", "服务正在停止", "系统",
|
||
["uptime_s"],
|
||
"平台进程收到退出信号", agg_window=0),
|
||
_e("user.login", "用户登录", "系统",
|
||
["username"],
|
||
"有人登录了后台", agg_window=0),
|
||
|
||
# ---------------- AI ----------------
|
||
_e("ai.audit.finished", "经验巡检完成", "AI",
|
||
["reviewed", "suggested", "summary"],
|
||
"每日经验库巡检跑完（或手动触发）", agg_window=0, recommend=True),
|
||
_e("ai.audit.failed", "经验巡检异常", "AI",
|
||
["error"],
|
||
"巡检过程中报错", agg_window=0),
|
||
_e("ai.audit.skipped", "经验巡检跳过", "AI",
|
||
["reason"],
|
||
"没配模型 Key 等原因跳过巡检", agg_window=0),
|
||
|
||
# ---------------- 其它 ----------------
|
||
_e("notify.test", "测试通知", "其它",
|
||
["hook_name", "operator"],
|
||
"「发送测试」按钮专用", agg_window=0),
|
||
]
|
||
|
||
BY_KEY = {e.key: e for e in EVENTS}
|
||
|
||
# 界面上的分组顺序
|
||
CATEGORIES = ("任务批次", "任务·单设备", "业务", "设备", "安装", "系统", "AI",
|
||
"Worker", "其它")
|
||
|
||
|
||
def get(key):
|
||
"""按 key 取事件定义；未知 key 返回 None（未知事件会被 notify 丢弃）。"""
|
||
return BY_KEY.get(key)
|
||
|
||
|
||
def match(pattern, key):
|
||
"""事件订阅匹配。
|
||
|
||
- `*` → 全部
|
||
- `task.*` → task 下所有（含 task.device.failed）
|
||
- `task.device.*` → 任意层级前缀匹配
|
||
- `task.device.failed` → 精确
|
||
"""
|
||
if not pattern:
|
||
return False
|
||
if pattern == "*" or pattern == key:
|
||
return True
|
||
if pattern.endswith(".*"):
|
||
return key.startswith(pattern[:-1]) # "task.*" → "task."
|
||
return False
|
||
|
||
|
||
def list_events():
|
||
"""给接口/前端用的事件目录（按 CATEGORIES 排序）。"""
|
||
order = {c: i for i, c in enumerate(CATEGORIES)}
|
||
rows = [{"key": e.key, "label": e.label, "category": e.category,
|
||
"default": e.default, "recommend": e.recommend,
|
||
"agg_window": e.agg_window, "agg_key": list(e.agg_key or ()),
|
||
"fields": list(e.fields), "desc": e.desc}
|
||
for e in EVENTS]
|
||
rows.sort(key=lambda r: (order.get(r["category"], 99), r["key"]))
|
||
return rows
|