fix: health check PID now advisory-only, remote services skip local port, add ?§ to task entries

gateway/scripts/agents_health_check.py

@@ -30,6 +30,7 @@ for svc in _SR:
         "fix_script": svc["script"],
         "fix_args": svc["args"],
         "fix_cwd": svc["workdir"],
+        "remote": svc.get("remote", False),
     })
 
 TODO_FILE = os.path.join(TEMP, "health_todos.jsonl")
@@ -126,13 +127,17 @@ def main():
         entry["pid"] = pid if pid_ok else 0
         entry["pid_ok"] = pid_ok
 
-        # 2. Port check
-        port_ok = port_open(svc["port"])
+        # 2. Port check（远程服务跳过本地端口检测）
+        is_remote = svc.get("remote", False)
+        if is_remote:
+            port_ok = True  # 远程服务不检查本地端口
+        else:
+            port_ok = port_open(svc["port"])
         entry["port_ok"] = port_ok
 
-        # 3. HTTP health check
+        # 3. HTTP health check（远程服务直接用 HTTP 判断）
         http_msg = ""
-        if port_ok:
+        if port_ok or is_remote:
             ok, detail = http_check(svc["health_url"])
             # accept 401/403 as healthy if configured
             if not ok and svc.get("accept_401") and ("401" in detail or "403" in detail):
@@ -146,20 +151,21 @@ def main():
             entry["http_ok"] = False
             entry["http_detail"] = "port_closed"
 
-        healthy = (pid_ok or not pid_file) and port_ok and entry.get("http_ok", False)
+        # 健康判定：port+HTTP 都 OK → 健康；PID 仅作为辅助信号，不阻塞判定
+        primary_ok = port_ok and entry.get("http_ok", False)
+        entry["status"] = "ok" if primary_ok else "fail"
 
-        if healthy:
-            entry["status"] = "ok"
+        if primary_ok:
+            if not pid_ok:
+                entry["pid_warning"] = True  # PID 异常但服务在正常运行
             report["summary"]["ok"] += 1
         else:
-            entry["status"] = "fail"
             report["summary"]["fail"] += 1
             dirty = True
             # 收集具体失败原因
             reasons = []
-            if not pid_ok: reasons.append("pid_dead")
-            if not port_ok: reasons.append("port_closed")
-            if not entry.get("http_ok") and port_ok: reasons.append(f"http_{http_msg}")
+            if not port_ok and not is_remote: reasons.append("port_closed")
+            if not entry.get("http_ok"): reasons.append(f"http_{http_msg}")
             issue = f"异常: {' + '.join(reasons)}"
             write_todo(name, issue, svc["fix_script"], svc.get("fix_args", []), svc["fix_cwd"])
 
@@ -170,7 +176,9 @@ def main():
         # 只输出异常服务到 stdout（silent-by-default）
         for svc in report["services"]:
             if svc["status"] != "ok":
-                print(f"[FAIL] {svc['name']}: PID={svc['pid_ok']} PORT={svc['port_ok']} HTTP={svc.get('http_detail','?')}")
+                print(f"[FAIL] {svc['name']}: PORT={svc['port_ok']} HTTP={svc.get('http_detail','?')}")
+            elif svc.get("pid_warning"):
+                print(f"[WARN] {svc['name']}: PID异常但服务正常")
     else:
         # 全正常→完全静默（silent-by-default）
         pass

gateway/scripts/service_registry.py

@@ -63,6 +63,7 @@ SERVICES = [
         "health_url": "http://192.168.1.246:8642/v1/health",
         "accept_401": True,      # may need auth
         "depends_on": ["ejabberd"],
+        "remote": True,             # 服务运行在 Linux 246，不检查本地端口
         "log_files": lambda: [],
         "description": "Hermes Gateway (mohe@yoin.fun)",
     },
@@ -76,6 +77,7 @@ SERVICES = [
         "health_url": "http://192.168.1.246:8643/v1/health",
         "accept_401": True,
         "depends_on": ["ejabberd"],
+        "remote": True,             # 服务运行在 Linux 246，不检查本地端口
         "log_files": lambda: [],
         "description": "Hermes Gateway (zhiwei@yoin.fun)",
     },

gateway/scripts/specs/daily_health.json

@@ -0,0 +1,46 @@
+{
+  "module": "daily_health",
+  "version": "1.0",
+  "purpose": "Tier 2 全面健康检查（每日 08:00）。在 Tier 1 基础上增加：磁盘空间检查、定时任务存活检查、数据新鲜度、跨服务依赖链检查、看门狗日志新鲜度。",
+  "ui_location": "F Tab → 定时任务 → 日报健康检查(每日)",
+
+  "human_help": {
+    "title": "Tier 2 每日健康检查",
+    "description": [
+      "每天早上 8:00 由 Windows Task Scheduler 触发，生成全面系统健康日报。",
+      "检查范围：服务状态（PID+端口+HTTP）、磁盘空间（C盘）、定时任务存活、看门狗日志新鲜度。",
+      "输出结构化 JSON 报告供 Dashboard 和人工审阅。",
+      "发现异常时生成推荐动作清单。"
+    ],
+    "usage": [
+      "报告文件：gateway/temp/last_daily_health.json",
+      "日志：gateway/logs/daily_health_report.log",
+      "Dashboard 通过 /api/monitor 读取报告显示 Tier2 统计"
+    ],
+    "troubleshooting": [
+      "如果定时任务未运行：检查 Windows Task Scheduler → agents-daily-health",
+      "磁盘警告阈值：剩余 < 10GB 警告，< 2GB 严重",
+      "看门狗日志超过 1 小时未更新 → 可能看门狗已死"
+    ],
+    "related": "依赖 service_registry.py（服务定义）; 参考 agents_health_check.py（Tier1 检查逻辑）"
+  },
+
+  "ai_spec": {
+    "dependencies": [
+      "service_registry.py — 服务定义",
+      "Python 3.10",
+      "Windows Task Scheduler — 每天 08:00 触发",
+      "wmic — Windows 磁盘空间查询"
+    ],
+    "constraints": [
+      "只在 Windows 上运行（使用 tasklist/netstat/wmic）",
+      "磁盘检查仅针对 C: 盘",
+      "定时任务检查通过 schtasks /Query 验证"
+    ],
+    "related_files": [
+      "gateway/scripts/agents_daily_health.py — 本脚本",
+      "gateway/scripts/service_registry.py — 服务注册表",
+      "gateway/scripts/specs/daily_health.json — 本 spec"
+    ]
+  }
+}

gateway/scripts/specs/health_check.json

@@ -0,0 +1,48 @@
+{
+  "module": "health_check",
+  "version": "1.0",
+  "purpose": "Tier 1 快速健康检查（每5分钟）。对 service_registry 中所有注册服务做三合一检查：PID存活 → 端口监听 → HTTP /health。异常时写 TODO 给 self_todo_executor 消费。正常时静默。",
+  "ui_location": "F Tab → 定时任务 → 健康检查(5min)",
+
+  "human_help": {
+    "title": "Tier 1 健康检查",
+    "description": [
+      "每 5 分钟由 Windows Task Scheduler 触发，检查所有注册服务的运行状态。",
+      "检查项：端口监听（netstat）+ HTTP /health 端点响应。远程服务跳过本地端口检测。",
+      "PID 检查仅作为辅助信号——端口+HTTP 都正常即判定为健康。",
+      "异常服务会写入 health_todos.jsonl，由自修复执行器消费。",
+      "全正常时静默运行，不输出任何噪音。"
+    ],
+    "usage": [
+      "报告文件：gateway/temp/last_health_check.json",
+      "TODO 文件：gateway/temp/health_todos.jsonl",
+      "日志：gateway/logs/health_check_report.log",
+      "Dashboard 通过 /api/monitor 读取报告显示 Tier1 统计"
+    ],
+    "troubleshooting": [
+      "如果定时任务未运行：检查 Windows Task Scheduler → agents-health-check",
+      "如果报告未更新：检查 C:\Users\hmo\AppData\Local\Programs\Python\Python310\python.exe 是否可用",
+      "如果 TODO 堆积：检查 self_todo_executor 是否正常消费"
+    ],
+    "related": "依赖 service_registry.py（服务定义）; 输出给 self_todo_executor.py（修复消费）; 报告被 dashboard.py 读取"
+  },
+
+  "ai_spec": {
+    "dependencies": [
+      "service_registry.py — 服务定义（名称/端口/health URL/fix_script）",
+      "Python 3.10 (C:\\Users\\hmo\\AppData\\Local\\Programs\\Python\\Python310\\python.exe)",
+      "Windows Task Scheduler — 每5分钟触发"
+    ],
+    "constraints": [
+      "PID 检查不是硬性失败条件——端口+HTTP OK 即视为健康",
+      "远程服务（remote:true）跳过本地 netstat 端口检测，直接用 HTTP 判断",
+      "静默原则：全正常时零输出；仅异常时写 TODO + 打印报告"
+    ],
+    "related_files": [
+      "gateway/scripts/agents_health_check.py — 本脚本",
+      "gateway/scripts/service_registry.py — 服务注册表",
+      "gateway/scripts/self_todo_executor.py — TODO 消费者",
+      "gateway/scripts/specs/health_check.json — 本 spec"
+    ]
+  }
+}

gateway/scripts/specs/todo_executor.json

@@ -0,0 +1,47 @@
+{
+  "module": "todo_executor",
+  "version": "1.0",
+  "purpose": "TODO 自修复执行器（每10分钟）。轮询 health_todos.jsonl 中 status=pending 的条目，执行 fix_script 重启异常服务。成功标记 completed，失败 escalation 到 XMPP 群聊。",
+  "ui_location": "F Tab → 定时任务 → TODO自修复",
+
+  "human_help": {
+    "title": "TODO 自修复执行器",
+    "description": [
+      "每 10 分钟由 Windows Task Scheduler 触发，自动修复健康检查发现的异常。",
+      "读取 health_todos.jsonl 中的 pending 条目，执行对应的 fix_script。",
+      "修复成功 → 标记 completed。",
+      "修复失败 → 通过 XMPP HTTP 桥（:5802/send）发送告警到群聊。",
+      "特殊情况：如果错误信息为 'another instance is already running'，说明服务已在正常运行，标记 completed 跳过。"
+    ],
+    "usage": [
+      "TODO 文件：gateway/temp/health_todos.jsonl（输入+输出）",
+      "日志：gateway/logs/todo_executor.log",
+      "告警通道：HTTP POST → 127.0.0.1:5802/send → XMPP 群聊"
+    ],
+    "troubleshooting": [
+      "如果 TODO 堆积不减少：检查 agents-todo-executor 定时任务是否运行",
+      "如果修复失败：检查 fix_script 路径是否存在（远程服务的 fix_script 为 null，无法自动修复）",
+      "如果告警未发出：检查 xmpp_bot HTTP 桥 :5802/send 是否可达"
+    ],
+    "related": "消费 agents_health_check.py 的 TODO 输出; 通过 xmpp_bot HTTP 桥发送告警"
+  },
+
+  "ai_spec": {
+    "dependencies": [
+      "health_todos.jsonl — TODO 输入（由 agents_health_check.py 写入）",
+      "xmpp_bot HTTP 桥 (:5802/send) — 告警输出通道",
+      "Python 3.10",
+      "Windows Task Scheduler — 每10分钟触发"
+    ],
+    "constraints": [
+      "只修复有 fix_script 的服务（远程服务无本地脚本，无法自动修复）",
+      "already_running 错误视为成功（服务已在运行）",
+      "单次执行超时 30s — 大脚本可能被截断"
+    ],
+    "related_files": [
+      "gateway/scripts/self_todo_executor.py — 本脚本",
+      "gateway/scripts/agents_health_check.py — TODO 生产者",
+      "gateway/scripts/specs/todo_executor.json — 本 spec"
+    ]
+  }
+}

gateway/scripts/templates/dashboard.html

@@ -176,8 +176,9 @@ async function fHealth(){try{
   M.dashboard={n:'管理门户',l:'辅助服务',i:'Dashboard无法访问',spec:'dashboard'};
   M.watchdog={n:'进程看门狗',l:'辅助服务',i:'异常进程无自动恢复'};
   M.ejabberd={n:'XMPP服务器',l:'通信层',i:'所有XMPP连接不可用',spec:'ejabberd'};
-  M['agents-health-check']={n:'健康检查(5min)',l:'定时任务',i:'5分钟级健康检测未运行'};
-  M['agents-daily-health']={n:'日报健康检查(每日)',l:'定时任务',i:'日报统计未运行'};
+  M['agents-health-check']={n:'健康检查(5min)',l:'定时任务',i:'5分钟级健康检测未运行',spec:'health_check'};
+  M['agents-daily-health']={n:'日报健康检查(每日)',l:'定时任务',i:'日报统计未运行',spec:'daily_health'};
+  M['agents-todo-executor']={n:'TODO自修复(10min)',l:'定时任务',i:'服务异常无法自动恢复',spec:'todo_executor'};
   var layerOrder=['通信层','AI网关','辅助服务','定时任务'];
 
   var h='<h2>F 系统健康<span class="help-btn" onclick="showModuleHelp(\'health\',\'human\')" title="使用说明">?</span><span class="help-btn ai" onclick="showModuleHelp(\'health\',\'ai\')" title="AI Spec">§</span></h2>';
@@ -263,11 +264,14 @@ async function fHealth(){try{
     h+='<div class=section><h2 style=margin-bottom:8px>定时任务</h2>';
     for(var i=0;i<mD.tasks.length;i++){var t=mD.tasks[i];
       var m=M[t.name]||{n:t.name};
+      var ms=m.spec||'';
       var tk=t.status==='ok'||t.status==='timer_ok'||t.status==='cron_ok'||t.status==='schtasks_ok';
       var tn=t.status==='not_deployed';
       h+='<div style="display:flex;align-items:center;gap:8px;padding:4px 8px;margin:2px 0">';
       h+='<span class="sd '+(tk?'g':tn?'y':'r')+'"></span>';
       h+='<span style=flex:1>'+esc(m.n)+'</span>';
+      if(ms){h+='<span class="help-btn" onclick="showModuleHelp(\''+ms+'\',\'human\')" title="使用说明">?</span>';
+      h+='<span class="help-btn ai" onclick="showModuleHelp(\''+ms+'\',\'ai\')" title="AI Spec">§</span>';}
       h+='<span class="tag '+(tk?'g':tn?'y':'r')+'">'+(tk?'正常':tn?'未部署':'异常')+'</span></div>';}
     h+='</div>';}