feat: auto-heal — gateway down auto-restart via cron every 5min

agents_health_check.py

@@ -120,6 +120,24 @@ def run():
     # ── XMPP 通道健康采集 ──
     _collect_xmpp_health(now)
 
+    # ── 自愈：Gateway 宕机自动重启 ──
+    gateway_ok = any(r["name"] == "zhiwei_gateway" and r["health"]["ok"] for r in results)
+    if not gateway_ok:
+        try:
+            import subprocess as _sp
+            _sp.run(["sudo", "-n", "pkill", "-9", "-f", "position-analyst.*gateway"],
+                   capture_output=True, timeout=5)
+            _sp.run(["rm", "-f", "/home/hmo/.hermes/profiles/position-analyst/gateway.lock",
+                    "/home/hmo/.hermes/profiles/position-analyst/gateway.pid"])
+            _sp.Popen(["/home/hmo/hermes-agent/.venv/bin/python", "-m", "hermes_cli.main",
+                      "-p", "position-analyst", "gateway", "run", "--replace"],
+                     stdout=open("/tmp/hermes_gw_auto.log", "a"),
+                     stderr=open("/tmp/hermes_gw_auto.log", "a"),
+                     start_new_session=True)
+            print(f"[{now.strftime('%H:%M')}] Auto-heal: Gateway DOWN → restarted")
+        except Exception as e:
+            print(f"[{now.strftime('%H:%M')}] Auto-heal error: {e}")
+
     # Handle issues
     if issues:
         with open(TODO_FILE, "a", encoding="utf-8") as f: