fix(bot+auto_heal): stop auto_heal from killing bot during slow LLM calls

Root cause of 'no response': auto_heal restart_zhiwei_bot fired whenever
inbound>0 + outbound=0 + errors>0 — which is exactly the normal state while
the bot waits for a slow LLM call (agent tool-use turns take 1-10 min).
Restart killed the in-flight LLM call -> user never got reply -> next cron
cycle saw same state -> restart again. Death loop (fired 22:30,22:35,22:40,
22:45,22:50,23:35).

Fixes:
- CALL_HERMES_TIMEOUT 180s -> 600s (agent tool calls need minutes)
- health(): parse last_inbound/outbound timestamps from journal
- auto_heal: restart bot ONLY if last inbound >600s old with no outbound
  since (truly stuck), plus 600s bot-restart cooldown
- slow-but-normal state now logs bot_busy_not_stuck instead of killing

deploy/bot/xmpp_agent_core.py

@@ -52,8 +52,8 @@ ACK_DELAY = 120  # 真卡死才提示（普通 LLM 冷启动 20-100s 不应触
 GATEWAY_URL = ""
 GATEWAY_API_KEY = ""
 GATEWAY_SESSION_ID = ""
-GATEWAY_DEADLINE_SECONDS = 180
+GATEWAY_DEADLINE_SECONDS = 600
-CALL_HERMES_TIMEOUT = 180
+CALL_HERMES_TIMEOUT = 600  # agent 带工具调用（读文件/查数据）需几分钟
 FALLBACK_REPLY = "请稍等，我在处理..."
 
 # ── 图片 OCR 配置（截图消息 → SenseNova vision）──

xmpp_logger.py

@@ -292,11 +292,27 @@ def health():
         inbound = [l for l in lines if "📩 收到" in l]
         outbound = [l for l in lines if "📤 发送" in l or "📤" in l]
         errors = [l for l in lines if "ERROR" in l or "TimeoutError" in l or "timed out" in l]
+
+        # 解析最后一条 inbound 的时间（行首 python logging 时间戳 "2026-07-19 23:31:07,342"）
+        import re as _re2
+        def _line_epoch(line):
+            m = _re2.search(r"(\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2})", line)
+            if not m:
+                return 0
+            try:
+                return datetime.strptime(m.group(1), "%Y-%m-%d %H:%M:%S").timestamp()
+            except Exception:
+                return 0
+
+        last_inbound_ts = max((_line_epoch(l) for l in inbound), default=0)
+        last_outbound_ts = max((_line_epoch(l) for l in outbound), default=0)
         result["bot_activity"] = {
             "inbound": len(inbound), "outbound": len(outbound), "errors": len(errors),
             "last_error": errors[-1][:250] if errors else None,
             "last_inbound": inbound[-1][:250] if inbound else None,
             "last_outbound": outbound[-1][:200] if outbound else None,
+            "last_inbound_age_sec": int(now_epoch - last_inbound_ts) if last_inbound_ts else -1,
+            "last_outbound_age_sec": int(now_epoch - last_outbound_ts) if last_outbound_ts else -1,
         }
         if errors and not outbound: result["status"] = "degraded"
         if inbound and not outbound: result["status"] = "degraded"
@@ -494,16 +510,44 @@ def auto_heal():
                                     "success": False,
                                     "detail": str(e)[:100]})
 
-    # 2. Bot 无出站 → 重启知微 Bot
+    # 2. Bot 真卡死才重启：最后一条 inbound 超过 10 分钟仍无出站，才算死（不是 LLM 慢）
+    # LLM 正常冷启动/工具调用需要 1-10 分钟，inbound 后立即重启会杀掉进行中的调用。
     ba = h.get("bot_activity", {})
-    if ba.get("inbound", 0) > 0 and ba.get("outbound", 0) == 0 and ba.get("errors", 0) > 0:
+    inbound_age = ba.get("last_inbound_age_sec", -1)
+    outbound_age = ba.get("last_outbound_age_sec", -1)
+    bot_truly_stuck = (
+        inbound_age > 600  # 最后一条用户消息已超过10分钟
+        and (outbound_age < 0 or outbound_age > inbound_age - 600)  # 且其间没有任何出站
+        and ba.get("inbound", 0) > 0
+    )
+    if bot_truly_stuck:
+        # cooldown: bot 重启不频繁于10分钟一次
         try:
-            r = _sp.run(["sudo", "-n", "systemctl", "restart", "xmpp-zhiwei"], capture_output=True, timeout=30, text=True)
+            last_bot_restart = 0
-            actions.append({"action": "restart_zhiwei_bot", "target": "xmpp-zhiwei",
+            bf = LOG_DIR / "last_bot_restart.txt"
-                          "success": r.returncode == 0, "detail": "restarted" if r.returncode == 0 else r.stderr[:100]})
+            if bf.exists():
-        except Exception as e:
+                last_bot_restart = float(bf.read_text().strip() or 0)
-            actions.append({"action": "restart_zhiwei_bot", "target": "xmpp-zhiwei",
+        except Exception:
-                          "success": False, "detail": str(e)[:100]})
+            last_bot_restart = 0
+        bot_elapsed = _time.time() - last_bot_restart
+        if bot_elapsed < 600:
+            actions.append({"action": "skip_bot_restart",
+                            "reason": f"bot restart cooldown {int(bot_elapsed)}s < 600s"})
+        else:
+            try:
+                _sp.Popen(["sudo", "-n", "systemctl", "restart", "xmpp-zhiwei"],
+                          stdout=_sp.DEVNULL, stderr=_sp.DEVNULL, start_new_session=True)
+                (LOG_DIR / "last_bot_restart.txt").write_text(str(_time.time()))
+                actions.append({"action": "restart_zhiwei_bot", "target": "xmpp-zhiwei",
+                              "success": True, "detail": "restart triggered (async)"})
+            except Exception as e:
+                actions.append({"action": "restart_zhiwei_bot", "target": "xmpp-zhiwei",
+                              "success": False, "detail": str(e)[:100]})
+    elif ba.get("inbound", 0) > 0 and ba.get("outbound", 0) == 0 and ba.get("errors", 0) > 0:
+        # inbound 后仍在处理窗口内（LLM 慢但正常），不动 bot
+        actions.append({"action": "bot_busy_not_stuck",
+                        "inbound_age_sec": inbound_age,
+                        "reason": "inbound < 10min ago — LLM still processing, no restart"})
 
     # 3. ejabberd → 重启容器
     if h.get("ejabberd", "") and "Up" not in str(h["ejabberd"]):