http://192.168.110.207:8090/        ← 207 服务入口页（所有可点服务的卡片）

探活定位 → docker → litellm → Dify compose → Langfuse → Xinference
        → Higress → 监控(hw8) → k3s + GitLab demo → 等 45s → 再探活

# /etc/systemd/system/host-portal.service
[Unit]
Description=Host service portal
After=network-online.target

[Service]
ExecStart=/usr/bin/python3 -m http.server 8090 --bind 0.0.0.0 --directory /var/www/portal
Restart=on-failure
RestartSec=2

[Install]
WantedBy=multi-user.target

# 落地
mkdir -p /var/www/portal
cp _portal_8090_index.html /var/www/portal/index.html
systemctl daemon-reload
systemctl enable --now host-portal
systemctl is-active host-portal          # active
curl -s -o /dev/null -w '%{http_code}\n' http://192.168.110.207:8090/    # 200

const host = location.hostname || "192.168.110.207";
const http = (port, path) => "http://" + host + (port === 80 ? "" : ":" + port) + (path || "/");

const groups = [
  { title: "应用", items: [
      ["Dify", "对话应用与工作流", http(80)],
      ["Langfuse", "LLM 调用观测", http(3000)],
      ["DeepSeek Harness", "DeepSeek 控制台", http(3080)],
      ["办公 IP 云同步", "办公网 IP 同步控制台", http(8765)],
      ["Hindsight", "记忆服务 API 文档", http(18888, "/docs")],
  ]},
  { title: "模型与网关", items: [
      ["LiteLLM 控制台", "模型网关管理界面", http(4000, "/ui")],
      ["LiteLLM API", "网关接口文档", http(4000)],
      ["Xinference", "本地模型推理", http(9997)],
      ["Higress 控制台", "网关配置", http(8001)],
      ["Higress 网关", "HTTP 入口", http(18080)],
      ["Nacos", "配置与注册中心", http(8888)],
  ]},
  { title: "监控", items: [
      ["Grafana", "指标看板", http(3001)],
      ["Prometheus", "指标查询", http(9092)],
  ]},
  { title: "代码与接口", items: [
      ["GitLab", "代码仓库", http(8929)],
      ["服务器状态 API", "接口文档", http(8777, "/docs")],
      ["工单 API", "接口文档", http(8778, "/docs")],
      ["vLLM 健康检查", "接口文档", http(8779, "/docs")],
      ["vLLM 自愈 API", "接口文档", http(8780, "/docs")],
  ]},
];

docker（一切容器的前提）
  └─ litellm（systemd，依赖网络）
  └─ Dify compose（api / worker / web / nginx / plugin_daemon / db / redis / weaviate / sandbox）
        └─ 依赖 LiteLLM(:4000) 与 Xinference(:9997)
  └─ Langfuse compose（web / worker / postgres / clickhouse / redis / minio）
  └─ Xinference（独立容器）
  └─ Higress compose（gateway / console / pilot / controller / nacos / prometheus / grafana / loki）
  └─ 监控 hw8（grafana / prometheus）
  └─ k3s（NodePort → Argo CD / demo-nginx）

# Windows 开发机（本机）
python _reboot_restore_207.py

python _reboot_check.py

# 1) 底盘
systemctl start docker; systemctl is-active docker

# 2) LiteLLM
systemctl restart litellm; curl -s -o /dev/null -w '%{http_code}\n' http://127.0.0.1:4000/health/liveliness

# 3) Dify
cd /dify/dify-main/docker && docker compose up -d && docker compose ps

# 4) Langfuse
cd /Langfuse/langfuse && docker compose up -d

# 5) Xinference
docker start xinference

# 6) Higress
cd /opt/higress/compose && docker compose up -d

# 7) 监控
cd /opt/workbuddy/patches_cursor/作业8_SLA监控大盘 && docker compose -f docker-compose.monitoring.yml up -d

# 8) k3s
systemctl start k3s
KUBECONFIG=/etc/rancher/k3s/k3s.yaml k3s kubectl get pods -A | head -30

# 9) 统一收口：11 项探活
for u in \
  http://192.168.110.207/ \
  http://192.168.110.207:4000/health/liveliness \
  http://192.168.110.207:3000/api/public/health \
  http://192.168.110.207:9997/v1/models \
  http://192.168.110.207:18080/ \
  http://192.168.110.207:8001/ \
  http://192.168.110.207:8929/users/sign_in \
  http://192.168.110.207:30080/ \
  http://192.168.110.207:30088/ \
  http://192.168.110.207:3001/ \
  http://192.168.110.207:9092/-/ready ; do
  printf '%-50s ' "$u"; curl -s -o /dev/null -w '%{http_code}\n' --max-time 8 "$u"
done

Dify 应用
   └─(1) LiteLLM :4000 ──(2)──> vLLM 116.62.10.183:8000

echo "网关: $(curl -s -o /dev/null -w '%{http_code}' --max-time 5 http://192.168.110.207:4000/health/liveliness)"
echo "上游: $(curl -s -o /dev/null -w '%{http_code}' --max-time 5 http://116.62.10.183:8000/v1/models)"
echo "模型: $(curl -s --max-time 5 -H 'Authorization: Bearer sk-litellm-bocheng-207' \
        http://192.168.110.207:4000/v1/models | head -c 200)"

# LiteLLM（systemd，最容易看出上游超时/429）
journalctl -u litellm -n 200 --no-pager | tail -60

# Dify api / worker
cd /dify/dify-main/docker && docker compose logs --tail=200 api worker

① 打开 http://192.168.110.207:8090/   → 确认入口页本身还在（207 活着）
② 跑三层定性那一组 curl               → 判是 ①网关 / ②上游 / ③应用
③ ①层挂 → restart litellm            ③层挂 → 查 Dify 模型供应商配置
   ②层挂 → 上 116（或作业5/7/13）
④ 恢复后看 Langfuse Trace 确认新请求正常，并记一条值班日志

docker exec docker-db_postgres-1 psql -U postgres -d dify -c \
  "select category, strategy_setting, upgrade_mode, exclude_plugins::text,
          to_char(updated_at,'YYYY-MM-DD HH24:MI') as updated
     from tenant_plugin_auto_upgrade_strategies where category='model';"

python scripts_cursor/Dify_专题十七_OpenAI插件升级后schema补丁丢失/专题十七_01_查看插件自动升级策略.py

# 全部探活
python _reboot_restore_207.py        # 末尾会再打一遍探活表

# 真实对话（走 LiteLLM）
curl -s http://192.168.110.207:4000/v1/chat/completions \
  -H 'Authorization: Bearer sk-litellm-bocheng-207' \
  -H 'Content-Type: application/json' \
  -d '{"model":"qwen3.6-27b","messages":[{"role":"user","content":"ping"}],"max_tokens":16}'