部署的Rasa开源虚拟助手无响应问题排查求助

环境与背景 #

• 部署环境：Google Cloud E2机器（Ubuntu 18.04），Docker容器集群
• 助手功能：旧金山探索馆虚拟讲解员，支持三类查询：
  • 展品概念/现象解释
  • 指定展品详情（描述、位置、创作者、年份）
  • 展品推荐
• 问题现象：
  • 对话中助手时常无响应
  • 本地rasa shell运行正常，但部署时COPY的模型被判无效，故在Dockerfile中加入RUN rasa train（训练数据/配置与本地一致）
  • M1芯片本地Docker测试时，助手几乎总是忽略第一个用户输入，日志显示BotUttered已触发但Web服务器无响应
  • 怀疑问题出在Nginx、Socket.io等网络组件，而非Rasa本身

配置详情 #

docker-compose.yml #

version: '3'
services:
  rasa_server:
    container_name: "rasa_server"
    build:
      context: backend
    user: root
    ports:
      - "5005:5005"
    expose:
      - 5005
  action_server:
    container_name: "action_server"
    build:
      context: actions
    ports:
      - "5055:5055"
    expose:
      - 5055
  web_server:
    container_name: "web_server"
    image: nginx
    depends_on:
      - "rasa_server"
      - "action_server"
    ports:
      - "80:80"
      - "443:443"
    restart: always
    volumes:
      - ./web/conf/default.conf:/etc/nginx/conf.d/default.conf
      - ./web/conf/explorer-ai.chat.conf:/etc/nginx/conf.d/explorer-ai.chat.conf
      - ./web/conf/nginx.conf:/etc/nginx/nginx.conf
      - ./web/html:/var/www/html
      - ./certbot/www:/var/www/certbot
      - ./certbot/conf:/etc/letsencrypt
  certbot:
    container_name: "certbot"
    image: certbot/certbot
    volumes:
      - ./certbot/www:/var/www/cerbot:rw
      - ./certbot/conf:/etc/letsencrypt:rw

backend Dockerfile #

FROM rasa/rasa:3.3.1
WORKDIR /app
USER root
COPY ./models /app/models
COPY ./data /app/data
COPY . /app
RUN rasa train
ENTRYPOINT ["rasa", "run", "-m", "models", "--enable-api", "--cors", "*", "--debug"]

actions Dockerfile #

FROM rasa/rasa-sdk:3.3.0
WORKDIR /app
COPY ./requirements-actions.txt requirements.txt
COPY . /app
USER root
RUN pip install --no-cache-dir -r requirements.txt
USER 1001
EXPOSE 5055
ENTRYPOINT ["python", "-m", "rasa_sdk", "--actions", "actions"]

credentials.yml Socket.io配置 #

socketio:
  user_message_evt: user_uttered
  bot_message_evt: bot_uttered
  session_persistence: false

index.html Socket.io相关代码 #

<script
  src="https://cdnjs.cloudflare.com/ajax/libs/socket.io/3.1.2/socket.io.js"
  integrity="sha512-YybopSVjZU0fe8TY4YDuQbP5bhwpGBE/T6eBUEZ0usM72IWBfWrgVI13qfX4V2A/W7Hdqnm7PIOYOwP9YHnICw=="
  crossorigin="anonymous" referrerpolicy="no-referrer">
</script>
<script>
  const socket = io("https://explorer-ai.chat");
  const messages = document.getElementById('messages');
  const form = document.getElementById('form');
  const messageInput = document.getElementById('message-input');
  function scrollToBottom() {
    window.scrollTo(0, document.body.scrollHeight);
  }
  function appendMessage(msg, type) {
    const item = document.createElement('div');
    item.textContent = msg;
    item.classList.add("message");
    item.classList.add(`message_${type}`);
    messages.appendChild(item);
    scrollToBottom();
  }
  const welcome = "Hello! I'm Explorer AI, a virtual assistant for visitors to the " +
    "Exploratorium in San Francisco. I've been trained to answer questions about the " + 
    "Exploratorium's exhibits and the concepts they illustrate.";
  const how_help = "How may I assist you?";
  appendMessage(welcome, "received");
  appendMessage(how_help, "received");
  form.addEventListener('submit', function (e) {
    e.preventDefault();
    const msg = messageInput.value;
    if (msg) {
      socket.emit('user_uttered', {
        "message": msg,
      });
      messageInput.value = '';
      appendMessage(msg, "sent");
    }
  });
  socket.on('connect', function () {
    console.log("Connected to Socket.io server");
  });
  socket.on('connect_error', (error) => {
    console.log("Connection to Socket.io FAILED");
    console.error(error);
  });
  socket.on('bot_uttered', function (response) {
    console.log("Bot uttered:", response);
    if (response.text) {
      appendMessage(response.text, "received");
    }
    if (response.attachment) {
      appendImage(response.attachment.payload.src, "received");
    }
    if (response.quick_replies) {
      appendQuickReplies(response.quick_replies);
    }
  });
</script>

Nginx配置 #

server {
    listen 80;
    listen [::]:80;
    server_name explorer-ai.chat www.explorer-ai.chat;
    location /.well-known/acme-challenge/ {
        root /var/www/certbot;
    }
    location / {
        return 301 https://explorer-ai.chat$request_uri;
    }
} 
server {
    listen 443 ssl;
    listen [::]:443 ssl;
    server_name explorer-ai.chat www.explorer-ai.chat;
    ssl_certificate         /etc/letsencrypt/live/explorer-ai.chat/fullchain.pem;
    ssl_certificate_key     /etc/letsencrypt/live/explorer-ai.chat/privkey.pem;
    include                 /etc/letsencrypt/options-ssl-nginx.conf;
    ssl_dhparam             /etc/letsencrypt/ssl-dhparams.pem;
    root /var/www/html;
    index index.html;
    location /socket.io/ {
        proxy_pass http://rasa_server:5005;
    }
}

排查建议 #

1. 补全Nginx Socket.io代理配置
   Socket.io依赖WebSocket协议，当前Nginx配置缺少必要的握手支持，在location /socket.io/块中添加：

   proxy_http_version 1.1;
   proxy_set_header Upgrade $http_upgrade;
   proxy_set_header Connection "upgrade";
   proxy_set_header Host $host;
   proxy_cache_bypass $http_upgrade;
   

   缺少这些配置会导致WebSocket连接失败，消息无法实时推送。

2. 优化服务启动顺序
   depends_on仅保证容器启动顺序，不保证服务就绪。给rasa_server添加健康检查，确保服务完全启动后再接收请求：
   在docker-compose.yml的rasa_server服务中加入：

   healthcheck:
     test: ["CMD", "curl", "-f", "http://localhost:5005/health"]
     interval: 10s
     timeout: 5s
     retries: 5
   

   同时修改web_server的depends_on：

   depends_on:
     rasa_server:
       condition: service_healthy
     action_server:
       condition: service_started
   

   避免前端连接时Rasa服务尚未初始化完成。

3. 检查Socket.io版本兼容性
   前端使用Socket.io 3.1.2，需确认Rasa容器中python-socketio版本与之兼容。在backend容器中执行pip list | grep python-socketio查看版本，若差异过大需调整前端或后端版本。

4. 保证会话ID一致性
   前端发送user_uttered事件时未携带会话ID，可能导致会话不匹配。修改前端逻辑：

   // 生成或获取持久化的会话ID
   let sessionId = localStorage.getItem('rasa_session_id') || self.crypto.randomUUID();
   localStorage.setItem('rasa_session_id', sessionId);
   
   // 发送消息时携带session_id
   socket.emit('user_uttered', {
     "message": msg,
     "session_id": sessionId
   });
   

   确保前后端会话关联正确。

5. 日志排查

   • 查看Rasa服务器日志：docker logs rasa_server，确认BotUttered事件是否正确发送
   • 查看Nginx日志：docker logs web_server，检查Socket.io请求的状态码
   • 查看浏览器控制台，确认Socket.io连接和消息发送是否有错误

内容的提问来源于stack exchange，提问作者vonbecker