XAgent的部署及运行

源代码clone

git clone

config 文件的修改

在XAgent源码目录,运行 vi .env, 修改以下配置条目

CONFIG_FILE=assets/gpt-3.5-turbo_config.yml

python环境

python >=3.10

安装conda,通过conda激活python3.10的环境

wget https://repo.anaconda.com/archive/Anaconda3-2023.03-1-Linux-x86_64.sh

# 修改执行权限
chmod 777 Anaconda3-2023.03-1-Linux-x86_64.sh

# 执行
./Anaconda3-2023.03-1-Linux-x86_64.sh

# 创建python3.10+ 的环境
conda create --name py310 python=3.10

#查看新创建的环境 
conda info --envs
# conda environments:
#
# base                  *  /root/anaconda3
# py310                    /root/anaconda3/envs/py310

# 激活py310环境
activate py310

# 重新进入shell后,查看python的版本
python --version

PIP

安装pip包

# 下载安装脚本
wget https://bootstrap.pypa.io/get-pip.py

# 运行安装脚本
python3 get-pip.py

docker的部署及运行

确保已安装docker,在XAgent源码目录,运行以下命令,从代码builddocker

docker compose build
# 如果不是第一次build,可以加上--no-cache参数,强制不使用缓存,否则build内容可能不会被更新
docker compose build --no-cache

看到以下提示时,运行完毕

=> [ 4/12] RUN apt update && apt install -y build-essential make openjdk-17-jdk-headless curl docker.io docker-compose psmisc sudo                                                                         41.5s
 => [ 5/12] RUN pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple                                                                                                                     1.9s
 => [ 6/12] RUN pip install playwright && playwright install chromium &&  playwright install-deps                                                                                                           83.0s
 => [ 7/12] COPY ToolServer/ToolServerNode/requirements.txt .                                                                                                                                                0.1s
 => [ 8/12] RUN pip install --no-cache-dir -r requirements.txt                                                                                                                                             144.4s
 => [ 9/12] RUN rm -rf /tmp/* /var/tmp/*                                                                                                                                                                     0.4s
 => [10/12] COPY ToolServer/ToolServerNode .                                                                                                                                                                 0.1s 
 => [11/12] RUN mkdir -p assets                                                                                                                                                                              0.2s 
 => [12/12] RUN mkdir -p workspace                                                                                                                                                                           0.2s 
 => exporting to image                                                                                                                                                                                       9.8s 
 => => exporting layers                                                                                                                                                                                      9.8s
 => => writing image sha256:45e525fd5f2f9fbbb7679e96359934cd4b4a0658f6aaa39c0987718c87758b64                                                                                                                 0.0s
 => => naming to docker.io/xagentteam/toolserver-node:latest       

使用 docker images 命令查看,可以看到xagent的几个images已经出现在列表中

REPOSITORY                      TAG       IMAGE ID       CREATED          SIZE
xagentteam/toolserver-node      latest    45e525fd5f2f   18 minutes ago   3.75GB
xagentteam/xagent-server        latest    829edddf670b   19 minutes ago   1.29GB
xagentteam/toolserver-manager   latest    1fab5aa2bb99   22 minutes ago   1.03GB
xagentteam/toolserver-monitor   latest    803d4cd44e84   23 minutes ago   1.02GB

再运行

#  拉取的依赖容器并启动相应的服务
docker compose up

再使用 docker images 命令查看,依赖的images,比如redis,mongo,mysql也出现在列表中。在这里需要注意,系统中的某些端口不能被占用,比如3306,否则就会导致相应的应用启动失败

REPOSITORY                      TAG       IMAGE ID       CREATED          SIZE
xagentteam/toolserver-node      latest    45e525fd5f2f   23 minutes ago   3.75GB
xagentteam/xagent-server        latest    829edddf670b   24 minutes ago   1.29GB
xagentteam/toolserver-manager   latest    1fab5aa2bb99   28 minutes ago   1.03GB
xagentteam/toolserver-monitor   latest    803d4cd44e84   28 minutes ago   1.02GB
redis                           latest    76506809a39f   9 days ago       138MB
mongo                           latest    5acb2131d51f   11 days ago      757MB
mysql                           latest    f7fdab215ab7   7 weeks ago      605MB

同时,一些容器也会在系统中创建成功,可以使用docker ps --no-trunc查看(我的这个列表中mysql容器因为端口原因就启动失败了,需要更改被占用的端口后重启)

[root@opencloudos XAgent-1.0.0]# docker ps --no-trunc
                                NAMES
0c28937576b3036d988d8075c23ca5fa2acd952b0719879e2fd5301d463192bb   xagentteam/toolserver-node:0.2         "uvicorn main:app --host 0.0.0.0 --port 31942"                                                                                                      19 minutes ago   Up 19 minutes (healthy)   31942/tcp                                                                              friendly_darwin
3cf4eb9a46caf8bc774ad731a2dae9f6dffaf67ef6da85b5f8198a117f766dd0   xagentteam/xagent-server:latest        "python start_server.py"                                                                                                                            21 minutes ago   Up 21 minutes             0.0.0.0:5173->5173/tcp, :::5173->5173/tcp, 0.0.0.0:8090->8090/tcp, :::8090->8090/tcp   XAgent-Server
ede01c59c67f61dbbe7e16b984447fcdb9c65df39da264085dc001abb9b31b83   xagentteam/toolserver-node:latest      "uvicorn main:app --host 0.0.0.0 --port 31942"                                                                                                      21 minutes ago   Up 21 minutes             31942/tcp                                                                              xagent-main-ToolServerNode-1
8c5b52f2d082ed74603fa01bf75fbbbdd44301d482843c6e59dbda4f03b25347   xagentteam/toolserver-manager:latest   "gunicorn main:app -b 0.0.0.0:8080 --worker-class uvicorn.workers.UvicornWorker --workers 2 -t 600"                                                 21 minutes ago   Up 21 minutes (healthy)   0.0.0.0:8080->8080/tcp, :::8080->8080/tcp                                              xagent-main-ToolServerManager-1
0c19c715098a7b719eeb427c13f3b0ddc6eacbf059df07bae36553aefc66170e   xagentteam/toolserver-node:0.2         "uvicorn main:app --host 0.0.0.0 --port 31942"                                                                                                      8 hours ago      Up 8 hours (healthy)      31942/tcp                                                                              nice_diffie
41f1b764cbdfce3fe7d2dfad75f90fdaca9e3abd660c8942b040268f4bcb1a34   xagentteam/toolserver-node:0.2         "uvicorn main:app --host 0.0.0.0 --port 31942"                                                                                                      11 hours ago     Up 11 hours (healthy)     31942/tcp                                                                              eager_greider
1a2bfe7c41fad33417ae9528135d3a46c9c24465fc934e5cf9228e8002762de7   mysql                                  "docker-entrypoint.sh --default-authentication-plugin=caching_sha2_password --character-set-server=utf8mb4 --collation-server=utf8mb4_unicode_ci"   11 hours ago     Up 8 hours (healthy)      0.0.0.0:3306->3306/tcp, :::3306->3306/tcp, 33060/tcp                                   xagent-main-xagent-mysql-1
379d82d0c10df5ef2933acffc6529569a390bbd5f79cc4abdd2149ee42ff435a   redis                                  "docker-entrypoint.sh redis-server --requirepass xagent"                                                                                            11 hours ago     Up 8 hours (healthy)      0.0.0.0:6379->6379/tcp, :::6379->6379/tcp                                              xagent-main-xagent-redis-1


如果顺利启动,则可以看到以下提示:

INFO:     Register env WebEnv with 2 tools!
xagent-100-ToolServerNode-1     | INFO:     Register env PythonNotebook with 2 tools!
xagent-100-ToolServerNode-1     | INFO:     Register env FileSystemEnv with 3 tools!
xagent-100-ToolServerNode-1     | INFO:     Register tool shell_command_executor!
xagent-100-ToolServerNode-1     | INFO:     Loaded 8 tools and 3 envs!
xagent-100-ToolServerNode-1     | INFO:     Application startup complete.
xagent-100-ToolServerNode-1     | INFO:     Uvicorn running on http://0.0.0.0:31942 (Press CTRL+C to quit)
xagent-100-xagent-mysql-1       | 2023-12-14T05:49:03.879382Z 8 [Warning] [MY-013360] [Server] Plugin mysql_native_password reported: ''mysql_native_password' is deprecated and will be removed in a future release. Please use caching_sha2_password instead'
XAgent-Server                   | INFO:     Will watch for changes in these directories: ['/app']
XAgent-Server                   | INFO:     Uvicorn running on http://0.0.0.0:8090 (Press CTRL+C to quit)
XAgent-Server                   | INFO:     Started reloader process [1] using StatReload
xagent-100-xagent-mysql-1       | 2023-12-14T05:49:06.192883Z 9 [Warning] [MY-013360] [Server] Plugin mysql_native_password reported: ''mysql_native_password' is deprecated and will be removed in a future release. Please use caching_sha2_password instead'
XAgent-Server                   | INFO:     Started server process [16]
XAgent-Server                   | INFO:     Waiting for application startup.
XAgent-Server                   | INFO:     Application startup complete.

访问XAgent-Server

XAgent-Server启动后,默认的访问地址是 http://localhost:5173, 要确保端口是对外开放的,首次访问该地址,会看到上个步骤的控制台上会打印访问的日志信息:


XAgent-Server                   |   XAgent Service Startup Param:
XAgent-Server                   |             app: app:app
XAgent-Server                   |             prod: False
XAgent-Server                   |             base_dir: XAgentServer
XAgent-Server                   |             use_redis: False
XAgent-Server                   |             recorder_root_dir: running_records
XAgent-Server                   |             default_login: True
XAgent-Server                   |             check_running: False
XAgent-Server                   |             host: 0.0.0.0
XAgent-Server                   |             port: 8090
XAgent-Server                   |             debug: True
XAgent-Server                   |             reload: True
XAgent-Server                   |             workers: 1
XAgent-Server                   |             share_url: https://x-agent.net/api/conv/community
XAgent-Server                   |             DB: <class 'XAgentServer.application.core.envs.XAgentServerEnv.DB'>
XAgent-Server                   |             Redis: <class 'XAgentServer.application.core.envs.XAgentServerEnv.Redis'>
XAgent-Server                   |             Email: <class 'XAgentServer.application.core.envs.XAgentServerEnv.Email'>
XAgent-Server                   |             Upload: <class 'XAgentServer.application.core.envs.XAgentServerEnv.Upload'>
XAgent-Server                   | XAgent Service Init Dependence.  
XAgent-Server                   | init a thread pool executor, max_workers: 1  
XAgent-Server                   | XAgent Service Init Dependence: Complete!  
XAgent-Server                   | Default user: Guest, token: xagent, you can use it to login  
XAgent-Server                   | XAgent Server Dependences:  
XAgent-Server                   |         Python: 3.10+ 
XAgent-Server                   |         FastAPI: Http server
XAgent-Server                   |         Websocket: long connect with client
XAgent-Server                   |         MySQL: save xagent data
XAgent-Server                   |         SqlAlchemy: ORM with MySQL
XAgent-Server                   |         Redis: save status of interaction
XAgent-Server                   |         Threading: run interaction
XAgent-Server                   |         APScheduler: send data to client and keep alive
XAgent-Server                   |         FastAPI APIRouter: manage websocket route
XAgent-Server                   |         XAgentError: XAgentServer.exts.exception_ext
XAgent-Server                   | XAgent Server Version:  
XAgent-Server                   |         V 1.1.0
XAgent-Server                   | Notes:  
XAgent-Server                   |         Since V 1.1.0, 
XAgent-Server                   |         Local storage will no longer be supported, replaced by Mysql.
XAgent-Server                   |         The service depends on Redis and Mysql, 
XAgent-Server                   |         so you need to install Redis and Mysql before using it.
XAgent-Server                   |         Before you use this service, please ensure that the following services are available:
XAgent-Server                   |             1. Redis on docker, port: 6379, you can start it by docker, default password: xagent
XAgent-Server                   |             2. Mysql on docker, port: 3306, you can start it by docker
XAgent-Server                   |             3. XAgent Tool Server is runnning on port 8080
XAgent-Server                   |             4. Port 8090 is not occupied

在这里插入图片描述

部署及启动过程中的一些错误解决

错误列表

ToolServerNode—but the ‘h2’ package is not installed

我是使用的是XAgent-1.0.0版本,启动过程中xagentteam/toolserver-node:latest这个image的容器碰到以下错误:

ImportError: Using http2=True, but the 'h2' package is not installed. Make sure to install httpx using `pip install httpx[http2]`.
xagent-100-ToolServerNode-1     | 
xagent-100-ToolServerNode-1     | ERROR:    Application startup failed. Exiting.

解决方法

# 到\XAgent-1.0.0\dockerfiles\ToolServerNode\目录下
# 修改Dockerfile文件
# 在RUN pip install playwright && playwright install chromium &&  playwright install-deps这段之后添加:
RUN pip install httpx[http2]
# 注意,添加上面这段脚本后,要重新运行docker compose build 和 docker compose up

mysql启动后不停闪现的warning信息

后台服务启动后,如果不停闪现以下错误:

 [Warning] [MY-013360] [Server] Plugin mysql_native_password reported: ''mysql_native_password' is deprecated and will be removed in a future release. Please use caching_sha2_password instead'

解决方法 请修改mysql的healthcheck段配置:

healthcheck:
     # test: [ "CMD", "mysqladmin","ping", "-h", "localhost" ]
     test: ["CMD-SHELL", "mysql -h localhost -u root -pxagent -e 'SELECT 1'"]
     #interval: 10s
     timeout: 20s
     retries: 20

如果以上配置修改不能解决您的问题,尝试将你的Xagent代码切换到master版本,我在1.00版本碰到该问题,切换到master版本后运行,以上问题应该已经被修复

max_workers must be greater than 0 错误

暂无解决方案,如果你是1.00版本,建议使用master版本尝试该问题是否能修复,总而言之,master版本似乎要比1.00版本稳定不少。

   |   File "/app/XAgent/function_handler.py", line 234, in handle_tool_call
XAgent-Server                   |     command_result = self.long_result_summary(
XAgent-Server                   |   File "/app/XAgent/function_handler.py", line 167, in long_result_summary
XAgent-Server                   |     with ThreadPoolExecutor(max_workers=len(result)) as pool:
XAgent-Server                   |   File "/usr/local/lib/python3.10/concurrent/futures/thread.py", line 144, in __init__
XAgent-Server                   |     raise ValueError("max_workers must be greater than 0")
XAgent-Server                   | ValueError: max_workers must be greater than 0

AttributeError: ‘ToolServerInterface’ object has no attribute ‘cookies’

伴随以上错误,同时会有以下异常信息:

 XAgentServer.exts.exception_ext.XAgentRunningError: HTTPConnectionPool(host='localhost', port=8080): Max retries exceeded with url: /get_cookie

解决方法 问题是由localhost的8080端口沟通失败引起的,尝试使用IP地址,请注意打开8080端口的访问权限

### 部署 AI Agent 的方法和最佳实践 部署 AI Agent 是一项复杂而重要的任务,涉及多个方面和技术细节。以下是关于如何有效部署 AI Agent 的一些关键点: #### 1. **理解需求并选择合适的架构** 在部署之前,需明确目标场景及其特定需求。如果仅需要单个代理来完成某项独立任务,则可以采用 Single-Agent 架构;而对于复杂的协作环境或者分布式系统,则应考虑 Multi-Agent 系统的设计[^1]。 #### 2. **定义清晰的功能边界** 无论是单一还是多代理设置下,都需要清楚界定每个AI Agent的具体职责范围以及它们之间可能存在的交互方式。这有助于减少不必要的计算开销,并提高整体效率[^3]。 #### 3. **利用现有工具与框架加速开发进程** 对于开发者而言,使用成熟的开源库或商业产品能够显著缩短研发周期。例如,在Python生态系统中有许多强大的库支持快速搭建对话型或其他类型的智能体应用[^3]。同时也可以借鉴其他成功案例中的经验教训来进行优化调整[^2]。 #### 4. **实施有效的测试策略** 为了确保所部署的AI Agents能够在实际运行环境中稳定表现良好,必须制定详尽全面的质量保证计划。包括但不限于单元测试、集成测试乃至压力测试等多个层面的工作内容安排[^2]。 #### 5. **持续监控性能指标并与业务目标保持一致** 即使完成了初步上线之后也不意味着结束工作——相反地开始了另一个长期维护改进的过程。定期收集反馈数据用于分析评估当前状况是否满足预期效果,并据此做出相应调整直至达到理想状态为止[^1]。 ```python import openai def initialize_agent(api_key, model="gpt-3.5-turbo"): """ 初始化一个简单的AI Agent. 参数: api_key (str): OpenAI API 密钥 model (str): 使用的模型名称,默认为 gpt-3.5-turbo 返回: function: 调用API函数 """ openai.api_key = api_key def call_api(prompt): response = openai.ChatCompletion.create( model=model, messages=[{"role": "user", "content": prompt}, {"role": "assistant"}], ) return response['choices'][0]['message']['content'] return call_api ``` 以上代码片段展示了一个非常基础版本的AI Agent初始化过程示例[^3]。 ---
评论 2
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值