# 网络环境极差时的MathModelAgent配置过程此文档用于记录2025/7/12时，在网络极差时的配置过程。试图在VMware Workstation17中的CentOS 7上部署。因为网络问题而失败。校园网能上GitHub上不了dockerhub，在外面挂VPN（不开启TUN）能上dockerhub上不了GitHub。改用Windows 11 家庭中文版 24H2 从零开始安装并使用Docker部署。 ## 1.下载并配置Docker Desktop 此步不需要使用VPN。官网上不去可以自行搜索安装包。运行下载的安装程序，按照安装向导的提示完成安装。安装结束后程序会要求重启计算机。安装完成后，启动 Docker Desktop。启动时需要登录 Docker 账号，连不上Docker官网所以选择免登录的方式。等待 Docker Desktop 启动完成，确保系统托盘中的 Docker 图标显示为运行状态。

此时会弹出命令行要求安装WSL。等待即可。 a33121ae9047c01153edebe7ed666a06

本机使用VPN的时候Github一直都能ping通的，但是WSL的安装进度会卡在30%左右， ![c4c65346b2914b38da56a7ee5065702f](https://github.com/user-attachments/assets/6d91a444-587c-4920-a2ef-ce2ba0bd0e25) 安装完毕。根据自己的网络状态自行配置。

## 2.下载 MathModelAgent 项目打开命令提示符或 PowerShell，执行以下命令克隆项目： ```bash git clone https://github.com/jihe520/MathModelAgent.git cd MathModelAgent ``` 连不上Github就找同学帮忙或者干脆tb代下。笔者直接下载的整个代码压缩包。 ## 3.配置环境变量 1. **后端配置** - 在`backend/`目录下，复制`.env.dev.example`为`.env.dev` - 编辑 `.env.dev`，填入模型 API 密钥等配置（如 OpenAI API 密钥）。 2. **前端配置** - 在`frontend/`目录下，复制`.env.example`为`.env.development` ## 4.使用 Docker 构建并启动服务 1. **打开终端** 进入项目根目录（包含 `docker-compose.yml` 的目录）。 2. **构建 Docker 镜像** 执行以下命令构建后端和前端镜像： ```bash docker-compose build ``` - 若提示 `docker-compose` 命令不存在，可尝试使用 `docker compose`（注意中间有空格）。实际上网不好就会有这种情况： ```bash D:\mmaS\MathModelAgent>docker-compose build [+] Building 21.6s (6/6) FINISHED => [internal] load local bake definitions 0.0s => => reading from stdin 678B 0.0s => [backend internal] load build definition from Dockerfile 0.1s => => transferring dockerfile: 799B 0.0s => [frontend internal] load build definition from Dockerfile 0.1s => => transferring dockerfile: 363B 0.0s => ERROR [frontend internal] load metadata for docker.io/library/node:20 21.1s => [backend internal] load metadata for ghcr.io/astral-sh/uv:latest 4.3s => ERROR [backend internal] load metadata for docker.io/library/python:3.12-slim 21.1s ------ > [frontend internal] load metadata for docker.io/library/node:20: ------ ------ > [backend internal] load metadata for docker.io/library/python:3.12-slim: ------ Dockerfile:2 -------------------- 1 | # 构建阶段 2 | >>> FROM node:20 AS build 3 | 4 | WORKDIR /app -------------------- target frontend: failed to solve: node:20: failed to resolve source metadata for docker.io/library/node:20: failed to do request: Head "https://registry-1.docker.io/v2/library/node/manifests/20": dialing registry-1.docker.io:443 container via direct connection because disabled has no HTTPS proxy: connecting to registry-1.docker.io:443: dial tcp 108.160.169.46:443: connectex: A connection attempt failed because the connected party did not properly respond after a period of time, or established connection failed because connected host has failed to respond. ``` Docker 在拉取镜像时无法连接到 Docker Hub 镜像仓库。笔者在修改Docker设置中的镜像源后仍然不行。可以尝试手动拉取。 ```bash docker pull node:20 docker pull python:3.12-slim docker pull ghcr.io/astral-sh/uv:latest ``` 网不好的时候手动拉取都不行。在自己的Windows上随意修改Dockerfile可能会有奇奇怪怪的错误。所以其实不妨 ```bash C:\Users\Nova>docker pull swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/library/node:20 20: Pulling from ddn-k8s/docker.io/library/node ca4e5d672725: Pull complete 30b93c12a9c9: Pull complete 10d643a5fa82: Pull complete d6dc1019d793: Pull complete ea8e6f2ca326: Pull complete 19951d6e9461: Pull complete 247e8d31d16f: Pull complete 282ebe8f0e48: Pull complete Digest: sha256:7f0e372e66623dd26aa7e7156dab18e3bb02f10f08c69aca1ecba7c538f4d6d4 Status: Downloaded newer image for swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/library/node:20 swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/library/node:20 C:\Users\Nova>docker pull swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/python:3.12-slim 3.12-slim: Pulling from ddn-k8s/docker.io/python f11c1adaa26e: Pull complete a64cae28bf14: Pull complete 0e903b6a67b0: Pull complete 808ac08b1607: Pull complete c4e9078483aa: Pull complete Digest: sha256:8d86cd5ae705baa369b0a74881b4811f28a60d2a2900d2a6221b65be7d481101 Status: Downloaded newer image for swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/python:3.12-slim swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/python:3.12-slim C:\Users\Nova>docker pull swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/astral-sh/uv:latest latest: Pulling from ddn-k8s/ghcr.io/astral-sh/uv d34f8c7c82d2: Pull complete ac8f40aa894b: Pull complete Digest: sha256:d6a9c6668bb8fc51bcb0b452a1d9e98252c36cd1e4705a056f92a97bcbabfb73 Status: Downloaded newer image for swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/astral-sh/uv:latest swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/astral-sh/uv:latest ``` 使用一般都可靠的华为云。此时拉取的镜像来自华为云仓库（`swr.cn-north-4.myhuaweicloud.com/ddn-k8s/...`），但项目的`Dockerfile`中引用的是默认镜像名（如`node:20`、`python:3.12-slim`）。需要给这些镜像添加与`Dockerfile`一致的标签，让 Docker 构建时直接使用本地镜像。 **在执行拉取的目录下**： ```powershell # 为node镜像添加标签（匹配Dockerfile中的node:20） docker tag swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/library/node:20 node:20 # 为python镜像添加标签（匹配Dockerfile中的python:3.12-slim） docker tag swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/python:3.12-slim python:3.12-slim # 为uv镜像添加标签（匹配Dockerfile中的ghcr.io/astral-sh/uv:latest） docker tag swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/astral-sh/uv:latest ghcr.io/astral-sh/uv:latest ``` 然后检查： ```powershell C:\Users\Nova>docker images | findstr "node python uv" ghcr.io/astral-sh/uv latest 25a2dfe6d423 2 months ago 40.9MB swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/astral-sh/uv latest 25a2dfe6d423 2 months ago 40.9MB node 20 1a8e51cfa7a5 11 months ago 1.1GB swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/library/node 20 1a8e51cfa7a5 11 months ago 1.1GB python 3.12-slim 36d84f5948d0 12 months ago 129MB swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/python 3.12-slim 36d84f5948d0 12 months ago 129MB ``` 华为云原镜像也会显示，不影响。把终端关了，切换到项目根目录（包含`docker-compose.yml`的文件夹）再打开终端。执行构建命令（强制使用本地镜像，不联网拉取）： ```powershell docker-compose build --no-cache ``` - `--no-cache`参数确保不使用缓存，直接基于本地镜像构建，避免网络请求。这个过程要5-15分钟，笔者用了11分钟左右。 ```powershell D:\mmaS\MathModelAgent>docker-compose build --no-cache [+] Building 686.8s (28/28) FINISHED => [internal] load local bake definitions 0.0s => => reading from stdin 726B 0.0s => [frontend internal] load build definition from Dockerfile 0.0s => => transferring dockerfile: 363B 0.0s => [backend internal] load build definition from Dockerfile 0.0s => => transferring dockerfile: 799B 0.0s => [frontend internal] load metadata for docker.io/library/node:20 0.0s => [frontend internal] load .dockerignore 0.1s => => transferring context: 633B 0.0s => [backend internal] load metadata for ghcr.io/astral-sh/uv:latest 0.0s => [backend internal] load metadata for docker.io/library/python:3.12-slim 0.0s => [backend internal] load .dockerignore 0.1s => => transferring context: 248B 0.0s => [frontend 1/6] FROM docker.io/library/node:20 0.6s => [frontend internal] load build context 0.5s => => transferring context: 9.92MB 0.4s => [backend] FROM ghcr.io/astral-sh/uv:latest 0.2s => [backend builder 1/7] FROM docker.io/library/python:3.12-slim 0.5s => [backend internal] load build context 2.1s => => transferring context: 89.14MB 2.0s => [backend builder 2/7] COPY --from=ghcr.io/astral-sh/uv:latest /uv /uvx /bin/ 0.3s => [frontend 2/6] WORKDIR /app 0.1s => [frontend 3/6] COPY package.json pnpm-lock.yaml ./ 0.2s => [backend builder 3/7] WORKDIR /app 0.2s => [frontend 4/6] RUN npm install -g pnpm 7.5s => [backend builder 4/7] COPY pyproject.toml uv.lock ./ 0.4s => [backend builder 5/7] RUN --mount=type=cache,target=/root/.cache/uv uv sync --locked --no-install-proje 157.1s => [frontend 5/6] RUN pnpm install 670.7s => [backend builder 6/7] COPY . . 0.2s => [backend builder 7/7] RUN --mount=type=cache,target=/root/.cache/uv uv sync --locked 0.3s => [backend] exporting to image 4.4s => => exporting layers 4.3s => => writing image sha256:d32f0070b82bad9a6a5b65eae10fe1aea6db669ef5f9d6a0bcf08964f337786e 0.0s => => naming to docker.io/library/mathmodelagent-backend 0.0s => [backend] resolving provenance for metadata file 0.0s => [frontend 6/6] COPY . . 0.1s => [frontend] exporting to image 7.0s => => exporting layers 6.9s => => writing image sha256:31813222c752a29ce4b26e2268abefa43ed5e224edb03a843955d0ca71085afe 0.0s => => naming to docker.io/library/mathmodelagent-frontend 0.0s => [frontend] resolving provenance for metadata file 0.0s [+] Building 2/2 ✔ backend Built 0.0s ✔ frontend Built 0.0s ``` 这样就算构建好了。 3. **构建完成后，启动容器** ```bash docker-compose up -d ``` - `-d` 表示后台运行。网络不好时就会有： ```powershell D:\mmaS\MathModelAgent>docker-compose up -d [+] Running 1/1 ✘ redis Error Get "https://registry-1.docker.io/v2/": context deadline exceeded 16.0s Error response from daemon: Get "https://registry-1.docker.io/v2/": context deadline exceeded ``` 即Docker 无法连接到 Docker 官方镜像仓库（`registry-1.docker.io`），导致拉取`redis`镜像时超时。还是只能使用镜像。 ```bash D:\mmaS\MathModelAgent>docker pull swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/redis:latest latest: Pulling from ddn-k8s/docker.io/redis d2eb42b4a5eb: Pull complete 5de2dd3ff2ef: Pull complete 6c334acf232e: Pull complete 3090e1a50a6c: Pull complete f5bc47c37726: Pull complete 20eea55b3ebb: Pull complete 4f4fb700ef54: Pull complete d128ccd842a6: Pull complete Digest: sha256:00d8139cc831c0cdbac83efe93b3089b1c714bda4ea6c4f533cd60b69a6ef8bc Status: Downloaded newer image for swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/redis:latest swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/redis:latest D:\mmaS\MathModelAgent>docker tag swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/redis:latest redis:latest D:\mmaS\MathModelAgent>docker images | grep redis 'grep' 不是内部或外部命令，也不是可运行的程序或批处理文件。 ``` 此时可能仍会有： ```powershell D:\mmaS\MathModelAgent>docker-compose up -d [+] Running 1/1 ✘ redis Error Get "https://registry-1.docker.io/v2/": net/http: request canceled while waiting fo... 15.9s Error response from daemon: Get "https://registry-1.docker.io/v2/": net/http: request canceled while waiting for connection (Client.Timeout exceeded while awaiting headers) ``` 即已经手动拉取并标签化了 Redis 镜像，但`docker-compose`仍在尝试从 Docker Hub 拉取 Redis（可能是配置缓存或镜像引用优先级问题）。解决办法是直接在`docker-compose.yml`中指定使用华为云的 Redis 镜像，跳过标签依赖。换一种方式验证镜像的存在： ```powershell D:\mmaS\MathModelAgent>docker images | findstr "redis" redis latest fa310398637f 6 months ago 117MB swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/redis latest fa310398637f 6 months ago 117MB ``` 能看到`redis:latest`，说明镜像已正确标签化。直接在配置文件中指定 Redis 镜像的完整地址，避免 Docker 尝试从默认仓库拉取： 1. 在项目根目录找到`docker-compose.yml`文件，用记事本或编辑器打开。 2. 找到`redis`服务的配置，在本项目中： ```yaml # docker-compose.yml services: redis: image: redis:alpine container_name: mathmodelagent_redis ports: - "6379:6379" # 将 Redis 容器的 6379 端口映射到主机的 6379 端口 volumes: - redis_data:/data # Redis 数据的持久化存储卷 backend: build: context: ./backend # Dockerfile 的上下文路径 dockerfile: Dockerfile # Dockerfile 文件名 container_name: mathmodelagent_backend ports: - "8000:8000" # 将后端容器的 8000 端口映射到主机的 8000 端口 env_file: - ./backend/.env.dev #从此文件加载环境变量 environment: - ENV=DEV # 为后端显式设置 ENV # .env.dev 文件中的 REDIS_URL 应为 redis://redis:6379/0 volumes: - ./backend:/app # 挂载后端代码以实现热重载 (开发环境) - ./backend/project/work_dir:/app/project/work_dir # 持久化生成的文件 - backend_venv:/app/.venv # 可选：持久化 venv 以在依赖不变时加快后续构建速度 depends_on: - redis # 确保 Redis 在后端启动前启动 stdin_open: true # 保持标准输入打开 tty: true # 分配一个伪终端 frontend: build: context: ./frontend dockerfile: Dockerfile container_name: mathmodelagent_frontend ports: - "5173:5173" # 将前端容器的 5173 端口 (Vite 默认) 映射到主机的 5173 端口 env_file: - ./frontend/.env.development # 加载前端的环境变量 volumes: - ./frontend:/app # 挂载前端代码以实现热重载 (开发环境) - /app/node_modules # 匿名卷，防止主机的 node_modules 覆盖容器内的 depends_on: - backend # 确保后端可用 (尽管前端通常只需要其 URL) stdin_open: true tty: true volumes: redis_data: # 定义用于 Redis 持久化的命名卷 backend_venv: # 定义用于后端 venv 的命名卷 ``` 3. 修改`image`字段为华为云镜像地址：这一段 ```yaml services: redis: image: redis:alpine # 这里就是需要修改的 image 字段 container_name: mathmodelagent_redis ports: - "6379:6379" volumes: - redis_data:/data ``` 修改为： ```yaml services: redis: image: swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/redis:alpine # 修改后的镜像地址 container_name: mathmodelagent_redis ports: - "6379:6379" volumes: - redis_data:/data ``` 然后保存。注意**YAML 要求 `key: value` 中冒号后必须有空格**，不然就会像我这样： ```powershell D:\mmaS\MathModelAgent>docker-compose down yaml: line 6: mapping values are not allowed in this context ``` yaml文件格式错误导致关不掉。改好了就再重启服务： ```powershell D:\mmaS\MathModelAgent>docker-compose down D:\mmaS\MathModelAgent>docker-compose up -d [+] Running 9/9 ✔ redis Pulled 5.3s ✔ da9db072f522 Pull complete 1.3s ✔ dd8d46bd4047 Pull complete 1.4s ✔ 5057e26f1a86 Pull complete 1.6s ✔ be83d0fd33a3 Pull complete 2.6s ✔ b3d150cb1b6c Pull complete 4.0s ✔ 369ad5b9119b Pull complete 4.0s ✔ 4f4fb700ef54 Pull complete 4.1s ✔ 37d63ae71d35 Pull complete 4.1s [+] Running 6/6 ✔ Network mathmodelagent_default Created 0.1s ✔ Volume "mathmodelagent_backend_venv" Created 0.0s ✔ Volume "mathmodelagent_redis_data" Created 0.0s ✔ Container mathmodelagent_redis Started 28.2s ✔ Container mathmodelagent_backend Started 28.4s ✔ Container mathmodelagent_frontend Started 11.9s D:\mmaS\MathModelAgent>docker ps | findstr "redis" 0947be7781cc swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/redis:alpine "docker-entrypoint.s鈥? 43 seconds ago Up 15 seconds 0.0.0.0:6379->6379/tcp, [::]:6379->6379/tcp mathmodelagent_redis FINDSTR: 写入错误 ``` 所有容器（`redis`、`backend`、`frontend`）都已成功启动，那个`FINDSTR: 写入错误`是 Windows 命令行的临时小问题，不影响服务运行，不用在意。 ## 5.启动项目 ### 1. 访问项目前端界面打开浏览器，输入以下地址：`http://localhost:5173` 如果一切正常，会看到`MathModelAgent`的前端界面，此时可以尝试使用项目功能（比如输入数学建模问题，让系统自动处理）。

### 2. 验证后端服务是否正常访问后端 API 地址，确认服务运行：`http://localhost:8000` 如果后端正常启动，会看到类似 API 文档或状态提示的页面（具体取决于项目设计）。

### 3. 检查服务日志（若有问题）如果访问时出现空白页、报错等情况，可以查看容器日志排查问题： ```powershell # 查看前端日志（比如前端无法加载、报错） docker logs mathmodelagent_frontend ```

```powershell # 查看Redis日志（比如启动失败） docker logs mathmodelagent_redis ```

日志中会显示具体错误信息（如依赖缺失、配置错误等），根据提示调整即可。 ### 4. 停止服务（如需）如果后续需要停止项目，执行： ```powershell docker-compose down ```