# docker搭建deepseek运行环境和open-webui

这篇文章没有什么实用价值，仅作为快速体验ollama运行环境。或者当做一个附加工具，用来管理和查看ollama环境。

#### docker-compose.yaml

这个脚本将同时启动**ollama** 和 **open-webui**

```yaml

version: "3"
services:
  ollama:
    image: harbor.iovhm.com/hub/ollama/ollama:0.5.12
    container_name: ollama
    restart: always
    privileged: true
    ports:
      - "11434:11434"
    volumes:
      - ./ollama:/root/.ollama
    networks:
      - vpclub-bridge

# docker-compose --profile open-webui up -d
  open-webui:
    # CPU版
    image: harbor.iovhm.com/public/open-webui/open-webui:main
    # GPU版
    # image: harbor.iovhm.com/public/open-webui/open-webui:main-gpu
    container_name: open-webui
    restart: always
    privileged: true
    ports:
      - "3000:8080"
    volumes:
      - ./open-webui:/app/backend/data
    environment:
      # 如果你的 ollama 服务器不在本机，请修改此地址，如果OLLAMA_BASE_URLS被设置，则使用OLLAMA_BASE_URLS
      # - OLLAMA_BASE_URL=http://ollama:11434
      # 如果你的 ollama 服务器不在本机，请修改此地址，可以使用分号分割多个地址提供负载均衡能力
      - OLLAMA_BASE_URLS=http://ollama:11434
      # 关闭 openai api 否则会因为连不上openai而卡界面
      - ENABLE_OPENAI_API=false
      # 禁用自带的 Arena Model模型（竞技场模型）
      - ENABLE_EVALUATION_ARENA_MODELS=false
      # 关闭社区分享功能
      - ENABLE_COMMUNITY_SHARING=false
      # 离线模式，不自动从Internet下载模型(非必须，有魔法的话不需要设置)
      # - HF_HUB_OFFLINE=true
      # 默认的语义向量模型引擎(非必须)
      # - RAG_EMBEDDING_ENGINE=ollama
      # 默认的语义向量模型(非必须)
      # - RAG_EMBEDDING_MODEL=nomic-embed-text:latest
    networks:
      - vpclub-bridge


networks:
  vpclub-bridge:
    external:
      name: vpclub-bridge  


```

#### 测试一下服务


打开浏览器，输入ollama的服务地址， **http://ip:11434** ,界面提示:**Ollama is running** ,则表示**ollama**安装成功了

[![](https://iovhm.com/book/uploads/images/gallery/2025-03/scaled-1680-/BiuUdBKkSZ84xqu1-image-1740998429287.png)](https://iovhm.com/book/uploads/images/gallery/2025-03/BiuUdBKkSZ84xqu1-image-1740998429287.png)


输入open-webui的地址，**http://ip:3000** , 则可以看到open-webui的地址。

[![](https://iovhm.com/book/uploads/images/gallery/2025-03/scaled-1680-/qVkNjWcJacl6qx45-image-1741150619335.png)](https://iovhm.com/book/uploads/images/gallery/2025-03/qVkNjWcJacl6qx45-image-1741150619335.png)


但是这个时候可能是白屏，什么都看不到；或者好不容易刷出来界面，设置好用户名密码进入系统后，也是白屏，此时应该去查看open-webui容器的出错提示，多半都是被墙拉取不到镜像的原因。那你需要魔法。


##### 界面打不开，或者进去了也很卡问题解决

- 打开open-webui等很久，不出来界面，白屏

因为软件有很多外部依赖要下载，如果被墙，下载不到、就一直卡住，可以先不使用**openai**和不使用内置的**语义向量模型模型**

```yaml

# 修改open-webui的环境变量设置

      # 关闭 openai api 否则会因为连不上openai而卡界面
      - ENABLE_OPENAI_API=false
      # 默认的RAG引擎(非必须)
      - RAG_EMBEDDING_ENGINE=ollama
      # 默认的语义向量模型(非必须)
      - RAG_EMBEDDING_MODEL=nomic-embed-text:latest

```

##### 进入软件后，每次刷新都要等很久

**关闭使用openai外部链接，修改语义向量模型**


[![](https://iovhm.com/book/uploads/images/gallery/2025-03/scaled-1680-/WpbjKcI2bviH0DgW-image-1741153227244.png)](https://iovhm.com/book/uploads/images/gallery/2025-03/WpbjKcI2bviH0DgW-image-1741153227244.png)

**修改语义向量模型**

[![](https://iovhm.com/book/uploads/images/gallery/2025-03/scaled-1680-/hzIZlpfesgHvWmYD-image-1741142128397.png)](https://iovhm.com/book/uploads/images/gallery/2025-03/hzIZlpfesgHvWmYD-image-1741142128397.png)


#### 进入到ollama容器下载模型

```shell

# 安装完成后进入ollma容器
docker exec -it ollama /bin/bash

# 查看ollama版本
ollama -v

# 下载deepseek模型
ollama run deepseek-r1:7b

# 语义向量模型
ollama pull nomic-embed-text

```


#### 查看系统当前都安装了什么模型



[![](https://iovhm.com/book/uploads/images/gallery/2025-03/scaled-1680-/rwA6k3Sa89tpKccp-image-1741152828292.png)](https://iovhm.com/book/uploads/images/gallery/2025-03/rwA6k3Sa89tpKccp-image-1741152828292.png)


#### 配置知识库



[![](https://iovhm.com/book/uploads/images/gallery/2025-03/scaled-1680-/fkz6DVGPKeSyYHBx-image-1741150695356.png)](https://iovhm.com/book/uploads/images/gallery/2025-03/fkz6DVGPKeSyYHBx-image-1741150695356.png)

点击上传文件，建议的文件类型为**markdown**,使用 **####** 进行段落区分，如果保存知识库出错，那是因为没有设置正确的语义向量模型

[![](https://iovhm.com/book/uploads/images/gallery/2025-03/scaled-1680-/Myex8Ea0Ex5FeqIr-image-1741150764668.png)](https://iovhm.com/book/uploads/images/gallery/2025-03/Myex8Ea0Ex5FeqIr-image-1741150764668.png)




##### 为模型关联知识库，或者在聊天界面使用 **#** 引用知识库

[![](https://iovhm.com/book/uploads/images/gallery/2025-03/scaled-1680-/jOeOgUNEL9rJmQo7-image-1741142780405.png)](https://iovhm.com/book/uploads/images/gallery/2025-03/jOeOgUNEL9rJmQo7-image-1741142780405.png)


##### 在聊天界面使用# 引用知识库

[![](https://iovhm.com/book/uploads/images/gallery/2025-03/scaled-1680-/h9hU4ioINq49Vl0m-image-1741142824988.png)](https://iovhm.com/book/uploads/images/gallery/2025-03/h9hU4ioINq49Vl0m-image-1741142824988.png)


那我们就可以开始体验deepseek能力啦，来看看deepseek静静的装B

[![](https://iovhm.com/book/uploads/images/gallery/2025-03/scaled-1680-/7Zobkh3mmisSRGgi-image-1741017091300.png)](https://iovhm.com/book/uploads/images/gallery/2025-03/7Zobkh3mmisSRGgi-image-1741017091300.png)