使用 Ollama 本地部署 DeepSeek 模型及 Chatbox 配置

Github

• https://github.com/ollama/ollama

官网

• https://ollama.com/

简介

Ollama 是一个本地化大语言模型运行框架，主要特点包括：

• 支持在本地运行 Llama 3、Mistral、DeepSeek 等主流大模型
• 提供REST API接口（如您代码中的 base_url 配置）
• 支持多平台（MacOS/Linux/Windows）
• 内置模型库管理功能

安装

• 验证是否安装成功

ollama -v
# 监听 Ollama 服务端口
lsof -i :11434 | grep LISTEN

基础命令

# 下载模型到本地
ollama pull deepseek-r1:1.5b
# 列出已有的模型
ollama list
# 运行指定模型
ollama run deepseek-r1:1.5b
ollama run deepseek-r1:7b
# 删除本地模型
ollama rm deepseek-r1:7b

模型下载

• https://ollama.com/library/deepseek-r1
  
  

注：本地资源有限就只下载了 1.5b 版本。

运行模型

ollama list
ollama run deepseek-r1:1.5b

注：首次执行会先下载模型；按 Ctr + d 停止运行

测试模型

API文档

• https://github.com/ollama/ollama/blob/main/docs/api.md

常用API

• http://127.0.0.1:11434
  
• http://127.0.0.1:11434/api/tags
  
• http://127.0.0.1:11434/api/ps
  

Generate 响应

• http://127.0.0.1:11434/api/generate

curl http://127.0.0.1:11434/api/generate \
 -H "Content-Type: application/json" \
 -d '{
      "model": "deepseek-r1:1.5b",
      "prompt": "你好",
      "stream": false
    }' \
 --max-time 30 \
 --no-buffer

Python Generate 示例

import requests
import json

API_URL = "http://localhost:11434/api/generate"

def Demo1():
    data = {
        "model": "deepseek-r1:1.5b",
        "system": "You are a helpful assistant, and you only speak Chinese.",
        "prompt": "你好",
        "stream": False
    }

    response = requests.post(API_URL, json=data)
    print(response)
    json_response = response.json()
    print(json.dumps(json_response, indent=2, ensure_ascii=False))
    print(json_response.get("response", "").strip())

if __name__ == "__main__":
    Demo1()

Chat 响应

• http://127.0.0.1:11434/api/chat

curl http://127.0.0.1:11434/api/chat \
 -H "Content-Type: application/json" \
 -d '{
      "model": "deepseek-r1:1.5b",
      "messages": [
        {"role": "system", "content": "You are a helpful assistant, and you only speak Chinese."},
        {"role": "user", "content": "你好"}
      ],
      "stream": false
    }' \
 --max-time 30 \
 --no-buffer

Python Chat 示例

import requests
import json

API_URL = "http://localhost:11434/api/chat"

def Demo1():
    data = {
        "model": "deepseek-r1:1.5b",
        "messages": [
            {"role": "system", "content": "You are a helpful assistant, and you only speak Chinese."},
            {"role": "user", "content": "你好"}
        ],
        "stream": False
    }

    response = requests.post(API_URL, json=data)
    print(response)
    json_response = response.json()
    print(json.dumps(json_response, indent=2, ensure_ascii=False))
    print(json_response['message']['content'])

if __name__ == "__main__":
    Demo1()

• 响应内容

{
  "model": "deepseek-r1:1.5b",
  "created_at": "2025-03-20T15:10:32.605318Z",
  "message": {
    "role": "assistant",
    "content": "<think>\n\n</think>\n\n你好！有什么我可以帮助你的吗？"
  },
  "done_reason": "stop",
  "done": true,
  "total_duration": 1046775688,
  "load_duration": 25081968,
  "prompt_eval_count": 16,
  "prompt_eval_duration": 69998462,
  "eval_count": 13,
  "eval_duration": 950691930
}

Chatbox

• https://chatboxai.app/zh

Chatbox AI 是一款 AI 客户端应用和智能助手，支持众多先进的 AI 模型和 API，可在 Windows、MacOS、Android、iOS、Linux 和网页版上使用。

Mac 开放局域网访问

# 强制关闭 Ollama
sudo pkill -9 ollama 

• 使用正确的Ollama配置创建plist文件

echo '<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
    <key>Label</key>
    <string>ollama</string>
    <key>ProgramArguments</key>
    <array>
        <string>/bin/bash</string>
        <string>-c</string>
        <string>export OLLAMA_HOST="0.0.0.0:11434" && /usr/local/bin/ollama serve</string>
    </array>
    <key>RunAtLoad</key>
    <true/>
    <key>KeepAlive</key>
    <true/>
</dict>
</plist>' > ~/Library/LaunchAgents/ollama.plist

• 设置权限并重新加载服务

chmod 644 ~/Library/LaunchAgents/ollama.plist
launchctl unload ~/Library/LaunchAgents/ollama.plist 2>/dev/null
launchctl load ~/Library/LaunchAgents/ollama.plist

• 启启 Ollama 服务后

lsof -i :11434 | grep LISTEN

• 局域网内访问测试

curl http://192.168.0.100:11434/api/generate \
 -H "Content-Type: application/json" \
 -d '{
      "model": "deepseek-r1:1.5b",
      "prompt": "你好",
      "stream": false
    }' \
 --max-time 30 \
 --no-buffer