openclaw + ollama 本地源码安装
openclaw 本地源码安装

openclaw + ollama 本地源码安装



本文采用源码安装openclaw,并结合ollama本地运行大模型的方式,运行openclaw。本文安装的版本为OpenClaw 2026.3.13

克隆代码

访问github,获取clone链接,直接clone代码。

链接:https://github.com/openclaw/openclaw

git clone https://github.com/openclaw/openclaw 

安装依赖

首先安装node,官方推荐 Node 24 (recommended) or Node 22.16+,然后安装pnpm

#安装pnpm 
npm install -g pnpm

#安装openclaw依赖
cd openclaw

pnpm install

安装openclaw

pnpm ui:build # auto-installs UI deps on first run
pnpm build

以下是详细配置步骤

1、配置 openclaw

以下步骤中出现选择的,都是按上下键切换选项,按空格键选中

# pnpm openclaw onboard --install-daemon

 I understand this is personal-by-default and shared/multi-user use requires lock-down. Continue?
   Yes /  No   # 选择yes

  Onboarding mode
  QuickStart     # 选择QuickStart

  Model/auth provider
  Cloudflare AI Gateway
  Custom Provider    # 所有的兼容openai的都可以选这个包括ollama
  API Base URL
  http://127.0.0.1:11434/v1   #注意此处没有带/v1的话需要手动填写上
  How do you want to provide this API key?
   Paste API key now   # ollama 默认没有认证 所以直接选这个然后随便填写几个字母就行
  API Key (leave blank if not required)
  ollama               # 我直接填写ollama
  Endpoint compatibility
   OpenAI-compatible (Uses /chat/completions)  # 直接选openai兼容接口
   Anthropic-compatible
   Unknown (detect automatically)
  Model ID
  qwen/qwen3-4b-thinking-2507          # 直接填写你想使用的模型
  Verification successful.

  Endpoint ID
  custom-127-0-0-1-12345    # 可以不修改 

  Model alias (optional)
  qwen3-4b-thinking-2507    # 可以随便写,注意: 这一步写完之后,不要操作,稍等一下
Configured custom provider: custom-127-0-0-1-12345/qwen/qwen3-4b-thinking-2507
  Select channel (QuickStart)
   Telegram (Bot API) (recommended · newcomer-friendly)
   Skip for now            # 配置channel 需要的话直接选择填写,没有的话直接Skip for now
  Web search ────────────────────────────────────────╮
                                                     
  Web search lets your agent look things up online.  
  Choose a provider and paste your API key.          
  Docs: https://docs.openclaw.ai/tools/web           
                                                     
├─────────────────────────────────────────────────────╯

  Search provider
   Brave Search (Structured results · country/language/time filters)
   Gemini (Google Search)
   Grok (xAI)
   Kimi (Moonshot)
   Perplexity Search
   Skip for now   # 需要的话直接选择填写,没有的话直接Skip for now
  Skills status ─────────────╮
                             
  Eligible: 4                
  Missing requirements: 39   
  Unsupported on this OS: 8  
  Blocked by allowlist: 0    
                             
├─────────────────────────────╯

  Configure skills now? (recommended)
   Yes /  No                 # 选择 yes
  Install missing skill dependencies
   Skip for now (Continue without installing dependencies)  #需要的话直接选择填写,没有的话直接Skip for now
  Set NOTION_API_KEY for notion?
  No                                 # 以下几个都选择No
  Set NOTION_API_KEY for notion?

  Set OPENAI_API_KEY for openai-image-gen?
  No

  Set OPENAI_API_KEY for openai-whisper-api?
  No

  Set ELEVENLABS_API_KEY for sag?
  No

  Hooks ──────────────────────────────────────────────────────────────────╮
                                                                          
  Hooks let you automate actions when agent commands are issued.          
  Example: Save session context to memory when you issue /new or /reset.  
                                                                          
  Learn more: https://docs.openclaw.ai/automation/hooks                   
                                                                          
├──────────────────────────────────────────────────────────────────────────╯

  Enable hooks?
   Skip for now                #需要的话直接选择填写,没有的话直接Skip for now
   🚀 boot-md
   📎 bootstrap-extra-files
   📝 command-logger
   💾 session-memory
  Gateway service runtime ────────────────────────────────────────────╮
                                                                      
  QuickStart uses Node for the Gateway service (stable + supported).  
                                                                      
├──────────────────────────────────────────────────────────────────────╯

2、运行网关

上面配置就结束了,下面另起一个命令窗口

# 运行gateway
pnpm gateway:watch

# 这时候打开浏览器访问 http://localhost:18789 

自此openclaw安装完成

实用技巧

如果需要访问令牌执行以下命令

pnpm openclaw dashboard --no-open

如果想要改变端口,执行以下命令

pnpm openclaw gateway --port 1878

如果出现以下告警,原因是ollama的上下文长度太小了,需要修改默认的上下文长度

 [agent/embedded] low context window: custom-127-0-0-1-12345/qwen/qwen3-4b-thinking-2507 ctx=16000 (warn<32000) source=modelsConfig  

 # 当你修改完之后,还会出现这个告警,这时候需要你修改openclaw.json的模型配置参数,将contextWindow,修改成正确的值

   "models": {
    "mode": "merge",
    "providers": {
      "custom-127-0-0-1-12345": {
        "baseUrl": "http://127.0.0.1:12345",
        "apiKey": "ollama",
        "api": "openai-completions",
        "models": [
          {
            "id": "qwen/qwen3-4b-thinking-2507",
            "name": "qwen/qwen3-4b-thinking-2507 (Custom Provider)",
            "reasoning": false,
            "input": [
              "text"
            ],
            "cost": {
              "input": 0,
              "output": 0,
              "cacheRead": 0,
              "cacheWrite": 0
            },
            "contextWindow": 64000,
            "maxTokens": 4096
          }
        ]
      }
    }