最近养小龙虾太费钱了，非常愚蠢的选择了按次数收费的Deepseek，应该选千问的code plan，包月模式，没办法了，只能自己硬扛，之前想的是安装qmd来解决，核心解决方案是利用向量检索，来减少每次和大模型的交互上下文，达到token减少的逻辑，这符合正常思路；但是memos更为轻量和柔和，打算装它了；

有cloud版本可以用，但是肯定也是会泄露一些个人信息或者存储限制，我翻阅了一下memos较为轻量，对cpu损耗也可控，就直接装本地了！

我的openclaw版本：3.2.6，不同版本不一样，需要注意下

官网教程：https://memos-claw.openmem.net/docs/index.html#quickstart；安装的过程很简单，但是和openclaw结合起来，很复杂；

1.安装memos

我本地的Mac M1太古老了，也没升级系统，只能是通过源代码模式安装，使用brew安装各种问题，趟坑了算是；
memos源码：https://github.com/usememos/memos/releases

cd ~/Downloads
tar -zxvf memos_0.23.0_darwin_arm64.tar.gz  # 替换为你下载的版本号
# 1. 创建 Memos 专用目录
mkdir -p ~/memos/{bin,data}

# 2. 移动可执行文件并添加权限
mv ~/Downloads/memos ~/memos/bin/
chmod +x ~/memos/bin/memos

# 3. 前台启动测试（M1 原生运行）
cd ~/memos
chmod -R 777 ./data  # 赋予权限
./bin/memos serve --data ./data --port 5230

源码模式安装很简单，没什么弯弯绕绕，打开【http://localhost:5230】看到下图即代表成功；

左侧菜单的底部有个language选择，选择【简体中文】体验更好

为了方便，我打算让它后台一直启动了

nohup ./bin/memos serve --data ./data --port 5230 > memos.log 2>&1 &

2.openclaw配置memos

openclaw plugins install @memtensor/memos-local-openclaw-plugin
openclaw gateway start

打开openclaw.json

"plugins": {
    "allow": [
      "feishu",
      "memos-local-openclaw-plugin"
    ],
    "slots": {
      "memory": "memos-local-openclaw-plugin"
    },
    "entries": {
      "memos-local-openclaw-plugin": {
        "enabled": true,
        "config": {
          "embedding": {
            "provider": "local"
          },
          "summarizer": {
            "provider": "openai_compatible",
            "endpoint": "https://api.deepseek.com/v1",
            "model": "deepseek-chat"
          }
        }
      },
      "memory-core": {
        "enabled": false
      },
      "memory-lancedb": {
        "enabled": false
      }
    },
    "installs": {
      "memos-local-openclaw-plugin": {
        "source": "npm",
        "spec": "@memtensor/memos-local-openclaw-plugin",
        "installPath": "/Users/a58/.openclaw/extensions/memos-local-openclaw-plugin",
        "version": "1.0.2",
        "resolvedName": "@memtensor/memos-local-openclaw-plugin",
        "resolvedVersion": "1.0.2",
        "resolvedSpec": "@memtensor/memos-local-openclaw-plugin@1.0.2",
        "integrity": "sha512-3D/kcQ4nzXr80IJhGwIZ6Tqz60AACjZFsN1cNKZUFXtzGcROpXiIf8AYKdoLSMGLR0/LwHWr3fBDEfPAqJ7HHQ==",
        "shasum": "a8407f617410793dadd2f1a73e20883fcac786e0",
        "resolvedAt": "2026-03-13T03:58:43.931Z",
        "installedAt": "2026-03-13T03:58:52.494Z"
      }
    }

4.排查手段

• 创建memos记忆以后，使用curl命令查看是否有记录; 存储memos笔记的时候千万存成公开笔记，否则查不到的

curl -X GET "http://localhost:5230/api/v1/memos?limit=100" \
  -H "Authorization: Bearer openclaw-memos" \
  -H "Content-Type: application/json"
{"memos":[{"name":"memos/f2y8psnYHfEMcuPcQJk5jF", "state":"NORMAL", "creator":"users/1", "createTime":"2026-03-09T11:28:24Z", "updateTime":"2026-03-09T11:28:24Z", "displayTime":"2026-03-09T11:28:24Z", "content":"DeepSeek 模型关键参数：\n\ncontextWindow：128000\nmaxTokens：4096\n输入 Token 成本：0.14/千Token\n", "visibility":"PUBLIC", "tags":[], "pinned":false, "attachments":[], "relations":[], "reactions":[], "property":{"hasLink":false, "hasTaskList":false, "hasCode":false, "hasIncompleteTasks":false}, "snippet":"DeepSeek 模型关键参数： contextWindow：128000 maxTokens：4096 输入 Token ..."}], "nextPageToken":""}%  

• openclaw logs |grep memos 反复查看日志，持续修正

5.验证效果

memos的ui界面：http://127.0.0.1:18799/