笔记笔记
  • Home
  • AI&ML
  • Example
  • Zoo
  • 关于
⌘ K
Light mode
简介
Ollama
ZeroClaw
OpenClaw
Claude Code
RAGFlow
Stable Diffusion
ComfyUI
PyTorch
Paddle
关注的模型
关注的组织
目标检测
研究
YOLO
其他
文档技巧
最后更新时间：
Copyright © 2023-2026 | Powered by dumi | GuoDapeng | 冀ICP备20004032号-2 | 冀公网安备 冀公网安备13020302001452号

TABLE OF CONTENTS

‌
‌
‌
‌

OpenClaw

在 Fedora Linux 43 (Workstation Edition) Linux 6.18.12-200.fc43.aarch64 虚拟机里测试。

Apple M3 Max 128 GB macOS Tahoe 上测试。

Ollama

qwen3.5:122b 模型可用 success：

似乎需要先让模型常驻内存，才能流畅的配置成功。

模型常驻内存：

shell
curl http://localhost:11434/api/generate -d '{"model": "qwen3.5:122b", "keep_alive": -1}'

安装 zsh （可选）

shell
sudo yum install -y zsh


# 自己解决网络问题
export https_proxy=http://192.168.30.202:7897 http_proxy=http://192.168.30.202:7890 all_proxy=socks5://192.168.30.202:7897


sh -c "$(curl -fsSL https://raw.githubusercontent.com/ohmyzsh/ohmyzsh/master/tools/install.sh)"


# 配置主题
sed -i 's/ZSH_THEME="robbyrussell"/ZSH_THEME="fino-time"/' ~/.zshrc

安装 OpenClaw

安装 nodejs。

shell
sudo yum install -y nodejs


node -v
# v22.22.0

安装 OpenClaw。

shell
mkdir openclaw && cd openclaw


npm config set prefix  ~/openclaw


npm install -g openclaw@latest


export PATH=~/openclaw/bin:$PATH


# 运行新手引导向导（并安装服务）
openclaw onboard --install-daemon


openclaw gateway status

Web UI

shell
ssh -L 18789:127.0.0.1:18789 fedora@192.168.30.35


# 重新生成 网关令牌
openclaw dashboard --no-open

Skills 技能

  • Skills 技能

我喜欢的 skill：

  • find-skills
  • skill-creator

多智能体

添加智能体：

shell
openclaw agents add work
openclaw agents add probe

配置会话：

bindings 需要配置。

json
{
  "agents": {
    "list": [
      {
        "id": "main"
      },
      {
        "id": "work",
        "name": "work",
        "workspace": "/home/fedora/.openclaw/workspace-work",
        "agentDir": "/home/fedora/.openclaw/agents/work/agent"
      },
      {
        "id": "probe",
        "name": "probe",
        "workspace": "/home/fedora/.openclaw/workspace-probe",
        "agentDir": "/home/fedora/.openclaw/agents/probe/agent"
      }
    ]
  },
  "bindings": [
    {
      "agentId": "work",
      "match": {
        "channel": "feishu",
        "accountId": "work",
        "peer": {
          "kind": "group",
          "id": "oc_50...532"
        }
      }
    },
    {
      "agentId": "probe",
      "match": {
        "channel": "feishu",
        "accountId": "probe",
        "peer": {
          "kind": "group",
          "id": "oc_50...532"
        }
      }
    }
  ],
  "channels": {
    "feishu": {
      "enabled": true,
      "accounts": {
        "work": {
          "appId": "cli_a92...385bb5",
          "appSecret": "...",
          "domain": "feishu",
          "groupPolicy": "open"
        },
        "probe": {
          "appId": "cli_a92...e9bbdb",
          "appSecret": "...",
          "domain": "feishu",
          "groupPolicy": "open"
        }
      },
      "groups": {
        "oc_50...532": {
          "requireMention": false
        }
      }
    }
  },
  "...": {},
  "plugins": {
    "...": {},
    "installs": {
      "feishu": {
        "...": "..."
      }
    }
  }
}

查看会话

shell
╭─fedora at MiWiFi-RC06-srv in ~ 26-03-05 - 14:21:37
╰─○ openclaw sessions --help
🦞 OpenClaw 2026.3.2 (85377a2) — I'll butter your workflow like a lobster roll: messy, delicious, effective.


Usage: openclaw sessions [options] [command]


List stored conversation sessions


Options:
  --active <minutes>  Only show sessions updated within the past N minutes
  --agent <id>        Agent id to inspect (default: configured default agent)
  --all-agents        Aggregate sessions across all configured agents (default: false)
  -h, --help          Display help for command
  --json              Output as JSON (default: false)
  --store <path>      Path to session store (default: resolved from config)
  --verbose           Verbose logging (default: false)


Commands:
  cleanup             Run session-store maintenance now


Examples:
  openclaw sessions
    List all sessions.
  openclaw sessions --agent work
    List sessions for one agent.
  openclaw sessions --all-agents
    Aggregate sessions across agents.
  openclaw sessions --active 120
    Only last 2 hours.
  openclaw sessions --json
    Machine-readable output.
  openclaw sessions --store ./tmp/sessions.json
    Use a specific session store.


Shows token usage per session when the agent reports it; set agents.defaults.contextTokens to cap the window and show %.


Docs: docs.openclaw.ai/cli/sessions

发送消息

shell
openclaw message send --target "chat:oc_14b...e99" --message "Hi" --account work
openclaw message send --target "user:ou_ee3...517" --message "Hi" --account work


openclaw agent --agent work --message "Hi" --deliver
Preview
Preview
Preview