准备工作

• debian虚拟机
• axonhub
• chrome

安装并运行

conda create -n pi conda-forge::nodejs
conda run -n pi npx @agegr/pi-web@latest --hostname 0.0.0.0

添加模型和插件

• 模型格式选择 response 启用缓存
• 一些好用的插件

npm:pi-hashline-edit-pro
npm:@juicesharp/rpiv-ask-user-question
npm:pi-rtk-optimizer
npm:@ff-labs/pi-fff
npm:pi-interactive-shell

安装skill

skills.sh/leeguooooo/chrome-use/chrome-use
skills.sh/badlogic/pi-skills/brave-search

配置插件外部依赖

• 安装好chrome，并完成初始设置

conda install -n pi conda-forge::rtk-cli
conda install -n pi reportlab pdfplumber pypdf
conda install -n pi pip

micromamba activate pi
curl -fsSL https://cli.tavily.com/install.sh | bash 
tvly login --api-key <你的api-key>
tvly search "limour's blog" --json
micromamba deactivate


AGENT_BROWSER_NO_SKILL=1 curl -fsSL https://raw.githubusercontent.com/leeguooooo/chrome-use/main/install.sh | sh

• nano ~/.local/bin/agent-browser 写入以下内容

#!/usr/bin/env sh
exec "$HOME/.local/bin/chrome-use" "$@"

• chmod +x ~/.local/bin/agent-browser
• chrome-use extension install

注意事项

• 接入格式用兼容 Anthropic 接口对工具识图更友好

效果图

附加 拓展推荐

npm:pi-mcp-adapter
git:github.com/justhil/pi-image-gen