Claude Skill
lasywolf/Learn-OpenClaw
Learn-OpenClaw is a Python tutorial that teaches AI agent fundamentals from scratch in about 9 hours. Perfect for beginners who want to understand how simple agents really are.
Overview
Repository
Install this Skill
git clone https://github.com/lasywolf/Learn-OpenClaw.gitRegistry
Summary
Learn-OpenClaw is a hands-on tutorial repository that teaches agent fundamentals from scratch in approximately 9 hours. Written in Python, it aims to demystify AI agents by showing how simple they really are, making it ideal for beginners.
零基础一天(约9小时)学完Agent基础!写这个教程就是想告诉大家,Agent其实非常简单!
Key features
- Zero-to-hero agent tutorial in ~9 hours
- Written entirely in Python for clarity
- Focuses on core agent fundamentals
- Designed for absolute beginners
- Practical, hands-on learning approach
Use cases
- Learning AI agent basics quickly
- Building a solid foundation for agent development
- Teaching agent concepts in workshops or classes
- Self-paced study for Python developers new to agents
- Preparing for more advanced agent frameworks
README excerpt
[中文](./README.md) | [English](./README-en.md) # 这个tutorial能干什么 零基础一天 (9小时)学完agent!写这个教程就是想告诉大家,Agent其实非常简单! 并且能帮助你找到 Agent相关工作/实习!目前有很多个同学看我的教程找到了实习,且本教程在同学群里备受好评,现在开源给大伙! ## 赞助商 [](https://myclaw.ai?utm_source=github&utm_campaign=learn-openclaw) **[MyClaw.ai](https://myclaw.ai?utm_source=github&utm_campaign=learn-openclaw)** · 不想自己折腾服务器的话,可以用云端托管的 OpenClaw 实例:一键设置、24/7 在线,并且保留完整代码控制权和数据所有权。[立即开始 ->](https://myclaw.ai?utm_source=github&utm_campaign=learn-openclaw) ## 总体内容展示 ### 学会 Agent(学习需约 1天 * 9小时) 1. 拥有你自己的llm api-key(阅读需约15分钟) - 你可能需要学会使用python和[uv](https://docs.astral.sh/uv/getting-started/installation/) 用 rust 编写,类似于 rust 里面的 cargo,非常非常快 - 为什么不用conda而是uv:uv 开源无商用风险,Conda 在超过 200 人的组织有[潜在商用授权问题](https://www.anaconda.com/blog/is-conda-free) - uv 可以像 pip 一样编辑镜像源,mac 和 linux 系统修改 `~/.config/uv/uv.toml` 并写入类似于下面的内容,Windows 系统可以自己查下怎么配置。项目初始化需要`uv sync`。 ``` [[index]] url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple" default = true ``` - 你可能需要llm api-key,推荐[DeepSeek](https://platform.deepseek.com/)、[kimi](https://platform.moonshot.cn/docs/overview)或者[智谱](https://open.bigmodel.cn) - 配置环境变量`OPENAI_API_KEY`和`OPENAI_BASE_URL`,并且尝试运行[`core/llm.py`](./core/llm.py) 2. 实现 Node / Workflow / Agent (阅读需约1小时) - 我们最终的目标是造一个Agent,能够联网搜索、运行命令行、文件编辑。 - Agent底层可以使用Node来抽象,我已经准备好了一个极简的实现,可以看[`core/node.py`](./core/node.py),不到60行就实现了一个Agent的轻框架,实在是太容易理解了!如果没有py基础看不懂,可以把代码复制给ai让它来解释。 - 但我们应该怎么去用Node呢,我们可以新建3个Node并把它们串起来实现功能`接收输入->上网搜索->大模型生成总结`,恭喜你已经实现了workflow,相关实现已经在[`examples/workflow`](./examples/workflow) - 现在新建个workflow,实现功能`接收用户输入->大模型回复`,并且loop反复调用这个workflow,恭喜你已经实现了chatbot,相关实现已经在[`examples/chatbot`](./examples/chatbot) - 现在尝试给chatbot一些tools(下文
Topics
No topics yet.