WenyuChiou/awesome-agentic-ai-zh

What is this project?

awesome-agentic-ai-zh is a free, structured learning roadmap for AI Agents — AI systems that can decide their own next step and take action to reach a goal you give them. The project is written in Traditional Chinese (with Simplified Chinese and English versions), and its goal is to take you from "what is an AI Agent?" to "I can build a reliable agent system."

It's not a textbook or encyclopedia. Instead, it's a curated path: it tells you what to learn, in what order, and points you to official docs and other resources when you need deep detail. It also includes small, runnable exercises so you practice as you go.

Who is it for?

Three kinds of readers, each with a suggested starting point:

• Complete beginners (never coded): start at Stage 0, which covers basic computer setup.
• People who already know Python, Git, and APIs: start at Stage 1 (LLM basics).
• People who just want to use AI safely in daily life without programming: there's a dedicated "everyday user" guide.

Beyond that, the roadmap offers role-based branches for researchers, developers, teachers, knowledge workers (consultants, PMs, analysts), and everyday users.

How is it structured?

The roadmap has 10 learning stations: a Stage 0 (preparation), Stages 1–8, plus a Stage 7.5 (advanced reading). The core path is:

1. Stage 0 – Basic preparation (computer setup, Python, Git)
2. Stage 1 – LLM basics (what models, tokens, and context are)
3. Stage 2 – Prompt design (how to clearly state goals, data, rules, and output)
4. Stage 3 – Tool use and your first Agent Loop (a loop where the model calls tools, checks results, and repeats)
5. Stage 4 – Workflow graphs and agent frameworks (drawing multi-step processes as maps)
6. Stage 5 – The Claude Code ecosystem (MCP, Skills, Plugins, Hooks, Subagents)
7. Stage 6 – Memory and RAG (how agents look up documents and remember important info)
8. Stage 7 – Production engineering (making agents testable, observable, stoppable, and recoverable)
9. Stage 7.5 – Advanced agentic concepts (optional deep dives)
10. Stage 8 – Agent interfaces (how agents operate browsers, screens, and sandboxes)

There are also two main tracks you choose after finishing Stages 0–2:

• Track A — CLI Power User (~8–10 weeks): learn to use existing command-line agents like Claude Code, Codex, or OpenCode to get work done.
• Track B — Agent Builder (~16–22 weeks, or 5–7 months at 5–8 hours/week): learn to build your own agents, tool loops, workflows, and services.

What makes it useful?

• Clear ordering: each stage answers one core question, and you're told exactly what you'll be able to do when you finish it.
• Plain-language explanations: technical terms are explained in simple words the first time they appear, then the formal English name is kept. A glossary is available for quick lookup.
• Hands-on exercises: each starter.py file is a runnable reference — you read the goal, change one thing, and re-run a test.
• Time estimates: rough guidance so you know what to expect (not deadlines).
• Role-based shortcuts: you don't have to read everything — pick the branch that fits your job.
• Curated resources: a short list of "bookmark first" links (setup guide, glossary, cookbook, course map) plus a full resource list.

How do you start?

You can either read it online (recommended for phone) or clone it locally with git clone. Then open the Stage 0 file, or jump straight to the track that fits your goal using the decision table in the README.

Project details

• License: MIT (free to use, modify, and share)
• Maintainer: WenyuChiou
• Languages: Traditional Chinese (primary), Simplified Chinese, English
• Contributions welcome: you can report errors, suggest resources (with a note on which stage they teach), or submit pull requests following the contribution guidelines
• Inspiration: the project credits Datawhale's Hello-Agents (for deep chapters) and liyupi/ai-guide (for breadth) as related projects — this one focuses on ordering the learning path.

In short: if you want to learn AI Agents but don't know where to start or what order to learn things in, this roadmap gives you a clear, sequenced path with practice built in.