Stanford's AI agents course, translated for Python devs who read Chinese

What it does This repo turns Stanford’s public CS329Z material (Engineering AI Agents) into a self-paced Chinese tutorial aimed at Python developers. It spans 19 lessons and 4 hands-on exercises, walking from LLM basics and RAG through tool calling, memory, multi-agent collaboration, evaluation, and safety. Everything ships as Markdown sources plus a web reader with search, a table of contents, and dark/light themes — and a single-file offline.html if you’d rather keep one file on a plane.

The interesting bit It’s not just translation: there’s a code-guide.md walking you through the HW1 and Lesson 5 companion code, and an editorial-notes.md that openly tracks which source materials are unpublished and what still needs fixing. That kind of errata honesty is rare in fan-made course notes.

Key highlights

• 19 lessons covering the full agent stack: LLMs, RAG, tools, memory, multi-agent, eval, safety
• 4 practical exercises, including building an agent from scratch and designing evaluations
• Suggested learning path: lessons 1–4 first, then HW1 hands-on
• Runs fully offline via a single self-contained HTML file
• Build tooling included (tools/build_site.mjs, tools/verify_publication.py) for regenerating the site from Markdown

Caveats

• Explicitly not an official Stanford Chinese edition — an independent study guide, with rights to originals retained by their authors
• The editorial notes flag ongoing revision issues, so expect some rough edges in specific sections

Verdict Worth a look if you write Python, read Chinese, and want a structured path into agent engineering without wading through English lecture notes. If you’re after the original course or official translations, this isn’t that — and it doesn’t pretend to be.