// the find
bojieli/ai-agent-book
《深入理解 AI Agent：设计原理与工程实践》（李博杰 著）开源主仓库：全书正文、编译版 PDF 与按章配套代码
An open-source book on AI agent design and engineering, written in Chinese by Li Bojie, with 10 chapters and 92 companion coding experiments covering context engineering, RAG, tool use, coding agents, evaluation, post-training, and multi-agent systems. It is free (PDF/EPUB), actively maintained, and has community translations in 7 languages. The target audience is developers who want the engineering reality behind agent systems, not just API wrappers.
The 92 companion experiments are the main differentiator — over 70 can run standalone, and they cover everything from building a coding agent to training a small LLM from scratch. The breadth is serious: chapters on evaluation methodology and SFT-vs-RL tradeoffs are not things you find in most agent tutorials. The build pipeline is solid — pandoc/xelatex with automated PDF/EPUB releases and a CI workflow that verifies fonts and i18n drift. The i18n setup with per-language source directories and a dedicated check workflow means translations are at least tracked and not just a README footnote.
The primary source is Chinese, and the community translations lag — if you're reading the English PDF you may be a chapter or two behind the original and there's no clear freshness indicator. Chapters 6, 7, 9, and 10 require manually cloning 19 separate external repos to run the experiments; the book is incomplete without them, which is an annoying setup burden the README glosses over with a single code block. The PDF build chain requires pandoc, xelatex, ElegantBook, and Chinese fonts — non-trivial to get working outside a pre-configured Linux environment. There's a sponsored referral link with a discount code embedded in the README body, which is fine but worth knowing before you share the link internally.