Read a structured 10-chapter book explaining how AI agents work, from context engineering to multi-agent collaboration.
Run hands-on coding projects tied to specific chapters, such as building a coding agent or evaluating agent performance.
Follow the book's guidance to set up API keys and clone external benchmark or training repositories for deeper exercises.
| bojieli/ai-agent-book | bbfamily/abu | kvcache-ai/ktransformers | |
|---|---|---|---|
| Stars | 17,133 | 17,126 | 17,156 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | moderate | hard |
| Complexity | 3/5 | 4/5 | 5/5 |
| Audience | researcher | data | researcher |
Figures from each repo's GitHub metadata at analysis time.
Reading the book online needs nothing, compiling the PDF yourself requires pandoc, xelatex, and a LaTeX document class, and some exercises need external repos cloned separately.
This repository is the open source home of a Chinese book called Understanding AI Agents: Design Principles and Engineering Practice, written by Li Bojie. It contains the full text of the book, a compiled PDF and EPUB, and code for 92 hands on projects that go along with each chapter, more than 70 of which can be run directly. The book is built around one core idea, that an agent equals a language model plus context plus tools, and it uses ten chapters to walk from the basics of that idea to real engineering practice. The chapters cover agent fundamentals, context engineering such as caching and prompt design, user memory and knowledge bases including retrieval and knowledge graphs, tool use through protocols like MCP, coding agents and code generation, how to evaluate agent performance, model post training methods such as supervised fine tuning and reinforcement learning, how agents can improve themselves without changing their underlying weights, multimodal and real time interaction covering voice and computer use, and finally multi agent collaboration. The main text is in Chinese, with community translated versions in Traditional Chinese, English, Russian, Tamil, and Vietnamese kept in separate folders, since translations may lag behind the original. There is an online reading site that rebuilds automatically on every update, plus downloadable PDF and EPUB files for offline reading. Readers who want to compile the PDF themselves need tools like pandoc and xelatex. Some hands on projects for the evaluation, training, and multimodal chapters rely on external repositories, such as benchmark suites and robotics platforms, that are not bundled in due to size and licensing, so the book provides a one line clone script for each of the nineteen external projects needed. The repository also lists several AI API providers a reader could sign up with to follow along with the exercises. Overall this is an educational resource for people who want to understand and build AI agents from first principles through to production level engineering.
The open-source companion repository for a Chinese book on AI agent design and engineering, containing the full text, PDF/EPUB downloads, and 92 hands-on coding projects across 10 chapters.
Mainly Python. The stack also includes Python, Pandoc, XeLaTeX.
Apache-2.0 license, free to use, modify, and redistribute, including commercially, as long as you keep the license and notices.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly researcher.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.