This project provides a comprehensive, practice-oriented tutorial roadmap for developing AI Agents, targeting developers, students, and engineers. It offers a structured learning path from LLM fundamentals to production-ready multi-agent systems, emphasizing hands-on coding and staying current with the latest research through daily arXiv updates.

How It Works

The project is structured as an mdBook tutorial, featuring a systematic progression through AI Agent concepts. Its core innovation lies in its daily automated arXiv paper tracking, integrating frontier research into relevant chapters to ensure content remains cutting-edge. The approach is "code-first," providing runnable Python examples for every concept, complemented by over 120 original SVG diagrams and 5 interactive HTML animations to clarify complex mechanisms like Perceive-Think-Act loops, ReAct, RAG, and multi-agent communication protocols.

Quick Start & Requirements

• Installation: Install mdBook (cargo install mdbook or brew install mdbook) and the mdbook-katex plugin (cargo install mdbook-katex). Clone the repository (git clone https://github.com/Haozhe-Xing/agent_learning.git). Run ./serve.sh to build and start a local server (default port 3000).
• Code Practice Environment: Python 3.11+, venv, pip install langchain langchain-openai langgraph openai anthropic. Requires OPENAI_API_KEY environment variable.
• Online Access: English version: https://Haozhe-Xing.github.io/agent_learning/en/. Chinese version: https://Haozhe-Xing.github.io/agent_learning/zh/.

Highlighted Details

• Daily automated arXiv paper tracking and content updates for frontier research.
• Over 120 original SVG architecture diagrams and flowcharts, plus 5 interactive HTML animations.
• Structured learning paths for beginners, engineers, researchers, and project-based learners.
• Covers advanced topics: Agentic RL (GRPO/DPO/PPO), MCP/A2A/ANP protocols, Context Engineering, and production essentials.
• Includes runnable Python code examples and 3 comprehensive hands-on projects.

Maintenance & Community

The project is actively maintained, with a commitment to daily updates from arXiv. Contributions are welcomed via GitHub Issues and Discussions.

Licensing & Compatibility

Licensed under the MIT License, permitting broad use, modification, and distribution, including for commercial purposes and integration into closed-source projects.

Limitations & Caveats

Several planned features, including runnable agent example projects, an agent glossary, an architecture diagram gallery, interview questions, and a production-ready agent template, are not yet implemented. Interactive animations are primarily accessible via the online e-book.