AI Agents from Beginner to Practice
A Chinese-language tutorial that builds an AI agent from scratch in plain Python, covering the agent loop, tools, memory and planning, before introducing frameworks such as LangChain, LangGraph and Dify.
The tutorial is organized into four parts: understanding what an agent is and its core components, the language model, its tools, its memory and its planning; hand-writing a first agent with a working ReAct loop and then adding memory, tool use and planning to it; agent behavior engineering followed by chapters on the LangChain, LangGraph and Dify frameworks; and a final practical part.
That practical part builds a retrieval-augmented-generation knowledge-base agent, a multi-agent collaboration system, and a capstone 'intelligent research assistant' project that combines everything covered earlier. Each chapter follows the same structure: a plain-language explanation of the concept, complete and annotated runnable code, a demonstration of the expected output, practice exercises, and notes on common mistakes and how to find them. It is licensed MIT.