Hello-Agents
Hello-Agents is a Chinese-language tutorial from the Datawhale community that teaches agent theory and practice, from classic paradigms such as ReAct through building a custom agent framework from scratch.
The sixteen chapters move from agent history and large-language-model basics, through hands-on implementations of ReAct, Plan-and-Solve and Reflection, a survey of low-code agent platforms such as Dify, Coze and n8n alongside code-first frameworks like AutoGen, AgentScope and LangGraph, and building an original framework called HelloAgents on the OpenAI API from zero.
Later chapters cover memory and retrieval, context engineering, agent communication protocols including MCP, A2A and ANP, agentic reinforcement learning from supervised fine-tuning through GRPO, and performance evaluation, before capstone projects such as a travel-planning assistant and a simulated multi-agent town.
The material is free and open source, readable online or as a downloadable PDF, with community write-ups supplementing the core chapters. Python.