83.5K ·GitHub
What is this?
Hello-Agents is Datawhale's free, open-source tutorial on building AI agent systems from scratch — theory and hands-on practice in equal measure, taking you from LLM user to agent builder.
Hello-Agents is Datawhale's free, open-source tutorial on building AI agent systems from scratch — theory and hands-on practice in equal measure, taking you from LLM user to agent builder.Is it a book to read, or code to run?Both: a written tutorial in 16 chapters across five parts, with all the companion code in the repo's code folder that you are meant to run and modify as you go.
Both: a written tutorial in 16 chapters across five parts, with all the companion code in the repo's code folder that you are meant to run and modify as you go.Do I need to download anything before I can start learning?No — there are online reading links, including a China-accelerated one, and a local option if you want to read or contribute on your own machine.
No — there are online reading links, including a China-accelerated one, and a local option if you want to read or contribute on your own machine.How much Python do I need to know beforehand?Basic Python and a conceptual grasp of LLMs — e.g., having called one through an API. Deep algorithm or model-training experience is explicitly not required.
Basic Python and a conceptual grasp of LLMs — e.g., having called one through an API. Deep algorithm or model-training experience is explicitly not required.Is the material in Chinese only, or can I follow it in English?The chapters, community posts, and discussions are all written in Chinese, so comfortable Chinese reading is a practical requirement.
The chapters, community posts, and discussions are all written in Chinese, so comfortable Chinese reading is a practical requirement.I'm a student teaching myself AI — is this meant for someone like me?Yes — the README names students and self-learners right alongside AI developers and software engineers as its ideal readers.
Yes — the README names students and self-learners right alongside AI developers and software engineers as its ideal readers.Can I use the material for commercial training at work?The project is free to learn, but reuse is non-commercial only — the README states a Creative Commons Attribution-NonCommercial license, so share and adapt with attribution, don't sell it.
The project is free to learn, but reuse is non-commercial only — the README states a Creative Commons Attribution-NonCommercial license, so share and adapt with attribution, don't sell it.Do I use existing frameworks, or build my own?Both — you first try low-code platforms like Coze, Dify, and n8n and frameworks like LangGraph and AutoGen, then chapter 7 has you build your own HelloAgents framework from zero on the OpenAI-native API.
Both — you first try low-code platforms like Coze, Dify, and n8n and frameworks like LangGraph and AutoGen, then chapter 7 has you build your own HelloAgents framework from zero on the OpenAI-native API.What real projects will I have built by the end?An intelligent travel assistant driven by multi-agent cooperation, a DeepResearch-style automated research agent, and a simulated cyber town showing social dynamics — capped by a graduation project: your own complete multi-agent application.
An intelligent travel assistant driven by multi-agent cooperation, a DeepResearch-style automated research agent, and a simulated cyber town showing social dynamics — capped by a graduation project: your own complete multi-agent application.How deep does it go — does it cover RAG and agent training?Yes: memory and retrieval with RAG, context engineering, agent communication protocols (MCP, A2A, ANP), Agentic RL training from SFT through GRPO, and performance evaluation against benchmarks.
Yes: memory and retrieval with RAG, context engineering, agent communication protocols (MCP, A2A, ANP), Agentic RL training from SFT through GRPO, and performance evaluation against benchmarks.Is the material still being updated?Yes — video courses are rolling out gradually, the HelloAgents framework reached V1.0.0 with more lightweight tools planned, a new learning route was slated for May 2026, and a follow-up book on training agents is on the way.
Yes — video courses are rolling out gradually, the HelloAgents framework reached V1.0.0 with more lightweight tools planned, a new learning route was slated for May 2026, and a follow-up book on training agents is on the way.What are similar projects?ai-agent-book, AgentGuide, ai-agents-from-zero, ai-agents-the-definitive-guide, learn-agent, ai-system-design, Agentic-Design-Patterns, llm-cookbook
ai-agent-book, AgentGuide, ai-agents-from-zero, ai-agents-the-definitive-guide, learn-agent, ai-system-design, Agentic-Design-Patterns, llm-cookbookWhich topics is it in?langgraph, llm-agents, context-engineering, curriculum, course, reinforcement-learning, evaluation, prompt-engineering, rag, multi-agent, agents, learning
langgraph, llm-agents, context-engineering, curriculum, course, reinforcement-learning, evaluation, prompt-engineering, rag, multi-agent, agents, learning