// the find
didilili/ai-agents-from-zero
🚀 2026 最系统的 AI Agent 速成指南｜智能体实战教程 · 完整学习路径 + 实战项目 + 面试题库 · 对标大模型应用开发工程师岗位 · 覆盖LangChain / LangGraph / Coze / Dify / MCP / skills / LLM / RAG / 提示词 · 企业级部署与微调 · 从0到企业级落地 + 从学习到上线项目 + 面试准备一体化
A Chinese-language tutorial repo covering the full AI agent engineering stack: LangChain, LangGraph, RAG, MCP, Coze/Dify low-code platforms, and LLM fine-tuning. The target audience is Chinese developers trying to land AI application engineering roles, and the curriculum is explicitly structured around what shows up in job descriptions and bootcamp syllabi. Two complete end-to-end projects (NL2SQL with LangGraph, multi-agent deep research) have separate source repos.
The two flagship projects are genuinely end-to-end — the NL2SQL one wires together MySQL, Qdrant, Elasticsearch, LangGraph, and FastAPI SSE in a single runnable pipeline, which is more complete than most tutorial projects. The interview question bank is a real differentiator: it's cross-linked to chapters by topic and explicitly aligned to what Chinese tech companies ask, not generic Q&A filler. The LangGraph coverage (chapters 22–26) goes deeper than most English-language resources — state machines, persistence, streaming, and multi-agent A2A are all addressed. Active maintenance through mid-2026 means the MCP and A2A content reflects the current ecosystem rather than a snapshot from 2023.
The repo is 100% Chinese — every file, comment, and example. Non-Chinese readers get nothing from this. The README admits the content derives from 尚硅谷's paid bootcamp materials; this is a structured retelling with additions, not original curriculum, and that ceiling shows in how the chapters mirror a typical training course outline rather than problem-driven learning. Twenty-eight chapters covering Transformer internals, Docker, LoRA fine-tuning, and Coze workflows means several sections are thin — breadth over depth is the real tradeoff here. The sponsorship banner (a GPU cloud referral link with a 5 yuan signup bonus) sits at the top of the README and the project openly references driving traffic to that platform, which muddies whether certain tool choices reflect genuine recommendations.