# Snailclimb/AIGuide: An Open Source Curriculum for AI Application Development > AIGuide is a Chinese-language guide to LLM, RAG, Agent, MCP and AI coding practice, aimed at backend and product engineers rather than algorithm researchers. It is a reading curriculum with two installable Agent Skills, not a library you import. **Snailclimb/AIGuide** — AI 应用开发、AI 编程实战与面试指南，涵盖 LLM、Agent、RAG、MCP、Claude Code、Codex 等核心技术与工程实践。 - Repository: https://github.com/Snailclimb/AIGuide - Website: https://javaguide.cn/ai/ - Stars: 658 · Forks: 74 - Language: Unknown - License: not declared - Published: 2026-09-10 · Updated: 2026-09-10 · Language: en - Canonical page: https://hysenlabs.com/projects/snailclimb-aiguide ## Who AIGuide Is Written For The README addresses one audience explicitly: people who want to learn AI application development and AI engineering without first switching to an algorithm role. It names backend, frontend, test, architect, technical manager and product-technical readers, and says you do not need to start with training frameworks or paper formulas. That framing matters, because most AI material published in Chinese assumes you will eventually fine-tune something. AIGuide assumes you will call something. The author is the maintainer of JavaGuide, a long-running Java learning site. The README leans on that lineage: if you already work in Java or Go, the argument goes, your experience with concurrency, caching, databases, message queues and observability transfers directly. Readers from frontend, testing or product backgrounds are pointed at Prompt, RAG, Agent, AI coding, evaluation and system design as entry points instead. The split is honest about what it is: a reading path, not a toolkit. ## What the Repository Actually Contains The top level of the repository holds .gitignore, README.md and a skills/ directory. There is no source tree for an application, no build file, no package manifest. The substance lives on the linked site at javaguide.cn/ai/, with a separate AI coding section at javaguide.cn/ai-coding/. The README is best understood as a table of contents with commentary. The reading order the README proposes is four stages. First, LLM basics: tokens, context windows, sampling parameters, API calls, structured output and evaluation. Second, RAG: document processing, vector retrieval, the update pipeline and evaluation, described as the most common enterprise use case and the one with the most pitfalls. Third, AI Agent: tool calling, memory, MCP, Skills, and the Workflow/Graph/Loop distinction. Fourth, AI system design: gateways, rate limiting, fallback, cost, observability, security and canary rollout. AI coding is presented as a parallel habit rather than a fifth stage, practiced with Claude Code, Codex, Cursor or Trae while you read. Within those stages the README lists individual articles with one-line descriptions. The RAG section alone covers document chunking, vector index algorithms and database choice (HNSW, IVFFLAT, pgvector, Milvus, Elasticsearch are named), knowledge base update strategies including incremental update and full rebuild, GraphRAG, and retrieval optimization through query rewrite, hybrid search and rerank. That level of granularity is the useful part: it tells you what a topic contains before you click. ## Installing the drawio-chart and java-coding-standards Skills The repository ships two Agent Skills under skills/. The README describes drawio-chart as generating draw.io diagrams from requirements, covering flowcharts, architecture diagrams, sequence diagrams and ER diagrams, with export to PNG, SVG or PDF. The second, java-coding-standards, encodes engineering conventions for Java and Spring Boot projects across layering, code style, transactions, exceptions, logging, performance and testing. The documented install path uses the skills CLI. Running it fetches the skill from the GitHub repository by path: ```bash npx skills add Snailclimb/AIGuide --path skills/drawio-chart ``` The README does not state where the CLI places the skill on disk or which agent runtimes it targets, so check your agent's skills directory after the command completes. For Codex users the README gives a second route that installs straight from GitHub with a bundled installer script: ```bash python3 ~/.codex/skills/.system/skill-installer/scripts/install-skill-from-github.py \ --repo Snailclimb/AIGuide \ --path skills/drawio-chart ``` Swap the --path value for skills/java-coding-standards to install the Java conventions skill instead. Both commands are copied verbatim from the README; the file path to the installer script is specific to a Codex installation, so it will not exist if you use a different agent. A first real use is to ask your agent for an architecture diagram and see whether the skill triggers, since the README describes Skills as triggered and lazily loaded rather than always resident in context. ## The Limits of a Guide With No Runnable Code The most important limitation is structural. AIGuide is prose and diagrams. The repository has no library, no CLI of its own, no importable package. If you arrived looking for a RAG framework or an agent runtime, you are in the wrong place, and the README never claims otherwise. Language is the second constraint. The README and the linked articles are written in Chinese. The repository description and topic tags are partly English, but there is no indication of an English translation, so a reader without Chinese is limited to the two skills and the link structure. There is also a scope gap worth naming. The README promises real engineering scenarios, key parameters and pitfalls, yet the topics list includes Go and Java while the sample articles shown are Java-centric (JVM diagnosis, IDEA plugins, Spring Boot conventions). A Go backend reader may find less directly applicable material than the README's framing suggests. Finally, the README says content is still being updated, and the listed article links are the only evidence of what exists; nothing in the repository guarantees that every listed page is published or current. ## How AIGuide Differs From Framework Documentation The natural comparison is with the documentation of a framework such as LangChain or LlamaIndex. Those projects answer a different question: how do I build this with our abstractions. Their docs are organized around APIs, classes and configuration, and they change with each release. AIGuide is organized around concepts and decisions instead, and it is not tied to any vendor's API surface. That difference cuts both ways. A guide cannot break when a library ships a major version, but it also cannot be executed. When AIGuide discusses vector index algorithms or a model gateway, you get the reasoning and the trade-offs, not a working implementation you can run. For a team deciding between pgvector and Milvus, that reasoning is the expensive part and the guide is a reasonable starting point. For a team that needs a retrieval pipeline running this week, framework documentation plus a prototype will get you further, and the guide is background reading. The two are complementary rather than competing. Read AIGuide for the problem framing, then go to the framework docs for the call signatures. ## Conclusion Adopt AIGuide if you are a working engineer who wants a structured reading path through LLM APIs, RAG, Agent loops, MCP and AI system design before you commit code to a side project, and if you read Chinese comfortably. Skip it if you need a runnable framework, an English-language reference, or a library with a versioned API. Before relying on it, open the online edition at javaguide.cn/ai/ and confirm the sections you need are actually published, then run the drawio-chart install command and check that the skill lands in your skills directory. ## FAQ ### What is Snailclimb/AIGuide? It is an open source Chinese-language guide to AI application development, AI coding practice and AI interviews, covering LLM basics, RAG, Agent, MCP, Prompt engineering, evaluation and system design. The repository itself contains a README and a skills directory, with the articles hosted at javaguide.cn/ai/. ### How do I use Snailclimb/AIGuide? Read it as a curriculum: the README recommends starting with LLM basics, then RAG, then AI Agent, then AI system design, while practicing AI coding tools alongside. The repository also provides installable Agent Skills, which you add with the npx skills add command or the Codex installer script. ### Is Snailclimb/AIGuide a library or framework I can install? No. The top level holds only .gitignore, README.md and skills/, with no application source or package manifest. The only installable pieces are the two Agent Skills, drawio-chart and java-coding-standards. ### What can the drawio-chart skill in Snailclimb/AIGuide do? According to the README it generates draw.io diagrams from requirements, including flowcharts, architecture diagrams, sequence diagrams and ER diagrams, and can export them to PNG, SVG or PDF. ### Does Snailclimb/AIGuide cover Claude Code? Yes. The repository is tagged claude-code and codex, and the AI coding section of the site covers Claude Code, Codex, AI IDEs and CLI agents, including a case study on connecting Claude Code to a third-party model for JVM diagnosis. ## Sources - [Issues](https://github.com/Snailclimb/AIGuide/issues) - [Project website](https://javaguide.cn/ai/) - [README](https://github.com/Snailclimb/AIGuide/blob/main/README.md) - [Snailclimb/AIGuide on GitHub](https://github.com/Snailclimb/AIGuide) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/snailclimb-aiguide