# LearnPrompt/LearnPrompt: a free Chinese AI course site built as a Docusaurus project > LearnPrompt is a permanently free, open source Chinese-language AI tutorial site covering Claude Code, Codex, ChatGPT, Midjourney, Stable Diffusion and more. Its repository is a Docusaurus site, not a library, so adoption means running the docs locally or reading them on the web. **LearnPrompt/LearnPrompt** — 永久免费开源的 AIGC 课程, 目前已支持Claude Code，Codex，Hermes，OpenClaw，Obsidian，Prompt Engineering, ChatGPT, Midjourney, Runway, Stable Diffusion, AI数字人，AI声音&音乐，开源大模型 - Repository: https://github.com/LearnPrompt/LearnPrompt - Website: https://www.learnprompt.pro/ - Stars: 2,668 · Forks: 223 - Language: MDX - License: NOASSERTION - Published: 2026-09-10 · Updated: 2026-09-10 · Language: en - Canonical page: https://hysenlabs.com/projects/learnprompt-learnprompt ## What LearnPrompt actually is, and who it is written for The repository is a Chinese-language AI tutorial site. The README describes it as a permanently free and open source Chinese AI practical course, and says the author, who signs as 卡尔, has been doing AI tool testing, content creation and agent workflows. The stated audience is people meeting these tools for the first time: the README says that if you are studying AI systematically for the first time, you can start here. The topic list is wide. The README names Claude Code, Codex, OpenClaw, Hermes, prompt engineering, ChatGPT, RAG, agents, AI coding, Agent Skills, Obsidian, Midjourney, Runway, Stable Diffusion, digital humans, AI voice and music, and large model fine-tuning. That breadth is the point and also the weakness: a single repository covering this many fast-moving tools will drift, and the README itself admits that many old tutorials are already out of date. The README also states that the repository doubles as the GitHub profile README for the LearnPrompt account and as the project repository for the tutorial site. So the front page you land on is partly a personal introduction and partly a changelog of where the site is heading. If you came looking for a library to import, this is the wrong repository. ## The mechanism: an MDX content tree rendered by Docusaurus The site is a Docusaurus 2 project. The package.json lists @docusaurus/core, @docusaurus/preset-classic, @docusaurus/plugin-client-redirects, @docusaurus/plugin-google-analytics, @docusaurus/plugin-ideal-image and @docusaurus/theme-mermaid, all at ^2.4.0. Content lives in MDX, which matches the repository's primary language field. The top-level layout is a normal Docusaurus tree plus extras. There is a docs directory for content, a blog directory, an i18n directory for translations, a static directory for assets, sidebars.js for navigation, and docusaurus.config.js for site configuration. The crowdin.yml file indicates the translation workflow is managed through Crowdin, which lines up with the i18n directory. Several dependencies are content tooling rather than UI. @benchmark-urbanism/remark-bibtex is pulled from a GitHub URL and @renatonagliati/remark-auto-glossary is pinned to a specific commit, so citations and glossary terms are generated at build time from bibliography.bib and glossary.yml. There is also docusaurus-plugin-mindmap and @kailualabs/multimodal-search. The practical consequence is that a content change can require a matching entry in bibliography.bib or glossary.yml, and a build that fails on a citation is a content problem, not an infrastructure one. One more detail worth noting: the package.json name field is promptgineering and the version is 0.0.0 with private set to true. That is a leftover from the project's origin and a clear signal that this is not published to a registry. ## Installing LearnPrompt locally and running a first build There are no install instructions in the README. The repository's package.json scripts are the source of truth for how the site is run, and they are standard Docusaurus commands. Because the package.json declares private: true, you clone the repository rather than installing it as a package. Start by installing dependencies from the repository root: ```bash npm install ``` Then start the development server. The start script maps to docusaurus start, which serves the site with hot reload so you can edit an MDX file under docs and see the page update. ```bash npm run start ``` The production build uses docusaurus build, and the serve script runs docusaurus serve to preview the built output locally. ```bash npm run build npm run serve ``` Two scripts are specific to content maintenance rather than deployment. write-translations generates the translation files that Crowdin consumes, and write-heading-ids generates stable anchor IDs for headings, which matters if you rename a section and want existing links to keep working. ```bash npm run write-translations npm run write-heading-ids ``` If you only want to read the course, the README points to learnprompt.pro as the official site, and that is the cheaper path. ## Where LearnPrompt is the wrong tool The first limitation is language. The README states the project keeps a Chinese learning path as its main line. There is a README_EN.md and an i18n directory, but the course content is written for Chinese readers, and the README explicitly frames the new version as continuing that main line. If your team needs English-first material, you are reading a translation or the wrong project. The second is scope drift. The README concedes that many old tutorials are out of date, and the repository's own topic list spans prompting, image generation, video generation, voice, music, digital humans and fine-tuning. A tutorial that covers everything ages faster than one that covers one thing, and the README's stated plan is to move from tool introductions toward real tasks, which means the current structure is mid-transition. The third is that this is not a dependency. There is no published package, no API, and no versioned release. The package.json version is 0.0.0 and private. You cannot pin a version of the course and expect it to behave the same next month; the content is the product, and the content changes. The fourth is that the README does not document rollback, contribution review or a deprecation policy for outdated pages. If you fork it for a team, you inherit that gap. ## How it compares with Learn Prompting and other prompt courses The obvious comparison is Learn Prompting, which the search data shows people confuse with this project. Both are free prompt-engineering learning resources, and both are web-based. The difference is in the approach: Learn Prompting is an English-first prompt engineering course, while LearnPrompt/LearnPrompt is a Chinese-language site whose current focus has moved toward agent tooling such as Claude Code, Codex, OpenClaw and Hermes, alongside the older ChatGPT, Midjourney and Stable Diffusion material. The second difference is the delivery format. Learn Prompting is a course you read. LearnPrompt is a Docusaurus repository you can clone, run and fork, with bibliography.bib, glossary.yml and Crowdin translation files as first-class parts of the tree. If you want to adapt the material for an internal team, the repository gives you the build pipeline; a hosted course does not. The third difference is maintenance shape. This repository's last push was on 2026-09-09, so the tree is recent. But the README describes a transition in progress, from a static AIGC tutorial collection toward an AI-native course maintained with wiki skills and reusable workflows. That means the structure you clone today is not the structure the README promises. A generic video course is a third option and differs in the opposite direction: no repository, no fork, no build step, but also no way to fix a page that has gone stale. ## Licence, maintenance and the cost of keeping a fork current The repository's licence is reported as NOASSERTION, which means GitHub could not match the LICENSE file to a known licence template. The LICENSE file exists at the top level, so read it directly before reusing content. This is not legal advice, but the practical implication is that you should not assume the terms match MIT or CC BY just because the README calls the project free and open source. Maintenance cost for a fork is mostly content cost. The Docusaurus dependencies are pinned to ^2.4.0, so a major Docusaurus upgrade is a separate project. Two dependencies are pinned to a GitHub URL and a specific commit, which means npm install resolves them from source and a fork inherits that fragility. The bibliography and glossary plugins mean every new page may need a matching bibliography.bib or glossary.yml entry, and the build will tell you when it does not. Translation cost is real if you fork for a non-Chinese audience. The i18n directory and crowdin.yml exist, so the machinery is there, but the README does not describe how complete the existing translations are. Check the i18n tree before budgeting that work. ## Conclusion Adopt LearnPrompt if you need a free Chinese-language starting path through prompting and current agent tooling, or if you want to fork a Docusaurus site and rewrite the content for your own team. Do not adopt it if you need English-first material with a certificate, or if you expect a versioned, API-stable package; the repository is a docs site whose package.json is marked private. Verify first that the topics you care about are still current in the docs directory, then run npm run start locally to see the actual page tree rather than trusting the README's project list. ## FAQ ### Is LearnPrompt free to use? The README describes LearnPrompt as a permanently free and open source Chinese AI course, and the site is published at learnprompt.pro. The repository also carries a LICENSE file, which is reported as NOASSERTION, so read it before reusing content. ### How do I learn to prompt AI with LearnPrompt? The README says the site is a continuously maintained Chinese AI learning path and that first-time systematic learners can start there. The topics it lists include prompt engineering, ChatGPT, RAG and agents, so the path begins with prompting and moves into tool and agent workflows. ### What does prompt mean? The repository does not define the term directly. It treats prompting as a subject area, listing prompt engineering as one of the topics the course covers, so the definition lives inside the docs directory rather than the README. ### Is LearnPrompt a good resource? The README positions it as a starting path for people new to AI, and states it is permanently free and open source. It also acknowledges that many older tutorials are out of date and that the project is being restructured around real tasks, so quality varies by topic. ### What is an example of a prompt? The README does not give a worked prompt example. It lists prompt engineering as a covered topic, so concrete examples sit in the docs directory and on learnprompt.pro rather than in the repository's front page. ### What is the best prompt for learning? The repository does not rank prompts or recommend a single one. Its approach, per the README, is to organize learning around real tasks such as writing an article, making a slide deck or building a knowledge base, rather than around a canonical prompt. ## Sources - [Issues](https://github.com/LearnPrompt/LearnPrompt/issues) - [LearnPrompt/LearnPrompt on GitHub](https://github.com/LearnPrompt/LearnPrompt) - [Project website](https://www.learnprompt.pro/) - [README](https://github.com/LearnPrompt/LearnPrompt/blob/main/README.md) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/learnprompt-learnprompt