LLM Docs

LLM Docs are plain-text exports of your approved knowledge for models: an index, the complete text in one file, and each article as raw Markdown. Every Context documentation site includes them automatically under LLM Docs in the top bar. If you need the text without a website — to paste into a model's context or import into a retrieval system — declare a standalone LLM text package.

flowchart LR
  approved["Approved articles<br/>and knowledge map"] --> site["Website build"]
  approved --> llms["llmsPackage (optional)"]
  site --> sitellms["llms/ inside the site"]
  llms --> pkg["dist/name/ text bundle"]

What you get

File	Content
llms.txt	Index that preserves the knowledge map's grouping, order and repeated placements
llms-full.txt	All selected approved text, each article once, in first-placement order
llms/pages/<identity>.txt	One raw Markdown file per approved article
llms/index.html	Landing page on the website: index and full-text links followed by the knowledge map

On a website these files sit inside the site folder and use the configured base; no extra declaration or command is needed. They stay in the primary language even when the site has translations. Text files carry a UTF-8 byte-order mark so browsers detect the encoding when a host omits the charset; serve them as text/plain; charset=utf-8.

A standalone text package

Choose an LLM text package when you want one consolidated file that is easy to copy, index or upload, without Agent Skill packaging:

llmsPackage({ name: "project-llms", template: "src/package-templates/llms" });

It produces llms.txt from its template, plus llms-full.txt and the per-article files, organized by the same knowledge map. It can sit alongside a knowledge package in the same declaration; ask the Agent to add it in the work-start report or later.

Freshness

Only approved, selected content is exported. Any change to the map or an article invalidates both the website and its LLM Docs, and a failed build keeps the previous output. For packages built for Agents with a query Skill, see Knowledge Packages.

LLM Docs

LLM Docs 是把已批准的知识导出为供模型使用的纯文本：一个索引、一个包含全部文本的文件，以及每篇文章的原始 Markdown。每个 Context 文档站点都会自动在顶栏的 LLM Docs 下提供它们。如果你需要文本而不需要网站——例如粘贴到模型上下文中，或导入检索系统——请声明一个独立的 LLM 文本包。

flowchart LR
  approved["已批准的文章<br/>和知识地图"] --> site["网站构建"]
  approved --> llms["llmsPackage（可选）"]
  site --> sitellms["站点内的 llms/"]
  llms --> pkg["dist/name/ 文本包"]

你会得到什么

文件	内容
llms.txt	索引，保留知识地图的分组、顺序和重复出现的位置
llms-full.txt	所有选中的已批准文本，每篇文章只出现一次，按首次出现的位置排序
llms/pages/<identity>.txt	每篇已批准文章对应一个原始 Markdown 文件
llms/index.html	网站上的入口页：先是索引和全文链接，后面是知识地图

在网站中，这些文件位于站点文件夹内，并使用配置的 base；无需额外的声明或命令。即使站点有翻译，它们也保持主语言。文本文件带有 UTF-8 字节顺序标记（BOM），这样即使托管服务省略了字符集，浏览器也能识别编码；请以 text/plain; charset=utf-8 提供它们。

独立的文本包

如果你想要一个便于复制、索引或上传的合并文件，而不需要 Agent Skill 打包，请选择 LLM 文本包：

llmsPackage({ name: "project-llms", template: "src/package-templates/llms" });

它会根据模板生成 llms.txt，另外生成 llms-full.txt 和逐篇文章的文件，并按同一份知识地图组织。它可以与知识包放在同一份声明中；你可以在开工报告中或之后让 Agent 添加它。

时效性

只有已批准且被选中的内容才会被导出。知识地图或任何文章发生变化，都会使网站及其 LLM Docs 失效，而构建失败时会保留之前的产出。关于为 Agent 构建、带查询 Skill 的知识包，请参阅知识包。