VitePress uses file-based routing, which means the generated HTML pages are mapped from the directory structure of the source Markdown files. For example, given the following directory structure:
Project root is where VitePress will try to look for the `.vitepress` special directory. The `.vitepress` directory is a reserved location for VitePress' config file, dev server cache, build output, and optional theme customization code.
When you run `vitepress dev` or `vitepress build` from the command line, VitePress will use the current working directory as project root. To specify a sub-directory as root, you will need to pass the relative path to the command. For example, if your VitePress project is located in `./docs`, you should run `vitepress dev docs`:
Source directory is where your Markdown source files live. By default, it is the same as the project root. However, you can configure it via the [`srcDir`](../reference/site-config#srcdir) config option.
You can use both absolute and relative paths when linking between pages. Note that although both `.md` and `.html` extensions will work, the best practice is to omit file extensions so that VitePress can generate the final URLs based on your config.
Learn more about linking to assets such images in [Asset Handling](./asset-handling).
### Linking to Non-VitePress Pages
If you want to link to a page in your site that is not generated by VitePress, you'll either need to use the full URL (opens in a new tab) or explicitly specify the target:
**Input**
```md
[Link to pure.html](/pure.html){target="_self"}
```
**Output**
[Link to pure.html](/pure.html){target="_self"}
::: tip Note
In Markdown links, the `base` is automatically prepended to the URL. This means that if you want to link to a page outside of your base, you'd need something like `../../pure.html` in the link (resolved relative to the current page by the browser).
Alternatively, you can directly use the anchor tag syntax:
```md
<ahref="/pure.html"target="_self">Link to pure.html</a>
By default, VitePress resolves inbound links to URLs ending with `.html`. However, some users may prefer "Clean URLs" without the `.html` extension - for example, `example.com/path` instead of `example.com/path.html`.
Some servers or hosting platforms (for example Netlify, Vercel, GitHub Pages) provide the ability to map a URL like `/foo` to `/foo.html` if it exists, without a redirect:
You can customize the mapping between the source directory structure and the generated pages. It's useful when you have a complex project structure. For example, let's say you have a monorepo with multiple packages, and would like to place documentations along with the source files like this:
The `rewrites` option also supports dynamic route parameters. In the above example, it would be verbose to list all the paths if you have many packages. Given that they all have the same file structure, you can simplify the config like this:
The rewrite paths are compiled using the `path-to-regexp` package - consult [its documentation](https://github.com/pillarjs/path-to-regexp#parameters) for more advanced syntax.
When rewrites are enabled, **relative links should be based on the rewritten paths**. For example, in order to create a relative link from `packages/pkg-a/src/pkg-a-code.md` to `packages/pkg-b/src/pkg-b-code.md`, you should use:
You can generate many pages using a single Markdown file and dynamic data. For example, you can create a `packages/[pkg].md` file that generates a corresponding page for every package in a project. Here, the `[pkg]` segment is a route **parameter** that differentiates each page from the others.
### Paths Loader File
Since VitePress is a static site generator, the possible page paths must be determined at build time. Therefore, a dynamic route page **must** be accompanied by a **paths loader file**. For `packages/[pkg].md`, we will need `packages/[pkg].paths.js` (`.ts` is also supported):
```
.
└─ packages
├─ [pkg].md # route template
└─ [pkg].paths.js # route paths loader
```
The paths loader should provide an object with a `paths` method as its default export. The `paths` method should return an array of objects with a `params` property. Each of these objects will generate a corresponding page.
Given the following `paths` array:
```js
// packages/[pkg].paths.js
export default {
paths() {
return [
{ params: { pkg: 'foo' }},
{ params: { pkg: 'bar' }}
]
}
}
```
The generated HTML pages will be:
```
.
└─ packages
├─ foo.html
└─ bar.html
```
### Multiple Params
A dynamic route can contain multiple params:
**File Structure**
```
.
└─ packages
├─ [pkg]-[version].md
└─ [pkg]-[version].paths.js
```
**Paths Loader**
```js
export default {
paths: () => [
{ params: { pkg: 'foo', version: '1.0.0' }},
{ params: { pkg: 'foo', version: '2.0.0' }},
{ params: { pkg: 'bar', version: '1.0.0' }},
{ params: { pkg: 'bar', version: '2.0.0' }}
]
}
```
**Output**
```
.
└─ packages
├─ foo-1.0.0.html
├─ foo-2.0.0.html
├─ bar-1.0.0.html
└─ bar-2.0.0.html
```
### Dynamically Generating Paths
The paths loader module is run in Node.js and only executed during build time. You can dynamically generate the paths array using any data, either local or remote.
You can use the params to pass additional data to each page. The Markdown route file can access the current page params in Vue expressions via the `$params` global property:
You can also access the current page's params via the [`useData`](../reference/runtime-api#usedata) runtime API. This is available in both Markdown files and Vue components:
Params passed to the page will be serialized in the client JavaScript payload, so you should avoid passing heavy data in params, for example raw Markdown or HTML content fetched from a remote CMS.
Instead, you can pass such content to each page using the `content` property on each path object: