Vite Plugin Tags
Use htmlTagsToHead() when your SSR integration already has Vite HTML tag descriptors. The helper does not collect descriptors.
Call prepareViteHead() once when you load a manifest. Call the returned function for each request. If the manifest changes, prepare the new descriptors.
import type { HtmlTagDescriptor } from 'unhead/vite'
import { createHead, prepareHeadInput } from 'unhead/server'
import { htmlTagsToHead } from 'unhead/vite'
export function prepareViteHead(tags: readonly HtmlTagDescriptor[]) {
const viteHead = prepareHeadInput(htmlTagsToHead(tags))
return () => createHead({ init: [viteHead] })
}
Keep the prepared input immutable so Unhead can reuse normalized tags across requests. prepareHeadInput() is experimental.
The helper accepts flat base, link, meta, noscript, script, and style descriptors. Use string children for inline content from trusted plugins.
Omitted injectTo and head-prepend give a tag high priority. head, body-prepend, and body map to the head, start of the body, and end of the body. Unhead applies its normal deduplication and CAPO ordering. For base, the first supplied href and target win in Vite's head placement order.
If conversion throws TypeError, send the original batch through your framework's HTML transform. Use that path for unsupported tags or placements, nested descriptors, and attributes that Unhead would reinterpret. Examples include key, class, style, and strings that look like HTML character references.
Conversion also throws when a batch contains module and nomodule scripts that Unhead would deduplicate.
Vite's manifest support is tracked in vitejs/ecosystem#15.