fix(docs): drop a highlighter registration that can never fire

fumadocs-openapi calls `renderCodeBlock` with a hard-coded `"json"` from both of
its call sites (`request-tabs.js:76`, `response-tabs.js:48`), so the docs
`CodeBlock` it routes through never receives a shell language. The
`getHighlighter('js', { langs: [curlJsonBodyGrammar] })` registering the
shell-scoped JSON-body injection therefore did nothing but await on every API
sample render, and the docblock claiming the grammar covers those samples was
wrong.

- Delete the call and its imports.
- State the grammar's real coverage: prose fences only, via `langs`. Both API
  reference paths are unreachable — samples are JSON, and the cURL usage tabs
  highlight client-side off fumadocs' own factory.
- Correct `code-block.tsx`'s TSDoc, which still said API samples come from
  fumadocs' own renderer. They come through this component; `UsageTab` is the
  renderer that bypasses it.
- Re-home a comment orphaned when two CSS rules merged — it had drifted onto
  the rule below and read as documenting it.
- Drop a `.nd-copy-ignore` claim about transformers emitting those nodes;
  nothing here does, and upstream parity is the reason the clone exists.

Author: waleedlatif1

apps/docs/app/[lang]/[[...slug]]/page.tsx

@@ -1,5 +1,5 @@
 import type React from 'react'
-import { getHighlighter, highlight } from 'fumadocs-core/highlight'
+import { highlight } from 'fumadocs-core/highlight'
 import type { Root } from 'fumadocs-core/page-tree'
 import { findNeighbour } from 'fumadocs-core/page-tree'
 import type { ApiPageProps } from 'fumadocs-openapi/ui'
@@ -18,7 +18,6 @@ import { Heading } from '@/components/ui/heading'
 import { ResponseSection } from '@/components/ui/response-section'
 import { i18n } from '@/lib/i18n'
 import { getApiSpecContent, getAuthenticatedCodeSamples, openapi } from '@/lib/openapi'
-import { curlJsonBodyGrammar } from '@/lib/shiki-curl-json'
 import { simShikiOptions } from '@/lib/shiki-theme'
 import { type PageData, source } from '@/lib/source'
 import { DOCS_BASE_URL } from '@/lib/urls'
@@ -77,13 +76,8 @@ function stripLocalePrefix(url: string, lang: string): string {
  * rather than fumadocs-openapi's built-in one, so those blocks get the emcn copy control
  * instead of fumadocs' lucide clipboard. Mirrors the default renderer — same `highlight` call,
  * same `Pre` component, same `my-0` — differing only in which shell wraps the result.
- *
- * The `getHighlighter` call registers the JSON-body injection on the shared highlighter
- * `highlight` resolves. An injection is a property of the highlighter, not a per-call option, and
- * already-loaded grammars are skipped.
  */
 async function ApiCodeBlock({ lang, code }: { lang: string; code: string }) {
-  await getHighlighter('js', { langs: [curlJsonBodyGrammar] })
   return (
     <CodeBlock className='my-0'>
       {await highlight(code, { lang, ...simShikiOptions, components: { pre: Pre } })}

apps/docs/app/global.css

@@ -1734,6 +1734,8 @@ figure.shiki code:has(.line ~ .line) .line::before {
    the code surface. The viewport is the one box all three renderers agree on. */
 figure.shiki > div[role="region"] {
   background-color: var(--code-surface);
+  /* fumadocs ships 14px of vertical padding, the platform's viewer 8px; 10px splits them and
+     keeps a single-line fence from looking hollow at the tighter 21px line box. */
   padding-block: 10px;
 }
 
@@ -1743,9 +1745,6 @@ figure.shiki pre {
   background-color: transparent;
 }
 
-/* Viewport padding — fumadocs ships 14px vertical; the platform's viewer uses 8px. Split the
-   difference at 10px, which keeps a single-line fence from looking hollow at the tighter
-   21px line box. */
 /* Untitled blocks float the copy control over the code, so the last column has to clear it:
    an 8px offset plus emcn's 20px icon button, with room to breathe. fumadocs reserves 32px,
    which the glyphs run into.

apps/docs/components/ui/code-block.tsx

@@ -33,23 +33,24 @@ function CopyButton({ getCode }: { getCode: () => string }) {
 }
 
 /**
- * Docs code block for prose fences, wired into the MDX `pre` mapping.
+ * Docs code block for prose fences and the API reference's request/response samples — the MDX
+ * `pre` mapping and fumadocs-openapi's `renderCodeBlock` both render it.
  *
- * The shell — radius, hairline, fill — is not set here. Request and response samples in the
- * API reference come from fumadocs-openapi's own renderer, so the two share chrome through a
- * `figure.shiki` rule in `global.css` instead; see the note there. What stays here is the
- * part only this path has: the copy control, and the `my-4` prose rhythm that API samples,
- * which sit flush in their panel, must not inherit.
+ * The shell — radius, hairline, fill — is not set here. A third renderer, fumadocs-openapi's
+ * `UsageTab`, emits these figures without going through any component, so all three share
+ * chrome through a `figure.shiki` rule in `global.css` instead; see the note there. What stays
+ * here is the copy control, and the `my-4` prose rhythm that API samples, which sit flush in
+ * their panel, override with `my-0`.
  */
 export function CodeBlock({ title, ...props }: React.ComponentProps<typeof FumadocsCodeBlock>) {
   const figureRef = useRef<HTMLElement>(null)
 
   /**
    * Reads the block's text the way fumadocs' own `CopyButton` does: from a clone, with
-   * `.nd-copy-ignore` nodes replaced by newlines. Shiki transformers emit those to mark content
-   * that is displayed but is not part of the code, so reading `pre.textContent` directly would
-   * copy it. (The line-number gutter here is a `::before`, and pseudo-element content never
-   * reaches `textContent` — it is not what this guards.)
+   * `.nd-copy-ignore` nodes replaced by newlines — kept in step with upstream so a fence that
+   * gains such a node copies the same text there and here. (The line-number gutter is a
+   * `::before`, and pseudo-element content never reaches `textContent`, so it is not what this
+   * guards.)
    */
   function getCode() {
     const pre = figureRef.current?.getElementsByTagName('pre').item(0)

apps/docs/lib/shiki-curl-json.ts

@@ -19,13 +19,13 @@ import type { LanguageRegistration } from 'shiki'
  *   directly to Client Components" takes down every API reference page. A grammar is plain data,
  *   so it survives that boundary.
  *
- * Where it applies: prose fences, via `langs` on the MDX pipeline, and the API reference's
- * request/response samples, via the `getHighlighter` call in the docs page. NOT the API
- * reference's cURL usage tabs — those highlight in the browser off fumadocs' own factory, and
- * `ClientCodeBlockProvider` sits in a `"use client"` module the package does not expose through
- * its `exports` map, so reaching it means importing `fumadocs-openapi/ui/base` from client code
- * and dragging `remark` and `@fumari/json-schema-ts` into the browser bundle. That broke the
- * deployment once; it is not worth a second attempt.
+ * Applies to prose fences only, via `langs` on the MDX pipeline. Not the API reference:
+ * fumadocs-openapi calls `renderCodeBlock` with a hard-coded `"json"` for request and response
+ * samples, so a shell injection can never fire there, and its cURL usage tabs highlight in the
+ * browser off fumadocs' own factory — `ClientCodeBlockProvider` sits in a `"use client"` module
+ * the package does not expose through its `exports` map, so reaching it means importing
+ * `fumadocs-openapi/ui/base` from client code and dragging `remark` and
+ * `@fumari/json-schema-ts` into the browser bundle. That broke the deployment once.
  *
  * The opening brace requires a `}`, a quoted key, or end-of-line after it. That is what keeps
  * `awk '{print $1}'` out, while still matching a body whose brace ends the line — Oniguruma