@fullstackdatasolutions/articles 1.3.0 → 1.3.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@fullstackdatasolutions/articles",
-  "version": "1.3.0",
+  "version": "1.3.1",
   "private": false,
   "license": "MIT",
   "funding": {

package/src/__tests__/server-articles.test.ts

@@ -1017,6 +1017,27 @@ describe('AI markdown helpers', () => {
     expect(blocked.status).toBe(404)
   })
 
+  it('links the markdown twin back to its HTML article, not to itself', async () => {
+    setupArticleTreeMock([
+      {
+        slug: 'allowed',
+        frontmatter: 'date: 2025-01-01\naiCrawl: true\n',
+        body: '# Allowed',
+      },
+    ])
+
+    const response = await getArticleMarkdownResponse('allowed', {
+      siteUrl: 'https://example.com/',
+      siteName: 'Example',
+    })
+
+    expect(response.headers.get('Link')).toBe(
+      '<https://example.com/articles/allowed>; rel="canonical"'
+    )
+    expect(response.headers.get('Link')).not.toContain('.md')
+    expect(response.headers.get('Link')).not.toContain('alternate')
+  })
+
   it('generates robots rules for AI crawlers and blocked articles only', async () => {
     setupArticleTreeMock([
       { slug: 'allowed', frontmatter: 'date: 2025-01-01\naiCrawl: true\n' },

package/src/server-articles.ts

@@ -775,11 +775,20 @@ export async function getArticleMarkdownResponse(
     article && config.markdownTwinHeader !== false
       ? `${buildMarkdownTwinHeader(article, config, markdown)}${markdown.trimStart()}`
       : markdown
+  // `getArticleAiHeaders` builds a `rel="alternate"` link *to* the twin, which
+  // is correct on the HTML page and wrong here - on the twin's own response it
+  // pointed at the URL being requested, telling a client the alternate of
+  // `/articles/x.md` is `/articles/x.md`. `getArticleMarkdown` already returned
+  // null for anything not opted in, so that helper's other branch
+  // (`X-Robots-Tag: noai`) was unreachable from this call site anyway. Point
+  // back at the HTML article instead, which is the relationship a client
+  // fetching the twin actually needs: this is a representation of that page.
+  const canonicalUrl = `${config.siteUrl.replace(/\/$/, '')}/articles/${slug}`
   return new Response(body, {
     headers: {
       'Content-Type': 'text/markdown; charset=utf-8',
       'Cache-Control': 'public, max-age=3600, s-maxage=3600',
-      ...(article ? getArticleAiHeaders(article, config) : {}),
+      Link: `<${canonicalUrl}>; rel="canonical"`,
     },
   })
 }