Add articles.json output for MCP corpus

Adds a new Hugo output format (Articles) that generates public/articles.json — a full-text JSON index of all non-draft post and page content. Includes fields expected by the MCP corpus builder (title, url, summary, content, date, type, tags, categories). Also adds a Node.js test suite to verify the output is correct after a build.

Author: potatoqualitee

config/_default/config.toml

@@ -47,5 +47,11 @@ baseName = "search"
 isPlainText = true
 notAlternative = true
 
+[outputFormats.Articles]
+mediaType = "application/json"
+baseName = "articles"
+isPlainText = true
+notAlternative = true
+
 [outputs]
-home = ["HTML", "RSS", "REDIRECTS", "SearchIndex"]
+home = ["HTML", "RSS", "REDIRECTS", "SearchIndex", "Articles"]

layouts/index.articles.json

@@ -0,0 +1,20 @@
+{{- $.Scratch.Add "articles" slice -}}
+{{- range where .Site.RegularPages "Type" "in" (slice "post" "page") -}}
+  {{- if not .Params.draft -}}
+    {{- $date := "" -}}
+    {{- if not .Date.IsZero -}}
+      {{- $date = .Date.Format "2006-01-02" -}}
+    {{- end -}}
+    {{- $.Scratch.Add "articles" (dict
+      "title" .Title
+      "url" .Permalink
+      "summary" (.Summary | plainify)
+      "content" .Plain
+      "date" $date
+      "type" .Type
+      "tags" (default (slice) .Params.tags)
+      "categories" (default (slice) .Params.categories)
+    ) -}}
+  {{- end -}}
+{{- end -}}
+{{- $.Scratch.Get "articles" | jsonify -}}

tests/articles-json.test.js

@@ -0,0 +1,72 @@
+const test = require('node:test');
+const assert = require('node:assert/strict');
+const fs = require('node:fs');
+const path = require('node:path');
+
+const root = path.resolve(__dirname, '..');
+const articlesPath = path.join(root, 'public', 'articles.json');
+
+function loadArticles() {
+  if (!fs.existsSync(articlesPath)) {
+    throw new Error('public/articles.json missing - run `npm run build` first');
+  }
+  return JSON.parse(fs.readFileSync(articlesPath, 'utf8'));
+}
+
+function draftSlugs() {
+  const postDir = path.join(root, 'content', 'post');
+  return fs
+    .readdirSync(postDir)
+    .filter((f) => f.endsWith('.md'))
+    .filter((f) => {
+      const body = fs.readFileSync(path.join(postDir, f), 'utf8');
+      const frontMatter = body.split('---')[1] || '';
+      return /^draft:\s*true\s*$/m.test(frontMatter);
+    })
+    .map((f) => f.replace(/\.md$/, ''));
+}
+
+test('articles.json is a non-empty array', () => {
+  const articles = loadArticles();
+  assert.ok(Array.isArray(articles), 'expected an array');
+  assert.ok(articles.length > 50, `expected more than 50 articles, got ${articles.length}`);
+});
+
+test('articles.json exposes the fields the MCP corpus builder expects', () => {
+  const articles = loadArticles();
+  for (const key of ['title', 'url', 'summary', 'content', 'date', 'type', 'tags', 'categories']) {
+    assert.ok(key in articles[0], `missing field: ${key}`);
+  }
+});
+
+test('articles.json contains no drafts', () => {
+  const articles = loadArticles();
+  const drafts = draftSlugs();
+
+  // The fixture must be meaningful: if nothing is a draft, this proves nothing.
+  assert.ok(drafts.length > 0, 'expected at least one draft post to exist');
+
+  for (const slug of drafts) {
+    const hit = articles.find((a) => a.url.includes(`/${slug}`));
+    assert.equal(hit, undefined, `draft leaked into articles.json: ${slug}`);
+  }
+});
+
+test('articles.json excludes command pages', () => {
+  const articles = loadArticles();
+  assert.ok(!articles.some((a) => a.type === 'commands'), 'command pages must not be indexed here');
+});
+
+test('articles.json does not truncate content the way search.json does', () => {
+  const articles = loadArticles();
+  const longest = articles.reduce((m, a) => Math.max(m, a.content.length), 0);
+  assert.ok(longest > 500, `expected full text, longest content was ${longest} chars`);
+});
+
+test('articles.json uses absolute URLs', () => {
+  const articles = loadArticles();
+  assert.ok(
+    articles.every((a) => a.url.startsWith('https://dbatools.io/')),
+    'every url must be absolute'
+  );
+});