SEO: multilingual homepage (homeSlug per locale in path/sitemap), Article + llms.txt generators, siteDescription fallback, composeTitle dedup

src/modules/seo/buildLlmsTxt.ts

@@ -0,0 +1,116 @@
+import type { BasePayload } from 'payload'
+
+import type { I18nConfig } from '../i18n/index.js'
+
+import { buildLocalizedPath } from '../i18n/index.js'
+
+type BuildLlmsTxtArgs = {
+  /** Absolute site URL (https://…). */
+  baseUrl: string
+  config: I18nConfig
+  /** Locale for names/descriptions. Defaults to config.defaultLocale. */
+  locale?: string
+  /** Pages collection slug. Defaults to 'pages'. */
+  pagesSlug?: string
+  payload: BasePayload
+  /** Site settings global slug. Defaults to 'site-settings'. */
+  settingsSlug?: string
+}
+
+type SettingsShape = {
+  siteDescription?: null | string
+  siteName?: null | string
+}
+
+type PageRow = {
+  _status?: string
+  id: number | string
+  meta?: { description?: unknown; noindex?: boolean } | null
+  slug?: unknown
+  title?: unknown
+}
+
+/**
+ * Builds the body of /llms.txt — a Markdown file describing the site for AI
+ * agents / LLM crawlers, per the llmstxt.org convention. Mirrors buildRobots /
+ * buildSitemapEntries: the plugin already knows the site's name, description,
+ * and pages, so it can generate this automatically.
+ *
+ * Structure (llmstxt.org): H1 site name, a blockquote/summary, then a list of
+ * key pages as Markdown links with short descriptions. Agents read this to
+ * understand the site quickly without crawling everything.
+ *
+ * Wire it as a route that returns text/plain:
+ *
+ *   // app/llms.txt/route.ts
+ *   import { buildLlmsTxt } from '@intecion/ipal-kit'
+ *   import { getCachedPayload } from '@/lib/content'
+ *   import { i18nConfig } from '@/i18n.config'
+ *   export const dynamic = 'force-dynamic'
+ *   export async function GET() {
+ *     const body = await buildLlmsTxt({
+ *       payload: await getCachedPayload(),
+ *       config: i18nConfig,
+ *       baseUrl: process.env.NEXT_PUBLIC_SERVER_URL!,
+ *     })
+ *     return new Response(body, { headers: { 'Content-Type': 'text/plain; charset=utf-8' } })
+ *   }
+ *
+ * Data comes from the panel (siteName, siteDescription, pages) — nothing
+ * hardcoded. Skips drafts and noindex pages (same as the sitemap).
+ */
+export async function buildLlmsTxt({
+  baseUrl,
+  config,
+  locale,
+  pagesSlug = 'pages',
+  payload,
+  settingsSlug = 'site-settings',
+}: BuildLlmsTxtArgs): Promise<string> {
+  const loc = locale ?? config.defaultLocale
+  const origin = baseUrl.replace(/\/$/, '')
+
+  const settings = (await payload.findGlobal({
+    slug: settingsSlug as never,
+    depth: 0,
+    locale: loc as never,
+  })) as SettingsShape
+
+  const name = settings.siteName?.trim() || 'Website'
+  const description = settings.siteDescription?.trim()
+
+  const result = await payload.find({
+    collection: pagesSlug as never,
+    depth: 0,
+    limit: 1000,
+    locale: loc as never,
+    where: { _status: { not_equals: 'draft' } } as never,
+  })
+
+  const lines: string[] = [`# ${name}`, '']
+  if (description) {
+    lines.push(`> ${description}`, '')
+  }
+
+  const pageLinks: string[] = []
+  for (const raw of result.docs as PageRow[]) {
+    if (raw._status && raw._status !== 'published') {continue}
+    if (raw.meta?.noindex) {continue}
+
+    const title = typeof raw.title === 'string' ? raw.title : undefined
+    const slug = typeof raw.slug === 'string' ? raw.slug : undefined
+    if (!title || !slug) {continue}
+
+    const path = buildLocalizedPath({ config, locale: loc, slugs: { [loc]: slug } })
+    if (!path) {continue}
+
+    const pageDesc = typeof raw.meta?.description === 'string' ? ` — ${raw.meta.description}` : ''
+    pageLinks.push(`- [${title}](${origin}${path})${pageDesc}`)
+  }
+
+  if (pageLinks.length > 0) {
+    lines.push('## Strony', '', ...pageLinks, '')
+  }
+
+  return lines.join('\n')
+}