Merge pull request #5744 from FlowFuse/fix/robots-txt-prerender

Fix robots.txt 502 and the crashing Nitro fallback function

Author: Yndira-E

nuxt/nuxt.config.ts

@@ -342,6 +342,20 @@ export default defineNuxtConfig({
 
     nitro: {
         preset: 'netlify',
+        // sanitize-html is CommonJS but depends on htmlparser2 ^12, which is ESM-only
+        // ("type": "module", main ./dist/index.js). Left external, Nitro require()s it from
+        // /var/task/node_modules in the deployed function, and require() of an ES module only
+        // works on Node >= 22.12 - which is why sanitize-html declares engines node >=22.12.0.
+        // The Lambda runtime under this site is older than that, so the require threw
+        // ERR_REQUIRE_ESM at module load. server/utils is auto-imported by Nitro, so that one
+        // throw took down every route the fallback function serves: robots.txt, /api/*, and
+        // every 404, all returning 502. Inlining it makes rollup resolve htmlparser2 at build
+        // time, so the function no longer require()s ESM and this holds on any runtime.
+        // The runtime version itself is a Netlify site setting (AWS_LAMBDA_JS_RUNTIME); it
+        // cannot be set from netlify.toml, so it is not fixable here.
+        externals: {
+            inline: ['sanitize-html']
+        },
         // Nitro emits a .mjs.map next to every server chunk, so the Netlify functions bundle
         // ships one map file per chunk. Skipping them keeps the bundle smaller and the
         // server build a little shorter. The cost is that a server-side stack trace in the
@@ -395,6 +409,13 @@ export default defineNuxtConfig({
                     // content-urls.get.ts) silently resolves to undefined. Explicitly listing
                     // it here bakes it at build time instead, inside the git checkout.
                     '/sitemap.xml',
+                    // Same hybrid-preset gate as /sitemap.xml above: @nuxtjs/robots only
+                    // writes robots.txt to nuxt/dist when isNuxtGenerate() is true, so on the
+                    // netlify preset it was served live by the fallback function instead. Any
+                    // fault in that function therefore turned robots.txt into a 5xx, and
+                    // Google reads a 5xx on robots.txt as "crawl nothing" for the first ~12h.
+                    // Listing it here makes it a static file, independent of the function.
+                    '/robots.txt',
                     '/contact-us',
                     '/book-demo',
                     '/support',