fix(studio): move AI generation behind short-lived job requests

scripts/verify-deployment.ts

@@ -36,6 +36,13 @@ try {
   ]) {
     const response = await fetch(origin + path);
     if (!response.ok) throw Error(`${path} returned ${response.status}`);
+    if (
+      path === "/design-studio/studio.js" &&
+      !(await response.clone().text()).includes("respond-async")
+    )
+      throw Error(
+        "The deployed Studio client is missing background generation polling.",
+      );
     if (path === "/design-studio/") {
       const html = await response.clone().text();
       const presetMenu =
@@ -97,6 +104,7 @@ try {
     process.env.APIMART_API_KEY?.trim() &&
     (generation.provider !== "APIMart" ||
       generation.model !== "gpt-6-luna" ||
+      generation.delivery !== "polling" ||
       generation.maxAttempts !== 2 ||
       generation.maxOutputTokens !== 8192 ||
       generation.timeoutMs !== 60_000)
@@ -104,6 +112,21 @@ try {
     throw Error(
       "The compiled app did not pick up its server-side AI configuration.",
     );
+  if (process.env.APIMART_API_KEY?.trim()) {
+    const missingJob = await fetch(
+      origin +
+        "/design-studio/api/generate/jobs/00000000-0000-4000-8000-000000000000",
+    );
+    if (
+      missingJob.status !== 410 ||
+      !(missingJob.headers.get("content-type") ?? "").includes(
+        "application/json",
+      )
+    )
+      throw Error(
+        "The compiled app did not serve the background generation job endpoint.",
+      );
+  }
   if (application.enabled !== false)
     throw Error("Hosted Studio must remain export-only.");
   const preview = await fetch(origin + "/design-studio/api/preview", {