fix(prod): resolve middleware piping error and AI model ID

middleware.ts

@@ -51,12 +51,26 @@ export default async function middleware(request: NextRequest) {
     const hostHeader = headers.get('x-forwarded-host') || headers.get('host') || fallbackHost;
 
     urlObj.protocol = proto;
-
-    effectiveRequest = new NextRequest(urlObj, {
-      headers: request.headers,
-      method: request.method,
-      body: request.body,
-    });
+    // Do NOT reconstruct NextRequest for GET/HEAD requests to avoid stream locking/piping errors.
+    // next-intl middleware primarily uses the URL and headers, so a simplified request-like object
+    // or a redirected URL is often enough. For robustness, we only clone if we really have to.
+    if (method === 'GET' || method === 'HEAD') {
+      // In Next.js 15+, we can clone headers but avoid re-passing the body stream.
+      effectiveRequest = new NextRequest(urlObj, {
+        headers: new Headers(request.headers),
+        method: request.method,
+        // No body for GET/HEAD
+      });
+    } else {
+      // For POST/PUT etc, we must be careful. Re-passing the body can only be done once.
+      // If we don't need to fix the URL for the body-reading part, we can skip it.
+      // But usually intl-middleware doesn't care about the body.
+      effectiveRequest = new NextRequest(urlObj, {
+        headers: new Headers(request.headers),
+        method: request.method,
+        body: request.body,
+      });
+    }
 
     if (process.env.NODE_ENV !== 'production' || !process.env.CI) {
       console.log(