Launch without server-side rendering

Gradio's launch check of the page through the Node proxy can time out while the model loads, which crashes the app at startup.

app.py

@@ -213,4 +213,7 @@ if __name__ == "__main__":
             process_fn=process_fn,
         )
 
-    demo.queue().launch(show_error=True, pwa=True)
+    # Server-side rendering puts a Node proxy in front of every request, and Gradio's launch
+    # check of the page through it can time out while the model loads, which crashes the app.
+    # HARP calls the API alone, so nothing is lost without it.
+    demo.queue().launch(show_error=True, pwa=True, ssr_mode=False)