Minor fix

app.py

@@ -52,10 +52,10 @@ def infer(message, history):
     combined_embeds = []
     
     with torch.no_grad():
-        if message['file']:
+        if message['files']:
             projected_image_embeds = None
             audio_text_embeds = None
-            for path in message['file']:
+            for path in message['files']:
 
                 if path.endswith(('.jpg', '.png', '.jpeg')):
                     image = clip_preprocess(Image.open(path)).unsqueeze(0).to(device)