Update app.py

app.py

@@ -53,8 +53,7 @@ def whisper_speech_to_text(audio):
         return None
     finally:
         # Clear CUDA cache
-        #torch.cuda.empty_cache()
-        print("1")
+        torch.cuda.empty_cache()
 
 async def chat_with_ai(message, history):
     global chat_history
@@ -109,21 +108,23 @@ with gr.Blocks() as demo:
     audio_button = gr.Button("Send Audio")
     text_button = gr.Button("Send Text")
 
-    # JavaScript for handling space bar events for recording and playing output
-    demo.load(_js="""
-        document.addEventListener('keydown', function(event) {
-            if (event.code === 'Space') {
-                document.getElementById('audio_input').click();
-            }
-        });
-        
-        document.addEventListener('gradioAudioLoaded', function(event) {
-            var audioElement = document.getElementById('audio_output');
-            if (audioElement) {
-                audioElement.play();
-            }
-        });
-    """)
+    # Add custom JavaScript to handle spacebar press and play audio automatically
+    demo.append(gr.HTML("""
+        <script>
+            document.addEventListener('keydown', function(event) {
+                if (event.code === 'Space') {
+                    document.querySelector('input[type="file"]').click();
+                }
+            });
+            
+            document.addEventListener('gradioAudioLoaded', function(event) {
+                var audioElement = document.querySelector('audio');
+                if (audioElement) {
+                    audioElement.play();
+                }
+            });
+        </script>
+    """))
 
     audio_button.click(transcribe_and_chat, inputs=audio_input, outputs=[chat_output, audio_output])
     text_button.click(lambda x: asyncio.run(chat_with_ai(x, [])), inputs=text_input, outputs=[chat_output, audio_output])