Update app.py

app.py

@@ -82,8 +82,13 @@ def send_audio(chunk, cid):
     if not cid or cid not in connections:
         return "Connecting..."
     sr, arr = chunk
+
+    # Reset transcript if it's been running long or restarted
+    if len(connections[cid].transcript) > 1000:
+        connections[cid].transcript = ""
+
     connections[cid].enqueue_audio_chunk(sr, arr)
-    return connections[cid].transcript
+    return connections[cid].transcript.strip()
 
 def clear_transcript(cid):
     if cid in connections:
@@ -126,9 +131,12 @@ def handle_chat(user_input, history, thread_id, image_url):
     except Exception as e:
         return f"❌ {e}", history, thread_id, image_url
 
-def send_transcript_to_assistant(transcript, history, thread_id, image_url):
+def send_transcript_to_assistant(transcript, history, thread_id, image_url, cid):
     if not transcript.strip():
         return gr.update(), history, thread_id, image_url
+    # Clear transcript after sending
+    if cid in connections:
+        connections[cid].transcript = ""
     return handle_chat(transcript, history, thread_id, image_url)
 
 def clear_chat_and_transcript(client_id):
@@ -142,17 +150,18 @@ with gr.Blocks(theme=gr.themes.Soft()) as app:
 
     gr.HTML("""
     <style>
-    #ask-btn, #clear-chat-btn {
-        font-size: 16px !important;
-        padding: 10px 24px !important;
-        margin-top: 6px;
-    }
-    #record-audio button {
+    #ask-btn, #clear-chat-btn, #record-audio button {
         font-size: 16px !important;
-        padding: 12px 24px !important;
-        background-color: #f2f2f2 !important;
+        padding: 12px 28px !important;
         border-radius: 6px;
-        margin-top: 6px;
+        margin-top: 10px;
+        background-color: #f2f2f2 !important;
+    }
+    button {
+        margin-right: 8px;
+    }
+    #record-audio button svg {
+        display: none !important;
     }
     </style>
     """)
@@ -185,13 +194,17 @@ with gr.Blocks(theme=gr.themes.Soft()) as app:
                    inputs=[user_prompt, chat_state, thread_state, image_state],
                    outputs=[user_prompt, chat, thread_state, image_state])
     image_state.change(fn=lambda x: x, inputs=image_state, outputs=image_display)
+
     voice_input.stream(fn=send_audio, inputs=[voice_input, client_id], outputs=voice_transcript, stream_every=0.5)
+
     ask_btn.click(fn=send_transcript_to_assistant,
-                  inputs=[voice_transcript, chat_state, thread_state, image_state],
+                  inputs=[voice_transcript, chat_state, thread_state, image_state, client_id],
                   outputs=[user_prompt, chat, thread_state, image_state])
+
     clear_chat_btn.click(fn=clear_chat_and_transcript,
                          inputs=[client_id],
                          outputs=[chat, voice_transcript, thread_state, image_state])
+
     app.load(fn=create_ws, outputs=[client_id])
 
 app.launch()