Update app.py

app.py

@@ -19,7 +19,7 @@ MODELS = {model_name: genai.GenerativeModel(model_name=model_name) for model_nam
 async def respond(message, history, selected_model):
     model = MODELS.get(selected_model)
     if not model:
-        yield {"role": "assistant", "content": "Error: Selected model not available."}, ""
+        yield {"role": "assistant", "content": "Error: Selected model not available."}
         return
 
     try:
@@ -28,9 +28,9 @@ async def respond(message, history, selected_model):
         full_response = ""
         for chunk in response_stream:
             full_response += (chunk.text or "")
-            yield {"role": "assistant", "content": full_response}, ""  # Формат messages
+            yield {"role": "assistant", "content": full_response}
     except Exception as e:
-        yield {"role": "assistant", "content": f"Error during API call: {e}"}, ""
+        yield {"role": "assistant", "content": f"Error during API call: {e}"}
 
 async def respond_thinking(message, history, selected_model):
     if "thinking" not in selected_model:
@@ -56,19 +56,16 @@ async def respond_thinking(message, history, selected_model):
                 else:
                     model_response_text += (part.text or "")
 
-        yield {"role": "assistant", "content": model_response_text}, thinking_process_text
+        return {"role": "assistant", "content": model_response_text}, thinking_process_text
     except Exception as e:
-        yield {"role": "assistant", "content": f"Error during API call: {e}"}, f"Error during API call: {e}"
+        return {"role": "assistant", "content": f"Error during API call: {e}"}, f"Error during API call: {e}"
 
 async def process_message(message, history, model_name):
     if "thinking" in model_name:
-        generator = respond_thinking(message, history, model_name)
-        thinking_output_content = ""
-        async for response, thinking in generator:
-            yield response, thinking
-        #  Здесь можно было бы дополнительно обработать thinking_output_content, если нужно
+        response, thinking = await respond_thinking(message, history, model_name)
+        yield response, thinking
     else:
-        async for response, _ in respond(message, history, model_name):
+        async for response in respond(message, history, model_name):
             yield response, ""
 
 def clear_thinking():
@@ -98,14 +95,27 @@ with gr.Blocks() as demo:
     def change_chatbot(model_name):
         return gr.ChatInterface.update() # No need to change the processing function here
 
+    async def update_thinking_output(history):
+        if history:
+            last_turn = history[-1]
+            if last_turn[0] is not None and last_turn[1] is not None:
+                return last_turn[1]
+        return ""
+
     model_selection.change(
         change_chatbot,
         inputs=[model_selection],
         outputs=[chatbot],
     )
 
+    chatbot.change(
+        update_thinking_output,
+        inputs=[chatbot],
+        outputs=[thinking_output]
+    )
+
+    clear_button.click(lambda: None, None, chatbot, queue=False)
     clear_button.click(clear_thinking, outputs=[thinking_output], queue=False)
-    clear_button.click(lambda: None, None, chatbot, queue=False) # Clear chatbot history
 
 if __name__ == "__main__":
     demo.launch()