Update app.py

app.py

@@ -1,61 +1,4 @@
-# ==============================================================================
-    # Optimized send_message handler with reduced UI updates
-    # ==============================================================================
-    def send_message(message, show_thinking, temperature, model_choice, max_tokens, show_raw):
-        global stop_generation
-        stop_generation.clear()
-        
-        if not message.strip():
-            return "", "", "⚡ 0.0 tok/s", gr.update(interactive=True), gr.update(interactive=False)
-        
-        # Disable send button, enable stop button (ONCE at start)
-        yield "", "", "⚡ Generating...", gr.update(interactive=False), gr.update(interactive=True)
-        
-        # Switch backend based on selection
-        backend = available_models[model_choice]
-        
-        # Create single-turn history
-        history = [{"role": "user", "content": message}]
-        
-        # Show user message immediately (no button updates)
-        yield "", render_history(history, show_thinking, show_raw), "⚡ Generating...", gr.update(), gr.update()
-        
-        # Generate prompt (single turn, no history)
-        prompt = f"User: {message}\nSam:   <think>"
-        
-        # Start assistant message
-        history.append({"role": "assistant", "content": "<think>"})
-        
-        # Stream response
-        last_tokens_per_sec = 0
-        was_stopped = False
-        
-        for chunk_data in generate_response_stream(prompt, temperature, backend, max_tokens):
-            if len(chunk_data) == 5:  # New format with stopped flag
-                new_chunk, in_thinking, tokens_per_sec, avg_tokens_per_sec, stopped = chunk_data
-                
-                if stopped:
-                    was_stopped = True
-                    print("   ✅ Generation stopped successfully")
-                    break
-                
-                if new_chunk:  # Only update if there's actual content
-                    history[-1]["content"] += new_chunk
-                
-                last_tokens_per_sec = avg_tokens_per_sec
-                
-                # Update UI WITHOUT button updates for speed (use gr.update() to skip)
-                speed_text = f"⚡ {tokens_per_sec:.1f} tok/s"
-                yield "", render_history(history, show_thinking, show_raw), speed_text, gr.update(), gr.update()
-        
-        # Final yield - enable send button, disable stop button (ONCE at end)
-        if was_stopped:
-            final_speed = f"🛑 Stopped at {last_tokens_per_sec:.1f} tok/s"
-        else:
-            final_speed = f"✅ {last_tokens_per_sec:.1f} tok/s (avg)"
-        
-        print(f"   📊 Final speed: {final_speed}")
-        yield "", render_history(history, show_thinking, show_raw), final_speed, gr.update(interactive=True), gr.update(interactive=False)import os
+import os
 os.environ['KERAS_BACKEND'] = 'tensorflow'
 os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"