try

app.py

@@ -97,9 +97,9 @@ class AbliterationProcessor:
             )
             
             device_info = f" on {device.upper()}" if device == "cuda" else ""
-            return f"✅ Model {model_id} loaded successfully{device_info}!"
+            return f"✅ Model {model_id} loaded successfully{device_info}!", model_id
         except Exception as e:
-            return f"❌ Model loading failed: {str(e)}"
+            return f"❌ Model loading failed: {str(e)}", "No model loaded"
     
     def process_abliteration(self, model_id, harmful_text, harmless_text, instructions, 
                            scale_factor, skip_begin, skip_end, layer_fraction, 
@@ -346,8 +346,14 @@ class AbliterationProcessor:
             generated_text = self.tokenizer.decode(gen[0][toks.shape[1]:], skip_special_tokens=True)
             print(f"DEBUG: Generated text length: {len(generated_text)}")
             print(f"DEBUG: Generated text preview: {generated_text[:200]}...")
+            print(f"DEBUG: Full generated text: {generated_text}")
             
-            return generated_text, history + [[message, generated_text]]
+            # Clean the text - remove any potential formatting issues
+            cleaned_text = generated_text.strip()
+            print(f"DEBUG: Cleaned text length: {len(cleaned_text)}")
+            print(f"DEBUG: Cleaned text: {cleaned_text}")
+            
+            return cleaned_text, history + [[message, cleaned_text]]
             
         except Exception as e:
             print(f"DEBUG: Exception occurred: {str(e)}")
@@ -439,6 +445,11 @@ def create_interface():
                         model_id.render()
                         load_model_btn = gr.Button("📥 Load Model", variant="primary")
                         load_status = gr.Textbox(label="Load Status", interactive=False)
+                        current_model_display = gr.Textbox(
+                            label="Currently Loaded Model", 
+                            interactive=False,
+                            value="No model loaded"
+                        )
                         
                         gr.Markdown("### ⚙️ Processing Parameters")
                         instructions = gr.Number(
@@ -506,6 +517,7 @@ def create_interface():
             with gr.TabItem("💬 Chat Test"):
                 with gr.Row():
                     with gr.Column(scale=3):
+                        gr.Markdown("**Note**: You are chatting with the currently loaded model. If you've just completed processing, you're testing the modified model. To test the original model, reload it in the Model Processing tab.")
                         chatbot = gr.Chatbot(
                             label="Chat Window",
                             height=400,
@@ -552,7 +564,7 @@ def create_interface():
         load_model_btn.click(
             processor.load_model,
             inputs=[model_id],
-            outputs=[load_status]
+            outputs=[load_status, current_model_display]
         )
         
         process_btn.click(
@@ -573,7 +585,11 @@ def create_interface():
             if history and history[-1]["role"] == "user":
                 # Get complete response first
                 response, _ = processor.chat(history[-1]["content"], history[:-1], max_new_tokens, temperature)
+                print(f"DEBUG: Bot function received response: {response[:200]}...")
+                print(f"DEBUG: Bot function full response: {response}")
+                print(f"DEBUG: Bot function history before append: {history}")
                 history.append({"role": "assistant", "content": response})
+                print(f"DEBUG: Bot function history after append: {history}")
             return history
         
         msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(