Update app.py

app.py

@@ -2,6 +2,7 @@ import gradio as gr
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 from threading import Thread
+import re
 
 # 1. Load weights explicitly
 MODEL_ID = "Axiom-AI/Quasar-1-0.8B"
@@ -46,7 +47,7 @@ def generate_response(message, history, system_prompt, temperature, top_p, top_k
     # Setup standard base template structures
     prompt = f"System: {system_prompt}\n" if system_prompt else ""
     
-    # Pack up chat context from Gradio 6's dictionary format
+    # Pack up chat context from the dictionary format
     for turn in history:
         role = turn.get("role")
         content = turn.get("content")
@@ -54,12 +55,9 @@ def generate_response(message, history, system_prompt, temperature, top_p, top_k
             prompt += f"User: {content}\n"
         elif role == "assistant":
             # Strip out our custom HTML formatting back to raw tokens for context history
-            # (Simple fallback: if the history already has html, you might want to keep a raw text history, 
-            # but for this generation block we clean basic tags if present)
             clean_content = content
             if "🤔" in content:
                 # Reconstruct rough raw tags for the model's history context
-                import re
                 clean_content = re.sub(r'<details.*?>.*?<\/summary>', '<think>', clean_content)
                 clean_content = clean_content.replace('</div></details>', '</think>')
                 clean_content = re.sub(r'<div.*?>', '', clean_content)
@@ -87,7 +85,6 @@ def generate_response(message, history, system_prompt, temperature, top_p, top_k
     partial_text = "<think>"
     for new_text in streamer:
         partial_text += new_text
-        # Yield the beautifully formatted text instead of raw markdown tags
         yield format_reasoning(partial_text)
 
 # 3. Clean Interface Assembly 
@@ -97,8 +94,8 @@ with gr.Blocks() as demo:
     with gr.Row():
         # Left Workspace Column: Chat Layout
         with gr.Column(scale=3):
-            # Line wrapping and markdown parsing enabled by default
-            chatbot = gr.Chatbot(height=500, type="messages")
+            # REMOVED type="messages" to solve the init crash
+            chatbot = gr.Chatbot(height=500)
             msg_input = gr.Textbox(placeholder="Enter reasoning prompt...", label="Your Message")
             with gr.Row():
                 submit_btn = gr.Button("Send", variant="primary")
@@ -117,7 +114,7 @@ with gr.Blocks() as demo:
             top_k = gr.Slider(minimum=1, maximum=100, value=40, step=1, label="Top-K")
             max_tokens = gr.Slider(minimum=64, maximum=2048, value=1024, step=64, label="Max New Tokens")
 
-    # Gradio 6 explicit dictionary append functions
+    # Gradio dictionary append functions
     def user_side_submit(user_message, history):
         history.append({"role": "user", "content": user_message})
         return "", history