update

app.py

@@ -27,11 +27,17 @@ def respond(
     max_tokens,
     temperature,
     top_p,
+    with_think,
 ):
     messages = [{"role": "system", "content": system_message}]
 
     messages.extend(history)
     
+    if with_think:
+        message = message + " /think"
+    else:
+        message = message + " /no_think"
+    
     messages.append({"role": "user", "content": message})
 
     headers = {
@@ -39,6 +45,9 @@ def respond(
         "Authorization": f"Bearer {API_TOKEN}"
     }
 
+    # print(f"[INFO] process user msg...")
+    # print(f"[INFO] userMsg: {message}")
+
     data = {
         "model": "/data/DMind-1-mini",
         "stream": True,
@@ -50,9 +59,6 @@ def respond(
         "max_tokens": 16384
     }
 
-    # print(f"[INFO] process user msg...")
-    # print(f"[INFO] userMsg: {message}")
-
     try:
         with requests.post(API_URL, headers=headers, json=data, stream=True) as r:
             if r.status_code == 200:
@@ -70,23 +76,30 @@ def respond(
                                         if content:
                                             current_response += content
                                             
-                                            if len(current_response) > 16:
-                                                if '<think>' in current_response:
-                                                    current_response = current_response.replace('<think>', '<details open><summary>Thinking</summary>\n\n```')
-                                                if '</think>' in current_response:
-                                                    current_response = current_response.replace('</think>', '```\n\n</details>')
-                                                if '**Final Answer**' in current_response:
-                                                    current_response = current_response.replace('**Final Answer**', '')
-                                                
-                                                formatted_response = current_response[:-16]
-                                                
-                                                formatted_response = formatted_response.replace('<', '&lt;').replace('>', '&gt;')
-                                                formatted_response = formatted_response.replace('&lt;details open&gt;', '<details open>')
-                                                formatted_response = formatted_response.replace('&lt;/details&gt;', '</details>')
-                                                formatted_response = formatted_response.replace('&lt;summary&gt;', '<summary>')
-                                                formatted_response = formatted_response.replace('&lt;/summary&gt;', '</summary>')
-                                                formatted_response = formatted_response.replace('*', '\\*')
-                                                yield formatted_response
+                                            if len(current_response) > 21:
+                                                if with_think:
+                                                    if '<think>' in current_response:
+                                                        current_response = current_response.replace('<think>', '<details open><summary>Thinking</summary>\n\n```')
+                                                    if '</think>' in current_response:
+                                                        current_response = current_response.replace('</think>', '```\n\n</details>')
+                                                    if '**Final Answer**' in current_response:
+                                                        current_response = current_response.replace('**Final Answer**', '')
+                                                    
+                                                    formatted_response = current_response[:-16]
+                                                    
+                                                    formatted_response = formatted_response.replace('<', '&lt;').replace('>', '&gt;')
+                                                    formatted_response = formatted_response.replace('&lt;details open&gt;', '<details open>')
+                                                    formatted_response = formatted_response.replace('&lt;/details&gt;', '</details>')
+                                                    formatted_response = formatted_response.replace('&lt;summary&gt;', '<summary>')
+                                                    formatted_response = formatted_response.replace('&lt;/summary&gt;', '</summary>')
+                                                    formatted_response = formatted_response.replace('*', '\\*')
+                                                    yield formatted_response
+                                                else:
+                                                    if '<think>' in current_response and '</think>\n' in current_response:
+                                                        start = current_response.find('<think>')
+                                                        end = current_response.find('</think>\n') + len('</think>\n')
+                                                        current_response = current_response[:start] + current_response[end:]
+                                                    yield current_response
                             except json.JSONDecodeError:
                                 continue
                 if current_response:
@@ -127,6 +140,7 @@ demo = gr.ChatInterface(
             step=0.05,
             label="Top-p (nucleus sampling)",
         ),
+        gr.Checkbox(value=False, label="With Think"),
     ],
     type="messages",
     css="""