Update app.py

app.py

@@ -368,10 +368,18 @@ def count_message_tokens(messages, model="gpt-3.5-turbo-0301"):
 def stream_notdiamond_response(response, model):
     """流式处理 notdiamond API 响应。"""
     buffer = ""
+    previous_content = ""
+    
     for chunk in response.iter_content(1024):
         if chunk:
             buffer += chunk.decode('utf-8')
-            yield create_openai_chunk(buffer, model)
+            current_content = buffer.split('\n\n')[-1]  # 获取最新的内容块
+            
+            if current_content:
+                new_content = current_content[len(previous_content):]  # 提取新内容
+                previous_content = current_content  # 更新已处理的内容
+                
+                yield create_openai_chunk(new_content, model)
     
     yield create_openai_chunk('', model, 'stop')
 
@@ -412,10 +420,14 @@ def handle_non_stream_response(response, model, prompt_tokens):
 def generate_stream_response(response, model, prompt_tokens):
     """生成流式 HTTP 响应。"""
     total_completion_tokens = 0
+    previous_content = ""
     
     for chunk in stream_notdiamond_response(response, model):
         content = chunk['choices'][0]['delta'].get('content', '')
-        total_completion_tokens += count_tokens(content, model)
+        new_content = content[len(previous_content):]  # 提取新内容
+        previous_content = content  # 更新已处理的内容
+        
+        total_completion_tokens += count_tokens(new_content, model)
         
         chunk['usage'] = {
             "prompt_tokens": prompt_tokens,