Update app.py
Browse files
app.py
CHANGED
|
@@ -96,19 +96,24 @@ def chat_completions():
|
|
| 96 |
|
| 97 |
elif '"final":true' in event.data:
|
| 98 |
final_data = json.loads(event.data)
|
| 99 |
-
|
|
|
|
|
|
|
| 100 |
if finish_reason == 'length':
|
| 101 |
messages.append({"role": "assistant", "content": full_response})
|
| 102 |
messages.append({"role": "user", "content": "请继续你的输出,不要重复之前的内容"})
|
| 103 |
break # 跳出当前循环,继续下一次请求
|
| 104 |
else:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 105 |
yield f"data: {json.dumps(format_openai_response('', finish_reason))}\n\n"
|
| 106 |
yield "data: [DONE]\n\n"
|
| 107 |
return # 完全结束生成
|
| 108 |
-
|
| 109 |
-
# 如果因为 length 而跳出内部循环,则继续外部循环
|
| 110 |
|
| 111 |
-
#
|
| 112 |
yield f"data: {json.dumps(format_openai_response('', 'stop'))}\n\n"
|
| 113 |
yield "data: [DONE]\n\n"
|
| 114 |
|
|
|
|
| 96 |
|
| 97 |
elif '"final":true' in event.data:
|
| 98 |
final_data = json.loads(event.data)
|
| 99 |
+
response_message = final_data.get('responseMessage', {})
|
| 100 |
+
finish_reason = response_message.get('finish_reason', 'stop')
|
| 101 |
+
|
| 102 |
if finish_reason == 'length':
|
| 103 |
messages.append({"role": "assistant", "content": full_response})
|
| 104 |
messages.append({"role": "user", "content": "请继续你的输出,不要重复之前的内容"})
|
| 105 |
break # 跳出当前循环,继续下一次请求
|
| 106 |
else:
|
| 107 |
+
# 正常结束,发送最后的内容(如果有的话)
|
| 108 |
+
last_content = response_message.get('text', '')
|
| 109 |
+
if last_content and last_content != full_response:
|
| 110 |
+
yield f"data: {json.dumps(format_openai_response(last_content[len(full_response):]))}\n\n"
|
| 111 |
+
|
| 112 |
yield f"data: {json.dumps(format_openai_response('', finish_reason))}\n\n"
|
| 113 |
yield "data: [DONE]\n\n"
|
| 114 |
return # 完全结束生成
|
|
|
|
|
|
|
| 115 |
|
| 116 |
+
# 如果因为多次长度限制而最终结束,发送一个停止信号
|
| 117 |
yield f"data: {json.dumps(format_openai_response('', 'stop'))}\n\n"
|
| 118 |
yield "data: [DONE]\n\n"
|
| 119 |
|