Media Engine罢工有办法让流程先走完吗

[ditto66]

错误信息

研究过程中发生错误: Error code: 400 - {'error': {'code': '402', 'message': '很抱歉，检测到有敏感词，互联网非法外之地，请注意言辞~', 'param': '', 'type': 'param_error'}}

错误详情

Traceback (most recent call last):
  File "/Users/she/Project/BettaFish/SingleEngineApp/media_engine_streamlit_app.py", line 182, in execute_research
    agent._initial_search_and_summary(i)
  File "/Users/she/Project/BettaFish/SingleEngineApp/../MediaEngine/agent.py", line 221, in _initial_search_and_summary
    search_output = self.first_search_node.run(search_input)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/she/Project/BettaFish/SingleEngineApp/../MediaEngine/nodes/search_node.py", line 79, in run
    raise e
  File "/Users/she/Project/BettaFish/SingleEngineApp/../MediaEngine/nodes/search_node.py", line 69, in run
    response = self.llm_client.stream_invoke_to_string(SYSTEM_PROMPT_FIRST_SEARCH, message)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/she/Project/BettaFish/utils/retry_helper.py", line 89, in wrapper
    raise e
  File "/Users/she/Project/BettaFish/utils/retry_helper.py", line 77, in wrapper
    result = func(*args, **kwargs)
             ^^^^^^^^^^^^^^^^^^^^^
  File "/Users/she/Project/BettaFish/SingleEngineApp/../MediaEngine/llms/base.py", line 153, in stream_invoke_to_string
    for chunk in self.stream_invoke(system_prompt, user_prompt, **kwargs):
  File "/Users/she/Project/BettaFish/SingleEngineApp/../MediaEngine/llms/base.py", line 136, in stream_invoke
    raise e
  File "/Users/she/Project/BettaFish/SingleEngineApp/../MediaEngine/llms/base.py", line 122, in stream_invoke
    stream = self.client.chat.completions.create(
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/she/Project/BettaFish/.venv/lib/python3.11/site-packages/openai/_utils/_utils.py", line 286, in wrapper
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/Users/she/Project/BettaFish/.venv/lib/python3.11/site-packages/openai/resources/chat/completions/completions.py", line 1211, in create
    return self._post(
           ^^^^^^^^^^^
  File "/Users/she/Project/BettaFish/.venv/lib/python3.11/site-packages/openai/_base_client.py", line 1297, in post
    return cast(ResponseT, self.request(cast_to, opts, stream=stream, stream_cls=stream_cls))
                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/she/Project/BettaFish/.venv/lib/python3.11/site-packages/openai/_base_client.py", line 1070, in request
    raise self._make_status_error_from_response(err.response) from None
openai.BadRequestError: Error code: 400 - {'error': {'code': '402', 'message': '很抱歉，检测到有敏感词，互联网非法外之地，请注意言辞~', 'param': '', 'type': 'param_error'}}

环境信息

• 应用: Media Engine Streamlit App
• 时间: 2026-03-11 01:26:49

[dosubot]

Hi @ditto66! I'm Dosu and I’m helping the BettaFish team.

目前 MediaEngine 遇到敏感词检测错误（402）会直接中断流程，系统没有内置的跳过机制。以下是几个解决方案：

最快的临时方案：切换 LLM 服务商

修改 .env 文件切换到内容审核更宽松的模型（如 Qwen）：

MEDIA_ENGINE_API_KEY=your_api_key
MEDIA_ENGINE_BASE_URL=https://api.example.com/v1
MEDIA_ENGINE_MODEL_NAME=qwen-max

代码层面实现容错

如果需要让流程继续执行，可以修改 MediaEngine/agent.py 中的 research 方法，在段落处理处添加 try-catch：

for i, paragraph_goal in enumerate(self.state.outline.sections):
    try:
        self._process_paragraph(paragraph_goal, i)
    except Exception as e:
        logger.error(f"段落 {i + 1} 处理失败: {e}, 跳过并继续")
        self.state.paragraphs[i].latest_summary = "该段落因内容审核失败而跳过"
        continue

根本原因

当前的重试机制基于异常类型而非 HTTP 状态码，无法区分敏感词错误和网络错误，所有错误都会重试 6 次后抛出异常。社区曾有类似 issue 讨论这个问题，建议检查输入内容中的敏感词或切换服务商。

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}