binary-husky/gpt_academic · error · RuntimeError

由于提问含不合规内容被Azure过滤。

Error message

由于提问含不合规内容被Azure过滤。

What it means

RuntimeError raised after the OpenRouter stream completes when the final chunk's finish_reason is 'content_filter': the upstream (typically Azure OpenAI content safety) blocked the response. No partial result is returned; the exception is the only signal. Despite living in bridge_openrouter, the message text is Azure-specific because the bridge reuses Azure handling logic.

Solutions

  1. Rephrase the sensitive part of the input and retry
  2. If you own the Azure deployment, relax the content filter categories/thresholds in Azure OpenAI Studio
  3. Switch to a non-Azure provider for that model via OpenRouter routing
  4. Catch RuntimeError and surface a friendly 'request blocked by content filter' message instead of a stack trace

Example fix

# before
resp = predict(inputs, llm_kwargs, ...)

# after
try:
    resp = predict(inputs, llm_kwargs, ...)
except RuntimeError as e:
    if '不合规内容' in str(e):
        chatbot[-1] = (chatbot[-1][0], '[Local Message] 请求被内容安全策略拦截，请改写后重试。')
        yield from update_ui(chatbot)
        return
Defensive patterns

Strategy: try-catch

Type guard

def finish_reason_is_filter(json_data) -> bool:
    return bool(json_data) and json_data.get('finish_reason') == 'content_filter'

Try / catch

try:
    result = predict(inputs, llm_kwargs, plugin_kwargs, chatbot, history)
except RuntimeError as e:
    if '不合规内容' in str(e):
        chatbot[-1] = (chatbot[-1][0], '[Local Message] 内容被安全过滤，请改写提问后重试。')
        yield from update_ui(chatbot)
        return
    raise

Prevention

When it happens

Trigger: Prompt or generated output trips Azure's content moderation filter (finish_reason='content_filter'); sending content in a regulated category for the deployment's filter policy; routing via OpenRouter to an Azure-backed provider slot.

Common situations: Academic/medical/violence-adjacent text triggering false positives; strict default Azure content filters on a new deployment; prompt-injection-like payloads being filtered.

Related errors

AI-assisted analysis of binary-husky/gpt_academic@d6bde0fa54 (2026-08-14). Data as JSON: /api/errors/89c8ad00aaaa1130. Report an issue: GitHub.

Appendix: source

Thrown at request_llms/bridge_openrouter.py:202

        json_data = chunkjson['choices'][0]
        delta = json_data["delta"]
        if len(delta) == 0: break
        if (not has_content) and has_role: continue
        if (not has_content) and (not has_role): continue # raise RuntimeError("发现不标准的第三方接口："+delta)
        if has_content: # has_role = True/False
            result += delta["content"]
            if not console_silence: print(delta["content"], end='')
            if observe_window is not None:
                # 观测窗，把已经获取的数据显示出去
                if len(observe_window) >= 1:
                    observe_window[0] += delta["content"]
                # 看门狗，如果超过期限没有喂狗，则终止
                if len(observe_window) >= 2:
                    if (time.time()-observe_window[1]) > watch_dog_patience:
                        raise RuntimeError("用户取消了程序。")
        else: raise RuntimeError("意外Json结构："+delta)
    if json_data and json_data['finish_reason'] == 'content_filter':
        raise RuntimeError("由于提问含不合规内容被Azure过滤。")
    if json_data and json_data['finish_reason'] == 'length':
        raise ConnectionAbortedError("正常结束，但显示Token不足，导致输出不完整，请削减单次输入的文本量。")
    return result


def predict(inputs:str, llm_kwargs:dict, plugin_kwargs:dict, chatbot:ChatBotWithCookies,
            history:list=[], system_prompt:str='', stream:bool=True, additional_fn:str=None):
    """
    发送至chatGPT，流式获取输出。
    用于基础的对话功能。
    inputs 是本次问询的输入
    top_p, temperature是chatGPT的内部调优参数
    history 是之前的对话列表（注意无论是inputs还是history，内容太长了都会触发token数量溢出的错误）
    chatbot 为WebUI中显示的对话列表，修改它，然后yield出去，可以直接修改对话界面内容
    additional_fn代表点击的哪个按钮，按钮见functional.py
    """
    from request_llms.bridge_all import model_info
    if is_any_api_key(inputs):

View on GitHub (pinned to d6bde0fa54)