binary-husky/gpt_academic · error · RuntimeError

Json解析不合常规

Error message

Json解析不合常规

What it means

Catch-all inside the try block that processes each SSE chunk in predict_no_ui_long_connection: any exception while accumulating result/handling finish (key errors from unexpected chunk shapes, decode issues, the watchdog raise being caught too) triggers get_full_error, logger.error of the raw chunk, and RuntimeError('Json解析不合常规'). Like 188/189, the real detail is in the log; the message just says the JSON shape was unexpected.

Source

Thrown at request_llms/oai_std_model_template.py:263

                            print(f"[response] {result}")
                        break
                    result += response_text
                    if reasoning:
                        reasoning_buffer += reasoning_content
                    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] += response_text
                        # 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。")
                except Exception as e:
                    chunk = get_full_error(chunk, stream_response)
                    chunk_decoded = chunk.decode()
                    error_msg = chunk_decoded
                    logger.error(error_msg)
                    raise RuntimeError("Json解析不合常规")
        if reasoning:
            paragraphs = ''.join([f'<p style="margin: 1.25em 0;">{line}</p>' for line in reasoning_buffer.split('\n')])
            return f'''<div class="reasoning_process" >{paragraphs}</div>\n\n''' + result
        return result

    def predict(
        inputs,
        llm_kwargs,
        plugin_kwargs,
        chatbot,
        history=[],
        system_prompt="",
        stream=True,
        additional_fn=None,
    ):
        """
        发送至chatGPT，流式获取输出。
        用于基础的对话功能。

View on GitHub (pinned to d6bde0fa54)

Solutions

1. Inspect logger.error output — the offending raw chunk is logged verbatim.
2. Point API_URL_ORIGINAL at a strictly OpenAI-compatible endpoint or fix the relay's response format.
3. Update decode_chunk to tolerate the provider's actual schema (e.g. missing reasoning_content).
4. Retry after provider-side errors resolve (502s are often transient).

try:
    result = predict_no_ui_long_connection(...)
except RuntimeError as e:
    if 'Json解析不合常规' in str(e):
        chunk = get_last_logged_chunk()  # logger.error has the raw payload
        if is_provider_5xx(chunk):
            backoff_and_retry()
        else:
            fix_or_report_schema(chunk)

When it happens

Trigger: A chunk whose structure violates decode_chunk/accumulation assumptions: provider returns an error object where content deltas are expected, missing 'choices' field, non-JSON relay responses (HTML error pages), or schema changes in a custom OpenAI-compatible backend.

Common situations: Third-party relays with slightly non-standard SSE payloads; gateways returning HTML 502 pages mid-stream; provider API version changes; this catch also swallows the user-cancel watchdog raise from error 189 when cancellation coincides with chunk processing.

Related errors

• 意外Json结构：
• 无法读取以下数据，请检查配置。 {chunk_decoded}
• API异常,请检测终端输出。可能的原因是:{finish_reason}
• OpenAI拒绝了请求:
• OpenAI拒绝了请求：

AI-assisted analysis of binary-husky/gpt_academic@d6bde0fa54 (2026-08-14). Data as JSON: /api/errors/3eeccaf5e6047bfd. Report an issue: GitHub.