binary-husky/gpt_academic · error · RuntimeError

程序终止。

Error message

程序终止。

What it means

Raised inside the Gemini streaming loop (predict_no_ui_long_connection) by a watchdog: if the time stored in observe_window[1] is more than watch_dog_patience seconds (5s) behind time.time(), the stream is considered stalled and the generator aborts with RuntimeError('程序终止。'). It is a client-side liveness check on the SSE stream, not a Gemini API error.

Source

Thrown at request_llms/bridge_google_gemini.py:42

    genai = GoogleChatInit(llm_kwargs)
    watch_dog_patience = 5  # 看门狗的耐心, 设置5秒即可
    gpt_replying_buffer = ''
    stream_response = genai.generate_chat(inputs, llm_kwargs, history, sys_prompt)
    for response in stream_response:
        results = response.decode()
        match = re.search(r'"text":\s*"((?:[^"\\]|\\.)*)"', results, flags=re.DOTALL)
        error_match = re.search(r'\"message\":\s*\"(.*?)\"', results, flags=re.DOTALL)
        if match:
            try:
                paraphrase = json.loads('{"text": "%s"}' % match.group(1))
            except:
                raise ValueError(f"解析GEMINI消息出错。")
            buffer = paraphrase['text']
            gpt_replying_buffer += buffer
            if len(observe_window) >= 1:
                observe_window[0] = gpt_replying_buffer
            if len(observe_window) >= 2:
                if (time.time() - observe_window[1]) > watch_dog_patience: raise RuntimeError("程序终止。")
        if error_match:
            raise RuntimeError(f'{gpt_replying_buffer} 对话错误')
    return gpt_replying_buffer

def make_media_input(inputs, image_paths):
    image_base64_array = []
    for image_path in image_paths:
        path = os.path.abspath(image_path)
        inputs = inputs + f'<br/><br/><div align="center"><img src="file={path}"></div>'
        base64 = encode_image(path)
        image_base64_array.append(base64)
    return inputs, image_base64_array

def predict(inputs:str, llm_kwargs:dict, plugin_kwargs:dict, chatbot:ChatBotWithCookies,
            history:list=[], system_prompt:str='', stream:bool=True, additional_fn:str=None):

    from .bridge_all import model_info

View on GitHub (pinned to d6bde0fa54)

Solutions

1. Check network/proxy reachability to the Gemini endpoint (curl a minimal streaming request) and retry once.
2. Raise watch_dog_patience (e.g. to 30s) or have the caller update observe_window[1] on each yield so only true stalls trip the dog.
3. If the stream legitimately produces no text for long periods (safety refusal), handle the empty-buffer path at the end of the loop instead of relying on the watchdog.
4. Upgrade/patch the bridge to use the official google-generativeai SDK instead of regex-scraping raw SSE bytes.

Example fix

// before
watch_dog_patience = 5
if (time.time() - observe_window[1]) > watch_dog_patience: raise RuntimeError("程序终止。")

# after
watch_dog_patience = 30
if (time.time() - observe_window[1]) > watch_dog_patience:
    raise RuntimeError("Gemini stream stalled for >30s; check proxy/network.")

import time
start = time.time()
# caller refreshes window[1] each iteration
window = ['', time.time()]
try:
    out = predict_no_ui_long_connection(..., observe_window=window)
except RuntimeError as e:
    if '程序终止' in str(e) and time.time()-start < 120:
        window[1] = time.time(); retry = True  # one retry with refreshed watchdog

try:
    result = predict_no_ui_long_connection(...)
except RuntimeError as e:
    if '程序终止' in str(e):
        # watchdog stall: refresh observe_window[1] and retry once
        ...

When it happens

Trigger: Calling predict_no_ui_long_connection with an observe_window list of length >= 2 where observe_window[1] was set to a start timestamp, and no successfully decoded '"text":' chunk arrives within 5 seconds (slow network, proxy stall, or the API stopped sending deltas).

Common situations: Flaky proxy/VPN to generativelanguage.googleapis.com, Gemini rate-limiting or hanging the connection, large prompt causing long silent thinking time, or the caller forgetting to refresh observe_window[1] between chunks.

Related errors

• {moonshot_bro_result} 程序终止。
• 程序终止。
• 用户取消了程序。
• 程序终止。
• 程序终止。

AI-assisted analysis of binary-husky/gpt_academic@d6bde0fa54 (2026-08-14). Data as JSON: /api/errors/4d72091315a10e47. Report an issue: GitHub.