What causes "无法下载论文 {self.doi}，所有重试都失败了" in binary-husky/gpt_academic?

Read the preceding log lines ('第 N 次尝试失败: ...') to identify which stage fails — mirror, extraction, or PDF fetch.

binary-husky/gpt_academic · error · Exception

无法下载论文 {self.doi}，所有重试都失败了

Error message

无法下载论文 {self.doi}，所有重试都失败了

What it means

Top-level retry exhaustion in the SciHub downloader: the download is attempted 3 times (with 3s/6s backoff between), and if every attempt raises, it fails with Exception('无法下载论文 {doi}，所有重试都失败了'). The per-attempt errors are only in logs; the raised message does not include the last cause.

Source

Thrown at crazy_functions/review_fns/data_sources/scihub_source.py:296

                    continue

                # 保存PDF文件
                pdf_name = f"{self.doi.replace('/', '_').replace(':', '_')}.pdf"
                pdf_path = self.path.joinpath(pdf_name)
                pdf_path.write_bytes(pdf_content)

                logger.info(f"成功下载论文: {pdf_name}，文件大小: {len(pdf_content)} bytes")
                return str(pdf_path)

            except Exception as e:
                logger.error(f"第 {attempt + 1} 次尝试失败: {str(e)}")
                if attempt < 2:  # 不是最后一次尝试
                    wait_time = (attempt + 1) * 3  # 递增等待时间
                    logger.info(f"等待 {wait_time} 秒后重试...")
                    time.sleep(wait_time)
                continue

        raise Exception(f"无法下载论文 {self.doi}，所有重试都失败了")

# Usage Example
if __name__ == '__main__':
    # 创建一个用于保存PDF的目录
    save_path = Path('./downloaded_papers')
    save_path.mkdir(exist_ok=True)

    # DOI示例
    sample_doi = '10.3897/rio.7.e67379'  # 这是一篇Nature的论文DOI

    try:
        # 初始化SciHub下载器，先尝试使用代理
        logger.info("尝试使用代理模式...")
        downloader = SciHub(doi=sample_doi, path=save_path, use_proxy=True)

        # 开始下载
        result = downloader.fetch()
        print(f"论文已保存到: {result}")

View on GitHub (pinned to d6bde0fa54)

Solutions

Read the preceding log lines ('第 N 次尝试失败: ...') to identify which stage fails — mirror, extraction, or PDF fetch.
Test the DOI manually in a browser on a live mirror to confirm SciHub actually has the paper.
Fix the environment: working proxy, updated mirror list, longer timeout.
If the paper is on arXiv, use ArxivSource.download_pdf instead of SciHub.

Defensive patterns

Strategy: fallback

Try / catch

try:
    pdf = scihub.download(doi, dirpath)
except Exception as e:
    if '所有重试都失败了' in str(e):
        logger.warning('SciHub failed for %s, trying open-access sources', doi)
        pdf = try_open_access(doi)  # arXiv / unpaywall / publisher OA
        if pdf is None:
            raise
    else:
        raise

Prevention

Treat SciHub as best-effort: chain it after legal/open-access sources with a fallback chain.
Check the per-attempt log lines — they carry the root cause the final message omits.
Confirm network/proxy health before retry loops; blind retries against a blocked network waste minutes.

When it happens

Trigger: All three attempts hit the failure chain: mirror discovery fails (error 87), all mirrors fail the POST (error 88), PDF link extraction from the returned HTML fails (no #pdf element/iframe), or downloading the extracted PDF content fails — then this exception surfaces from the download-by-DOI entry point.

Common situations: Blocked/unstable network to SciHub over the whole retry window; DOI not in SciHub's holdings so every page lacks a PDF element; captcha interstitials on every attempt; proxy timeouts under 30s while SciHub is slow.

Related errors

AI-assisted analysis of binary-husky/gpt_academic@d6bde0fa54 (2026-08-14). Data as JSON: /api/errors/b5032667703b9bc4. Report an issue: GitHub.