binary-husky/gpt_academic · error · RuntimeError

请先将.doc文档转换为.docx文档。

Error message

请先将.doc文档转换为.docx文档。

What it means

For files whose extension is not docx, Word_Summary tries to open them through pywin32's Word.Application COM automation. The bare except hides the real error and raises a request to convert .doc to .docx. Therefore it fires for every non-docx path where COM is unavailable or Word cannot open the file, not only for actual .doc files.

Source

Thrown at crazy_functions/Word_Summary.py:30

    for index, fp in enumerate(file_manifest):
        if fp.split(".")[-1] == "docx":
            from docx import Document
            doc = Document(fp)
            file_content = "\n".join([para.text for para in doc.paragraphs])
        else:
            try:
                import win32com.client
                word = win32com.client.Dispatch("Word.Application")
                word.visible = False
                # 打开文件
                doc = word.Documents.Open(os.getcwd() + '/' + fp)
                # file_content = doc.Content.Text
                doc = word.ActiveDocument
                file_content = doc.Range().Text
                doc.Close()
                word.Quit()
            except:
                raise RuntimeError('请先将.doc文档转换为.docx文档。')

        # private_upload里面的文件名在解压zip后容易出现乱码（rar和7z格式正常），故可以只分析文章内容，不输入文件名
        from crazy_functions.pdf_fns.breakdown_txt import breakdown_text_to_satisfy_token_limit
        from request_llms.bridge_all import model_info
        max_token = model_info[llm_kwargs['llm_model']]['max_token']
        TOKEN_LIMIT_PER_FRAGMENT = max_token * 3 // 4
        paper_fragments = breakdown_text_to_satisfy_token_limit(txt=file_content, limit=TOKEN_LIMIT_PER_FRAGMENT, llm_model=llm_kwargs['llm_model'])
        this_paper_history = []
        for i, paper_frag in enumerate(paper_fragments):
            i_say = f'请对下面的文章片段用中文做概述，文件名是{os.path.relpath(fp, project_folder)}，文章内容是 ```{paper_frag}```'
            i_say_show_user = f'请对下面的文章片段做概述: {os.path.abspath(fp)}的第{i+1}/{len(paper_fragments)}个片段。'
            gpt_say = yield from request_gpt_model_in_new_thread_with_ui_alive(
                inputs=i_say,
                inputs_show_user=i_say_show_user,
                llm_kwargs=llm_kwargs,
                chatbot=chatbot,
                history=[],
                sys_prompt="总结文章。"

View on GitHub (pinned to d6bde0fa54)

Solutions

1. Convert the file to .docx before upload, preferably on a machine with Word or with LibreOffice.
2. On a server, run: soffice --headless --convert-to docx file.doc, then upload the result.
3. On Windows, install Microsoft Word and pywin32 and ensure Word can open without a modal dialog.
4. Use os.path.abspath(fp) instead of os.getcwd() + '/' + fp.
5. Log the original COM exception so the actual cause is visible.

Example fix

# before
doc = word.Documents.Open(os.getcwd() + '/' + fp)

# after
doc_path = os.path.abspath(fp)
doc = word.Documents.Open(doc_path)

import sys
ext = os.path.splitext(fp.lower())[1]
if ext != ".docx":
    if sys.platform != "win32":
        raise ValueError("Convert .doc to .docx before upload on non-Windows systems")
    try:
        import win32com.client  # noqa
    except ImportError as e:
        raise ValueError("pywin32 is required for .doc support") from e

def is_supported_word_input(fp: str) -> bool:
    return os.path.isfile(fp) and os.path.splitext(fp.lower())[1] == ".docx"

try:
    yield from 解析docx(...)
except RuntimeError as e:
    if "转换为.docx" in str(e):
        convert_with_libreoffice_or_prompt_user(fp)
    else:
        raise

When it happens

Trigger: Running on Linux/macOS or in Docker without Microsoft Word; pywin32 is not installed; Word is not activated or cannot start headless; os.getcwd()+'/'+fp is not a valid absolute/relative path; the document is protected, corrupt, or not really a .doc.

Common situations: Deploying gpt_academic in Docker; uploading legacy .doc files to a server; uppercase .DOC handling; another Word COM instance or dialog blocks Dispatch/Open; insufficient filesystem permissions.

AI-assisted analysis of binary-husky/gpt_academic@d6bde0fa54 (2026-08-14). Data as JSON: /api/errors/617aaf11e7a61808. Report an issue: GitHub.