docling-project/docling

Docling cannot obtain the images inserted into the text box of docx file

Open

#3,314 建立於 2026年4月16日

在 GitHub 查看
 (1 留言) (0 反應) (1 負責人)Python (59,751 star) (4,140 fork)batch import
bugdocxgood first issue

描述

Bug

Docling cannot obtain the images inserted into the text box of docx file

Steps to reproduce

  1. download the zip from https://www.3gpp.org/ftp/tsg_ran/WG1_RL1/TSGR1_124b/Docs/R1-2602023.zip , unzip this zip file and then you can get
    a file named R1-2602023_Discussion on other aspects of CSI acquisition for 6GR.docx
  2. use docling to convert the docx to markdown file
            cmd = [
                "docling",
                "--from", "docx",
                "--to", "md",
                "--output", os.path.dirname(md_path),
                doc_path
            ]

            print(f"run command: {' '.join(cmd)}")
            result = subprocess.run(
                cmd,
                capture_output=True,
                text=True,
                check=True
            )
  1. I get the markdown, I have obtained the markdown file and carefully compared the contents of the docx file and the markdown file (I use typora to read markdown file).
  2. I found the error. The details are as follows: The display effect of the markdown file is as follows:
The display effect of the docx file is as follows:

Docling version

Docling version: 2.88.0 Docling Core version: 2.72.0 Docling IBM Models version: 3.13.0 Docling Parse version: 5.8.0 Python: cpython-312 (3.12.12) Platform: Linux-6.8.0-90-generic-x86_64-with-glibc2.39

Python version

Python 3.12.12

貢獻者指南

技術棧
python
領域
backend
議題類型
bug
難度面向新貢獻者的預計實作難度，1 表示很小改動，5 表示專家級工作。
3
預計時間有經驗貢獻者完成調查、實作、測試並準備 pull request 的粗略時間範圍。
1-3 hours
活動狀態議題目前的可參與程度：新鮮、活躍、陳舊、阻塞或等待維護者輸入。
blocked
清晰度議題是否清楚說明預期改動、驗收標準和下一步。
clear
前置要求
Pythondocx formatdocling codebase
新手友善度1-100 的估計分數，表示該議題對首次貢獻者的友善程度。
45
研究方向
Investigate how Docling extracts images from docx files, particularly images inside text boxes. Start by examining the docx parser module in the docling codebase (likely in docling/document/converter.py or similar). Look for existing image extraction logic and identify where text boxes are handled. Use the provided sample document (R1-2602023 Discussion on other aspects of CSI acquisition for 6GR.docx) to reproduce the issue and trace the parsing flow. Check if there is a missing handler for images within text boxes compared to regular inline images.