github.devotd.
ReleaseKorean tech

chrisryugj/kordoc v4.22.5 — Fixes for table and symbol handling in layout and PDF

The v4.22.5 release of chrisryugj/kordoc, a CLI and MCP server that converts HWP, HWPX, PDF and Office documents into Markdown, fixes table and symbol handling in layout and PDF extraction and updates the reproduction and quality materials.

Repositorychrisryugj/kordoc ↗

The v4.22.5 release of chrisryugj/kordoc, a CLI and MCP server that converts HWP, HWPX, PDF and Office documents into Markdown, fixes table and symbol handling in layout and PDF extraction and updates the reproduction and quality materials.

GitHub description: 모두 파싱해버리겠다 — HWP·HWPX·PDF·Office 문서를 Markdown으로. 양식 자동 채우기와 신구대조를 갖춘 CLI·MCP 서버 | Convert Korean documents (HWP, HWPX, PDF, Office) to Markdown — CLI and MCP server with form filling and diff

stars
★ 2.5K
Forks
457
Language
TypeScript
License
MIT
Latest release
v4.22.5
Read original source ↗View on GitHub ↗

The GitHub release v4.22.5 of chrisryugj/kordoc was published on October 16, 2026. This repository is a tool that turns HWP, HWPX, PDF and Office documents into Markdown, shipping both a CLI that runs from the command line and an MCP server that AI applications load. This version fixes table and symbol handling in document layout and PDF extraction, and updates the reproduction and quality materials.

Working directly with HWP and HWPX, the Korean document formats, is the scope of this repository. On top of the document-to-Markdown conversion feature, it also provides automatic form filling that puts values into the designated blanks of a form, and a new-old comparison that diffs two documents. The release also presents the changes in the layout and PDF evaluation metrics as numbers. Updating the reproduction and quality materials, so that layout and PDF extraction results can be looked at again under the same conditions, is also part of this version's changes. The repository has 2502 GitHub stars.

However, the release states that complete handling of PDF, layout and optical character recognition (the technology that reads character images as text) is not yet in place. Which document samples the layout and PDF evaluation metrics came from, and how far the table and symbol fixes reach, cannot be confirmed from the given information. The fact that the metrics are presented as numbers alone does not allow them to be taken as general reference values for document conversion quality. Unless real documents with many tables and symbols are fed in directly, there is no way to know how far it works. It is better to fix the formats and samples to be handled and check them separately.