manulint

Catch manuscript reporting and formatting issues before submission — locally, with no API key.

简体中文 · Example report · Report a bug

manulint reads Word (.docx), Markdown, and plain-text manuscripts. It flags expressions such as p = 0.000, repeated references, broken reference numbering, and low-resolution embedded figures. Select an original-research, review, or case-report profile instead of applying IMRaD to every article.

manulint rule check | score 84/100 | 4 flagged rules / 4 warnings; score is unvalidated
[FLAG] Found 1 rounded p = 0.000 expressions; report p < 0.001 instead.
[FLAG] image1.png (occurrence 1): effective resolution 60 dpi (minimum axis).
[FLAG] Missing reference numbers (first 10): [3].
[FLAG] Repeated reference labels: [4].
[WARN] Found 1 potentially repeated reference entries after removing number labels.

This is an abbreviated excerpt from the synthetic demonstration. The example contains no patient data and supports no scientific claim. The score counts rule flags; it is not a validated measure of publication readiness.

Try it in one command

Python 3.9 or later is required. From an installed package:

python -m manulint.demo --out-dir demo-output

This creates a fictional Markdown manuscript, a Word version with a deliberately low-resolution figure, and demo-output/report.md. It does not download data. Existing example files are preserved; choose a new output directory to run it again.

Install

Install the latest published version:

python -m pip install --upgrade manulint

This README documents 0.1.1. If the latest published version is older, install the repository version to use the profiles and new demonstration:

git clone https://github.com/chenmei24/manulint.git
cd manulint
python -m pip install .

Check your manuscript

manulint paper.docx --md report.md
manulint paper.md --article-type review --language en --md review-report.md
manulint case.docx --article-type case-report --language zh
manulint paper.docx --p-style upper --md report.md

Exit codes: 0 means checks completed, including when flags exist. 2 means invalid/empty/unreadable input, an invalid option, or a report-write failure. Exit 0 is not a scientific validation result. No report or score is generated for an empty manuscript.

What is checked

Important extraction limits

• Word extraction covers body paragraphs, table text, and modern embedded body drawings, in document order. Headers, footers, text boxes, footnotes, equations, charts, legacy VML pictures, and tracked-change text are not fully assessed.
• Reusing the same image at different displayed sizes creates separate resolution findings. DPI checks use displayed width and height, not image-file metadata. Cropping and journal-specific figure requirements still need manual review.
• Section detection expects recognizable English/Chinese headings. A structured abstract is separated using a later Keywords or Introduction heading; without those boundaries its end can be ambiguous. Unusual headings may be missed.
• Numbered references support [1], (1), 1., and 1). Wrapped continuation lines are joined; author-year duplicates are compared per non-empty line. References stop at recognized later sections or another Markdown heading.
• Files stay on the local computer. Remote Markdown images are not downloaded. Reports may contain excerpts and filenames from your manuscript, so review them before sharing.

Scoring and interpretation

The score is max(0, 100 - 3 × flagged rules - warnings). These weights are a convenience, have not been validated, and can count related issues more than once. Use the individual findings and the target journal's author instructions. A score of 100 does not establish completeness, scientific quality, ethical adequacy, or acceptance likelihood.

Contribute

Please report a reproducible rule failure with a small synthetic example. Do not upload confidential manuscripts or patient data. See CONTRIBUTING.md for local checks and CHANGELOG.md for changes.

python -m pip install .
python -m unittest discover -s tests -v
python -m manulint.demo --out-dir demo-output

CI is configured for Linux and Windows on Python 3.9, 3.11, and 3.13. Tagged PyPI publication runs the regression suite and distribution checks before trusted publishing.

License

MIT.

Related tools / 相关工具

• Medical-imaging manifest audits
• Traceable clinical CSV QA

Launch tutorials and share images / 推广教程与展示图