file2md 0.1.0

Convert PDF and DOCX to clean, grep-friendly Markdown for AI/IDE workflows

file2md

Convert PDF and DOCX files to clean, grep-friendly Markdown optimized for AI tools and IDE workflows.

Features

• PDF conversion — text extraction with PyMuPDF, page separators, scanned PDF detection
• DOCX conversion — headings, lists, tables, bold/italic preserved as Markdown
• Grep-friendly output — paragraph reflow, hyphenation fix, whitespace normalization
• Drag-and-drop web UI — upload and convert files in your browser
• Full CLI — single file, batch, and web server modes
• Table extraction — PDF tables converted to GitHub-flavored Markdown
• Header/footer removal — heuristic detection of repeating headers/footers
• Metadata headers — source filename and conversion timestamp in output
• YAML frontmatter — optional structured metadata for downstream tools

Installation

# CLI only (lightweight)
pip install file2md

# With web UI
pip install file2md[web]

# Development (all dependencies)
pip install file2md[all]

Quick Start

Web UI

file2md serve
# Open http://127.0.0.1:8000 and drag your files

CLI — Single File

# Basic conversion
file2md convert document.pdf -o document.md

# With all enhancements
file2md convert report.pdf -o report.md --clean --frontmatter --page-labels --extract-tables

CLI — Batch

# Convert all PDFs and DOCXs in a directory
file2md batch ./documents --out-dir ./markdown --recursive

CLI Reference

file2md convert

Convert a single PDF or DOCX file to Markdown.

Flag	Description
-o / --output	Output file path (defaults to input name with .md)
--clean	Normalize whitespace, reflow paragraphs, fix hyphenation
--frontmatter	Add YAML frontmatter (source, timestamp, converter)
--page-labels	Add ## Page N headings (PDF only)
--extract-tables	Detect and convert tables to GFM (PDF)
--max-chars N	Truncate output at N characters
--overwrite	Overwrite existing output file
--quiet	Suppress warnings
--verbose	Show detailed progress
--json	Machine-readable JSON output

file2md batch

Batch convert all PDF/DOCX files in a directory.

Flag	Description
--out-dir	Output directory (required)
--recursive	Process subdirectories
All flags from convert	Same options available

file2md serve

Start the web UI server.

Flag	Description
--host	Host to bind to (default: 127.0.0.1)
--port	Port to listen on (default: 8000)

Exit Codes

Code	Meaning
0	Success
2	Unsupported file type
3	Extraction failed
4	Scanned PDF detected (no OCR)

Output Conventions

Metadata

Every converted file includes a metadata comment:

<!-- source: document.pdf | converted: 2026-02-26T12:00:00Z | converter: file2md v0.1.0 -->

With --frontmatter:

---
source: document.pdf
converted: 2026-02-26T12:00:00Z
converter: file2md v0.1.0
---

PDF Page Separators

Pages are separated by ---. With --page-labels:

## Page 1

Content of page 1...

---

## Page 2

Content of page 2...

Clean Mode (--clean)

• Paragraph reflow — undoes hard line wraps from PDF extraction
• Hyphenation fix — merges hyphen-\nated words across lines
• Header/footer removal — detects and removes repeating page headers/footers
• Whitespace normalization — collapses extra spaces, limits blank lines

Architecture

src/file2md/
├── convert.py          # Main entry point — dispatches by file type
├── pdf.py              # PDF → Markdown (PyMuPDF)
├── docx_converter.py   # DOCX → Markdown (python-docx)
├── normalize.py        # Text cleanup (reflow, hyphenation, headers/footers)
├── cli.py              # Click CLI (convert, batch, serve)
├── web.py              # FastAPI web server
├── utils.py            # Shared types, validation, metadata
└── templates/
    └── index.html      # Drag-and-drop web UI

Development

git clone https://github.com/faizkhairi/file2md.git
cd file2md
python -m venv .venv
source .venv/bin/activate  # Linux/macOS
# .venv\Scripts\activate   # Windows
pip install -e ".[all]"

# Run tests
pytest

# Lint
ruff check src/ tests/

# Build
python -m build

Known Limitations

• No OCR — scanned/image-only PDFs are detected and rejected with a clear error (exit code 4). OCR support is planned for a future release.
• Complex PDF layouts — multi-column documents, sidebars, and footnotes may produce text in unexpected order.
• Nested DOCX lists — only flat bullet/numbered lists are supported. Nested and mixed lists are not preserved.
• Merged table cells — may produce duplicated or empty cells in the Markdown output.

Troubleshooting

"All pages appear to be scanned images" — The PDF contains only images, no extractable text. You need to OCR the PDF first using a tool like ocrmypdf before converting.

Tables not appearing (PDF) — Use the --extract-tables flag. Table detection is off by default to keep output clean for text-heavy documents.

Output has hard line breaks — Use the --clean flag to enable paragraph reflow, which joins lines that were artificially broken by PDF formatting.

License

MIT