Convert PDF, DjVu, and DOCX into clean Markdown for RAG and LLM pipelines, with an engine that runs on a 4 GB GPU and deterministic cleaning.
uv tool install --python ">=3.11,<3.15" raw2md --torch-backend cu130raw2md is an open-source Python command-line tool that converts PDF, DjVu, and DOCX documents into clean Markdown for retrieval-augmented generation (RAG) and other large language model (LLM) pipelines. It recognizes PDF and DjVu with its own engine raw2md-engine, which is based on marker 1.10.2 by Datalab. It converts DOCX with pandoc. Deterministic rules then repair the defects in headings, tables, formulas, lists, and text. Markdown from another converter goes through the same cleaning. The rules use no LLM, so the same input always gives the same cleaned text.
raw2md-engine runs on a low-end GPU: 4 GB of video memory is enough, for example on an NVIDIA T1200 laptop card. On the MPDocBench-Parse benchmark, raw2md without the LLM stages gets an Overall score of 69.30.
Two optional LLM stages repair what the rules leave: inspection corrects recognition errors against the source pages, and post-processing repairs damaged markup.
Why raw2md?
raw2md is built for large, mixed collections of long documents:
- The tool focuses on the structure of the document: the heading levels, the tables, the lists, and the formulas. A RAG pipeline can then split the text by sections.
- One command converts a whole folder. The tool picks the engine for each format:
raw2md-enginefor PDF,pandocfor DOCX, and DjVuLibre withraw2md-enginefor DjVu. - A rule does not guess: it leaves an ambiguous defect in place and reports it.
- Each result starts with a header that holds a quality score from 0 to 100 and a verdict. The verdict is
okat a score of 50 or more by default, andbadbelow 50 or after a severe defect. The verdict shows which files need a review. - Without the LLM stages, the documents do not leave your machine.
- An interrupted run continues with
raw2md resumeand skips the finished files.
What’s the impact?
These are the scores of raw2md 0.3.0 without the LLM stages on the full MPDocBench-Parse set. The columns and the number format are the same as in the leaderboard of the benchmark:
| Model | Overall | Text Edit | Truncated Text Edit | Formula CDM | Table TEDS | Truncated Table TEDS | Figure F1 | Read Order | Heading TEDS |
|---|---|---|---|---|---|---|---|---|---|
| raw2md 0.3.0 | 69.30 | 0.130 | 0.299 | 82.58 | 69.56 | 55.26 | 65.18 | 0.206 | 45.30 |
The three edit distances (Text Edit, Truncated Text Edit, and Read Order) are better when they are lower. The other columns are on a 0–100 scale and are better when they are higher. Overall is the mean of the eight task scores, with each edit distance x counted as (1 − x) · 100. The benchmark results give every metric, the scores on the English part, and the way to run the benchmark.
The benchmark, however, tells only part of the story: raw2md is designed primarily for long, complex documents. The examples show how it copes with entire books:
| Result | Source type | Pages | Engine | Source |
|---|---|---|---|---|
book.md |
scanned PDF | 344 | raw2md-engine |
Wikimedia Commons |
thesis.md |
born-digital PDF | 230 | raw2md-engine |
University of Cambridge |
report.md |
DOCX | – | pandoc |
University of Derby |
How does it work?
The first step, the conversion itself, runs on one of these engines:
raw2md-enginefor PDF: an improved copy ofmarker1.10.2,surya-ocr0.17.1, andpdftext0.6.3 by Datalab, bundled with raw2mdpandocfor DOCX- Gemini for OCR of PDF and DjVu files (the
-eflag) - DjVuLibre to turn a DjVu file into a PDF before recognition
The cleaning of a file then goes through three stages, and two of them are optional:
| Stage | Formats | Turned on by | What the stage does | Guide |
|---|---|---|---|---|
cleaning |
all | always | removes typical artifacts with deterministic rules | cleaning.md |
inspection |
PDF, DjVu | the -i flag |
checks the text against the source | inspection.md |
post |
all | the -p flag |
repairs the structure without the source | post.md |
The engines leave typical defects in Markdown. In a RAG pipeline, these defects can split chunks in the wrong places and hide words from search. The cleaning stage repairs more than 70 types of defects (full list), but only where the defect is unambiguous. For example:
- running heads and page numbersremoved
- wrong heading levelscorrected from the PDF bookmarks, the table of contents, or the numbering
- tables cut by a page breakmerged
- words broken across linesjoined
- paragraphs broken into short linesjoined
- formulas cut into partsjoined
- recognition loopscut to one copy
- navigation links of
raw2md-engineandmarkerreplaced with their text
After cleaning, and again after each LLM stage, a deterministic quality evaluation grades the text with a score from 0 to 100. The score, the verdict ok or bad, and the checks that lowered the score most go into the header of the result.
For now, inspection runs on Gemini only, while post-processing can use either Gemini or Claude Code through claude -p.
Questions
Does raw2md need a GPU?
For raw2md-engine, yes: an NVIDIA GPU of the Turing generation or newer (GTX 16xx, RTX 20xx, and later cards) with 4 GB of VRAM or more and driver 580 or newer. Without such a card, the engine runs on the CPU. This mode works, but it is much slower.
Do my documents leave my machine?
Not without the LLM stages. Inspection sends the source pages and the recognized text to the provider, -e sends the page images, and post-processing sends fragments of the text.
Does the same input always give the same output?
The cleaning does. It uses deterministic rules, and a second pass changes nothing. An ambiguous defect stays in the text and goes into the findings.