Lightweight PDF parser with layout, tables, formulas and bounding boxes

Developer Beatriz Almeida released papero, an open-source PDF parser that extracts reading order, tables, formulas, figures and bounding boxes from PDFs using geometry alone, with no ML models and CPU-only execution. The tool installs via `pip install pdf-text-api`, runs in the browser, Python or as a REST API, and exports to Markdown, JSON, Word, Excel, HTML and CSV, with OCR for scanned pages and Apache Tika support for DOCX, PPTX, XLSX, EPUB and HTML. papero is aimed at RAG, search and LLM pipelines that need document structure rather than raw text.

PDF → Markdown · JSON · Word · Excel — with reading order, tables, formulas, figures and the position of every block. CPU only. No ML models. Runs in your browser, in Python, or as an API. ▶ Try it in your browser https://beatrizalmeidaf.github.io/papero-pdf-text-extractor/ · Quick start quick-start · Benchmarks <sub 30 seconds in the browser app: load a PDF, inspect any block, check tables and formulas, export to Word. Your PDF never leaves your machine. MP4 https://github.com/beatrizalmeidaf/papero-pdf-text-extractor/blob/main/web/assets/demo.mp4 </sub Getting the text out of a PDF is easy. Getting its structure back — which column comes first, which lines are a table, where the formula is — is what makes the output usable for RAG, search and LLMs. papero does that with plain geometry, so it stays fast on a laptop CPU. | 📖 Reading order Two- and three-column papers read column by column. Headers, footers, page numbers and repeated logos are set aside. | ▦ Real tables Ruled, borderless and LaTeX booktabs tables come back as rows and columns — multi-line cells included. Export to CSV or Excel. | ∑ Formulas Superscripts, subscripts and math symbols become LaTeX E = mc^{2} , plus a cropped image of the formula. | | 📍 Position of everything Every block has a bounding box — cite the exact spot in a RAG answer, draw over the page, or crop it. | 🖼 Figures & charts Images and vector charts are cropped to PNG, with their caption, axis labels and legend kept together. | 📝 Back to Word Alignment, indents, line spacing, bold runs and fonts are kept, so a .docx export looks like the original page. | Also: accents drawn as separate glyphs in LaTeX PDFs Computa¸ca˜o → Computação , invisible white text used by form generators is dropped, scanned pages go through OCR, and DOCX/PPTX/XLSX/EPUB/HTML are read through Apache Tika. pip install pdf-text-api python from pdf text api import extract doc = extract "paper.pdf" print doc.to markdown Or skip the install: open the browser app https://beatrizalmeidaf.github.io/papero-pdf-text-extractor/ , drop a PDF, export to the format you need. More Python — tables, formulas, positions, images, options python from pdf text api import extract, extract text doc = extract "paper.pdf", images=True doc.tables 0 .rows "Model", "Accuracy" , "Base", "0.81" , ... doc.formulas 0 .latex "E = mc^{2}" doc.figures 0 .image.data PNG bytes for block in doc.pages 0 .blocks: reading order, with positions print block.type, block.bbox, block.text :60 doc.to html keeps alignment and indents doc.to dict the full JSON extract "slides.pptx" .to markdown any format Apache Tika reads extract text "contract.pdf" .text fastest: clean text only | Option | Default | | |---|---|---| | pages | all | "1-3,5,10-" | | images | False | crop figures, tables and formulas to PNG | | tables / formulas | True | detection on/off | | ocr | "auto" | "auto" scanned pages only , "force" , "off" | | ocr language | "por+eng" | Tesseract languages | | tika | True | False runs the layout engine alone no Java | | workers | 1 | processes for long documents | CLI pdf-text-api extract paper.pdf -o paper.md --images Markdown + images/ folder pdf-text-api extract paper.pdf -o paper.json format from the extension pdf-text-api extract paper.pdf -f csv -o tables.csv tables only pdf-text-api extract paper.pdf -p 1-5 -f html pdf-text-api extract paper.pdf --fast clean text only pdf-text-api serve --port 8000 API + browser app REST API & Docker docker compose up API + Apache Tika + Tesseract + browser app on :8000 curl -F "file=@paper.pdf" "localhost:8000/v1/extract?format=markdown" curl -F "file=@paper.pdf" "localhost:8000/v1/extract?format=zip&images=true" -o paper.zip curl -F "file=@paper.pdf" "localhost:8000/v1/extract?per page=true" blocks + positions One endpoint, POST /v1/extract ; interactive docs at /docs . | Parameter | Default | | |---|---|---| | mode | structured | structured layout + Tika or fast text only | | format | json | json , markdown , text , html , csv , zip | | pages | all | 1-3,5,10- | | per page | false | include pages, blocks and positions in the JSON | | images | false | crop figures, tables and formulas | | ocr | auto | auto , force , off | Configuration through environment variables — see .env.example https://github.com/beatrizalmeidaf/papero-pdf-text-extractor/blob/main/.env.example . Every block knows what it is and where it was: { "type": "table", "bbox": 56.7, 294.8, 481.9, 374.2 , "rows": "Model", "Accuracy" , "Base", "0.81" , "caption": "Table 1: Comparison between models." } | Output | Python · CLI · API | Browser app | |---|---|---| | Markdown, plain text, JSON | ✓ | ✓ | | HTML keeps alignment and indents | ✓ | ✓ | | CSV of the tables, ZIP with images | ✓ | ✓ | | Word .docx that keeps the page's look | — | ✓ | | Excel .xlsx , one sheet per table | — | ✓ | Full JSON schema and block types { "schema": "pdf-text-api/document@1", "engine": "tika+pdfium", "page count": 12, "metadata": { "title": "...", "author": "...", "language": "en" }, "pages": { "number": 1, "width": 595.3, "height": 841.9, "blocks": { "id": "p1-b4", "type": "paragraph", "bbox": 74.0, 217.0, 522.0, 275.0 , "text": "Atestamos que a estudante ...", "style": { "pt": 11.0, "font": "Arial", "bold": false }, "format": { "align": "justify", "first line": 42.7, "line spacing": 1.8 }, "runs": { "text": "FULANA DE TAL", "bold": true, "italic": false, "script": null } } } } Block types: heading with level , paragraph , list item with marker , table with rows , figure , formula with latex , caption , code , and — kept apart from the text — header , footer , page number . Bounding boxes are x0, y0, x1, y1 in points, origin at the top-left of the page. Dense arXiv papers multi-column, formulas, tables, figures on one laptop CPU, no GPU. papero · fast returns clean text; papero · structured also rebuilds reading order, tables, formulas and figures — 0 failures on 54 papers, 39 ms per page median . Reproduce with benchmarks/ https://github.com/beatrizalmeidaf/papero-pdf-text-extractor/blob/main/benchmarks . | | papero | PyMuPDF | pdfplumber | pypdf | Docling | Marker | |---|---|---|---|---|---|---| | License | MIT | AGPL | MIT | BSD | MIT | GPL | | Needs ML models / PyTorch | no | no | no | no | yes | yes | | Multi-column reading order | ✓ | partial | — | — | ✓ | ✓ | | Structured tables | ✓ | ✓ | ✓ | — | ✓ | ✓ | | Formulas | LaTeX from glyphs + image | — | — | — | ✓ | ✓ | | Bounding boxes | ✓ | ✓ | ✓ | — | ✓ | ✓ | | DOCX / PPTX / XLSX / EPUB | ✓ | partial | — | — | ✓ | partial | | Runs entirely in the browser | ✓ | — | — | — | — | — | ML-based tools still win on very irregular layouts and complex math stacked fractions, matrices — papero gives you the formula as approximate LaTeX and as an image so nothing is lost. Two engines run on the same file at the same time : - A layout engine on PDFium reads every glyph with its position, font and size, plus every rule and image, and rebuilds columns, tables, formulas, lists and figures with a column-aware XY-cut. - Apache Tika adds metadata, tagged-PDF headings, OCR Tesseract and every non-PDF format. The browser app runs the same algorithm ported to JavaScript on pdf.js, and CI checks block by block that both engines agree. Limitations - Math: LaTeX is rebuilt from glyphs — stacked fractions, matrices and big radicals come out linear the cropped image is always there . - Borderless tables with very narrow gaps between columns can read as text. - Scanned PDFs need OCR, which runs on the server path Tesseract is in the Docker image . - Word/Excel export is in the browser app for now. Development git clone https://github.com/beatrizalmeidaf/papero-pdf-text-extractor.git && cd pdf-text-extractor pip install -e ". dev " pytest -q includes real-world regressions ruff check src tests && ruff format --check src tests npm install --prefix tests/js && python tests/js/expected.py tests/js/out && node tests/js/parity.mjs tests/js/out python -m http.server -d web browser app at http://localhost:8000 src/pdf text api/ is the Python engine, API and CLI · web/ is the browser app GitHub Pages · tests/js/ checks the two engines agree · benchmarks/ downloads the dataset and draws the chart. Found a PDF papero gets wrong? That's the most useful issue you can open — attach the file or a page of it and say what you expected. Reading order, tables, formulas, encoding, OCR and browser/server differences are all fair game. If papero saves you time, a ⭐ helps other people find it. Keywords: PDF to Markdown · PDF to JSON · PDF to Word · PDF to Excel · PDF table extraction · PDF parser · document parsing · layout analysis · reading order · multi-column PDF · formula extraction · LaTeX · bounding boxes · OCR · Apache Tika · PDFium · pdf.js · RAG preprocessing · LLM document loader · Docling alternative · PyMuPDF alternative · converter PDF para Markdown, Word e Excel · extrair tabelas de PDF · extrair texto de PDF mantendo a formatação · OCR de PDF escaneado <sub MIT © Beatriz Almeida · package and imports keep the name pdf-text-api / pdf text api for compatibility.</sub