PAVÉS: Bajo los adoquines, la PLAYA 🏖️

PLAYA is intended to get objects out of PDF, with no dependencies or further analysis. So, over top of PLAYA, this package provides PDF, Analyse et Visualisation simplifiÉeS.

Or, if you prefer, PDF Analysis and Visualization for dummiES.

The goal here is not to provide elaborate, enterprise-grade, battle-tested, cloud and AI-native, completely configurable and confoundingly complex classes for ETL. It's to give you some helpful functions that you can use to poke around in PDFs and get useful things out of them, often but not exclusively in the context of a Jupyter notebook.

Installation

Install it from PyPI (as paves) with pip or uv, preferably in a environment. That's all. If you want to play around in the source code you can use hatch or uv (your choice).

Quick start

To use it, I recommend importing playa and then abbreviating the PAVÉS modules like this:

import python
import paves.image as pi
import paves.tables as pt
import paves.text as px

The major use case (well, my major use case) for PAVÉS is looking at stuff in PDFs, generally in a Jupyter notebook. Basically pi.show renders pages to images, pi.box draws boxes around objects, and pi.mark highlights objects with semi-transparent colour. Images are PIL.Image.Image objects everywhere so you can, for example, crop them, save them, etc.

pdf = playa.open("awesome-document.pdf")
page = pdf.pages[0]
image = pi.show(page)
image_with_text_objects = pi.box(page.texts)
image_with_marked_content = pi.box(page.structure)
image_with_highlighted_images = pi.mark(page.images)

Another use case is to extract words and lines of text, with associated styling and graphics state. So for instance you could look at text with associated fonts:

pi.show(px.words(page), label=px.textfont)

It can also detect tables. Contrary to the current zeitgeist, PAVÉS always tries to do everything the simplest and most efficient way, taking advantage of explicit structure when possible (yes, PDFs can have explicit structure). This will fall back to Deep Learning if there isn't any such structure, and if you have the necessary PyTorch and HuggingFace junk installed.

Once you've found these tables you can... do stuff with them. More on this soon.

License

PAVÉS is distributed under the terms of the MIT license.