Cosma at FOSDEM 24


These are my (slightly edited) notes for my presentation of Cosma at the Open Research devroom of FOSDEM 24. I presented on behalf of a team: first and foremost Guillaume Brioudes, who authors the code; and Olivier Le Deuff, Clément Borel, David Pucheu, Rayya Roumanos, our senior colleagues who have advised us over the years. The presentation was about the architecture of the program, which could interest people working on interactive publications; as well as the features of the program, which could interest scientists with information management needs (which is to say, scientists).

Portrait of Paul Otlet (colorized) Cosma came as part of HyperOtlet, a research programme on Paul Otlet. I’m glad to mention him at FOSDEM 24 in Brussels because he was born here. He was a figure of Belgian intellectual life, a pioneer of knowledge organization, a precursor of information science. He’s the one who popularized the words document and documentation.

Archival document from the Mundaneum showing a book’s contents being extracted and organized into the drawers of a card-based encyclopedia. Otlet’s main idea was to go beyond the book: to extract facts from documents and organize them into a universal encyclopedia. He worked all his life on tools to achieve this, such as bibliography, classification, index cards, and so on. The Mundaneum is a museum and archival center dedicated to Otlet in Mons, Belgium.

Screenshot of our relational map of Paul Otlet. In 2018 and 2019 we worked on this map of Paul Otlet’s professional network. It was our take on the idea of combining a graph view and a card view. One day, I asked my developer colleague, Guillaume Brioudes, if he could make the same interface but for my research notes. A bunch of plain text files, each one with notes on a specific thing. The files reference eachother with links: title or identifier of the target file between double brackets, just like in a wiki.

Example of plain text notes, based on Andy Matuschak’s public notes.

Cosma renders these as a standalone HTML file, containing a rendering of the text of each file, a graph view and a few navigational tools.

A “cosmoscope”, the HTML file generated by Cosma based on a directory of plain text files.

This could be any kind of knowledge base, a network of people, concepts, events, anything. Any kind of knowledge work that involves describing things and the relationships between them. Conceptually, this can be thought as a commonplace book, a wiki, a Zettelkasten, a glossary, even a mind map to some extent. What distinguishes Cosma is its architecture, and the fact that it was designed specifically around the needs of scientific writing.

Examples of cosmoscopes generated with Cosma. Architecture: Cosma is only a visualization program. And most of the features are actually located within the files that it creates. It is a command-line tool. It is used to generate HTML files. We put all the features in the export. It’s a bit like a wiki made with TiddlyWiki, in the sense that it’s a single HTML file, except that these are read-only. So it’s less like a web application and more like an “augmented document”, or an “interactive essay”, where the data cannot be edited from the file, but you many options to interact with the data. You can share this file, and people who open it will be able to navigate the data in the same way as you. This is great when collaborating, or when working with students.

Using filters to hide certain types of notes: Screenshot of a cosmoscope showing how filters are used to hide certain types of notes. Cosma’s features are geared towards information retrieval/emergence: it is mostly a memory aid. Cosma encourages knowledge organization by providing visual cues and filters for categories.

It also encourages link-based knowledge organization by generating contextualized backlinks: incoming links with their surrounding context.

An example of contextualized backlink in Cosma. This is an old hypertext idea that is absent from the Web and is coming back with the “tools for thought” movement.

Backlinks can also be generated from citations. If you have bibliographic data in a file and you cite it using Pandoc’s citation syntax, Cosma generates a bibliographic note in the HTML file, with backlinks corresponding to where it’s been cited (and again, with the surrounding context). You can use this to check where you’ve cited something and how.

An example of contextualized backlink to a bibliographic reference in Cosma. Contrary to other notes, this one has been generated automatically by the program.

I will close on the idea of network synthesis. In my PhD dissertationPerret, De l’héritage épistémologique de Paul Otlet à une théorie relationnelle de l’organisation des connaissances, 2022.
I argue that linking is a knowledge organization process: it can be used to index, to classify, to assign, to tag… but also for ideation, by composing existing things into new ideas. The purpose of Cosma is to encourage this process of network synthesis (= assemble and expand document graphs).


Question: Can it be used to view data from Obsidian, Logseq, etc.?

Answer: Yes, our colleague from CNRS Kévin Polisano has made an Obsidian to Cosma converter.

Q: What kind of notes is it best suited for?

A: Evergreen notes (atomic, conceptual, densely linked; see Andy Matuschak’s definition); glossaries; biographical notes; reading notes… See the examples on Cosma’s website.


Thanks to the volunteers who organized FOSDEM, especially the Open Research devroom, and in particular Mathieu Jacomy for suggesting I propose a talk.


Perret, Arthur. De l’héritage épistémologique de Paul Otlet à une théorie relationnelle de l’organisation des connaissances. Thèse de doctorat. Université Bordeaux Montaigne, 2022.
Perret, Arthur, Brioudes, Guillaume, Borel, Clément and Le Deuff, Olivier. Cosma. Zenodo, 2021.