Bruno Rodrigues

Today was my first day as "Senior Vibe Coder" at Cloudflare , I pushed some updates to prod, people should start seeing improvements soon!

Luxembourg used to be safe but now not so much! went to my favourite coffee place and left to go to the bathroom and when I came back someone had installed windows 11 on my laptop!!!

stay safe out there

I like to use the chat interfaces wich LLMs and this is the quickest way I found to upload a code base to an LLM before starting a coding session

I'm really happy to learn that it makes your life easier! now try rixpress 😁

yes the rstats-on-nix on Nix fork gets daily Cran bumps, but packages that don't get updated in a long time would come from nixpkgs

no they don't and not all of cran Bioconductor gets built, only a small number of popular packages but it helps

that's right but packages that are the same across the fork and upstream get cached by the official binary cache + we have a 50gb dedicated cache for new packages that only exist on rstats-on-nix

`treemerge`: github.com/b-rodrigues/...

treemerge scans a directory tree, detects all plain-text files, and concatenates them into a single output file with clear per-file headers. It offers a configurable way to assemble large text corpora for supplying contextual input to LLMs.

renv2nix — renv2nix renv2nix

And it's easy to test out! It's def worth knowing it's an option. The vignettes @brodriguesco.bsky.social wrote are pretty good, too.

It might not be for you, but it might be 😁

docs.ropensci.org/rix/referenc...

rix's author here, if you do try it out let me know how it goes!

WHERE IS THE BISCUIT

what this does is that anyone can clone my repository, run `nix develop` and drop into a development shell with all the required dev dependencies to start working.

It's just so easy.

🔥 {torchgnn} 🕸️ is a WIP implementation of Graph Neural Networks in R.

- GCN
- GCN (generalized)
- GraphSAGE
- Aggregators

Development based on personal needs, ofc :)

#rstats
github.com/JosiahParry/...

thank you #RStats 🙏

rix/default.nix at main · ropensci/rix Reproducible Data Science environments for R with Nix - ropensci/rix

you'd define an environment with the required deps versions, for example github.com/ropensci/rix...

this is the default.nix to develop rix. If I needed a specific version of some dependency, I'd add it there.

new blog post:

Of course, someone has to write imperative code to build reproducible data science pipelines. It doesn’t have to be you.

brodrigues.co/posts/2025-1...

NOOOOOOOOOO DON'T START EXPLORING DATA IN A JUPYTER NOTEBOOK!

I'm not sure I understand what you're trying to do, could you describe what you're trying to achieve?

rixpress_demos/r_multi_envs at master · b-rodrigues/rixpress_demos Contribute to b-rodrigues/rixpress_demos development by creating an account on GitHub.

you can have a specific environment with specific packages per node, see github.com/b-rodrigues/...

that being said mojo is not included in Nix for now

rixpress_demos/r_python_quarto at master · b-rodrigues/rixpress_demos Contribute to b-rodrigues/rixpress_demos development by creating an account on GitHub.

here's an example where both R and Python are used github.com/b-rodrigues/...

wanna try the latest R 4.5.2 and bioc 3.22?

rix::rix(
date = "2025-11-01",
r_pkgs = c("dplyr")
)

#RStats #Nix

I find Luxembourguish fascinating. This word is constituted of the German "Gesundheit" (health) and the French "Donnée" (data). In French, pluralized words get an 'S' at the end, but in German 'en'. In LB, even though "donnée" comes from FR it gets pluralized as in DE

new blog post:

Of course, someone has to write imperative code to build reproducible data science pipelines. It doesn’t have to be you.

brodrigues.co/posts/2025-1...

thanks! of course one could ask why one would need to use machine learning for predictions since the solved model is right there, but the point was to show how easy rixpress makes using different languages for one single project 😁

It's difficult not to act arrogant when you see how #Nix elegantly solves so many problems (comprehensive dependency management and universal binaries to name a few) when you see this being re-solved by other half-baked solutions.

#Nix for Polyglot, Reproducible Data Science Workflows

Many thanks for the feedback, I think that this version is now much better: github.com/b-rodrigues/...
if you want to read it online instead: b-rodrigues.github.io/rix_paper/

I aim for a submission before the end of the year!

thanks for the feedback, really cool to see it being useful to others!

cool! if you have questions don't hesitate !

btw, how long have you been using Nix? you use it only inside Docker? why not directly on your machine?