Valve: R belongs in production

[@kennedymwavu]

This is the content I signed up for!👌

[@flonga2302]

Absolutely awesome! Thanks for building this, decreasing the overhead to putting r in production is key to the growth of the community. Love this work.

[@blaisepascal3905]

Amazing content, thank you so much!

[@nischal92]

Great work with `{valve}` and very well presented! Good example of using Rust as well. I'm digging this direction towards using `{plumber}` for web apps in general.

[@aswanijehangeer]

Great job, Josiah! Your content is truly impressive. Keep up the excellent work and continue delivering such high-quality content.

[@vacon04]

Hey thank you for the video. I saw that you also already have the docker instructions in the package website. I'm hoping to run a simple app with plumber in production in the next few weeks. I was going to just use caddy as a load balancer for multiple plumber instances but I think using Valve should provide better functionality out of the box. Thanks again!

[@josiahparry]

Just make sure that your Docker container has enough vCPU to handle multiple threads and ensure that you don't over-allocate them as well! And you can also use a load balancer in front of multiple Valve apps

[@vacon04]

@@josiahparry thanks! I have a cloud instance with 8 vcpus so it should be good. I'm not expecting massive traffic either and the computations that R has to do are very quick. I just want to make the UX as good as possible. I think 4 workers max should be good enough for my purposes. In case more is needed well that's the beauty of your package, I can just reload with more workers and if should be good to go =)

[@brodriguesco]

Very cool, I'll surely give valve a spin!

[@josiahparry]

wonderful! If you're feeling spicy, it would be cool to have a Docker template using nix like github.com/JosiahParry/valve/blob/main/inst/minimal-docker/Dockerfile

[@jensharbers5620]

BTW: you can also export some models as pmml data. These can be directly imported in KNIME or Python

[@josiahparry]

im not familiar with that! Links?

[@sinan_islam]

I think if Google decided to make TensorFlow and Scikit-Learn commercial, the Python hype will burst.

[@nevinthomas5130]

Great Content!!

[@semigsedem]

nice graphics zooming in-out on your architecture sketches! What tools do you use?

[@josiahparry]

i use excalidraw.com/ and davinci resolve 18

[@joshstat8114]

This `rextendr` is a great feature for future R. Would this be added in RStudio in the future, just like Rcpp for C++ in RStudio?

[@josiahparry]

i dont think that will happen. its much like usethis which is an R package to help create the scaffolding for an r package. On the other side is the rust components which uses extendr. There's not anything else you'll need!!

[@joshstat8114]

@@josiahparry are there any documentations of Rust in R aside from what I found in GitHub?

[@sinan_islam]

This project sounds like polars?

[@josiahparry]

Entirely different! This is not a data frame library. This is about scaling existing plumber APIs using Rust. So if you can write a function in R you can make a REST API, if you can do that, you can deploy it to production as a backend web service or as a serverless function. Or whatever you want! Valve helps make sure that you use all available resources on a container or machine to scale that effectively .

[@jmartindelasierra]

Can this be used to achieve concurrence in Shiny?

[@josiahparry]

Unfortunately it cannot. I hope to be able to do this some day. But i don’t know enough about how shiny works to concoct it myself