rx is a binary that you can use as a prefix to any other command (e.g., `rx ls` instead of `ls`) to run a command on the cloud. It sets up your project on the cloud and transparently copies what you're working on back and forth from your local machine.
Background: I was trying to build an app that used stable diffusion and got very frustrated by what should have been simple: I wanted to develop my program locally, but run it on a different machine. Setting up, copying over, and running the program ended up being an exercise in frustration (Careful documentation of the process: https://blog.run-rx.com/cloud-development-the-board-game).
To make life easier, I made a simple binary to copy everything to the cloud and back. Then I added streaming stdio and made it keep dependencies up-to-date and automatic installation of the languages the project uses and on and on.
Current state: give the CLI a project and it'll set up a persistent remote environment that is synced to your local environment (all dependencies installed and automatically kept up-to-date). This lets you develop locally using any tools you like, while still running remotely for anything where you don't want using your CPU.
Please give it a try, I'd love any and all feedback!
Looks very cool! Love the conceptual simplicity of it. I’m wondering if I develop in Python on my MacBook and use a library like PIL (Pillow) which on my Mac is ARM-specific, will rx install the Intel version in the cloud?
Thanks! Yeah, it takes a look at your requirements.txt (or environment.yml) and uses that to install on the remote machine. So as long as that doesn't specify a particular arch, it'll chose whatever it's running on.
I'm hosting the machines so that users don't have to set anything up on the cloud side. It could have on prem or open source versions if people are interested in setting up their own clusters.
So far I'm implemented pretty coarse-grained control over machine specs: it's $20/month for a MacBook-like machine and $3/hour if you want to use "a GPU" (whatever Amazon has available for cheap). I figured the config could specify exact disk/ram/instance type requirements in the future, it sounds like something you're interested in?
Background: I was trying to build an app that used stable diffusion and got very frustrated by what should have been simple: I wanted to develop my program locally, but run it on a different machine. Setting up, copying over, and running the program ended up being an exercise in frustration (Careful documentation of the process: https://blog.run-rx.com/cloud-development-the-board-game).
To make life easier, I made a simple binary to copy everything to the cloud and back. Then I added streaming stdio and made it keep dependencies up-to-date and automatic installation of the languages the project uses and on and on.
Current state: give the CLI a project and it'll set up a persistent remote environment that is synced to your local environment (all dependencies installed and automatically kept up-to-date). This lets you develop locally using any tools you like, while still running remotely for anything where you don't want using your CPU.
Please give it a try, I'd love any and all feedback!