On Bazzite’s website they wrote “Bazzite is developed on Bazzite”, but it’s not possible to download a -dx variant as with Bluefin or Aurora. Does it mean all those tools are already included in Bazzite out of the box?
Apart from the gaming aspect, which I’m interested in, is it a good daily driver OS for development (AI, machine learning, deep learning) out of the box? Or does it need to be tweaked for it?
There is a long-standing thread that is open for comment for bazzite-dx: Open for comment: DX Images (Developer Experience) · Issue #533 · ublue-os/bazzite · GitHub
I’d imagine that people just layer or add packages via github image builder to make bazzite better for developing.
Bluefin GDX is probably better for that: Call for Testing: Bluefin GDX (Alpha)
That’s the same as bluefin-dx with a CentOS Stream base instead of Fedora, so either way there are some tweaks to perform. In that GitHub issue the dev talks about game development anyway.
In that case it’s probably easier to layer the development stuff on top of Bazzite? Slapping gaming related stuff on top of Bluefin seems much more complicated
With Bazzite-Nvidia Desktop I utilized some of the included base functions to build out the script and submit the changes to Bazzite to add the ujust bazzite-cli command. Most of the components are built into the base. I just had to install vscodium and pull the repo to begin working on everything.
For AI development I used boxbuddy to set-up a fedora-40 base in which I installed ollama using the ollama using the script provided on Download Ollama on Linux I also utilized that instance to install open-webui “GitHub - open-webui/open-webui: User-friendly AI Interface (Supports Ollama, OpenAI API, ...)” for ollama to add a web interface for my local AI’s. For this to work I open the terminal and click the carrot drop down to open one instance of ollama “ollama serve” then open a second instance using the carrot dropdown to open the python version of open-webui “open-webui serve” and then log into the AI with the web link “http://localhost:8080/” and with that I can utilize any of the AI’s through ollama with an easy to access interface.
There are a few tricks you can do like in the first instance of the boxbuddy fedora “box” to launch ollama using “OLLAMA_HOST=0.0.0.0 ollama serve” to launch it for the local network. I use this function to make the base available for other PC’s on my home network.
Most of the -dx components can be added through various methods via boxbuddy. On my actual bluefin instance I used boxbuddy to install the Zed IDE which reached out to the main computer on my network’s ollama instance to call on the coding focused AI for development tips. All the inter-connectivity is neat. I’m still exploring all my options but have had fun doing so.
For a bit of additional context.
My main PC is a desktop running Bazzite. It does my gaming and AI hosting with it’s stronger NVIDIA card.
My laptop is a Framework 16 which runs Bluefin-DX and it leans on the Bazzite machine for Game Streaming and AI management.
I have a third Windows PC which is dumb and doesn’t do much aside from stuff for my day job so I won’t elaborate on it further ;p