Create fedora rpms build artifacts for release#7346
Create fedora rpms build artifacts for release#7346SidShetye wants to merge 3 commits intojanhq:mainfrom
Conversation
|
Please keep the commit of changing |
…ntion)" This reverts commit e1f659a.
|
Hey there, wish I had remembered to search through PRs to find this before submitting an issue and figuring it out on my own, but I got the job done with this little hack: #4543 (comment) Out of curiosity, is this of any use, or would the script have found the libraries for Fedora's repos automatically once you added |
Code ReviewSummaryAdds Key Findings
Recommendation: fix neededAdd a CI step verifying RPM build success, test install/run on Fedora, and update release documentation. At minimum, document RPM as experimental. |
Describe Your Changes
This change creates rpm build artifacts for Fedora and related distros.
Why rpm ?
This was briefly raised at #6115 (I was absent there, just found this project a couple of hours back) but that was closed as providing flatpak as an alternative. However flatpak's sandbox and portals actually get in the way when trying to manage various custom llamas and models, while a native rpm is able to work very easily with the existing model and runtime libraries on a system.
Example:
In
~/.var/app/ai.jan.Jan/data/Jan/data/llamacpp/models/some-model.gguf/model.ymlyou might haveThose entries are very opaque, leading to a poor experience. Plus those
/run/user/1000/doc/<id>/entries are ephemeral mount points created by the xdg-document-portal service, so they are on their own lifecycle.All said and done, the native deb or rpm application run is great for power users on Linux.
Other considerations
I kept the commit to have the binary called
janas opposed toJanby default as a separate commit to allow easy picking. I personally think aJanbinary is weird on linux but so you guys can decide :) !