Ramalama Container needs updating on the quay.io to use new llama-simple-chat

[bmahabirbu]

Hey @ericcurtin I was testing out the new changes and I noticed the ramalama container file on quay.io needs to be updated to have llama-simple-chat

Everything worked when I built the container from scratch. Thanks!

[ericcurtin]

Yeah we really need to automate our container builds somehow.

I'll be away for the next while...

@rhatdan can we hold off on new pip releases, etc. until we have rebuild and push all our containers?

llama-simple-chat binary will probably change again to be ramalama-core or some other name. I'm discussing with llama.cpp folk how we are going to do that.

[ericcurtin]

Long-term we probably have to version all our containers:

ramalama:latest (basically rawhide, expect breakages)
ramalama:0.1.0
ramalama:0.1.1
ramalama:0.1.2
etc.

The other thing to worry about is cleaning up the old container images when one upgrades.

[debarshiray]

Yeah we really need to automate our container builds somehow.

We have been doing this for a while for the built-in Arch Linux and Ubuntu images for Toolbx:
https://github.com/containers/toolbox/tree/main/.github/workflows

We rebuild the images every Monday, even if their sources haven't changed, and we rebuild them when their sources have been changed.

[rhatdan]

Care to open a new PR to add support for this on RamaLama?

[debarshiray]

It was @travier who did all the heavy lifting and I still have one Podman pull request that I should wrap up before I can get to this one.

In short, you can start by copy pasting one of the YAML files from https://github.com/containers/toolbox/tree/main/.github/workflows and replacing the specifics, like the paths to the image sources, etc.. QUAY_ROBOT_TOKEN is added to the GitHub project's settings under secrets and variables. I forgot if there was something that needed to be done on Quay, but I am sure @travier will remember.

In short, it's simple, but one needs to devote some time to iron out the annoying details, and I am struggling to find the time to own that work. :/

[debarshiray]

Hey @ericcurtin I was testing out the new changes and I noticed the ramalama container file on quay.io needs to be updated to have llama-simple-chat

The original problem reported here went away because the change to llama-simple-chat was reverted:
#477
dec1738

[ericcurtin]

We will move to llama-run next time I guess. That was recently upstreamed to llama.cpp . We need to version our images at some point. Having only "latest" images really restricts our ability to make breaking changes.