Enable galaxyAI on qa1

[ljocha]

Done by adding inference_services to galaxy: section of galaxy.yml, according to galaxyproject/galaxy#21434

Ansible secret file requires something like:

chatgxy:
  model: agentic
  token: sk-XXXXXXX
  endpoint: https://llm.ai.e-infra.cz/v1

I did it for QA2 (where it works now), I can't find the valut password for usegalaxy.cz now ... Kindly asking @martenson or @martindemko to add it.

The token can be the same as llm_hub, but it needn't be the same, hence introducing another config entry.

[martindemko]

I think we should reformat this because, as it is now, all our instances will expect to have these Ansible variables. If we want to enable this functionality on QA2 and GCZ, it should be moved to host_vars/INSTANCE/vars.yml. Otherwise, the playbook won't be compatible.

[ljocha]

I'm afraid that due to the way Ansible works, you would have to move the whole galaxy_config: from group_vars/galaxyservers.yml to host_vars/..., which you don't want.

But you are right things get broken now, IMHO we need placeholders like

usegalaxy/group_vars/galaxyservers.yml

Line 49 in 1bb1375

llm_hub:

, don't we?

[martenson]

we'd make csnt_inference_services and propagate through that, see existing examples with the prefix in the file