Allow to Specify Additional, Task and/or dataset Specific Resources

[mam10eks]

When we now want to make LLMs available in TIRA (such as Alpaca or other instruction-based models), we should pay attention that a software can be executed against different (especially future, simulated, or commercial) LLMs without adoption.

Therefore, Johannes had the idea that one can select a second type of resources that might be dependent on the task/dataset. I.e., that I can select in the first select box the resources on which my software runs on (e.g., "4CPU's + 40GB RAM") and in a second select box I can select which ChatNoir Chat Access token and endpoint configuration I want to have injected into my software execution. For security reasons, the ChatNoir Chat version accessible from within TIRA should only be available within TIRA, no one else (besides admins) should be able to access it.

@potthast I think a previous version of TIRA already allowed the injection of API keys if I remember correctly for the plagiarism source retrieval task?

[mam10eks]

This might be a good motivation that the public version of ChatNoir chat is not even running in our own infrastructure but within OWS?

[mam10eks]

cc @MattiWe, @kaaage

[github-actions]

This issue has been marked stale because it has been open 60 days with no activity.