Adding YAML config support for 'profiles' which mean kwargs we pass into the LLM

[ndebard]

This is a small change so that the YAML workflow allows new config options for passing in kwargs (e.g. reasoning effort high, medium, low, etc). Also changed the 2 YAML examples to include some different uses of this. The workflow also does some pretty-printing to try and validate that what you sent in was actually read back / accepted by the LLM (though technically I don't think we have control over what the provider ends up using). This just helps give you some validation that you got what you asked for. Example:

[ndebard]

don't merge yet, i've got a bug i'm fixing.

[ndebard]

@mikegros and @luiarthur and @awadell1 - this small PR is ready for review. Quick quest - we're having this CI/test fail b/c it doesn't have an API key. What's our solution to this?

[mikegros]

@mikegros and @luiarthur and @awadell1 - this small PR is ready for review. Quick quest - we're having this CI/test fail b/c it doesn't have an API key. What's our solution to this?

I'll pull and run the tests locally before merging.

If you want it to run as CI, you can set an API key as "Secret" on your fork and use that key. That is how they work on the main repo. But forks cannot access it directly (because someone could just fork and put in a pull request that prints the key and exfiltrate it).

[awadell1]

The easy fix would be to push your branch to the LANL repo (instead of a fork), but that would require a rejiggering of the repo config (i.e. adding branch protection rules for main) and giving contributors push access

The other option would be to switch to environment secrets, but that would require approval before each workflow runs (to prevent the exfiltration issue above).

[mikegros]

Just waiting on our debugging of the dangling tool errors you saw @ndebard. Should merge this soon though.