Add support for unwrapping jsonpath array output / support jq

[mgcrea]

Describe the feature or potential improvement

I am playing a bit trying to replicate the OpenAI fine-tuning guide using Langfuse, my dataset is:

"{""ticket_text"":""My monitor won't turn on!""}","{""correct_label"":""Hardware""}"
"{""ticket_text"":""I'm in vim and I can't quit!""}","{""correct_label"":""Software""}"
"{""ticket_text"":""Best restaurants in Cleveland?""}","{""correct_label"":""Other""}"

So when configuring an LLM-as-a-judge evaluator I have a prompt looking like:

Task:
Compare a model output string to a ground truth

Ground truth: {{ground_truth}}
Model output: {{model_output}}

And I bind ground_truth to a "Dataset Item > Expected output" with a $.correct_label JsonPath, however since jsonpath always output arrays I'm left with a clunky:

Ground truth: ["Other"]
Model output: Other

That won't match and need to workaround by adding normalisation rules like:

Normalize BOTH sides:
- trim whitespace
- lowercase
- remove surrounding quotes if present
- if the ground truth looks like ["label"], extract that element
- then compare the normalized strings

Which works but isn't great!

Adding a checkbox to unwrap the array to a string, or better yet directly support jq syntax which is more expressive would be nice.

Tried $.correct_label[0] but it returns the first letter, still wrapped in an array.

Thanks!

Additional information

No response