Hey, I have a question for which I'm looking for best practices. If answers are in the docs, I didn't find them.
How to input my data where I have questions and expected answers, which are different for each question? Ideally I'd load them from files / folder structure, without entering each path to the yaml. The CSV input seems promising. I can also generate the yaml programmatically. But maybe there is built-in way?
I am testing a RAG pipeline. In one file I have a prompt that I ask from a custom provider that implements RAG. I have instructions for scoring the response in another file, which I run through a scoring model. How to best specify this when the input and output pairs are always different?