Hi all, is there an official, graceful way to convert between phoenix.client.types.PromptVersion and phoenix.evals.templating.Template?
phoenix.client.types.PromptVersion has a _dumps which allows me to get the raw template, but it's an internal method. Otherwise format only produce the final prompt with all variables provided. I can probably do something like this: prompt.format(variables={"text": "{text}"}) but that seems a bit hacky.
LLMEvaluator class receives a prompt_template parameter, so I need to have the raw prompt available.
Any help is greatly appreciated. Thank you!
Hi Xander S., that's exactly it!! my usecase is basically:
fetch a saved prompt in phoenix server
then:
either setup an experiment and pass several LLMEvaluators (such as ClassificationEvaluator) to the experiment
or use evaluate_dataframe for local experimentation
Hope that makes sense! Also I'm on arize-phoenix-evals==2.6.0 and arize-phoenix-client==1.24.0
Truong H. right now the best way is to use the _dumps method, like you mentioned. We will look into adding a utility for more graceful conversion, though!
thanks, I was going to create an issue on github but saw that you guys already did that!
