Getting raw inputs and outputs along with token count statistics

[Mirc224]

Hi, I'm new to the instructor library. I'm doing research where I want to find out the effectiveness of different approaches to extracting structured data from unstructured texts. I would need to get the total number of input and output tokens even after doing retries.

Is it possible to retrieve the raw input that is sent to the model after it is modified by the library (entities, their fields, descriptions, types, etc. are added)? Is it also possible to retrieve it for repeated attempts? Can similar statistics be obtained for the output as well?
Thank you.