Feature suggestion: Allow extracting more fields than are used for renaming

It might improve accuracy if we could instruct the LLM to extract several fields and then use only a subset for renaming.

For instance, extract: First_Author_Last_Name, First_Author_First_Name, Year, Title, Subtitle, First_sentence_of_the_first_paragraph, Header, and then use for naming the file only: [YYYY]_[First_Author_Last_Name]_[Title].

This might improve accuracy. In my sample of PDFs LLMs struggle with extracting Title when a subtitle or a name of the series is close to the title (incorrectly inferred title 20-30% of the time).

If I am not mistaken, currently the “unused” fields to extract are being automatically removed from a template.

🧑‍🔬