* feat: process pdfs synthetically im model does not support file natively * fix: small code optimisations