are you taking the piss.. you didn't change anything, it's still showing the same hard coded values.. did you implement a vision model to analyse the uploaded image?

it seems to be using harcoded values instead of actually analysing the image using a vision model and then defining the fieldnames based on powers of deduction

analyse an image which contains regulatory data and ghs symbols and everytihng else that is on a chemical label and put bounding boxes around everything and give them fieldnames that make sense, or come from a provided list of fieldnames, so when you see "Best Before: 04/2025" you would create two fieldnames, lbl_best_before_date which contains "Best Before:" and best_before_date which contains "04/2025" and so on.. basically everything on the label needs either a static label name ( starting with lbl_ ) or it's a variable name. I can provide the list of fieldnames and suffixes and prefixes if needed.