[gls2ro]

Not the OP but I think in case of scanning and tagging/summarization you can run a local LLM and it will work with a good enough accuracy for this case.