REPCO is translated (batch) for the vector search DB

we understand messy the repco data is and what to do about it (where to clean up what and what to leave as-is)
we translated all of repco (ContentItem.{content,title,summary}) into the reference language
we added the translation in the agreed upon format (for the columns title, summary, content.)
we added the translation to an easier view or table to select it easier
we did some sanity checks on the translations (for example: first check for error codes. Then the number of tokens original vs. number of tokens translated . Or benchmark on how well it was translated, or cross check selectively with DeepL (some samples), and/or eye-balling the translations manually)
we did a short write-up document on the translation
we have a function which can translate new texts on the fly (RESTful API which can be easily integrated into Transposer)
Have a cron script which a) checks the new contentItem rows which are not translated yet b) creates a batch file with those c) runs the batch d) queries the results and e) report back any errors (where?? (initially monitor via email alerts)) f) stores valid results back into repco and mark these rows as translated.