An interesting solution one of our engineers put forward! https://github.com/ObservedObserver/chatgpt-i18n

Hi Paul, Obviously, there are various way to simplify the translations, but the base condition is having the texts separated from the logic, etc. - should it be JSON configuration files, Unix message files, resource files, etc. Considering the variety technologies in use, one common base format plus conversion tools to different native formats for the different technologies would be probably useful (e.g. the CLI would hardly use the same format of such resource files directly, but it would still benefit from the possibility of sharing some texts and their translations with the front-end, etc.).

We’re very interesting in adding configurability here- there are questions to solve but would be a good topic for a community RFC