Is it possible for the detection to distinguish between a line break and a para break with a bit of intelligence? Encountering a larger spacing between lines can be treated as a para break, else line break. This would be invaluable when dealing with documents that have multiple paragraphs of text. The current approach of sending a \n (or \n\r) obviously causes reflow problems due to the hard breaks.
These can of course be deleted after the API response is received, but that’s even worse off since then it would result in 1 confusing chunk of text.