The API is returned by streaming, and sometimes the stream is interrupted due to network problems, so how does openai charge?
Only charge for what has already been produced (not including post-interruption parts)?
The cost of the entire answer (including the part after the interruption)?
1 Like