Chat Output thinking Poll

[dixonSolutions]

Chat output thinking comes natively into to the chat, built in for the model, but harnesses expect thinking chunks, so should we package the chat thinking to the harness as thinking chunks, if the harness already expects them to provide it? A lot of our issues have been mentioning it, so we want your opinion to settle the beef. Also some of our issues mention the chat thinking interfering with responses, that needs a fix regardless.
So pretty much regardless, actual raw output, text or tools calls will need to be separate or else it interferes, especially for initial session title #9