brauliogusmao steadies streamed AI answers before polishing them
The fork stops a live assistant reply from breaking mid-stream, then applies richer formatting once the answer is complete.
When an assistant answer arrives word by word, rich formatting can sometimes destabilise the message on screen. brauliogusmao's fork takes a deliberately conservative route: it shows the incoming response plainly while it is still arriving, then refreshes it with the full formatting once complete.
That means tables, mathematical notation and other Markdown niceties appear after the stream finishes, rather than competing with a changing live response. The tradeoff is a brief delay before the polished version appears, but the practical gain is a more dependable chat experience without disappearing or broken messages.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?