My only concern is whether buffering the Markdown could make streamed LLM output feel slower. How does Yoke handle that?
webofbits
mudasobwa
Creator of Cure
webofbits
Got it, that makes sense. Streaming stays responsive, then the final response gets formatted. I’ll prototype that approach in Draught maybe with a plain-text fallback for terminals that cannot redraw cleanly.
0
mudasobwa
Creator of Cure
webofbits
webofbits
This is archived now. Good experiment but I won’t have the time to invest on evolving this until it gets more mature. Will try other CLI tools to run local models.
1
mudasobwa
Creator of Cure
