The page bounces and glitches as the model types if the text box reaches the bottom of page.
Pretty much what the title says haha. Its glitchy as the model types and the whole page bounces up and down!
its down to this mfer not liking the rapid changed while parsing markdown.
https://github.com/nilsherzig/LLocalSearch/blob/1e19940ea5079fdd00e922bf7e7712f79eb9a47b/src/lib/log_item.svelte#L14
While using a slow model you can see how the markdown parser gets confused by the streaming text from time to time (for example everything becomes a heading while an empty list item exists). I think i need to implement some kind of buffer for this. I still want the streaming responses be in rendered markdown. Especially tables look very cool while getting created haha
Ohhhhh gotcha!
I still want the streaming responses be in rendered markdown.
Yes certainly! I think it's really cool too!
i really dont know how to fix the glitching, so i guess i should just debounce the page scrolling?
Thats fine!
nvm i think i got it