Repository navigation
一个单词一个单词的输出? #24
Description
Activity
The problem is that markdown parsing has to have a look-ahead. There's a way to make it not require it but certain features of commonmark has to be disabled. I had an option to do that at one point, I'll have to look and see if it's still there
Checked ... I removed that option. It'll have to be done carefully. Things like tables can't reasonably be guessed because the number of columns is unknown. Other things, like links for instance need the full url before emitting. Then the H1/H2 markdown can have ==== on the next line.
I mean the real question is what to do there? if the option is "no buffering" then I think it's just maybe disable those features? Or is the option "no buffering, I don't care about H1/H2, and you can buffer for urls and tables ... don't care."
I dunno
Here's the doc: https://spec.commonmark.org/0.31.2/
这个工具是只能凑满了一行才能输出么?可以一个单词一个单词的输出么?因为llm回复的内容可能速度比较满,终端比较宽凑满一行需要等待太久了.