3 ms·

I would also love to see a token budget use for the chats -- to know when the model is about to run out of context. It's crazy this is not there.

• joquarky 9mo ago
  It would have to be intentionally vague since there is no hard cutoff threshold.
  • bspammer 9mo ago
    Claude code does it, to a precision of 1 digit of a percentage. That’s more than enough to be useful.