Currently, every time a character is entered into the input field, the token calculator restarts a full calculation cycle—note, for every single character entered.
This leads to an issue: even in newly created conversations, after just a few rounds of simple dialogue, when token usage approaches 40,000, inputting characters becomes severely laggy. It becomes completely impossible to type at a normal, fluid pace.
Could the calculation frequency be reduced, or could computation be deferred until submission?
(Is token calculation truly necessary during the input phase? 
)
