allow passing -1 max to automatically set it

fix 'no instruction' check

update example

allow reading/writing to file

fix token estimation

output the length of the code

fallback to returning raw response

add code fence extraction

shorter flag names

change the default temp to 0.6

use gpt-4

switch to use gpt-4

remove pointless conversion

check if there's anything being piped in before reading

default top_p to 1