This package defines a set of functions and variables for running
instruction-following language models like GPT-4 and Claude 3.5
Sonnet.  It allows the user to enter a command with history and
completion, and optionally use the current region as input.  The
output of the command is displayed in a temporary buffer with the
same major mode as the original buffer.  The output is streamed as
it is produced by the GPT process.  The user can enter a follow-up
command in the output buffer, which will provide the output, the
follow-up command to GPT as a new prompt.  The follow-up output
will be appended to the output buffer.  The user can view and
export the command history to a file.