Update prompts / outputs for models to not use JSON format to prevent weirdness with JSON escaped characters.

After executing a command using developerGPT, I would want to see it behave as if I had manually written and executed it myself.

Specifically, I would prefer to have the executed (generated) commands entered in .zsh_history/.bash_history as well so that they show up when I type history in my terminal or when I perform reverse-lookup (aka bck-i-search) using Ctrl+r

This will make it more convenient to reuse and recall already executed commands (with slight changes for subsequent use cases) without invoking developerGPT and reentering the prompt, and incurring the usage on the API again

Is your feature request related to a problem? Please describe.
As a power user of DeveloperGPT, I would like to be able to directly type my question into the command line instead of waiting for the prompt

Describe the solution you'd like
Command line integration with commands to execute (ex developergpt cmd "install anaconda")

This is an amazing project! Can't wait to see what's next.

Improve DeveloperGPT cmd response speed by having the model first send over the commands, then send over the explanation after (so we can display the commands much faster)

Currently 'tab' is used to cycle between tab completion options. Should change this to 'tab' to accept completion so the 'enter' key does not also trigger the command to be run