tiacode / gptkitten

ChatGPT Kitten for kitty terminal

Access ChatGPT from your kitty terminal with terminal text sent as data. Ask questions or use autocomplete.

Features

Chat

Autocomplete

Usage

• Autocomplete: This needs openai python library to be accessible to your kitty, so run pip install --user openai. Edit your OpenAI API key to gpt_kitten.py and copy it to to your kitty config dir (~/.config/kitty/). Add config line map kitty_mod+i kitten gpt_kitten.py to your kitty.conf (change the keys though.) You can add parameter -e to select what data will be sent from the screen - default is "selection,lastcmd,screen" which basically tries to get your selected text, and falls back to output of last or currently running command that isn't empty.
• Chat: either create venv with OpenAI python library or install it userwide like with autocomplete. edit OpenAI api key to kitty_assistant.py. Add line map kitty_mod+x launch --copy-env --hold --stdin-source=@selection /usr/bin/python /path_to/kitty_assistant.py to your config (or use the venv python binary). You can choose what terminal data is used with --stdin-source - @selection, screen, @last_cmd_output, @screen_scrollback are good options, others can be found at kitty docs. You can also skip that argument if you want to just chat without terminal data. Press your chosen shortcut and start chatting.

Drawbacks

• Code is absolutely horrible, I used a hammer and made this in like five different ways before coming to this solution, I can't be bothered to make it clean for a few days at least. Bugs are likely, pull requests are welcome.