gptalk is a command-line grab bag of workflows I often use with LLMs.
ChatGPT <https://chat.openai.com/>__ is the only supported back-end.
GPTalk's commands can perform pre-processing. For example, if asked to summarize a YouTube video, it will quietly extract the subtitles to the video and print a summary:
|asciicast|
.. |asciicast| image:: https://asciinema.org/a/UDieN0MgIUfBBJluaiKl4aY8z.svg :target: https://asciinema.org/a/UDieN0MgIUfBBJluaiKl4aY8z
I recommend pipx <https://pipx.pypa.io/stable/>__ for managing python
modules in your interactive shell.
.. code:: console
$ poetry build && pipx install .
.. code:: console
$ gptalk --help Usage: gptalk [OPTIONS] COMMAND [ARGS]...
Do some stuff with ChatGPT.Options: --version Show the version and exit. --help Show this message and exit.
Commands:
howdoi Like howdoi but ChatGPT instead of StackOverflow.
outline Generate MECE outline of an arbitrary topic.
subsfix Fix subtitles generated with speech to text.
ticket Create a task ticket using a basic WHAT/WHY/AC format.
tldr Provide a tl;dr on a stream.
vcard Read unstructured data and output a contact card (VCF).
gptalk has various subcommands. Data is read from standard input; if
not found, your $EDITOR will be opened.
To install development dependencies, you will need poetry and
pre-commit:
.. code:: console
$ pipx install poetry pre-commit [...] $ cd gptalk $ pre-commit install --install-hooks [...] $ poetry install && poetry shell