Show HN: Shell-AI, run shell commands with natural language (opens in new tab)

Hello HN! I know this project is a super simple wrapper around LangChain/OpenAI but I just found myself wanting this badly myself: a super simple `pip install` package that I can use to get command suggestions within the terminal as I'm being productive doing other things.

Example usage:

`shai git diff but without the big json blobs`

`shai 'copy everything in data/ via ssh to $SSH_HOST using scp'`

The implementation is literally one short glue of LangChain and InquirerPy for interactive CLI.

I'm curious which ideas you all have to make this smarter/better. MIT licensed, if you're keen on contributing please feel free to do so. It's a pure hobby project for me.

Some key objectives: never automatically run shell code, I want to see what I run before I run it, present me with some alternatives, a simple path to using local models in the future (Llama 2 Code soon?).

Will add I was inspired by the great https://github.com/gorilla-llm/gorilla-cli project, but didn't like that it sends the prompt to some IP based endpoint the author presumably runs.