dense-analysis/neural

Use :Neural on visually selected words

ckangnz opened this issue · 5 comments

ckangnz commented

Is there any way to write up the question or instruction first, and then ask the chat GPT rather than writing down everything in :Neural ... prompt?

w0rp commented

I'll keep this issue open, as it's separate, but see #16. First and foremost I would like to alleviate concerns over sharing secrets/IP with third parties, and then look at other solutions regarding sending snippets of text from the editor to third parties.

  • 🚀1
Konfekt commented

A natural use would be asking ChatGPT to expand a visually selected comment, say describing a function, into the actual code (template).

w0rp commented

A natural use would be asking ChatGPT to expand a visually selected comment, say describing a function, into the actual code (template).

Yes, this is a good use of that functionality. Bear in mind our requirements for privacy outlined in #16.

Konfekt commented

In the meanwhile, a mapping such as nnoremap <leader>n :<c-u>Neural <c-r>* would do.

  • 👍4
w0rp commented

I have implemented :NeuralExplain for visual mode, which sends visually selected text to the selected data source and prints the explanation in the preview window. I've put in nearly bare minimum effort to redact passwords and secrets, as explained in the README. More work can be done on redacting passwords and secrets.

#31 will cover the use case of sending lines to an LLM that can be refined in a scratch buffer.

  • 🎉1