Closed minad closed 1 year ago
@minad Hi!
I'm almost done with org support, I should be adding it in 30 minutes or so.
For the last couple of days I was trying to get ChatGPT to reply with org-formatted responses by using a prompt, but the results were inconsistent and non-reproducible. I guess this element of randomness is expected with LLMs.
So I wrote a basic converter for the responses instead. (Suggested markdown
-> org
solutions all seemed to involve pandoc, so I wrote a very basic parser in elisp.)
BTW the readme doesn't mention this yet, but you can use gptel in any buffer by selecting a region and calling gptel-send
, the special buffer is not needed. You can then select a region that includes your prompt and its response and call gptel-send
again to continue the conversation. (Responses are tracked using a text property.)
Org support added in 9db679dd.
You can enable it with
(setq gptel-default-mode 'org-mode
gptel-prompt-string "*** ") ;<-- or anything you'd like
Alternatively, you can directly call gptel-send
from an org-mode buffer (with a region of text selected) and it should work correctly.
Great, I will try this now!
@karthink
(Suggested markdown -> org solutions all seemed to involve pandoc, so I wrote a very basic parser in elisp.)
Why avoid pandoc
? IMO reuse of a much more complete and robust implementation seems like too enormous of a benefit.
I saw that you consider adding this. What are the obstacles? Could you use some generic regexp-based machinery to support different file types? Maybe you also need some converter for the responses, which I suspect are in markdown?