Andrea Arcangeli

You should be able to reproduce the reversal with C-c C-c if you first delete an entry with "d". After C-c C-c commits the deletion of an entry any further...

Another detail relevant to this pull request, gptel-add-file does not add the absolute path name without this commit if the path picked is ~/.. however buffer-file-name in my hooks above...

You can ignore the email and only look at the PR. I did the PR after a few days because I wasn't sure if the git-send-email went to the gmail...

Hello, I probably didn't flag this properly but this is a regression fix so it's technically higher priority than other optimizations. Without fixing this bug ripgrep-edit.el doesn't work, but it's...

Sorry, I didn't notice I could select oneshot presets with @preset-name in the prompt. Another way I thought of was to switch back to the orig buffer during the callback...

Hello, what's the status of the fix? I think the gptel-rewrite-directives-hook working again is more important than the @preset feature I didn't know about. Perhaps you shall consider dropping the...

Thanks for the update! I may have misunderstood, I thought the cons of evaluating the gptel--rewrite-directive before switching buffer was purely that @rewrite-extra couldn't be used in the minibuffer anymore,...

Hi @karthink, Thanks for the thoughtful responses and for taking the time to consider this carefully. I wanted to share that I’ve found a working workaround for the longstanding regression...

> @aagit wait, I didn't know that `gptel-rewrite-directives-hook` is non-functional. This is different from the above bug report, that the directive expansion happens in the wrong buffer: > > >...

More details: to reproduce it's sufficient to use buffer-name in your hook. The whole point of the fix is to run the hook *before* changing buffer name... so it is...