Merging gptel branch and main branch #29
Conversation
Simplify response handling with gptel-request and gptel-fsm
|
I have tested this with |
|
Forced pushed with a cleaner history and minor cleanup |
|
Before merging this one I would like for #28 to be reviewed since it's such a huge change that fixes critical bugs. |
|
Any luck on this? |
It works perfectly for me and @douo hasn't followed up. Please check it out and post what you think. I may merge it if you don't encounter any issues. |
|
Thanks for you quick response, I will try this PR on my machine. |
Oh my bad! I thought this was a comment for the other PR. |
Aha, no concerns, I'll proceed with this PR nonetheless. |
The douo#28 PR introduces some new functions and adds extensive debugging. Move the common function and adding appropriate debug calls within gptel functions as well.
|
Before merging, can you guys check if this works as intended? |
(use-package magit-gptcommit
:load-path "~/misc/magit-gptcommit/"
:bind (:map git-commit-mode-map
("C-c C-g" . magit-gptcommit-commit-accept))
:commands (magit-gptcommit-generate)
:custom
(magit-gptcommit-backend 'gptel)
:hook (magit-status-mode . 'my/enable-gptcommit)
:init
(defun my/enable-gptcommit ()
(interactive)
(magit-gptcommit-mode 1)
(magit-gptcommit-status-buffer-setup)))This is config I tested, and this PR works totally fine for me! |
Require gptel-request and introduce a local handler alist so magit-gptcommit can merge its custom gptel handlers with the handlers provided by the gptel-request module. Instead of hardcoding the handler list, compute the unique set of handler keys from both alists and assemble a combined handler entry for each key, preserving magit-specific handlers while reusing gptel-request handlers.
5 participants
This PR merges the separated gptel branch with the main branch. To support both
llmandgptel, this also adds a customizable variablemagit-gptcommit-backendwhich is expected to be eitherllmorgptel. It also removes the respectiverequirecalls, as this would expect the user to have one of the packages configured (seemagit-gptcommit--ensure-backend-loaded). I have also updated the readme accordingly. Let me know what you think.Note: to keep up with the main branch, I had been cherry-picking commits and applying them to a branch maintained in my fork. Hence why the merge history shows some of the commits already merged to master.