Its possibly use ollama as backend instead chatGPT ?

[kascesar]

Hi, i have been use gptel with ollama, this package is based on gptel if i knew well, so its possible use ollama as backend ?

I've tried this, but not working:

  (use-package magit-gptcommit
    :ensure t
    :demand t
    :after magit
    :bind (:map git-commit-mode-map
                ("C-c C-g" . magit-gptcommit-commit-accept))
    :custom
    (magit-gptcommit-llm-provider (make-llm-ollama
  				 :host "localhost:11434"
  				 :chat-model "qwen2.5-coder:1.5b"
  				 :embedding-model "qwen2.5-coder:1.5b"))

    :config
    ;; Enable magit-gptcommit-mode to watch staged changes and generate commit message automatically in magit status buffer
    ;; This mode is optional, you can also use `magit-gptcommit-generate' to generate commit message manually
    ;; `magit-gptcommit-generate' should only execute on magit status buffer currently
    ;; (magit-gptcommit-mode 1)

    ;; Add gptcommit transient commands to `magit-commit'
    ;; Eval (transient-remove-suffix 'magit-commit '(1 -1)) to remove gptcommit transient commands
    (magit-gptcommit-status-buffer-setup))

[douo]

Of course, it's possible! Could you provide more details about the error you're encountering? If you're using the default configuration, try removing the :host parameter. Additionally, note that :port should be configured separately and cannot be included within :host.

I tested the same model locally, and it works fine. However, for better results, you might want to adjust the prompt by modifying the magit-gptcommit-prompt variable.