Feature Request: AI generate commit message with custom model provider, and allow to edit the system prompt

[patrick-fu]

The new commit message auto-generation feature is great!

However, I have noticed that currently only OpenAI is available as a provider. I hope to introduce an endpoint URL parameter and a model parameter, allowing users to switch to any model provider which compatible with the OpenAI API format and allow to choose different models

Additionally, I hope the fork app can also allow to edit the system prompt used for generating commit messages. Different projects may require different commit message formats, therefore, it is better to read the system prompt from a specified configuration file, for example, first reading from the .fork/xxxxx.json file in the project's root directory, if not found there, search in the home directory; if still unavailable, use an editable prompt in the fork app settings

[fredli74]

+1 on this. I use Fork on proprietary code where sending the source to external providers is out of the question.

I use this on a daily basis, as a custom command in Fork, but as it's not integrated nicely with the GUI, I can't see the message before committing. I have to go back and amend if the comment is wrong instead, which is not the best workflow.

git diff --staged | ollama run qwen3 --hidethinking 'Summarize the changes into a commit message. Keep reducing until you have a short commit message, ASCII only, first line max 50 characters, wrap at 72. Only respond with the message in raw format as it is piped into the commit tool.' | sed 's/^[`"\x27]//; s/[`"\x27]$//' | git commit -F -