Incorrect author attribution in preset prompts library

[danbarr]

Describe the issue

We appear to be mis-attributing the prompts in our preset library sourced from awesome-cursorrules.

The actual rule author's name is in the README.md file that accompanies every rule. But we appear to be sourcing from the last committer(s) to the file itself, which often is not the same person.

For example: the python-containerization-cursorrules-prompt-file has the author noted as Chakshu Gautam. But in our UI, we're showing PatrickJS who authored the last commit just making some formatting changes.

Steps to Reproduce

Compare the author listed in the source repo with our displayed author in the dashboard.

Operating System

MacOS (Arm)

IDE and Version

NA

Extension and Version

NA

Provider

Anthropic

Model

NA

Codegate version

v0.1.16

Logs

No response

Additional Context

No response

[kantord]

This one is tricky. What we can easily programatically extract is the metadata from the git commit. Even this is limited, because associating it with a GitHub user is based on some heuristic.

So if we want to be technical, we could say right now this data shows all the GitHub users who uploaded/edited these files to the repository, not necessarily who wrote the original text, so cannot be considered a 100% accurate author attribution, but (almost) perfectly matched what you would see in a git blame view.

How to fix this? Ultimately I think that the only way to fix it is to have a proper data source that requires the cursor rule to be attributed to a GitHub user in a structured data format. So I think we would have to fork the original repo (the license allows it), and make these changes (such as adding a metadata.json file for each rule).

But even in this perfect solution, if there was no GitHub account associated with the credited author, we cannot add one, even if we could find the author's GitHub profile, since the author might not wish this to happen. (Unlikely that anyone would be too bothered by this, but still it's a possibility. That's a bigger problem than not having the "proper" attribution, because actually the license does not require any attribution)

[kantord]

I guess one thing we could do is replace the "attribution" (which is more meant as "committer"/uploader anyways) with a "Readme" link when a Readme file is present. This way we don't risk involuntarily suggesting that the prompt was created by X, when the readme says it was created by Y.

This could be a simple solution to an otherwise difficult problem.

What do you think @jtenniswood ?

[jtenniswood]

Makes sense to me.