Parser for both long and short forms of the AI Attribution spec (and nonstandard uses)

[MoralCode]

This is probably a more involved project, but we should have a parser for https://aiattribution.github.io/

I have seen at least one project place these attributions in their README

the spec-defined formats are in the link above. I have seen examples like the following used as well:

AI Attributuion: AIA EIA Hin R Claude Code v1.0 (Claude Sonnet and Opus)

AI Attribution (AIA): EAI Hin R Gemini v1.0

looking at the spec, I suspect we could parse these just by ignoring any text before AIA, any symbols (parenthesis, colons) within the AIA, and everything after.

Maybe we could also capture either the raw line, or just the text before/after and store it in an additional field, such as comments/context as it may contain specifiers for specific models, or even specific works within a larger repository that may be AI generated

[omkar-foss]

@MoralCode are you or anyone looking into this one, if not may I pick this up?

[MoralCode]

I'm not currently working on this, so go for it!

Right now I'm not sure if we have a use for the parser, but I think itll be very helpful for input/output of attribution information in the near future as we develop this whole project more.

[omkar-foss]

Got it, I'll look into this one. Thanks.

[omkar-foss]

I checked this out, found that the aiattribution project isn't active since last few months and was released as a part of IBM's research last year, but unfortunately the attribution method hasn't gained much practical traction as Assisted-By is now becoming the standard.

The format is certainly useful (AIA Ph SeCeNc Hin R v1.0 captures great amount of info), so I suppose we could discuss in what exact way we'll use this for attribution info I/O before proceeding, because it'll need maintenance if there happen to be future versions of the format spec :)

[MoralCode]

I have seen this used in the wild in a few repos, although mainly for citing images generated by AI, rather than commits.

My main idea for one way we could use it is as a standard "canonical" way to capture AI content and include it in such a way that tools like CollectOSS can gather that data effectively. Using AIA seems like it would also have baked in support for providing information about the model and version being used (per #29 ) but we would need to consider how various AIA outputs would map to our tooling (should a maintainer decide they want the action to automatically post a comment with the AIA result from a PR or issue).

Open to thoughts on whether another format for metrics tools to read various metadata about AI that we generate would be better.