How does Ossie think about unstructured data?

[Yuki-Matsuyama-a]

I've been reading the current core specification and wanted to clarify the intended scope.

The current model seems primarily designed around structured datasets, fields, relationships, and metrics.

How does the project currently think about unstructured data such as documents, transcripts, or other content used by AI systems?

Is the expectation that these remain outside the core model for now, or could they eventually be represented through extensions or a dedicated resource abstraction?

I'm mainly trying to understand the intended boundary of the specification rather than propose a concrete change at this stage.

[flyrain]

Your reading is right: the current core spec mainly targets structured data. #238 explores a bit on unstructured files and AI context(5.2.2), while #338 proposes richer sources for structured file-backed datasets. Technically, extraction will be needed before the unstructured data is usable.

My take is that the community focuses on structured data now. Would you mind providing more details of the motivation? Especially, solid use cases will be helpful.