Articles keep their authors, tags and abstract

Creating an article and supplying its metadata — who wrote it, how it is tagged, its abstract, its thumbnail — could end with all of that missing. Nothing reported a problem: the article was created, it opened, and the absent metadata looked exactly as if the author had never filled it in.

The cause was a mismatch in how the Article type was described. A content type declares the shape of what it holds, and the platform trusts that declaration when it needs to interpret stored content. Article's declaration named a shape that nothing actually used — one that had no place to put an abstract at all. So on the way back out, the content was reinterpreted into that shape: the abstract had nowhere to land and was dropped, and the rest no longer matched what the article's own views were looking for.

It only bit content created directly — through an assistant, an integration, or the API — because articles written as files carry an explicit note of their shape, and that note took precedence. That is why existing articles look right and the problem stayed hidden.

The declaration now matches what articles really contain, so the four fields survive being written and read back.

Alongside the fix, the platform no longer stays quiet when this happens. Whenever stored content is interpreted into a shape that cannot hold part of it, that is now reported — naming the fields that could not be carried and why — instead of returning a plausible-looking result with values silently missing. Content whose type is simply an older version of the right one is unaffected: those extra fields were already preserved and still are.