Get File

Metadata for one file — OpenAI's "Retrieve file" / Anthropic's "Get File Metadata". Ownership-scoped like every other read here (wrong member, wrong kind, and unknown file_id are indistinguishable 404s), but reading metadata is not a "use": unlike GET /files/{file_id}/content, this never mints a signed URL and never bumps custom_time/last_used_at.

Reports pending files too (state visible), same as GET /files's list — a client's "all files" view needs a single-file fetch that behaves like the list it paginates, not a narrower one.