Native fields, shown as they are
Wherever the engine answers, the clients show Native’s own field names and values, unchanged: engine_version, native_directory_format, product_api_version, visible_csn, root_digest, directory_lineage, checkpoint_digest, max_sql_rows… Nothing is renamed, rounded or interpreted. Two consequences:
- Anything you see in the Studio can be reproduced with
curlorhcloudand compared byte for byte. - The Cloud never becomes an authority over the engine’s data by paraphrasing it.
HYPRSP01 and the one kind that is decoded
Section titled “HYPRSP01 and the one kind that is decoded”Native’s /v2 speaks a binary envelope, application/vnd.hyphae.product-v1 (HYPRSP01). The Studio, hcloud and @hyphae/cloud decode exactly one kind: kind 1, capabilities, using the layout published in Native’s contract. The decoded table shows the engine’s field names; the raw bytes stay one click away. Any other kind — including /v2/sql answers — is shown as bytes or refused, never guessed at.
That is also why there is no SQL editor: /v2/sql accepts only Native’s binary request envelope, and the Cloud does not invent a grammar on top of it. Use the official Native SDK through the Cloud proxy for data work (@hyphae/cloud exposes data(projectId, { key }).raw(path) for that).