vercel.internal._gw_images_input
(internal) The rendered prompt lowered to the endpoint's two inputs.
Reference version
Signature
function vercel.internal._gw_images_input(
prompt: ai.Prompt,
journal: ai.Journal,
preview: bool = …
) -> vercel.internal.GwImagesInput throws ai.errors.InvalidRequest | ai.errors.NetworkFailure | ai.errors.PreviewUnsupported | baml.errors.InvalidArgument | baml.errors.Io | baml.errors.Timeout
(internal) The rendered prompt lowered to the endpoint's two inputs.
Text: every text part, trimmed, blanks dropped, joined with a blank line —
sys_llm's prompt_to_text, roles discarded. Journal user messages follow so
a re-ask still reaches the model.
Files: image parts become files[] entries, which is how this endpoint does
image editing / variations (image-model protocol v4). The other media kinds
have no representation and are still rejected.
Source:<builtin>/vercel/ns_internal/images.bamlbytes 6588–8286