A provider-neutral image or video generation request.
This is a separate request from LlmCall.Request on purpose: dedicated media
endpoints take a prompt and media parameters, not messages, a system prompt, or
tools. Flattening a completion request into {prompt, n, size} would be a hidden
convention.
references are images or videos the generation is conditioned on. An
adapter may still refuse a source it cannot carry (a {:base64, _} image
without a media type, a video reference to an image-only endpoint); those are
validation errors, not silent drops.
Summary
Types
@type modality() :: :image | :video
@type params() :: %{ optional(:n) => pos_integer() | nil, optional(:size) => String.t() | nil, optional(:aspect_ratio) => String.t() | nil, optional(:duration_s) => pos_integer() | nil, optional(:seed) => non_neg_integer() | nil, optional(:negative_prompt) => String.t() | nil }
@type t() :: %LlmCall.Media.Request{ modality: modality(), model: String.t(), params: params(), prompt: String.t(), provider_options: map(), references: nil | [LlmCall.Content.Image.t() | LlmCall.Content.Video.t()] }