GenAI.Media.Request (GenAI Core v0.3.2)

A media-generation request (ADR-016). The consumer facade GenAI.generate_media/2 takes one of these, the Router picks a provider, and the provider's generate_media/2 fulfils it.

prompt - the generation prompt: a string, or content parts (incl. an image part
```
         for image+text -> image edits).
```

output - the target modality (GenAI.InferenceProviderBehaviour.modality):

         `:image | :speech | :music | :sfx | :video | :document | :text`. Most
         media generation outputs non-text; `:text` covers transcription
         (speech -> text), which also rides this path.

input - optional explicit input modality list (e.g. [:speech] for

         transcription). When set, the Router uses it verbatim instead of
         inferring from `prompt` — needed when the source media rides in
         `settings` (e.g. audio bytes) rather than in the prompt.

provider - optional explicit provider (module or config-key atom); nil => the
```
         Router picks by declared capability.
```
model - optional model override (e.g. "gpt-image-1").
settings - provider-interpreted knobs (size, quality, voice, duration, audio, ...).
api_key - optional per-request key; providers otherwise read their config/env.

Summary

Types

modality()

t()

Types

modality()

@type modality() :: GenAI.InferenceProviderBehaviour.modality()

t()

@type t() :: %GenAI.Media.Request{
  api_key: String.t() | nil,
  input: [modality()] | nil,
  model: String.t() | nil,
  output: modality(),
  prompt: term(),
  provider: module() | atom() | nil,
  settings: map()
}