Extract audio bytes from OpenAI-style content parts, rewriting the parts
to the template's {type:"audio"} form. Accepts the OpenAI-canonical
{type:"input_audio", input_audio:{data:, format:...}} plus optiq's
{type:"audio"} (base64 data) and {type:"audio_url"} aliases.
Extract audio bytes from OpenAI-style content parts, rewriting the parts to the template's {type:"audio"} form. Accepts the OpenAI-canonical {type:"input_audio", input_audio:{data:, format:...}} plus optiq's
{type:"audio"} (base64
data) and {type:"audio_url"} aliases.