Skip to main content

MultimodalModel

lc_core::language_models

Trait MultimodalModel 

Source 
pub trait MultimodalModel:
    BaseChatModel
    + Send
    + Sync {
    // Provided methods
    fn transcribe<'life0, 'async_trait>(
        &'life0 self,
        _audio: AudioContent,
    ) -> Pin<Box<dyn Future<Output = Result<String, MultimodalError>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait { ... }
    fn generate_speech<'life0, 'life1, 'async_trait>(
        &'life0 self,
        _text: &'life1 str,
    ) -> Pin<Box<dyn Future<Output = Result<Vec<u8>, MultimodalError>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait { ... }
    fn generate_image<'life0, 'life1, 'async_trait>(
        &'life0 self,
        _prompt: &'life1 str,
    ) -> Pin<Box<dyn Future<Output = Result<ImageContent, MultimodalError>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait { ... }
}
Expand description

Multimodal model trait — extends BaseChatModel with audio, speech, and image generation.

Providers implement only the methods they support. The default implementation of every method returns an explicit Err(MultimodalError::Unsupported(..)) — this is a hard capability limit reported as an error, never a silent no-op or placeholder output. Callers MUST match on MultimodalError::Unsupported (or propagate via ?) before assuming the operation produced meaningful content.

§Example

ⓘ
use lc_core::language_models::MultimodalModel;

// Transcribe audio
let transcript = llm.transcribe(AudioContent::from_url("https://...")).await?;

// Generate speech
let audio_bytes = llm.generate_speech("Hello, world!").await?;

// Generate image
let image = llm.generate_image("A cat wearing a hat").await?;

Provided Methods§

Source

fn transcribe<'life0, 'async_trait>( &'life0 self, _audio: AudioContent, ) -> Pin<Box<dyn Future<Output = Result<String, MultimodalError>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait,

Transcribes audio to text (Speech-to-Text).

Returns the transcribed text.

Source

fn generate_speech<'life0, 'life1, 'async_trait>( &'life0 self, _text: &'life1 str, ) -> Pin<Box<dyn Future<Output = Result<Vec<u8>, MultimodalError>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Generates audio from text (Text-to-Speech).

Returns the raw audio bytes (format depends on provider, typically MP3 or PCM).

Source

fn generate_image<'life0, 'life1, 'async_trait>( &'life0 self, _prompt: &'life1 str, ) -> Pin<Box<dyn Future<Output = Result<ImageContent, MultimodalError>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Generates an image from a text prompt.

Returns the generated image content.

Dyn Compatibility§

This trait is dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety".

Implementors§