Skip to main content

LanguageModel

llmsdk_provider::language_model

Trait LanguageModel 

Source 
pub trait LanguageModel:
    Send
    + Sync
    + Debug {
    // Required methods
    fn provider(&self) -> &str;
    fn model_id(&self) -> &str;
    fn do_generate<'life0, 'async_trait>(
        &'life0 self,
        options: CallOptions,
    ) -> Pin<Box<dyn Future<Output = Result<GenerateResult>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait;
    fn do_stream<'life0, 'async_trait>(
        &'life0 self,
        options: CallOptions,
    ) -> Pin<Box<dyn Future<Output = Result<StreamResult>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait;

    // Provided methods
    fn specification_version(&self) -> &'static str { ... }
    fn supported_urls<'life0, 'async_trait>(
        &'life0 self,
    ) -> Pin<Box<dyn Future<Output = SupportedUrls> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait { ... }
}
Expand description

Contract every chat / completion model implements.

Mirrors LanguageModelV4. Method names keep the do_ prefix from ai-sdk to discourage direct end-user usage; downstream llmsdk crates wrap these into ergonomic helpers.

Required Methods§

Source

fn provider(&self) -> &str

Provider id, e.g. "openai".

Source

fn model_id(&self) -> &str

Provider-specific model id, e.g. "gpt-4o-mini".

Source

fn do_generate<'life0, 'async_trait>( &'life0 self, options: CallOptions, ) -> Pin<Box<dyn Future<Output = Result<GenerateResult>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait,

Run a non-streaming generation.

§Errors

Returns a crate::ProviderError when the upstream call fails, the response is malformed, or the prompt is rejected.

Source

fn do_stream<'life0, 'async_trait>( &'life0 self, options: CallOptions, ) -> Pin<Box<dyn Future<Output = Result<StreamResult>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait,

Run a streaming generation.

The returned stream yields StreamParts. Errors come in two flavors:

  • Outer Result::Err — the call itself failed before any data flowed.
  • Inner StreamPart::Error — the stream is alive but the provider reported a recoverable issue (content filter, partial failure).
§Errors

Same conditions as Self::do_generate.

Provided Methods§

Source

fn specification_version(&self) -> &'static str

Specification version (currently "v4").

Mirrors LanguageModelV4.specificationVersion (ai-sdk language-model-v4.ts). Acts as the trait-version tag for middleware composing across versions; provider impls inherit the default.

Source

fn supported_urls<'life0, 'async_trait>( &'life0 self, ) -> Pin<Box<dyn Future<Output = SupportedUrls> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait,

URL patterns the model can ingest natively, by media type.

Returning a pattern tells the SDK not to download a matching URL before calling the model. Defaults to an empty map (no native URL support).

Dyn Compatibility§

This trait is dyn compatible.

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

Implementors§