Skip to main content

LlmProvider

tkach::provider

Trait LlmProvider 

Source 
pub trait LlmProvider: Send + Sync {
    // Required methods
    fn complete<'life0, 'async_trait>(
        &'life0 self,
        request: Request,
    ) -> Pin<Box<dyn Future<Output = Result<Response, ProviderError>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait;
    fn stream<'life0, 'async_trait>(
        &'life0 self,
        request: Request,
    ) -> Pin<Box<dyn Future<Output = Result<ProviderEventStream, ProviderError>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait;
}
Expand description

Trait for LLM providers (Anthropic, OpenAI, etc.).

Implement this trait to add support for a new LLM provider.

Two API surfaces, two transports:

  • complete — single-shot HTTP, fully buffered response. Lowest overhead when the caller wants the final answer in one go.
  • stream — Server-Sent Events HTTP, emits incremental StreamEvents as the model produces them.

Implementations are independent code paths: streaming is not derived from complete() and vice-versa. Errors that happen before the stream begins (auth, malformed request, connection refused) surface from the stream(...) async fn itself; errors that surface mid-stream (parse failures, mid-body HTTP errors) arrive as Err items inside the stream.

Required Methods§

Source

fn complete<'life0, 'async_trait>( &'life0 self, request: Request, ) -> Pin<Box<dyn Future<Output = Result<Response, ProviderError>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait,

Source

fn stream<'life0, 'async_trait>( &'life0 self, request: Request, ) -> Pin<Box<dyn Future<Output = Result<ProviderEventStream, ProviderError>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait,

Implementors§