Trait LlmBackend 

pub trait LlmBackend: Send + Sync {
    // Required methods
    fn id(&self) -> String;
    fn complete<'life0, 'life1, 'life2, 'async_trait>(
        &'life0 self,
        model: &'life1 str,
        messages: &'life2 [Message],
    ) -> Pin<Box<dyn Future<Output = Result<Message>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait,
             'life2: 'async_trait;
    fn stream<'life0, 'life1, 'life2, 'async_trait>(
        &'life0 self,
        model: &'life1 str,
        messages: &'life2 [Message],
    ) -> Pin<Box<dyn Future<Output = Result<BoxStream<'static, Result<Message>>>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait,
             'life2: 'async_trait;
}

A streamable LLM backend (ollama HTTP, llama.cpp in-process, etc).

Required Methods

fn id(&self) -> String

Human-readable backend identifier (e.g. "ollama:llama3.1").

fn complete<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, model: &'life1 str, messages: &'life2 [Message], ) -> Pin<Box<dyn Future<Output = Result<Message>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Single-shot completion. Implementations MAY buffer a streamed response internally — see stream for the streamed API.

fn stream<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, model: &'life1 str, messages: &'life2 [Message], ) -> Pin<Box<dyn Future<Output = Result<BoxStream<'static, Result<Message>>>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Streamed completion. Yields incremental Message deltas; the last delta has the final assistant turn.

Implementors