Skip to main content

Provider

pi::provider

Trait Provider 

Source 
pub trait Provider: Send + Sync {
    // Required methods
    fn name(&self) -> &str;
    fn api(&self) -> &str;
    fn model_id(&self) -> &str;
    fn stream<'life0, 'life1, 'life2, 'life3, 'async_trait>(
        &'life0 self,
        context: &'life1 Context<'life2>,
        options: &'life3 StreamOptions,
    ) -> Pin<Box<dyn Future<Output = Result<Pin<Box<dyn Stream<Item = Result<StreamEvent>> + Send>>>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait,
             'life2: 'async_trait,
             'life3: 'async_trait;
}
Expand description

An LLM backend capable of streaming assistant output (and tool calls).

A Provider is typically configured for a specific API + model and is used by the agent loop to produce a stream of StreamEvent updates.

Required Methods§

Source

fn name(&self) -> &str

Get the provider name.

Source

fn api(&self) -> &str

Get the API type.

Source

fn model_id(&self) -> &str

Get the model identifier used by this provider.

Source

fn stream<'life0, 'life1, 'life2, 'life3, 'async_trait>( &'life0 self, context: &'life1 Context<'life2>, options: &'life3 StreamOptions, ) -> Pin<Box<dyn Future<Output = Result<Pin<Box<dyn Stream<Item = Result<StreamEvent>> + Send>>>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait, 'life3: 'async_trait,

Start streaming a completion.

Implementations should yield StreamEvent items as soon as they are decoded, and should stop promptly when the request is cancelled.

Implementors§