Struct RetryingProvider 

pub struct RetryingProvider<P> { /* private fields */ }

Wraps any LlmProvider with automatic retry + exponential backoff.

Retries on:

• HTTP 429 (rate limit)
• HTTP 500, 502, 503, 529 (server errors)
• Network errors (Error::Http)

Does NOT retry on:

• HTTP 400, 401, 403, 404 (client errors — retrying won’t help)
• JSON/SSE parse errors (deterministic failures)
• Agent/Config/Memory/Store errors (not LLM-related)

Implementations

impl<P> RetryingProvider<P>

pub fn new(inner: P, config: RetryConfig) -> Self

Wrap inner with the given retry configuration.

pub fn with_defaults(inner: P) -> Self

Wrap a provider with default retry config (3 retries, 500ms base delay).

pub fn with_on_retry(self, callback: Arc<OnRetry>) -> Self

Set a callback invoked before each retry attempt.

The callback receives (attempt, max_retries, delay_ms, error_class).

Trait Implementations

impl<P: LlmProvider> LlmProvider for RetryingProvider<P>

fn model_name(&self) -> Option<&str>

Return the model identifier, if known.

async fn complete( &self, request: CompletionRequest, ) -> Result<CompletionResponse, Error>

Send a completion request and wait for the full response.

async fn stream_complete( &self, request: CompletionRequest, on_text: &OnText, ) -> Result<CompletionResponse, Error>

Stream a completion, calling on_text for each text delta as it arrives.