Struct AnthropicProvider 

pub struct AnthropicProvider { /* private fields */ }

Built-in provider for the Anthropic Messages API.

Requires the anthropic feature flag.

Example

use llm_agent_runtime::providers::AnthropicProvider;
let provider = AnthropicProvider::new("sk-ant-...");

Implementations

impl AnthropicProvider

pub fn new(api_key: impl Into<String>) -> Self

Create a new Anthropic provider with the given API key.

pub fn with_base_url( api_key: impl Into<String>, api_url: impl Into<String>, ) -> Self

Create a provider pointing to a custom API endpoint.

Useful for testing against a mock server or routing through a proxy.

pub fn with_max_concurrent_streams( api_key: impl Into<String>, max: usize, ) -> Self

Create a provider with a custom limit on concurrent streaming tasks.

Useful when many agents share a single provider and you want to cap the number of simultaneous SSE parse tasks.

pub fn with_stream_max_tokens(self, max_tokens: u32) -> Self

Set the maximum output tokens used by stream_complete.

By default stream_complete uses the hard-coded MAX_TOKENS constant (1024). Call this method to override that value, e.g. when a task requires longer streamed responses.

Trait Implementations

impl LlmProvider for AnthropicProvider

fn complete_with_options<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, prompt: &'life1 str, options: CompletionOptions<'life2>, ) -> Pin<Box<dyn Future<Output = Result<String, AgentRuntimeError>> + Send + 'async_trait>>

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

Complete with per-request options (max_tokens, temperature).

Falls back to Self::MAX_TOKENS when options.max_tokens is not set.

fn stream_complete<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, prompt: &'life1 str, model: &'life2 str, ) -> Pin<Box<dyn Future<Output = Result<Receiver<Result<String, AgentRuntimeError>>, AgentRuntimeError>> + Send + 'async_trait>>

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

Stream the completion token-by-token using chunk() for true streaming.

Uses response.chunk() to consume the SSE body incrementally so tokens are emitted as they arrive rather than after the entire response has been buffered, fixing the original bytes().await implementation.

fn complete<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, prompt: &'life1 str, model: &'life2 str, ) -> Pin<Box<dyn Future<Output = Result<String, AgentRuntimeError>> + Send + 'async_trait>>

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

Send a prompt to the model and return the completion text.