Struct LlmClient 

pub struct LlmClient { /* private fields */ }

LLM 客户端实现

Implementations

impl LlmClient

pub fn new(provider: LlmProvider, api_key: String) -> LlmClient

pub fn with_base_url(self, base_url: String) -> LlmClient

创建客户端并指定自定义 base_url（覆盖 provider 默认值）

pub async fn send_stream_with_callback( &self, messages: Vec<Message>, config: &LlmRequestConfig, on_chunk: Box<dyn Fn(&str) + Send + Sync>, ) -> Result<LlmResponse, Error>

发送带回调的流式请求

pub async fn send_stream_abortable( &self, messages: Vec<Message>, config: &LlmRequestConfig, abort_flag: Arc<Atomic<bool>>, ) -> Result<LlmResponse, Error>

发送可中断的流式请求

pub async fn send_with_retry( &self, messages: Vec<Message>, config: &LlmRequestConfig, max_retries: u32, ) -> Result<LlmResponse, Error>

带错误恢复的请求重试

pub async fn send_stream_with_retry( &self, messages: Vec<Message>, config: &LlmRequestConfig, max_retries: u32, ) -> Result<LlmResponse, Error>

带错误恢复的流式请求重试

impl LlmClient

pub fn build_anthropic_messages_url(base_url: &str) -> String

Construct the messages endpoint URL for Anthropic API

Handles three cases:

1. Official Anthropic API: https://api.anthropic.com -> https://api.anthropic.com/v1/messages
2. Already contains full path: https://api.example.com/anthropic/messages -> unchanged
3. Anthropic-compatible endpoint (contains /anthropic): https://api.example.com/anthropic -> /messages
4. Already contains v1: https://api.example.com/v1 -> https://api.example.com/v1/messages

Trait Implementations

impl LlmClientTrait for LlmClient

fn send<'life0, 'life1, 'async_trait>( &'life0 self, messages: Vec<Message>, config: &'life1 LlmRequestConfig, ) -> Pin<Box<dyn Future<Output = Result<LlmResponse, Error>> + Send + 'async_trait>>

where 'life0: 'async_trait, 'life1: 'async_trait, LlmClient: 'async_trait,

发送请求并获取响应

fn send_stream<'life0, 'life1, 'async_trait>( &'life0 self, messages: Vec<Message>, config: &'life1 LlmRequestConfig, ) -> Pin<Box<dyn Future<Output = Result<MessageStream, Error>> + Send + 'async_trait>>

where 'life0: 'async_trait, 'life1: 'async_trait, LlmClient: 'async_trait,

发送请求并流式获取响应