Struct Client 

pub struct Client { /* private fields */ }

Unified LLM client that wraps llm-connector for all providers

Implementations

impl Client

pub async fn chat( &self, model: &str, messages: Vec<Message>, tools: Option<Vec<Tool>>, ) -> Result<Response>

Send a non-streaming chat request to the LLM

impl Client

pub async fn list_models(&self) -> Result<Vec<Model>>

List available models

impl Client

pub async fn chat_stream_with_format( &self, model: &str, messages: Vec<Message>, format: StreamFormat, ) -> Result<UnboundedReceiverStream<String>>

Send a streaming chat request with specified format (Ollama-style response)

This method returns streaming responses in Ollama API format, which is used by Ollama-compatible clients like Zed.dev.

pub async fn chat_stream_openai( &self, model: &str, messages: Vec<Message>, tools: Option<Vec<Tool>>, format: StreamFormat, ) -> Result<UnboundedReceiverStream<String>>

Send a streaming chat request for OpenAI API (OpenAI-style response)

This method returns streaming responses in OpenAI API format, which is used by OpenAI-compatible clients like Codex CLI.

Key feature: Automatically corrects finish_reason from “stop” to “tool_calls” when tool_calls are detected in the stream.

impl Client

pub fn new(config: &LlmBackendSettings) -> Result<Self>

Create a new client with the specified backend configuration