Struct AgentLoop 

pub struct AgentLoop { /* private fields */ }

Agent loop that processes messages with tool support

Implementations

impl AgentLoop

pub fn new( provider: Arc<dyn LLMProvider>, tools: Arc<dyn ToolExecutor>, config: AgentLoopConfig, ) -> AgentLoop

Create a new agent loop

pub fn with_defaults( provider: Arc<dyn LLMProvider>, tools: Arc<dyn ToolExecutor>, ) -> AgentLoop

Create with default configuration

pub async fn process_message( &self, context: Vec<ChatMessage>, content: &str, media: Option<Vec<String>>, ) -> Result<String, Error>

Process a single message with tool execution loop

pub async fn process_with_model( &self, context: Vec<ChatMessage>, content: &str, media: Option<Vec<String>>, model: &str, ) -> Result<String, Error>

Process with custom model

pub async fn process_with_options( &self, context: Vec<ChatMessage>, content: &str, media: Option<Vec<String>>, model: Option<&str>, ) -> Result<String, Error>

Process with custom context builder and options

pub async fn process_with_context_builder( &self, context_builder: &AgentContextBuilder, history: Vec<ChatMessage>, content: &str, media: Option<Vec<String>>, model: Option<&str>, ) -> Result<String, Error>

Process with a context builder (system prompt + history) and optional model.

pub async fn process_with_session( &self, session: &mut ChatSession, content: &str, media: Option<Vec<String>>, context_builder: Option<&AgentContextBuilder>, model: Option<&str>, ) -> Result<String, Error>

Process with a session, updating history automatically.

If a context builder is provided, it will be used to construct the LLM prompt (system + history + user). The session will still record only the user/assistant messages.

pub fn config(&self) -> &AgentLoopConfig

Get the configuration