Struct AgentPool 

pub struct AgentPool { /* private fields */ }

Manages a pool of background TaskAgents.

All agents share the same provider, tool executor, communication hub, file lock manager, and working directory. Each agent gets its own conversation history and working set.

Implementations

impl AgentPool

pub fn new( max_agents: usize, provider: Arc<dyn Provider>, tool_executor: Arc<dyn ToolExecutor>, communication_hub: Arc<CommunicationHub>, file_lock_manager: Arc<FileLockManager>, working_directory: impl Into<String>, ) -> Self

Create a new agent pool.

Parameters

• max_agents: maximum number of concurrently running agents.
• provider: AI provider shared by all agents.
• tool_executor: tool executor shared by all agents.
• communication_hub: inter-agent message bus.
• file_lock_manager: file coordination across agents.
• working_directory: default working directory for spawned agents.

pub async fn spawn_agent( &self, task: Task, config: Option<TaskAgentConfig>, ) -> Result<String>

Spawn a new task agent and start it on a Tokio background task.

Returns the agent ID. Use await_completion to wait for the result.

Returns an error if the pool is already at capacity.

pub async fn spawn_agent_with_context( &self, task: Task, context: Arc<AgentContext>, config: Option<TaskAgentConfig>, ) -> Result<String>

Spawn a new task agent with a custom AgentContext.

Unlike spawn_agent which uses the pool’s default working directory, this method accepts a pre-built context. This is useful for workers that run in isolated worktrees with per-agent working directories.

Returns the agent ID.

pub async fn get_status(&self, agent_id: &str) -> Option<TaskAgentStatus>

Get the current status of an agent.

Returns None if the agent is not in the pool.

pub async fn get_task(&self, agent_id: &str) -> Option<Task>

Get a snapshot of the task assigned to an agent.

pub async fn stop_agent(&self, agent_id: &str) -> Result<()>

Abort an agent and remove it from the pool.

File locks held by the agent are released immediately.

pub async fn await_completion(&self, agent_id: &str) -> Result<TaskAgentResult>

Wait for an agent to finish and return its result.

The agent is removed from the pool once it completes.

pub async fn list_active(&self) -> Vec<(String, TaskAgentStatus)>

List all agents currently in the pool with their status.

pub async fn active_count(&self) -> usize

Number of agents currently in the pool (running or pending cleanup).

pub async fn is_running(&self, agent_id: &str) -> bool

Returns true if the agent is still running (join handle not finished).

pub async fn cleanup_completed(&self) -> Vec<(String, Result<TaskAgentResult>)>

Remove all finished agents from the pool and return their results.

pub async fn await_all(&self) -> Vec<(String, Result<TaskAgentResult>)>

Wait for every agent in the pool to finish.

pub async fn shutdown(&self)

Abort all agents and clear the pool.

pub async fn stats(&self) -> AgentPoolStats

Get a statistical snapshot of the pool.

pub fn file_lock_manager(&self) -> Arc<FileLockManager>

Get the shared file lock manager.

pub fn communication_hub(&self) -> Arc<CommunicationHub>

Get the shared communication hub.