Struct ImpSession 

pub struct ImpSession { /* private fields */ }

A fully wired agent session.

Manages the lifecycle of a single agent: config resolution, model selection, session persistence, and the event/command channels.

Implementations

impl ImpSession

pub async fn create(options: SessionOptions) -> Result<Self>

Create a new session by resolving config, auth, model, and tools.

This is the main factory — mirrors pi’s createAgentSession().

pub async fn prompt(&mut self, text: &str) -> Result<()>

Send a prompt and run the agent loop.

The agent runs on a background task. Use [recv_event] to consume events, and [steer] / [follow_up] / [cancel] to control it.

Returns an error if the agent is already running.

pub async fn prompt_and_wait(&mut self, text: &str) -> Result<()>

Send a prompt and block until the agent finishes.

Events are still emitted via [recv_event], but this method does not return until the agent loop completes.

pub async fn wait(&mut self) -> Result<()>

Wait for the running agent to finish.

pub async fn steer(&self, text: &str) -> Result<()>

Interrupt the agent: delivered after the current tool finishes, remaining queued tools are skipped.

pub async fn follow_up(&self, text: &str) -> Result<()>

Follow-up: delivered only after the agent finishes all current work.

pub async fn cancel(&self) -> Result<()>

Cancel the current agent run.

pub fn abort(&mut self)

Force-abort the current agent task when graceful cancellation does not finish.

pub async fn recv_event(&mut self) -> Option<AgentEvent>

Receive the next event from the agent.

Returns None when the agent has finished and all events have been consumed.

pub fn event_rx(&mut self) -> &mut Receiver<AgentEvent>

Get mutable access to the raw event receiver.

Use this when you need select! or other channel combinators.

pub async fn set_model(&mut self, hint: &str) -> Result<()>

Switch the model for subsequent prompts.

The change takes effect on the next prompt() call.

pub fn set_thinking(&mut self, level: ThinkingLevel)

Set the thinking level for subsequent prompts.

pub fn model(&self) -> &Model

The current model.

pub fn config(&self) -> &Config

The resolved config.

pub fn session_manager(&self) -> &SessionManager

The session manager (tree, entries, persistence).

pub fn session_manager_mut(&mut self) -> &mut SessionManager

Mutable access to the session manager.

pub fn cwd(&self) -> &PathBuf

The working directory.

pub fn auth_store(&self) -> &AuthStore

The auth store (for checking credentials, OAuth status, etc).

pub fn auth_store_mut(&mut self) -> &mut AuthStore

Mutable access to the auth store.

pub fn model_registry(&self) -> &ModelRegistry

The model registry.

pub fn is_running(&self) -> bool

Whether the agent is currently running a prompt.

pub fn command_tx(&self) -> &Sender<AgentCommand>

Get the raw command sender for advanced use cases.