Struct NodeTransport 

pub struct NodeTransport { /* private fields */ }

Implementations

impl NodeTransport

pub fn start( child: Child, stdin: ChildStdin, stdout: ChildStdout, ) -> Result<(Self, Value), String>

pub fn start_with_cleanup( child: Child, stdin: ChildStdin, stdout: ChildStdout, cleanup_path: Option<PathBuf>, ) -> Result<(Self, Value), String>

Start a transport and remove an optional host script when the child is dropped. Windows cannot carry the embedded Node source in node -e once the host grows beyond the CreateProcess command-line limit, so the caller may launch a temporary .mjs file and hand its path to us.

pub fn start_with_cleanup_and_handlers( child: Child, stdin: ChildStdin, stdout: ChildStdout, cleanup_path: Option<PathBuf>, initial_handlers: Vec<RuntimeHandler>, ) -> Result<(Self, Value), String>

Start a transport with runtime handlers that are available while the Node host is loading extension factories and lifecycle hooks. The host may issue runtime_request messages before it publishes its id:0 initialization response, so the reader must be running before waiting for that response.

pub fn start_pending_with_cleanup_and_handlers( child: Child, stdin: ChildStdin, stdout: ChildStdout, cleanup_path: Option<PathBuf>, initial_handlers: Vec<RuntimeHandler>, ) -> Result<NodeTransportStartup, String>

Spawn the transport and start its reader, returning before the Node host’s initialization response arrives. This is the primitive used by the lazy runtime so shutdown can kill a host whose extension factory is stuck before it emits id:0.

pub fn shutdown(&self)

Stop the Node host and resolve all currently pending requests with an error. Safe to call repeatedly and while other transport clones exist.

pub fn is_shutdown(&self) -> bool

Whether the reader or an explicit shutdown has made this transport unusable. Lazy owners use this to discard a stale slot and retry on the next request.

pub fn same_instance(&self, other: &Self) -> bool

Whether two handles refer to the same child process. Lazy startup can briefly expose one handle through both its starting and transport slots; callers updating shared runtime handlers must avoid appending the same callback twice in that window.

pub fn begin_request( &self, method: &str, payload: Value, ) -> Result<PendingRequest, String>

pub fn request(&self, method: &str, payload: Value) -> Result<Value, String>

pub fn send_event(&self, event: Value) -> Result<(), String>

pub fn request_event(&self, event: Value) -> Result<Value, String>

Send a host event and wait for a normal id-correlated response. This is used for terminal input listeners, whose consume result must be known before the outer Rust TUI dispatches the same key.

pub fn cancel(&self, id: u64) -> Result<(), String>

pub fn register_tool_update_handler( &self, tool_call_id: impl Into<String>, handler: ToolUpdateHandler, ) -> Result<(), String>

Register a callback for partial updates emitted by one JS tool call.

The callback is intentionally keyed by the Pi tool-call id rather than the transport request id: the former is what Node receives and what remains stable when a request is cancelled or retried.

pub fn unregister_tool_update_handler(&self, tool_call_id: &str)

Stop routing partial updates after a tool call settles. This also suppresses updates sent by a JS tool after its execute promise resolves, matching the agent loop’s late-update gate.

pub fn replace_runtime_handlers( &self, handler: RuntimeHandler, ) -> Result<(), String>

pub fn add_runtime_handler(&self, handler: RuntimeHandler) -> Result<(), String>

Trait Implementations

impl Clone for NodeTransport

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.