runtime: route notifications through worker inbox

This commit is contained in:
2026-08-05 01:44:57 +09:00
parent dd2ca54874
commit f98a123e40
15 changed files with 614 additions and 811 deletions
+51 -10
View File
@@ -285,6 +285,7 @@ impl WorkerController {
worker.push_notify(
"Restored Worker state contained unreachable delegated child Workers; their delegated write scopes were reclaimed before resume."
.to_string(),
false,
);
}
@@ -883,7 +884,7 @@ async fn controller_loop<C, St>(
)
.await;
let parent_originated = run.is_parent_originated();
let (new_status, shutdown) = match run {
let (mut new_status, shutdown) = match run {
PendingRun::Run(input) => {
drive_turn(
worker.run(input),
@@ -930,6 +931,11 @@ async fn controller_loop<C, St>(
.await
}
};
if !shutdown && new_status == WorkerStatus::Idle && notify_buffer.has_auto_run_pending()
{
pending = Some(PendingRun::RunForNotification(protocol::InvokeKind::Notify));
new_status = WorkerStatus::Running;
}
finish_controller_run(
&mut worker,
&shared_state,
@@ -985,13 +991,13 @@ async fn controller_loop<C, St>(
// `LogEntry::SystemItem` entry — drained out of the
// notify buffer + broadcast through the sink. No
// separate echo here.
worker.push_notify(message);
// RUNNING / Paused: the buffer push is the entire
// operation; an in-flight turn (or the next
// Resume/Run) will drain it at its next
// pending_history_appends. IDLE: only `auto_run`
// notifications stage RunForNotification; weak progress
// notices stay queued until an explicit run/resume.
worker.push_notify(message, auto_run);
// RUNNING: the in-flight turn drains the buffer at its next
// pending_history_appends; if an auto-run notification remains
// at turn end, the Controller stages a follow-up notification
// turn. Paused notifications remain queued until Resume/Run.
// IDLE: `auto_run` notifications stage RunForNotification;
// weak progress notices stay queued until an explicit run.
if should_auto_run_notification(shared_state.get_status(), auto_run) {
pending = Some(PendingRun::RunForNotification(protocol::InvokeKind::Notify));
}
@@ -1385,11 +1391,11 @@ where
.into(),
});
}
Some(Method::Notify { message, .. }) => {
Some(Method::Notify { message, auto_run }) => {
// Live echo arrives via `Event::SystemItem` once
// the in-flight turn's next `pending_history_appends`
// drains this entry through the interceptor.
notify_buffer.push_notify(message);
notify_buffer.push_notify(message, auto_run);
}
Some(Method::ListCompletions { .. }) => {}
Some(Method::ListWorkers | Method::RestoreWorker { .. } | Method::RegisterPeer { .. }) => {
@@ -1904,6 +1910,41 @@ mod tests {
assert_eq!(env.notify_buffer.len(), 1);
}
#[tokio::test]
async fn running_auto_run_notify_remains_staged_for_followup_turn() {
let mut env = make_env().await;
env._method_tx
.send(Method::Notify {
message: "continue".into(),
auto_run: true,
})
.await
.expect("send notify");
let worker_future = async {
tokio::time::sleep(Duration::from_millis(50)).await;
Ok::<_, WorkerError>(WorkerRunResult::Finished)
};
let (status, shutdown) = drive_turn(
worker_future,
&mut env.method_rx,
&env.event_tx,
&env.cancel_tx,
&env.shared_state,
&env.notify_buffer,
Some(&env.parent_socket_path),
"parent",
&env.spawned_registry,
false,
)
.await;
assert_eq!(status, WorkerStatus::Idle);
assert!(!shutdown);
assert_eq!(env.notify_buffer.len(), 1);
assert!(env.notify_buffer.has_auto_run_pending());
}
#[tokio::test]
async fn compact_method_is_rejected_while_running() {
let mut env = make_env().await;
+6 -4
View File
@@ -220,7 +220,9 @@ impl Interceptor for WorkerInterceptor {
// simply be skipped from the SystemItem batch.
warn!(error = %e, "failed to render notify_wrapper; using raw message");
let fallback = match &entry {
super::notify_buffer::PendingNotify::Notify { message } => message.clone(),
super::notify_buffer::PendingNotify::Notify { message, .. } => {
message.clone()
}
super::notify_buffer::PendingNotify::WorkerEvent { event } => {
session_store::render_worker_event(event)
}
@@ -1019,8 +1021,8 @@ mod tests {
async fn pending_history_appends_drains_buffer_into_items() {
let registry = Arc::new(HookRegistryBuilder::new().build());
let buffer = NotifyBuffer::new();
buffer.push_notify("first".into());
buffer.push_notify("second".into());
buffer.push_notify("first".into(), false);
buffer.push_notify("second".into(), false);
let interceptor = WorkerInterceptor::new(
registry,
@@ -1057,7 +1059,7 @@ mod tests {
// anything itself.
let registry = Arc::new(HookRegistryBuilder::new().build());
let buffer = NotifyBuffer::new();
buffer.push_notify("msg".into());
buffer.push_notify("msg".into(), false);
let interceptor = WorkerInterceptor::new(
registry,
+22 -9
View File
@@ -41,7 +41,7 @@ const CAPACITY: usize = 128;
/// is available.
#[derive(Debug, Clone)]
pub enum PendingNotify {
Notify { message: String },
Notify { message: String, auto_run: bool },
WorkerEvent { event: WorkerEvent },
}
@@ -61,8 +61,8 @@ impl NotifyBuffer {
/// Push a notify entry onto the queue. If the queue is full, the
/// oldest entry is dropped and a `tracing::warn` is emitted — the
/// caller should never hit this in normal operation.
pub fn push_notify(&self, message: String) {
self.push_entry(PendingNotify::Notify { message });
pub fn push_notify(&self, message: String, auto_run: bool) {
self.push_entry(PendingNotify::Notify { message, auto_run });
}
/// Push a typed worker-event entry onto the queue.
@@ -89,6 +89,15 @@ impl NotifyBuffer {
q.drain(..).collect()
}
/// Whether an undrained `Method::Notify { auto_run: true }` remains.
pub fn has_auto_run_pending(&self) -> bool {
self.inner
.lock()
.expect("notify buffer poisoned")
.iter()
.any(|entry| matches!(entry, PendingNotify::Notify { auto_run: true, .. }))
}
/// Number of pending entries. Primarily for tests.
pub fn len(&self) -> usize {
self.inner.lock().expect("notify buffer poisoned").len()
@@ -107,7 +116,7 @@ pub(crate) fn build_system_item(
prompts: &PromptCatalog,
) -> Result<SystemItem, CatalogError> {
match entry {
PendingNotify::Notify { message } => {
PendingNotify::Notify { message, .. } => {
let body = prompts.notify_wrapper(message)?;
Ok(SystemItem::Notification {
message: message.clone(),
@@ -132,12 +141,15 @@ mod tests {
#[test]
fn push_then_drain_preserves_order() {
let buf = NotifyBuffer::new();
buf.push_notify("one".into());
buf.push_notify("two".into());
buf.push_notify("one".into(), false);
assert!(!buf.has_auto_run_pending());
buf.push_notify("two".into(), true);
assert!(buf.has_auto_run_pending());
let drained = buf.drain();
assert!(!buf.has_auto_run_pending());
assert_eq!(drained.len(), 2);
match &drained[0] {
PendingNotify::Notify { message } => assert_eq!(message, "one"),
PendingNotify::Notify { message, .. } => assert_eq!(message, "one"),
other => panic!("unexpected: {other:?}"),
}
assert!(buf.is_empty());
@@ -147,12 +159,12 @@ mod tests {
fn capacity_drops_oldest() {
let buf = NotifyBuffer::new();
for i in 0..(CAPACITY + 5) {
buf.push_notify(format!("msg{i}"));
buf.push_notify(format!("msg{i}"), false);
}
let drained = buf.drain();
assert_eq!(drained.len(), CAPACITY);
match &drained[0] {
PendingNotify::Notify { message } => assert_eq!(message, "msg5"),
PendingNotify::Notify { message, .. } => assert_eq!(message, "msg5"),
other => panic!("unexpected: {other:?}"),
}
}
@@ -161,6 +173,7 @@ mod tests {
fn build_system_item_for_notify_carries_wrapper_body() {
let entry = PendingNotify::Notify {
message: "hello".into(),
auto_run: false,
};
let catalog = PromptCatalog::builtins_only().unwrap();
let item = build_system_item(&entry, &catalog).unwrap();
+3 -2
View File
@@ -1597,8 +1597,8 @@ impl<C: LlmClient, St: Store> Worker<C, St> {
/// `Item::system_message` just before the next LLM request, via
/// `WorkerInterceptor::pending_history_appends`. See [`NotifyBuffer`]
/// for overflow behaviour and the lane-of-record rationale.
pub fn push_notify(&self, message: String) {
self.pending_notifies.push_notify(message);
pub fn push_notify(&self, message: String, auto_run: bool) {
self.pending_notifies.push_notify(message, auto_run);
}
/// Push an agent-visible typed `WorkerEvent` entry onto the pending buffer.
@@ -4364,6 +4364,7 @@ where
self.push_notify(
"Restored Worker state contained missing or unreachable delegated child Workers; their delegated write scopes were reclaimed before resume."
.to_string(),
false,
);
Ok(())
}