feat(macos-app): add custom namespace UI for cluster isolation

Add Advanced Options section with custom namespace field that allows
users to override EXO_LIBP2P_NAMESPACE environment variable. This
enables splitting machines that can see each other into separate
clusters.

- Added customNamespace property with UserDefaults persistence
- Added Advanced Options collapsible section with text field
- Added Save & Restart button that auto-restarts exo process
- Namespace replaces buildTag when custom value is set
- Falls back to buildTag (version) when namespace is empty

app/EXO/EXO/ContentView.swift

@@ -20,6 +20,8 @@ struct ContentView: View {
     @State private var showDebugInfo = false
     @State private var bugReportInFlight = false
     @State private var bugReportMessage: String?
+    @State private var showAdvancedOptions = false
+    @State private var pendingNamespace: String = ""
 
     var body: some View {
         VStack(alignment: .leading, spacing: 12) {
@@ -197,6 +199,8 @@ struct ContentView: View {
                 updater.checkForUpdates()
             }
             .padding(.bottom, 8)
+            advancedOptionsSection
+                .padding(.bottom, 8)
             debugSection
                 .padding(.bottom, 8)
             controlButton(title: "Quit", tint: .secondary) {
@@ -327,6 +331,47 @@ struct ContentView: View {
         }
     }
 
+    private var advancedOptionsSection: some View {
+        VStack(alignment: .leading, spacing: 6) {
+            HStack {
+                Text("Advanced Options")
+                    .font(.caption)
+                    .foregroundColor(.secondary)
+                Spacer()
+                collapseButton(isExpanded: $showAdvancedOptions)
+            }
+            .animation(nil, value: showAdvancedOptions)
+            if showAdvancedOptions {
+                VStack(alignment: .leading, spacing: 8) {
+                    VStack(alignment: .leading, spacing: 4) {
+                        Text("Cluster Namespace")
+                            .font(.caption2)
+                            .foregroundColor(.secondary)
+                        HStack {
+                            TextField("optional", text: $pendingNamespace)
+                                .textFieldStyle(.roundedBorder)
+                                .font(.caption2)
+                                .onAppear {
+                                    pendingNamespace = controller.customNamespace
+                                }
+                            Button("Save & Restart") {
+                                controller.customNamespace = pendingNamespace
+                                if controller.status == .running || controller.status == .starting {
+                                    controller.restart()
+                                }
+                            }
+                            .font(.caption2)
+                            .disabled(pendingNamespace == controller.customNamespace)
+                        }
+
+                    }
+                }
+                .transition(.opacity)
+            }
+        }
+        .animation(.easeInOut(duration: 0.25), value: showAdvancedOptions)
+    }
+
     private var debugSection: some View {
         VStack(alignment: .leading, spacing: 6) {
             HStack {

app/EXO/EXO/ExoProcessController.swift

@@ -2,6 +2,8 @@ import AppKit
 import Combine
 import Foundation
 
+private let customNamespaceKey = "EXOCustomNamespace"
+
 @MainActor
 final class ExoProcessController: ObservableObject {
     enum Status: Equatable {
@@ -27,6 +29,13 @@ final class ExoProcessController: ObservableObject {
     @Published private(set) var status: Status = .stopped
     @Published private(set) var lastError: String?
     @Published private(set) var launchCountdownSeconds: Int?
+    @Published var customNamespace: String = {
+        return UserDefaults.standard.string(forKey: customNamespaceKey) ?? ""
+    }() {
+        didSet {
+            UserDefaults.standard.set(customNamespace, forKey: customNamespaceKey)
+        }
+    }
 
     private var process: Process?
     private var runtimeDirectoryURL: URL?
@@ -180,7 +189,7 @@ final class ExoProcessController: ObservableObject {
     private func makeEnvironment(for runtimeURL: URL) -> [String: String] {
         var environment = ProcessInfo.processInfo.environment
         environment["EXO_RUNTIME_DIR"] = runtimeURL.path
-        environment["EXO_LIBP2P_NAMESPACE"] = buildTag()
+        environment["EXO_LIBP2P_NAMESPACE"] = computeNamespace()
 
         var paths: [String] = []
         if let existing = environment["PATH"], !existing.isEmpty {
@@ -217,6 +226,12 @@ final class ExoProcessController: ObservableObject {
         }
         return "dev"
     }
+
+    private func computeNamespace() -> String {
+        let base = buildTag()
+        let custom = customNamespace.trimmingCharacters(in: .whitespaces)
+        return custom.isEmpty ? base : custom
+    }
 }
 
 struct RuntimeError: LocalizedError {

dashboard/src/lib/components/ChatForm.svelte

@@ -139,6 +139,11 @@
 	}
 
 	function handleKeydown(event: KeyboardEvent) {
+		// Prevent form submission during IME composition (e.g., Chinese, Japanese, Korean input)
+		if (event.isComposing || event.keyCode === 229) {
+			return;
+		}
+		
 		if (event.key === 'Enter' && !event.shiftKey) {
 			event.preventDefault();
 			handleSubmit();

dashboard/src/routes/+page.svelte

@@ -51,6 +51,59 @@ const sidebarVisible = $derived(chatSidebarVisible());
 	let selectedSharding = $state<'Pipeline' | 'Tensor'>('Pipeline');
 	type InstanceMeta = 'MlxRing' | 'MlxIbv' | 'MlxJaccl';
 	
+	// Launch defaults persistence
+	const LAUNCH_DEFAULTS_KEY = 'exo-launch-defaults';
+	interface LaunchDefaults {
+		modelId: string | null;
+		sharding: 'Pipeline' | 'Tensor';
+		instanceType: InstanceMeta;
+		minNodes: number;
+	}
+	
+	function saveLaunchDefaults(): void {
+		const defaults: LaunchDefaults = {
+			modelId: selectedPreviewModelId(),
+			sharding: selectedSharding,
+			instanceType: selectedInstanceType,
+			minNodes: selectedMinNodes,
+		};
+		try {
+			localStorage.setItem(LAUNCH_DEFAULTS_KEY, JSON.stringify(defaults));
+		} catch (e) {
+			console.warn('Failed to save launch defaults:', e);
+		}
+	}
+	
+	function loadLaunchDefaults(): LaunchDefaults | null {
+		try {
+			const stored = localStorage.getItem(LAUNCH_DEFAULTS_KEY);
+			if (!stored) return null;
+			return JSON.parse(stored) as LaunchDefaults;
+		} catch (e) {
+			console.warn('Failed to load launch defaults:', e);
+			return null;
+		}
+	}
+	
+	function applyLaunchDefaults(availableModels: Array<{id: string}>, maxNodes: number): void {
+		const defaults = loadLaunchDefaults();
+		if (!defaults) return;
+		
+		// Apply sharding and instance type unconditionally
+		selectedSharding = defaults.sharding;
+		selectedInstanceType = defaults.instanceType;
+		
+		// Apply minNodes if valid (between 1 and maxNodes)
+		if (defaults.minNodes && defaults.minNodes >= 1 && defaults.minNodes <= maxNodes) {
+			selectedMinNodes = defaults.minNodes;
+		}
+		
+		// Only apply model if it exists in the available models
+		if (defaults.modelId && availableModels.some(m => m.id === defaults.modelId)) {
+			selectPreviewModel(defaults.modelId);
+		}
+	}
+	
 	let selectedInstanceType = $state<InstanceMeta>('MlxRing');
 	let selectedMinNodes = $state<number>(1);
 	let minNodesInitialized = $state(false);
@@ -298,6 +351,9 @@ function toggleInstanceDownloadDetails(nodeId: string): void {
 				const data = await response.json();
 				// API returns { data: [{ id, name }] } format
 				models = data.data || [];
+				// Restore last launch defaults if available
+				const currentNodeCount = topologyData() ? Object.keys(topologyData()!.nodes).length : 1;
+				applyLaunchDefaults(models, currentNodeCount);
 			}
 		} catch (error) {
 			console.error('Failed to fetch models:', error);
@@ -988,6 +1044,7 @@ function toggleInstanceDownloadDetails(nodeId: string): void {
 
 	function handleSliderMouseUp() {
 		isDraggingSlider = false;
+		saveLaunchDefaults();
 	}
 
 	// Handle touch events for mobile
@@ -1007,6 +1064,7 @@ function toggleInstanceDownloadDetails(nodeId: string): void {
 
 	function handleSliderTouchEnd() {
 		isDraggingSlider = false;
+		saveLaunchDefaults();
 	}
 
 	const nodeCount = $derived(data ? Object.keys(data.nodes).length : 0);
@@ -1464,6 +1522,7 @@ function toggleInstanceDownloadDetails(nodeId: string): void {
 												onclick={() => {
 													if (modelCanFit) {
 														selectPreviewModel(model.id);
+														saveLaunchDefaults();
 														isModelDropdownOpen = false;
 														modelDropdownSearch = '';
 													}
@@ -1497,7 +1556,7 @@ function toggleInstanceDownloadDetails(nodeId: string): void {
 								<div class="text-xs text-white/70 font-mono mb-2">Sharding:</div>
 								<div class="flex gap-2">
 									<button 
-										onclick={() => selectedSharding = 'Pipeline'}
+										onclick={() => { selectedSharding = 'Pipeline'; saveLaunchDefaults(); }}
 										class="flex items-center gap-2 py-2 px-4 text-sm font-mono border rounded transition-all duration-200 cursor-pointer {selectedSharding === 'Pipeline' ? 'bg-transparent text-exo-yellow border-exo-yellow' : 'bg-transparent text-white/70 border-exo-medium-gray/50 hover:border-exo-yellow/50'}"
 									>
 										<span class="w-4 h-4 rounded-full border-2 flex items-center justify-center {selectedSharding === 'Pipeline' ? 'border-exo-yellow' : 'border-exo-medium-gray'}">
@@ -1508,7 +1567,7 @@ function toggleInstanceDownloadDetails(nodeId: string): void {
 										Pipeline
 									</button>
 									<button 
-										onclick={() => selectedSharding = 'Tensor'}
+										onclick={() => { selectedSharding = 'Tensor'; saveLaunchDefaults(); }}
 										class="flex items-center gap-2 py-2 px-4 text-sm font-mono border rounded transition-all duration-200 cursor-pointer {selectedSharding === 'Tensor' ? 'bg-transparent text-exo-yellow border-exo-yellow' : 'bg-transparent text-white/70 border-exo-medium-gray/50 hover:border-exo-yellow/50'}"
 									>
 										<span class="w-4 h-4 rounded-full border-2 flex items-center justify-center {selectedSharding === 'Tensor' ? 'border-exo-yellow' : 'border-exo-medium-gray'}">
@@ -1526,7 +1585,7 @@ function toggleInstanceDownloadDetails(nodeId: string): void {
 								<div class="text-xs text-white/70 font-mono mb-2">Instance Type:</div>
 								<div class="flex gap-2">
 									<button 
-										onclick={() => selectedInstanceType = 'MlxRing'}
+										onclick={() => { selectedInstanceType = 'MlxRing'; saveLaunchDefaults(); }}
 										class="flex items-center gap-2 py-2 px-4 text-sm font-mono border rounded transition-all duration-200 cursor-pointer {selectedInstanceType === 'MlxRing' ? 'bg-transparent text-exo-yellow border-exo-yellow' : 'bg-transparent text-white/70 border-exo-medium-gray/50 hover:border-exo-yellow/50'}"
 									>
 										<span class="w-4 h-4 rounded-full border-2 flex items-center justify-center {selectedInstanceType === 'MlxRing' ? 'border-exo-yellow' : 'border-exo-medium-gray'}">
@@ -1537,7 +1596,7 @@ function toggleInstanceDownloadDetails(nodeId: string): void {
 										MLX Ring
 									</button>
 									<button 
-										onclick={() => selectedInstanceType = 'MlxIbv'}
+										onclick={() => { selectedInstanceType = 'MlxIbv'; saveLaunchDefaults(); }}
 										class="flex items-center gap-2 py-2 px-4 text-sm font-mono border rounded transition-all duration-200 cursor-pointer {selectedInstanceType === 'MlxIbv' ? 'bg-transparent text-exo-yellow border-exo-yellow' : 'bg-transparent text-white/70 border-exo-medium-gray/50 hover:border-exo-yellow/50'}"
 									>
 										<span class="w-4 h-4 rounded-full border-2 flex items-center justify-center {selectedInstanceType === 'MlxIbv' ? 'border-exo-yellow' : 'border-exo-medium-gray'}">

src/exo/main.py

@@ -28,7 +28,7 @@ from exo.worker.main import Worker
 @dataclass
 class Node:
     router: Router
-    worker: Worker
+    worker: Worker | None
     election: Election  # Every node participates in election, as we do want a node to become master even if it isn't a master candidate if no master candidates are present.
     election_result_receiver: Receiver[ElectionResult]
     master: Master | None
@@ -62,15 +62,19 @@ class Node:
         else:
             api = None
 
-        worker = Worker(
-            node_id,
-            session_id,
-            exo_shard_downloader(),
-            connection_message_receiver=router.receiver(topics.CONNECTION_MESSAGES),
-            global_event_receiver=router.receiver(topics.GLOBAL_EVENTS),
-            local_event_sender=router.sender(topics.LOCAL_EVENTS),
-            command_sender=router.sender(topics.COMMANDS),
-        )
+        if not args.no_worker:
+            worker = Worker(
+                node_id,
+                session_id,
+                exo_shard_downloader(),
+                connection_message_receiver=router.receiver(topics.CONNECTION_MESSAGES),
+                global_event_receiver=router.receiver(topics.GLOBAL_EVENTS),
+                local_event_sender=router.sender(topics.LOCAL_EVENTS),
+                command_sender=router.sender(topics.COMMANDS),
+            )
+        else:
+            worker = None
+
         # We start every node with a master
         master = Master(
             node_id,
@@ -100,8 +104,9 @@ class Node:
         async with self._tg as tg:
             signal.signal(signal.SIGINT, lambda _, __: self.shutdown())
             tg.start_soon(self.router.run)
-            tg.start_soon(self.worker.run)
             tg.start_soon(self.election.run)
+            if self.worker:
+                tg.start_soon(self.worker.run)
             if self.master:
                 tg.start_soon(self.master.run)
             if self.api:
@@ -209,6 +214,7 @@ class Args(CamelCaseModel):
     spawn_api: bool = False
     api_port: PositiveInt = 52415
     tb_only: bool = False
+    no_worker: bool = False
 
     @classmethod
     def parse(cls) -> Self:
@@ -246,6 +252,10 @@ class Args(CamelCaseModel):
             dest="api_port",
             default=52415,
         )
+        parser.add_argument(
+            "--no-worker",
+            action="store_true",
+        )
 
         args = parser.parse_args()
         return cls(**vars(args))  # pyright: ignore[reportAny] - We are intentionally validating here, we can't do it statically

src/exo/shared/types/worker/runners.py

@@ -53,6 +53,10 @@ class RunnerRunning(BaseRunnerStatus):
     pass
 
 
+class RunnerShuttingDown(BaseRunnerStatus):
+    pass
+
+
 class RunnerShutdown(BaseRunnerStatus):
     pass
 
@@ -70,6 +74,7 @@ RunnerStatus = (
     | RunnerWarmingUp
     | RunnerReady
     | RunnerRunning
+    | RunnerShuttingDown
     | RunnerShutdown
     | RunnerFailed
 )

src/exo/worker/plan.py

@@ -274,6 +274,12 @@ def _pending_tasks(
             if task.instance_id != runner.bound_instance.instance.instance_id:
                 continue
 
+            # I have a design point here; this is a state race in disguise as the task status doesn't get updated to completed fast enough
+            # however, realistically the task status should be set to completed by the LAST runner, so this is a true race
+            # the actual solution is somewhat deeper than this bypass - TODO!
+            if task.task_id in runner.completed:
+                continue
+
             # TODO: Check ordering aligns with MLX distributeds expectations.
 
             if isinstance(runner.status, RunnerReady) and all(

src/exo/worker/runner/runner.py

@@ -32,6 +32,7 @@ from exo.shared.types.worker.runners import (
     RunnerReady,
     RunnerRunning,
     RunnerShutdown,
+    RunnerShuttingDown,
     RunnerStatus,
     RunnerWarmingUp,
 )
@@ -187,13 +188,14 @@ def main(
                         current_status = RunnerReady()
                         logger.info("runner ready")
                     case Shutdown():
+                        current_status = RunnerShuttingDown()
                         logger.info("runner shutting down")
                         event_sender.send(
-                            TaskStatusUpdated(
-                                task_id=task.task_id, task_status=TaskStatus.Complete
+                            RunnerStatusUpdated(
+                                runner_id=runner_id, runner_status=current_status
                             )
                         )
-                        break
+                        current_status = RunnerShutdown()
                     case _:
                         raise ValueError(
                             f"Received {task.__class__.__name__} outside of state machine in {current_status=}"
@@ -208,9 +210,8 @@ def main(
                         runner_id=runner_id, runner_status=current_status
                     )
                 )
-        event_sender.send(
-            RunnerStatusUpdated(runner_id=runner_id, runner_status=RunnerShutdown())
-        )
+                if isinstance(current_status, RunnerShutdown):
+                    break
     except ClosedResourceError:
         logger.warning("runner communication closed unexpectedly")
     except Exception as e:

src/exo/worker/runner/runner_supervisor.py

@@ -14,13 +14,23 @@ from anyio import (
 from anyio.abc import TaskGroup
 from loguru import logger
 
-from exo.shared.types.events import Event, RunnerStatusUpdated, TaskAcknowledged
-from exo.shared.types.tasks import Task, TaskId
+from exo.shared.types.events import (
+    Event,
+    RunnerStatusUpdated,
+    TaskAcknowledged,
+    TaskStatusUpdated,
+)
+from exo.shared.types.tasks import Task, TaskId, TaskStatus
 from exo.shared.types.worker.instances import BoundInstance
 from exo.shared.types.worker.runners import (
+    RunnerConnecting,
     RunnerFailed,
     RunnerIdle,
+    RunnerLoading,
+    RunnerRunning,
+    RunnerShuttingDown,
     RunnerStatus,
+    RunnerWarmingUp,
 )
 from exo.shared.types.worker.shards import ShardMetadata
 from exo.utils.channels import MpReceiver, MpSender, Sender, mp_channel
@@ -39,10 +49,10 @@ class RunnerSupervisor:
     _ev_recv: MpReceiver[Event]
     _task_sender: MpSender[Task]
     _event_sender: Sender[Event]
-    # err_path: str
     _tg: TaskGroup | None = field(default=None, init=False)
     status: RunnerStatus = field(default_factory=RunnerIdle, init=False)
     pending: dict[TaskId, anyio.Event] = field(default_factory=dict, init=False)
+    completed: set[TaskId] = field(default_factory=set, init=False)
 
     @classmethod
     def create(
@@ -77,7 +87,6 @@ class RunnerSupervisor:
             _ev_recv=ev_recv,
             _task_sender=task_sender,
             _event_sender=event_sender,
-            # err_path=err_path,
         )
 
         return self
@@ -118,6 +127,10 @@ class RunnerSupervisor:
         self._tg.cancel_scope.cancel()
 
     async def start_task(self, task: Task):
+        if task.task_id in self.completed:
+            logger.info(
+                f"Skipping invalid task {task} as it has already been completed"
+            )
         logger.info(f"Starting task {task}")
         event = anyio.Event()
         self.pending[task.task_id] = event
@@ -138,6 +151,22 @@ class RunnerSupervisor:
                     if isinstance(event, TaskAcknowledged):
                         self.pending.pop(event.task_id).set()
                         continue
+                    if (
+                        isinstance(event, TaskStatusUpdated)
+                        and event.task_status == TaskStatus.Complete
+                    ):
+                        # If a task has just been completed, we should be working on it.
+                        assert isinstance(
+                            self.status,
+                            (
+                                RunnerRunning,
+                                RunnerWarmingUp,
+                                RunnerLoading,
+                                RunnerConnecting,
+                                RunnerShuttingDown,
+                            ),
+                        )
+                        self.completed.add(event.task_id)
                     await self._event_sender.send(event)
             except (ClosedResourceError, BrokenResourceError) as e:
                 await self._check_runner(e)

src/exo/worker/tests/unittests/conftest.py

@@ -1,11 +1,9 @@
-from __future__ import annotations
-
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 
 from exo.shared.types.common import NodeId
 from exo.shared.types.memory import Memory
 from exo.shared.types.models import ModelId, ModelMetadata
-from exo.shared.types.tasks import BaseTask
+from exo.shared.types.tasks import BaseTask, TaskId
 from exo.shared.types.worker.instances import (
     BoundInstance,
     Instance,
@@ -21,6 +19,7 @@ from exo.shared.types.worker.shards import PipelineShardMetadata, ShardMetadata
 class FakeRunnerSupervisor:
     bound_instance: BoundInstance
     status: RunnerStatus
+    completed: set[TaskId] = field(default_factory=set)
 
 
 class OtherTask(BaseTask):

src/exo/worker/tests/unittests/test_runner/test_event_ordering.py

@@ -34,6 +34,7 @@ from exo.shared.types.worker.runners import (
     RunnerReady,
     RunnerRunning,
     RunnerShutdown,
+    RunnerShuttingDown,
     RunnerWarmingUp,
 )
 from exo.utils.channels import mp_channel
@@ -199,6 +200,9 @@ def test_events_processed_in_correct_order(patch_out_mlx: pytest.MonkeyPatch):
             RunnerStatusUpdated(runner_id=RUNNER_1_ID, runner_status=RunnerReady()),
             TaskStatusUpdated(task_id=SHUTDOWN_TASK_ID, task_status=TaskStatus.Running),
             TaskAcknowledged(task_id=SHUTDOWN_TASK_ID),
+            RunnerStatusUpdated(
+                runner_id=RUNNER_1_ID, runner_status=RunnerShuttingDown()
+            ),
             TaskStatusUpdated(
                 task_id=SHUTDOWN_TASK_ID, task_status=TaskStatus.Complete
             ),