feat(KB-157): add soft-pause (enginePaused) alongside hard-stop (globalPause)

- Add enginePaused field to Settings type with soft-pause semantics (drain queue, don't kill agents)
- Gate scheduler, triage, auto-merge, and periodic merge retry on enginePaused
- Add Pause and Stop buttons to dashboard Header replacing single toggle
- Wire enginePaused state through App with optimistic toggle and unpause resume logic
- Add tests for scheduler, triage, dashboard CLI, Header, and App covering pause behavior
This commit is contained in:
Dustin Byrne
2026-03-28 03:50:42 -04:00
parent 3dc741c56e
commit cac10af5e1
14 changed files with 692 additions and 38 deletions

View File

@@ -1069,6 +1069,142 @@ describe("Scheduler in-review worktrees do not count against maxWorktrees", () =
});
});
describe("Scheduler enginePaused (soft pause)", () => {
beforeEach(() => {
vi.clearAllMocks();
});
async function runSchedule(scheduler: Scheduler): Promise<void> {
(scheduler as any).running = true;
await scheduler.schedule();
}
it("does not move any tasks when enginePaused is true", async () => {
const tasks = [
makeTask({ id: "KB-001", column: "todo" }),
makeTask({ id: "KB-002", column: "todo" }),
];
const store = createMockStore(tasks);
store.getSettings.mockResolvedValue({
maxConcurrent: 2,
maxWorktrees: 4,
pollIntervalMs: 15000,
groupOverlappingFiles: false,
autoMerge: false,
enginePaused: true,
});
const scheduler = new Scheduler(store, { maxConcurrent: 2 });
await runSchedule(scheduler);
expect(store.moveTask).not.toHaveBeenCalled();
expect(store.updateTask).not.toHaveBeenCalled();
});
it("resumes scheduling when enginePaused is toggled back to false", async () => {
const tasks = [
makeTask({ id: "KB-001", column: "todo" }),
];
const store = createMockStore(tasks);
store.getSettings.mockResolvedValue({
maxConcurrent: 2,
maxWorktrees: 4,
pollIntervalMs: 15000,
groupOverlappingFiles: false,
autoMerge: false,
enginePaused: true,
});
const scheduler = new Scheduler(store, { maxConcurrent: 2 });
await runSchedule(scheduler);
expect(store.moveTask).not.toHaveBeenCalled();
// Toggle enginePaused off
store.getSettings.mockResolvedValue({
maxConcurrent: 2,
maxWorktrees: 4,
pollIntervalMs: 15000,
groupOverlappingFiles: false,
autoMerge: false,
enginePaused: false,
});
await runSchedule(scheduler);
expect(store.moveTask).toHaveBeenCalledWith("KB-001", "in-progress");
});
it("logs once when entering engine paused state", async () => {
const tasks = [makeTask({ id: "KB-001", column: "todo" })];
const store = createMockStore(tasks);
store.getSettings.mockResolvedValue({
maxConcurrent: 2,
maxWorktrees: 4,
pollIntervalMs: 15000,
groupOverlappingFiles: false,
autoMerge: false,
enginePaused: true,
});
const scheduler = new Scheduler(store, { maxConcurrent: 2 });
const logSpy = vi.spyOn(console, "log").mockImplementation(() => {});
await runSchedule(scheduler);
await runSchedule(scheduler);
await runSchedule(scheduler);
const pauseMessages = logSpy.mock.calls.filter(
(args) =>
typeof args[0] === "string" &&
args[0].includes("Engine paused"),
);
expect(pauseMessages).toHaveLength(1);
logSpy.mockRestore();
});
it("calls schedule() immediately when enginePaused transitions from true to false", async () => {
const tasks = [
makeTask({ id: "KB-001", column: "todo" }),
];
const store = createMockStore(tasks);
store.getSettings.mockResolvedValue({
maxConcurrent: 2,
maxWorktrees: 4,
pollIntervalMs: 15000,
groupOverlappingFiles: false,
autoMerge: false,
enginePaused: false,
});
const scheduler = new Scheduler(store, { maxConcurrent: 2 });
(scheduler as any).running = true;
// Fire the settings:updated event: enginePaused true → false
store._trigger("settings:updated", {
settings: { enginePaused: false },
previous: { enginePaused: true },
});
await new Promise((r) => setTimeout(r, 50));
expect(store.moveTask).toHaveBeenCalledWith("KB-001", "in-progress");
});
it("does NOT call schedule() when enginePaused stays false (false → false)", async () => {
const store = createMockStore();
const scheduler = new Scheduler(store, { maxConcurrent: 2 });
(scheduler as any).running = true;
store._trigger("settings:updated", {
settings: { enginePaused: false },
previous: { enginePaused: false },
});
await new Promise((r) => setTimeout(r, 50));
expect(store.listTasks).not.toHaveBeenCalled();
expect(store.moveTask).not.toHaveBeenCalled();
});
});
describe("Scheduler semaphore-aware slot counting", () => {
beforeEach(() => {
vi.clearAllMocks();

View File

@@ -76,6 +76,7 @@ export class Scheduler {
private scheduling = false;
private wasWorktreeLimited = false;
private wasGlobalPaused = false;
private wasEnginePaused = false;
private pollInterval: ReturnType<typeof setInterval> | null = null;
/** The interval (ms) of the currently active `setInterval` timer. */
private activePollMs: number | null = null;
@@ -98,6 +99,18 @@ export class Scheduler {
this.schedule();
}
});
/**
* Immediate soft-unpause resume: when `enginePaused` transitions from
* `true` to `false`, trigger a scheduling pass right away instead of
* waiting for the next poll interval. Same pattern as the globalPause
* unpause handler above.
*/
this.store.on("settings:updated", ({ settings, previous }) => {
if (previous.enginePaused && !settings.enginePaused && this.running) {
this.schedule();
}
});
}
start(): void {
@@ -197,7 +210,7 @@ export class Scheduler {
// Refresh the poll interval if the persisted setting has changed
this.refreshPollInterval(settings.pollIntervalMs);
// Global pause: halt all scheduling activity
// Global pause (hard stop): halt all scheduling activity
if (settings.globalPause) {
if (!this.wasGlobalPaused) {
schedulerLog.log("Global pause active — scheduling halted");
@@ -207,6 +220,16 @@ export class Scheduler {
}
this.wasGlobalPaused = false;
// Engine paused (soft pause): halt new work dispatch, but let agents finish
if (settings.enginePaused) {
if (!this.wasEnginePaused) {
schedulerLog.log("Engine paused — scheduling halted (in-flight agents continue)");
this.wasEnginePaused = true;
}
return;
}
this.wasEnginePaused = false;
// Count only in-progress tasks toward the worktree limit.
// In-review tasks with worktrees are idle (waiting to merge) and
// should not block new tasks from starting.

View File

@@ -823,6 +823,164 @@ describe("TriageProcessor immediate resume on unpause via settings:updated", ()
});
});
describe("TriageProcessor enginePaused (soft pause)", () => {
beforeEach(() => {
vi.clearAllMocks();
});
it("does not specify any tasks when enginePaused is true", async () => {
const triageTask = {
id: "KB-001",
title: "Test",
description: "Test task",
column: "triage" as const,
dependencies: [],
steps: [],
currentStep: 0,
log: [],
createdAt: new Date().toISOString(),
updatedAt: new Date().toISOString(),
};
const store = createMockStore([triageTask]);
store.getSettings.mockResolvedValue({
maxConcurrent: 2,
maxWorktrees: 4,
pollIntervalMs: 15000,
groupOverlappingFiles: false,
autoMerge: false,
enginePaused: true,
});
mockedCreateHaiAgent.mockResolvedValue({
session: {
prompt: vi.fn().mockResolvedValue(undefined),
dispose: vi.fn(),
},
} as any);
const triage = new TriageProcessor(store, "/tmp/test");
(triage as any).running = true;
await (triage as any).poll();
// Agent should never be created when engine is soft-paused
expect(mockedCreateHaiAgent).not.toHaveBeenCalled();
expect(store.updateTask).not.toHaveBeenCalled();
});
it("resumes triage when enginePaused is toggled back to false", async () => {
const triageTask = {
id: "KB-002",
title: "Normal",
description: "Normal task",
column: "triage" as const,
dependencies: [],
steps: [],
currentStep: 0,
log: [],
createdAt: new Date().toISOString(),
updatedAt: new Date().toISOString(),
};
const store = createMockStore([triageTask]);
store.getSettings.mockResolvedValue({
maxConcurrent: 2,
maxWorktrees: 4,
pollIntervalMs: 15000,
groupOverlappingFiles: false,
autoMerge: false,
enginePaused: true,
});
mockedCreateHaiAgent.mockResolvedValue({
session: {
prompt: vi.fn().mockResolvedValue(undefined),
dispose: vi.fn(),
},
} as any);
const triage = new TriageProcessor(store, "/tmp/test");
(triage as any).running = true;
// First poll — engine paused, nothing happens
await (triage as any).poll();
expect(mockedCreateHaiAgent).not.toHaveBeenCalled();
// Toggle enginePaused off
store.getSettings.mockResolvedValue({
maxConcurrent: 2,
maxWorktrees: 4,
pollIntervalMs: 15000,
groupOverlappingFiles: false,
autoMerge: false,
enginePaused: false,
});
// Second poll — should process tasks
await (triage as any).poll();
await new Promise((r) => setTimeout(r, 50));
expect(store.updateTask).toHaveBeenCalledWith("KB-002", { status: "specifying" });
});
it("calls poll() immediately when enginePaused transitions from true to false", async () => {
const triageTask = {
id: "KB-001",
title: "Test",
description: "Test task",
column: "triage" as const,
dependencies: [],
steps: [],
currentStep: 0,
log: [],
createdAt: new Date().toISOString(),
updatedAt: new Date().toISOString(),
};
const store = createMockStore([triageTask]);
store.getSettings.mockResolvedValue({
maxConcurrent: 2,
maxWorktrees: 4,
pollIntervalMs: 15000,
groupOverlappingFiles: false,
autoMerge: false,
enginePaused: false,
});
mockedCreateHaiAgent.mockResolvedValue({
session: {
prompt: vi.fn().mockResolvedValue(undefined),
dispose: vi.fn(),
},
} as any);
const triage = new TriageProcessor(store, "/tmp/test");
(triage as any).running = true;
// Fire the settings:updated event: enginePaused true → false
store._trigger("settings:updated", {
settings: { enginePaused: false },
previous: { enginePaused: true },
});
await new Promise((r) => setTimeout(r, 50));
// poll() should have been called → triage task processed
expect(store.updateTask).toHaveBeenCalledWith("KB-001", { status: "specifying" });
});
it("does NOT call poll() when enginePaused stays false (false → false)", async () => {
const store = createMockStore([]);
const triage = new TriageProcessor(store, "/tmp/test");
(triage as any).running = true;
store._trigger("settings:updated", {
settings: { enginePaused: false },
previous: { enginePaused: false },
});
await new Promise((r) => setTimeout(r, 50));
expect(store.listTasks).not.toHaveBeenCalled();
});
});
describe("buildSpecificationPrompt", () => {
it("includes project commands when testCommand is set", () => {
const task = createMockTaskDetail();

View File

@@ -184,6 +184,7 @@ export class TriageProcessor {
private activePollMs: number | null = null;
private processing = new Set<string>();
private wasGlobalPaused = false;
private wasEnginePaused = false;
/** Active agent sessions per task, used to terminate on global pause. */
private activeSessions = new Map<string, { dispose: () => void }>();
/** Tasks that were aborted due to global pause (to avoid reporting as errors). */
@@ -228,6 +229,22 @@ export class TriageProcessor {
this.poll();
}
});
/**
* Immediate soft-unpause resume: when `enginePaused` transitions from
* `true` to `false`, trigger a triage poll right away instead of
* waiting for the next poll interval. Same pattern as the globalPause
* unpause handler above.
*
* Note: the agent-kill listener above only fires on `globalPause`
* transitions (hard stop). `enginePaused` (soft pause) lets in-flight
* agents finish gracefully.
*/
store.on("settings:updated", ({ settings, previous }) => {
if (previous.enginePaused && !settings.enginePaused && this.running) {
this.poll();
}
});
}
start(): void {
@@ -286,7 +303,7 @@ export class TriageProcessor {
const settings = await this.store.getSettings();
this.refreshPollInterval(settings.pollIntervalMs);
// Global pause: halt all triage activity
// Global pause (hard stop): halt all triage activity
if (settings.globalPause) {
if (!this.wasGlobalPaused) {
triageLog.log("Global pause active — triage halted");
@@ -296,6 +313,16 @@ export class TriageProcessor {
}
this.wasGlobalPaused = false;
// Engine paused (soft pause): halt new triage work, but let agents finish
if (settings.enginePaused) {
if (!this.wasEnginePaused) {
triageLog.log("Engine paused — triage halted (in-flight agents continue)");
this.wasEnginePaused = true;
}
return;
}
this.wasEnginePaused = false;
const tasks = await this.store.listTasks();
const triageTasks = tasks.filter(
(t) => t.column === "triage" && !this.processing.has(t.id) && !t.paused,