feat(v0.7): Walk-away Workbench (Band A + B) — multi-agent parallel build

Implements docs/PLAN_WALKAWAY_WORKBENCH.md (27 tasks, waves R0→W0→W1×14→W2→W3→W4) via module-builder agents. 23 tasks built, 0 blocked. Band A (finish the walk-away loop): A1 Web Push + lock-screen approve/deny (web-push dep), A2 voice dictation, A3 quick-reply chips + saved-prompt palette, A4 activity timeline, A5 stuck/idle alert. Band B (workbench above the terminal): B1 read-only git diff viewer, B2 statusLine telemetry → per-tab cost/context/PR gauges, B3 create git worktrees from the UI, B4 plan-mode / permission-mode relay. New: src/push/* (subscription store + VAPID push), src/http/{diff,statusline}.ts, src/session/timeline.ts, public/{diff,timeline,quickreply,push-ui,...}.ts, sw-push, statusLine script; extends hook intake, manager, server routes (Origin/CSRF guards + per-IP rate limits on state-changing ones; loopback-only ingest), terminal-session, tabs, projects detail, service worker, setup-hooks (statusLine + ntfy bridge). Orchestrator reconciled a W0 contract gap: added the 21 v0.7 Config fields to the Config interface in types.ts (T-types had left them only in config.ts's return). Verified: both tsc clean, full vitest + coverage 91.4/84.1/92.2/93.4 (≥80×4), build:web OK. W4 review: no CRITICAL/HIGH; all security checks pass. Follow-ups (non-blocking): move approve.mode validation into parseClientMessage, drop CSP ws:/wss: wildcard, validate worktree base ref, +2 targeted tests.
2026-06-30 17:42:18 +02:00
parent 4f1d3ebc6b
commit d6809c65c4
54 changed files with 13171 additions and 200 deletions
--- a/test/manager.test.ts
+++ b/test/manager.test.ts
@@ -13,7 +13,14 @@

 import { describe, it, expect, beforeEach, vi } from 'vitest';

-import type { Config, Dims, Session, WebSocketLike } from '../src/types.js';
+import type {
+  Config,
+  Dims,
+  NotifyService,
+  Session,
+  StatusTelemetry,
+  WebSocketLike,
+} from '../src/types.js';
 import { WS_OPEN } from '../src/types.js';
 import { createMockPty, type MockIPty } from './helpers/mock-pty.js';

@@ -53,10 +60,53 @@ const CFG: Config = {
  previewBytes: 24 * 1024,
  useTmux: false,
  allowedOrigins: [],
+  // v0.6 project manager
+  projectRoots: ['/home/tester'],
+  projectScanDepth: 4,
+  projectScanTtlMs: 10_000,
+  projectDirtyCheck: true,
+  editorCmd: 'code',
+  // v0.7 Walk-away Workbench
+  vapidPublicKey: undefined,
+  vapidPrivateKey: undefined,
+  vapidSubject: undefined,
+  pushStorePath: '/home/tester/.web-terminal-push-subs.json',
+  pushMaxSubs: 50,
+  notifyDone: true,
+  notifyDnd: false,
+  decisionTokenTtlMs: 300_000,
+  timelineMax: 200,
+  timelineEnabled: true,
+  stuckTtlMs: 10_000, // 10 s for easy arithmetic in tests
+  stuckAlert: true,
+  diffTimeoutMs: 2000,
+  diffMaxBytes: 2 * 1024 * 1024,
+  diffMaxFiles: 300,
+  statuslineTtlMs: 30_000,
+  worktreeEnabled: true,
+  worktreeRoot: undefined,
+  worktreeTimeoutMs: 10_000,
+  defaultPermissionMode: 'default',
+  allowAutoMode: false,
 };

 const DIMS: Dims = { cols: 80, rows: 24 };

+/** A telemetry sample for B2 tests. */
+function makeTelemetry(at = 1_000): StatusTelemetry {
+  return { contextUsedPct: 42, costUsd: 1.23, model: 'sonnet', at };
+}
+
+/** A mock NotifyService (DI) recording every notify(session, cls, token?) call. */
+function createMockNotify() {
+  const notify = vi.fn(async (_session: Session, _cls: string, _token?: string) => {});
+  const service: NotifyService = {
+    isEnabled: () => true,
+    notify,
+  };
+  return { service, notify };
+}
+
 interface MockWs extends WebSocketLike {
  readyState: number;
  sent: string[];
@@ -646,3 +696,356 @@ describe('onExit removes session from table when ws attached at exit time (L2)',
    expect(session.exitedAt).not.toBeNull();
  });
 });
+
+// ── handleHookEvent — A4 timeline + B4 gate (T-manager) ───────────────────────
+describe('handleHookEvent — timeline append (A4)', () => {
+  it('appends a derived timeline event when eventClass is supplied + timeline enabled', () => {
+    const mgr = createSessionManager(CFG);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+
+    mgr.handleHookEvent(s.meta.id, 'working', 'Bash', false, undefined, 'PreToolUse', 'Bash');
+
+    expect(s.timeline).toHaveLength(1);
+    expect(s.timeline[0]).toMatchObject({ class: 'tool', toolName: 'Bash', label: 'using Bash' });
+    expect(typeof s.timeline[0]?.at).toBe('number');
+  });
+
+  it('replaces the timeline array immutably (never mutates the previous one)', () => {
+    const mgr = createSessionManager(CFG);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+    const before = s.timeline;
+
+    mgr.handleHookEvent(s.meta.id, 'working', undefined, false, undefined, 'PreToolUse', 'Bash');
+
+    expect(s.timeline).not.toBe(before);
+    expect(before).toHaveLength(0);
+  });
+
+  it('does NOT append when timeline is disabled', () => {
+    const mgr = createSessionManager({ ...CFG, timelineEnabled: false });
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+
+    mgr.handleHookEvent(s.meta.id, 'working', undefined, false, undefined, 'PreToolUse', 'Bash');
+
+    expect(s.timeline).toHaveLength(0);
+  });
+
+  it('does NOT append when eventClass is omitted', () => {
+    const mgr = createSessionManager(CFG);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+
+    mgr.handleHookEvent(s.meta.id, 'waiting', 'Bash', true);
+
+    expect(s.timeline).toHaveLength(0);
+  });
+
+  it('drops non-whitelisted event classes (makeTimelineEvent → null)', () => {
+    const mgr = createSessionManager(CFG);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+
+    mgr.handleHookEvent(s.meta.id, 'working', undefined, false, undefined, 'BogusEvent', 'X');
+
+    expect(s.timeline).toHaveLength(0);
+  });
+
+  it('evicts oldest events past cfg.timelineMax', () => {
+    const mgr = createSessionManager({ ...CFG, timelineMax: 3 });
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+
+    for (let i = 0; i < 5; i += 1) {
+      mgr.handleHookEvent(s.meta.id, 'working', undefined, false, undefined, 'PreToolUse', `Tool${i}`);
+    }
+
+    expect(s.timeline).toHaveLength(3);
+    expect(s.timeline.map((e) => e.toolName)).toEqual(['Tool2', 'Tool3', 'Tool4']);
+  });
+});
+
+describe('handleHookEvent — gate broadcast (B4)', () => {
+  it('broadcasts the status frame with gate when supplied', () => {
+    const mgr = createSessionManager(CFG);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+    ws.sent.length = 0;
+
+    mgr.handleHookEvent(s.meta.id, 'waiting', 'ExitPlanMode', true, 'plan', 'PermissionRequest', 'ExitPlanMode');
+
+    const status = parseSent(ws).find((m) => m.type === 'status');
+    expect(status).toMatchObject({
+      type: 'status',
+      status: 'waiting',
+      detail: 'ExitPlanMode',
+      pending: true,
+      gate: 'plan',
+    });
+  });
+
+  it('omits gate when not supplied', () => {
+    const mgr = createSessionManager(CFG);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+    ws.sent.length = 0;
+
+    mgr.handleHookEvent(s.meta.id, 'waiting', 'Bash', true);
+
+    const status = parseSent(ws).find((m) => m.type === 'status');
+    expect(status).not.toHaveProperty('gate');
+  });
+});
+
+// ── handleStatusLine (B2 telemetry) ───────────────────────────────────────────
+describe('handleStatusLine', () => {
+  it('stores telemetry on the session and broadcasts it to all clients', () => {
+    const mgr = createSessionManager(CFG);
+    const a = createMockWs();
+    const b = createMockWs();
+    const s = mgr.handleAttach(a, null, DIMS, 1_000);
+    mgr.handleAttach(b, s.meta.id, DIMS, 2_000);
+    a.sent.length = 0;
+    b.sent.length = 0;
+
+    const telemetry = makeTelemetry();
+    mgr.handleStatusLine(s.meta.id, telemetry);
+
+    expect(s.telemetry).toBe(telemetry);
+    for (const ws of [a, b]) {
+      const msg = parseSent(ws).find((m) => m.type === 'telemetry');
+      expect(msg).toMatchObject({ type: 'telemetry', telemetry: { costUsd: 1.23, model: 'sonnet' } });
+    }
+  });
+
+  it('is a no-op for an unknown session id', () => {
+    const mgr = createSessionManager(CFG);
+    expect(() => mgr.handleStatusLine('00000000-0000-4000-8000-0000000000cc', makeTelemetry())).not.toThrow();
+  });
+});
+
+// ── handleAttach Case 2 — late-join telemetry/status replay (M3 / AC-B2.3) ─────
+describe('handleAttach — late-join replay (M3)', () => {
+  it('sends the current telemetry to a device joining a live session', () => {
+    const mgr = createSessionManager(CFG);
+    const a = createMockWs();
+    const s = mgr.handleAttach(a, null, DIMS, 1_000);
+    mgr.handleStatusLine(s.meta.id, makeTelemetry());
+
+    const b = createMockWs();
+    mgr.handleAttach(b, s.meta.id, DIMS, 2_000);
+
+    const msg = parseSent(b).find((m) => m.type === 'telemetry');
+    expect(msg).toMatchObject({ type: 'telemetry', telemetry: { model: 'sonnet' } });
+  });
+
+  it('sends the current status to a late-joining device', () => {
+    const mgr = createSessionManager(CFG);
+    const a = createMockWs();
+    const s = mgr.handleAttach(a, null, DIMS, 1_000);
+    mgr.handleHookEvent(s.meta.id, 'working');
+
+    const b = createMockWs();
+    mgr.handleAttach(b, s.meta.id, DIMS, 2_000);
+
+    const status = parseSent(b).find((m) => m.type === 'status');
+    expect(status).toMatchObject({ type: 'status', status: 'working' });
+  });
+
+  it('does NOT send a telemetry frame when the session has no telemetry yet', () => {
+    const mgr = createSessionManager(CFG);
+    const a = createMockWs();
+    const s = mgr.handleAttach(a, null, DIMS, 1_000);
+
+    const b = createMockWs();
+    mgr.handleAttach(b, s.meta.id, DIMS, 2_000);
+
+    expect(parseSent(b).some((m) => m.type === 'telemetry')).toBe(false);
+  });
+});
+
+// ── sweepStuck (A5) ───────────────────────────────────────────────────────────
+describe('sweepStuck (A5)', () => {
+  it('marks a silent live session stuck, broadcasts, and notifies once', () => {
+    const { service, notify } = createMockNotify();
+    const mgr = createSessionManager(CFG, service);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000); // lastOutputAt = 1000
+    ws.sent.length = 0;
+
+    mgr.sweepStuck(1_000 + CFG.stuckTtlMs + 1);
+
+    expect(s.claudeStatus).toBe('stuck');
+    expect(s.stuckNotified).toBe(true);
+    const status = parseSent(ws).find((m) => m.type === 'status');
+    expect(status).toMatchObject({ type: 'status', status: 'stuck' });
+    expect(notify).toHaveBeenCalledTimes(1);
+    expect(notify).toHaveBeenCalledWith(s, 'stuck');
+  });
+
+  it('covers ATTACHED sessions, not only detached ones (AC-A5.3)', () => {
+    const { service, notify } = createMockNotify();
+    const mgr = createSessionManager(CFG, service);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+    // still attached: clients.size === 1, detachedAt === null
+
+    mgr.sweepStuck(1_000 + CFG.stuckTtlMs + 1);
+
+    expect(s.detachedAt).toBeNull();
+    expect(s.claudeStatus).toBe('stuck');
+    expect(notify).toHaveBeenCalledTimes(1);
+  });
+
+  it('alerts at most once per silent round (AC-A5.1)', () => {
+    const { service, notify } = createMockNotify();
+    const mgr = createSessionManager(CFG, service);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+
+    mgr.sweepStuck(1_000 + CFG.stuckTtlMs + 1);
+    mgr.sweepStuck(1_000 + CFG.stuckTtlMs + 9_999);
+
+    expect(notify).toHaveBeenCalledTimes(1);
+    expect(s.claudeStatus).toBe('stuck');
+  });
+
+  it('re-arms after output recovery so a later silence alerts again (AC-A5.2)', () => {
+    const { service, notify } = createMockNotify();
+    const mgr = createSessionManager(CFG, service);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+
+    mgr.sweepStuck(1_000 + CFG.stuckTtlMs + 1);
+    expect(notify).toHaveBeenCalledTimes(1);
+
+    // Output recovery (session.ts onData) re-arms the flag + status.
+    (s.pty as MockIPty).emitData('back to work');
+    expect(s.stuckNotified).toBe(false);
+    expect(s.claudeStatus).toBe('working');
+    s.lastOutputAt = 100_000; // pin to a known cursor
+
+    mgr.sweepStuck(100_000 + CFG.stuckTtlMs + 1);
+    expect(notify).toHaveBeenCalledTimes(2);
+    expect(s.claudeStatus).toBe('stuck');
+  });
+
+  it('skips idle sessions', () => {
+    const { service, notify } = createMockNotify();
+    const mgr = createSessionManager(CFG, service);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+    s.claudeStatus = 'idle';
+
+    mgr.sweepStuck(1_000 + CFG.stuckTtlMs + 1);
+
+    expect(s.claudeStatus).toBe('idle');
+    expect(notify).not.toHaveBeenCalled();
+  });
+
+  it('skips already-exited sessions (kept in table after detach-then-exit)', () => {
+    const { service, notify } = createMockNotify();
+    const mgr = createSessionManager(CFG, service);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+    s.clients.clear();
+    s.detachedAt = 2_000;
+    (s.pty as MockIPty).emitExit(0); // detached-then-exit → stays in table (L1)
+
+    mgr.sweepStuck(1_000 + CFG.stuckTtlMs + 1);
+
+    expect(s.claudeStatus).not.toBe('stuck');
+    expect(notify).not.toHaveBeenCalled();
+  });
+
+  it('does not mark a session whose output is within the TTL window', () => {
+    const { service, notify } = createMockNotify();
+    const mgr = createSessionManager(CFG, service);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+
+    mgr.sweepStuck(1_000 + CFG.stuckTtlMs - 1);
+
+    expect(s.claudeStatus).not.toBe('stuck');
+    expect(notify).not.toHaveBeenCalled();
+  });
+
+  it('is disabled when stuckAlert is false', () => {
+    const { service, notify } = createMockNotify();
+    const mgr = createSessionManager({ ...CFG, stuckAlert: false }, service);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+
+    mgr.sweepStuck(1_000 + CFG.stuckTtlMs + 1);
+
+    expect(s.claudeStatus).not.toBe('stuck');
+    expect(notify).not.toHaveBeenCalled();
+  });
+
+  it('is disabled when stuckTtlMs is 0', () => {
+    const { service, notify } = createMockNotify();
+    const mgr = createSessionManager({ ...CFG, stuckTtlMs: 0 }, service);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+
+    mgr.sweepStuck(99_999_999);
+
+    expect(s.claudeStatus).not.toBe('stuck');
+    expect(notify).not.toHaveBeenCalled();
+  });
+
+  it('swallows a notifyService rejection (logs via console.error, never throws)', async () => {
+    const errSpy = vi.spyOn(console, 'error').mockImplementation(() => {});
+    const notify = vi.fn(async () => {
+      throw new Error('push transport down');
+    });
+    const service: NotifyService = { isEnabled: () => true, notify };
+    const mgr = createSessionManager(CFG, service);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+
+    expect(() => mgr.sweepStuck(1_000 + CFG.stuckTtlMs + 1)).not.toThrow();
+    await Promise.resolve();
+    await Promise.resolve();
+
+    expect(s.claudeStatus).toBe('stuck');
+    expect(errSpy).toHaveBeenCalled();
+    errSpy.mockRestore();
+  });
+
+  it('works without an injected notifyService (broadcasts, no throw)', () => {
+    const mgr = createSessionManager(CFG); // no notifyService
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+    ws.sent.length = 0;
+
+    expect(() => mgr.sweepStuck(1_000 + CFG.stuckTtlMs + 1)).not.toThrow();
+    expect(s.claudeStatus).toBe('stuck');
+    expect(parseSent(ws).some((m) => m.type === 'status' && m.status === 'stuck')).toBe(true);
+  });
+});
+
+// ── list() carries telemetry (B2 thumbnail wall) ──────────────────────────────
+describe('list — telemetry field', () => {
+  it('includes the latest telemetry for each session', () => {
+    const mgr = createSessionManager(CFG);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+    const telemetry = makeTelemetry();
+    mgr.handleStatusLine(s.meta.id, telemetry);
+
+    const entry = mgr.list().find((e) => e.id === s.meta.id);
+    expect(entry?.telemetry).toBe(telemetry);
+  });
+
+  it('reports null telemetry before the first statusLine report', () => {
+    const mgr = createSessionManager(CFG);
+    const ws = createMockWs();
+    const s = mgr.handleAttach(ws, null, DIMS, 1_000);
+
+    const entry = mgr.list().find((e) => e.id === s.meta.id);
+    expect(entry?.telemetry).toBeNull();
+  });
+});