newmark-agent 0.4.6 → 0.4.8

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (43) hide show
  1. package/assets/app-icon-dark.svg +6 -0
  2. package/dist/assets/app-icon-dark.svg +6 -0
  3. package/dist/cli-commands.js +2 -1
  4. package/dist/cli-discovery.js +2 -0
  5. package/dist/conversation-utility-host.bundle.cjs +577 -166
  6. package/dist/conversation-utility-host.js +3 -0
  7. package/dist/core/agent.d.ts +40 -4
  8. package/dist/core/agent.js +251 -49
  9. package/dist/core/agentKernelRunner.d.ts +3 -0
  10. package/dist/core/agentKernelRunner.js +74 -91
  11. package/dist/core/config.js +1 -1
  12. package/dist/core/conversationKernel.d.ts +40 -0
  13. package/dist/core/conversationKernel.js +219 -8
  14. package/dist/core/electronUtilityAgentClient.d.ts +8 -0
  15. package/dist/core/electronUtilityAgentClient.js +5 -1
  16. package/dist/core/electronUtilityRuntimePool.d.ts +15 -0
  17. package/dist/core/electronUtilityRuntimePool.js +11 -0
  18. package/dist/core/installUpdate.js +11 -8
  19. package/dist/core/mobilePairing.d.ts +1 -0
  20. package/dist/core/mobilePairing.js +15 -1
  21. package/dist/core/subagent.d.ts +10 -3
  22. package/dist/core/subagent.js +23 -8
  23. package/dist/core/toolPolicy.js +14 -3
  24. package/dist/core/utilityAgentProtocol.d.ts +29 -1
  25. package/dist/core/utilityHostToolRouter.js +18 -0
  26. package/dist/core/wslAgentClient.d.ts +8 -0
  27. package/dist/core/wslAgentClient.js +5 -1
  28. package/dist/core/wslAgentProtocol.d.ts +18 -1
  29. package/dist/core/wslAgentRuntimePool.d.ts +15 -0
  30. package/dist/core/wslAgentRuntimePool.js +11 -0
  31. package/dist/launcher.js +14 -11
  32. package/dist/main.js +207 -9
  33. package/dist/providers/chat-completions.adapter.js +6 -2
  34. package/dist/providers/responses.adapter.js +1 -0
  35. package/dist/server.d.ts +33 -1
  36. package/dist/server.js +696 -48
  37. package/dist/toolchain/registry-seeder.js +3 -1
  38. package/dist/tools/index.js +57 -6
  39. package/dist/tools/nativeTools.js +2 -1
  40. package/dist/ui/index.html +88 -101
  41. package/dist/wsl-agent-host.bundle.cjs +577 -166
  42. package/dist/wsl-agent-host.js +3 -0
  43. package/package.json +4 -2
@@ -1,6 +1,6 @@
1
1
  import { BrowserControlRequest, BrowserControlResult } from './browserControl';
2
2
  import { BrowserUseRequest } from './browserUse';
3
- import { AgentPromptMessage, ConversationContextCompressOptions, ConversationKernelRunOptions, ConversationKernelRunResult, ConversationQueueMode, ConversationRuntimeState, ConversationStopResult } from './conversationKernel';
3
+ import { AgentPromptMessage, ConversationContextCompressOptions, ConversationKernelRunOptions, ConversationKernelRunResult, ConversationQueueAction, ConversationQueueMode, ConversationRuntimeState, ConversationStopResult } from './conversationKernel';
4
4
  import { ConversationRuntimeTarget, NormalizedConversationTarget } from './conversationTarget';
5
5
  import { AgentMode, AgentWorkEvent, ConversationInputEnvelope, GuideReceipt } from './types';
6
6
  import type { AutoRouteRatingResult, ConversationSnapshot } from './agent';
@@ -40,6 +40,10 @@ export type UtilityHostToolRequest = (UtilityHostToolRequestBase & {
40
40
  tool: 'browser_use';
41
41
  args: BrowserUseRequest;
42
42
  context: UtilityHostToolContext;
43
+ }) | (UtilityHostToolRequestBase & {
44
+ tool: 'screen_capture';
45
+ args: Record<string, unknown>;
46
+ context: UtilityHostToolContext;
43
47
  }) | (UtilityHostToolRequestBase & {
44
48
  tool: 'computer_use';
45
49
  args: Record<string, unknown>;
@@ -96,6 +100,20 @@ export type UtilityAgentRequest = {
96
100
  target: ConversationRuntimeTarget;
97
101
  envelope: ConversationInputEnvelope;
98
102
  };
103
+ } | {
104
+ id: string;
105
+ method: 'queue_action';
106
+ params: {
107
+ target: ConversationRuntimeTarget;
108
+ action: ConversationQueueAction;
109
+ input?: {
110
+ id?: string;
111
+ text?: string;
112
+ requestedMode?: string;
113
+ goalObjective?: string;
114
+ createdAt?: string;
115
+ };
116
+ };
99
117
  } | {
100
118
  id: string;
101
119
  method: 'checkpoint';
@@ -210,6 +228,16 @@ export interface UtilityAgentSnapshotResult {
210
228
  steering: string[];
211
229
  followUp: string[];
212
230
  };
231
+ queueItems?: Array<{
232
+ id: string;
233
+ text: string;
234
+ queueMode: 'steer' | 'followUp';
235
+ requestedMode?: string;
236
+ goalObjective?: string;
237
+ runId?: string;
238
+ createdAt: string;
239
+ }>;
240
+ queuePaused?: boolean;
213
241
  workEvents: AgentWorkEvent[];
214
242
  [key: string]: unknown;
215
243
  }
@@ -126,6 +126,24 @@ function createUtilityHostToolHandler(options) {
126
126
  throwIfAborted(signal);
127
127
  return result;
128
128
  }
129
+ if (request.tool === 'screen_capture') {
130
+ const args = request.args || {};
131
+ const target = String(args.target || '').toLowerCase() === 'application' ? 'application' : 'desktop';
132
+ const result = await (options.runComputer || computerUse_1.runComputerUse)({
133
+ action: target === 'application' ? 'app_observe' : 'observe',
134
+ appTarget: String(args.app_target || args.appTarget || ''),
135
+ windowHandle: target === 'application' ? String(args.app_target || args.appTarget || '') : '',
136
+ maxChars: Number(args.max_chars || args.maxChars || 30_000),
137
+ captureMaxWidth: Number(args.capture_max_width || args.captureMaxWidth),
138
+ captureMaxHeight: Number(args.capture_max_height || args.captureMaxHeight),
139
+ allowEphemeralVisionImage: request.context.allowEphemeralVisionImage === true,
140
+ workspacePath: request.target.workspacePath,
141
+ invocation: 'agent',
142
+ ownerId: `screen-capture:${request.target.runtimeKey}:${String(request.context.actorId || ROOT_AGENT_ACTOR_ID)}`,
143
+ });
144
+ throwIfAborted(signal);
145
+ return result;
146
+ }
129
147
  const args = request.args || {};
130
148
  const action = String(args.action || '').trim().toLowerCase();
131
149
  const trustedComputerUseContext = request.context;
@@ -1,4 +1,5 @@
1
1
  import { AgentMode, AgentWorkEvent, ConversationInputEnvelope, GuideReceipt } from './types';
2
+ import { ConversationQueueAction } from './conversationKernel';
2
3
  import { TerminalTakeoverEvent, TerminalTakeoverOwnerFilter, TerminalTakeoverState } from '../tools/terminalTakeover';
3
4
  import { WslAgentPromptRequest, WslAgentPromptResult, WslAutoRouteRatingResult, WslAgentWorkspace, WslAgentStopResult, WslConversationRewindResult, WslHostToolRequest } from './wslAgentProtocol';
4
5
  import { ConversationRuntimeTarget, NormalizedConversationTarget } from './conversationTarget';
@@ -71,6 +72,13 @@ export declare class WslAgentClient {
71
72
  snapshotTarget(target: ConversationRuntimeTarget): Promise<Record<string, unknown>>;
72
73
  rewind(target: ConversationRuntimeTarget, messageIndex: number): Promise<WslConversationRewindResult>;
73
74
  enqueueGuide(target: ConversationRuntimeTarget, envelope: ConversationInputEnvelope): Promise<GuideReceipt>;
75
+ queueAction(target: ConversationRuntimeTarget, action: ConversationQueueAction, input?: {
76
+ id?: string;
77
+ text?: string;
78
+ requestedMode?: string;
79
+ goalObjective?: string;
80
+ createdAt?: string;
81
+ }): Promise<Record<string, unknown>>;
74
82
  checkpoint(target: ConversationRuntimeTarget): Promise<Record<string, unknown>>;
75
83
  contextCompress(target: ConversationRuntimeTarget, options?: {
76
84
  keepRecent?: number;
@@ -313,6 +313,10 @@ class WslAgentClient {
313
313
  await this.start();
314
314
  return await this.request('guide', { target: await this.mapTarget(target), envelope }, 5_000);
315
315
  }
316
+ async queueAction(target, action, input = {}) {
317
+ await this.start();
318
+ return await this.request('queue_action', { target: await this.mapTarget(target), action, input }, 5_000);
319
+ }
316
320
  async checkpoint(target) {
317
321
  await this.start();
318
322
  return await this.request('checkpoint', { target: await this.mapTarget(target) }, 5_000);
@@ -536,7 +540,7 @@ class WslAgentClient {
536
540
  else if (!this.hostToolHandler) {
537
541
  result = { requestId: request.requestId, ok: false, error: 'No Windows host tool handler is registered' };
538
542
  }
539
- else if (!['browser_control', 'computer_use', 'browser_use', 'automation', 'terminal_takeover'].includes(request.tool)) {
543
+ else if (!['browser_control', 'screen_capture', 'computer_use', 'browser_use', 'automation', 'terminal_takeover'].includes(request.tool)) {
540
544
  result = { requestId: request.requestId, ok: false, error: `WSL host tool is not allowed: ${String(request.tool)}` };
541
545
  }
542
546
  else {
@@ -1,5 +1,5 @@
1
1
  import { AgentMode, AgentWorkEvent, ConversationInputEnvelope, GuideReceipt } from './types';
2
- import { AgentPromptMessage, ConversationContextCompressOptions, ConversationKernelRunOptions, ConversationKernelRunResult, ConversationQueueMode, ConversationStopResult } from './conversationKernel';
2
+ import { AgentPromptMessage, ConversationContextCompressOptions, ConversationKernelRunOptions, ConversationKernelRunResult, ConversationQueueAction, ConversationQueueMode, ConversationStopResult } from './conversationKernel';
3
3
  import { ConversationRuntimeTarget } from './conversationTarget';
4
4
  import { TerminalTakeoverEvent, TerminalTakeoverOwnerFilter, TerminalTakeoverState } from '../tools/terminalTakeover';
5
5
  import { BrowserUseRequest } from './browserUse';
@@ -42,6 +42,9 @@ interface WslHostToolRequestBase {
42
42
  export type WslHostToolRequest = (WslHostToolRequestBase & {
43
43
  tool: 'browser_control';
44
44
  args: BrowserControlRequest;
45
+ }) | (WslHostToolRequestBase & {
46
+ tool: 'screen_capture';
47
+ args: Record<string, unknown>;
45
48
  }) | (WslHostToolRequestBase & {
46
49
  tool: 'computer_use';
47
50
  args: Record<string, unknown>;
@@ -108,6 +111,20 @@ export type WslAgentRequest = {
108
111
  target: ConversationRuntimeTarget;
109
112
  envelope: ConversationInputEnvelope;
110
113
  };
114
+ } | {
115
+ id: string;
116
+ method: 'queue_action';
117
+ params: {
118
+ target: ConversationRuntimeTarget;
119
+ action: ConversationQueueAction;
120
+ input?: {
121
+ id?: string;
122
+ text?: string;
123
+ requestedMode?: string;
124
+ goalObjective?: string;
125
+ createdAt?: string;
126
+ };
127
+ };
111
128
  } | {
112
129
  id: string;
113
130
  method: 'checkpoint';
@@ -1,4 +1,5 @@
1
1
  import { AgentMode, AgentWorkEvent, ConversationInputEnvelope, GuideReceipt } from './types';
2
+ import { ConversationQueueAction } from './conversationKernel';
2
3
  import { ConversationRuntimeTarget, NormalizedConversationTarget } from './conversationTarget';
3
4
  import { WslHostToolHandler } from './wslAgentClient';
4
5
  import { WslAgentPromptRequest, WslAgentPromptResult, WslAutoRouteRatingResult, WslAgentStopResult, WslConversationRewindResult } from './wslAgentProtocol';
@@ -10,6 +11,13 @@ export interface WslTargetRuntimeClient {
10
11
  rewind(target: ConversationRuntimeTarget, messageIndex: number): Promise<WslConversationRewindResult>;
11
12
  requestStop(target: ConversationRuntimeTarget, runId?: string): Promise<WslAgentStopResult>;
12
13
  enqueueGuide(target: ConversationRuntimeTarget, envelope: ConversationInputEnvelope): Promise<GuideReceipt>;
14
+ queueAction?(target: ConversationRuntimeTarget, action: ConversationQueueAction, input?: {
15
+ id?: string;
16
+ text?: string;
17
+ requestedMode?: string;
18
+ goalObjective?: string;
19
+ createdAt?: string;
20
+ }): Promise<Record<string, unknown>>;
13
21
  checkpoint(target: ConversationRuntimeTarget): Promise<Record<string, unknown>>;
14
22
  contextCompress?(target: ConversationRuntimeTarget, options?: {
15
23
  keepRecent?: number;
@@ -73,6 +81,13 @@ export declare class WslAgentRuntimePool {
73
81
  rewind(target: ConversationRuntimeTarget, messageIndex: number): Promise<WslConversationRewindResult>;
74
82
  requestStop(target: ConversationRuntimeTarget, runId?: string): Promise<WslPoolStopResult>;
75
83
  enqueueGuide(envelope: ConversationInputEnvelope): Promise<GuideReceipt>;
84
+ queueAction(target: ConversationRuntimeTarget, action: ConversationQueueAction, input?: {
85
+ id?: string;
86
+ text?: string;
87
+ requestedMode?: string;
88
+ goalObjective?: string;
89
+ createdAt?: string;
90
+ }): Promise<Record<string, unknown>>;
76
91
  checkpoint(target: ConversationRuntimeTarget): Promise<Record<string, unknown>>;
77
92
  contextCompress(target: ConversationRuntimeTarget, options?: {
78
93
  keepRecent?: number;
@@ -195,6 +195,17 @@ class WslAgentRuntimePool {
195
195
  this.release(entry, true);
196
196
  }
197
197
  }
198
+ async queueAction(target, action, input = {}) {
199
+ const entry = await this.acquireExisting(target);
200
+ if (!entry || !entry.client.queueAction)
201
+ throw new Error('Target conversation is not running');
202
+ try {
203
+ return await entry.client.queueAction(entry.target, action, input);
204
+ }
205
+ finally {
206
+ this.release(entry, true);
207
+ }
208
+ }
198
209
  async checkpoint(target) {
199
210
  const normalized = (0, conversationTarget_1.normalizeConversationTarget)(target);
200
211
  const entry = await this.acquireExisting(normalized);
package/dist/launcher.js CHANGED
@@ -312,6 +312,19 @@ if (isGui) {
312
312
  launchGui();
313
313
  }
314
314
  else if (isTui) {
315
+ // 远程触及开关开启 → 后端托管启动 mobile server(不阻塞 TUI 其他功能;TUI 退出即随终端托管结束)
316
+ try {
317
+ const { ConfigManager } = require('./core/config');
318
+ const tuiConfig = new ConfigManager(root);
319
+ if (tuiConfig.getBool('remote', 'touch_enabled')) {
320
+ const { runServer } = require('./server');
321
+ runServer(root);
322
+ console.log('[Newmark] mobile server hosted (remote touch enabled)');
323
+ }
324
+ }
325
+ catch (error) {
326
+ console.error('[Newmark] TUI hosted mobile server failed:', error instanceof Error ? error.message : String(error));
327
+ }
315
328
  const { start } = require('./tui/src/app');
316
329
  start({ root, workspacePath: resolveTuiWorkspacePath(args, root), desktopDist: __dirname });
317
330
  }
@@ -363,18 +376,8 @@ else if (isCli || (!isServer && isTerminal())) {
363
376
  runCli(root);
364
377
  }
365
378
  else {
366
- // Server mode - start HTTP server and open browser
379
+ // Server mode - start HTTP server(不绑定启动浏览器)
367
380
  const { runServer } = require('./server');
368
381
  runServer(root);
369
- if (!args.includes('--no-browser')) {
370
- const { exec } = require('child_process');
371
- const port = 47890;
372
- const cmd = process.platform === 'win32'
373
- ? `start http://localhost:${port}`
374
- : process.platform === 'darwin'
375
- ? `open http://localhost:${port}`
376
- : `xdg-open http://localhost:${port}`;
377
- exec(cmd);
378
- }
379
382
  }
380
383
  //# sourceMappingURL=launcher.js.map
package/dist/main.js CHANGED
@@ -120,6 +120,7 @@ let _forceQuit = false;
120
120
  let forcedExitTimer = null;
121
121
  let electronBrowserUseHost = null;
122
122
  let browserUseEngine = null;
123
+ const mobileServerWorkEventSubscribers = new Set();
123
124
  // A single renderer can host several hidden conversation-bound guests. The
124
125
  // legacy host map is retained for focused-window discovery, while this map is
125
126
  // the authoritative Browser-Use/right-sidebar binding.
@@ -288,7 +289,7 @@ async function resetWslAgentClient() {
288
289
  if (wslAgentClient)
289
290
  await wslAgentClient.resetAgent();
290
291
  }
291
- function broadcastAgentWorkEvent(event) {
292
+ function broadcastAgentWorkEvent(event, mirrorToMobile = true) {
292
293
  const workEvent = event;
293
294
  if ((workEvent.type === 'done' || workEvent.type === 'error') && workEvent.runtimeKey) {
294
295
  browserUseEngine?.clearRuntime(workEvent.runtimeKey);
@@ -299,6 +300,14 @@ function broadcastAgentWorkEvent(event) {
299
300
  continue;
300
301
  win.webContents.send('agent:workEvent', event);
301
302
  }
303
+ if (mirrorToMobile) {
304
+ for (const listener of mobileServerWorkEventSubscribers) {
305
+ try {
306
+ listener(event);
307
+ }
308
+ catch { /* ignore disconnected mobile bridge */ }
309
+ }
310
+ }
302
311
  }
303
312
  function ensureConversationKernel(root) {
304
313
  if (!agent)
@@ -420,9 +429,10 @@ function relayFlowAgentWorkEvents(flowAgent, target) {
420
429
  });
421
430
  }
422
431
  function themedAppIconPath() {
423
- const themeName = electron_1.nativeTheme.shouldUseDarkColors ? 'light' : 'dark';
424
- const compactPath = path.join(__dirname, 'assets', `app-icon-${themeName}-64.png`);
425
- return fs.existsSync(compactPath) ? compactPath : appAssetPath(`app-icon-${themeName}.png`);
432
+ // Fixed black-on-white icon across every theme: the titlebar, taskbar, tray,
433
+ // and Windows executable all share the same dark icon asset.
434
+ const compactPath = path.join(__dirname, 'assets', 'app-icon-dark-64.png');
435
+ return fs.existsSync(compactPath) ? compactPath : appAssetPath('app-icon-dark.png');
426
436
  }
427
437
  function createAppIconImage(size) {
428
438
  const image = electron_1.nativeImage.createFromPath(themedAppIconPath());
@@ -1881,6 +1891,190 @@ else {
1881
1891
  ensureWorkspaceRegistryWatcher();
1882
1892
  restoreStoredFlowSuspension();
1883
1893
  recordStartup('agent-ready');
1894
+ // 远程触及开关开启 → GUI 进程内托管启动 mobile server(托盘常驻不中断)
1895
+ if (agent.config.getBool('remote', 'touch_enabled')) {
1896
+ try {
1897
+ const { runServer } = require('./server');
1898
+ runServer(root, {
1899
+ agent,
1900
+ automation,
1901
+ onWorkEvent: event => broadcastAgentWorkEvent(event, false),
1902
+ subscribeWorkEvents: listener => {
1903
+ mobileServerWorkEventSubscribers.add(listener);
1904
+ return () => mobileServerWorkEventSubscribers.delete(listener);
1905
+ },
1906
+ conversationUiState: async (requested) => {
1907
+ const target = conversationRuntimeTarget(requested);
1908
+ const snapshot = wslBackendEnabled()
1909
+ ? await ensureWslConversationPool().snapshot(target)
1910
+ : await ensureElectronUtilityPool().snapshot(target);
1911
+ const flowRunning = flowRunningForTarget(target);
1912
+ const flowSuspension = flowSuspensionForTarget(target);
1913
+ return {
1914
+ ...snapshot,
1915
+ flow: flowRunning ? {
1916
+ running: true,
1917
+ paused: false,
1918
+ name: flowRunning.name,
1919
+ promptText: String(activeFlowStateFor(target)?.input || ''),
1920
+ message: '',
1921
+ } : flowSuspension ? {
1922
+ running: true,
1923
+ paused: true,
1924
+ name: flowSuspension.workflowName || '',
1925
+ promptText: String(flowSuspension.input || ''),
1926
+ message: String(flowSuspension.message || ''),
1927
+ reason: flowSuspension.reason || '',
1928
+ } : null,
1929
+ };
1930
+ },
1931
+ conversationUiAction: async (requested, action, value, input) => {
1932
+ const target = conversationRuntimeTarget(requested);
1933
+ if (action.startsWith('queue_')) {
1934
+ const queueAction = action === 'queue_enqueue' ? 'enqueue'
1935
+ : action === 'queue_update' ? 'update'
1936
+ : action === 'queue_delete' ? 'delete'
1937
+ : action === 'queue_toggle_pause' ? 'toggle_pause'
1938
+ : 'guide';
1939
+ const queueInput = {
1940
+ id: String(input?.id || ''),
1941
+ text: String(input?.text || value || ''),
1942
+ requestedMode: String(input?.requestedMode || 'build'),
1943
+ goalObjective: String(input?.goalObjective || ''),
1944
+ createdAt: String(input?.createdAt || ''),
1945
+ };
1946
+ return wslBackendEnabled()
1947
+ ? await ensureWslConversationPool().queueAction(target, queueAction, queueInput)
1948
+ : await ensureElectronUtilityPool().queueAction(target, queueAction, queueInput);
1949
+ }
1950
+ if (action === 'goal_update') {
1951
+ return { goal: await mutateTargetConversation(target, () => {
1952
+ const isolated = isolatedConversationAgent(target);
1953
+ isolated.updateGoal(String(value || '').trim());
1954
+ isolated.setMode('goal');
1955
+ isolated.saveWorkspaceConversationState(true);
1956
+ return isolated.getConversationSnapshot(target.conversationId).goal;
1957
+ }) };
1958
+ }
1959
+ if (action === 'goal_guide' || action === 'conversation_guide') {
1960
+ const objective = action === 'goal_guide' ? String(value || '').trim() : '';
1961
+ const guideText = String(value || '').trim();
1962
+ const snapshot = wslBackendEnabled()
1963
+ ? await ensureWslConversationPool().snapshot(target)
1964
+ : await ensureElectronUtilityPool().snapshot(target);
1965
+ const runtime = snapshot.runtime;
1966
+ if (!guideText || !runtime?.running || !runtime.runId) {
1967
+ return { ok: false, error: action === 'goal_guide'
1968
+ ? 'There is no active Build for this Goal Guide.'
1969
+ : 'Target conversation is not running.' };
1970
+ }
1971
+ const prefix = 'Goal for the current Build:\n';
1972
+ const now = new Date().toISOString();
1973
+ const envelope = {
1974
+ clientMessageId: (0, crypto_1.randomUUID)(),
1975
+ guideId: (0, crypto_1.randomUUID)(),
1976
+ target: { workspaceId: target.workspaceId, conversationId: target.conversationId },
1977
+ runId: runtime.runId,
1978
+ deliveryMode: 'steer',
1979
+ text: objective ? prefix + objective : guideText,
1980
+ goalObjective: objective || undefined,
1981
+ createdAt: now,
1982
+ };
1983
+ const receipt = wslBackendEnabled()
1984
+ ? await ensureWslConversationPool().enqueueGuide(envelope)
1985
+ : await ensureElectronUtilityPool().enqueueGuide(envelope);
1986
+ return { ok: receipt.status !== 'rejected', receipt };
1987
+ }
1988
+ if (action === 'goal_toggle_pause') {
1989
+ const resident = wslBackendEnabled()
1990
+ ? await ensureWslConversationPool().toggleGoalPause(target)
1991
+ : await ensureElectronUtilityPool().toggleGoalPause(target);
1992
+ const paused = resident !== null ? resident : await mutateTargetConversation(target, () => isolatedConversationAgent(target).toggleGoalPause());
1993
+ return { ok: true, paused };
1994
+ }
1995
+ if (action === 'goal_clear') {
1996
+ const resident = wslBackendEnabled()
1997
+ ? await ensureWslConversationPool().clearGoal(target)
1998
+ : await ensureElectronUtilityPool().clearGoal(target);
1999
+ if (resident === null)
2000
+ await mutateTargetConversation(target, () => isolatedConversationAgent(target).clearGoal());
2001
+ return { ok: true, cleared: true };
2002
+ }
2003
+ if (action === 'flow_pause')
2004
+ return await stopFlowForTarget(target);
2005
+ if (action === 'flow_resume')
2006
+ return await resumeFlowForTarget(String(value || ''), target);
2007
+ if (action === 'conversation_stop') {
2008
+ const snapshot = wslBackendEnabled()
2009
+ ? await ensureWslConversationPool().snapshot(target)
2010
+ : await ensureElectronUtilityPool().snapshot(target);
2011
+ const runtime = snapshot.runtime;
2012
+ return wslBackendEnabled()
2013
+ ? await ensureWslConversationPool().requestStop(target, runtime?.runId)
2014
+ : await ensureElectronUtilityPool().requestStop(target, runtime?.runId);
2015
+ }
2016
+ if (action === 'input_mode') {
2017
+ const mode = String(value || '') === 'next' ? 'next' : 'guide';
2018
+ const selected = wslBackendEnabled()
2019
+ ? await ensureWslConversationPool().setInputMode(target, mode)
2020
+ : await ensureElectronUtilityPool().setInputMode(target, mode);
2021
+ return { ok: true, inputMode: selected || mode };
2022
+ }
2023
+ const flow = activeFlowStateFor(target);
2024
+ if (!flow?.abortController || !flow.flowAgent)
2025
+ return { ok: false, error: 'No active Flow accepts Guide input.' };
2026
+ const accepted = flow.flowAgent.queueActiveKernelMessage(String(value || '').trim(), 'steer') || false;
2027
+ return accepted ? { ok: true, accepted: true, flow: flow.name } : { ok: false, error: 'The current Flow Build is not accepting Guide input.' };
2028
+ },
2029
+ conversationPrompt: async (requested, message, requestedOptions) => {
2030
+ const target = conversationRuntimeTarget(requested);
2031
+ if (activeFlowStateFor(target))
2032
+ clearFlowSuspensionForNewWork(target);
2033
+ const snapshot = wslBackendEnabled()
2034
+ ? await ensureWslConversationPool().snapshot(target)
2035
+ : await ensureElectronUtilityPool().snapshot(target);
2036
+ const requestedMode = String(requestedOptions?.requestedMode || '').toLowerCase();
2037
+ const goalObjective = String(requestedOptions?.goalObjective || '').trim();
2038
+ const mode = (goalObjective || requestedMode === 'goal') ? 'build' : String(snapshot.mode || 'build');
2039
+ const inputMode = String(requestedOptions?.inputMode || snapshot.inputMode || 'guide') === 'next' ? 'next' : 'guide';
2040
+ const options = {
2041
+ mode,
2042
+ model: String(snapshot.model || agent.ensureUsableModelSelection()),
2043
+ intelligence: String(snapshot.intelligence || agent.intelligence),
2044
+ inputMode,
2045
+ engine: agent.engine,
2046
+ };
2047
+ const queueMode = inputMode === 'guide' ? 'steer' : 'followUp';
2048
+ const promptMessage = goalObjective ? {
2049
+ text: message,
2050
+ visibleMode: 'goal',
2051
+ goalObjective,
2052
+ } : message;
2053
+ const result = wslBackendEnabled()
2054
+ ? await ensureWslConversationPool().prompt({
2055
+ message: promptMessage,
2056
+ target,
2057
+ conversationId: target.conversationId,
2058
+ options,
2059
+ queueMode,
2060
+ workspace: target.workspace ? {
2061
+ id: target.workspace.id,
2062
+ name: target.workspace.name,
2063
+ path: target.workspace.path,
2064
+ isInternal: target.workspace.isInternal,
2065
+ kind: target.workspace.kind,
2066
+ } : null,
2067
+ })
2068
+ : await ensureElectronUtilityPool().prompt({ message: promptMessage, target, options, queueMode });
2069
+ return { ...result };
2070
+ },
2071
+ });
2072
+ recordStartup('mobile-server-hosted');
2073
+ }
2074
+ catch (error) {
2075
+ console.error('[Newmark] hosted mobile server failed:', error instanceof Error ? error.message : String(error));
2076
+ }
2077
+ }
1884
2078
  }
1885
2079
  };
1886
2080
  const localConversationSnapshotForStartup = (target) => {
@@ -1964,6 +2158,8 @@ else {
1964
2158
  return result.tokens.map(token => token.text).join('');
1965
2159
  });
1966
2160
  startupAgent.setAutomationManager(automation);
2161
+ const { setHostedServerAutomation } = require('./server');
2162
+ setHostedServerAutomation(automation);
1967
2163
  automation.onChange(items => {
1968
2164
  setTimeout(() => {
1969
2165
  try {
@@ -2664,7 +2860,7 @@ else {
2664
2860
  return await utilityHostToolHandler({ ...request, target: routedTarget, context: routedContext }, signal);
2665
2861
  }
2666
2862
  const result = await utilityHostToolHandler({ ...request, target: routedTarget, context: routedContext }, signal);
2667
- return request.tool === 'computer_use' ? prepareWslComputerUseVisionResult(result) : result;
2863
+ return request.tool === 'computer_use' || request.tool === 'screen_capture' ? prepareWslComputerUseVisionResult(result) : result;
2668
2864
  };
2669
2865
  runWslHostTool.cancelTarget = (runtimeKey) => utilityHostToolHandler.cancelTarget(runtimeKey);
2670
2866
  const ensureWslConversationPool = () => {
@@ -3024,7 +3220,7 @@ else {
3024
3220
  }
3025
3221
  }
3026
3222
  });
3027
- electron_1.ipcMain.handle('flow:resume', async (_event, response, targetInput) => {
3223
+ const resumeFlowForTarget = async (response, targetInput) => {
3028
3224
  if (!agent)
3029
3225
  return { ok: false, error: 'No suspended Flow is waiting for user input.' };
3030
3226
  const requestedTarget = targetInput ? conversationRuntimeTarget(targetInput) : null;
@@ -3158,7 +3354,8 @@ else {
3158
3354
  await setTargetFlowMode(flowTarget, suspension.previousMode);
3159
3355
  }
3160
3356
  }
3161
- });
3357
+ };
3358
+ electron_1.ipcMain.handle('flow:resume', async (_event, response, targetInput) => await resumeFlowForTarget(response, targetInput));
3162
3359
  electron_1.ipcMain.handle('agent:setMode', async (_event, mode) => {
3163
3360
  if (agent) {
3164
3361
  const nextMode = mode;
@@ -3731,7 +3928,7 @@ else {
3731
3928
  const accepted = state.flowAgent.queueActiveKernelMessage(text, 'steer') || false;
3732
3929
  return accepted ? { ok: true, accepted: true, flow: state.name } : { ok: false, error: 'The current Flow Build is not accepting Guide input.' };
3733
3930
  });
3734
- electron_1.ipcMain.handle('flow:stop', async (_event, targetInput) => {
3931
+ const stopFlowForTarget = async (targetInput) => {
3735
3932
  if (!agent)
3736
3933
  return { ok: true, action: 'not_running' };
3737
3934
  const target = targetInput
@@ -3769,7 +3966,8 @@ else {
3769
3966
  controller.abort(new Error(`Flow interrupted by user: ${flowName}`));
3770
3967
  state.flowAgent?.abortActiveKernelRun();
3771
3968
  return { ok: true, action: 'stopping', flow: flowName };
3772
- });
3969
+ };
3970
+ electron_1.ipcMain.handle('flow:stop', async (_event, targetInput) => await stopFlowForTarget(targetInput));
3773
3971
  electron_1.ipcMain.handle('agent:readGlobalPrompt', async () => {
3774
3972
  if (!agent)
3775
3973
  return { error: 'Agent is not initialized' };
@@ -38,14 +38,18 @@ class ChatCompletionsAdapter {
38
38
  ...(request.systemPrompt ? [{ role: CHAT_SYSTEM_ROLE, content: request.systemPrompt }] : []),
39
39
  ...(0, chat_messages_1.openAIChatMessages)(request.messages),
40
40
  ];
41
+ const tools = this.serializeTools(request.tools);
41
42
  const body = {
42
43
  model: request.model,
43
44
  messages,
44
45
  temperature: request.temperature,
45
46
  max_tokens: request.maxOutputTokens,
46
- tools: this.serializeTools(request.tools),
47
- tool_choice: 'auto',
48
47
  };
48
+ if (tools.length) {
49
+ body.tools = tools;
50
+ body.tool_choice = 'auto';
51
+ body.parallel_tool_calls = true;
52
+ }
49
53
  if (request.reasoningEffort)
50
54
  body.reasoning_effort = request.reasoningEffort;
51
55
  // 会话标识透传:仅当上层(支持 session_id 语义的 provider)显式填充时写进
@@ -78,6 +78,7 @@ class ResponsesAdapter {
78
78
  if (tools.length) {
79
79
  body.tools = tools;
80
80
  body.tool_choice = 'auto';
81
+ body.parallel_tool_calls = true;
81
82
  }
82
83
  const base = request.baseUrl.replace(/\/+$/, '');
83
84
  return {
package/dist/server.d.ts CHANGED
@@ -1,2 +1,34 @@
1
- export declare function runServer(root: string): void;
1
+ import { Agent, AgentWorkEvent } from './core/agent';
2
+ import { AutomationManager } from './core/automation';
3
+ export interface HostedMobileServerOptions {
4
+ agent?: Agent;
5
+ automation?: AutomationManager | null;
6
+ /** Mobile-originated runs must reach the already-open desktop renderer. */
7
+ onWorkEvent?: (event: AgentWorkEvent) => void;
8
+ /** Desktop GUI/runtime-pool events must reach mobile SSE subscribers. */
9
+ subscribeWorkEvents?: (listener: (event: AgentWorkEvent) => void) => (() => void);
10
+ /** Read the GUI runtime-pool state for one exact workspace/conversation. */
11
+ conversationUiState?: (target: {
12
+ workspaceId: string;
13
+ conversationId: string;
14
+ }) => Promise<Record<string, unknown>>;
15
+ /** Mutate Goal/Flow state owned by the GUI runtime pool for one exact target. */
16
+ conversationUiAction?: (target: {
17
+ workspaceId: string;
18
+ conversationId: string;
19
+ }, action: 'goal_update' | 'goal_guide' | 'conversation_guide' | 'goal_toggle_pause' | 'goal_clear' | 'flow_pause' | 'flow_resume' | 'flow_guide' | 'conversation_stop' | 'input_mode' | 'queue_enqueue' | 'queue_update' | 'queue_delete' | 'queue_toggle_pause' | 'queue_guide', value?: string, input?: Record<string, unknown>) => Promise<Record<string, unknown>>;
20
+ /** Send through the same GUI runtime pool used by the desktop renderer. */
21
+ conversationPrompt?: (target: {
22
+ workspaceId: string;
23
+ conversationId: string;
24
+ }, message: string, options?: {
25
+ requestedMode?: string;
26
+ goalObjective?: string;
27
+ inputMode?: string;
28
+ }) => Promise<Record<string, unknown>>;
29
+ }
30
+ /** 托管启动 server(GUI/TUI 内嵌调用;幂等防重入,进程常驻即服务常驻) */
31
+ export declare function runServer(root: string, options?: HostedMobileServerOptions): void;
32
+ /** Attach the GUI-owned automation manager after deferred startup. */
33
+ export declare function setHostedServerAutomation(manager: AutomationManager): void;
2
34
  //# sourceMappingURL=server.d.ts.map