@midscene/core 0.18.1-beta-20250611082446.0 → 0.18.1

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 (36) hide show
  1. package/dist/es/ai-model.d.ts +52 -4
  2. package/dist/es/ai-model.js +9 -1
  3. package/dist/es/{chunk-D4ZWVRRG.js → chunk-A7OTPIZJ.js} +19 -6
  4. package/dist/es/chunk-A7OTPIZJ.js.map +1 -0
  5. package/dist/es/{chunk-X2UALJ3B.js → chunk-EYIL4AHP.js} +353 -3
  6. package/dist/es/chunk-EYIL4AHP.js.map +1 -0
  7. package/dist/es/index.d.ts +4 -4
  8. package/dist/es/index.js +19 -26
  9. package/dist/es/index.js.map +1 -1
  10. package/dist/es/{llm-planning-453ffce5.d.ts → llm-planning-a951deb9.d.ts} +2 -2
  11. package/dist/es/{types-00c9086f.d.ts → types-dce56c26.d.ts} +6 -6
  12. package/dist/es/utils.d.ts +1 -1
  13. package/dist/es/utils.js +1 -1
  14. package/dist/lib/ai-model.d.ts +52 -4
  15. package/dist/lib/ai-model.js +10 -2
  16. package/dist/lib/{chunk-D4ZWVRRG.js → chunk-A7OTPIZJ.js} +18 -5
  17. package/dist/lib/chunk-A7OTPIZJ.js.map +1 -0
  18. package/dist/lib/{chunk-X2UALJ3B.js → chunk-EYIL4AHP.js} +371 -21
  19. package/dist/lib/chunk-EYIL4AHP.js.map +1 -0
  20. package/dist/lib/index.d.ts +4 -4
  21. package/dist/lib/index.js +37 -44
  22. package/dist/lib/index.js.map +1 -1
  23. package/dist/lib/{llm-planning-453ffce5.d.ts → llm-planning-a951deb9.d.ts} +2 -2
  24. package/dist/{types/types-00c9086f.d.ts → lib/types-dce56c26.d.ts} +6 -6
  25. package/dist/lib/utils.d.ts +1 -1
  26. package/dist/lib/utils.js +2 -2
  27. package/dist/types/ai-model.d.ts +52 -4
  28. package/dist/types/index.d.ts +4 -4
  29. package/dist/types/{llm-planning-453ffce5.d.ts → llm-planning-a951deb9.d.ts} +2 -2
  30. package/dist/{lib/types-00c9086f.d.ts → types/types-dce56c26.d.ts} +6 -6
  31. package/dist/types/utils.d.ts +1 -1
  32. package/package.json +4 -3
  33. package/dist/es/chunk-D4ZWVRRG.js.map +0 -1
  34. package/dist/es/chunk-X2UALJ3B.js.map +0 -1
  35. package/dist/lib/chunk-D4ZWVRRG.js.map +0 -1
  36. package/dist/lib/chunk-X2UALJ3B.js.map +0 -1
package/dist/lib/index.js CHANGED
@@ -1,6 +1,6 @@
1
1
  "use strict";Object.defineProperty(exports, "__esModule", {value: true}); function _optionalChain(ops) { let lastAccessLHS = undefined; let value = ops[0]; let i = 1; while (i < ops.length) { const op = ops[i]; const fn = ops[i + 1]; i += 2; if ((op === 'optionalAccess' || op === 'optionalCall') && value == null) { return undefined; } if (op === 'access' || op === 'optionalAccess') { lastAccessLHS = value; value = fn(value); } else if (op === 'call' || op === 'optionalCall') { value = fn((...args) => value.call(lastAccessLHS, ...args)); lastAccessLHS = undefined; } } return value; }
2
2
 
3
- var _chunkD4ZWVRRGjs = require('./chunk-D4ZWVRRG.js');
3
+ var _chunkA7OTPIZJjs = require('./chunk-A7OTPIZJ.js');
4
4
 
5
5
 
6
6
 
@@ -11,7 +11,7 @@ var _chunkD4ZWVRRGjs = require('./chunk-D4ZWVRRG.js');
11
11
 
12
12
 
13
13
 
14
- var _chunkX2UALJ3Bjs = require('./chunk-X2UALJ3B.js');
14
+ var _chunkEYIL4AHPjs = require('./chunk-EYIL4AHP.js');
15
15
 
16
16
  // src/ai-model/action-executor.ts
17
17
 
@@ -121,11 +121,10 @@ ${_optionalChain([this, 'access', _7 => _7.latestErrorTask, 'call', _8 => _8(),
121
121
  task.status = "finished";
122
122
  task.timing.end = Date.now();
123
123
  task.timing.cost = task.timing.end - task.timing.start;
124
- task.timing.aiCost = _optionalChain([returnValue, 'optionalAccess', _12 => _12.aiCost]) || 0;
125
124
  taskIndex++;
126
125
  } catch (e) {
127
126
  successfullyCompleted = false;
128
- task.error = _optionalChain([e, 'optionalAccess', _13 => _13.message]) || (typeof e === "string" ? e : "error-without-message");
127
+ task.error = _optionalChain([e, 'optionalAccess', _12 => _12.message]) || (typeof e === "string" ? e : "error-without-message");
129
128
  task.errorStack = e.stack;
130
129
  task.status = "failed";
131
130
  task.timing.end = Date.now();
@@ -172,7 +171,7 @@ ${_optionalChain([this, 'access', _7 => _7.latestErrorTask, 'call', _8 => _8(),
172
171
  }
173
172
  }
174
173
  const dumpData = {
175
- sdkVersion: _chunkD4ZWVRRGjs.getVersion.call(void 0, ),
174
+ sdkVersion: _chunkA7OTPIZJjs.getVersion.call(void 0, ),
176
175
  model_name: _env.getAIConfig.call(void 0, _env.MIDSCENE_MODEL_NAME) || "",
177
176
  model_description: modelDescription,
178
177
  logTime: Date.now(),
@@ -186,25 +185,21 @@ ${_optionalChain([this, 'access', _7 => _7.latestErrorTask, 'call', _8 => _8(),
186
185
  // src/ai-model/prompt/describe.ts
187
186
 
188
187
  var elementDescriberInstruction = () => {
189
- return `Tell what is the content of the element wrapped by the read rectangle in the screenshot. Your description is expected to be used to precisely locate the element from other similar elements on screenshot. Use ${_env.getPreferredLanguage.call(void 0, )} in the description.
190
-
191
- Please follow the following rules:
192
- 1. The description should be start with a brief description, like "a button for confirming the action".
193
-
194
- 2. Include these information in the description to distinguish the element from its siblings and other similar elements, as much as possible:
195
- - The text of the element, like "with text 'Confirm'"
196
- - What the element looks like if it's an image, like "with image '...'"
197
- - The relative position of the element, like "on the left of ..., around ..."
198
- - How to distinguish the element from its siblings elements, like "it is the icon instead of the text"
199
-
200
- 3. Do NOT mention the red rectangle in the description.
201
-
202
- 4. Use the error field to describe the unexpected situations, if any. If not, put null.
203
-
204
- Return in JSON:
188
+ return `Describe the element in the red rectangle for precise identification. Use ${_env.getPreferredLanguage.call(void 0, )}.
189
+
190
+ Rules:
191
+ 1. Start with element type (button, input, link, etc.)
192
+ 2. Include key identifiers:
193
+ - Text content: "with text 'Submit'"
194
+ - Visual features: "blue background", "icon only"
195
+ - Position: "top-right", "below search bar"
196
+ 3. Keep description under 20 words
197
+ 4. Don't mention the red rectangle
198
+
199
+ Return JSON:
205
200
  {
206
- "description": "[{brief description}]: {text of the element} {image of the element} {relative position of the element} ... ",
207
- "error"?: "..."
201
+ "description": "brief element type with key identifiers",
202
+ "error"?: "error message if any"
208
203
  }`;
209
204
  };
210
205
 
@@ -222,12 +217,9 @@ var _logger = require('@midscene/shared/logger');
222
217
  // src/insight/utils.ts
223
218
 
224
219
 
225
-
226
-
227
-
228
220
  function emitInsightDump(data, dumpSubscriber) {
229
221
  const baseData = {
230
- sdkVersion: _chunkD4ZWVRRGjs.getVersion.call(void 0, ),
222
+ sdkVersion: _chunkA7OTPIZJjs.getVersion.call(void 0, ),
231
223
  logTime: Date.now(),
232
224
  model_name: _env.getAIConfig.call(void 0, _env.MIDSCENE_MODEL_NAME) || ""
233
225
  };
@@ -236,24 +228,24 @@ function emitInsightDump(data, dumpSubscriber) {
236
228
  ...baseData,
237
229
  ...data
238
230
  };
239
- _optionalChain([dumpSubscriber, 'optionalCall', _14 => _14(finalData)]);
231
+ _optionalChain([dumpSubscriber, 'optionalCall', _13 => _13(finalData)]);
240
232
  }
241
233
 
242
234
  // src/insight/index.ts
243
235
  var debug = _logger.getDebug.call(void 0, "ai:insight");
244
236
  var Insight = class {
245
237
  constructor(context, opt) {
246
- this.aiVendorFn = _chunkX2UALJ3Bjs.callAiFn;
238
+ this.aiVendorFn = _chunkEYIL4AHPjs.callAiFn;
247
239
  _utils.assert.call(void 0, context, "context is required for Insight");
248
240
  if (typeof context === "function") {
249
241
  this.contextRetrieverFn = context;
250
242
  } else {
251
243
  this.contextRetrieverFn = () => Promise.resolve(context);
252
244
  }
253
- if (typeof _optionalChain([opt, 'optionalAccess', _15 => _15.aiVendorFn]) !== "undefined") {
245
+ if (typeof _optionalChain([opt, 'optionalAccess', _14 => _14.aiVendorFn]) !== "undefined") {
254
246
  this.aiVendorFn = opt.aiVendorFn;
255
247
  }
256
- if (typeof _optionalChain([opt, 'optionalAccess', _16 => _16.taskInfo]) !== "undefined") {
248
+ if (typeof _optionalChain([opt, 'optionalAccess', _15 => _15.taskInfo]) !== "undefined") {
257
249
  this.taskInfo = opt.taskInfo;
258
250
  }
259
251
  }
@@ -280,13 +272,13 @@ var Insight = class {
280
272
  );
281
273
  searchAreaPrompt = void 0;
282
274
  }
283
- const context = _optionalChain([opt, 'optionalAccess', _17 => _17.context]) || await this.contextRetrieverFn("locate");
275
+ const context = _optionalChain([opt, 'optionalAccess', _16 => _16.context]) || await this.contextRetrieverFn("locate");
284
276
  let searchArea = void 0;
285
277
  let searchAreaRawResponse = void 0;
286
278
  let searchAreaUsage = void 0;
287
279
  let searchAreaResponse = void 0;
288
280
  if (searchAreaPrompt) {
289
- searchAreaResponse = await _chunkX2UALJ3Bjs.AiLocateSection.call(void 0, {
281
+ searchAreaResponse = await _chunkEYIL4AHPjs.AiLocateSection.call(void 0, {
290
282
  context,
291
283
  sectionDescription: searchAreaPrompt
292
284
  });
@@ -299,7 +291,7 @@ var Insight = class {
299
291
  searchArea = searchAreaResponse.rect;
300
292
  }
301
293
  const startTime = Date.now();
302
- const { parseResult, rect, elementById, rawResponse, usage } = await _chunkX2UALJ3Bjs.AiLocateElement.call(void 0, {
294
+ const { parseResult, rect, elementById, rawResponse, usage } = await _chunkEYIL4AHPjs.AiLocateElement.call(void 0, {
303
295
  callAI: callAI || this.aiVendorFn,
304
296
  context,
305
297
  targetElementDescription: queryPrompt,
@@ -317,7 +309,7 @@ var Insight = class {
317
309
  searchAreaUsage
318
310
  };
319
311
  let errorLog;
320
- if (_optionalChain([parseResult, 'access', _18 => _18.errors, 'optionalAccess', _19 => _19.length])) {
312
+ if (_optionalChain([parseResult, 'access', _17 => _17.errors, 'optionalAccess', _18 => _18.length])) {
321
313
  errorLog = `AI model failed to locate:
322
314
  ${parseResult.errors.join("\n")}`;
323
315
  }
@@ -336,7 +328,7 @@ ${parseResult.errors.join("\n")}`;
336
328
  const elements = [];
337
329
  (parseResult.elements || []).forEach((item) => {
338
330
  if ("id" in item) {
339
- const element = elementById(_optionalChain([item, 'optionalAccess', _20 => _20.id]));
331
+ const element = elementById(_optionalChain([item, 'optionalAccess', _19 => _19.id]));
340
332
  if (!element) {
341
333
  console.warn(
342
334
  `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`
@@ -387,7 +379,7 @@ ${parseResult.errors.join("\n")}`;
387
379
  this.onceDumpUpdatedFn = void 0;
388
380
  const context = await this.contextRetrieverFn("extract");
389
381
  const startTime = Date.now();
390
- const { parseResult, usage } = await _chunkX2UALJ3Bjs.AiExtractElementInfo.call(void 0, {
382
+ const { parseResult, usage } = await _chunkEYIL4AHPjs.AiExtractElementInfo.call(void 0, {
391
383
  context,
392
384
  dataQuery: dataDemand,
393
385
  extractOption: opt
@@ -399,7 +391,7 @@ ${parseResult.errors.join("\n")}`;
399
391
  rawResponse: JSON.stringify(parseResult)
400
392
  };
401
393
  let errorLog;
402
- if (_optionalChain([parseResult, 'access', _21 => _21.errors, 'optionalAccess', _22 => _22.length])) {
394
+ if (_optionalChain([parseResult, 'access', _20 => _20.errors, 'optionalAccess', _21 => _21.length])) {
403
395
  errorLog = `AI response error:
404
396
  ${parseResult.errors.join("\n")}`;
405
397
  }
@@ -439,7 +431,7 @@ ${parseResult.errors.join("\n")}`;
439
431
  this.onceDumpUpdatedFn = void 0;
440
432
  const context = await this.contextRetrieverFn("assert");
441
433
  const startTime = Date.now();
442
- const assertResult = await _chunkX2UALJ3Bjs.AiAssert.call(void 0, {
434
+ const assertResult = await _chunkEYIL4AHPjs.AiAssert.call(void 0, {
443
435
  assertion,
444
436
  context
445
437
  });
@@ -472,7 +464,7 @@ ${parseResult.errors.join("\n")}`;
472
464
  async describe(target, opt) {
473
465
  _utils.assert.call(void 0, target, "target is required for insight.describe");
474
466
  const context = await this.contextRetrieverFn("describe");
475
- const { screenshotBase64 } = context;
467
+ const { screenshotBase64, size } = context;
476
468
  _utils.assert.call(void 0, screenshotBase64, "screenshot is required for insight.describe");
477
469
  const systemPrompt = elementDescriberInstruction();
478
470
  const defaultRectSize = 30;
@@ -484,6 +476,7 @@ ${parseResult.errors.join("\n")}`;
484
476
  } : target;
485
477
  let imagePayload = await _img.compositeElementInfoImg.call(void 0, {
486
478
  inputImgBase64: screenshotBase64,
479
+ size,
487
480
  elementsPositionInfo: [
488
481
  {
489
482
  rect: targetRect
@@ -491,8 +484,8 @@ ${parseResult.errors.join("\n")}`;
491
484
  ],
492
485
  borderThickness: 3
493
486
  });
494
- if (_optionalChain([opt, 'optionalAccess', _23 => _23.deepThink])) {
495
- const searchArea = _chunkX2UALJ3Bjs.expandSearchArea.call(void 0, targetRect, context.size);
487
+ if (_optionalChain([opt, 'optionalAccess', _22 => _22.deepThink])) {
488
+ const searchArea = _chunkEYIL4AHPjs.expandSearchArea.call(void 0, targetRect, context.size);
496
489
  debug("describe: set searchArea", searchArea);
497
490
  imagePayload = await _img.cropByRect.call(void 0,
498
491
  imagePayload,
@@ -515,7 +508,7 @@ ${parseResult.errors.join("\n")}`;
515
508
  ]
516
509
  }
517
510
  ];
518
- const callAIFn = this.aiVendorFn || _chunkX2UALJ3Bjs.callToGetJSONObject;
511
+ const callAIFn = this.aiVendorFn || _chunkEYIL4AHPjs.callToGetJSONObject;
519
512
  const res = await callAIFn(msgs, 4 /* DESCRIBE_ELEMENT */);
520
513
  const { content } = res;
521
514
  _utils.assert.call(void 0, !content.error, `describe failed: ${content.error}`);
@@ -538,6 +531,6 @@ var src_default = Insight;
538
531
 
539
532
 
540
533
 
541
- exports.AiAssert = _chunkX2UALJ3Bjs.AiAssert; exports.AiLocateElement = _chunkX2UALJ3Bjs.AiLocateElement; exports.Executor = Executor; exports.Insight = Insight; exports.MIDSCENE_MODEL_NAME = _env.MIDSCENE_MODEL_NAME; exports.default = src_default; exports.describeUserPage = _chunkX2UALJ3Bjs.describeUserPage; exports.getAIConfig = _env.getAIConfig; exports.getVersion = _chunkD4ZWVRRGjs.getVersion; exports.plan = _chunkX2UALJ3Bjs.plan;
534
+ exports.AiAssert = _chunkEYIL4AHPjs.AiAssert; exports.AiLocateElement = _chunkEYIL4AHPjs.AiLocateElement; exports.Executor = Executor; exports.Insight = Insight; exports.MIDSCENE_MODEL_NAME = _env.MIDSCENE_MODEL_NAME; exports.default = src_default; exports.describeUserPage = _chunkEYIL4AHPjs.describeUserPage; exports.getAIConfig = _env.getAIConfig; exports.getVersion = _chunkA7OTPIZJjs.getVersion; exports.plan = _chunkEYIL4AHPjs.plan;
542
535
 
543
536
  //# sourceMappingURL=index.js.map
@@ -1 +1 @@
1
- {"version":3,"mappings":";;;;;;;;;;;;;;;;AAUA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,cAAc;AAEhB,IAAM,WAAN,MAAe;AAAA,EAUpB,YACE,MACA,SAGA;AACA,SAAK,SACH,SAAS,SAAS,QAAQ,MAAM,SAAS,IAAI,YAAY;AAC3D,SAAK,OAAO;AACZ,SAAK,SAAS,SAAS,SAAS,CAAC,GAAG;AAAA,MAAI,CAAC,SACvC,KAAK,kBAAkB,IAAI;AAAA,IAC7B;AACA,SAAK,cAAc,SAAS;AAAA,EAC9B;AAAA,EAEQ,kBAAkB,MAAyC;AACjE,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,MAAgE;AAC3E;AAAA,MACE,KAAK,WAAW;AAAA,MAChB;AAAA,QAAyD,KAAK,gBAAgB,GAAG,KAAK;AAAA,EAAK,KAAK,gBAAgB,GAAG,UAAU;AAAA,IAC/H;AACA,QAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,WAAK,MAAM,KAAK,GAAG,KAAK,IAAI,CAAC,SAAS,KAAK,kBAAkB,IAAI,CAAC,CAAC;AAAA,IACrE,OAAO;AACL,WAAK,MAAM,KAAK,KAAK,kBAAkB,IAAI,CAAC;AAAA,IAC9C;AACA,QAAI,KAAK,WAAW,WAAW;AAC7B,WAAK,SAAS;AAAA,IAChB;AAAA,EACF;AAAA,EAEA,MAAM,QAAsB;AAC1B,QAAI,KAAK,WAAW,UAAU,KAAK,MAAM,SAAS,GAAG;AACnD,cAAQ;AAAA,QACN;AAAA,MACF;AAAA,IACF;AAEA,WAAO,KAAK,WAAW,WAAW,6BAA6B;AAC/D,WAAO,KAAK,WAAW,aAAa,+BAA+B;AACnE,WAAO,KAAK,WAAW,SAAS,4BAA4B;AAE5D,UAAM,mBAAmB,KAAK,MAAM;AAAA,MAClC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,mBAAmB,GAAG;AAExB;AAAA,IACF;AAEA,SAAK,SAAS;AACd,QAAI,YAAY;AAChB,QAAI,wBAAwB;AAE5B,QAAI;AAEJ,WAAO,YAAY,KAAK,MAAM,QAAQ;AACpC,YAAM,OAAO,KAAK,MAAM,SAAS;AACjC;AAAA,QACE,KAAK,WAAW;AAAA,QAChB,2CAA2C,KAAK,MAAM;AAAA,MACxD;AACA,WAAK,SAAS;AAAA,QACZ,OAAO,KAAK,IAAI;AAAA,MAClB;AACA,UAAI;AACF,aAAK,SAAS;AACd,YAAI;AACF,cAAI,KAAK,aAAa;AACpB,kBAAM,KAAK,YAAY,IAAI;AAAA,UAC7B;AAAA,QACF,SAAS,GAAG;AACV,kBAAQ,MAAM,wBAAwB,CAAC;AAAA,QACzC;AACA;AAAA,UACE,CAAC,WAAW,UAAU,UAAU,EAAE,QAAQ,KAAK,IAAI,KAAK;AAAA,UACxD,0BAA0B,KAAK,IAAI;AAAA,QACrC;AAEA,cAAM,EAAE,UAAU,MAAM,IAAI;AAC5B,eAAO,UAAU,uCAAuC,KAAK,IAAI,EAAE;AAEnE,YAAI;AACJ,cAAM,kBAAmC;AAAA,UACvC;AAAA,UACA,SAAS,oBAAoB;AAAA,QAC/B;AAEA,YAAI,KAAK,SAAS,WAAW;AAC3B;AAAA,YACE,KAAK,YAAY,YACf,KAAK,YAAY,WACjB,KAAK,YAAY,YACjB,KAAK,YAAY,aACjB,KAAK,YAAY,YACjB,KAAK,YAAY;AAAA,YACnB,gCAAgC,KAAK,OAAO;AAAA,UAC9C;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AACxD,cAAI,KAAK,YAAY,UAAU;AAC7B,iCACE,aACC;AAAA,UACL;AAAA,QACF,WAAW,KAAK,SAAS,YAAY,KAAK,SAAS,YAAY;AAC7D,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D,OAAO;AACL,kBAAQ;AAAA,YACN,0BAA0B,KAAK,IAAI;AAAA,UACrC;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D;AAEA,eAAO,OAAO,MAAM,WAAW;AAC/B,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD,aAAK,OAAO,SAAU,aAAqB,UAAU;AACrD;AAAA,MACF,SAAS,GAAQ;AACf,gCAAwB;AACxB,aAAK,QACH,GAAG,YAAY,OAAO,MAAM,WAAW,IAAI;AAC7C,aAAK,aAAa,EAAE;AAEpB,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF;AAAA,IACF;AAGA,aAAS,IAAI,YAAY,GAAG,IAAI,KAAK,MAAM,QAAQ,KAAK;AACtD,WAAK,MAAM,CAAC,EAAE,SAAS;AAAA,IACzB;AAEA,QAAI,uBAAuB;AACzB,WAAK,SAAS;AAAA,IAChB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAEA,QAAI,KAAK,MAAM,QAAQ;AAErB,YAAM,cAAc,KAAK,IAAI,WAAW,KAAK,MAAM,SAAS,CAAC;AAC7D,aAAO,KAAK,MAAM,WAAW,EAAE;AAAA,IACjC;AAAA,EACF;AAAA,EAEA,iBAA0B;AACxB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA,EAEA,kBAAwC;AACtC,QAAI,KAAK,WAAW,SAAS;AAC3B,aAAO;AAAA,IACT;AACA,UAAM,iBAAiB,KAAK,MAAM;AAAA,MAChC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,kBAAkB,GAAG;AACvB,aAAO,KAAK,MAAM,cAAc;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EAEA,OAAsB;AACpB,QAAI,mBAAmB;AAEvB,QAAI,aAAa,GAAG;AAClB,YAAM,iBAAiB,mBAAmB;AAC1C,UAAI,gBAAgB;AAClB,2BAAmB,WAAW,cAAc;AAAA,MAC9C,OAAO;AACL,2BAAmB,GAAG,aAAa,CAAC;AAAA,MACtC;AAAA,IACF;AACA,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,mBAAmB;AAAA,MACnB,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC9NA,SAAS,4BAA4B;AAE9B,IAAM,8BAA8B,MAAM;AAC/C,SAAO,mNAAmN,qBAAqB,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAoBlP;;;ACQA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA,gBAAAA;AAAA,OACK;AACP,SAAS,yBAAyB,kBAAkB;AACpD,SAAS,gBAAgB;AACzB,SAAS,UAAAC,eAAc;;;AChCvB;AAAA,EACE,uBAAAC;AAAA,EACA,eAAAC;AAAA,OAGK;AACP,SAAS,YAAY;AAEd,SAAS,gBACd,MACA,gBACA;AACA,QAAM,WAAqB;AAAA,IACzB,YAAY,WAAW;AAAA,IACvB,SAAS,KAAK,IAAI;AAAA,IAClB,YAAYA,aAAYD,oBAAmB,KAAK;AAAA,EAClD;AACA,QAAM,YAAyB;AAAA,IAC7B,OAAO,KAAK;AAAA,IACZ,GAAG;AAAA,IACH,GAAG;AAAA,EACL;AAEA,mBAAiB,SAAS;AAC5B;;;ADoBA,IAAM,QAAQ,SAAS,YAAY;AACnC,IAAqB,UAArB,MAGE;AAAA,EAWA,YACE,SAGA,KACA;AAXF,sBAAoD;AAYlD,IAAAD,QAAO,SAAS,iCAAiC;AACjD,QAAI,OAAO,YAAY,YAAY;AACjC,WAAK,qBAAqB;AAAA,IAC5B,OAAO;AACL,WAAK,qBAAqB,MAAM,QAAQ,QAAQ,OAAO;AAAA,IACzD;AAEA,QAAI,OAAO,KAAK,eAAe,aAAa;AAC1C,WAAK,aAAa,IAAI;AAAA,IACxB;AACA,QAAI,OAAO,KAAK,aAAa,aAAa;AACxC,WAAK,WAAW,IAAI;AAAA,IACtB;AAAA,EACF;AAAA,EAEA,MAAM,OACJ,OACA,KACuB;AACvB,UAAM,EAAE,OAAO,IAAI,OAAO,CAAC;AAC3B,UAAM,cAAc,OAAO,UAAU,WAAW,QAAQ,MAAM;AAC9D,IAAAA,QAAO,aAAa,8BAA8B;AAClD,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,IAAAA,QAAO,OAAO,UAAU,UAAU,sCAAsC;AAExE,UAAM,wBAAwB;AAAA,MAC5B;AAAA,IACF;AACA,QAAI,uBAAuB;AACzB,YAAM,yBAAyB,qBAAqB;AAAA,IACtD;AACA,QAAI;AACJ,QAAI,MAAM,aAAa,uBAAuB;AAC5C,yBAAmB,MAAM;AAAA,IAC3B;AAEA,QAAI,oBAAoB,CAACD,cAAa,GAAG;AACvC,cAAQ;AAAA,QACN;AAAA,MACF;AACA,yBAAmB;AAAA,IACrB;AAEA,UAAM,UAAU,KAAK,WAAY,MAAM,KAAK,mBAAmB,QAAQ;AAEvE,QAAI,aAA+B;AACnC,QAAI,wBAA4C;AAChD,QAAI,kBAA2C;AAC/C,QAAI,qBAEY;AAChB,QAAI,kBAAkB;AACpB,2BAAqB,MAAM,gBAAgB;AAAA,QACzC;AAAA,QACA,oBAAoB;AAAA,MACtB,CAAC;AACD,MAAAC;AAAA,QACE,mBAAmB;AAAA,QACnB,gCAAgC,gBAAgB,IAC9C,mBAAmB,QAAQ,KAAK,mBAAmB,KAAK,KAAK,EAC/D;AAAA,MACF;AACA,8BAAwB,mBAAmB;AAC3C,wBAAkB,mBAAmB;AACrC,mBAAa,mBAAmB;AAAA,IAClC;AAEA,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,aAAa,aAAa,MAAM,IACzD,MAAM,gBAAgB;AAAA,MACpB,QAAQ,UAAU,KAAK;AAAA,MACvB;AAAA,MACA,0BAA0B;AAAA,MAC1B,cAAc;AAAA,IAChB,CAAC;AAEH,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,MACvC,gBAAgB,KAAK,UAAU,WAAW;AAAA,MAC1C;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAgC,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAC1E;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT,SAAS;AAAA,MACX;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,aAAa;AAAA,MACb,MAAM;AAAA,MACN;AAAA,MACA,WAAW,CAAC,CAAC;AAAA,MACb,OAAO;AAAA,IACT;AAEA,UAAM,WAA0B,CAAC;AACjC,KAAC,YAAY,YAAY,CAAC,GAAG,QAAQ,CAAC,SAAS;AAC7C,UAAI,QAAQ,MAAM;AAChB,cAAM,UAAU,YAAY,MAAM,EAAE;AAEpC,YAAI,CAAC,SAAS;AACZ,kBAAQ;AAAA,YACN,kCAAkC,KAAK,EAAE;AAAA,UAC3C;AACA;AAAA,QACF;AACA,iBAAS,KAAK,OAAO;AAAA,MACvB;AAAA,IACF,CAAC;AAED;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,IACF;AAEA,QAAI,UAAU;AACZ,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,IAAAA;AAAA,MACE,SAAS,UAAU;AAAA,MACnB,6CAA6C,SAAS,MAAM;AAAA,IAC9D;AAEA,QAAI,SAAS,WAAW,GAAG;AACzB,aAAO;AAAA,QACL,SAAS;AAAA,UACP,IAAI,SAAS,CAAC,EAAG;AAAA,UACjB,SAAS,SAAS,CAAC,EAAG;AAAA,UACtB,QAAQ,SAAS,CAAC,EAAG;AAAA,UACrB,MAAM,SAAS,CAAC,EAAG;AAAA,UACnB,QAAQ,SAAS,CAAC,EAAG,UAAU,CAAC;AAAA,UAChC,YAAY,SAAS,CAAC,EAAG;AAAA,QAC3B;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,WAAO;AAAA,MACL,SAAS;AAAA,MACT;AAAA,IACF;AAAA,EACF;AAAA,EAYA,MAAM,QACJ,YACA,KACc;AACd,IAAAA;AAAA,MACE,OAAO,eAAe,YAAY,OAAO,eAAe;AAAA,MACxD,kDAAkD,OAAO,UAAU;AAAA,IACrE;AACA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,SAAS;AAEvD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,IAAI,MAAM,qBAAwB;AAAA,MAC3D;AAAA,MACA,WAAW;AAAA,MACX,eAAe;AAAA,IACjB,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,IACzC;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAwB,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAClE;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,OAAO;AAAA,IACT;AAEA,UAAM,EAAE,KAAK,IAAI,eAAe,CAAC;AAGjC;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH;AAAA,MACF;AAAA,MACA;AAAA,IACF;AAEA,QAAI,YAAY,CAAC,MAAM;AACrB,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,WAAsD;AACjE,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,QAAQ;AACtD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,eAAe,MAAM,SAAS;AAAA,MAClC;AAAA,MACA;AAAA,IACF,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,aAAa,OAAO;AAAA,IAClD;AAEA,UAAM,EAAE,SAAS,KAAK,IAAI,aAAa;AACvC,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,eAAe;AAAA,MACf,kBAAkB;AAAA,MAClB,OAAO,OAAO,SAAY;AAAA,IAC5B;AACA,oBAAgB,UAAU,cAAc;AAExC,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA,OAAO,aAAa;AAAA,IACtB;AAAA,EACF;AAAA,EACA,MAAM,SACJ,QACA,KAGyD;AACzD,IAAAA,QAAO,QAAQ,yCAAyC;AACxD,UAAM,UAAU,MAAM,KAAK,mBAAmB,UAAU;AACxD,UAAM,EAAE,iBAAiB,IAAI;AAC7B,IAAAA,QAAO,kBAAkB,6CAA6C;AAEtE,UAAM,eAAe,4BAA4B;AAGjD,UAAM,kBAAkB;AACxB,UAAM,aAAmB,MAAM,QAAQ,MAAM,IACzC;AAAA,MACE,MAAM,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAChD,KAAK,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAC/C,OAAO;AAAA,MACP,QAAQ;AAAA,IACV,IACA;AAEJ,QAAI,eAAe,MAAM,wBAAwB;AAAA,MAC/C,gBAAgB;AAAA,MAChB,sBAAsB;AAAA,QACpB;AAAA,UACE,MAAM;AAAA,QACR;AAAA,MACF;AAAA,MACA,iBAAiB;AAAA,IACnB,CAAC;AAED,QAAI,KAAK,WAAW;AAClB,YAAM,aAAa,iBAAiB,YAAY,QAAQ,IAAI;AAC5D,YAAM,4BAA4B,UAAU;AAC5C,qBAAe,MAAM;AAAA,QACnB;AAAA,QACA;AAAA,QACA,qBAAqB,oBAAoB;AAAA,MAC3C;AAAA,IACF;AAEA,UAAM,OAAe;AAAA,MACnB,EAAE,MAAM,UAAU,SAAS,aAAa;AAAA,MACxC;AAAA,QACE,MAAM;AAAA,QACN,SAAS;AAAA,UACP;AAAA,YACE,MAAM;AAAA,YACN,WAAW;AAAA,cACT,KAAK;AAAA,cACL,QAAQ;AAAA,YACV;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,WACJ,KAAK,cAAc;AAErB,UAAM,MAAM,MAAM,SAAS,8BAAmC;AAE9D,UAAM,EAAE,QAAQ,IAAI;AACpB,IAAAA,QAAO,CAAC,QAAQ,OAAO,oBAAoB,QAAQ,KAAK,EAAE;AAC1D,IAAAA,QAAO,QAAQ,aAAa,gCAAgC;AAC5D,WAAO;AAAA,EACT;AACF;;;AEzZA,SAAS,eAAAE,cAAa,uBAAAD,4BAA2B;AAGjD,IAAO,cAAQ","names":["vlLocateMode","assert","MIDSCENE_MODEL_NAME","getAIConfig"],"ignoreList":[],"sources":["../../src/ai-model/action-executor.ts","../../src/ai-model/prompt/describe.ts","../../src/insight/index.ts","../../src/insight/utils.ts","../../src/index.ts"],"sourcesContent":["import type {\n ExecutionDump,\n ExecutionTask,\n ExecutionTaskApply,\n ExecutionTaskInsightLocateOutput,\n ExecutionTaskProgressOptions,\n ExecutionTaskReturn,\n ExecutorContext,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n MIDSCENE_MODEL_NAME,\n getAIConfig,\n uiTarsModelVersion,\n vlLocateMode,\n} from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\n\nexport class Executor {\n name: string;\n\n tasks: ExecutionTask[];\n\n // status of executor\n status: 'init' | 'pending' | 'running' | 'completed' | 'error';\n\n onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n\n constructor(\n name: string,\n options?: ExecutionTaskProgressOptions & {\n tasks?: ExecutionTaskApply[];\n },\n ) {\n this.status =\n options?.tasks && options.tasks.length > 0 ? 'pending' : 'init';\n this.name = name;\n this.tasks = (options?.tasks || []).map((item) =>\n this.markTaskAsPending(item),\n );\n this.onTaskStart = options?.onTaskStart;\n }\n\n private markTaskAsPending(task: ExecutionTaskApply): ExecutionTask {\n return {\n status: 'pending',\n ...task,\n };\n }\n\n async append(task: ExecutionTaskApply[] | ExecutionTaskApply): Promise<void> {\n assert(\n this.status !== 'error',\n `executor is in error state, cannot append task\\nerror=${this.latestErrorTask()?.error}\\n${this.latestErrorTask()?.errorStack}`,\n );\n if (Array.isArray(task)) {\n this.tasks.push(...task.map((item) => this.markTaskAsPending(item)));\n } else {\n this.tasks.push(this.markTaskAsPending(task));\n }\n if (this.status !== 'running') {\n this.status = 'pending';\n }\n }\n\n async flush(): Promise<any> {\n if (this.status === 'init' && this.tasks.length > 0) {\n console.warn(\n 'illegal state for executor, status is init but tasks are not empty',\n );\n }\n\n assert(this.status !== 'running', 'executor is already running');\n assert(this.status !== 'completed', 'executor is already completed');\n assert(this.status !== 'error', 'executor is in error state');\n\n const nextPendingIndex = this.tasks.findIndex(\n (task) => task.status === 'pending',\n );\n if (nextPendingIndex < 0) {\n // all tasks are completed\n return;\n }\n\n this.status = 'running';\n let taskIndex = nextPendingIndex;\n let successfullyCompleted = true;\n\n let previousFindOutput: ExecutionTaskInsightLocateOutput | undefined;\n\n while (taskIndex < this.tasks.length) {\n const task = this.tasks[taskIndex];\n assert(\n task.status === 'pending',\n `task status should be pending, but got: ${task.status}`,\n );\n task.timing = {\n start: Date.now(),\n };\n try {\n task.status = 'running';\n try {\n if (this.onTaskStart) {\n await this.onTaskStart(task);\n }\n } catch (e) {\n console.error('error in onTaskStart', e);\n }\n assert(\n ['Insight', 'Action', 'Planning'].indexOf(task.type) >= 0,\n `unsupported task type: ${task.type}`,\n );\n\n const { executor, param } = task;\n assert(executor, `executor is required for task type: ${task.type}`);\n\n let returnValue;\n const executorContext: ExecutorContext = {\n task,\n element: previousFindOutput?.element,\n };\n\n if (task.type === 'Insight') {\n assert(\n task.subType === 'Locate' ||\n task.subType === 'Query' ||\n task.subType === 'Assert' ||\n task.subType === 'Boolean' ||\n task.subType === 'Number' ||\n task.subType === 'String',\n `unsupported insight subType: ${task.subType}`,\n );\n returnValue = await task.executor(param, executorContext);\n if (task.subType === 'Locate') {\n previousFindOutput = (\n returnValue as ExecutionTaskReturn<ExecutionTaskInsightLocateOutput>\n )?.output;\n }\n } else if (task.type === 'Action' || task.type === 'Planning') {\n returnValue = await task.executor(param, executorContext);\n } else {\n console.warn(\n `unsupported task type: ${task.type}, will try to execute it directly`,\n );\n returnValue = await task.executor(param, executorContext);\n }\n\n Object.assign(task, returnValue);\n task.status = 'finished';\n task.timing.end = Date.now();\n task.timing.cost = task.timing.end - task.timing.start;\n task.timing.aiCost = (returnValue as any)?.aiCost || 0;\n taskIndex++;\n } catch (e: any) {\n successfullyCompleted = false;\n task.error =\n e?.message || (typeof e === 'string' ? e : 'error-without-message');\n task.errorStack = e.stack;\n\n task.status = 'failed';\n task.timing.end = Date.now();\n task.timing.cost = task.timing.end - task.timing.start;\n break;\n }\n }\n\n // set all remaining tasks as cancelled\n for (let i = taskIndex + 1; i < this.tasks.length; i++) {\n this.tasks[i].status = 'cancelled';\n }\n\n if (successfullyCompleted) {\n this.status = 'completed';\n } else {\n this.status = 'error';\n }\n\n if (this.tasks.length) {\n // return the last output\n const outputIndex = Math.min(taskIndex, this.tasks.length - 1);\n return this.tasks[outputIndex].output;\n }\n }\n\n isInErrorState(): boolean {\n return this.status === 'error';\n }\n\n latestErrorTask(): ExecutionTask | null {\n if (this.status !== 'error') {\n return null;\n }\n const errorTaskIndex = this.tasks.findIndex(\n (task) => task.status === 'failed',\n );\n if (errorTaskIndex >= 0) {\n return this.tasks[errorTaskIndex];\n }\n return null;\n }\n\n dump(): ExecutionDump {\n let modelDescription = '';\n\n if (vlLocateMode()) {\n const uiTarsModelVer = uiTarsModelVersion();\n if (uiTarsModelVer) {\n modelDescription = `UI-TARS=${uiTarsModelVer}`;\n } else {\n modelDescription = `${vlLocateMode()} mode`;\n }\n }\n const dumpData: ExecutionDump = {\n sdkVersion: getVersion(),\n model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n model_description: modelDescription,\n logTime: Date.now(),\n name: this.name,\n tasks: this.tasks,\n };\n return dumpData;\n }\n}\n","import { getPreferredLanguage } from '@midscene/shared/env';\n\nexport const elementDescriberInstruction = () => {\n return `Tell what is the content of the element wrapped by the read rectangle in the screenshot. Your description is expected to be used to precisely locate the element from other similar elements on screenshot. Use ${getPreferredLanguage()} in the description.\n\nPlease follow the following rules:\n1. The description should be start with a brief description, like \"a button for confirming the action\".\n\n2. Include these information in the description to distinguish the element from its siblings and other similar elements, as much as possible:\n- The text of the element, like \"with text 'Confirm'\"\n- What the element looks like if it's an image, like \"with image '...'\"\n- The relative position of the element, like \"on the left of ..., around ...\"\n- How to distinguish the element from its siblings elements, like \"it is the icon instead of the text\"\n\n3. Do NOT mention the red rectangle in the description.\n\n4. Use the error field to describe the unexpected situations, if any. If not, put null.\n\nReturn in JSON:\n{\n \"description\": \"[{brief description}]: {text of the element} {image of the element} {relative position of the element} ... \",\n \"error\"?: \"...\"\n}`;\n};\n","import {\n AIActionType,\n type AIArgs,\n callAiFn,\n expandSearchArea,\n} from '@/ai-model/common';\nimport {\n AiExtractElementInfo,\n AiLocateElement,\n callToGetJSONObject,\n} from '@/ai-model/index';\nimport { AiAssert, AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n AIDescribeElementResponse,\n AIElementResponse,\n AIUsageInfo,\n BaseElement,\n DetailedLocateParam,\n DumpSubscriber,\n InsightAction,\n InsightAssertionResponse,\n InsightExtractOption,\n InsightExtractParam,\n InsightOptions,\n InsightTaskInfo,\n LocateResult,\n PartialInsightDumpFromSDK,\n Rect,\n UIContext,\n} from '@/types';\nimport {\n MIDSCENE_FORCE_DEEP_THINK,\n MIDSCENE_USE_QWEN_VL,\n getAIConfigInBoolean,\n vlLocateMode,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n context?: UIContext<BaseElement>;\n callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n ElementType extends BaseElement = BaseElement,\n ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n contextRetrieverFn: (\n action: InsightAction,\n ) => Promise<ContextType> | ContextType;\n\n aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n onceDumpUpdatedFn?: DumpSubscriber;\n\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n constructor(\n context:\n | ContextType\n | ((action: InsightAction) => Promise<ContextType> | ContextType),\n opt?: InsightOptions,\n ) {\n assert(context, 'context is required for Insight');\n if (typeof context === 'function') {\n this.contextRetrieverFn = context;\n } else {\n this.contextRetrieverFn = () => Promise.resolve(context);\n }\n\n if (typeof opt?.aiVendorFn !== 'undefined') {\n this.aiVendorFn = opt.aiVendorFn;\n }\n if (typeof opt?.taskInfo !== 'undefined') {\n this.taskInfo = opt.taskInfo;\n }\n }\n\n async locate(\n query: DetailedLocateParam,\n opt?: LocateOpts,\n ): Promise<LocateResult> {\n const { callAI } = opt || {};\n const queryPrompt = typeof query === 'string' ? query : query.prompt;\n assert(queryPrompt, 'query is required for locate');\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n assert(typeof query === 'object', 'query should be an object for locate');\n\n const globalDeepThinkSwitch = getAIConfigInBoolean(\n MIDSCENE_FORCE_DEEP_THINK,\n );\n if (globalDeepThinkSwitch) {\n debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n }\n let searchAreaPrompt;\n if (query.deepThink || globalDeepThinkSwitch) {\n searchAreaPrompt = query.prompt;\n }\n\n if (searchAreaPrompt && !vlLocateMode()) {\n console.warn(\n 'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n );\n searchAreaPrompt = undefined;\n }\n\n const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n let searchArea: Rect | undefined = undefined;\n let searchAreaRawResponse: string | undefined = undefined;\n let searchAreaUsage: AIUsageInfo | undefined = undefined;\n let searchAreaResponse:\n | Awaited<ReturnType<typeof AiLocateSection>>\n | undefined = undefined;\n if (searchAreaPrompt) {\n searchAreaResponse = await AiLocateSection({\n context,\n sectionDescription: searchAreaPrompt,\n });\n assert(\n searchAreaResponse.rect,\n `cannot find search area for \"${searchAreaPrompt}\"${\n searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n }`,\n );\n searchAreaRawResponse = searchAreaResponse.rawResponse;\n searchAreaUsage = searchAreaResponse.usage;\n searchArea = searchAreaResponse.rect;\n }\n\n const startTime = Date.now();\n const { parseResult, rect, elementById, rawResponse, usage } =\n await AiLocateElement({\n callAI: callAI || this.aiVendorFn,\n context,\n targetElementDescription: queryPrompt,\n searchConfig: searchAreaResponse,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(rawResponse),\n formatResponse: JSON.stringify(parseResult),\n usage,\n searchArea,\n searchAreaRawResponse,\n searchAreaUsage,\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'locate',\n userQuery: {\n element: queryPrompt,\n },\n matchedElement: [],\n matchedRect: rect,\n data: null,\n taskInfo,\n deepThink: !!searchArea,\n error: errorLog,\n };\n\n const elements: BaseElement[] = [];\n (parseResult.elements || []).forEach((item) => {\n if ('id' in item) {\n const element = elementById(item?.id);\n\n if (!element) {\n console.warn(\n `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n );\n return;\n }\n elements.push(element);\n }\n });\n\n emitInsightDump(\n {\n ...dumpData,\n matchedElement: elements,\n },\n dumpSubscriber,\n );\n\n if (errorLog) {\n throw new Error(errorLog);\n }\n\n assert(\n elements.length <= 1,\n `locate: multiple elements found, length = ${elements.length}`,\n );\n\n if (elements.length === 1) {\n return {\n element: {\n id: elements[0]!.id,\n indexId: elements[0]!.indexId,\n center: elements[0]!.center,\n rect: elements[0]!.rect,\n xpaths: elements[0]!.xpaths || [],\n attributes: elements[0]!.attributes,\n },\n rect,\n };\n }\n return {\n element: null,\n rect,\n };\n }\n\n async extract<T = any>(input: string, opt?: InsightExtractOption): Promise<T>;\n async extract<T extends Record<string, string>>(\n input: T,\n opt?: InsightExtractOption,\n ): Promise<Record<keyof T, any>>;\n async extract<T extends object>(\n input: Record<keyof T, string>,\n opt?: InsightExtractOption,\n ): Promise<T>;\n\n async extract<T>(\n dataDemand: InsightExtractParam,\n opt?: InsightExtractOption,\n ): Promise<any> {\n assert(\n typeof dataDemand === 'object' || typeof dataDemand === 'string',\n `dataDemand should be object or string, but get ${typeof dataDemand}`,\n );\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n const context = await this.contextRetrieverFn('extract');\n\n const startTime = Date.now();\n const { parseResult, usage } = await AiExtractElementInfo<T>({\n context,\n dataQuery: dataDemand,\n extractOption: opt,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(parseResult),\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'extract',\n userQuery: {\n dataDemand,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n error: errorLog,\n };\n\n const { data } = parseResult || {};\n\n // 4\n emitInsightDump(\n {\n ...dumpData,\n data,\n },\n dumpSubscriber,\n );\n\n if (errorLog && !data) {\n throw new Error(errorLog);\n }\n\n return {\n data,\n usage,\n };\n }\n\n async assert(assertion: string): Promise<InsightAssertionResponse> {\n if (typeof assertion !== 'string') {\n throw new Error(\n 'This is the assert method for Midscene, the first argument should be a string. If you want to use the assert method from Node.js, please import it from the Node.js assert module.',\n );\n }\n\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n const context = await this.contextRetrieverFn('assert');\n const startTime = Date.now();\n const assertResult = await AiAssert({\n assertion,\n context,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(assertResult.content),\n };\n\n const { thought, pass } = assertResult.content;\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'assert',\n userQuery: {\n assertion,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n assertionPass: pass,\n assertionThought: thought,\n error: pass ? undefined : thought,\n };\n emitInsightDump(dumpData, dumpSubscriber);\n\n return {\n pass,\n thought,\n usage: assertResult.usage,\n };\n }\n async describe(\n target: Rect | [number, number],\n opt?: {\n deepThink?: boolean;\n },\n ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n assert(target, 'target is required for insight.describe');\n const context = await this.contextRetrieverFn('describe');\n const { screenshotBase64 } = context;\n assert(screenshotBase64, 'screenshot is required for insight.describe');\n\n const systemPrompt = elementDescriberInstruction();\n\n // Convert [x,y] center point to Rect if needed\n const defaultRectSize = 30;\n const targetRect: Rect = Array.isArray(target)\n ? {\n left: Math.floor(target[0] - defaultRectSize / 2),\n top: Math.floor(target[1] - defaultRectSize / 2),\n width: defaultRectSize,\n height: defaultRectSize,\n }\n : target;\n\n let imagePayload = await compositeElementInfoImg({\n inputImgBase64: screenshotBase64,\n elementsPositionInfo: [\n {\n rect: targetRect,\n },\n ],\n borderThickness: 3,\n });\n\n if (opt?.deepThink) {\n const searchArea = expandSearchArea(targetRect, context.size);\n debug('describe: set searchArea', searchArea);\n imagePayload = await cropByRect(\n imagePayload,\n searchArea,\n getAIConfigInBoolean(MIDSCENE_USE_QWEN_VL),\n );\n }\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n detail: 'high',\n },\n },\n ],\n },\n ];\n\n const callAIFn =\n this.aiVendorFn || callToGetJSONObject<AIDescribeElementResponse>;\n\n const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT);\n\n const { content } = res;\n assert(!content.error, `describe failed: ${content.error}`);\n assert(content.description, 'failed to describe the element');\n return content;\n }\n}\n","import type {\n DumpMeta,\n DumpSubscriber,\n InsightDump,\n PartialInsightDumpFromSDK,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n MIDSCENE_MODEL_NAME,\n getAIConfig,\n uiTarsModelVersion,\n vlLocateMode,\n} from '@midscene/shared/env';\nimport { uuid } from '@midscene/shared/utils';\n\nexport function emitInsightDump(\n data: PartialInsightDumpFromSDK,\n dumpSubscriber?: DumpSubscriber,\n) {\n const baseData: DumpMeta = {\n sdkVersion: getVersion(),\n logTime: Date.now(),\n model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n };\n const finalData: InsightDump = {\n logId: uuid(),\n ...baseData,\n ...data,\n };\n\n dumpSubscriber?.(finalData);\n}\n","import { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n plan,\n describeUserPage,\n AiLocateElement,\n AiAssert,\n} from './ai-model/index';\n\nexport { getAIConfig, MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n MidsceneYamlScript,\n MidsceneYamlTask,\n MidsceneYamlFlowItem,\n MidsceneYamlFlowItemAIRightClick,\n} from './yaml';\n"]}
1
+ {"version":3,"mappings":";;;;;;;;;;;;;;;;AAUA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,cAAc;AAEhB,IAAM,WAAN,MAAe;AAAA,EAUpB,YACE,MACA,SAGA;AACA,SAAK,SACH,SAAS,SAAS,QAAQ,MAAM,SAAS,IAAI,YAAY;AAC3D,SAAK,OAAO;AACZ,SAAK,SAAS,SAAS,SAAS,CAAC,GAAG;AAAA,MAAI,CAAC,SACvC,KAAK,kBAAkB,IAAI;AAAA,IAC7B;AACA,SAAK,cAAc,SAAS;AAAA,EAC9B;AAAA,EAEQ,kBAAkB,MAAyC;AACjE,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,MAAgE;AAC3E;AAAA,MACE,KAAK,WAAW;AAAA,MAChB;AAAA,QAAyD,KAAK,gBAAgB,GAAG,KAAK;AAAA,EAAK,KAAK,gBAAgB,GAAG,UAAU;AAAA,IAC/H;AACA,QAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,WAAK,MAAM,KAAK,GAAG,KAAK,IAAI,CAAC,SAAS,KAAK,kBAAkB,IAAI,CAAC,CAAC;AAAA,IACrE,OAAO;AACL,WAAK,MAAM,KAAK,KAAK,kBAAkB,IAAI,CAAC;AAAA,IAC9C;AACA,QAAI,KAAK,WAAW,WAAW;AAC7B,WAAK,SAAS;AAAA,IAChB;AAAA,EACF;AAAA,EAEA,MAAM,QAAsB;AAC1B,QAAI,KAAK,WAAW,UAAU,KAAK,MAAM,SAAS,GAAG;AACnD,cAAQ;AAAA,QACN;AAAA,MACF;AAAA,IACF;AAEA,WAAO,KAAK,WAAW,WAAW,6BAA6B;AAC/D,WAAO,KAAK,WAAW,aAAa,+BAA+B;AACnE,WAAO,KAAK,WAAW,SAAS,4BAA4B;AAE5D,UAAM,mBAAmB,KAAK,MAAM;AAAA,MAClC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,mBAAmB,GAAG;AAExB;AAAA,IACF;AAEA,SAAK,SAAS;AACd,QAAI,YAAY;AAChB,QAAI,wBAAwB;AAE5B,QAAI;AAEJ,WAAO,YAAY,KAAK,MAAM,QAAQ;AACpC,YAAM,OAAO,KAAK,MAAM,SAAS;AACjC;AAAA,QACE,KAAK,WAAW;AAAA,QAChB,2CAA2C,KAAK,MAAM;AAAA,MACxD;AACA,WAAK,SAAS;AAAA,QACZ,OAAO,KAAK,IAAI;AAAA,MAClB;AACA,UAAI;AACF,aAAK,SAAS;AACd,YAAI;AACF,cAAI,KAAK,aAAa;AACpB,kBAAM,KAAK,YAAY,IAAI;AAAA,UAC7B;AAAA,QACF,SAAS,GAAG;AACV,kBAAQ,MAAM,wBAAwB,CAAC;AAAA,QACzC;AACA;AAAA,UACE,CAAC,WAAW,UAAU,UAAU,EAAE,QAAQ,KAAK,IAAI,KAAK;AAAA,UACxD,0BAA0B,KAAK,IAAI;AAAA,QACrC;AAEA,cAAM,EAAE,UAAU,MAAM,IAAI;AAC5B,eAAO,UAAU,uCAAuC,KAAK,IAAI,EAAE;AAEnE,YAAI;AACJ,cAAM,kBAAmC;AAAA,UACvC;AAAA,UACA,SAAS,oBAAoB;AAAA,QAC/B;AAEA,YAAI,KAAK,SAAS,WAAW;AAC3B;AAAA,YACE,KAAK,YAAY,YACf,KAAK,YAAY,WACjB,KAAK,YAAY,YACjB,KAAK,YAAY,aACjB,KAAK,YAAY,YACjB,KAAK,YAAY;AAAA,YACnB,gCAAgC,KAAK,OAAO;AAAA,UAC9C;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AACxD,cAAI,KAAK,YAAY,UAAU;AAC7B,iCACE,aACC;AAAA,UACL;AAAA,QACF,WAAW,KAAK,SAAS,YAAY,KAAK,SAAS,YAAY;AAC7D,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D,OAAO;AACL,kBAAQ;AAAA,YACN,0BAA0B,KAAK,IAAI;AAAA,UACrC;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D;AAEA,eAAO,OAAO,MAAM,WAAW;AAC/B,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF,SAAS,GAAQ;AACf,gCAAwB;AACxB,aAAK,QACH,GAAG,YAAY,OAAO,MAAM,WAAW,IAAI;AAC7C,aAAK,aAAa,EAAE;AAEpB,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF;AAAA,IACF;AAGA,aAAS,IAAI,YAAY,GAAG,IAAI,KAAK,MAAM,QAAQ,KAAK;AACtD,WAAK,MAAM,CAAC,EAAE,SAAS;AAAA,IACzB;AAEA,QAAI,uBAAuB;AACzB,WAAK,SAAS;AAAA,IAChB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAEA,QAAI,KAAK,MAAM,QAAQ;AAErB,YAAM,cAAc,KAAK,IAAI,WAAW,KAAK,MAAM,SAAS,CAAC;AAC7D,aAAO,KAAK,MAAM,WAAW,EAAE;AAAA,IACjC;AAAA,EACF;AAAA,EAEA,iBAA0B;AACxB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA,EAEA,kBAAwC;AACtC,QAAI,KAAK,WAAW,SAAS;AAC3B,aAAO;AAAA,IACT;AACA,UAAM,iBAAiB,KAAK,MAAM;AAAA,MAChC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,kBAAkB,GAAG;AACvB,aAAO,KAAK,MAAM,cAAc;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EAEA,OAAsB;AACpB,QAAI,mBAAmB;AAEvB,QAAI,aAAa,GAAG;AAClB,YAAM,iBAAiB,mBAAmB;AAC1C,UAAI,gBAAgB;AAClB,2BAAmB,WAAW,cAAc;AAAA,MAC9C,OAAO;AACL,2BAAmB,GAAG,aAAa,CAAC;AAAA,MACtC;AAAA,IACF;AACA,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,mBAAmB;AAAA,MACnB,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC7NA,SAAS,4BAA4B;AAE9B,IAAM,8BAA8B,MAAM;AAC/C,SAAO,6EAA6E,qBAAqB,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAgB5G;;;ACYA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA,gBAAAA;AAAA,OACK;AACP,SAAS,yBAAyB,kBAAkB;AACpD,SAAS,gBAAgB;AACzB,SAAS,UAAAC,eAAc;;;AChCvB,SAAS,uBAAAC,sBAAqB,eAAAC,oBAAmB;AACjD,SAAS,YAAY;AAEd,SAAS,gBACd,MACA,gBACA;AACA,QAAM,WAAqB;AAAA,IACzB,YAAY,WAAW;AAAA,IACvB,SAAS,KAAK,IAAI;AAAA,IAClB,YAAYA,aAAYD,oBAAmB,KAAK;AAAA,EAClD;AACA,QAAM,YAAyB;AAAA,IAC7B,OAAO,KAAK;AAAA,IACZ,GAAG;AAAA,IACH,GAAG;AAAA,EACL;AAEA,mBAAiB,SAAS;AAC5B;;;ADyBA,IAAM,QAAQ,SAAS,YAAY;AACnC,IAAqB,UAArB,MAGE;AAAA,EAWA,YACE,SAGA,KACA;AAXF,sBAAoD;AAYlD,IAAAD,QAAO,SAAS,iCAAiC;AACjD,QAAI,OAAO,YAAY,YAAY;AACjC,WAAK,qBAAqB;AAAA,IAC5B,OAAO;AACL,WAAK,qBAAqB,MAAM,QAAQ,QAAQ,OAAO;AAAA,IACzD;AAEA,QAAI,OAAO,KAAK,eAAe,aAAa;AAC1C,WAAK,aAAa,IAAI;AAAA,IACxB;AACA,QAAI,OAAO,KAAK,aAAa,aAAa;AACxC,WAAK,WAAW,IAAI;AAAA,IACtB;AAAA,EACF;AAAA,EAEA,MAAM,OACJ,OACA,KACuB;AACvB,UAAM,EAAE,OAAO,IAAI,OAAO,CAAC;AAC3B,UAAM,cAAc,OAAO,UAAU,WAAW,QAAQ,MAAM;AAC9D,IAAAA,QAAO,aAAa,8BAA8B;AAClD,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,IAAAA,QAAO,OAAO,UAAU,UAAU,sCAAsC;AAExE,UAAM,wBAAwB;AAAA,MAC5B;AAAA,IACF;AACA,QAAI,uBAAuB;AACzB,YAAM,yBAAyB,qBAAqB;AAAA,IACtD;AACA,QAAI;AACJ,QAAI,MAAM,aAAa,uBAAuB;AAC5C,yBAAmB,MAAM;AAAA,IAC3B;AAEA,QAAI,oBAAoB,CAACD,cAAa,GAAG;AACvC,cAAQ;AAAA,QACN;AAAA,MACF;AACA,yBAAmB;AAAA,IACrB;AAEA,UAAM,UAAU,KAAK,WAAY,MAAM,KAAK,mBAAmB,QAAQ;AAEvE,QAAI,aAA+B;AACnC,QAAI,wBAA4C;AAChD,QAAI,kBAA2C;AAC/C,QAAI,qBAEY;AAChB,QAAI,kBAAkB;AACpB,2BAAqB,MAAM,gBAAgB;AAAA,QACzC;AAAA,QACA,oBAAoB;AAAA,MACtB,CAAC;AACD,MAAAC;AAAA,QACE,mBAAmB;AAAA,QACnB,gCAAgC,gBAAgB,IAC9C,mBAAmB,QAAQ,KAAK,mBAAmB,KAAK,KAAK,EAC/D;AAAA,MACF;AACA,8BAAwB,mBAAmB;AAC3C,wBAAkB,mBAAmB;AACrC,mBAAa,mBAAmB;AAAA,IAClC;AAEA,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,aAAa,aAAa,MAAM,IACzD,MAAM,gBAAgB;AAAA,MACpB,QAAQ,UAAU,KAAK;AAAA,MACvB;AAAA,MACA,0BAA0B;AAAA,MAC1B,cAAc;AAAA,IAChB,CAAC;AAEH,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,MACvC,gBAAgB,KAAK,UAAU,WAAW;AAAA,MAC1C;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAgC,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAC1E;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT,SAAS;AAAA,MACX;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,aAAa;AAAA,MACb,MAAM;AAAA,MACN;AAAA,MACA,WAAW,CAAC,CAAC;AAAA,MACb,OAAO;AAAA,IACT;AAEA,UAAM,WAA0B,CAAC;AACjC,KAAC,YAAY,YAAY,CAAC,GAAG,QAAQ,CAAC,SAAS;AAC7C,UAAI,QAAQ,MAAM;AAChB,cAAM,UAAU,YAAY,MAAM,EAAE;AAEpC,YAAI,CAAC,SAAS;AACZ,kBAAQ;AAAA,YACN,kCAAkC,KAAK,EAAE;AAAA,UAC3C;AACA;AAAA,QACF;AACA,iBAAS,KAAK,OAAO;AAAA,MACvB;AAAA,IACF,CAAC;AAED;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,IACF;AAEA,QAAI,UAAU;AACZ,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,IAAAA;AAAA,MACE,SAAS,UAAU;AAAA,MACnB,6CAA6C,SAAS,MAAM;AAAA,IAC9D;AAEA,QAAI,SAAS,WAAW,GAAG;AACzB,aAAO;AAAA,QACL,SAAS;AAAA,UACP,IAAI,SAAS,CAAC,EAAG;AAAA,UACjB,SAAS,SAAS,CAAC,EAAG;AAAA,UACtB,QAAQ,SAAS,CAAC,EAAG;AAAA,UACrB,MAAM,SAAS,CAAC,EAAG;AAAA,UACnB,QAAQ,SAAS,CAAC,EAAG,UAAU,CAAC;AAAA,UAChC,YAAY,SAAS,CAAC,EAAG;AAAA,QAC3B;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,WAAO;AAAA,MACL,SAAS;AAAA,MACT;AAAA,IACF;AAAA,EACF;AAAA,EAYA,MAAM,QACJ,YACA,KACc;AACd,IAAAA;AAAA,MACE,OAAO,eAAe,YAAY,OAAO,eAAe;AAAA,MACxD,kDAAkD,OAAO,UAAU;AAAA,IACrE;AACA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,SAAS;AAEvD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,IAAI,MAAM,qBAAwB;AAAA,MAC3D;AAAA,MACA,WAAW;AAAA,MACX,eAAe;AAAA,IACjB,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,IACzC;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAwB,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAClE;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,OAAO;AAAA,IACT;AAEA,UAAM,EAAE,KAAK,IAAI,eAAe,CAAC;AAGjC;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH;AAAA,MACF;AAAA,MACA;AAAA,IACF;AAEA,QAAI,YAAY,CAAC,MAAM;AACrB,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,WAAsD;AACjE,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,QAAQ;AACtD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,eAAe,MAAM,SAAS;AAAA,MAClC;AAAA,MACA;AAAA,IACF,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,aAAa,OAAO;AAAA,IAClD;AAEA,UAAM,EAAE,SAAS,KAAK,IAAI,aAAa;AACvC,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,eAAe;AAAA,MACf,kBAAkB;AAAA,MAClB,OAAO,OAAO,SAAY;AAAA,IAC5B;AACA,oBAAgB,UAAU,cAAc;AAExC,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA,OAAO,aAAa;AAAA,IACtB;AAAA,EACF;AAAA,EACA,MAAM,SACJ,QACA,KAGyD;AACzD,IAAAA,QAAO,QAAQ,yCAAyC;AACxD,UAAM,UAAU,MAAM,KAAK,mBAAmB,UAAU;AACxD,UAAM,EAAE,kBAAkB,KAAK,IAAI;AACnC,IAAAA,QAAO,kBAAkB,6CAA6C;AAEtE,UAAM,eAAe,4BAA4B;AAGjD,UAAM,kBAAkB;AACxB,UAAM,aAAmB,MAAM,QAAQ,MAAM,IACzC;AAAA,MACE,MAAM,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAChD,KAAK,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAC/C,OAAO;AAAA,MACP,QAAQ;AAAA,IACV,IACA;AAEJ,QAAI,eAAe,MAAM,wBAAwB;AAAA,MAC/C,gBAAgB;AAAA,MAChB;AAAA,MACA,sBAAsB;AAAA,QACpB;AAAA,UACE,MAAM;AAAA,QACR;AAAA,MACF;AAAA,MACA,iBAAiB;AAAA,IACnB,CAAC;AAED,QAAI,KAAK,WAAW;AAClB,YAAM,aAAa,iBAAiB,YAAY,QAAQ,IAAI;AAC5D,YAAM,4BAA4B,UAAU;AAC5C,qBAAe,MAAM;AAAA,QACnB;AAAA,QACA;AAAA,QACA,qBAAqB,oBAAoB;AAAA,MAC3C;AAAA,IACF;AAEA,UAAM,OAAe;AAAA,MACnB,EAAE,MAAM,UAAU,SAAS,aAAa;AAAA,MACxC;AAAA,QACE,MAAM;AAAA,QACN,SAAS;AAAA,UACP;AAAA,YACE,MAAM;AAAA,YACN,WAAW;AAAA,cACT,KAAK;AAAA,cACL,QAAQ;AAAA,YACV;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,WACJ,KAAK,cAAc;AAErB,UAAM,MAAM,MAAM,SAAS,8BAAmC;AAE9D,UAAM,EAAE,QAAQ,IAAI;AACpB,IAAAA,QAAO,CAAC,QAAQ,OAAO,oBAAoB,QAAQ,KAAK,EAAE;AAC1D,IAAAA,QAAO,QAAQ,aAAa,gCAAgC;AAC5D,WAAO;AAAA,EACT;AACF;;;AE1ZA,SAAS,eAAAE,cAAa,uBAAAD,4BAA2B;AAGjD,IAAO,cAAQ","names":["vlLocateMode","assert","MIDSCENE_MODEL_NAME","getAIConfig"],"ignoreList":[],"sources":["../../src/ai-model/action-executor.ts","../../src/ai-model/prompt/describe.ts","../../src/insight/index.ts","../../src/insight/utils.ts","../../src/index.ts"],"sourcesContent":["import type {\n ExecutionDump,\n ExecutionTask,\n ExecutionTaskApply,\n ExecutionTaskInsightLocateOutput,\n ExecutionTaskProgressOptions,\n ExecutionTaskReturn,\n ExecutorContext,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n MIDSCENE_MODEL_NAME,\n getAIConfig,\n uiTarsModelVersion,\n vlLocateMode,\n} from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\n\nexport class Executor {\n name: string;\n\n tasks: ExecutionTask[];\n\n // status of executor\n status: 'init' | 'pending' | 'running' | 'completed' | 'error';\n\n onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n\n constructor(\n name: string,\n options?: ExecutionTaskProgressOptions & {\n tasks?: ExecutionTaskApply[];\n },\n ) {\n this.status =\n options?.tasks && options.tasks.length > 0 ? 'pending' : 'init';\n this.name = name;\n this.tasks = (options?.tasks || []).map((item) =>\n this.markTaskAsPending(item),\n );\n this.onTaskStart = options?.onTaskStart;\n }\n\n private markTaskAsPending(task: ExecutionTaskApply): ExecutionTask {\n return {\n status: 'pending',\n ...task,\n };\n }\n\n async append(task: ExecutionTaskApply[] | ExecutionTaskApply): Promise<void> {\n assert(\n this.status !== 'error',\n `executor is in error state, cannot append task\\nerror=${this.latestErrorTask()?.error}\\n${this.latestErrorTask()?.errorStack}`,\n );\n if (Array.isArray(task)) {\n this.tasks.push(...task.map((item) => this.markTaskAsPending(item)));\n } else {\n this.tasks.push(this.markTaskAsPending(task));\n }\n if (this.status !== 'running') {\n this.status = 'pending';\n }\n }\n\n async flush(): Promise<any> {\n if (this.status === 'init' && this.tasks.length > 0) {\n console.warn(\n 'illegal state for executor, status is init but tasks are not empty',\n );\n }\n\n assert(this.status !== 'running', 'executor is already running');\n assert(this.status !== 'completed', 'executor is already completed');\n assert(this.status !== 'error', 'executor is in error state');\n\n const nextPendingIndex = this.tasks.findIndex(\n (task) => task.status === 'pending',\n );\n if (nextPendingIndex < 0) {\n // all tasks are completed\n return;\n }\n\n this.status = 'running';\n let taskIndex = nextPendingIndex;\n let successfullyCompleted = true;\n\n let previousFindOutput: ExecutionTaskInsightLocateOutput | undefined;\n\n while (taskIndex < this.tasks.length) {\n const task = this.tasks[taskIndex];\n assert(\n task.status === 'pending',\n `task status should be pending, but got: ${task.status}`,\n );\n task.timing = {\n start: Date.now(),\n };\n try {\n task.status = 'running';\n try {\n if (this.onTaskStart) {\n await this.onTaskStart(task);\n }\n } catch (e) {\n console.error('error in onTaskStart', e);\n }\n assert(\n ['Insight', 'Action', 'Planning'].indexOf(task.type) >= 0,\n `unsupported task type: ${task.type}`,\n );\n\n const { executor, param } = task;\n assert(executor, `executor is required for task type: ${task.type}`);\n\n let returnValue;\n const executorContext: ExecutorContext = {\n task,\n element: previousFindOutput?.element,\n };\n\n if (task.type === 'Insight') {\n assert(\n task.subType === 'Locate' ||\n task.subType === 'Query' ||\n task.subType === 'Assert' ||\n task.subType === 'Boolean' ||\n task.subType === 'Number' ||\n task.subType === 'String',\n `unsupported insight subType: ${task.subType}`,\n );\n returnValue = await task.executor(param, executorContext);\n if (task.subType === 'Locate') {\n previousFindOutput = (\n returnValue as ExecutionTaskReturn<ExecutionTaskInsightLocateOutput>\n )?.output;\n }\n } else if (task.type === 'Action' || task.type === 'Planning') {\n returnValue = await task.executor(param, executorContext);\n } else {\n console.warn(\n `unsupported task type: ${task.type}, will try to execute it directly`,\n );\n returnValue = await task.executor(param, executorContext);\n }\n\n Object.assign(task, returnValue);\n task.status = 'finished';\n task.timing.end = Date.now();\n task.timing.cost = task.timing.end - task.timing.start;\n taskIndex++;\n } catch (e: any) {\n successfullyCompleted = false;\n task.error =\n e?.message || (typeof e === 'string' ? e : 'error-without-message');\n task.errorStack = e.stack;\n\n task.status = 'failed';\n task.timing.end = Date.now();\n task.timing.cost = task.timing.end - task.timing.start;\n break;\n }\n }\n\n // set all remaining tasks as cancelled\n for (let i = taskIndex + 1; i < this.tasks.length; i++) {\n this.tasks[i].status = 'cancelled';\n }\n\n if (successfullyCompleted) {\n this.status = 'completed';\n } else {\n this.status = 'error';\n }\n\n if (this.tasks.length) {\n // return the last output\n const outputIndex = Math.min(taskIndex, this.tasks.length - 1);\n return this.tasks[outputIndex].output;\n }\n }\n\n isInErrorState(): boolean {\n return this.status === 'error';\n }\n\n latestErrorTask(): ExecutionTask | null {\n if (this.status !== 'error') {\n return null;\n }\n const errorTaskIndex = this.tasks.findIndex(\n (task) => task.status === 'failed',\n );\n if (errorTaskIndex >= 0) {\n return this.tasks[errorTaskIndex];\n }\n return null;\n }\n\n dump(): ExecutionDump {\n let modelDescription = '';\n\n if (vlLocateMode()) {\n const uiTarsModelVer = uiTarsModelVersion();\n if (uiTarsModelVer) {\n modelDescription = `UI-TARS=${uiTarsModelVer}`;\n } else {\n modelDescription = `${vlLocateMode()} mode`;\n }\n }\n const dumpData: ExecutionDump = {\n sdkVersion: getVersion(),\n model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n model_description: modelDescription,\n logTime: Date.now(),\n name: this.name,\n tasks: this.tasks,\n };\n return dumpData;\n }\n}\n","import { getPreferredLanguage } from '@midscene/shared/env';\n\nexport const elementDescriberInstruction = () => {\n return `Describe the element in the red rectangle for precise identification. Use ${getPreferredLanguage()}.\n\nRules:\n1. Start with element type (button, input, link, etc.) \n2. Include key identifiers:\n - Text content: \"with text 'Submit'\"\n - Visual features: \"blue background\", \"icon only\"\n - Position: \"top-right\", \"below search bar\"\n3. Keep description under 20 words\n4. Don't mention the red rectangle\n\nReturn JSON:\n{\n \"description\": \"brief element type with key identifiers\",\n \"error\"?: \"error message if any\"\n}`;\n};\n","import {\n AIActionType,\n type AIArgs,\n callAiFn,\n expandSearchArea,\n} from '@/ai-model/common';\nimport {\n AiExtractElementInfo,\n AiLocateElement,\n callToGetJSONObject,\n} from '@/ai-model/index';\nimport { AiAssert, AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n AIDescribeElementResponse,\n AIElementResponse,\n AIUsageInfo,\n BaseElement,\n DetailedLocateParam,\n DumpSubscriber,\n InsightAction,\n InsightAssertionResponse,\n InsightExtractOption,\n InsightExtractParam,\n InsightOptions,\n InsightTaskInfo,\n LocateResult,\n PartialInsightDumpFromSDK,\n Rect,\n UIContext,\n} from '@/types';\nimport {\n MIDSCENE_FORCE_DEEP_THINK,\n MIDSCENE_USE_QWEN_VL,\n getAIConfigInBoolean,\n vlLocateMode,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n context?: UIContext<BaseElement>;\n callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n ElementType extends BaseElement = BaseElement,\n ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n contextRetrieverFn: (\n action: InsightAction,\n ) => Promise<ContextType> | ContextType;\n\n aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n onceDumpUpdatedFn?: DumpSubscriber;\n\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n constructor(\n context:\n | ContextType\n | ((action: InsightAction) => Promise<ContextType> | ContextType),\n opt?: InsightOptions,\n ) {\n assert(context, 'context is required for Insight');\n if (typeof context === 'function') {\n this.contextRetrieverFn = context;\n } else {\n this.contextRetrieverFn = () => Promise.resolve(context);\n }\n\n if (typeof opt?.aiVendorFn !== 'undefined') {\n this.aiVendorFn = opt.aiVendorFn;\n }\n if (typeof opt?.taskInfo !== 'undefined') {\n this.taskInfo = opt.taskInfo;\n }\n }\n\n async locate(\n query: DetailedLocateParam,\n opt?: LocateOpts,\n ): Promise<LocateResult> {\n const { callAI } = opt || {};\n const queryPrompt = typeof query === 'string' ? query : query.prompt;\n assert(queryPrompt, 'query is required for locate');\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n assert(typeof query === 'object', 'query should be an object for locate');\n\n const globalDeepThinkSwitch = getAIConfigInBoolean(\n MIDSCENE_FORCE_DEEP_THINK,\n );\n if (globalDeepThinkSwitch) {\n debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n }\n let searchAreaPrompt;\n if (query.deepThink || globalDeepThinkSwitch) {\n searchAreaPrompt = query.prompt;\n }\n\n if (searchAreaPrompt && !vlLocateMode()) {\n console.warn(\n 'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n );\n searchAreaPrompt = undefined;\n }\n\n const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n let searchArea: Rect | undefined = undefined;\n let searchAreaRawResponse: string | undefined = undefined;\n let searchAreaUsage: AIUsageInfo | undefined = undefined;\n let searchAreaResponse:\n | Awaited<ReturnType<typeof AiLocateSection>>\n | undefined = undefined;\n if (searchAreaPrompt) {\n searchAreaResponse = await AiLocateSection({\n context,\n sectionDescription: searchAreaPrompt,\n });\n assert(\n searchAreaResponse.rect,\n `cannot find search area for \"${searchAreaPrompt}\"${\n searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n }`,\n );\n searchAreaRawResponse = searchAreaResponse.rawResponse;\n searchAreaUsage = searchAreaResponse.usage;\n searchArea = searchAreaResponse.rect;\n }\n\n const startTime = Date.now();\n const { parseResult, rect, elementById, rawResponse, usage } =\n await AiLocateElement({\n callAI: callAI || this.aiVendorFn,\n context,\n targetElementDescription: queryPrompt,\n searchConfig: searchAreaResponse,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(rawResponse),\n formatResponse: JSON.stringify(parseResult),\n usage,\n searchArea,\n searchAreaRawResponse,\n searchAreaUsage,\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'locate',\n userQuery: {\n element: queryPrompt,\n },\n matchedElement: [],\n matchedRect: rect,\n data: null,\n taskInfo,\n deepThink: !!searchArea,\n error: errorLog,\n };\n\n const elements: BaseElement[] = [];\n (parseResult.elements || []).forEach((item) => {\n if ('id' in item) {\n const element = elementById(item?.id);\n\n if (!element) {\n console.warn(\n `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n );\n return;\n }\n elements.push(element);\n }\n });\n\n emitInsightDump(\n {\n ...dumpData,\n matchedElement: elements,\n },\n dumpSubscriber,\n );\n\n if (errorLog) {\n throw new Error(errorLog);\n }\n\n assert(\n elements.length <= 1,\n `locate: multiple elements found, length = ${elements.length}`,\n );\n\n if (elements.length === 1) {\n return {\n element: {\n id: elements[0]!.id,\n indexId: elements[0]!.indexId,\n center: elements[0]!.center,\n rect: elements[0]!.rect,\n xpaths: elements[0]!.xpaths || [],\n attributes: elements[0]!.attributes,\n },\n rect,\n };\n }\n return {\n element: null,\n rect,\n };\n }\n\n async extract<T = any>(input: string, opt?: InsightExtractOption): Promise<T>;\n async extract<T extends Record<string, string>>(\n input: T,\n opt?: InsightExtractOption,\n ): Promise<Record<keyof T, any>>;\n async extract<T extends object>(\n input: Record<keyof T, string>,\n opt?: InsightExtractOption,\n ): Promise<T>;\n\n async extract<T>(\n dataDemand: InsightExtractParam,\n opt?: InsightExtractOption,\n ): Promise<any> {\n assert(\n typeof dataDemand === 'object' || typeof dataDemand === 'string',\n `dataDemand should be object or string, but get ${typeof dataDemand}`,\n );\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n const context = await this.contextRetrieverFn('extract');\n\n const startTime = Date.now();\n const { parseResult, usage } = await AiExtractElementInfo<T>({\n context,\n dataQuery: dataDemand,\n extractOption: opt,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(parseResult),\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'extract',\n userQuery: {\n dataDemand,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n error: errorLog,\n };\n\n const { data } = parseResult || {};\n\n // 4\n emitInsightDump(\n {\n ...dumpData,\n data,\n },\n dumpSubscriber,\n );\n\n if (errorLog && !data) {\n throw new Error(errorLog);\n }\n\n return {\n data,\n usage,\n };\n }\n\n async assert(assertion: string): Promise<InsightAssertionResponse> {\n if (typeof assertion !== 'string') {\n throw new Error(\n 'This is the assert method for Midscene, the first argument should be a string. If you want to use the assert method from Node.js, please import it from the Node.js assert module.',\n );\n }\n\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n const context = await this.contextRetrieverFn('assert');\n const startTime = Date.now();\n const assertResult = await AiAssert({\n assertion,\n context,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(assertResult.content),\n };\n\n const { thought, pass } = assertResult.content;\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'assert',\n userQuery: {\n assertion,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n assertionPass: pass,\n assertionThought: thought,\n error: pass ? undefined : thought,\n };\n emitInsightDump(dumpData, dumpSubscriber);\n\n return {\n pass,\n thought,\n usage: assertResult.usage,\n };\n }\n async describe(\n target: Rect | [number, number],\n opt?: {\n deepThink?: boolean;\n },\n ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n assert(target, 'target is required for insight.describe');\n const context = await this.contextRetrieverFn('describe');\n const { screenshotBase64, size } = context;\n assert(screenshotBase64, 'screenshot is required for insight.describe');\n\n const systemPrompt = elementDescriberInstruction();\n\n // Convert [x,y] center point to Rect if needed\n const defaultRectSize = 30;\n const targetRect: Rect = Array.isArray(target)\n ? {\n left: Math.floor(target[0] - defaultRectSize / 2),\n top: Math.floor(target[1] - defaultRectSize / 2),\n width: defaultRectSize,\n height: defaultRectSize,\n }\n : target;\n\n let imagePayload = await compositeElementInfoImg({\n inputImgBase64: screenshotBase64,\n size,\n elementsPositionInfo: [\n {\n rect: targetRect,\n },\n ],\n borderThickness: 3,\n });\n\n if (opt?.deepThink) {\n const searchArea = expandSearchArea(targetRect, context.size);\n debug('describe: set searchArea', searchArea);\n imagePayload = await cropByRect(\n imagePayload,\n searchArea,\n getAIConfigInBoolean(MIDSCENE_USE_QWEN_VL),\n );\n }\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n detail: 'high',\n },\n },\n ],\n },\n ];\n\n const callAIFn =\n this.aiVendorFn || callToGetJSONObject<AIDescribeElementResponse>;\n\n const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT);\n\n const { content } = res;\n assert(!content.error, `describe failed: ${content.error}`);\n assert(content.description, 'failed to describe the element');\n return content;\n }\n}\n","import type {\n DumpMeta,\n DumpSubscriber,\n InsightDump,\n PartialInsightDumpFromSDK,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport { MIDSCENE_MODEL_NAME, getAIConfig } from '@midscene/shared/env';\nimport { uuid } from '@midscene/shared/utils';\n\nexport function emitInsightDump(\n data: PartialInsightDumpFromSDK,\n dumpSubscriber?: DumpSubscriber,\n) {\n const baseData: DumpMeta = {\n sdkVersion: getVersion(),\n logTime: Date.now(),\n model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n };\n const finalData: InsightDump = {\n logId: uuid(),\n ...baseData,\n ...data,\n };\n\n dumpSubscriber?.(finalData);\n}\n","import { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n plan,\n describeUserPage,\n AiLocateElement,\n AiAssert,\n} from './ai-model/index';\n\nexport { getAIConfig, MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n MidsceneYamlScript,\n MidsceneYamlTask,\n MidsceneYamlFlowItem,\n MidsceneYamlFlowItemAIRightClick,\n} from './yaml';\n"]}
@@ -1,4 +1,4 @@
1
- import { l as AIUsageInfo, U as UIContext, ax as ReferenceImage, q as AIElementLocatorResponse, K as ElementById, g as InsightExtractOption, t as AIDataExtractionResponse, v as AIAssertionResponse, av as PageType, V as PlanningAIResponse } from './types-00c9086f.js';
1
+ import { l as AIUsageInfo, U as UIContext, ax as ReferenceImage, q as AIElementLocatorResponse, K as ElementById, g as InsightExtractOption, t as AIDataExtractionResponse, v as AIAssertionResponse, av as PageType, V as PlanningAIResponse } from './types-dce56c26.js';
2
2
  import { Rect, ElementTreeNode, BaseElement } from '@midscene/shared/types';
3
3
  import { ChatCompletionSystemMessageParam, ChatCompletionUserMessageParam } from 'openai/resources';
4
4
 
@@ -29,7 +29,7 @@ declare function elementByPositionWithElementInfo(treeRoot: ElementTreeNode<Base
29
29
  declare function describeUserPage<ElementType extends BaseElement = BaseElement>(context: Omit<UIContext<ElementType>, 'describer'>, opt?: {
30
30
  truncateTextLength?: number;
31
31
  filterNonTextContent?: boolean;
32
- domIncluded?: boolean;
32
+ domIncluded?: boolean | 'visible-only';
33
33
  visibleOnly?: boolean;
34
34
  }): Promise<{
35
35
  description: string;
@@ -8,7 +8,7 @@ interface LocateOption {
8
8
  cacheable?: boolean;
9
9
  }
10
10
  interface InsightExtractOption {
11
- domIncluded?: boolean;
11
+ domIncluded?: boolean | 'visible-only';
12
12
  screenshotIncluded?: boolean;
13
13
  }
14
14
  interface ReferenceImage {
@@ -37,6 +37,7 @@ interface MidsceneYamlTask {
37
37
  }
38
38
  interface MidsceneYamlScriptEnvBase {
39
39
  output?: string;
40
+ unstableLogContent?: boolean | string;
40
41
  aiActionContext?: string;
41
42
  }
42
43
  interface MidsceneYamlScriptWebEnv extends MidsceneYamlScriptEnvBase {
@@ -141,9 +142,10 @@ interface ScriptPlayerTaskStatus extends MidsceneYamlTask {
141
142
  type ScriptPlayerStatusValue = 'init' | 'running' | 'done' | 'error';
142
143
 
143
144
  type AIUsageInfo = Record<string, any> & {
144
- prompt_tokens: number;
145
- completion_tokens: number;
146
- total_tokens: number;
145
+ prompt_tokens: number | undefined;
146
+ completion_tokens: number | undefined;
147
+ total_tokens: number | undefined;
148
+ time_cost: number | undefined;
147
149
  };
148
150
  /**
149
151
  * openai
@@ -220,7 +222,6 @@ interface AgentDescribeElementAtPointResult {
220
222
  */
221
223
  declare abstract class UIContext<ElementType extends BaseElement = BaseElement> {
222
224
  abstract screenshotBase64: string;
223
- abstract content: ElementType[];
224
225
  abstract tree: ElementTreeNode<ElementType>;
225
226
  abstract size: Size;
226
227
  }
@@ -413,7 +414,6 @@ type ExecutionTask<E extends ExecutionTaskApply<any, any, any> = ExecutionTaskAp
413
414
  start: number;
414
415
  end?: number;
415
416
  cost?: number;
416
- aiCost?: number;
417
417
  };
418
418
  usage?: AIUsageInfo;
419
419
  };
@@ -1,4 +1,4 @@
1
- import { R as ReportDumpWithAttributes } from './types-00c9086f.js';
1
+ import { R as ReportDumpWithAttributes } from './types-dce56c26.js';
2
2
  import { Rect } from '@midscene/shared/types';
3
3
  import '@midscene/shared/constants';
4
4
  import 'openai/resources';
package/dist/lib/utils.js CHANGED
@@ -12,7 +12,7 @@
12
12
 
13
13
 
14
14
 
15
- var _chunkD4ZWVRRGjs = require('./chunk-D4ZWVRRG.js');
15
+ var _chunkA7OTPIZJjs = require('./chunk-A7OTPIZJ.js');
16
16
 
17
17
 
18
18
 
@@ -27,4 +27,4 @@ var _chunkD4ZWVRRGjs = require('./chunk-D4ZWVRRG.js');
27
27
 
28
28
 
29
29
 
30
- exports.getTmpDir = _chunkD4ZWVRRGjs.getTmpDir; exports.getTmpFile = _chunkD4ZWVRRGjs.getTmpFile; exports.getVersion = _chunkD4ZWVRRGjs.getVersion; exports.groupedActionDumpFileExt = _chunkD4ZWVRRGjs.groupedActionDumpFileExt; exports.overlapped = _chunkD4ZWVRRGjs.overlapped; exports.replaceStringWithFirstAppearance = _chunkD4ZWVRRGjs.replaceStringWithFirstAppearance; exports.replacerForPageObject = _chunkD4ZWVRRGjs.replacerForPageObject; exports.reportHTMLContent = _chunkD4ZWVRRGjs.reportHTMLContent; exports.sleep = _chunkD4ZWVRRGjs.sleep; exports.stringifyDumpData = _chunkD4ZWVRRGjs.stringifyDumpData; exports.uploadTestInfoToServer = _chunkD4ZWVRRGjs.uploadTestInfoToServer; exports.writeDumpReport = _chunkD4ZWVRRGjs.writeDumpReport; exports.writeLogFile = _chunkD4ZWVRRGjs.writeLogFile;
30
+ exports.getTmpDir = _chunkA7OTPIZJjs.getTmpDir; exports.getTmpFile = _chunkA7OTPIZJjs.getTmpFile; exports.getVersion = _chunkA7OTPIZJjs.getVersion; exports.groupedActionDumpFileExt = _chunkA7OTPIZJjs.groupedActionDumpFileExt; exports.overlapped = _chunkA7OTPIZJjs.overlapped; exports.replaceStringWithFirstAppearance = _chunkA7OTPIZJjs.replaceStringWithFirstAppearance; exports.replacerForPageObject = _chunkA7OTPIZJjs.replacerForPageObject; exports.reportHTMLContent = _chunkA7OTPIZJjs.reportHTMLContent; exports.sleep = _chunkA7OTPIZJjs.sleep; exports.stringifyDumpData = _chunkA7OTPIZJjs.stringifyDumpData; exports.uploadTestInfoToServer = _chunkA7OTPIZJjs.uploadTestInfoToServer; exports.writeDumpReport = _chunkA7OTPIZJjs.writeDumpReport; exports.writeLogFile = _chunkA7OTPIZJjs.writeLogFile;
@@ -1,13 +1,19 @@
1
- import { l as AIUsageInfo, T as PlanningAction, j as MidsceneYamlFlowItem } from './types-00c9086f.js';
1
+ import { l as AIUsageInfo, T as PlanningAction, j as MidsceneYamlFlowItem } from './types-dce56c26.js';
2
+ import OpenAI from 'openai';
2
3
  import { ChatCompletionMessageParam } from 'openai/resources';
3
4
  export { ChatCompletionMessageParam } from 'openai/resources';
4
- import { b as AIActionType } from './llm-planning-453ffce5.js';
5
- export { a as AiAssert, f as AiExtractElementInfo, A as AiLocateElement, g as AiLocateSection, h as adaptBboxToRect, c as callAiFn, d as describeUserPage, e as elementByPositionWithElementInfo, p as plan } from './llm-planning-453ffce5.js';
5
+ import { b as AIActionType } from './llm-planning-a951deb9.js';
6
+ export { a as AiAssert, f as AiExtractElementInfo, A as AiLocateElement, g as AiLocateSection, h as adaptBboxToRect, c as callAiFn, d as describeUserPage, e as elementByPositionWithElementInfo, p as plan } from './llm-planning-a951deb9.js';
6
7
  import { vlLocateMode } from '@midscene/shared/env';
8
+ import { ChromeRecordedEvent as ChromeRecordedEvent$1 } from '@midscene/recorder';
7
9
  import { actionParser } from '@ui-tars/action-parser';
8
10
  import { Size } from '@midscene/shared/types';
9
11
  import '@midscene/shared/constants';
10
12
 
13
+ declare function call(messages: ChatCompletionMessageParam[], AIActionTypeValue: AIActionType, responseFormat?: OpenAI.ChatCompletionCreateParams['response_format'] | OpenAI.ResponseFormatJSONObject): Promise<{
14
+ content: string;
15
+ usage?: AIUsageInfo;
16
+ }>;
11
17
  declare function callToGetJSONObject<T>(messages: ChatCompletionMessageParam[], AIActionTypeValue: AIActionType): Promise<{
12
18
  content: T;
13
19
  usage?: AIUsageInfo;
@@ -15,6 +21,48 @@ declare function callToGetJSONObject<T>(messages: ChatCompletionMessageParam[],
15
21
 
16
22
  declare function systemPromptToLocateElement(vlMode: ReturnType<typeof vlLocateMode>): string;
17
23
 
24
+ interface PlaywrightGenerationOptions {
25
+ testName?: string;
26
+ includeScreenshots?: boolean;
27
+ includeTimestamps?: boolean;
28
+ maxScreenshots?: number;
29
+ description?: string;
30
+ viewportSize?: {
31
+ width: number;
32
+ height: number;
33
+ };
34
+ waitForNetworkIdle?: boolean;
35
+ waitForNetworkIdleTimeout?: number;
36
+ }
37
+ interface ChromeRecordedEvent {
38
+ type: string;
39
+ timestamp: number;
40
+ url?: string;
41
+ title?: string;
42
+ elementDescription?: string;
43
+ value?: string;
44
+ pageInfo?: any;
45
+ elementRect?: any;
46
+ screenshotBefore?: string;
47
+ screenshotAfter?: string;
48
+ screenshotWithBox?: string;
49
+ }
50
+ /**
51
+ * Generates Playwright test code from recorded events
52
+ */
53
+ declare const generatePlaywrightTest: (events: ChromeRecordedEvent[], options?: PlaywrightGenerationOptions) => Promise<string>;
54
+
55
+ interface YamlGenerationOptions {
56
+ testName?: string;
57
+ includeTimestamps?: boolean;
58
+ maxScreenshots?: number;
59
+ description?: string;
60
+ }
61
+ /**
62
+ * Generates YAML test configuration from recorded events using AI
63
+ */
64
+ declare const generateYamlTest: (events: ChromeRecordedEvent$1[], options?: YamlGenerationOptions) => Promise<string>;
65
+
18
66
  declare function vlmPlanning(options: {
19
67
  userInstruction: string;
20
68
  conversationHistory: ChatCompletionMessageParam[];
@@ -30,4 +78,4 @@ declare function vlmPlanning(options: {
30
78
  }>;
31
79
  declare function resizeImageForUiTars(imageBase64: string, size: Size): Promise<string>;
32
80
 
33
- export { callToGetJSONObject, resizeImageForUiTars, systemPromptToLocateElement, vlmPlanning };
81
+ export { AIActionType, call as callAi, callToGetJSONObject, generatePlaywrightTest, generateYamlTest, resizeImageForUiTars, systemPromptToLocateElement, vlmPlanning };
@@ -1,7 +1,7 @@
1
- import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightExtractOption, h as InsightAssertionResponse, A as AIDescribeElementResponse } from './types-00c9086f.js';
2
- export { v as AIAssertionResponse, t as AIDataExtractionResponse, r as AIElementCoordinatesResponse, q as AIElementLocatorResponse, s as AIElementResponse, m as AIResponseFormat, u as AISectionLocatorResponse, p as AISingleElementResponse, n as AISingleElementResponseById, o as AISingleElementResponseByPosition, l as AIUsageInfo, Q as AgentAssertOpt, y as AgentDescribeElementAtPointResult, N as AgentWaitForOpt, a4 as BaseAgentParserOpt, C as CallAIFn, a3 as Color, G as DumpMeta, K as ElementById, z as EnsureObject, a7 as ExecutionRecorderItem, ap as ExecutionTaskAction, ao as ExecutionTaskActionApply, an as ExecutionTaskInsightAssertion, am as ExecutionTaskInsightAssertionApply, al as ExecutionTaskInsightAssertionParam, ae as ExecutionTaskInsightDumpLog, ag as ExecutionTaskInsightLocate, af as ExecutionTaskInsightLocateApply, ad as ExecutionTaskInsightLocateOutput, ac as ExecutionTaskInsightLocateParam, ak as ExecutionTaskInsightQuery, aj as ExecutionTaskInsightQueryApply, ai as ExecutionTaskInsightQueryOutput, ah as ExecutionTaskInsightQueryParam, ar as ExecutionTaskLog, aq as ExecutionTaskLogApply, at as ExecutionTaskPlanning, as as ExecutionTaskPlanningApply, ab as ExecutionTaskReturn, a8 as ExecutionTaskType, a9 as ExecutorContext, aT as FreeFn, au as GroupedActionDump, H as InsightDump, B as InsightExtractParam, J as LiteUISection, aw as LocateOption, F as LocateResultElement, x as LocateValidatorResult, w as LocatorValidatorOption, j as MidsceneYamlFlowItem, aD as MidsceneYamlFlowItemAIAction, aE as MidsceneYamlFlowItemAIAssert, aI as MidsceneYamlFlowItemAIBoolean, aM as MidsceneYamlFlowItemAIHover, aN as MidsceneYamlFlowItemAIInput, aO as MidsceneYamlFlowItemAIKeyboardPress, aJ as MidsceneYamlFlowItemAILocate, aH as MidsceneYamlFlowItemAINString, aG as MidsceneYamlFlowItemAINumber, aF as MidsceneYamlFlowItemAIQuery, k as MidsceneYamlFlowItemAIRightClick, aP as MidsceneYamlFlowItemAIScroll, aL as MidsceneYamlFlowItemAITap, aK as MidsceneYamlFlowItemAIWaitFor, aQ as MidsceneYamlFlowItemEvaluateJavaScript, aS as MidsceneYamlFlowItemLogScreenshot, aR as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aB as MidsceneYamlScriptAndroidEnv, aC as MidsceneYamlScriptEnv, az as MidsceneYamlScriptEnvBase, aA as MidsceneYamlScriptWebEnv, i as MidsceneYamlTask, O as OnTaskStartTip, av as PageType, P as PartialInsightDumpFromSDK, V as PlanningAIResponse, T as PlanningAction, $ as PlanningActionParamAssert, a1 as PlanningActionParamError, X as PlanningActionParamHover, Z as PlanningActionParamInputOrKeyPress, Y as PlanningActionParamRightClick, _ as PlanningActionParamScroll, a0 as PlanningActionParamSleep, W as PlanningActionParamTap, a2 as PlanningActionParamWaitFor, S as PlanningLocateParam, a6 as PlaywrightParserOpt, a5 as PuppeteerParserOpt, ax as ReferenceImage, R as ReportDumpWithAttributes, aV as ScriptPlayerStatusValue, aU as ScriptPlayerTaskStatus, aa as TaskCacheInfo, ay as scrollParam } from './types-00c9086f.js';
3
- import { c as callAiFn } from './llm-planning-453ffce5.js';
4
- export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-453ffce5.js';
1
+ import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightExtractOption, h as InsightAssertionResponse, A as AIDescribeElementResponse } from './types-dce56c26.js';
2
+ export { v as AIAssertionResponse, t as AIDataExtractionResponse, r as AIElementCoordinatesResponse, q as AIElementLocatorResponse, s as AIElementResponse, m as AIResponseFormat, u as AISectionLocatorResponse, p as AISingleElementResponse, n as AISingleElementResponseById, o as AISingleElementResponseByPosition, l as AIUsageInfo, Q as AgentAssertOpt, y as AgentDescribeElementAtPointResult, N as AgentWaitForOpt, a4 as BaseAgentParserOpt, C as CallAIFn, a3 as Color, G as DumpMeta, K as ElementById, z as EnsureObject, a7 as ExecutionRecorderItem, ap as ExecutionTaskAction, ao as ExecutionTaskActionApply, an as ExecutionTaskInsightAssertion, am as ExecutionTaskInsightAssertionApply, al as ExecutionTaskInsightAssertionParam, ae as ExecutionTaskInsightDumpLog, ag as ExecutionTaskInsightLocate, af as ExecutionTaskInsightLocateApply, ad as ExecutionTaskInsightLocateOutput, ac as ExecutionTaskInsightLocateParam, ak as ExecutionTaskInsightQuery, aj as ExecutionTaskInsightQueryApply, ai as ExecutionTaskInsightQueryOutput, ah as ExecutionTaskInsightQueryParam, ar as ExecutionTaskLog, aq as ExecutionTaskLogApply, at as ExecutionTaskPlanning, as as ExecutionTaskPlanningApply, ab as ExecutionTaskReturn, a8 as ExecutionTaskType, a9 as ExecutorContext, aT as FreeFn, au as GroupedActionDump, H as InsightDump, B as InsightExtractParam, J as LiteUISection, aw as LocateOption, F as LocateResultElement, x as LocateValidatorResult, w as LocatorValidatorOption, j as MidsceneYamlFlowItem, aD as MidsceneYamlFlowItemAIAction, aE as MidsceneYamlFlowItemAIAssert, aI as MidsceneYamlFlowItemAIBoolean, aM as MidsceneYamlFlowItemAIHover, aN as MidsceneYamlFlowItemAIInput, aO as MidsceneYamlFlowItemAIKeyboardPress, aJ as MidsceneYamlFlowItemAILocate, aH as MidsceneYamlFlowItemAINString, aG as MidsceneYamlFlowItemAINumber, aF as MidsceneYamlFlowItemAIQuery, k as MidsceneYamlFlowItemAIRightClick, aP as MidsceneYamlFlowItemAIScroll, aL as MidsceneYamlFlowItemAITap, aK as MidsceneYamlFlowItemAIWaitFor, aQ as MidsceneYamlFlowItemEvaluateJavaScript, aS as MidsceneYamlFlowItemLogScreenshot, aR as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aB as MidsceneYamlScriptAndroidEnv, aC as MidsceneYamlScriptEnv, az as MidsceneYamlScriptEnvBase, aA as MidsceneYamlScriptWebEnv, i as MidsceneYamlTask, O as OnTaskStartTip, av as PageType, P as PartialInsightDumpFromSDK, V as PlanningAIResponse, T as PlanningAction, $ as PlanningActionParamAssert, a1 as PlanningActionParamError, X as PlanningActionParamHover, Z as PlanningActionParamInputOrKeyPress, Y as PlanningActionParamRightClick, _ as PlanningActionParamScroll, a0 as PlanningActionParamSleep, W as PlanningActionParamTap, a2 as PlanningActionParamWaitFor, S as PlanningLocateParam, a6 as PlaywrightParserOpt, a5 as PuppeteerParserOpt, ax as ReferenceImage, R as ReportDumpWithAttributes, aV as ScriptPlayerStatusValue, aU as ScriptPlayerTaskStatus, aa as TaskCacheInfo, ay as scrollParam } from './types-dce56c26.js';
3
+ import { c as callAiFn } from './llm-planning-a951deb9.js';
4
+ export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-a951deb9.js';
5
5
  import { BaseElement, Rect } from '@midscene/shared/types';
6
6
  export { BaseElement, ElementTreeNode, Point, Rect, Size } from '@midscene/shared/types';
7
7
  export { getVersion } from './utils.js';
@@ -1,4 +1,4 @@
1
- import { l as AIUsageInfo, U as UIContext, ax as ReferenceImage, q as AIElementLocatorResponse, K as ElementById, g as InsightExtractOption, t as AIDataExtractionResponse, v as AIAssertionResponse, av as PageType, V as PlanningAIResponse } from './types-00c9086f.js';
1
+ import { l as AIUsageInfo, U as UIContext, ax as ReferenceImage, q as AIElementLocatorResponse, K as ElementById, g as InsightExtractOption, t as AIDataExtractionResponse, v as AIAssertionResponse, av as PageType, V as PlanningAIResponse } from './types-dce56c26.js';
2
2
  import { Rect, ElementTreeNode, BaseElement } from '@midscene/shared/types';
3
3
  import { ChatCompletionSystemMessageParam, ChatCompletionUserMessageParam } from 'openai/resources';
4
4
 
@@ -29,7 +29,7 @@ declare function elementByPositionWithElementInfo(treeRoot: ElementTreeNode<Base
29
29
  declare function describeUserPage<ElementType extends BaseElement = BaseElement>(context: Omit<UIContext<ElementType>, 'describer'>, opt?: {
30
30
  truncateTextLength?: number;
31
31
  filterNonTextContent?: boolean;
32
- domIncluded?: boolean;
32
+ domIncluded?: boolean | 'visible-only';
33
33
  visibleOnly?: boolean;
34
34
  }): Promise<{
35
35
  description: string;