@midscene/core 1.4.5 → 1.4.7

package/dist/es/agent/utils.mjs

@@ -2,31 +2,60 @@ import { ScreenshotItem } from "../screenshot-item.mjs";
 import { uploadTestInfoToServer } from "../utils.mjs";
 import { MIDSCENE_REPORT_QUIET, MIDSCENE_REPORT_TAG_NAME, globalConfigManager } from "@midscene/shared/env";
 import { generateElementByRect } from "@midscene/shared/extractor";
+import { imageInfoOfBase64, resizeImgBase64 } from "@midscene/shared/img";
 import { getDebug } from "@midscene/shared/logger";
 import { assert, logMsg, uuid } from "@midscene/shared/utils";
 import dayjs from "dayjs";
-import { debug } from "./task-cache.mjs";
-const debugProfile = getDebug('web:tool:profile');
+import { debug as external_task_cache_mjs_debug } from "./task-cache.mjs";
 async function commonContextParser(interfaceInstance, _opt) {
+    const debug = getDebug('commonContextParser');
     assert(interfaceInstance, 'interfaceInstance is required');
-    debugProfile("Getting interface description");
+    debug("Getting interface description");
     const description = interfaceInstance.describe?.() || '';
-    debugProfile("Interface description end");
-    debugProfile('Uploading test info to server');
+    debug("Interface description end");
+    debug('Uploading test info to server');
     uploadTestInfoToServer({
         testUrl: description,
         serverUrl: _opt.uploadServerUrl
     });
-    debugProfile('UploadTestInfoToServer end');
+    debug('UploadTestInfoToServer end');
+    debug('will get size');
+    const interfaceSize = await interfaceInstance.size();
+    const { width: logicalWidth, height: logicalHeight } = interfaceSize;
+    if (interfaceSize.dpr) console.warn('Warning: return value of interface.size() include a dpr property, which is not expected and ignored. ');
+    if (!Number.isFinite(logicalWidth) || !Number.isFinite(logicalHeight)) throw new Error(`Invalid interface size: width and height must be finite numbers. Received width: ${logicalWidth}, height: ${logicalHeight}`);
+    debug(`size: ${logicalWidth}x${logicalHeight}`);
     const screenshotBase64 = await interfaceInstance.screenshotBase64();
     assert(screenshotBase64, 'screenshotBase64 is required');
-    debugProfile('will get size');
-    const size = await interfaceInstance.size();
-    debugProfile(`size: ${size.width}x${size.height} dpr: ${size.dpr}`);
-    const screenshot = ScreenshotItem.create(screenshotBase64);
+    debug('will get screenshot dimensions');
+    const { width: imgWidth, height: imgHeight } = await imageInfoOfBase64(screenshotBase64);
+    debug('screenshot dimensions', imgWidth, 'x', imgHeight);
+    const shrinkFactor = imgWidth / logicalWidth;
+    debug('calculated shrink factor:', shrinkFactor);
+    if (1 !== shrinkFactor) {
+        const targetWidth = Math.round(imgWidth / shrinkFactor);
+        const targetHeight = Math.round(imgHeight / shrinkFactor);
+        debug(`Applying screenshot shrink factor: ${shrinkFactor} (physical: ${imgWidth}x${imgHeight} -> target: ${targetWidth}x${targetHeight})`);
+        const resizedBase64 = await resizeImgBase64(screenshotBase64, {
+            width: targetWidth,
+            height: targetHeight
+        });
+        return {
+            shotSize: {
+                width: targetWidth,
+                height: targetHeight
+            },
+            deprecatedDpr: shrinkFactor,
+            screenshot: ScreenshotItem.create(resizedBase64)
+        };
+    }
     return {
-        size,
-        screenshot
+        shotSize: {
+            width: imgWidth,
+            height: imgHeight
+        },
+        deprecatedDpr: 1,
+        screenshot: ScreenshotItem.create(screenshotBase64)
     };
 }
 function getReportFileName(tag = 'web') {
@@ -86,9 +115,9 @@ function matchElementFromPlan(planLocateParam) {
 }
 async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable) {
     if (!cacheEntry) return;
-    if (false === cacheable) return void debug('cache disabled for prompt: %s', cachePrompt);
+    if (false === cacheable) return void external_task_cache_mjs_debug('cache disabled for prompt: %s', cachePrompt);
     if (!context.taskCache?.isCacheResultUsed) return;
-    if (!context.interfaceInstance.rectMatchesCacheFeature) return void debug('interface does not implement rectMatchesCacheFeature, skip cache');
+    if (!context.interfaceInstance.rectMatchesCacheFeature) return void external_task_cache_mjs_debug('interface does not implement rectMatchesCacheFeature, skip cache');
     try {
         const rect = await context.interfaceInstance.rectMatchesCacheFeature(cacheEntry);
         const element = {
@@ -99,14 +128,14 @@ async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable
             rect,
             description: 'string' == typeof cachePrompt ? cachePrompt : cachePrompt.prompt || ''
         };
-        debug('cache hit, prompt: %s', cachePrompt);
+        external_task_cache_mjs_debug('cache hit, prompt: %s', cachePrompt);
         return element;
     } catch (error) {
-        debug('rectMatchesCacheFeature error: %s', error);
+        external_task_cache_mjs_debug('rectMatchesCacheFeature error: %s', error);
         return;
     }
 }
-const getMidsceneVersion = ()=>"1.4.5";
+const getMidsceneVersion = ()=>"1.4.7";
 const parsePrompt = (prompt)=>{
     if ('string' == typeof prompt) return {
         textPrompt: prompt,
@@ -120,6 +149,50 @@ const parsePrompt = (prompt)=>{
         } : void 0
     };
 };
-export { commonContextParser, generateCacheId, getCurrentExecutionFile, getMidsceneVersion, getReportFileName, ifPlanLocateParamIsBbox, matchElementFromCache, matchElementFromPlan, parsePrompt, printReportMsg };
+const transformScreenshotElementToLogical = (element, shrunkShotToLogicalRatio)=>{
+    if (1 === shrunkShotToLogicalRatio) return element;
+    return {
+        ...element,
+        center: [
+            Math.round(element.center[0] / shrunkShotToLogicalRatio),
+            Math.round(element.center[1] / shrunkShotToLogicalRatio)
+        ],
+        rect: {
+            ...element.rect,
+            left: Math.round(element.rect.left / shrunkShotToLogicalRatio),
+            top: Math.round(element.rect.top / shrunkShotToLogicalRatio),
+            width: Math.round(element.rect.width / shrunkShotToLogicalRatio),
+            height: Math.round(element.rect.height / shrunkShotToLogicalRatio)
+        }
+    };
+};
+const transformLogicalElementToScreenshot = (element, shrunkShotToLogicalRatio)=>{
+    if (1 === shrunkShotToLogicalRatio) return element;
+    return {
+        ...element,
+        center: [
+            Math.round(element.center[0] * shrunkShotToLogicalRatio),
+            Math.round(element.center[1] * shrunkShotToLogicalRatio)
+        ],
+        rect: {
+            ...element.rect,
+            left: Math.round(element.rect.left * shrunkShotToLogicalRatio),
+            top: Math.round(element.rect.top * shrunkShotToLogicalRatio),
+            width: Math.round(element.rect.width * shrunkShotToLogicalRatio),
+            height: Math.round(element.rect.height * shrunkShotToLogicalRatio)
+        }
+    };
+};
+const transformLogicalRectToScreenshotRect = (rect, shrunkShotToLogicalRatio)=>{
+    if (1 === shrunkShotToLogicalRatio) return rect;
+    return {
+        ...rect,
+        left: Math.round(rect.left * shrunkShotToLogicalRatio),
+        top: Math.round(rect.top * shrunkShotToLogicalRatio),
+        width: Math.round(rect.width * shrunkShotToLogicalRatio),
+        height: Math.round(rect.height * shrunkShotToLogicalRatio)
+    };
+};
+export { commonContextParser, generateCacheId, getCurrentExecutionFile, getMidsceneVersion, getReportFileName, ifPlanLocateParamIsBbox, matchElementFromCache, matchElementFromPlan, parsePrompt, printReportMsg, transformLogicalElementToScreenshot, transformLogicalRectToScreenshotRect, transformScreenshotElementToLogical };
 
 //# sourceMappingURL=utils.mjs.map

package/dist/es/agent/utils.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"agent/utils.mjs","sources":["../../../src/agent/utils.ts"],"sourcesContent":["import type { TMultimodalPrompt, TUserPrompt } from '@/common';\nimport type { AbstractInterface } from '@/device';\nimport { ScreenshotItem } from '@/screenshot-item';\nimport type {\n  ElementCacheFeature,\n  LocateResultElement,\n  PlanningLocateParam,\n  UIContext,\n} from '@/types';\nimport { uploadTestInfoToServer } from '@/utils';\nimport {\n  MIDSCENE_REPORT_QUIET,\n  MIDSCENE_REPORT_TAG_NAME,\n  globalConfigManager,\n} from '@midscene/shared/env';\nimport { generateElementByRect } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { assert, logMsg, uuid } from '@midscene/shared/utils';\nimport dayjs from 'dayjs';\nimport type { TaskCache } from './task-cache';\nimport { debug as cacheDebug } from './task-cache';\n\nconst debugProfile = getDebug('web:tool:profile');\n\nexport async function commonContextParser(\n  interfaceInstance: AbstractInterface,\n  _opt: { uploadServerUrl?: string },\n): Promise<UIContext> {\n  assert(interfaceInstance, 'interfaceInstance is required');\n\n  debugProfile('Getting interface description');\n  const description = interfaceInstance.describe?.() || '';\n  debugProfile('Interface description end');\n\n  debugProfile('Uploading test info to server');\n  uploadTestInfoToServer({\n    testUrl: description,\n    serverUrl: _opt.uploadServerUrl,\n  });\n  debugProfile('UploadTestInfoToServer end');\n\n  const screenshotBase64 = await interfaceInstance.screenshotBase64();\n  assert(screenshotBase64!, 'screenshotBase64 is required');\n\n  debugProfile('will get size');\n  const size = await interfaceInstance.size();\n  debugProfile(`size: ${size.width}x${size.height} dpr: ${size.dpr}`);\n\n  const screenshot = ScreenshotItem.create(screenshotBase64!);\n\n  return {\n    size,\n    screenshot,\n  };\n}\n\nexport function getReportFileName(tag = 'web') {\n  const reportTagName = globalConfigManager.getEnvConfigValue(\n    MIDSCENE_REPORT_TAG_NAME,\n  );\n  const dateTimeInFileName = dayjs().format('YYYY-MM-DD_HH-mm-ss');\n  // ensure uniqueness at the same time\n  const uniqueId = uuid().substring(0, 8);\n  return `${reportTagName || tag}-${dateTimeInFileName}-${uniqueId}`;\n}\n\nexport function printReportMsg(filepath: string) {\n  if (globalConfigManager.getEnvConfigInBoolean(MIDSCENE_REPORT_QUIET)) {\n    return;\n  }\n  logMsg(`Midscene - report file updated: ${filepath}`);\n}\n\n/**\n * Get the current execution file name\n * @returns The name of the current execution file\n */\nexport function getCurrentExecutionFile(trace?: string): string | false {\n  const error = new Error();\n  const stackTrace = trace || error.stack;\n  const pkgDir = process.cwd() || '';\n  if (stackTrace) {\n    const stackLines = stackTrace.split('\\n');\n    for (const line of stackLines) {\n      if (\n        line.includes('.spec.') ||\n        line.includes('.test.') ||\n        line.includes('.ts') ||\n        line.includes('.js')\n      ) {\n        const match = line.match(/(?:at\\s+)?(.*?\\.(?:spec|test)\\.[jt]s)/);\n        if (match?.[1]) {\n          const targetFileName = match[1]\n            .replace(pkgDir, '')\n            .trim()\n            .replace('at ', '');\n          return targetFileName;\n        }\n      }\n    }\n  }\n  return false;\n}\n\nconst testFileIndex = new Map<string, number>();\n\nexport function generateCacheId(fileName?: string): string {\n  let taskFile = fileName || getCurrentExecutionFile();\n  if (!taskFile) {\n    taskFile = uuid();\n    console.warn(\n      'Midscene - using random UUID for cache id. Cache may be invalid.',\n    );\n  }\n\n  if (testFileIndex.has(taskFile)) {\n    const currentIndex = testFileIndex.get(taskFile);\n    if (currentIndex !== undefined) {\n      testFileIndex.set(taskFile, currentIndex + 1);\n    }\n  } else {\n    testFileIndex.set(taskFile, 1);\n  }\n  return `${taskFile}-${testFileIndex.get(taskFile)}`;\n}\n\nexport function ifPlanLocateParamIsBbox(\n  planLocateParam: PlanningLocateParam,\n): boolean {\n  return !!(\n    planLocateParam.bbox &&\n    Array.isArray(planLocateParam.bbox) &&\n    planLocateParam.bbox.length === 4\n  );\n}\n\nexport function matchElementFromPlan(\n  planLocateParam: PlanningLocateParam,\n): LocateResultElement | undefined {\n  if (!planLocateParam) {\n    return undefined;\n  }\n\n  if (planLocateParam.bbox) {\n    // Convert bbox [x1, y1, x2, y2] to rect {left, top, width, height}\n    const rect = {\n      left: planLocateParam.bbox[0],\n      top: planLocateParam.bbox[1],\n      width: planLocateParam.bbox[2] - planLocateParam.bbox[0] + 1,\n      height: planLocateParam.bbox[3] - planLocateParam.bbox[1] + 1,\n    };\n\n    const element = generateElementByRect(\n      rect,\n      typeof planLocateParam.prompt === 'string'\n        ? planLocateParam.prompt\n        : planLocateParam.prompt?.prompt || '',\n    );\n    return element;\n  }\n\n  return undefined;\n}\n\nexport async function matchElementFromCache(\n  context: {\n    taskCache?: TaskCache;\n    interfaceInstance: AbstractInterface;\n  },\n  cacheEntry: ElementCacheFeature | undefined,\n  cachePrompt: TUserPrompt,\n  cacheable: boolean | undefined,\n): Promise<LocateResultElement | undefined> {\n  if (!cacheEntry) {\n    return undefined;\n  }\n\n  if (cacheable === false) {\n    cacheDebug('cache disabled for prompt: %s', cachePrompt);\n    return undefined;\n  }\n\n  if (!context.taskCache?.isCacheResultUsed) {\n    return undefined;\n  }\n\n  if (!context.interfaceInstance.rectMatchesCacheFeature) {\n    cacheDebug(\n      'interface does not implement rectMatchesCacheFeature, skip cache',\n    );\n    return undefined;\n  }\n\n  try {\n    const rect =\n      await context.interfaceInstance.rectMatchesCacheFeature(cacheEntry);\n    const element: LocateResultElement = {\n      center: [\n        Math.round(rect.left + rect.width / 2),\n        Math.round(rect.top + rect.height / 2),\n      ],\n      rect,\n      description:\n        typeof cachePrompt === 'string'\n          ? cachePrompt\n          : cachePrompt.prompt || '',\n    };\n\n    cacheDebug('cache hit, prompt: %s', cachePrompt);\n    return element;\n  } catch (error) {\n    cacheDebug('rectMatchesCacheFeature error: %s', error);\n    return undefined;\n  }\n}\n\ndeclare const __VERSION__: string | undefined;\n\nexport const getMidsceneVersion = (): string => {\n  if (typeof __VERSION__ !== 'undefined') {\n    return __VERSION__;\n  } else if (\n    process.env.__VERSION__ &&\n    process.env.__VERSION__ !== 'undefined'\n  ) {\n    return process.env.__VERSION__;\n  }\n  throw new Error('__VERSION__ inject failed during build');\n};\n\nexport const parsePrompt = (\n  prompt: TUserPrompt,\n): {\n  textPrompt: string;\n  multimodalPrompt?: TMultimodalPrompt;\n} => {\n  if (typeof prompt === 'string') {\n    return {\n      textPrompt: prompt,\n      multimodalPrompt: undefined,\n    };\n  }\n  return {\n    textPrompt: prompt.prompt,\n    multimodalPrompt: prompt.images\n      ? {\n          images: prompt.images,\n          convertHttpImage2Base64: !!prompt.convertHttpImage2Base64,\n        }\n      : undefined,\n  };\n};\n"],"names":["debugProfile","getDebug","commonContextParser","interfaceInstance","_opt","assert","description","uploadTestInfoToServer","screenshotBase64","size","screenshot","ScreenshotItem","getReportFileName","tag","reportTagName","globalConfigManager","MIDSCENE_REPORT_TAG_NAME","dateTimeInFileName","dayjs","uniqueId","uuid","printReportMsg","filepath","MIDSCENE_REPORT_QUIET","logMsg","getCurrentExecutionFile","trace","error","Error","stackTrace","pkgDir","process","stackLines","line","match","targetFileName","testFileIndex","Map","generateCacheId","fileName","taskFile","console","currentIndex","undefined","ifPlanLocateParamIsBbox","planLocateParam","Array","matchElementFromPlan","rect","element","generateElementByRect","matchElementFromCache","context","cacheEntry","cachePrompt","cacheable","cacheDebug","Math","getMidsceneVersion","__VERSION__","parsePrompt","prompt"],"mappings":";;;;;;;;AAuBA,MAAMA,eAAeC,SAAS;AAEvB,eAAeC,oBACpBC,iBAAoC,EACpCC,IAAkC;IAElCC,OAAOF,mBAAmB;IAE1BH,aAAa;IACb,MAAMM,cAAcH,kBAAkB,QAAQ,QAAQ;IACtDH,aAAa;IAEbA,aAAa;IACbO,uBAAuB;QACrB,SAASD;QACT,WAAWF,KAAK,eAAe;IACjC;IACAJ,aAAa;IAEb,MAAMQ,mBAAmB,MAAML,kBAAkB,gBAAgB;IACjEE,OAAOG,kBAAmB;IAE1BR,aAAa;IACb,MAAMS,OAAO,MAAMN,kBAAkB,IAAI;IACzCH,aAAa,CAAC,MAAM,EAAES,KAAK,KAAK,CAAC,CAAC,EAAEA,KAAK,MAAM,CAAC,MAAM,EAAEA,KAAK,GAAG,EAAE;IAElE,MAAMC,aAAaC,eAAe,MAAM,CAACH;IAEzC,OAAO;QACLC;QACAC;IACF;AACF;AAEO,SAASE,kBAAkBC,MAAM,KAAK;IAC3C,MAAMC,gBAAgBC,oBAAoB,iBAAiB,CACzDC;IAEF,MAAMC,qBAAqBC,QAAQ,MAAM,CAAC;IAE1C,MAAMC,WAAWC,OAAO,SAAS,CAAC,GAAG;IACrC,OAAO,GAAGN,iBAAiBD,IAAI,CAAC,EAAEI,mBAAmB,CAAC,EAAEE,UAAU;AACpE;AAEO,SAASE,eAAeC,QAAgB;IAC7C,IAAIP,oBAAoB,qBAAqB,CAACQ,wBAC5C;IAEFC,OAAO,CAAC,gCAAgC,EAAEF,UAAU;AACtD;AAMO,SAASG,wBAAwBC,KAAc;IACpD,MAAMC,QAAQ,IAAIC;IAClB,MAAMC,aAAaH,SAASC,MAAM,KAAK;IACvC,MAAMG,SAASC,QAAQ,GAAG,MAAM;IAChC,IAAIF,YAAY;QACd,MAAMG,aAAaH,WAAW,KAAK,CAAC;QACpC,KAAK,MAAMI,QAAQD,WACjB,IACEC,KAAK,QAAQ,CAAC,aACdA,KAAK,QAAQ,CAAC,aACdA,KAAK,QAAQ,CAAC,UACdA,KAAK,QAAQ,CAAC,QACd;YACA,MAAMC,QAAQD,KAAK,KAAK,CAAC;YACzB,IAAIC,OAAO,CAAC,EAAE,EAAE;gBACd,MAAMC,iBAAiBD,KAAK,CAAC,EAAE,CAC5B,OAAO,CAACJ,QAAQ,IAChB,IAAI,GACJ,OAAO,CAAC,OAAO;gBAClB,OAAOK;YACT;QACF;IAEJ;IACA,OAAO;AACT;AAEA,MAAMC,gBAAgB,IAAIC;AAEnB,SAASC,gBAAgBC,QAAiB;IAC/C,IAAIC,WAAWD,YAAYd;IAC3B,IAAI,CAACe,UAAU;QACbA,WAAWpB;QACXqB,QAAQ,IAAI,CACV;IAEJ;IAEA,IAAIL,cAAc,GAAG,CAACI,WAAW;QAC/B,MAAME,eAAeN,cAAc,GAAG,CAACI;QACvC,IAAIE,AAAiBC,WAAjBD,cACFN,cAAc,GAAG,CAACI,UAAUE,eAAe;IAE/C,OACEN,cAAc,GAAG,CAACI,UAAU;IAE9B,OAAO,GAAGA,SAAS,CAAC,EAAEJ,cAAc,GAAG,CAACI,WAAW;AACrD;AAEO,SAASI,wBACdC,eAAoC;IAEpC,OAAO,CAAC,CACNA,CAAAA,gBAAgB,IAAI,IACpBC,MAAM,OAAO,CAACD,gBAAgB,IAAI,KAClCA,AAAgC,MAAhCA,gBAAgB,IAAI,CAAC,MAAM,AAAK;AAEpC;AAEO,SAASE,qBACdF,eAAoC;IAEpC,IAAI,CAACA,iBACH;IAGF,IAAIA,gBAAgB,IAAI,EAAE;QAExB,MAAMG,OAAO;YACX,MAAMH,gBAAgB,IAAI,CAAC,EAAE;YAC7B,KAAKA,gBAAgB,IAAI,CAAC,EAAE;YAC5B,OAAOA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAE,GAAG;YAC3D,QAAQA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAE,GAAG;QAC9D;QAEA,MAAMI,UAAUC,sBACdF,MACA,AAAkC,YAAlC,OAAOH,gBAAgB,MAAM,GACzBA,gBAAgB,MAAM,GACtBA,gBAAgB,MAAM,EAAE,UAAU;QAExC,OAAOI;IACT;AAGF;AAEO,eAAeE,sBACpBC,OAGC,EACDC,UAA2C,EAC3CC,WAAwB,EACxBC,SAA8B;IAE9B,IAAI,CAACF,YACH;IAGF,IAAIE,AAAc,UAAdA,WAAqB,YACvBC,MAAW,iCAAiCF;IAI9C,IAAI,CAACF,QAAQ,SAAS,EAAE,mBACtB;IAGF,IAAI,CAACA,QAAQ,iBAAiB,CAAC,uBAAuB,EAAE,YACtDI,MACE;IAKJ,IAAI;QACF,MAAMR,OACJ,MAAMI,QAAQ,iBAAiB,CAAC,uBAAuB,CAACC;QAC1D,MAAMJ,UAA+B;YACnC,QAAQ;gBACNQ,KAAK,KAAK,CAACT,KAAK,IAAI,GAAGA,KAAK,KAAK,GAAG;gBACpCS,KAAK,KAAK,CAACT,KAAK,GAAG,GAAGA,KAAK,MAAM,GAAG;aACrC;YACDA;YACA,aACE,AAAuB,YAAvB,OAAOM,cACHA,cACAA,YAAY,MAAM,IAAI;QAC9B;QAEAE,MAAW,yBAAyBF;QACpC,OAAOL;IACT,EAAE,OAAOtB,OAAO;QACd6B,MAAW,qCAAqC7B;QAChD;IACF;AACF;AAIO,MAAM+B,qBAAqB,IAEvBC;AAUJ,MAAMC,cAAc,CACzBC;IAKA,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAO;QACL,YAAYA;QACZ,kBAAkBlB;IACpB;IAEF,OAAO;QACL,YAAYkB,OAAO,MAAM;QACzB,kBAAkBA,OAAO,MAAM,GAC3B;YACE,QAAQA,OAAO,MAAM;YACrB,yBAAyB,CAAC,CAACA,OAAO,uBAAuB;QAC3D,IACAlB;IACN;AACF"}
+{"version":3,"file":"agent/utils.mjs","sources":["../../../src/agent/utils.ts"],"sourcesContent":["import type { TMultimodalPrompt, TUserPrompt } from '@/common';\nimport type { AbstractInterface } from '@/device';\nimport { ScreenshotItem } from '@/screenshot-item';\nimport type {\n  ElementCacheFeature,\n  LocateResultElement,\n  PlanningLocateParam,\n  Rect,\n  UIContext,\n} from '@/types';\nimport { uploadTestInfoToServer } from '@/utils';\nimport {\n  MIDSCENE_REPORT_QUIET,\n  MIDSCENE_REPORT_TAG_NAME,\n  globalConfigManager,\n} from '@midscene/shared/env';\nimport { generateElementByRect } from '@midscene/shared/extractor';\nimport { imageInfoOfBase64, resizeImgBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { assert, logMsg, uuid } from '@midscene/shared/utils';\nimport dayjs from 'dayjs';\nimport type { TaskCache } from './task-cache';\nimport { debug as cacheDebug } from './task-cache';\n\nexport async function commonContextParser(\n  interfaceInstance: AbstractInterface,\n  _opt: { uploadServerUrl?: string },\n): Promise<UIContext> {\n  const debug = getDebug('commonContextParser');\n\n  assert(interfaceInstance, 'interfaceInstance is required');\n\n  debug('Getting interface description');\n  const description = interfaceInstance.describe?.() || '';\n  debug('Interface description end');\n\n  debug('Uploading test info to server');\n  uploadTestInfoToServer({\n    testUrl: description,\n    serverUrl: _opt.uploadServerUrl,\n  });\n  debug('UploadTestInfoToServer end');\n\n  debug('will get size');\n  const interfaceSize = await interfaceInstance.size();\n  const { width: logicalWidth, height: logicalHeight } = interfaceSize;\n\n  if ((interfaceSize as unknown as { dpr: number }).dpr) {\n    console.warn(\n      'Warning: return value of interface.size() include a dpr property, which is not expected and ignored. ',\n    );\n  }\n\n  if (!Number.isFinite(logicalWidth) || !Number.isFinite(logicalHeight)) {\n    throw new Error(\n      `Invalid interface size: width and height must be finite numbers. Received width: ${logicalWidth}, height: ${logicalHeight}`,\n    );\n  }\n  debug(`size: ${logicalWidth}x${logicalHeight}`);\n\n  const screenshotBase64 = await interfaceInstance.screenshotBase64();\n  assert(screenshotBase64!, 'screenshotBase64 is required');\n\n  // Get physical screenshot dimensions\n  debug('will get screenshot dimensions');\n  const { width: imgWidth, height: imgHeight } =\n    await imageInfoOfBase64(screenshotBase64);\n  debug('screenshot dimensions', imgWidth, 'x', imgHeight);\n\n  const shrinkFactor = imgWidth / logicalWidth;\n\n  debug('calculated shrink factor:', shrinkFactor);\n\n  if (shrinkFactor !== 1) {\n    const targetWidth = Math.round(imgWidth / shrinkFactor);\n    const targetHeight = Math.round(imgHeight / shrinkFactor);\n\n    debug(\n      `Applying screenshot shrink factor: ${shrinkFactor} (physical: ${imgWidth}x${imgHeight} -> target: ${targetWidth}x${targetHeight})`,\n    );\n\n    const resizedBase64 = await resizeImgBase64(screenshotBase64, {\n      width: targetWidth,\n      height: targetHeight,\n    });\n    return {\n      shotSize: {\n        width: targetWidth,\n        height: targetHeight,\n      },\n      deprecatedDpr: shrinkFactor,\n      screenshot: ScreenshotItem.create(resizedBase64),\n    };\n  }\n  return {\n    shotSize: {\n      width: imgWidth,\n      height: imgHeight,\n    },\n    deprecatedDpr: 1,\n    screenshot: ScreenshotItem.create(screenshotBase64),\n  };\n}\n\nexport function getReportFileName(tag = 'web') {\n  const reportTagName = globalConfigManager.getEnvConfigValue(\n    MIDSCENE_REPORT_TAG_NAME,\n  );\n  const dateTimeInFileName = dayjs().format('YYYY-MM-DD_HH-mm-ss');\n  // ensure uniqueness at the same time\n  const uniqueId = uuid().substring(0, 8);\n  return `${reportTagName || tag}-${dateTimeInFileName}-${uniqueId}`;\n}\n\nexport function printReportMsg(filepath: string) {\n  if (globalConfigManager.getEnvConfigInBoolean(MIDSCENE_REPORT_QUIET)) {\n    return;\n  }\n  logMsg(`Midscene - report file updated: ${filepath}`);\n}\n\n/**\n * Get the current execution file name\n * @returns The name of the current execution file\n */\nexport function getCurrentExecutionFile(trace?: string): string | false {\n  const error = new Error();\n  const stackTrace = trace || error.stack;\n  const pkgDir = process.cwd() || '';\n  if (stackTrace) {\n    const stackLines = stackTrace.split('\\n');\n    for (const line of stackLines) {\n      if (\n        line.includes('.spec.') ||\n        line.includes('.test.') ||\n        line.includes('.ts') ||\n        line.includes('.js')\n      ) {\n        const match = line.match(/(?:at\\s+)?(.*?\\.(?:spec|test)\\.[jt]s)/);\n        if (match?.[1]) {\n          const targetFileName = match[1]\n            .replace(pkgDir, '')\n            .trim()\n            .replace('at ', '');\n          return targetFileName;\n        }\n      }\n    }\n  }\n  return false;\n}\n\nconst testFileIndex = new Map<string, number>();\n\nexport function generateCacheId(fileName?: string): string {\n  let taskFile = fileName || getCurrentExecutionFile();\n  if (!taskFile) {\n    taskFile = uuid();\n    console.warn(\n      'Midscene - using random UUID for cache id. Cache may be invalid.',\n    );\n  }\n\n  if (testFileIndex.has(taskFile)) {\n    const currentIndex = testFileIndex.get(taskFile);\n    if (currentIndex !== undefined) {\n      testFileIndex.set(taskFile, currentIndex + 1);\n    }\n  } else {\n    testFileIndex.set(taskFile, 1);\n  }\n  return `${taskFile}-${testFileIndex.get(taskFile)}`;\n}\n\nexport function ifPlanLocateParamIsBbox(\n  planLocateParam: PlanningLocateParam,\n): boolean {\n  return !!(\n    planLocateParam.bbox &&\n    Array.isArray(planLocateParam.bbox) &&\n    planLocateParam.bbox.length === 4\n  );\n}\n\nexport function matchElementFromPlan(\n  planLocateParam: PlanningLocateParam,\n): LocateResultElement | undefined {\n  if (!planLocateParam) {\n    return undefined;\n  }\n\n  if (planLocateParam.bbox) {\n    // Convert bbox [x1, y1, x2, y2] to rect {left, top, width, height}\n    const rect = {\n      left: planLocateParam.bbox[0],\n      top: planLocateParam.bbox[1],\n      width: planLocateParam.bbox[2] - planLocateParam.bbox[0] + 1,\n      height: planLocateParam.bbox[3] - planLocateParam.bbox[1] + 1,\n    };\n\n    const element = generateElementByRect(\n      rect,\n      typeof planLocateParam.prompt === 'string'\n        ? planLocateParam.prompt\n        : planLocateParam.prompt?.prompt || '',\n    );\n    return element;\n  }\n\n  return undefined;\n}\n\nexport async function matchElementFromCache(\n  context: {\n    taskCache?: TaskCache;\n    interfaceInstance: AbstractInterface;\n  },\n  cacheEntry: ElementCacheFeature | undefined,\n  cachePrompt: TUserPrompt,\n  cacheable: boolean | undefined,\n): Promise<LocateResultElement | undefined> {\n  if (!cacheEntry) {\n    return undefined;\n  }\n\n  if (cacheable === false) {\n    cacheDebug('cache disabled for prompt: %s', cachePrompt);\n    return undefined;\n  }\n\n  if (!context.taskCache?.isCacheResultUsed) {\n    return undefined;\n  }\n\n  if (!context.interfaceInstance.rectMatchesCacheFeature) {\n    cacheDebug(\n      'interface does not implement rectMatchesCacheFeature, skip cache',\n    );\n    return undefined;\n  }\n\n  try {\n    const rect =\n      await context.interfaceInstance.rectMatchesCacheFeature(cacheEntry);\n    const element: LocateResultElement = {\n      center: [\n        Math.round(rect.left + rect.width / 2),\n        Math.round(rect.top + rect.height / 2),\n      ],\n      rect,\n      description:\n        typeof cachePrompt === 'string'\n          ? cachePrompt\n          : cachePrompt.prompt || '',\n    };\n\n    cacheDebug('cache hit, prompt: %s', cachePrompt);\n    return element;\n  } catch (error) {\n    cacheDebug('rectMatchesCacheFeature error: %s', error);\n    return undefined;\n  }\n}\n\ndeclare const __VERSION__: string | undefined;\n\nexport const getMidsceneVersion = (): string => {\n  if (typeof __VERSION__ !== 'undefined') {\n    return __VERSION__;\n  } else if (\n    process.env.__VERSION__ &&\n    process.env.__VERSION__ !== 'undefined'\n  ) {\n    return process.env.__VERSION__;\n  }\n  throw new Error('__VERSION__ inject failed during build');\n};\n\nexport const parsePrompt = (\n  prompt: TUserPrompt,\n): {\n  textPrompt: string;\n  multimodalPrompt?: TMultimodalPrompt;\n} => {\n  if (typeof prompt === 'string') {\n    return {\n      textPrompt: prompt,\n      multimodalPrompt: undefined,\n    };\n  }\n  return {\n    textPrompt: prompt.prompt,\n    multimodalPrompt: prompt.images\n      ? {\n          images: prompt.images,\n          convertHttpImage2Base64: !!prompt.convertHttpImage2Base64,\n        }\n      : undefined,\n  };\n};\n\n/**\n * Transform coordinates from screenshot coordinate system to logical coordinate system.\n * When shrunkShotToLogicalRatio > 1, the screenshot is larger than logical size,\n * so we need to divide coordinates by shrunkShotToLogicalRatio.\n *\n * @param element - The locate result element with coordinates in screenshot space\n * @param shrunkShotToLogicalRatio - The ratio of screenshot size to logical size\n * @returns A new element with coordinates transformed to logical space\n */\nexport const transformScreenshotElementToLogical = (\n  element: LocateResultElement,\n  shrunkShotToLogicalRatio: number,\n): LocateResultElement => {\n  if (shrunkShotToLogicalRatio === 1) {\n    return element;\n  }\n\n  return {\n    ...element,\n    center: [\n      Math.round(element.center[0] / shrunkShotToLogicalRatio),\n      Math.round(element.center[1] / shrunkShotToLogicalRatio),\n    ],\n    rect: {\n      ...element.rect,\n      left: Math.round(element.rect.left / shrunkShotToLogicalRatio),\n      top: Math.round(element.rect.top / shrunkShotToLogicalRatio),\n      width: Math.round(element.rect.width / shrunkShotToLogicalRatio),\n      height: Math.round(element.rect.height / shrunkShotToLogicalRatio),\n    },\n  };\n};\n\nexport const transformLogicalElementToScreenshot = (\n  element: LocateResultElement,\n  shrunkShotToLogicalRatio: number,\n): LocateResultElement => {\n  if (shrunkShotToLogicalRatio === 1) {\n    return element;\n  }\n\n  return {\n    ...element,\n    center: [\n      Math.round(element.center[0] * shrunkShotToLogicalRatio),\n      Math.round(element.center[1] * shrunkShotToLogicalRatio),\n    ],\n    rect: {\n      ...element.rect,\n      left: Math.round(element.rect.left * shrunkShotToLogicalRatio),\n      top: Math.round(element.rect.top * shrunkShotToLogicalRatio),\n      width: Math.round(element.rect.width * shrunkShotToLogicalRatio),\n      height: Math.round(element.rect.height * shrunkShotToLogicalRatio),\n    },\n  };\n};\n\nexport const transformLogicalRectToScreenshotRect = (\n  rect: Rect,\n  shrunkShotToLogicalRatio: number,\n): Rect => {\n  if (shrunkShotToLogicalRatio === 1) {\n    return rect;\n  }\n\n  return {\n    ...rect,\n    left: Math.round(rect.left * shrunkShotToLogicalRatio),\n    top: Math.round(rect.top * shrunkShotToLogicalRatio),\n    width: Math.round(rect.width * shrunkShotToLogicalRatio),\n    height: Math.round(rect.height * shrunkShotToLogicalRatio),\n  };\n};\n"],"names":["commonContextParser","interfaceInstance","_opt","debug","getDebug","assert","description","uploadTestInfoToServer","interfaceSize","logicalWidth","logicalHeight","console","Number","Error","screenshotBase64","imgWidth","imgHeight","imageInfoOfBase64","shrinkFactor","targetWidth","Math","targetHeight","resizedBase64","resizeImgBase64","ScreenshotItem","getReportFileName","tag","reportTagName","globalConfigManager","MIDSCENE_REPORT_TAG_NAME","dateTimeInFileName","dayjs","uniqueId","uuid","printReportMsg","filepath","MIDSCENE_REPORT_QUIET","logMsg","getCurrentExecutionFile","trace","error","stackTrace","pkgDir","process","stackLines","line","match","targetFileName","testFileIndex","Map","generateCacheId","fileName","taskFile","currentIndex","undefined","ifPlanLocateParamIsBbox","planLocateParam","Array","matchElementFromPlan","rect","element","generateElementByRect","matchElementFromCache","context","cacheEntry","cachePrompt","cacheable","cacheDebug","getMidsceneVersion","__VERSION__","parsePrompt","prompt","transformScreenshotElementToLogical","shrunkShotToLogicalRatio","transformLogicalElementToScreenshot","transformLogicalRectToScreenshotRect"],"mappings":";;;;;;;;;AAyBO,eAAeA,oBACpBC,iBAAoC,EACpCC,IAAkC;IAElC,MAAMC,QAAQC,SAAS;IAEvBC,OAAOJ,mBAAmB;IAE1BE,MAAM;IACN,MAAMG,cAAcL,kBAAkB,QAAQ,QAAQ;IACtDE,MAAM;IAENA,MAAM;IACNI,uBAAuB;QACrB,SAASD;QACT,WAAWJ,KAAK,eAAe;IACjC;IACAC,MAAM;IAENA,MAAM;IACN,MAAMK,gBAAgB,MAAMP,kBAAkB,IAAI;IAClD,MAAM,EAAE,OAAOQ,YAAY,EAAE,QAAQC,aAAa,EAAE,GAAGF;IAEvD,IAAKA,cAA6C,GAAG,EACnDG,QAAQ,IAAI,CACV;IAIJ,IAAI,CAACC,OAAO,QAAQ,CAACH,iBAAiB,CAACG,OAAO,QAAQ,CAACF,gBACrD,MAAM,IAAIG,MACR,CAAC,iFAAiF,EAAEJ,aAAa,UAAU,EAAEC,eAAe;IAGhIP,MAAM,CAAC,MAAM,EAAEM,aAAa,CAAC,EAAEC,eAAe;IAE9C,MAAMI,mBAAmB,MAAMb,kBAAkB,gBAAgB;IACjEI,OAAOS,kBAAmB;IAG1BX,MAAM;IACN,MAAM,EAAE,OAAOY,QAAQ,EAAE,QAAQC,SAAS,EAAE,GAC1C,MAAMC,kBAAkBH;IAC1BX,MAAM,yBAAyBY,UAAU,KAAKC;IAE9C,MAAME,eAAeH,WAAWN;IAEhCN,MAAM,6BAA6Be;IAEnC,IAAIA,AAAiB,MAAjBA,cAAoB;QACtB,MAAMC,cAAcC,KAAK,KAAK,CAACL,WAAWG;QAC1C,MAAMG,eAAeD,KAAK,KAAK,CAACJ,YAAYE;QAE5Cf,MACE,CAAC,mCAAmC,EAAEe,aAAa,YAAY,EAAEH,SAAS,CAAC,EAAEC,UAAU,YAAY,EAAEG,YAAY,CAAC,EAAEE,aAAa,CAAC,CAAC;QAGrI,MAAMC,gBAAgB,MAAMC,gBAAgBT,kBAAkB;YAC5D,OAAOK;YACP,QAAQE;QACV;QACA,OAAO;YACL,UAAU;gBACR,OAAOF;gBACP,QAAQE;YACV;YACA,eAAeH;YACf,YAAYM,eAAe,MAAM,CAACF;QACpC;IACF;IACA,OAAO;QACL,UAAU;YACR,OAAOP;YACP,QAAQC;QACV;QACA,eAAe;QACf,YAAYQ,eAAe,MAAM,CAACV;IACpC;AACF;AAEO,SAASW,kBAAkBC,MAAM,KAAK;IAC3C,MAAMC,gBAAgBC,oBAAoB,iBAAiB,CACzDC;IAEF,MAAMC,qBAAqBC,QAAQ,MAAM,CAAC;IAE1C,MAAMC,WAAWC,OAAO,SAAS,CAAC,GAAG;IACrC,OAAO,GAAGN,iBAAiBD,IAAI,CAAC,EAAEI,mBAAmB,CAAC,EAAEE,UAAU;AACpE;AAEO,SAASE,eAAeC,QAAgB;IAC7C,IAAIP,oBAAoB,qBAAqB,CAACQ,wBAC5C;IAEFC,OAAO,CAAC,gCAAgC,EAAEF,UAAU;AACtD;AAMO,SAASG,wBAAwBC,KAAc;IACpD,MAAMC,QAAQ,IAAI3B;IAClB,MAAM4B,aAAaF,SAASC,MAAM,KAAK;IACvC,MAAME,SAASC,QAAQ,GAAG,MAAM;IAChC,IAAIF,YAAY;QACd,MAAMG,aAAaH,WAAW,KAAK,CAAC;QACpC,KAAK,MAAMI,QAAQD,WACjB,IACEC,KAAK,QAAQ,CAAC,aACdA,KAAK,QAAQ,CAAC,aACdA,KAAK,QAAQ,CAAC,UACdA,KAAK,QAAQ,CAAC,QACd;YACA,MAAMC,QAAQD,KAAK,KAAK,CAAC;YACzB,IAAIC,OAAO,CAAC,EAAE,EAAE;gBACd,MAAMC,iBAAiBD,KAAK,CAAC,EAAE,CAC5B,OAAO,CAACJ,QAAQ,IAChB,IAAI,GACJ,OAAO,CAAC,OAAO;gBAClB,OAAOK;YACT;QACF;IAEJ;IACA,OAAO;AACT;AAEA,MAAMC,gBAAgB,IAAIC;AAEnB,SAASC,gBAAgBC,QAAiB;IAC/C,IAAIC,WAAWD,YAAYb;IAC3B,IAAI,CAACc,UAAU;QACbA,WAAWnB;QACXtB,QAAQ,IAAI,CACV;IAEJ;IAEA,IAAIqC,cAAc,GAAG,CAACI,WAAW;QAC/B,MAAMC,eAAeL,cAAc,GAAG,CAACI;QACvC,IAAIC,AAAiBC,WAAjBD,cACFL,cAAc,GAAG,CAACI,UAAUC,eAAe;IAE/C,OACEL,cAAc,GAAG,CAACI,UAAU;IAE9B,OAAO,GAAGA,SAAS,CAAC,EAAEJ,cAAc,GAAG,CAACI,WAAW;AACrD;AAEO,SAASG,wBACdC,eAAoC;IAEpC,OAAO,CAAC,CACNA,CAAAA,gBAAgB,IAAI,IACpBC,MAAM,OAAO,CAACD,gBAAgB,IAAI,KAClCA,AAAgC,MAAhCA,gBAAgB,IAAI,CAAC,MAAM,AAAK;AAEpC;AAEO,SAASE,qBACdF,eAAoC;IAEpC,IAAI,CAACA,iBACH;IAGF,IAAIA,gBAAgB,IAAI,EAAE;QAExB,MAAMG,OAAO;YACX,MAAMH,gBAAgB,IAAI,CAAC,EAAE;YAC7B,KAAKA,gBAAgB,IAAI,CAAC,EAAE;YAC5B,OAAOA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAE,GAAG;YAC3D,QAAQA,gBAAgB,IAAI,CAAC,EAAE,GAAGA,gBAAgB,IAAI,CAAC,EAAE,GAAG;QAC9D;QAEA,MAAMI,UAAUC,sBACdF,MACA,AAAkC,YAAlC,OAAOH,gBAAgB,MAAM,GACzBA,gBAAgB,MAAM,GACtBA,gBAAgB,MAAM,EAAE,UAAU;QAExC,OAAOI;IACT;AAGF;AAEO,eAAeE,sBACpBC,OAGC,EACDC,UAA2C,EAC3CC,WAAwB,EACxBC,SAA8B;IAE9B,IAAI,CAACF,YACH;IAGF,IAAIE,AAAc,UAAdA,WAAqB,YACvBC,8BAAW,iCAAiCF;IAI9C,IAAI,CAACF,QAAQ,SAAS,EAAE,mBACtB;IAGF,IAAI,CAACA,QAAQ,iBAAiB,CAAC,uBAAuB,EAAE,YACtDI,8BACE;IAKJ,IAAI;QACF,MAAMR,OACJ,MAAMI,QAAQ,iBAAiB,CAAC,uBAAuB,CAACC;QAC1D,MAAMJ,UAA+B;YACnC,QAAQ;gBACNxC,KAAK,KAAK,CAACuC,KAAK,IAAI,GAAGA,KAAK,KAAK,GAAG;gBACpCvC,KAAK,KAAK,CAACuC,KAAK,GAAG,GAAGA,KAAK,MAAM,GAAG;aACrC;YACDA;YACA,aACE,AAAuB,YAAvB,OAAOM,cACHA,cACAA,YAAY,MAAM,IAAI;QAC9B;QAEAE,8BAAW,yBAAyBF;QACpC,OAAOL;IACT,EAAE,OAAOpB,OAAO;QACd2B,8BAAW,qCAAqC3B;QAChD;IACF;AACF;AAIO,MAAM4B,qBAAqB,IAEvBC;AAUJ,MAAMC,cAAc,CACzBC;IAKA,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAO;QACL,YAAYA;QACZ,kBAAkBjB;IACpB;IAEF,OAAO;QACL,YAAYiB,OAAO,MAAM;QACzB,kBAAkBA,OAAO,MAAM,GAC3B;YACE,QAAQA,OAAO,MAAM;YACrB,yBAAyB,CAAC,CAACA,OAAO,uBAAuB;QAC3D,IACAjB;IACN;AACF;AAWO,MAAMkB,sCAAsC,CACjDZ,SACAa;IAEA,IAAIA,AAA6B,MAA7BA,0BACF,OAAOb;IAGT,OAAO;QACL,GAAGA,OAAO;QACV,QAAQ;YACNxC,KAAK,KAAK,CAACwC,QAAQ,MAAM,CAAC,EAAE,GAAGa;YAC/BrD,KAAK,KAAK,CAACwC,QAAQ,MAAM,CAAC,EAAE,GAAGa;SAChC;QACD,MAAM;YACJ,GAAGb,QAAQ,IAAI;YACf,MAAMxC,KAAK,KAAK,CAACwC,QAAQ,IAAI,CAAC,IAAI,GAAGa;YACrC,KAAKrD,KAAK,KAAK,CAACwC,QAAQ,IAAI,CAAC,GAAG,GAAGa;YACnC,OAAOrD,KAAK,KAAK,CAACwC,QAAQ,IAAI,CAAC,KAAK,GAAGa;YACvC,QAAQrD,KAAK,KAAK,CAACwC,QAAQ,IAAI,CAAC,MAAM,GAAGa;QAC3C;IACF;AACF;AAEO,MAAMC,sCAAsC,CACjDd,SACAa;IAEA,IAAIA,AAA6B,MAA7BA,0BACF,OAAOb;IAGT,OAAO;QACL,GAAGA,OAAO;QACV,QAAQ;YACNxC,KAAK,KAAK,CAACwC,QAAQ,MAAM,CAAC,EAAE,GAAGa;YAC/BrD,KAAK,KAAK,CAACwC,QAAQ,MAAM,CAAC,EAAE,GAAGa;SAChC;QACD,MAAM;YACJ,GAAGb,QAAQ,IAAI;YACf,MAAMxC,KAAK,KAAK,CAACwC,QAAQ,IAAI,CAAC,IAAI,GAAGa;YACrC,KAAKrD,KAAK,KAAK,CAACwC,QAAQ,IAAI,CAAC,GAAG,GAAGa;YACnC,OAAOrD,KAAK,KAAK,CAACwC,QAAQ,IAAI,CAAC,KAAK,GAAGa;YACvC,QAAQrD,KAAK,KAAK,CAACwC,QAAQ,IAAI,CAAC,MAAM,GAAGa;QAC3C;IACF;AACF;AAEO,MAAME,uCAAuC,CAClDhB,MACAc;IAEA,IAAIA,AAA6B,MAA7BA,0BACF,OAAOd;IAGT,OAAO;QACL,GAAGA,IAAI;QACP,MAAMvC,KAAK,KAAK,CAACuC,KAAK,IAAI,GAAGc;QAC7B,KAAKrD,KAAK,KAAK,CAACuC,KAAK,GAAG,GAAGc;QAC3B,OAAOrD,KAAK,KAAK,CAACuC,KAAK,KAAK,GAAGc;QAC/B,QAAQrD,KAAK,KAAK,CAACuC,KAAK,MAAM,GAAGc;IACnC;AACF"}

package/dist/es/ai-model/auto-glm/planning.mjs

@@ -45,7 +45,7 @@ async function autoGLMPlanning(userInstruction, options) {
         debug('action in response:', parsedResponse.content);
         const parsedAction = parseAction(parsedResponse);
         debug('Parsed action object:', parsedAction);
-        transformedActions = transformAutoGLMAction(parsedAction, context.size);
+        transformedActions = transformAutoGLMAction(parsedAction, context.shotSize);
         debug('Transformed actions:', transformedActions);
     } catch (parseError) {
         const errorMessage = parseError instanceof Error ? parseError.message : String(parseError);

package/dist/es/ai-model/auto-glm/planning.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/auto-glm/planning.mjs","sources":["../../../../src/ai-model/auto-glm/planning.ts"],"sourcesContent":["import type { PlanningAIResponse, UIContext } from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { ConversationHistory } from '../conversation-history';\nimport {\n  AIResponseParseError,\n  callAIWithStringResponse,\n} from '../service-caller/index';\nimport { transformAutoGLMAction } from './actions';\nimport { parseAction, parseAutoGLMResponse } from './parser';\nimport { getAutoGLMPlanPrompt } from './prompt';\n\nconst debug = getDebug('auto-glm-planning');\n\nexport async function autoGLMPlanning(\n  userInstruction: string,\n  options: {\n    conversationHistory: ConversationHistory;\n    context: UIContext;\n    modelConfig: IModelConfig;\n    actionContext?: string;\n  },\n): Promise<PlanningAIResponse> {\n  const { conversationHistory, context, modelConfig, actionContext } = options;\n\n  const systemPrompt =\n    getAutoGLMPlanPrompt(modelConfig.modelFamily) +\n    (actionContext\n      ? `<high_priority_knowledge>${actionContext}</high_priority_knowledge>`\n      : '');\n\n  const imagePayloadBase64 = context.screenshot.base64;\n\n  conversationHistory.append({\n    role: 'user',\n    content: [{ type: 'text', text: userInstruction }],\n  });\n  conversationHistory.append({\n    role: 'user',\n    content: [{ type: 'image_url', image_url: { url: imagePayloadBase64 } }],\n  });\n\n  const msgs: ChatCompletionMessageParam[] = [\n    { role: 'system', content: systemPrompt },\n    ...conversationHistory.snapshot(1),\n  ];\n\n  const { content: rawResponse, usage } = await callAIWithStringResponse(\n    msgs,\n    modelConfig,\n  );\n\n  debug('autoGLMPlanning rawResponse:', rawResponse);\n\n  let parsedResponse: ReturnType<typeof parseAutoGLMResponse>;\n  let transformedActions: ReturnType<typeof transformAutoGLMAction>;\n\n  try {\n    parsedResponse = parseAutoGLMResponse(rawResponse);\n    debug('thinking in response:', parsedResponse.think);\n    debug('action in response:', parsedResponse.content);\n\n    const parsedAction = parseAction(parsedResponse);\n    debug('Parsed action object:', parsedAction);\n    transformedActions = transformAutoGLMAction(parsedAction, context.size);\n    debug('Transformed actions:', transformedActions);\n  } catch (parseError) {\n    // Throw AIResponseParseError with usage and rawResponse preserved\n    const errorMessage =\n      parseError instanceof Error ? parseError.message : String(parseError);\n    throw new AIResponseParseError(\n      `Parse error: ${errorMessage}`,\n      JSON.stringify(rawResponse, undefined, 2),\n      usage,\n    );\n  }\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: `<think>${parsedResponse.think}</think><answer>${parsedResponse.content}</answer>`,\n  });\n\n  const shouldContinuePlanning = !parsedResponse.content.startsWith('finish(');\n\n  return {\n    actions: transformedActions,\n    log: rawResponse,\n    usage,\n    shouldContinuePlanning,\n    rawResponse: JSON.stringify(rawResponse, undefined, 2),\n  };\n}\n"],"names":["debug","getDebug","autoGLMPlanning","userInstruction","options","conversationHistory","context","modelConfig","actionContext","systemPrompt","getAutoGLMPlanPrompt","imagePayloadBase64","msgs","rawResponse","usage","callAIWithStringResponse","parsedResponse","transformedActions","parseAutoGLMResponse","parsedAction","parseAction","transformAutoGLMAction","parseError","errorMessage","Error","String","AIResponseParseError","JSON","undefined","shouldContinuePlanning"],"mappings":";;;;;AAaA,MAAMA,QAAQC,SAAS;AAEhB,eAAeC,gBACpBC,eAAuB,EACvBC,OAKC;IAED,MAAM,EAAEC,mBAAmB,EAAEC,OAAO,EAAEC,WAAW,EAAEC,aAAa,EAAE,GAAGJ;IAErE,MAAMK,eACJC,qBAAqBH,YAAY,WAAW,IAC3CC,CAAAA,gBACG,CAAC,yBAAyB,EAAEA,cAAc,0BAA0B,CAAC,GACrE,EAAC;IAEP,MAAMG,qBAAqBL,QAAQ,UAAU,CAAC,MAAM;IAEpDD,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YAAC;gBAAE,MAAM;gBAAQ,MAAMF;YAAgB;SAAE;IACpD;IACAE,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YAAC;gBAAE,MAAM;gBAAa,WAAW;oBAAE,KAAKM;gBAAmB;YAAE;SAAE;IAC1E;IAEA,MAAMC,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASH;QAAa;WACrCJ,oBAAoB,QAAQ,CAAC;KACjC;IAED,MAAM,EAAE,SAASQ,WAAW,EAAEC,KAAK,EAAE,GAAG,MAAMC,yBAC5CH,MACAL;IAGFP,MAAM,gCAAgCa;IAEtC,IAAIG;IACJ,IAAIC;IAEJ,IAAI;QACFD,iBAAiBE,qBAAqBL;QACtCb,MAAM,yBAAyBgB,eAAe,KAAK;QACnDhB,MAAM,uBAAuBgB,eAAe,OAAO;QAEnD,MAAMG,eAAeC,YAAYJ;QACjChB,MAAM,yBAAyBmB;QAC/BF,qBAAqBI,uBAAuBF,cAAcb,QAAQ,IAAI;QACtEN,MAAM,wBAAwBiB;IAChC,EAAE,OAAOK,YAAY;QAEnB,MAAMC,eACJD,sBAAsBE,QAAQF,WAAW,OAAO,GAAGG,OAAOH;QAC5D,MAAM,IAAII,qBACR,CAAC,aAAa,EAAEH,cAAc,EAC9BI,KAAK,SAAS,CAACd,aAAae,QAAW,IACvCd;IAEJ;IAEAT,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS,CAAC,OAAO,EAAEW,eAAe,KAAK,CAAC,gBAAgB,EAAEA,eAAe,OAAO,CAAC,SAAS,CAAC;IAC7F;IAEA,MAAMa,yBAAyB,CAACb,eAAe,OAAO,CAAC,UAAU,CAAC;IAElE,OAAO;QACL,SAASC;QACT,KAAKJ;QACLC;QACAe;QACA,aAAaF,KAAK,SAAS,CAACd,aAAae,QAAW;IACtD;AACF"}
+{"version":3,"file":"ai-model/auto-glm/planning.mjs","sources":["../../../../src/ai-model/auto-glm/planning.ts"],"sourcesContent":["import type { PlanningAIResponse, UIContext } from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { ConversationHistory } from '../conversation-history';\nimport {\n  AIResponseParseError,\n  callAIWithStringResponse,\n} from '../service-caller/index';\nimport { transformAutoGLMAction } from './actions';\nimport { parseAction, parseAutoGLMResponse } from './parser';\nimport { getAutoGLMPlanPrompt } from './prompt';\n\nconst debug = getDebug('auto-glm-planning');\n\nexport async function autoGLMPlanning(\n  userInstruction: string,\n  options: {\n    conversationHistory: ConversationHistory;\n    context: UIContext;\n    modelConfig: IModelConfig;\n    actionContext?: string;\n  },\n): Promise<PlanningAIResponse> {\n  const { conversationHistory, context, modelConfig, actionContext } = options;\n\n  const systemPrompt =\n    getAutoGLMPlanPrompt(modelConfig.modelFamily) +\n    (actionContext\n      ? `<high_priority_knowledge>${actionContext}</high_priority_knowledge>`\n      : '');\n\n  const imagePayloadBase64 = context.screenshot.base64;\n\n  conversationHistory.append({\n    role: 'user',\n    content: [{ type: 'text', text: userInstruction }],\n  });\n  conversationHistory.append({\n    role: 'user',\n    content: [{ type: 'image_url', image_url: { url: imagePayloadBase64 } }],\n  });\n\n  const msgs: ChatCompletionMessageParam[] = [\n    { role: 'system', content: systemPrompt },\n    ...conversationHistory.snapshot(1),\n  ];\n\n  const { content: rawResponse, usage } = await callAIWithStringResponse(\n    msgs,\n    modelConfig,\n  );\n\n  debug('autoGLMPlanning rawResponse:', rawResponse);\n\n  let parsedResponse: ReturnType<typeof parseAutoGLMResponse>;\n  let transformedActions: ReturnType<typeof transformAutoGLMAction>;\n\n  try {\n    parsedResponse = parseAutoGLMResponse(rawResponse);\n    debug('thinking in response:', parsedResponse.think);\n    debug('action in response:', parsedResponse.content);\n\n    const parsedAction = parseAction(parsedResponse);\n    debug('Parsed action object:', parsedAction);\n    transformedActions = transformAutoGLMAction(parsedAction, context.shotSize);\n    debug('Transformed actions:', transformedActions);\n  } catch (parseError) {\n    // Throw AIResponseParseError with usage and rawResponse preserved\n    const errorMessage =\n      parseError instanceof Error ? parseError.message : String(parseError);\n    throw new AIResponseParseError(\n      `Parse error: ${errorMessage}`,\n      JSON.stringify(rawResponse, undefined, 2),\n      usage,\n    );\n  }\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: `<think>${parsedResponse.think}</think><answer>${parsedResponse.content}</answer>`,\n  });\n\n  const shouldContinuePlanning = !parsedResponse.content.startsWith('finish(');\n\n  return {\n    actions: transformedActions,\n    log: rawResponse,\n    usage,\n    shouldContinuePlanning,\n    rawResponse: JSON.stringify(rawResponse, undefined, 2),\n  };\n}\n"],"names":["debug","getDebug","autoGLMPlanning","userInstruction","options","conversationHistory","context","modelConfig","actionContext","systemPrompt","getAutoGLMPlanPrompt","imagePayloadBase64","msgs","rawResponse","usage","callAIWithStringResponse","parsedResponse","transformedActions","parseAutoGLMResponse","parsedAction","parseAction","transformAutoGLMAction","parseError","errorMessage","Error","String","AIResponseParseError","JSON","undefined","shouldContinuePlanning"],"mappings":";;;;;AAaA,MAAMA,QAAQC,SAAS;AAEhB,eAAeC,gBACpBC,eAAuB,EACvBC,OAKC;IAED,MAAM,EAAEC,mBAAmB,EAAEC,OAAO,EAAEC,WAAW,EAAEC,aAAa,EAAE,GAAGJ;IAErE,MAAMK,eACJC,qBAAqBH,YAAY,WAAW,IAC3CC,CAAAA,gBACG,CAAC,yBAAyB,EAAEA,cAAc,0BAA0B,CAAC,GACrE,EAAC;IAEP,MAAMG,qBAAqBL,QAAQ,UAAU,CAAC,MAAM;IAEpDD,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YAAC;gBAAE,MAAM;gBAAQ,MAAMF;YAAgB;SAAE;IACpD;IACAE,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YAAC;gBAAE,MAAM;gBAAa,WAAW;oBAAE,KAAKM;gBAAmB;YAAE;SAAE;IAC1E;IAEA,MAAMC,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASH;QAAa;WACrCJ,oBAAoB,QAAQ,CAAC;KACjC;IAED,MAAM,EAAE,SAASQ,WAAW,EAAEC,KAAK,EAAE,GAAG,MAAMC,yBAC5CH,MACAL;IAGFP,MAAM,gCAAgCa;IAEtC,IAAIG;IACJ,IAAIC;IAEJ,IAAI;QACFD,iBAAiBE,qBAAqBL;QACtCb,MAAM,yBAAyBgB,eAAe,KAAK;QACnDhB,MAAM,uBAAuBgB,eAAe,OAAO;QAEnD,MAAMG,eAAeC,YAAYJ;QACjChB,MAAM,yBAAyBmB;QAC/BF,qBAAqBI,uBAAuBF,cAAcb,QAAQ,QAAQ;QAC1EN,MAAM,wBAAwBiB;IAChC,EAAE,OAAOK,YAAY;QAEnB,MAAMC,eACJD,sBAAsBE,QAAQF,WAAW,OAAO,GAAGG,OAAOH;QAC5D,MAAM,IAAII,qBACR,CAAC,aAAa,EAAEH,cAAc,EAC9BI,KAAK,SAAS,CAACd,aAAae,QAAW,IACvCd;IAEJ;IAEAT,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS,CAAC,OAAO,EAAEW,eAAe,KAAK,CAAC,gBAAgB,EAAEA,eAAe,OAAO,CAAC,SAAS,CAAC;IAC7F;IAEA,MAAMa,yBAAyB,CAACb,eAAe,OAAO,CAAC,UAAU,CAAC;IAElE,OAAO;QACL,SAASC;QACT,KAAKJ;QACLC;QACAe;QACA,aAAaF,KAAK,SAAS,CAACd,aAAae,QAAW;IACtD;AACF"}

package/dist/es/ai-model/inspect.mjs

@@ -65,8 +65,8 @@ async function AiLocateElement(options) {
     const userInstructionPrompt = findElementPrompt(targetElementDescriptionText);
     const systemPrompt = isAutoGLM(modelFamily) ? getAutoGLMLocatePrompt(modelFamily) : systemPromptToLocateElement(modelFamily);
     let imagePayload = screenshotBase64;
-    let imageWidth = context.size.width;
-    let imageHeight = context.size.height;
+    let imageWidth = context.shotSize.width;
+    let imageHeight = context.shotSize.height;
     let originalImageWidth = imageWidth;
     let originalImageHeight = imageHeight;
     if (options.searchConfig) {
@@ -270,18 +270,18 @@ async function AiLocateSection(options) {
     let sectionRect;
     const sectionBbox = result.content.bbox;
     if (sectionBbox) {
-        const targetRect = adaptBboxToRect(sectionBbox, context.size.width, context.size.height, 0, 0, context.size.width, context.size.height, modelFamily);
+        const targetRect = adaptBboxToRect(sectionBbox, context.shotSize.width, context.shotSize.height, 0, 0, context.shotSize.width, context.shotSize.height, modelFamily);
         debugSection('original targetRect %j', targetRect);
         const referenceBboxList = result.content.references_bbox || [];
         debugSection('referenceBboxList %j', referenceBboxList);
-        const referenceRects = referenceBboxList.filter((bbox)=>Array.isArray(bbox)).map((bbox)=>adaptBboxToRect(bbox, context.size.width, context.size.height, 0, 0, context.size.width, context.size.height, modelFamily));
+        const referenceRects = referenceBboxList.filter((bbox)=>Array.isArray(bbox)).map((bbox)=>adaptBboxToRect(bbox, context.shotSize.width, context.shotSize.height, 0, 0, context.shotSize.width, context.shotSize.height, modelFamily));
         debugSection('referenceRects %j', referenceRects);
         const mergedRect = mergeRects([
             targetRect,
             ...referenceRects
         ]);
         debugSection('mergedRect %j', mergedRect);
-        sectionRect = expandSearchArea(mergedRect, context.size);
+        sectionRect = expandSearchArea(mergedRect, context.shotSize);
         debugSection('expanded sectionRect %j', sectionRect);
     }
     let imageBase64 = screenshotBase64;

package/dist/es/ai-model/inspect.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/inspect.mjs","sources":["../../../src/ai-model/inspect.ts"],"sourcesContent":["import type {\n  AIDataExtractionResponse,\n  AIElementResponse,\n  AISectionLocatorResponse,\n  AIUsageInfo,\n  Rect,\n  ServiceExtractOption,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport {\n  generateElementByPoint,\n  generateElementByRect,\n} from '@midscene/shared/extractor/dom-util';\nimport {\n  cropByRect,\n  paddingToMatchBlockByBase64,\n  preProcessImageUrl,\n  scaleImage,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { LocateResultElement } from '@midscene/shared/types';\nimport { assert } from '@midscene/shared/utils';\nimport type {\n  ChatCompletionSystemMessageParam,\n  ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport type { TMultimodalPrompt, TUserPrompt } from '../common';\nimport { adaptBboxToRect, expandSearchArea, mergeRects } from '../common';\nimport { parseAutoGLMLocateResponse } from './auto-glm/parser';\nimport { getAutoGLMLocatePrompt } from './auto-glm/prompt';\nimport { isAutoGLM } from './auto-glm/util';\nimport {\n  extractDataQueryPrompt,\n  parseXMLExtractionResponse,\n  systemPromptToExtract,\n} from './prompt/extraction';\nimport {\n  findElementPrompt,\n  systemPromptToLocateElement,\n} from './prompt/llm-locator';\nimport {\n  sectionLocatorInstruction,\n  systemPromptToLocateSection,\n} from './prompt/llm-section-locator';\nimport {\n  orderSensitiveJudgePrompt,\n  systemPromptToJudgeOrderSensitive,\n} from './prompt/order-sensitive-judge';\nimport {\n  AIResponseParseError,\n  callAI,\n  callAIWithObjectResponse,\n  callAIWithStringResponse,\n} from './service-caller/index';\n\nexport type AIArgs = [\n  ChatCompletionSystemMessageParam,\n  ...ChatCompletionUserMessageParam[],\n];\n\nconst debugInspect = getDebug('ai:inspect');\nconst debugSection = getDebug('ai:section');\n\nconst extraTextFromUserPrompt = (prompt: TUserPrompt): string => {\n  if (typeof prompt === 'string') {\n    return prompt;\n  } else {\n    return prompt.prompt;\n  }\n};\n\nconst promptsToChatParam = async (\n  multimodalPrompt: TMultimodalPrompt,\n): Promise<ChatCompletionUserMessageParam[]> => {\n  const msgs: ChatCompletionUserMessageParam[] = [];\n  if (multimodalPrompt?.images?.length) {\n    msgs.push({\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: 'Next, I will provide all the reference images.',\n        },\n      ],\n    });\n\n    for (const item of multimodalPrompt.images) {\n      const base64 = await preProcessImageUrl(\n        item.url,\n        !!multimodalPrompt.convertHttpImage2Base64,\n      );\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'text',\n            text: `this is the reference image named '${item.name}':`,\n          },\n        ],\n      });\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: base64,\n              detail: 'high',\n            },\n          },\n        ],\n      });\n    }\n  }\n  return msgs;\n};\n\nexport async function AiLocateElement(options: {\n  context: UIContext;\n  targetElementDescription: TUserPrompt;\n  searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n  modelConfig: IModelConfig;\n}): Promise<{\n  parseResult: {\n    elements: LocateResultElement[];\n    errors?: string[];\n  };\n  rect?: Rect;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}> {\n  const { context, targetElementDescription, modelConfig } = options;\n  const { modelFamily } = modelConfig;\n  const screenshotBase64 = context.screenshot.base64;\n\n  assert(\n    targetElementDescription,\n    'cannot find the target element description',\n  );\n  const targetElementDescriptionText = extraTextFromUserPrompt(\n    targetElementDescription,\n  );\n  const userInstructionPrompt = findElementPrompt(targetElementDescriptionText);\n  const systemPrompt = isAutoGLM(modelFamily)\n    ? getAutoGLMLocatePrompt(modelFamily)\n    : systemPromptToLocateElement(modelFamily);\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = context.size.width;\n  let imageHeight = context.size.height;\n  let originalImageWidth = imageWidth;\n  let originalImageHeight = imageHeight;\n\n  if (options.searchConfig) {\n    assert(\n      options.searchConfig.rect,\n      'searchArea is provided but its rect cannot be found. Failed to locate element',\n    );\n    assert(\n      options.searchConfig.imageBase64,\n      'searchArea is provided but its imageBase64 cannot be found. Failed to locate element',\n    );\n\n    imagePayload = options.searchConfig.imageBase64;\n    imageWidth = options.searchConfig.rect?.width;\n    imageHeight = options.searchConfig.rect?.height;\n    originalImageWidth = imageWidth;\n    originalImageHeight = imageHeight;\n  } else if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\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          type: 'text',\n          text: isAutoGLM(modelFamily)\n            ? `Tap: ${userInstructionPrompt}`\n            : userInstructionPrompt,\n        },\n      ],\n    },\n  ];\n\n  if (typeof targetElementDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: targetElementDescription.images,\n      convertHttpImage2Base64: targetElementDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  if (isAutoGLM(modelFamily)) {\n    const { content: rawResponseContent, usage } =\n      await callAIWithStringResponse(msgs, modelConfig);\n\n    debugInspect('auto-glm rawResponse:', rawResponseContent);\n\n    const parsed = parseAutoGLMLocateResponse(rawResponseContent);\n\n    debugInspect('auto-glm thinking:', parsed.think);\n    debugInspect('auto-glm coordinates:', parsed.coordinates);\n\n    let resRect: Rect | undefined;\n    let matchedElements: LocateResultElement[] = [];\n    let errors: string[] = [];\n\n    if (parsed.error || !parsed.coordinates) {\n      errors = [parsed.error || 'Failed to parse auto-glm response'];\n      debugInspect('auto-glm parse error:', errors[0]);\n    } else {\n      const { x, y } = parsed.coordinates;\n\n      debugInspect('auto-glm coordinates [0-999]:', { x, y });\n\n      // Convert auto-glm coordinates [0,999] to pixel bbox\n      // Map from [0,999] to pixel coordinates\n      const pixelX = Math.round((x * imageWidth) / 1000);\n      const pixelY = Math.round((y * imageHeight) / 1000);\n\n      debugInspect('auto-glm pixel coordinates:', { pixelX, pixelY });\n\n      // Apply offset if searching in a cropped area\n      let finalX = pixelX;\n      let finalY = pixelY;\n      if (options.searchConfig?.rect) {\n        finalX += options.searchConfig.rect.left;\n        finalY += options.searchConfig.rect.top;\n      }\n\n      const element: LocateResultElement = generateElementByPoint(\n        [finalX, finalY],\n        targetElementDescriptionText as string,\n      );\n\n      resRect = element.rect;\n      debugInspect('auto-glm resRect:', resRect);\n\n      if (element) {\n        matchedElements = [element];\n      }\n    }\n\n    return {\n      rect: resRect,\n      parseResult: {\n        elements: matchedElements,\n        errors,\n      },\n      rawResponse: rawResponseContent,\n      usage,\n      reasoning_content: parsed.think,\n    };\n  }\n\n  let res: Awaited<\n    ReturnType<\n      typeof callAIWithObjectResponse<AIElementResponse | [number, number]>\n    >\n  >;\n  try {\n    res = await callAIWithObjectResponse<AIElementResponse | [number, number]>(\n      msgs,\n      modelConfig,\n    );\n  } catch (callError) {\n    // Return error with usage and rawResponse if available\n    const errorMessage =\n      callError instanceof Error ? callError.message : String(callError);\n    const rawResponse =\n      callError instanceof AIResponseParseError\n        ? callError.rawResponse\n        : errorMessage;\n    const usage =\n      callError instanceof AIResponseParseError ? callError.usage : undefined;\n    return {\n      rect: undefined,\n      parseResult: {\n        elements: [],\n        errors: [`AI call error: ${errorMessage}`],\n      },\n      rawResponse,\n      usage,\n      reasoning_content: undefined,\n    };\n  }\n\n  const rawResponse = JSON.stringify(res.content);\n\n  let resRect: Rect | undefined;\n  let matchedElements: LocateResultElement[] = [];\n  let errors: string[] | undefined =\n    'errors' in res.content ? res.content.errors : [];\n  try {\n    if (\n      'bbox' in res.content &&\n      Array.isArray(res.content.bbox) &&\n      res.content.bbox.length >= 1\n    ) {\n      resRect = adaptBboxToRect(\n        res.content.bbox,\n        imageWidth,\n        imageHeight,\n        options.searchConfig?.rect?.left,\n        options.searchConfig?.rect?.top,\n        originalImageWidth,\n        originalImageHeight,\n        modelFamily,\n        options.searchConfig?.scale,\n      );\n\n      debugInspect('resRect', resRect);\n\n      const element: LocateResultElement = generateElementByRect(\n        resRect,\n        targetElementDescriptionText as string,\n      );\n      errors = [];\n\n      if (element) {\n        matchedElements = [element];\n      }\n    }\n  } catch (e) {\n    const msg =\n      e instanceof Error\n        ? `Failed to parse bbox: ${e.message}`\n        : 'unknown error in locate';\n    if (!errors || errors?.length === 0) {\n      errors = [msg];\n    } else {\n      errors.push(`(${msg})`);\n    }\n  }\n\n  return {\n    rect: resRect,\n    parseResult: {\n      elements: matchedElements as LocateResultElement[],\n      errors: errors as string[],\n    },\n    rawResponse,\n    usage: res.usage,\n    reasoning_content: res.reasoning_content,\n  };\n}\n\nexport async function AiLocateSection(options: {\n  context: UIContext;\n  sectionDescription: TUserPrompt;\n  modelConfig: IModelConfig;\n}): Promise<{\n  rect?: Rect;\n  imageBase64?: string;\n  scale?: number;\n  error?: string;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n}> {\n  const { context, sectionDescription, modelConfig } = options;\n  const { modelFamily } = modelConfig;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const systemPrompt = systemPromptToLocateSection(modelFamily);\n  const sectionLocatorInstructionText = sectionLocatorInstruction(\n    extraTextFromUserPrompt(sectionDescription),\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: screenshotBase64,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: sectionLocatorInstructionText,\n        },\n      ],\n    },\n  ];\n\n  if (typeof sectionDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: sectionDescription.images,\n      convertHttpImage2Base64: sectionDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  let result: Awaited<\n    ReturnType<typeof callAIWithObjectResponse<AISectionLocatorResponse>>\n  >;\n  try {\n    result = await callAIWithObjectResponse<AISectionLocatorResponse>(\n      msgs,\n      modelConfig,\n    );\n  } catch (callError) {\n    // Return error with usage and rawResponse if available\n    const errorMessage =\n      callError instanceof Error ? callError.message : String(callError);\n    const rawResponse =\n      callError instanceof AIResponseParseError\n        ? callError.rawResponse\n        : errorMessage;\n    const usage =\n      callError instanceof AIResponseParseError ? callError.usage : undefined;\n    return {\n      rect: undefined,\n      imageBase64: undefined,\n      error: `AI call error: ${errorMessage}`,\n      rawResponse,\n      usage,\n    };\n  }\n\n  let sectionRect: Rect | undefined;\n  const sectionBbox = result.content.bbox;\n  if (sectionBbox) {\n    const targetRect = adaptBboxToRect(\n      sectionBbox,\n      context.size.width,\n      context.size.height,\n      0,\n      0,\n      context.size.width,\n      context.size.height,\n      modelFamily,\n    );\n    debugSection('original targetRect %j', targetRect);\n\n    const referenceBboxList = result.content.references_bbox || [];\n    debugSection('referenceBboxList %j', referenceBboxList);\n\n    const referenceRects = referenceBboxList\n      .filter((bbox) => Array.isArray(bbox))\n      .map((bbox) => {\n        return adaptBboxToRect(\n          bbox,\n          context.size.width,\n          context.size.height,\n          0,\n          0,\n          context.size.width,\n          context.size.height,\n          modelFamily,\n        );\n      });\n    debugSection('referenceRects %j', referenceRects);\n\n    // merge the sectionRect and referenceRects\n    const mergedRect = mergeRects([targetRect, ...referenceRects]);\n    debugSection('mergedRect %j', mergedRect);\n\n    sectionRect = expandSearchArea(mergedRect, context.size);\n    debugSection('expanded sectionRect %j', sectionRect);\n  }\n\n  let imageBase64 = screenshotBase64;\n  let scale: number | undefined;\n\n  if (sectionRect) {\n    const originalWidth = sectionRect.width;\n    const originalHeight = sectionRect.height;\n\n    const croppedResult = await cropByRect(\n      screenshotBase64,\n      sectionRect,\n      modelFamily === 'qwen2.5-vl',\n    );\n\n    const scaleRatio = 2;\n    const scaledResult = await scaleImage(\n      croppedResult.imageBase64,\n      scaleRatio,\n    );\n\n    imageBase64 = scaledResult.imageBase64;\n    scale = scaleRatio;\n    sectionRect.width = scaledResult.width;\n    sectionRect.height = scaledResult.height;\n\n    debugSection(\n      'scaled sectionRect from %dx%d to %dx%d (scale=%d)',\n      originalWidth,\n      originalHeight,\n      sectionRect.width,\n      sectionRect.height,\n      scale,\n    );\n  }\n\n  return {\n    rect: sectionRect,\n    imageBase64,\n    scale,\n    error: result.content.error,\n    rawResponse: JSON.stringify(result.content),\n    usage: result.usage,\n  };\n}\n\nexport async function AiExtractElementInfo<T>(options: {\n  dataQuery: string | Record<string, string>;\n  multimodalPrompt?: TMultimodalPrompt;\n  context: UIContext;\n  pageDescription?: string;\n  extractOption?: ServiceExtractOption;\n  modelConfig: IModelConfig;\n}) {\n  const { dataQuery, context, extractOption, multimodalPrompt, modelConfig } =\n    options;\n  const systemPrompt = systemPromptToExtract();\n  const screenshotBase64 = context.screenshot.base64;\n\n  const extractDataPromptText = extractDataQueryPrompt(\n    options.pageDescription || '',\n    dataQuery,\n  );\n\n  const userContent: ChatCompletionUserMessageParam['content'] = [];\n\n  if (extractOption?.screenshotIncluded !== false) {\n    userContent.push({\n      type: 'image_url',\n      image_url: {\n        url: screenshotBase64,\n        detail: 'high',\n      },\n    });\n  }\n\n  userContent.push({\n    type: 'text',\n    text: extractDataPromptText,\n  });\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userContent,\n    },\n  ];\n\n  if (multimodalPrompt) {\n    const addOns = await promptsToChatParam({\n      images: multimodalPrompt.images,\n      convertHttpImage2Base64: multimodalPrompt.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const {\n    content: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAI(msgs, modelConfig);\n\n  // Parse XML response to JSON object\n  let parseResult: AIDataExtractionResponse<T>;\n  try {\n    parseResult = parseXMLExtractionResponse<T>(rawResponse);\n  } catch (parseError) {\n    // Throw AIResponseParseError with usage and rawResponse preserved\n    const errorMessage =\n      parseError instanceof Error ? parseError.message : String(parseError);\n    throw new AIResponseParseError(\n      `XML parse error: ${errorMessage}`,\n      rawResponse,\n      usage,\n    );\n  }\n\n  return {\n    parseResult,\n    rawResponse,\n    usage,\n    reasoning_content,\n  };\n}\n\nexport async function AiJudgeOrderSensitive(\n  description: string,\n  callAIFn: typeof callAIWithObjectResponse<{ isOrderSensitive: boolean }>,\n  modelConfig: IModelConfig,\n): Promise<{\n  isOrderSensitive: boolean;\n  usage?: AIUsageInfo;\n}> {\n  const systemPrompt = systemPromptToJudgeOrderSensitive();\n  const userPrompt = orderSensitiveJudgePrompt(description);\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userPrompt,\n    },\n  ];\n\n  const result = await callAIFn(msgs, modelConfig);\n\n  return {\n    isOrderSensitive: result.content.isOrderSensitive ?? false,\n    usage: result.usage,\n  };\n}\n"],"names":["debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","modelConfig","modelFamily","screenshotBase64","assert","targetElementDescriptionText","userInstructionPrompt","findElementPrompt","systemPrompt","isAutoGLM","getAutoGLMLocatePrompt","systemPromptToLocateElement","imagePayload","imageWidth","imageHeight","originalImageWidth","originalImageHeight","paddedResult","paddingToMatchBlockByBase64","addOns","rawResponseContent","usage","callAIWithStringResponse","parsed","parseAutoGLMLocateResponse","resRect","matchedElements","errors","x","y","pixelX","Math","pixelY","finalX","finalY","element","generateElementByPoint","res","callAIWithObjectResponse","callError","errorMessage","Error","String","rawResponse","AIResponseParseError","undefined","JSON","Array","adaptBboxToRect","generateElementByRect","e","msg","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","sectionRect","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandSearchArea","imageBase64","scale","originalWidth","originalHeight","croppedResult","cropByRect","scaleRatio","scaledResult","scaleImage","AiExtractElementInfo","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent","reasoning_content","callAI","parseResult","parseXMLExtractionResponse","parseError","AiJudgeOrderSensitive","description","callAIFn","systemPromptToJudgeOrderSensitive","userPrompt","orderSensitiveJudgePrompt"],"mappings":";;;;;;;;;;;;;AA6DA,MAAMA,eAAeC,SAAS;AAC9B,MAAMC,eAAeD,SAAS;AAE9B,MAAME,0BAA0B,CAACC;IAC/B,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAEP,OAAOA,OAAO,MAAM;AAExB;AAEA,MAAMC,qBAAqB,OACzBC;IAEA,MAAMC,OAAyC,EAAE;IACjD,IAAID,kBAAkB,QAAQ,QAAQ;QACpCC,KAAK,IAAI,CAAC;YACR,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM;gBACR;aACD;QACH;QAEA,KAAK,MAAMC,QAAQF,iBAAiB,MAAM,CAAE;YAC1C,MAAMG,SAAS,MAAMC,mBACnBF,KAAK,GAAG,EACR,CAAC,CAACF,iBAAiB,uBAAuB;YAG5CC,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,MAAM,CAAC,mCAAmC,EAAEC,KAAK,IAAI,CAAC,EAAE,CAAC;oBAC3D;iBACD;YACH;YAEAD,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKE;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;QACF;IACF;IACA,OAAOF;AACT;AAEO,eAAeI,gBAAgBC,OAKrC;IAUC,MAAM,EAAEC,OAAO,EAAEC,wBAAwB,EAAEC,WAAW,EAAE,GAAGH;IAC3D,MAAM,EAAEI,WAAW,EAAE,GAAGD;IACxB,MAAME,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElDK,OACEJ,0BACA;IAEF,MAAMK,+BAA+BhB,wBACnCW;IAEF,MAAMM,wBAAwBC,kBAAkBF;IAChD,MAAMG,eAAeC,UAAUP,eAC3BQ,uBAAuBR,eACvBS,4BAA4BT;IAEhC,IAAIU,eAAeT;IACnB,IAAIU,aAAad,QAAQ,IAAI,CAAC,KAAK;IACnC,IAAIe,cAAcf,QAAQ,IAAI,CAAC,MAAM;IACrC,IAAIgB,qBAAqBF;IACzB,IAAIG,sBAAsBF;IAE1B,IAAIhB,QAAQ,YAAY,EAAE;QACxBM,OACEN,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFM,OACEN,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGFc,eAAed,QAAQ,YAAY,CAAC,WAAW;QAC/Ce,aAAaf,QAAQ,YAAY,CAAC,IAAI,EAAE;QACxCgB,cAAchB,QAAQ,YAAY,CAAC,IAAI,EAAE;QACzCiB,qBAAqBF;QACrBG,sBAAsBF;IACxB,OAAO,IAAIZ,AAAgB,iBAAhBA,aAA8B;QACvC,MAAMe,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAMxB,OAAe;QACnB;YAAE,MAAM;YAAU,SAASe;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKI;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMH,UAAUP,eACZ,CAAC,KAAK,EAAEI,uBAAuB,GAC/BA;gBACN;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAON,0BAAuC;QAChD,MAAMmB,SAAS,MAAM5B,mBAAmB;YACtC,QAAQS,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAP,KAAK,IAAI,IAAI0B;IACf;IAEA,IAAIV,UAAUP,cAAc;QAC1B,MAAM,EAAE,SAASkB,kBAAkB,EAAEC,KAAK,EAAE,GAC1C,MAAMC,yBAAyB7B,MAAMQ;QAEvCf,aAAa,yBAAyBkC;QAEtC,MAAMG,SAASC,2BAA2BJ;QAE1ClC,aAAa,sBAAsBqC,OAAO,KAAK;QAC/CrC,aAAa,yBAAyBqC,OAAO,WAAW;QAExD,IAAIE;QACJ,IAAIC,kBAAyC,EAAE;QAC/C,IAAIC,SAAmB,EAAE;QAEzB,IAAIJ,OAAO,KAAK,IAAI,CAACA,OAAO,WAAW,EAAE;YACvCI,SAAS;gBAACJ,OAAO,KAAK,IAAI;aAAoC;YAC9DrC,aAAa,yBAAyByC,MAAM,CAAC,EAAE;QACjD,OAAO;YACL,MAAM,EAAEC,CAAC,EAAEC,CAAC,EAAE,GAAGN,OAAO,WAAW;YAEnCrC,aAAa,iCAAiC;gBAAE0C;gBAAGC;YAAE;YAIrD,MAAMC,SAASC,KAAK,KAAK,CAAEH,IAAIf,aAAc;YAC7C,MAAMmB,SAASD,KAAK,KAAK,CAAEF,IAAIf,cAAe;YAE9C5B,aAAa,+BAA+B;gBAAE4C;gBAAQE;YAAO;YAG7D,IAAIC,SAASH;YACb,IAAII,SAASF;YACb,IAAIlC,QAAQ,YAAY,EAAE,MAAM;gBAC9BmC,UAAUnC,QAAQ,YAAY,CAAC,IAAI,CAAC,IAAI;gBACxCoC,UAAUpC,QAAQ,YAAY,CAAC,IAAI,CAAC,GAAG;YACzC;YAEA,MAAMqC,UAA+BC,uBACnC;gBAACH;gBAAQC;aAAO,EAChB7B;YAGFoB,UAAUU,QAAQ,IAAI;YACtBjD,aAAa,qBAAqBuC;YAElC,IAAIU,SACFT,kBAAkB;gBAACS;aAAQ;QAE/B;QAEA,OAAO;YACL,MAAMV;YACN,aAAa;gBACX,UAAUC;gBACVC;YACF;YACA,aAAaP;YACbC;YACA,mBAAmBE,OAAO,KAAK;QACjC;IACF;IAEA,IAAIc;IAKJ,IAAI;QACFA,MAAM,MAAMC,yBACV7C,MACAQ;IAEJ,EAAE,OAAOsC,WAAW;QAElB,MAAMC,eACJD,qBAAqBE,QAAQF,UAAU,OAAO,GAAGG,OAAOH;QAC1D,MAAMI,cACJJ,qBAAqBK,uBACjBL,UAAU,WAAW,GACrBC;QACN,MAAMnB,QACJkB,qBAAqBK,uBAAuBL,UAAU,KAAK,GAAGM;QAChE,OAAO;YACL,MAAMA;YACN,aAAa;gBACX,UAAU,EAAE;gBACZ,QAAQ;oBAAC,CAAC,eAAe,EAAEL,cAAc;iBAAC;YAC5C;YACAG;YACAtB;YACA,mBAAmBwB;QACrB;IACF;IAEA,MAAMF,cAAcG,KAAK,SAAS,CAACT,IAAI,OAAO;IAE9C,IAAIZ;IACJ,IAAIC,kBAAyC,EAAE;IAC/C,IAAIC,SACF,YAAYU,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IACE,UAAUA,IAAI,OAAO,IACrBU,MAAM,OAAO,CAACV,IAAI,OAAO,CAAC,IAAI,KAC9BA,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,IAAI,GAC3B;YACAZ,UAAUuB,gBACRX,IAAI,OAAO,CAAC,IAAI,EAChBxB,YACAC,aACAhB,QAAQ,YAAY,EAAE,MAAM,MAC5BA,QAAQ,YAAY,EAAE,MAAM,KAC5BiB,oBACAC,qBACAd,aACAJ,QAAQ,YAAY,EAAE;YAGxBZ,aAAa,WAAWuC;YAExB,MAAMU,UAA+Bc,sBACnCxB,SACApB;YAEFsB,SAAS,EAAE;YAEX,IAAIQ,SACFT,kBAAkB;gBAACS;aAAQ;QAE/B;IACF,EAAE,OAAOe,GAAG;QACV,MAAMC,MACJD,aAAaT,QACT,CAAC,sBAAsB,EAAES,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACvB,UAAUA,QAAQ,WAAW,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAEwB,IAAI,CAAC,CAAC;aAFtBxB,SAAS;YAACwB;SAAI;IAIlB;IAEA,OAAO;QACL,MAAM1B;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAgB;QACA,OAAON,IAAI,KAAK;QAChB,mBAAmBA,IAAI,iBAAiB;IAC1C;AACF;AAEO,eAAee,gBAAgBtD,OAIrC;IAQC,MAAM,EAAEC,OAAO,EAAEsD,kBAAkB,EAAEpD,WAAW,EAAE,GAAGH;IACrD,MAAM,EAAEI,WAAW,EAAE,GAAGD;IACxB,MAAME,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAMS,eAAe8C,4BAA4BpD;IACjD,MAAMqD,gCAAgCC,0BACpCnE,wBAAwBgE;IAE1B,MAAM5D,OAAe;QACnB;YAAE,MAAM;YAAU,SAASe;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKL;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMoD;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAMlC,SAAS,MAAM5B,mBAAmB;YACtC,QAAQ8D,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACA5D,KAAK,IAAI,IAAI0B;IACf;IAEA,IAAIsC;IAGJ,IAAI;QACFA,SAAS,MAAMnB,yBACb7C,MACAQ;IAEJ,EAAE,OAAOsC,WAAW;QAElB,MAAMC,eACJD,qBAAqBE,QAAQF,UAAU,OAAO,GAAGG,OAAOH;QAC1D,MAAMI,cACJJ,qBAAqBK,uBACjBL,UAAU,WAAW,GACrBC;QACN,MAAMnB,QACJkB,qBAAqBK,uBAAuBL,UAAU,KAAK,GAAGM;QAChE,OAAO;YACL,MAAMA;YACN,aAAaA;YACb,OAAO,CAAC,eAAe,EAAEL,cAAc;YACvCG;YACAtB;QACF;IACF;IAEA,IAAIqC;IACJ,MAAMC,cAAcF,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIE,aAAa;QACf,MAAMC,aAAaZ,gBACjBW,aACA5D,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBG;QAEFd,aAAa,0BAA0BwE;QAEvC,MAAMC,oBAAoBJ,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9DrE,aAAa,wBAAwByE;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAAShB,MAAM,OAAO,CAACgB,OAC/B,GAAG,CAAC,CAACA,OACGf,gBACLe,MACAhE,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBG;QAGNd,aAAa,qBAAqB0E;QAGlC,MAAME,aAAaC,WAAW;YAACL;eAAeE;SAAe;QAC7D1E,aAAa,iBAAiB4E;QAE9BN,cAAcQ,iBAAiBF,YAAYjE,QAAQ,IAAI;QACvDX,aAAa,2BAA2BsE;IAC1C;IAEA,IAAIS,cAAchE;IAClB,IAAIiE;IAEJ,IAAIV,aAAa;QACf,MAAMW,gBAAgBX,YAAY,KAAK;QACvC,MAAMY,iBAAiBZ,YAAY,MAAM;QAEzC,MAAMa,gBAAgB,MAAMC,WAC1BrE,kBACAuD,aACAxD,AAAgB,iBAAhBA;QAGF,MAAMuE,aAAa;QACnB,MAAMC,eAAe,MAAMC,WACzBJ,cAAc,WAAW,EACzBE;QAGFN,cAAcO,aAAa,WAAW;QACtCN,QAAQK;QACRf,YAAY,KAAK,GAAGgB,aAAa,KAAK;QACtChB,YAAY,MAAM,GAAGgB,aAAa,MAAM;QAExCtF,aACE,qDACAiF,eACAC,gBACAZ,YAAY,KAAK,EACjBA,YAAY,MAAM,EAClBU;IAEJ;IAEA,OAAO;QACL,MAAMV;QACNS;QACAC;QACA,OAAOX,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAaX,KAAK,SAAS,CAACW,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAemB,qBAAwB9E,OAO7C;IACC,MAAM,EAAE+E,SAAS,EAAE9E,OAAO,EAAE+E,aAAa,EAAEtF,gBAAgB,EAAES,WAAW,EAAE,GACxEH;IACF,MAAMU,eAAeuE;IACrB,MAAM5E,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAMiF,wBAAwBC,uBAC5BnF,QAAQ,eAAe,IAAI,IAC3B+E;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,eAAe,uBAAuB,OACxCI,YAAY,IAAI,CAAC;QACf,MAAM;QACN,WAAW;YACT,KAAK/E;YACL,QAAQ;QACV;IACF;IAGF+E,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAMvF,OAAe;QACnB;YAAE,MAAM;YAAU,SAASe;QAAa;QACxC;YACE,MAAM;YACN,SAAS0E;QACX;KACD;IAED,IAAI1F,kBAAkB;QACpB,MAAM2B,SAAS,MAAM5B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAC,KAAK,IAAI,IAAI0B;IACf;IAEA,MAAM,EACJ,SAASwB,WAAW,EACpBtB,KAAK,EACL8D,iBAAiB,EAClB,GAAG,MAAMC,OAAO3F,MAAMQ;IAGvB,IAAIoF;IACJ,IAAI;QACFA,cAAcC,2BAA8B3C;IAC9C,EAAE,OAAO4C,YAAY;QAEnB,MAAM/C,eACJ+C,sBAAsB9C,QAAQ8C,WAAW,OAAO,GAAG7C,OAAO6C;QAC5D,MAAM,IAAI3C,qBACR,CAAC,iBAAiB,EAAEJ,cAAc,EAClCG,aACAtB;IAEJ;IAEA,OAAO;QACLgE;QACA1C;QACAtB;QACA8D;IACF;AACF;AAEO,eAAeK,sBACpBC,WAAmB,EACnBC,QAAwE,EACxEzF,WAAyB;IAKzB,MAAMO,eAAemF;IACrB,MAAMC,aAAaC,0BAA0BJ;IAE7C,MAAMhG,OAAe;QACnB;YAAE,MAAM;YAAU,SAASe;QAAa;QACxC;YACE,MAAM;YACN,SAASoF;QACX;KACD;IAED,MAAMnC,SAAS,MAAMiC,SAASjG,MAAMQ;IAEpC,OAAO;QACL,kBAAkBwD,OAAO,OAAO,CAAC,gBAAgB,IAAI;QACrD,OAAOA,OAAO,KAAK;IACrB;AACF"}
+{"version":3,"file":"ai-model/inspect.mjs","sources":["../../../src/ai-model/inspect.ts"],"sourcesContent":["import type {\n  AIDataExtractionResponse,\n  AIElementResponse,\n  AISectionLocatorResponse,\n  AIUsageInfo,\n  Rect,\n  ServiceExtractOption,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport {\n  generateElementByPoint,\n  generateElementByRect,\n} from '@midscene/shared/extractor/dom-util';\nimport {\n  cropByRect,\n  paddingToMatchBlockByBase64,\n  preProcessImageUrl,\n  scaleImage,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { LocateResultElement } from '@midscene/shared/types';\nimport { assert } from '@midscene/shared/utils';\nimport type {\n  ChatCompletionSystemMessageParam,\n  ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport type { TMultimodalPrompt, TUserPrompt } from '../common';\nimport { adaptBboxToRect, expandSearchArea, mergeRects } from '../common';\nimport { parseAutoGLMLocateResponse } from './auto-glm/parser';\nimport { getAutoGLMLocatePrompt } from './auto-glm/prompt';\nimport { isAutoGLM } from './auto-glm/util';\nimport {\n  extractDataQueryPrompt,\n  parseXMLExtractionResponse,\n  systemPromptToExtract,\n} from './prompt/extraction';\nimport {\n  findElementPrompt,\n  systemPromptToLocateElement,\n} from './prompt/llm-locator';\nimport {\n  sectionLocatorInstruction,\n  systemPromptToLocateSection,\n} from './prompt/llm-section-locator';\nimport {\n  orderSensitiveJudgePrompt,\n  systemPromptToJudgeOrderSensitive,\n} from './prompt/order-sensitive-judge';\nimport {\n  AIResponseParseError,\n  callAI,\n  callAIWithObjectResponse,\n  callAIWithStringResponse,\n} from './service-caller/index';\n\nexport type AIArgs = [\n  ChatCompletionSystemMessageParam,\n  ...ChatCompletionUserMessageParam[],\n];\n\nconst debugInspect = getDebug('ai:inspect');\nconst debugSection = getDebug('ai:section');\n\nconst extraTextFromUserPrompt = (prompt: TUserPrompt): string => {\n  if (typeof prompt === 'string') {\n    return prompt;\n  } else {\n    return prompt.prompt;\n  }\n};\n\nconst promptsToChatParam = async (\n  multimodalPrompt: TMultimodalPrompt,\n): Promise<ChatCompletionUserMessageParam[]> => {\n  const msgs: ChatCompletionUserMessageParam[] = [];\n  if (multimodalPrompt?.images?.length) {\n    msgs.push({\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: 'Next, I will provide all the reference images.',\n        },\n      ],\n    });\n\n    for (const item of multimodalPrompt.images) {\n      const base64 = await preProcessImageUrl(\n        item.url,\n        !!multimodalPrompt.convertHttpImage2Base64,\n      );\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'text',\n            text: `this is the reference image named '${item.name}':`,\n          },\n        ],\n      });\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: base64,\n              detail: 'high',\n            },\n          },\n        ],\n      });\n    }\n  }\n  return msgs;\n};\n\nexport async function AiLocateElement(options: {\n  context: UIContext;\n  targetElementDescription: TUserPrompt;\n  searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n  modelConfig: IModelConfig;\n}): Promise<{\n  parseResult: {\n    elements: LocateResultElement[];\n    errors?: string[];\n  };\n  rect?: Rect;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}> {\n  const { context, targetElementDescription, modelConfig } = options;\n  const { modelFamily } = modelConfig;\n  const screenshotBase64 = context.screenshot.base64;\n\n  assert(\n    targetElementDescription,\n    'cannot find the target element description',\n  );\n  const targetElementDescriptionText = extraTextFromUserPrompt(\n    targetElementDescription,\n  );\n  const userInstructionPrompt = findElementPrompt(targetElementDescriptionText);\n  const systemPrompt = isAutoGLM(modelFamily)\n    ? getAutoGLMLocatePrompt(modelFamily)\n    : systemPromptToLocateElement(modelFamily);\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = context.shotSize.width;\n  let imageHeight = context.shotSize.height;\n  let originalImageWidth = imageWidth;\n  let originalImageHeight = imageHeight;\n\n  if (options.searchConfig) {\n    assert(\n      options.searchConfig.rect,\n      'searchArea is provided but its rect cannot be found. Failed to locate element',\n    );\n    assert(\n      options.searchConfig.imageBase64,\n      'searchArea is provided but its imageBase64 cannot be found. Failed to locate element',\n    );\n\n    imagePayload = options.searchConfig.imageBase64;\n    imageWidth = options.searchConfig.rect?.width;\n    imageHeight = options.searchConfig.rect?.height;\n    originalImageWidth = imageWidth;\n    originalImageHeight = imageHeight;\n  } else if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\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          type: 'text',\n          text: isAutoGLM(modelFamily)\n            ? `Tap: ${userInstructionPrompt}`\n            : userInstructionPrompt,\n        },\n      ],\n    },\n  ];\n\n  if (typeof targetElementDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: targetElementDescription.images,\n      convertHttpImage2Base64: targetElementDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  if (isAutoGLM(modelFamily)) {\n    const { content: rawResponseContent, usage } =\n      await callAIWithStringResponse(msgs, modelConfig);\n\n    debugInspect('auto-glm rawResponse:', rawResponseContent);\n\n    const parsed = parseAutoGLMLocateResponse(rawResponseContent);\n\n    debugInspect('auto-glm thinking:', parsed.think);\n    debugInspect('auto-glm coordinates:', parsed.coordinates);\n\n    let resRect: Rect | undefined;\n    let matchedElements: LocateResultElement[] = [];\n    let errors: string[] = [];\n\n    if (parsed.error || !parsed.coordinates) {\n      errors = [parsed.error || 'Failed to parse auto-glm response'];\n      debugInspect('auto-glm parse error:', errors[0]);\n    } else {\n      const { x, y } = parsed.coordinates;\n\n      debugInspect('auto-glm coordinates [0-999]:', { x, y });\n\n      // Convert auto-glm coordinates [0,999] to pixel bbox\n      // Map from [0,999] to pixel coordinates\n      const pixelX = Math.round((x * imageWidth) / 1000);\n      const pixelY = Math.round((y * imageHeight) / 1000);\n\n      debugInspect('auto-glm pixel coordinates:', { pixelX, pixelY });\n\n      // Apply offset if searching in a cropped area\n      let finalX = pixelX;\n      let finalY = pixelY;\n      if (options.searchConfig?.rect) {\n        finalX += options.searchConfig.rect.left;\n        finalY += options.searchConfig.rect.top;\n      }\n\n      const element: LocateResultElement = generateElementByPoint(\n        [finalX, finalY],\n        targetElementDescriptionText as string,\n      );\n\n      resRect = element.rect;\n      debugInspect('auto-glm resRect:', resRect);\n\n      if (element) {\n        matchedElements = [element];\n      }\n    }\n\n    return {\n      rect: resRect,\n      parseResult: {\n        elements: matchedElements,\n        errors,\n      },\n      rawResponse: rawResponseContent,\n      usage,\n      reasoning_content: parsed.think,\n    };\n  }\n\n  let res: Awaited<\n    ReturnType<\n      typeof callAIWithObjectResponse<AIElementResponse | [number, number]>\n    >\n  >;\n  try {\n    res = await callAIWithObjectResponse<AIElementResponse | [number, number]>(\n      msgs,\n      modelConfig,\n    );\n  } catch (callError) {\n    // Return error with usage and rawResponse if available\n    const errorMessage =\n      callError instanceof Error ? callError.message : String(callError);\n    const rawResponse =\n      callError instanceof AIResponseParseError\n        ? callError.rawResponse\n        : errorMessage;\n    const usage =\n      callError instanceof AIResponseParseError ? callError.usage : undefined;\n    return {\n      rect: undefined,\n      parseResult: {\n        elements: [],\n        errors: [`AI call error: ${errorMessage}`],\n      },\n      rawResponse,\n      usage,\n      reasoning_content: undefined,\n    };\n  }\n\n  const rawResponse = JSON.stringify(res.content);\n\n  let resRect: Rect | undefined;\n  let matchedElements: LocateResultElement[] = [];\n  let errors: string[] | undefined =\n    'errors' in res.content ? res.content.errors : [];\n  try {\n    if (\n      'bbox' in res.content &&\n      Array.isArray(res.content.bbox) &&\n      res.content.bbox.length >= 1\n    ) {\n      resRect = adaptBboxToRect(\n        res.content.bbox,\n        imageWidth,\n        imageHeight,\n        options.searchConfig?.rect?.left,\n        options.searchConfig?.rect?.top,\n        originalImageWidth,\n        originalImageHeight,\n        modelFamily,\n        options.searchConfig?.scale,\n      );\n\n      debugInspect('resRect', resRect);\n\n      const element: LocateResultElement = generateElementByRect(\n        resRect,\n        targetElementDescriptionText as string,\n      );\n      errors = [];\n\n      if (element) {\n        matchedElements = [element];\n      }\n    }\n  } catch (e) {\n    const msg =\n      e instanceof Error\n        ? `Failed to parse bbox: ${e.message}`\n        : 'unknown error in locate';\n    if (!errors || errors?.length === 0) {\n      errors = [msg];\n    } else {\n      errors.push(`(${msg})`);\n    }\n  }\n\n  return {\n    rect: resRect,\n    parseResult: {\n      elements: matchedElements as LocateResultElement[],\n      errors: errors as string[],\n    },\n    rawResponse,\n    usage: res.usage,\n    reasoning_content: res.reasoning_content,\n  };\n}\n\nexport async function AiLocateSection(options: {\n  context: UIContext;\n  sectionDescription: TUserPrompt;\n  modelConfig: IModelConfig;\n}): Promise<{\n  rect?: Rect;\n  imageBase64?: string;\n  scale?: number;\n  error?: string;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n}> {\n  const { context, sectionDescription, modelConfig } = options;\n  const { modelFamily } = modelConfig;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const systemPrompt = systemPromptToLocateSection(modelFamily);\n  const sectionLocatorInstructionText = sectionLocatorInstruction(\n    extraTextFromUserPrompt(sectionDescription),\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: screenshotBase64,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: sectionLocatorInstructionText,\n        },\n      ],\n    },\n  ];\n\n  if (typeof sectionDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: sectionDescription.images,\n      convertHttpImage2Base64: sectionDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  let result: Awaited<\n    ReturnType<typeof callAIWithObjectResponse<AISectionLocatorResponse>>\n  >;\n  try {\n    result = await callAIWithObjectResponse<AISectionLocatorResponse>(\n      msgs,\n      modelConfig,\n    );\n  } catch (callError) {\n    // Return error with usage and rawResponse if available\n    const errorMessage =\n      callError instanceof Error ? callError.message : String(callError);\n    const rawResponse =\n      callError instanceof AIResponseParseError\n        ? callError.rawResponse\n        : errorMessage;\n    const usage =\n      callError instanceof AIResponseParseError ? callError.usage : undefined;\n    return {\n      rect: undefined,\n      imageBase64: undefined,\n      error: `AI call error: ${errorMessage}`,\n      rawResponse,\n      usage,\n    };\n  }\n\n  let sectionRect: Rect | undefined;\n  const sectionBbox = result.content.bbox;\n  if (sectionBbox) {\n    const targetRect = adaptBboxToRect(\n      sectionBbox,\n      context.shotSize.width,\n      context.shotSize.height,\n      0,\n      0,\n      context.shotSize.width,\n      context.shotSize.height,\n      modelFamily,\n    );\n    debugSection('original targetRect %j', targetRect);\n\n    const referenceBboxList = result.content.references_bbox || [];\n    debugSection('referenceBboxList %j', referenceBboxList);\n\n    const referenceRects = referenceBboxList\n      .filter((bbox) => Array.isArray(bbox))\n      .map((bbox) => {\n        return adaptBboxToRect(\n          bbox,\n          context.shotSize.width,\n          context.shotSize.height,\n          0,\n          0,\n          context.shotSize.width,\n          context.shotSize.height,\n          modelFamily,\n        );\n      });\n    debugSection('referenceRects %j', referenceRects);\n\n    // merge the sectionRect and referenceRects\n    const mergedRect = mergeRects([targetRect, ...referenceRects]);\n    debugSection('mergedRect %j', mergedRect);\n\n    sectionRect = expandSearchArea(mergedRect, context.shotSize);\n    debugSection('expanded sectionRect %j', sectionRect);\n  }\n\n  let imageBase64 = screenshotBase64;\n  let scale: number | undefined;\n\n  if (sectionRect) {\n    const originalWidth = sectionRect.width;\n    const originalHeight = sectionRect.height;\n\n    const croppedResult = await cropByRect(\n      screenshotBase64,\n      sectionRect,\n      modelFamily === 'qwen2.5-vl',\n    );\n\n    const scaleRatio = 2;\n    const scaledResult = await scaleImage(\n      croppedResult.imageBase64,\n      scaleRatio,\n    );\n\n    imageBase64 = scaledResult.imageBase64;\n    scale = scaleRatio;\n    sectionRect.width = scaledResult.width;\n    sectionRect.height = scaledResult.height;\n\n    debugSection(\n      'scaled sectionRect from %dx%d to %dx%d (scale=%d)',\n      originalWidth,\n      originalHeight,\n      sectionRect.width,\n      sectionRect.height,\n      scale,\n    );\n  }\n\n  return {\n    rect: sectionRect,\n    imageBase64,\n    scale,\n    error: result.content.error,\n    rawResponse: JSON.stringify(result.content),\n    usage: result.usage,\n  };\n}\n\nexport async function AiExtractElementInfo<T>(options: {\n  dataQuery: string | Record<string, string>;\n  multimodalPrompt?: TMultimodalPrompt;\n  context: UIContext;\n  pageDescription?: string;\n  extractOption?: ServiceExtractOption;\n  modelConfig: IModelConfig;\n}) {\n  const { dataQuery, context, extractOption, multimodalPrompt, modelConfig } =\n    options;\n  const systemPrompt = systemPromptToExtract();\n  const screenshotBase64 = context.screenshot.base64;\n\n  const extractDataPromptText = extractDataQueryPrompt(\n    options.pageDescription || '',\n    dataQuery,\n  );\n\n  const userContent: ChatCompletionUserMessageParam['content'] = [];\n\n  if (extractOption?.screenshotIncluded !== false) {\n    userContent.push({\n      type: 'image_url',\n      image_url: {\n        url: screenshotBase64,\n        detail: 'high',\n      },\n    });\n  }\n\n  userContent.push({\n    type: 'text',\n    text: extractDataPromptText,\n  });\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userContent,\n    },\n  ];\n\n  if (multimodalPrompt) {\n    const addOns = await promptsToChatParam({\n      images: multimodalPrompt.images,\n      convertHttpImage2Base64: multimodalPrompt.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const {\n    content: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAI(msgs, modelConfig);\n\n  // Parse XML response to JSON object\n  let parseResult: AIDataExtractionResponse<T>;\n  try {\n    parseResult = parseXMLExtractionResponse<T>(rawResponse);\n  } catch (parseError) {\n    // Throw AIResponseParseError with usage and rawResponse preserved\n    const errorMessage =\n      parseError instanceof Error ? parseError.message : String(parseError);\n    throw new AIResponseParseError(\n      `XML parse error: ${errorMessage}`,\n      rawResponse,\n      usage,\n    );\n  }\n\n  return {\n    parseResult,\n    rawResponse,\n    usage,\n    reasoning_content,\n  };\n}\n\nexport async function AiJudgeOrderSensitive(\n  description: string,\n  callAIFn: typeof callAIWithObjectResponse<{ isOrderSensitive: boolean }>,\n  modelConfig: IModelConfig,\n): Promise<{\n  isOrderSensitive: boolean;\n  usage?: AIUsageInfo;\n}> {\n  const systemPrompt = systemPromptToJudgeOrderSensitive();\n  const userPrompt = orderSensitiveJudgePrompt(description);\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userPrompt,\n    },\n  ];\n\n  const result = await callAIFn(msgs, modelConfig);\n\n  return {\n    isOrderSensitive: result.content.isOrderSensitive ?? false,\n    usage: result.usage,\n  };\n}\n"],"names":["debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","modelConfig","modelFamily","screenshotBase64","assert","targetElementDescriptionText","userInstructionPrompt","findElementPrompt","systemPrompt","isAutoGLM","getAutoGLMLocatePrompt","systemPromptToLocateElement","imagePayload","imageWidth","imageHeight","originalImageWidth","originalImageHeight","paddedResult","paddingToMatchBlockByBase64","addOns","rawResponseContent","usage","callAIWithStringResponse","parsed","parseAutoGLMLocateResponse","resRect","matchedElements","errors","x","y","pixelX","Math","pixelY","finalX","finalY","element","generateElementByPoint","res","callAIWithObjectResponse","callError","errorMessage","Error","String","rawResponse","AIResponseParseError","undefined","JSON","Array","adaptBboxToRect","generateElementByRect","e","msg","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","sectionRect","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandSearchArea","imageBase64","scale","originalWidth","originalHeight","croppedResult","cropByRect","scaleRatio","scaledResult","scaleImage","AiExtractElementInfo","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent","reasoning_content","callAI","parseResult","parseXMLExtractionResponse","parseError","AiJudgeOrderSensitive","description","callAIFn","systemPromptToJudgeOrderSensitive","userPrompt","orderSensitiveJudgePrompt"],"mappings":";;;;;;;;;;;;;AA6DA,MAAMA,eAAeC,SAAS;AAC9B,MAAMC,eAAeD,SAAS;AAE9B,MAAME,0BAA0B,CAACC;IAC/B,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAEP,OAAOA,OAAO,MAAM;AAExB;AAEA,MAAMC,qBAAqB,OACzBC;IAEA,MAAMC,OAAyC,EAAE;IACjD,IAAID,kBAAkB,QAAQ,QAAQ;QACpCC,KAAK,IAAI,CAAC;YACR,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM;gBACR;aACD;QACH;QAEA,KAAK,MAAMC,QAAQF,iBAAiB,MAAM,CAAE;YAC1C,MAAMG,SAAS,MAAMC,mBACnBF,KAAK,GAAG,EACR,CAAC,CAACF,iBAAiB,uBAAuB;YAG5CC,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,MAAM,CAAC,mCAAmC,EAAEC,KAAK,IAAI,CAAC,EAAE,CAAC;oBAC3D;iBACD;YACH;YAEAD,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKE;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;QACF;IACF;IACA,OAAOF;AACT;AAEO,eAAeI,gBAAgBC,OAKrC;IAUC,MAAM,EAAEC,OAAO,EAAEC,wBAAwB,EAAEC,WAAW,EAAE,GAAGH;IAC3D,MAAM,EAAEI,WAAW,EAAE,GAAGD;IACxB,MAAME,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElDK,OACEJ,0BACA;IAEF,MAAMK,+BAA+BhB,wBACnCW;IAEF,MAAMM,wBAAwBC,kBAAkBF;IAChD,MAAMG,eAAeC,UAAUP,eAC3BQ,uBAAuBR,eACvBS,4BAA4BT;IAEhC,IAAIU,eAAeT;IACnB,IAAIU,aAAad,QAAQ,QAAQ,CAAC,KAAK;IACvC,IAAIe,cAAcf,QAAQ,QAAQ,CAAC,MAAM;IACzC,IAAIgB,qBAAqBF;IACzB,IAAIG,sBAAsBF;IAE1B,IAAIhB,QAAQ,YAAY,EAAE;QACxBM,OACEN,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFM,OACEN,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGFc,eAAed,QAAQ,YAAY,CAAC,WAAW;QAC/Ce,aAAaf,QAAQ,YAAY,CAAC,IAAI,EAAE;QACxCgB,cAAchB,QAAQ,YAAY,CAAC,IAAI,EAAE;QACzCiB,qBAAqBF;QACrBG,sBAAsBF;IACxB,OAAO,IAAIZ,AAAgB,iBAAhBA,aAA8B;QACvC,MAAMe,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAMxB,OAAe;QACnB;YAAE,MAAM;YAAU,SAASe;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKI;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMH,UAAUP,eACZ,CAAC,KAAK,EAAEI,uBAAuB,GAC/BA;gBACN;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAON,0BAAuC;QAChD,MAAMmB,SAAS,MAAM5B,mBAAmB;YACtC,QAAQS,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAP,KAAK,IAAI,IAAI0B;IACf;IAEA,IAAIV,UAAUP,cAAc;QAC1B,MAAM,EAAE,SAASkB,kBAAkB,EAAEC,KAAK,EAAE,GAC1C,MAAMC,yBAAyB7B,MAAMQ;QAEvCf,aAAa,yBAAyBkC;QAEtC,MAAMG,SAASC,2BAA2BJ;QAE1ClC,aAAa,sBAAsBqC,OAAO,KAAK;QAC/CrC,aAAa,yBAAyBqC,OAAO,WAAW;QAExD,IAAIE;QACJ,IAAIC,kBAAyC,EAAE;QAC/C,IAAIC,SAAmB,EAAE;QAEzB,IAAIJ,OAAO,KAAK,IAAI,CAACA,OAAO,WAAW,EAAE;YACvCI,SAAS;gBAACJ,OAAO,KAAK,IAAI;aAAoC;YAC9DrC,aAAa,yBAAyByC,MAAM,CAAC,EAAE;QACjD,OAAO;YACL,MAAM,EAAEC,CAAC,EAAEC,CAAC,EAAE,GAAGN,OAAO,WAAW;YAEnCrC,aAAa,iCAAiC;gBAAE0C;gBAAGC;YAAE;YAIrD,MAAMC,SAASC,KAAK,KAAK,CAAEH,IAAIf,aAAc;YAC7C,MAAMmB,SAASD,KAAK,KAAK,CAAEF,IAAIf,cAAe;YAE9C5B,aAAa,+BAA+B;gBAAE4C;gBAAQE;YAAO;YAG7D,IAAIC,SAASH;YACb,IAAII,SAASF;YACb,IAAIlC,QAAQ,YAAY,EAAE,MAAM;gBAC9BmC,UAAUnC,QAAQ,YAAY,CAAC,IAAI,CAAC,IAAI;gBACxCoC,UAAUpC,QAAQ,YAAY,CAAC,IAAI,CAAC,GAAG;YACzC;YAEA,MAAMqC,UAA+BC,uBACnC;gBAACH;gBAAQC;aAAO,EAChB7B;YAGFoB,UAAUU,QAAQ,IAAI;YACtBjD,aAAa,qBAAqBuC;YAElC,IAAIU,SACFT,kBAAkB;gBAACS;aAAQ;QAE/B;QAEA,OAAO;YACL,MAAMV;YACN,aAAa;gBACX,UAAUC;gBACVC;YACF;YACA,aAAaP;YACbC;YACA,mBAAmBE,OAAO,KAAK;QACjC;IACF;IAEA,IAAIc;IAKJ,IAAI;QACFA,MAAM,MAAMC,yBACV7C,MACAQ;IAEJ,EAAE,OAAOsC,WAAW;QAElB,MAAMC,eACJD,qBAAqBE,QAAQF,UAAU,OAAO,GAAGG,OAAOH;QAC1D,MAAMI,cACJJ,qBAAqBK,uBACjBL,UAAU,WAAW,GACrBC;QACN,MAAMnB,QACJkB,qBAAqBK,uBAAuBL,UAAU,KAAK,GAAGM;QAChE,OAAO;YACL,MAAMA;YACN,aAAa;gBACX,UAAU,EAAE;gBACZ,QAAQ;oBAAC,CAAC,eAAe,EAAEL,cAAc;iBAAC;YAC5C;YACAG;YACAtB;YACA,mBAAmBwB;QACrB;IACF;IAEA,MAAMF,cAAcG,KAAK,SAAS,CAACT,IAAI,OAAO;IAE9C,IAAIZ;IACJ,IAAIC,kBAAyC,EAAE;IAC/C,IAAIC,SACF,YAAYU,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IACE,UAAUA,IAAI,OAAO,IACrBU,MAAM,OAAO,CAACV,IAAI,OAAO,CAAC,IAAI,KAC9BA,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,IAAI,GAC3B;YACAZ,UAAUuB,gBACRX,IAAI,OAAO,CAAC,IAAI,EAChBxB,YACAC,aACAhB,QAAQ,YAAY,EAAE,MAAM,MAC5BA,QAAQ,YAAY,EAAE,MAAM,KAC5BiB,oBACAC,qBACAd,aACAJ,QAAQ,YAAY,EAAE;YAGxBZ,aAAa,WAAWuC;YAExB,MAAMU,UAA+Bc,sBACnCxB,SACApB;YAEFsB,SAAS,EAAE;YAEX,IAAIQ,SACFT,kBAAkB;gBAACS;aAAQ;QAE/B;IACF,EAAE,OAAOe,GAAG;QACV,MAAMC,MACJD,aAAaT,QACT,CAAC,sBAAsB,EAAES,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACvB,UAAUA,QAAQ,WAAW,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAEwB,IAAI,CAAC,CAAC;aAFtBxB,SAAS;YAACwB;SAAI;IAIlB;IAEA,OAAO;QACL,MAAM1B;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAgB;QACA,OAAON,IAAI,KAAK;QAChB,mBAAmBA,IAAI,iBAAiB;IAC1C;AACF;AAEO,eAAee,gBAAgBtD,OAIrC;IAQC,MAAM,EAAEC,OAAO,EAAEsD,kBAAkB,EAAEpD,WAAW,EAAE,GAAGH;IACrD,MAAM,EAAEI,WAAW,EAAE,GAAGD;IACxB,MAAME,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAMS,eAAe8C,4BAA4BpD;IACjD,MAAMqD,gCAAgCC,0BACpCnE,wBAAwBgE;IAE1B,MAAM5D,OAAe;QACnB;YAAE,MAAM;YAAU,SAASe;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKL;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMoD;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAMlC,SAAS,MAAM5B,mBAAmB;YACtC,QAAQ8D,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACA5D,KAAK,IAAI,IAAI0B;IACf;IAEA,IAAIsC;IAGJ,IAAI;QACFA,SAAS,MAAMnB,yBACb7C,MACAQ;IAEJ,EAAE,OAAOsC,WAAW;QAElB,MAAMC,eACJD,qBAAqBE,QAAQF,UAAU,OAAO,GAAGG,OAAOH;QAC1D,MAAMI,cACJJ,qBAAqBK,uBACjBL,UAAU,WAAW,GACrBC;QACN,MAAMnB,QACJkB,qBAAqBK,uBAAuBL,UAAU,KAAK,GAAGM;QAChE,OAAO;YACL,MAAMA;YACN,aAAaA;YACb,OAAO,CAAC,eAAe,EAAEL,cAAc;YACvCG;YACAtB;QACF;IACF;IAEA,IAAIqC;IACJ,MAAMC,cAAcF,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIE,aAAa;QACf,MAAMC,aAAaZ,gBACjBW,aACA5D,QAAQ,QAAQ,CAAC,KAAK,EACtBA,QAAQ,QAAQ,CAAC,MAAM,EACvB,GACA,GACAA,QAAQ,QAAQ,CAAC,KAAK,EACtBA,QAAQ,QAAQ,CAAC,MAAM,EACvBG;QAEFd,aAAa,0BAA0BwE;QAEvC,MAAMC,oBAAoBJ,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9DrE,aAAa,wBAAwByE;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAAShB,MAAM,OAAO,CAACgB,OAC/B,GAAG,CAAC,CAACA,OACGf,gBACLe,MACAhE,QAAQ,QAAQ,CAAC,KAAK,EACtBA,QAAQ,QAAQ,CAAC,MAAM,EACvB,GACA,GACAA,QAAQ,QAAQ,CAAC,KAAK,EACtBA,QAAQ,QAAQ,CAAC,MAAM,EACvBG;QAGNd,aAAa,qBAAqB0E;QAGlC,MAAME,aAAaC,WAAW;YAACL;eAAeE;SAAe;QAC7D1E,aAAa,iBAAiB4E;QAE9BN,cAAcQ,iBAAiBF,YAAYjE,QAAQ,QAAQ;QAC3DX,aAAa,2BAA2BsE;IAC1C;IAEA,IAAIS,cAAchE;IAClB,IAAIiE;IAEJ,IAAIV,aAAa;QACf,MAAMW,gBAAgBX,YAAY,KAAK;QACvC,MAAMY,iBAAiBZ,YAAY,MAAM;QAEzC,MAAMa,gBAAgB,MAAMC,WAC1BrE,kBACAuD,aACAxD,AAAgB,iBAAhBA;QAGF,MAAMuE,aAAa;QACnB,MAAMC,eAAe,MAAMC,WACzBJ,cAAc,WAAW,EACzBE;QAGFN,cAAcO,aAAa,WAAW;QACtCN,QAAQK;QACRf,YAAY,KAAK,GAAGgB,aAAa,KAAK;QACtChB,YAAY,MAAM,GAAGgB,aAAa,MAAM;QAExCtF,aACE,qDACAiF,eACAC,gBACAZ,YAAY,KAAK,EACjBA,YAAY,MAAM,EAClBU;IAEJ;IAEA,OAAO;QACL,MAAMV;QACNS;QACAC;QACA,OAAOX,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAaX,KAAK,SAAS,CAACW,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAemB,qBAAwB9E,OAO7C;IACC,MAAM,EAAE+E,SAAS,EAAE9E,OAAO,EAAE+E,aAAa,EAAEtF,gBAAgB,EAAES,WAAW,EAAE,GACxEH;IACF,MAAMU,eAAeuE;IACrB,MAAM5E,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAMiF,wBAAwBC,uBAC5BnF,QAAQ,eAAe,IAAI,IAC3B+E;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,eAAe,uBAAuB,OACxCI,YAAY,IAAI,CAAC;QACf,MAAM;QACN,WAAW;YACT,KAAK/E;YACL,QAAQ;QACV;IACF;IAGF+E,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAMvF,OAAe;QACnB;YAAE,MAAM;YAAU,SAASe;QAAa;QACxC;YACE,MAAM;YACN,SAAS0E;QACX;KACD;IAED,IAAI1F,kBAAkB;QACpB,MAAM2B,SAAS,MAAM5B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAC,KAAK,IAAI,IAAI0B;IACf;IAEA,MAAM,EACJ,SAASwB,WAAW,EACpBtB,KAAK,EACL8D,iBAAiB,EAClB,GAAG,MAAMC,OAAO3F,MAAMQ;IAGvB,IAAIoF;IACJ,IAAI;QACFA,cAAcC,2BAA8B3C;IAC9C,EAAE,OAAO4C,YAAY;QAEnB,MAAM/C,eACJ+C,sBAAsB9C,QAAQ8C,WAAW,OAAO,GAAG7C,OAAO6C;QAC5D,MAAM,IAAI3C,qBACR,CAAC,iBAAiB,EAAEJ,cAAc,EAClCG,aACAtB;IAEJ;IAEA,OAAO;QACLgE;QACA1C;QACAtB;QACA8D;IACF;AACF;AAEO,eAAeK,sBACpBC,WAAmB,EACnBC,QAAwE,EACxEzF,WAAyB;IAKzB,MAAMO,eAAemF;IACrB,MAAMC,aAAaC,0BAA0BJ;IAE7C,MAAMhG,OAAe;QACnB;YAAE,MAAM;YAAU,SAASe;QAAa;QACxC;YACE,MAAM;YACN,SAASoF;QACX;KACD;IAED,MAAMnC,SAAS,MAAMiC,SAASjG,MAAMQ;IAEpC,OAAO;QACL,kBAAkBwD,OAAO,OAAO,CAAC,gBAAgB,IAAI;QACrD,OAAOA,OAAO,KAAK;IACrB;AACF"}

package/dist/es/ai-model/llm-planning.mjs

@@ -62,7 +62,7 @@ function parseXMLPlanningResponse(xmlString, modelFamily) {
 }
 async function plan(userInstruction, opts) {
     const { context, modelConfig, conversationHistory } = opts;
-    const { size } = context;
+    const { shotSize } = context;
     const screenshotBase64 = context.screenshot.base64;
     const { modelFamily } = modelConfig;
     const systemPrompt = await systemPromptToTaskPlanning({
@@ -73,8 +73,8 @@ async function plan(userInstruction, opts) {
         deepThink: true === opts.deepThink
     });
     let imagePayload = screenshotBase64;
-    let imageWidth = size.width;
-    let imageHeight = size.height;
+    let imageWidth = shotSize.width;
+    let imageHeight = shotSize.height;
     if ('qwen2.5-vl' === modelFamily) {
         const paddedResult = await paddingToMatchBlockByBase64(imagePayload);
         imageWidth = paddedResult.width;

package/dist/es/ai-model/llm-planning.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/llm-planning.mjs","sources":["../../../src/ai-model/llm-planning.ts"],"sourcesContent":["import type {\n  DeepThinkOption,\n  DeviceAction,\n  InterfaceType,\n  PlanningAIResponse,\n  RawResponsePlanningAIResponse,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig, TModelFamily } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport {\n  buildYamlFlowFromPlans,\n  fillBboxParam,\n  findAllMidsceneLocatorField,\n} from '../common';\nimport type { ConversationHistory } from './conversation-history';\nimport { systemPromptToTaskPlanning } from './prompt/llm-planning';\nimport { extractXMLTag } from './prompt/util';\nimport {\n  AIResponseParseError,\n  callAI,\n  safeParseJson,\n} from './service-caller/index';\n\nconst debug = getDebug('planning');\nconst warnLog = getDebug('planning', { console: true });\n\n/**\n * Parse XML response from LLM and convert to RawResponsePlanningAIResponse\n */\nexport function parseXMLPlanningResponse(\n  xmlString: string,\n  modelFamily: TModelFamily | undefined,\n): RawResponsePlanningAIResponse {\n  const thought = extractXMLTag(xmlString, 'thought');\n  const memory = extractXMLTag(xmlString, 'memory');\n  const log = extractXMLTag(xmlString, 'log') || '';\n  const error = extractXMLTag(xmlString, 'error');\n  const actionType = extractXMLTag(xmlString, 'action-type');\n  const actionParamStr = extractXMLTag(xmlString, 'action-param-json');\n\n  // Parse <complete> tag with success attribute\n  const completeGoalRegex =\n    /<complete\\s+success=\"(true|false)\">([\\s\\S]*?)<\\/complete>/i;\n  const completeGoalMatch = xmlString.match(completeGoalRegex);\n  let finalizeMessage: string | undefined;\n  let finalizeSuccess: boolean | undefined;\n\n  if (completeGoalMatch) {\n    finalizeSuccess = completeGoalMatch[1] === 'true';\n    finalizeMessage = completeGoalMatch[2]?.trim() || undefined;\n  }\n\n  // Parse action\n  let action: any = null;\n  if (actionType && actionType.toLowerCase() !== 'null') {\n    const type = actionType.trim();\n    let param: any = undefined;\n\n    if (actionParamStr) {\n      try {\n        // Parse the JSON string in action-param-json\n        param = safeParseJson(actionParamStr, modelFamily);\n      } catch (e) {\n        throw new Error(`Failed to parse action-param-json: ${e}`);\n      }\n    }\n\n    action = {\n      type,\n      ...(param !== undefined ? { param } : {}),\n    };\n  }\n\n  return {\n    ...(thought ? { thought } : {}),\n    ...(memory ? { memory } : {}),\n    log,\n    ...(error ? { error } : {}),\n    action,\n    ...(finalizeMessage !== undefined ? { finalizeMessage } : {}),\n    ...(finalizeSuccess !== undefined ? { finalizeSuccess } : {}),\n  };\n}\n\nexport async function plan(\n  userInstruction: string,\n  opts: {\n    context: UIContext;\n    interfaceType: InterfaceType;\n    actionSpace: DeviceAction<any>[];\n    actionContext?: string;\n    modelConfig: IModelConfig;\n    conversationHistory: ConversationHistory;\n    includeBbox: boolean;\n    imagesIncludeCount?: number;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<PlanningAIResponse> {\n  const { context, modelConfig, conversationHistory } = opts;\n  const { size } = context;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const { modelFamily } = modelConfig;\n\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    modelFamily,\n    includeBbox: opts.includeBbox,\n    includeThought: true, // always include thought\n    deepThink: opts.deepThink === true,\n  });\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = size.width;\n  let imageHeight = size.height;\n  const rightLimit = imageWidth;\n  const bottomLimit = imageHeight;\n\n  // Process image based on VL mode requirements\n  if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const actionContext = opts.actionContext\n    ? `<high_priority_knowledge>${opts.actionContext}</high_priority_knowledge>\\n`\n    : '';\n\n  const instruction: ChatCompletionMessageParam[] = [\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${actionContext}<user_instruction>${userInstruction}</user_instruction>`,\n        },\n      ],\n    },\n  ];\n\n  let latestFeedbackMessage: ChatCompletionMessageParam;\n\n  // Build historical execution logs text to include in the message\n  const historicalLogsText = conversationHistory.historicalLogsToText();\n  const historicalLogsSection = historicalLogsText\n    ? `\\n\\n${historicalLogsText}`\n    : '';\n\n  // Build memories text to include in the message\n  const memoriesText = conversationHistory.memoriesToText();\n  const memoriesSection = memoriesText ? `\\n\\n${memoriesText}` : '';\n\n  if (conversationHistory.pendingFeedbackMessage) {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${conversationHistory.pendingFeedbackMessage}. The previous action has been executed, here is the latest screenshot. Please continue according to the instruction.${memoriesSection}${historicalLogsSection}`,\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n\n    conversationHistory.resetPendingFeedbackMessageIfExists();\n  } else {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `this is the latest screenshot${memoriesSection}${historicalLogsSection}`,\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n  }\n  conversationHistory.append(latestFeedbackMessage);\n\n  // Compress history if it exceeds the threshold to avoid context overflow\n  conversationHistory.compressHistory(50, 20);\n\n  const historyLog = conversationHistory.snapshot(opts.imagesIncludeCount);\n\n  const msgs: ChatCompletionMessageParam[] = [\n    { role: 'system', content: systemPrompt },\n    ...instruction,\n    ...historyLog,\n  ];\n\n  const {\n    content: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAI(msgs, modelConfig, {\n    deepThink: opts.deepThink === 'unset' ? undefined : opts.deepThink,\n  });\n\n  // Parse XML response to JSON object, capture parsing errors\n  let planFromAI: RawResponsePlanningAIResponse;\n  try {\n    planFromAI = parseXMLPlanningResponse(rawResponse, modelFamily);\n\n    if (planFromAI.action && planFromAI.finalizeSuccess !== undefined) {\n      warnLog(\n        'Planning response included both an action and <complete>; ignoring <complete> output.',\n      );\n      planFromAI.finalizeMessage = undefined;\n      planFromAI.finalizeSuccess = undefined;\n    }\n\n    const actions = planFromAI.action ? [planFromAI.action] : [];\n    let shouldContinuePlanning = true;\n\n    // Check if task is completed via <complete> tag\n    if (planFromAI.finalizeSuccess !== undefined) {\n      debug('task completed via <complete> tag, stop planning');\n      shouldContinuePlanning = false;\n    }\n\n    const returnValue: PlanningAIResponse = {\n      ...planFromAI,\n      actions,\n      rawResponse,\n      usage,\n      reasoning_content,\n      yamlFlow: buildYamlFlowFromPlans(actions, opts.actionSpace),\n      shouldContinuePlanning,\n    };\n\n    assert(planFromAI, \"can't get plans from AI\");\n\n    actions.forEach((action) => {\n      const type = action.type;\n      const actionInActionSpace = opts.actionSpace.find(\n        (action) => action.name === type,\n      );\n\n      debug('actionInActionSpace matched', actionInActionSpace);\n      const locateFields = actionInActionSpace\n        ? findAllMidsceneLocatorField(actionInActionSpace.paramSchema)\n        : [];\n\n      debug('locateFields', locateFields);\n\n      locateFields.forEach((field) => {\n        const locateResult = action.param[field];\n        if (locateResult && modelFamily !== undefined) {\n          // Always use model family to fill bbox parameters\n          action.param[field] = fillBboxParam(\n            locateResult,\n            imageWidth,\n            imageHeight,\n            modelFamily,\n          );\n        }\n      });\n    });\n\n    // Accumulate logs as historical execution steps\n    if (planFromAI.log) {\n      conversationHistory.appendHistoricalLog(planFromAI.log);\n    }\n\n    // Append memory to conversation history if present\n    if (planFromAI.memory) {\n      conversationHistory.appendMemory(planFromAI.memory);\n    }\n\n    conversationHistory.append({\n      role: 'assistant',\n      content: [\n        {\n          type: 'text',\n          text: rawResponse,\n        },\n      ],\n    });\n\n    return returnValue;\n  } catch (parseError) {\n    // Throw AIResponseParseError with usage and rawResponse preserved\n    const errorMessage =\n      parseError instanceof Error ? parseError.message : String(parseError);\n    throw new AIResponseParseError(\n      `XML parse error: ${errorMessage}`,\n      rawResponse,\n      usage,\n    );\n  }\n}\n"],"names":["debug","getDebug","warnLog","parseXMLPlanningResponse","xmlString","modelFamily","thought","extractXMLTag","memory","log","error","actionType","actionParamStr","completeGoalRegex","completeGoalMatch","finalizeMessage","finalizeSuccess","undefined","action","type","param","safeParseJson","e","Error","plan","userInstruction","opts","context","modelConfig","conversationHistory","size","screenshotBase64","systemPrompt","systemPromptToTaskPlanning","imagePayload","imageWidth","imageHeight","paddedResult","paddingToMatchBlockByBase64","actionContext","instruction","latestFeedbackMessage","historicalLogsText","historicalLogsSection","memoriesText","memoriesSection","historyLog","msgs","rawResponse","usage","reasoning_content","callAI","planFromAI","actions","shouldContinuePlanning","returnValue","buildYamlFlowFromPlans","assert","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam","parseError","errorMessage","String","AIResponseParseError"],"mappings":";;;;;;;AA2BA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,UAAUD,SAAS,YAAY;IAAE,SAAS;AAAK;AAK9C,SAASE,yBACdC,SAAiB,EACjBC,WAAqC;IAErC,MAAMC,UAAUC,cAAcH,WAAW;IACzC,MAAMI,SAASD,cAAcH,WAAW;IACxC,MAAMK,MAAMF,cAAcH,WAAW,UAAU;IAC/C,MAAMM,QAAQH,cAAcH,WAAW;IACvC,MAAMO,aAAaJ,cAAcH,WAAW;IAC5C,MAAMQ,iBAAiBL,cAAcH,WAAW;IAGhD,MAAMS,oBACJ;IACF,MAAMC,oBAAoBV,UAAU,KAAK,CAACS;IAC1C,IAAIE;IACJ,IAAIC;IAEJ,IAAIF,mBAAmB;QACrBE,kBAAkBF,AAAyB,WAAzBA,iBAAiB,CAAC,EAAE;QACtCC,kBAAkBD,iBAAiB,CAAC,EAAE,EAAE,UAAUG;IACpD;IAGA,IAAIC,SAAc;IAClB,IAAIP,cAAcA,AAA6B,WAA7BA,WAAW,WAAW,IAAe;QACrD,MAAMQ,OAAOR,WAAW,IAAI;QAC5B,IAAIS;QAEJ,IAAIR,gBACF,IAAI;YAEFQ,QAAQC,cAAcT,gBAAgBP;QACxC,EAAE,OAAOiB,GAAG;YACV,MAAM,IAAIC,MAAM,CAAC,mCAAmC,EAAED,GAAG;QAC3D;QAGFJ,SAAS;YACPC;YACA,GAAIC,AAAUH,WAAVG,QAAsB;gBAAEA;YAAM,IAAI,CAAC,CAAC;QAC1C;IACF;IAEA,OAAO;QACL,GAAId,UAAU;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAC9B,GAAIE,SAAS;YAAEA;QAAO,IAAI,CAAC,CAAC;QAC5BC;QACA,GAAIC,QAAQ;YAAEA;QAAM,IAAI,CAAC,CAAC;QAC1BQ;QACA,GAAIH,AAAoBE,WAApBF,kBAAgC;YAAEA;QAAgB,IAAI,CAAC,CAAC;QAC5D,GAAIC,AAAoBC,WAApBD,kBAAgC;YAAEA;QAAgB,IAAI,CAAC,CAAC;IAC9D;AACF;AAEO,eAAeQ,KACpBC,eAAuB,EACvBC,IAUC;IAED,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAEC,mBAAmB,EAAE,GAAGH;IACtD,MAAM,EAAEI,IAAI,EAAE,GAAGH;IACjB,MAAMI,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM,EAAEtB,WAAW,EAAE,GAAGuB;IAExB,MAAMI,eAAe,MAAMC,2BAA2B;QACpD,aAAaP,KAAK,WAAW;QAC7BrB;QACA,aAAaqB,KAAK,WAAW;QAC7B,gBAAgB;QAChB,WAAWA,AAAmB,SAAnBA,KAAK,SAAS;IAC3B;IAEA,IAAIQ,eAAeH;IACnB,IAAII,aAAaL,KAAK,KAAK;IAC3B,IAAIM,cAAcN,KAAK,MAAM;IAK7B,IAAIzB,AAAgB,iBAAhBA,aAA8B;QAChC,MAAMgC,eAAe,MAAMC,4BAA4BJ;QACvDC,aAAaE,aAAa,KAAK;QAC/BD,cAAcC,aAAa,MAAM;QACjCH,eAAeG,aAAa,WAAW;IACzC;IAEA,MAAME,gBAAgBb,KAAK,aAAa,GACpC,CAAC,yBAAyB,EAAEA,KAAK,aAAa,CAAC,4BAA4B,CAAC,GAC5E;IAEJ,MAAMc,cAA4C;QAChD;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGD,cAAc,kBAAkB,EAAEd,gBAAgB,mBAAmB,CAAC;gBACjF;aACD;QACH;KACD;IAED,IAAIgB;IAGJ,MAAMC,qBAAqBb,oBAAoB,oBAAoB;IACnE,MAAMc,wBAAwBD,qBAC1B,CAAC,IAAI,EAAEA,oBAAoB,GAC3B;IAGJ,MAAME,eAAef,oBAAoB,cAAc;IACvD,MAAMgB,kBAAkBD,eAAe,CAAC,IAAI,EAAEA,cAAc,GAAG;IAE/D,IAAIf,oBAAoB,sBAAsB,EAAE;QAC9CY,wBAAwB;YACtB,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGZ,oBAAoB,sBAAsB,CAAC,qHAAqH,EAAEgB,kBAAkBF,uBAAuB;gBACtN;gBACA;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKT;wBACL,QAAQ;oBACV;gBACF;aACD;QACH;QAEAL,oBAAoB,mCAAmC;IACzD,OACEY,wBAAwB;QACtB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAM,CAAC,6BAA6B,EAAEI,kBAAkBF,uBAAuB;YACjF;YACA;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKT;oBACL,QAAQ;gBACV;YACF;SACD;IACH;IAEFL,oBAAoB,MAAM,CAACY;IAG3BZ,oBAAoB,eAAe,CAAC,IAAI;IAExC,MAAMiB,aAAajB,oBAAoB,QAAQ,CAACH,KAAK,kBAAkB;IAEvE,MAAMqB,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASf;QAAa;WACrCQ;WACAM;KACJ;IAED,MAAM,EACJ,SAASE,WAAW,EACpBC,KAAK,EACLC,iBAAiB,EAClB,GAAG,MAAMC,OAAOJ,MAAMnB,aAAa;QAClC,WAAWF,AAAmB,YAAnBA,KAAK,SAAS,GAAeT,SAAYS,KAAK,SAAS;IACpE;IAGA,IAAI0B;IACJ,IAAI;QACFA,aAAajD,yBAAyB6C,aAAa3C;QAEnD,IAAI+C,WAAW,MAAM,IAAIA,AAA+BnC,WAA/BmC,WAAW,eAAe,EAAgB;YACjElD,QACE;YAEFkD,WAAW,eAAe,GAAGnC;YAC7BmC,WAAW,eAAe,GAAGnC;QAC/B;QAEA,MAAMoC,UAAUD,WAAW,MAAM,GAAG;YAACA,WAAW,MAAM;SAAC,GAAG,EAAE;QAC5D,IAAIE,yBAAyB;QAG7B,IAAIF,AAA+BnC,WAA/BmC,WAAW,eAAe,EAAgB;YAC5CpD,MAAM;YACNsD,yBAAyB;QAC3B;QAEA,MAAMC,cAAkC;YACtC,GAAGH,UAAU;YACbC;YACAL;YACAC;YACAC;YACA,UAAUM,uBAAuBH,SAAS3B,KAAK,WAAW;YAC1D4B;QACF;QAEAG,OAAOL,YAAY;QAEnBC,QAAQ,OAAO,CAAC,CAACnC;YACf,MAAMC,OAAOD,OAAO,IAAI;YACxB,MAAMwC,sBAAsBhC,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACR,SAAWA,OAAO,IAAI,KAAKC;YAG9BnB,MAAM,+BAA+B0D;YACrC,MAAMC,eAAeD,sBACjBE,4BAA4BF,oBAAoB,WAAW,IAC3D,EAAE;YAEN1D,MAAM,gBAAgB2D;YAEtBA,aAAa,OAAO,CAAC,CAACE;gBACpB,MAAMC,eAAe5C,OAAO,KAAK,CAAC2C,MAAM;gBACxC,IAAIC,gBAAgBzD,AAAgBY,WAAhBZ,aAElBa,OAAO,KAAK,CAAC2C,MAAM,GAAGE,cACpBD,cACA3B,YACAC,aACA/B;YAGN;QACF;QAGA,IAAI+C,WAAW,GAAG,EAChBvB,oBAAoB,mBAAmB,CAACuB,WAAW,GAAG;QAIxD,IAAIA,WAAW,MAAM,EACnBvB,oBAAoB,YAAY,CAACuB,WAAW,MAAM;QAGpDvB,oBAAoB,MAAM,CAAC;YACzB,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAMmB;gBACR;aACD;QACH;QAEA,OAAOO;IACT,EAAE,OAAOS,YAAY;QAEnB,MAAMC,eACJD,sBAAsBzC,QAAQyC,WAAW,OAAO,GAAGE,OAAOF;QAC5D,MAAM,IAAIG,qBACR,CAAC,iBAAiB,EAAEF,cAAc,EAClCjB,aACAC;IAEJ;AACF"}
+{"version":3,"file":"ai-model/llm-planning.mjs","sources":["../../../src/ai-model/llm-planning.ts"],"sourcesContent":["import type {\n  DeepThinkOption,\n  DeviceAction,\n  InterfaceType,\n  PlanningAIResponse,\n  RawResponsePlanningAIResponse,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig, TModelFamily } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport {\n  buildYamlFlowFromPlans,\n  fillBboxParam,\n  findAllMidsceneLocatorField,\n} from '../common';\nimport type { ConversationHistory } from './conversation-history';\nimport { systemPromptToTaskPlanning } from './prompt/llm-planning';\nimport { extractXMLTag } from './prompt/util';\nimport {\n  AIResponseParseError,\n  callAI,\n  safeParseJson,\n} from './service-caller/index';\n\nconst debug = getDebug('planning');\nconst warnLog = getDebug('planning', { console: true });\n\n/**\n * Parse XML response from LLM and convert to RawResponsePlanningAIResponse\n */\nexport function parseXMLPlanningResponse(\n  xmlString: string,\n  modelFamily: TModelFamily | undefined,\n): RawResponsePlanningAIResponse {\n  const thought = extractXMLTag(xmlString, 'thought');\n  const memory = extractXMLTag(xmlString, 'memory');\n  const log = extractXMLTag(xmlString, 'log') || '';\n  const error = extractXMLTag(xmlString, 'error');\n  const actionType = extractXMLTag(xmlString, 'action-type');\n  const actionParamStr = extractXMLTag(xmlString, 'action-param-json');\n\n  // Parse <complete> tag with success attribute\n  const completeGoalRegex =\n    /<complete\\s+success=\"(true|false)\">([\\s\\S]*?)<\\/complete>/i;\n  const completeGoalMatch = xmlString.match(completeGoalRegex);\n  let finalizeMessage: string | undefined;\n  let finalizeSuccess: boolean | undefined;\n\n  if (completeGoalMatch) {\n    finalizeSuccess = completeGoalMatch[1] === 'true';\n    finalizeMessage = completeGoalMatch[2]?.trim() || undefined;\n  }\n\n  // Parse action\n  let action: any = null;\n  if (actionType && actionType.toLowerCase() !== 'null') {\n    const type = actionType.trim();\n    let param: any = undefined;\n\n    if (actionParamStr) {\n      try {\n        // Parse the JSON string in action-param-json\n        param = safeParseJson(actionParamStr, modelFamily);\n      } catch (e) {\n        throw new Error(`Failed to parse action-param-json: ${e}`);\n      }\n    }\n\n    action = {\n      type,\n      ...(param !== undefined ? { param } : {}),\n    };\n  }\n\n  return {\n    ...(thought ? { thought } : {}),\n    ...(memory ? { memory } : {}),\n    log,\n    ...(error ? { error } : {}),\n    action,\n    ...(finalizeMessage !== undefined ? { finalizeMessage } : {}),\n    ...(finalizeSuccess !== undefined ? { finalizeSuccess } : {}),\n  };\n}\n\nexport async function plan(\n  userInstruction: string,\n  opts: {\n    context: UIContext;\n    interfaceType: InterfaceType;\n    actionSpace: DeviceAction<any>[];\n    actionContext?: string;\n    modelConfig: IModelConfig;\n    conversationHistory: ConversationHistory;\n    includeBbox: boolean;\n    imagesIncludeCount?: number;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<PlanningAIResponse> {\n  const { context, modelConfig, conversationHistory } = opts;\n  const { shotSize } = context;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const { modelFamily } = modelConfig;\n\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    modelFamily,\n    includeBbox: opts.includeBbox,\n    includeThought: true, // always include thought\n    deepThink: opts.deepThink === true,\n  });\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = shotSize.width;\n  let imageHeight = shotSize.height;\n  const rightLimit = imageWidth;\n  const bottomLimit = imageHeight;\n\n  // Process image based on VL mode requirements\n  if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const actionContext = opts.actionContext\n    ? `<high_priority_knowledge>${opts.actionContext}</high_priority_knowledge>\\n`\n    : '';\n\n  const instruction: ChatCompletionMessageParam[] = [\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${actionContext}<user_instruction>${userInstruction}</user_instruction>`,\n        },\n      ],\n    },\n  ];\n\n  let latestFeedbackMessage: ChatCompletionMessageParam;\n\n  // Build historical execution logs text to include in the message\n  const historicalLogsText = conversationHistory.historicalLogsToText();\n  const historicalLogsSection = historicalLogsText\n    ? `\\n\\n${historicalLogsText}`\n    : '';\n\n  // Build memories text to include in the message\n  const memoriesText = conversationHistory.memoriesToText();\n  const memoriesSection = memoriesText ? `\\n\\n${memoriesText}` : '';\n\n  if (conversationHistory.pendingFeedbackMessage) {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${conversationHistory.pendingFeedbackMessage}. The previous action has been executed, here is the latest screenshot. Please continue according to the instruction.${memoriesSection}${historicalLogsSection}`,\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n\n    conversationHistory.resetPendingFeedbackMessageIfExists();\n  } else {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `this is the latest screenshot${memoriesSection}${historicalLogsSection}`,\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n  }\n  conversationHistory.append(latestFeedbackMessage);\n\n  // Compress history if it exceeds the threshold to avoid context overflow\n  conversationHistory.compressHistory(50, 20);\n\n  const historyLog = conversationHistory.snapshot(opts.imagesIncludeCount);\n\n  const msgs: ChatCompletionMessageParam[] = [\n    { role: 'system', content: systemPrompt },\n    ...instruction,\n    ...historyLog,\n  ];\n\n  const {\n    content: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAI(msgs, modelConfig, {\n    deepThink: opts.deepThink === 'unset' ? undefined : opts.deepThink,\n  });\n\n  // Parse XML response to JSON object, capture parsing errors\n  let planFromAI: RawResponsePlanningAIResponse;\n  try {\n    planFromAI = parseXMLPlanningResponse(rawResponse, modelFamily);\n\n    if (planFromAI.action && planFromAI.finalizeSuccess !== undefined) {\n      warnLog(\n        'Planning response included both an action and <complete>; ignoring <complete> output.',\n      );\n      planFromAI.finalizeMessage = undefined;\n      planFromAI.finalizeSuccess = undefined;\n    }\n\n    const actions = planFromAI.action ? [planFromAI.action] : [];\n    let shouldContinuePlanning = true;\n\n    // Check if task is completed via <complete> tag\n    if (planFromAI.finalizeSuccess !== undefined) {\n      debug('task completed via <complete> tag, stop planning');\n      shouldContinuePlanning = false;\n    }\n\n    const returnValue: PlanningAIResponse = {\n      ...planFromAI,\n      actions,\n      rawResponse,\n      usage,\n      reasoning_content,\n      yamlFlow: buildYamlFlowFromPlans(actions, opts.actionSpace),\n      shouldContinuePlanning,\n    };\n\n    assert(planFromAI, \"can't get plans from AI\");\n\n    actions.forEach((action) => {\n      const type = action.type;\n      const actionInActionSpace = opts.actionSpace.find(\n        (action) => action.name === type,\n      );\n\n      debug('actionInActionSpace matched', actionInActionSpace);\n      const locateFields = actionInActionSpace\n        ? findAllMidsceneLocatorField(actionInActionSpace.paramSchema)\n        : [];\n\n      debug('locateFields', locateFields);\n\n      locateFields.forEach((field) => {\n        const locateResult = action.param[field];\n        if (locateResult && modelFamily !== undefined) {\n          // Always use model family to fill bbox parameters\n          action.param[field] = fillBboxParam(\n            locateResult,\n            imageWidth,\n            imageHeight,\n            modelFamily,\n          );\n        }\n      });\n    });\n\n    // Accumulate logs as historical execution steps\n    if (planFromAI.log) {\n      conversationHistory.appendHistoricalLog(planFromAI.log);\n    }\n\n    // Append memory to conversation history if present\n    if (planFromAI.memory) {\n      conversationHistory.appendMemory(planFromAI.memory);\n    }\n\n    conversationHistory.append({\n      role: 'assistant',\n      content: [\n        {\n          type: 'text',\n          text: rawResponse,\n        },\n      ],\n    });\n\n    return returnValue;\n  } catch (parseError) {\n    // Throw AIResponseParseError with usage and rawResponse preserved\n    const errorMessage =\n      parseError instanceof Error ? parseError.message : String(parseError);\n    throw new AIResponseParseError(\n      `XML parse error: ${errorMessage}`,\n      rawResponse,\n      usage,\n    );\n  }\n}\n"],"names":["debug","getDebug","warnLog","parseXMLPlanningResponse","xmlString","modelFamily","thought","extractXMLTag","memory","log","error","actionType","actionParamStr","completeGoalRegex","completeGoalMatch","finalizeMessage","finalizeSuccess","undefined","action","type","param","safeParseJson","e","Error","plan","userInstruction","opts","context","modelConfig","conversationHistory","shotSize","screenshotBase64","systemPrompt","systemPromptToTaskPlanning","imagePayload","imageWidth","imageHeight","paddedResult","paddingToMatchBlockByBase64","actionContext","instruction","latestFeedbackMessage","historicalLogsText","historicalLogsSection","memoriesText","memoriesSection","historyLog","msgs","rawResponse","usage","reasoning_content","callAI","planFromAI","actions","shouldContinuePlanning","returnValue","buildYamlFlowFromPlans","assert","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam","parseError","errorMessage","String","AIResponseParseError"],"mappings":";;;;;;;AA2BA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,UAAUD,SAAS,YAAY;IAAE,SAAS;AAAK;AAK9C,SAASE,yBACdC,SAAiB,EACjBC,WAAqC;IAErC,MAAMC,UAAUC,cAAcH,WAAW;IACzC,MAAMI,SAASD,cAAcH,WAAW;IACxC,MAAMK,MAAMF,cAAcH,WAAW,UAAU;IAC/C,MAAMM,QAAQH,cAAcH,WAAW;IACvC,MAAMO,aAAaJ,cAAcH,WAAW;IAC5C,MAAMQ,iBAAiBL,cAAcH,WAAW;IAGhD,MAAMS,oBACJ;IACF,MAAMC,oBAAoBV,UAAU,KAAK,CAACS;IAC1C,IAAIE;IACJ,IAAIC;IAEJ,IAAIF,mBAAmB;QACrBE,kBAAkBF,AAAyB,WAAzBA,iBAAiB,CAAC,EAAE;QACtCC,kBAAkBD,iBAAiB,CAAC,EAAE,EAAE,UAAUG;IACpD;IAGA,IAAIC,SAAc;IAClB,IAAIP,cAAcA,AAA6B,WAA7BA,WAAW,WAAW,IAAe;QACrD,MAAMQ,OAAOR,WAAW,IAAI;QAC5B,IAAIS;QAEJ,IAAIR,gBACF,IAAI;YAEFQ,QAAQC,cAAcT,gBAAgBP;QACxC,EAAE,OAAOiB,GAAG;YACV,MAAM,IAAIC,MAAM,CAAC,mCAAmC,EAAED,GAAG;QAC3D;QAGFJ,SAAS;YACPC;YACA,GAAIC,AAAUH,WAAVG,QAAsB;gBAAEA;YAAM,IAAI,CAAC,CAAC;QAC1C;IACF;IAEA,OAAO;QACL,GAAId,UAAU;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAC9B,GAAIE,SAAS;YAAEA;QAAO,IAAI,CAAC,CAAC;QAC5BC;QACA,GAAIC,QAAQ;YAAEA;QAAM,IAAI,CAAC,CAAC;QAC1BQ;QACA,GAAIH,AAAoBE,WAApBF,kBAAgC;YAAEA;QAAgB,IAAI,CAAC,CAAC;QAC5D,GAAIC,AAAoBC,WAApBD,kBAAgC;YAAEA;QAAgB,IAAI,CAAC,CAAC;IAC9D;AACF;AAEO,eAAeQ,KACpBC,eAAuB,EACvBC,IAUC;IAED,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAEC,mBAAmB,EAAE,GAAGH;IACtD,MAAM,EAAEI,QAAQ,EAAE,GAAGH;IACrB,MAAMI,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM,EAAEtB,WAAW,EAAE,GAAGuB;IAExB,MAAMI,eAAe,MAAMC,2BAA2B;QACpD,aAAaP,KAAK,WAAW;QAC7BrB;QACA,aAAaqB,KAAK,WAAW;QAC7B,gBAAgB;QAChB,WAAWA,AAAmB,SAAnBA,KAAK,SAAS;IAC3B;IAEA,IAAIQ,eAAeH;IACnB,IAAII,aAAaL,SAAS,KAAK;IAC/B,IAAIM,cAAcN,SAAS,MAAM;IAKjC,IAAIzB,AAAgB,iBAAhBA,aAA8B;QAChC,MAAMgC,eAAe,MAAMC,4BAA4BJ;QACvDC,aAAaE,aAAa,KAAK;QAC/BD,cAAcC,aAAa,MAAM;QACjCH,eAAeG,aAAa,WAAW;IACzC;IAEA,MAAME,gBAAgBb,KAAK,aAAa,GACpC,CAAC,yBAAyB,EAAEA,KAAK,aAAa,CAAC,4BAA4B,CAAC,GAC5E;IAEJ,MAAMc,cAA4C;QAChD;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGD,cAAc,kBAAkB,EAAEd,gBAAgB,mBAAmB,CAAC;gBACjF;aACD;QACH;KACD;IAED,IAAIgB;IAGJ,MAAMC,qBAAqBb,oBAAoB,oBAAoB;IACnE,MAAMc,wBAAwBD,qBAC1B,CAAC,IAAI,EAAEA,oBAAoB,GAC3B;IAGJ,MAAME,eAAef,oBAAoB,cAAc;IACvD,MAAMgB,kBAAkBD,eAAe,CAAC,IAAI,EAAEA,cAAc,GAAG;IAE/D,IAAIf,oBAAoB,sBAAsB,EAAE;QAC9CY,wBAAwB;YACtB,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGZ,oBAAoB,sBAAsB,CAAC,qHAAqH,EAAEgB,kBAAkBF,uBAAuB;gBACtN;gBACA;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKT;wBACL,QAAQ;oBACV;gBACF;aACD;QACH;QAEAL,oBAAoB,mCAAmC;IACzD,OACEY,wBAAwB;QACtB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAM,CAAC,6BAA6B,EAAEI,kBAAkBF,uBAAuB;YACjF;YACA;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKT;oBACL,QAAQ;gBACV;YACF;SACD;IACH;IAEFL,oBAAoB,MAAM,CAACY;IAG3BZ,oBAAoB,eAAe,CAAC,IAAI;IAExC,MAAMiB,aAAajB,oBAAoB,QAAQ,CAACH,KAAK,kBAAkB;IAEvE,MAAMqB,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASf;QAAa;WACrCQ;WACAM;KACJ;IAED,MAAM,EACJ,SAASE,WAAW,EACpBC,KAAK,EACLC,iBAAiB,EAClB,GAAG,MAAMC,OAAOJ,MAAMnB,aAAa;QAClC,WAAWF,AAAmB,YAAnBA,KAAK,SAAS,GAAeT,SAAYS,KAAK,SAAS;IACpE;IAGA,IAAI0B;IACJ,IAAI;QACFA,aAAajD,yBAAyB6C,aAAa3C;QAEnD,IAAI+C,WAAW,MAAM,IAAIA,AAA+BnC,WAA/BmC,WAAW,eAAe,EAAgB;YACjElD,QACE;YAEFkD,WAAW,eAAe,GAAGnC;YAC7BmC,WAAW,eAAe,GAAGnC;QAC/B;QAEA,MAAMoC,UAAUD,WAAW,MAAM,GAAG;YAACA,WAAW,MAAM;SAAC,GAAG,EAAE;QAC5D,IAAIE,yBAAyB;QAG7B,IAAIF,AAA+BnC,WAA/BmC,WAAW,eAAe,EAAgB;YAC5CpD,MAAM;YACNsD,yBAAyB;QAC3B;QAEA,MAAMC,cAAkC;YACtC,GAAGH,UAAU;YACbC;YACAL;YACAC;YACAC;YACA,UAAUM,uBAAuBH,SAAS3B,KAAK,WAAW;YAC1D4B;QACF;QAEAG,OAAOL,YAAY;QAEnBC,QAAQ,OAAO,CAAC,CAACnC;YACf,MAAMC,OAAOD,OAAO,IAAI;YACxB,MAAMwC,sBAAsBhC,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACR,SAAWA,OAAO,IAAI,KAAKC;YAG9BnB,MAAM,+BAA+B0D;YACrC,MAAMC,eAAeD,sBACjBE,4BAA4BF,oBAAoB,WAAW,IAC3D,EAAE;YAEN1D,MAAM,gBAAgB2D;YAEtBA,aAAa,OAAO,CAAC,CAACE;gBACpB,MAAMC,eAAe5C,OAAO,KAAK,CAAC2C,MAAM;gBACxC,IAAIC,gBAAgBzD,AAAgBY,WAAhBZ,aAElBa,OAAO,KAAK,CAAC2C,MAAM,GAAGE,cACpBD,cACA3B,YACAC,aACA/B;YAGN;QACF;QAGA,IAAI+C,WAAW,GAAG,EAChBvB,oBAAoB,mBAAmB,CAACuB,WAAW,GAAG;QAIxD,IAAIA,WAAW,MAAM,EACnBvB,oBAAoB,YAAY,CAACuB,WAAW,MAAM;QAGpDvB,oBAAoB,MAAM,CAAC;YACzB,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAMmB;gBACR;aACD;QACH;QAEA,OAAOO;IACT,EAAE,OAAOS,YAAY;QAEnB,MAAMC,eACJD,sBAAsBzC,QAAQyC,WAAW,OAAO,GAAGE,OAAOF;QAC5D,MAAM,IAAIG,qBACR,CAAC,iBAAiB,EAAEF,cAAc,EAClCjB,aACAC;IAEJ;AACF"}