@midscene/core 1.9.8 → 1.10.0

package/dist/es/element-describer.mjs

@@ -0,0 +1,180 @@
+import { assert } from "@midscene/shared/utils";
+import { createScreenshotBoundUIContext } from "./agent/utils.mjs";
+import { buildDetailedLocateParam } from "./yaml/utils.mjs";
+const distanceOfTwoPoints = (p1, p2)=>{
+    const [x1, y1] = p1;
+    const [x2, y2] = p2;
+    return Math.round(Math.sqrt((x1 - x2) ** 2 + (y1 - y2) ** 2));
+};
+const includedInRect = (point, rect)=>{
+    const [x, y] = point;
+    const { left, top, width, height } = rect;
+    return x >= left && x <= left + width && y >= top && y <= top + height;
+};
+const buildLocateValidatorResult = (expectCenter, located, verifyLocateOption)=>{
+    const distance = distanceOfTwoPoints(expectCenter, located.center);
+    const included = includedInRect(expectCenter, located.rect);
+    const pass = distance <= (verifyLocateOption?.centerDistanceThreshold || 20) || included;
+    return {
+        pass,
+        rect: located.rect,
+        center: located.center,
+        centerDistance: distance,
+        includedInRect: included
+    };
+};
+function assertPositiveSize(size, label) {
+    assert(size && Number.isFinite(size.width) && Number.isFinite(size.height) && size.width > 0 && size.height > 0, `${label} must include positive width and height`);
+}
+const mapPointToScreenshotSpace = (center, screenshotSize, opt)=>{
+    const coordinateSpace = opt.coordinateSpace || 'screenshot';
+    if ('screenshot' === coordinateSpace) return center;
+    assertPositiveSize(opt.logicalSize, 'logicalSize is required when coordinateSpace is logical');
+    return [
+        center[0] * screenshotSize.width / opt.logicalSize.width,
+        center[1] * screenshotSize.height / opt.logicalSize.height
+    ];
+};
+const createScreenshotBoundLocatorContext = async (center, opt)=>{
+    const screenshotContext = opt?.screenshotBase64 ? await createScreenshotBoundUIContext(opt.screenshotBase64, opt) : void 0;
+    const targetCenter = screenshotContext ? mapPointToScreenshotSpace(center, screenshotContext.shotSize, opt || {}) : center;
+    return {
+        screenshotContext,
+        locateOpt: screenshotContext ? {
+            uiContext: screenshotContext
+        } : void 0,
+        targetCenter
+    };
+};
+async function verifyLocator(runtime, prompt, locateOpt, expectCenter, verifyLocateOption) {
+    return locateAndVerify(runtime, prompt, expectCenter, {
+        centerDistanceThreshold: verifyLocateOption?.centerDistanceThreshold,
+        uiContext: locateOpt?.uiContext
+    });
+}
+function applyLocatorVerifyFn(locatorVerifyFn, input) {
+    if (!locatorVerifyFn) return input.verifyResult;
+    const customResult = locatorVerifyFn(input);
+    if ('boolean' == typeof customResult) return {
+        ...input.verifyResult,
+        pass: customResult
+    };
+    return customResult;
+}
+function errorMessage(error) {
+    return error instanceof Error ? error.message : String(error);
+}
+async function describeElementAtPoint(runtime, center, opt) {
+    const { verifyPrompt = true, retryLimit = 3 } = opt || {};
+    const { screenshotContext, locateOpt, targetCenter } = await createScreenshotBoundLocatorContext(center, opt);
+    let success = false;
+    let retryCount = 0;
+    let resultPrompt = '';
+    let deepLocate = opt?.deepLocate || false;
+    let verifyResult;
+    let lastError;
+    let failureStage;
+    while(!success && retryCount < retryLimit){
+        if (retryCount >= 2) deepLocate = true;
+        const describeOpt = screenshotContext ? {
+            deepLocate,
+            context: screenshotContext
+        } : {
+            deepLocate
+        };
+        let text;
+        try {
+            text = await runtime.service.describe(targetCenter, runtime.describeModelRuntime, describeOpt);
+        } catch (error) {
+            return {
+                prompt: resultPrompt,
+                deepLocate,
+                verifyResult,
+                success: false,
+                error: errorMessage(error),
+                failureStage: 'describe'
+            };
+        }
+        if (!text.description) return {
+            prompt: resultPrompt,
+            deepLocate,
+            verifyResult,
+            success: false,
+            error: `failed to describe element at [${targetCenter}]`,
+            failureStage: 'describe'
+        };
+        resultPrompt = text.description;
+        if (!verifyPrompt) {
+            opt?.onProgress?.({
+                prompt: resultPrompt,
+                deepLocate
+            });
+            success = true;
+            break;
+        }
+        try {
+            verifyResult = await verifyLocator(runtime, resultPrompt, locateOpt, targetCenter, opt);
+            verifyResult = applyLocatorVerifyFn(opt?.locatorVerifyFn, {
+                prompt: resultPrompt,
+                expectCenter: targetCenter,
+                deepLocate,
+                retryCount,
+                verifyResult
+            });
+            opt?.onProgress?.({
+                prompt: resultPrompt,
+                deepLocate,
+                verifyResult
+            });
+            if (verifyResult.pass) success = true;
+            else {
+                lastError = void 0;
+                failureStage = 'verify';
+                retryCount++;
+            }
+        } catch (error) {
+            lastError = errorMessage(error);
+            failureStage = 'verify';
+            opt?.onProgress?.({
+                prompt: resultPrompt,
+                deepLocate
+            });
+            retryCount++;
+        }
+    }
+    return {
+        prompt: resultPrompt,
+        deepLocate,
+        verifyResult,
+        success,
+        error: success || !verifyPrompt ? void 0 : lastError || 'describeElementAtPoint verify failed',
+        failureStage: success ? void 0 : failureStage
+    };
+}
+async function verifyElementDescriptionAtPoint(runtime, description, center, opt) {
+    assert(description?.trim(), "description must not be empty");
+    const { locateOpt, targetCenter } = await createScreenshotBoundLocatorContext(center, opt);
+    return verifyLocator(runtime, description, locateOpt, targetCenter, opt);
+}
+async function verifyElementByServiceLocate(runtime, description, center, opt) {
+    return locateAndVerify(runtime, description, center, opt);
+}
+async function locateAndVerify(runtime, description, center, opt) {
+    assert(description?.trim(), "description must not be empty");
+    const { screenshotContext, targetCenter } = await createScreenshotBoundLocatorContext(center, opt);
+    const context = opt?.uiContext || screenshotContext;
+    const locateParam = buildDetailedLocateParam(description, {
+        cacheable: opt?.cacheable,
+        deepLocate: opt?.deepLocate,
+        xpath: opt?.xpath
+    });
+    assert(locateParam, 'cannot get locate param for service locate');
+    const locateResult = await runtime.service.locate(locateParam, context ? {
+        context
+    } : {}, runtime.locateModelRuntime, opt?.abortSignal);
+    assert(locateResult.element, `Element not found: ${description}`);
+    return buildLocateValidatorResult(targetCenter, locateResult.element, opt);
+}
+export { describeElementAtPoint, verifyElementByServiceLocate, verifyElementDescriptionAtPoint, verifyLocator };
+
+//# sourceMappingURL=element-describer.mjs.map

package/dist/es/element-describer.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"element-describer.mjs","sources":["../../src/element-describer.ts"],"sourcesContent":["import type { ModelRuntime } from '@/ai-model/models';\nimport { assert } from '@midscene/shared/utils';\nimport { createScreenshotBoundUIContext } from './agent/utils';\nimport type Service from './service';\nimport type {\n  AgentDescribeElementAtPointResult,\n  LocateOption,\n  LocateResultElement,\n  LocateValidatorResult,\n  LocatorValidatorOption,\n  Rect,\n  Size,\n  UIContext,\n} from './types';\nimport { buildDetailedLocateParam } from './yaml/utils';\n\nexport type DescribeElementCoordinateSpace = 'screenshot' | 'logical';\n\nexport type LocatorVerifyFn = (input: {\n  prompt: string;\n  expectCenter: [number, number];\n  deepLocate: boolean;\n  retryCount: number;\n  verifyResult: LocateValidatorResult;\n}) => LocateValidatorResult | boolean;\n\nexport type DescribeElementAtPointOptions = {\n  verifyPrompt?: boolean;\n  retryLimit?: number;\n  deepLocate?: boolean;\n  locatorVerifyFn?: LocatorVerifyFn;\n  screenshotBase64?: string;\n  screenshotSize?: Size;\n  coordinateSpace?: DescribeElementCoordinateSpace;\n  logicalSize?: Size;\n  onProgress?: (progress: {\n    prompt?: string;\n    deepLocate?: boolean;\n    verifyResult?: LocateValidatorResult;\n  }) => void;\n} & LocatorValidatorOption;\n\ntype ScreenshotBoundContextOptions = {\n  screenshotBase64?: string;\n  screenshotSize?: Size;\n  coordinateSpace?: DescribeElementCoordinateSpace;\n  logicalSize?: Size;\n};\n\nexport type VerifyElementDescriptionAtPointOptions =\n  ScreenshotBoundContextOptions & LocatorValidatorOption;\n\nexport type VerifyElementByServiceLocateOptions =\n  VerifyElementDescriptionAtPointOptions &\n    Pick<LocateOption, 'cacheable' | 'deepLocate' | 'xpath'> & {\n      abortSignal?: AbortSignal;\n    };\n\nexport type ElementDescriberRuntime = {\n  service: Pick<Service, 'describe' | 'locate'>;\n  describeModelRuntime: ModelRuntime;\n  locateModelRuntime: ModelRuntime;\n};\n\nconst distanceOfTwoPoints = (p1: [number, number], p2: [number, number]) => {\n  const [x1, y1] = p1;\n  const [x2, y2] = p2;\n  return Math.round(Math.sqrt((x1 - x2) ** 2 + (y1 - y2) ** 2));\n};\n\nconst includedInRect = (point: [number, number], rect: Rect) => {\n  const [x, y] = point;\n  const { left, top, width, height } = rect;\n  return x >= left && x <= left + width && y >= top && y <= top + height;\n};\n\nconst buildLocateValidatorResult = (\n  expectCenter: [number, number],\n  located: Pick<LocateResultElement, 'center' | 'rect'>,\n  verifyLocateOption?: LocatorValidatorOption,\n): LocateValidatorResult => {\n  const distance = distanceOfTwoPoints(expectCenter, located.center);\n  const included = includedInRect(expectCenter, located.rect);\n  const pass =\n    distance <= (verifyLocateOption?.centerDistanceThreshold || 20) || included;\n  return {\n    pass,\n    rect: located.rect,\n    center: located.center,\n    centerDistance: distance,\n    includedInRect: included,\n  };\n};\n\nfunction assertPositiveSize(\n  size: Size | undefined,\n  label: string,\n): asserts size is Size {\n  assert(\n    size &&\n      Number.isFinite(size.width) &&\n      Number.isFinite(size.height) &&\n      size.width > 0 &&\n      size.height > 0,\n    `${label} must include positive width and height`,\n  );\n}\n\nconst mapPointToScreenshotSpace = (\n  center: [number, number],\n  screenshotSize: Size,\n  opt: ScreenshotBoundContextOptions,\n): [number, number] => {\n  const coordinateSpace = opt.coordinateSpace || 'screenshot';\n  if (coordinateSpace === 'screenshot') {\n    return center;\n  }\n\n  assertPositiveSize(\n    opt.logicalSize,\n    'logicalSize is required when coordinateSpace is logical',\n  );\n  return [\n    (center[0] * screenshotSize.width) / opt.logicalSize.width,\n    (center[1] * screenshotSize.height) / opt.logicalSize.height,\n  ];\n};\n\nconst createScreenshotBoundLocatorContext = async (\n  center: [number, number],\n  opt?: ScreenshotBoundContextOptions,\n): Promise<{\n  screenshotContext?: UIContext;\n  locateOpt?: LocateOption;\n  targetCenter: [number, number];\n}> => {\n  const screenshotContext = opt?.screenshotBase64\n    ? await createScreenshotBoundUIContext(opt.screenshotBase64, opt)\n    : undefined;\n  const targetCenter = screenshotContext\n    ? mapPointToScreenshotSpace(center, screenshotContext.shotSize, opt || {})\n    : center;\n  return {\n    screenshotContext,\n    locateOpt: screenshotContext ? { uiContext: screenshotContext } : undefined,\n    targetCenter,\n  };\n};\n\nexport async function verifyLocator(\n  runtime: Pick<ElementDescriberRuntime, 'service' | 'locateModelRuntime'>,\n  prompt: string,\n  locateOpt: LocateOption | undefined,\n  expectCenter: [number, number],\n  verifyLocateOption?: LocatorValidatorOption,\n): Promise<LocateValidatorResult> {\n  return locateAndVerify(runtime, prompt, expectCenter, {\n    centerDistanceThreshold: verifyLocateOption?.centerDistanceThreshold,\n    uiContext: locateOpt?.uiContext,\n  });\n}\n\nfunction applyLocatorVerifyFn(\n  locatorVerifyFn: LocatorVerifyFn | undefined,\n  input: Parameters<LocatorVerifyFn>[0],\n): LocateValidatorResult {\n  if (!locatorVerifyFn) {\n    return input.verifyResult;\n  }\n  const customResult = locatorVerifyFn(input);\n  if (typeof customResult === 'boolean') {\n    return {\n      ...input.verifyResult,\n      pass: customResult,\n    };\n  }\n  return customResult;\n}\n\nfunction errorMessage(error: unknown): string {\n  return error instanceof Error ? error.message : String(error);\n}\n\nexport async function describeElementAtPoint(\n  runtime: ElementDescriberRuntime,\n  center: [number, number],\n  opt?: DescribeElementAtPointOptions,\n): Promise<AgentDescribeElementAtPointResult> {\n  const { verifyPrompt = true, retryLimit = 3 } = opt || {};\n  const { screenshotContext, locateOpt, targetCenter } =\n    await createScreenshotBoundLocatorContext(center, opt);\n\n  let success = false;\n  let retryCount = 0;\n  let resultPrompt = '';\n  let deepLocate = opt?.deepLocate || false;\n  let verifyResult: LocateValidatorResult | undefined;\n  let lastError: string | undefined;\n  let failureStage: AgentDescribeElementAtPointResult['failureStage'];\n\n  while (!success && retryCount < retryLimit) {\n    if (retryCount >= 2) {\n      deepLocate = true;\n    }\n    const describeOpt = screenshotContext\n      ? { deepLocate, context: screenshotContext }\n      : { deepLocate };\n\n    let text: Awaited<\n      ReturnType<ElementDescriberRuntime['service']['describe']>\n    >;\n    try {\n      text = await runtime.service.describe(\n        targetCenter,\n        runtime.describeModelRuntime,\n        describeOpt,\n      );\n    } catch (error) {\n      return {\n        prompt: resultPrompt,\n        deepLocate,\n        verifyResult,\n        success: false,\n        error: errorMessage(error),\n        failureStage: 'describe',\n      };\n    }\n    if (!text.description) {\n      return {\n        prompt: resultPrompt,\n        deepLocate,\n        verifyResult,\n        success: false,\n        error: `failed to describe element at [${targetCenter}]`,\n        failureStage: 'describe',\n      };\n    }\n    resultPrompt = text.description;\n\n    if (!verifyPrompt) {\n      opt?.onProgress?.({ prompt: resultPrompt, deepLocate });\n      success = true;\n      break;\n    }\n\n    try {\n      verifyResult = await verifyLocator(\n        runtime,\n        resultPrompt,\n        locateOpt,\n        targetCenter,\n        opt,\n      );\n      verifyResult = applyLocatorVerifyFn(opt?.locatorVerifyFn, {\n        prompt: resultPrompt,\n        expectCenter: targetCenter,\n        deepLocate,\n        retryCount,\n        verifyResult,\n      });\n      opt?.onProgress?.({ prompt: resultPrompt, deepLocate, verifyResult });\n      if (verifyResult.pass) {\n        success = true;\n      } else {\n        lastError = undefined;\n        failureStage = 'verify';\n        retryCount++;\n      }\n    } catch (error) {\n      lastError = errorMessage(error);\n      failureStage = 'verify';\n      opt?.onProgress?.({ prompt: resultPrompt, deepLocate });\n      retryCount++;\n    }\n  }\n\n  return {\n    prompt: resultPrompt,\n    deepLocate,\n    verifyResult,\n    success,\n    error:\n      success || !verifyPrompt\n        ? undefined\n        : lastError || 'describeElementAtPoint verify failed',\n    failureStage: success ? undefined : failureStage,\n  };\n}\n\nexport async function verifyElementDescriptionAtPoint(\n  runtime: ElementDescriberRuntime,\n  description: string,\n  center: [number, number],\n  opt?: VerifyElementDescriptionAtPointOptions,\n): Promise<LocateValidatorResult> {\n  assert(description?.trim(), 'description must not be empty');\n  const { locateOpt, targetCenter } = await createScreenshotBoundLocatorContext(\n    center,\n    opt,\n  );\n\n  return verifyLocator(runtime, description, locateOpt, targetCenter, opt);\n}\n\nexport async function verifyElementByServiceLocate(\n  runtime: Pick<ElementDescriberRuntime, 'service' | 'locateModelRuntime'>,\n  description: string,\n  center: [number, number],\n  opt?: VerifyElementByServiceLocateOptions,\n): Promise<LocateValidatorResult> {\n  return locateAndVerify(runtime, description, center, opt);\n}\n\nasync function locateAndVerify(\n  runtime: Pick<ElementDescriberRuntime, 'service' | 'locateModelRuntime'>,\n  description: string,\n  center: [number, number],\n  opt?: VerifyElementByServiceLocateOptions & { uiContext?: UIContext },\n): Promise<LocateValidatorResult> {\n  assert(description?.trim(), 'description must not be empty');\n  const { screenshotContext, targetCenter } =\n    await createScreenshotBoundLocatorContext(center, opt);\n  const context = opt?.uiContext || screenshotContext;\n  const locateParam = buildDetailedLocateParam(description, {\n    cacheable: opt?.cacheable,\n    deepLocate: opt?.deepLocate,\n    xpath: opt?.xpath,\n  });\n  assert(locateParam, 'cannot get locate param for service locate');\n\n  const locateResult = await runtime.service.locate(\n    locateParam,\n    context ? { context } : {},\n    runtime.locateModelRuntime,\n    opt?.abortSignal,\n  );\n  assert(locateResult.element, `Element not found: ${description}`);\n  return buildLocateValidatorResult(targetCenter, locateResult.element, opt);\n}\n"],"names":["distanceOfTwoPoints","p1","p2","x1","y1","x2","y2","Math","includedInRect","point","rect","x","y","left","top","width","height","buildLocateValidatorResult","expectCenter","located","verifyLocateOption","distance","included","pass","assertPositiveSize","size","label","assert","Number","mapPointToScreenshotSpace","center","screenshotSize","opt","coordinateSpace","createScreenshotBoundLocatorContext","screenshotContext","createScreenshotBoundUIContext","undefined","targetCenter","verifyLocator","runtime","prompt","locateOpt","locateAndVerify","applyLocatorVerifyFn","locatorVerifyFn","input","customResult","errorMessage","error","Error","String","describeElementAtPoint","verifyPrompt","retryLimit","success","retryCount","resultPrompt","deepLocate","verifyResult","lastError","failureStage","describeOpt","text","verifyElementDescriptionAtPoint","description","verifyElementByServiceLocate","context","locateParam","buildDetailedLocateParam","locateResult"],"mappings":";;;AAgEA,MAAMA,sBAAsB,CAACC,IAAsBC;IACjD,MAAM,CAACC,IAAIC,GAAG,GAAGH;IACjB,MAAM,CAACI,IAAIC,GAAG,GAAGJ;IACjB,OAAOK,KAAK,KAAK,CAACA,KAAK,IAAI,CAAEJ,AAAAA,CAAAA,KAAKE,EAAC,KAAM,IAAKD,AAAAA,CAAAA,KAAKE,EAAC,KAAM;AAC5D;AAEA,MAAME,iBAAiB,CAACC,OAAyBC;IAC/C,MAAM,CAACC,GAAGC,EAAE,GAAGH;IACf,MAAM,EAAEI,IAAI,EAAEC,GAAG,EAAEC,KAAK,EAAEC,MAAM,EAAE,GAAGN;IACrC,OAAOC,KAAKE,QAAQF,KAAKE,OAAOE,SAASH,KAAKE,OAAOF,KAAKE,MAAME;AAClE;AAEA,MAAMC,6BAA6B,CACjCC,cACAC,SACAC;IAEA,MAAMC,WAAWrB,oBAAoBkB,cAAcC,QAAQ,MAAM;IACjE,MAAMG,WAAWd,eAAeU,cAAcC,QAAQ,IAAI;IAC1D,MAAMI,OACJF,YAAaD,CAAAA,oBAAoB,2BAA2B,EAAC,KAAME;IACrE,OAAO;QACLC;QACA,MAAMJ,QAAQ,IAAI;QAClB,QAAQA,QAAQ,MAAM;QACtB,gBAAgBE;QAChB,gBAAgBC;IAClB;AACF;AAEA,SAASE,mBACPC,IAAsB,EACtBC,KAAa;IAEbC,OACEF,QACEG,OAAO,QAAQ,CAACH,KAAK,KAAK,KAC1BG,OAAO,QAAQ,CAACH,KAAK,MAAM,KAC3BA,KAAK,KAAK,GAAG,KACbA,KAAK,MAAM,GAAG,GAChB,GAAGC,MAAM,uCAAuC,CAAC;AAErD;AAEA,MAAMG,4BAA4B,CAChCC,QACAC,gBACAC;IAEA,MAAMC,kBAAkBD,IAAI,eAAe,IAAI;IAC/C,IAAIC,AAAoB,iBAApBA,iBACF,OAAOH;IAGTN,mBACEQ,IAAI,WAAW,EACf;IAEF,OAAO;QACJF,MAAM,CAAC,EAAE,GAAGC,eAAe,KAAK,GAAIC,IAAI,WAAW,CAAC,KAAK;QACzDF,MAAM,CAAC,EAAE,GAAGC,eAAe,MAAM,GAAIC,IAAI,WAAW,CAAC,MAAM;KAC7D;AACH;AAEA,MAAME,sCAAsC,OAC1CJ,QACAE;IAMA,MAAMG,oBAAoBH,KAAK,mBAC3B,MAAMI,+BAA+BJ,IAAI,gBAAgB,EAAEA,OAC3DK;IACJ,MAAMC,eAAeH,oBACjBN,0BAA0BC,QAAQK,kBAAkB,QAAQ,EAAEH,OAAO,CAAC,KACtEF;IACJ,OAAO;QACLK;QACA,WAAWA,oBAAoB;YAAE,WAAWA;QAAkB,IAAIE;QAClEC;IACF;AACF;AAEO,eAAeC,cACpBC,OAAwE,EACxEC,MAAc,EACdC,SAAmC,EACnCxB,YAA8B,EAC9BE,kBAA2C;IAE3C,OAAOuB,gBAAgBH,SAASC,QAAQvB,cAAc;QACpD,yBAAyBE,oBAAoB;QAC7C,WAAWsB,WAAW;IACxB;AACF;AAEA,SAASE,qBACPC,eAA4C,EAC5CC,KAAqC;IAErC,IAAI,CAACD,iBACH,OAAOC,MAAM,YAAY;IAE3B,MAAMC,eAAeF,gBAAgBC;IACrC,IAAI,AAAwB,aAAxB,OAAOC,cACT,OAAO;QACL,GAAGD,MAAM,YAAY;QACrB,MAAMC;IACR;IAEF,OAAOA;AACT;AAEA,SAASC,aAAaC,KAAc;IAClC,OAAOA,iBAAiBC,QAAQD,MAAM,OAAO,GAAGE,OAAOF;AACzD;AAEO,eAAeG,uBACpBZ,OAAgC,EAChCV,MAAwB,EACxBE,GAAmC;IAEnC,MAAM,EAAEqB,eAAe,IAAI,EAAEC,aAAa,CAAC,EAAE,GAAGtB,OAAO,CAAC;IACxD,MAAM,EAAEG,iBAAiB,EAAEO,SAAS,EAAEJ,YAAY,EAAE,GAClD,MAAMJ,oCAAoCJ,QAAQE;IAEpD,IAAIuB,UAAU;IACd,IAAIC,aAAa;IACjB,IAAIC,eAAe;IACnB,IAAIC,aAAa1B,KAAK,cAAc;IACpC,IAAI2B;IACJ,IAAIC;IACJ,IAAIC;IAEJ,MAAO,CAACN,WAAWC,aAAaF,WAAY;QAC1C,IAAIE,cAAc,GAChBE,aAAa;QAEf,MAAMI,cAAc3B,oBAChB;YAAEuB;YAAY,SAASvB;QAAkB,IACzC;YAAEuB;QAAW;QAEjB,IAAIK;QAGJ,IAAI;YACFA,OAAO,MAAMvB,QAAQ,OAAO,CAAC,QAAQ,CACnCF,cACAE,QAAQ,oBAAoB,EAC5BsB;QAEJ,EAAE,OAAOb,OAAO;YACd,OAAO;gBACL,QAAQQ;gBACRC;gBACAC;gBACA,SAAS;gBACT,OAAOX,aAAaC;gBACpB,cAAc;YAChB;QACF;QACA,IAAI,CAACc,KAAK,WAAW,EACnB,OAAO;YACL,QAAQN;YACRC;YACAC;YACA,SAAS;YACT,OAAO,CAAC,+BAA+B,EAAErB,aAAa,CAAC,CAAC;YACxD,cAAc;QAChB;QAEFmB,eAAeM,KAAK,WAAW;QAE/B,IAAI,CAACV,cAAc;YACjBrB,KAAK,aAAa;gBAAE,QAAQyB;gBAAcC;YAAW;YACrDH,UAAU;YACV;QACF;QAEA,IAAI;YACFI,eAAe,MAAMpB,cACnBC,SACAiB,cACAf,WACAJ,cACAN;YAEF2B,eAAef,qBAAqBZ,KAAK,iBAAiB;gBACxD,QAAQyB;gBACR,cAAcnB;gBACdoB;gBACAF;gBACAG;YACF;YACA3B,KAAK,aAAa;gBAAE,QAAQyB;gBAAcC;gBAAYC;YAAa;YACnE,IAAIA,aAAa,IAAI,EACnBJ,UAAU;iBACL;gBACLK,YAAYvB;gBACZwB,eAAe;gBACfL;YACF;QACF,EAAE,OAAOP,OAAO;YACdW,YAAYZ,aAAaC;YACzBY,eAAe;YACf7B,KAAK,aAAa;gBAAE,QAAQyB;gBAAcC;YAAW;YACrDF;QACF;IACF;IAEA,OAAO;QACL,QAAQC;QACRC;QACAC;QACAJ;QACA,OACEA,WAAW,CAACF,eACRhB,SACAuB,aAAa;QACnB,cAAcL,UAAUlB,SAAYwB;IACtC;AACF;AAEO,eAAeG,gCACpBxB,OAAgC,EAChCyB,WAAmB,EACnBnC,MAAwB,EACxBE,GAA4C;IAE5CL,OAAOsC,aAAa,QAAQ;IAC5B,MAAM,EAAEvB,SAAS,EAAEJ,YAAY,EAAE,GAAG,MAAMJ,oCACxCJ,QACAE;IAGF,OAAOO,cAAcC,SAASyB,aAAavB,WAAWJ,cAAcN;AACtE;AAEO,eAAekC,6BACpB1B,OAAwE,EACxEyB,WAAmB,EACnBnC,MAAwB,EACxBE,GAAyC;IAEzC,OAAOW,gBAAgBH,SAASyB,aAAanC,QAAQE;AACvD;AAEA,eAAeW,gBACbH,OAAwE,EACxEyB,WAAmB,EACnBnC,MAAwB,EACxBE,GAAqE;IAErEL,OAAOsC,aAAa,QAAQ;IAC5B,MAAM,EAAE9B,iBAAiB,EAAEG,YAAY,EAAE,GACvC,MAAMJ,oCAAoCJ,QAAQE;IACpD,MAAMmC,UAAUnC,KAAK,aAAaG;IAClC,MAAMiC,cAAcC,yBAAyBJ,aAAa;QACxD,WAAWjC,KAAK;QAChB,YAAYA,KAAK;QACjB,OAAOA,KAAK;IACd;IACAL,OAAOyC,aAAa;IAEpB,MAAME,eAAe,MAAM9B,QAAQ,OAAO,CAAC,MAAM,CAC/C4B,aACAD,UAAU;QAAEA;IAAQ,IAAI,CAAC,GACzB3B,QAAQ,kBAAkB,EAC1BR,KAAK;IAEPL,OAAO2C,aAAa,OAAO,EAAE,CAAC,mBAAmB,EAAEL,aAAa;IAChE,OAAOhD,2BAA2BqB,cAAcgC,aAAa,OAAO,EAAEtC;AACxE"}

package/dist/es/index.mjs

@@ -6,6 +6,7 @@ import { AiLocateElement, PointSchema, RectSchema, SizeSchema, TMultimodalPrompt
 import { MIDSCENE_MODEL_NAME } from "@midscene/shared/env";
 import { ExecutionDump, GroupedActionDump, ReportActionDump, ServiceError } from "./types.mjs";
 import { Agent, createAgent } from "./agent/index.mjs";
+import { describeElementAtPoint, verifyElementByServiceLocate, verifyElementDescriptionAtPoint, verifyLocator } from "./element-describer.mjs";
 import { escapeContent, generateDumpScriptTag, generateImageScriptTag, parseDumpScript, parseDumpScriptAttributes, parseImageScripts, restoreImageReferences, unescapeContent } from "./dump/index.mjs";
 import { getTaskSearchArea, getTaskServiceDump } from "./dump/task-service-dump.mjs";
 import { ReportGenerator, nullReportGenerator } from "./report-generator.mjs";
@@ -15,6 +16,6 @@ import { ScreenshotItem } from "./screenshot-item.mjs";
 import { ScreenshotStore } from "./dump/screenshot-store.mjs";
 import { executionToMarkdown, reportToMarkdown } from "./report-markdown.mjs";
 const src = service;
-export { Agent, AiLocateElement, ExecutionDump, GroupedActionDump, MIDSCENE_MODEL_NAME, PointSchema, RectSchema, ReportActionDump, ReportGenerator, ReportMergingTool, ScreenshotItem, ScreenshotStore, service as Service, ServiceError, SizeSchema, TMultimodalPromptSchema, TUserPromptSchema, TaskRunner, collectDedupedExecutions, createAgent, createReportCliCommands, dedupeExecutionsKeepLatest, src as default, escapeContent, executionToMarkdown, generateDumpScriptTag, generateImageScriptTag, getMidsceneLocationSchema, getTaskSearchArea, getTaskServiceDump, getVersion, mergeReportFiles, nullReportGenerator, parseDumpScript, parseDumpScriptAttributes, parseImageScripts, plan, reportFileToMarkdown, reportToMarkdown, restoreImageReferences, runConnectivityTest, splitReportFile, splitReportHtmlByExecution, unescapeContent, z };
+export { Agent, AiLocateElement, ExecutionDump, GroupedActionDump, MIDSCENE_MODEL_NAME, PointSchema, RectSchema, ReportActionDump, ReportGenerator, ReportMergingTool, ScreenshotItem, ScreenshotStore, service as Service, ServiceError, SizeSchema, TMultimodalPromptSchema, TUserPromptSchema, TaskRunner, collectDedupedExecutions, createAgent, createReportCliCommands, dedupeExecutionsKeepLatest, src as default, describeElementAtPoint, escapeContent, executionToMarkdown, generateDumpScriptTag, generateImageScriptTag, getMidsceneLocationSchema, getTaskSearchArea, getTaskServiceDump, getVersion, mergeReportFiles, nullReportGenerator, parseDumpScript, parseDumpScriptAttributes, parseImageScripts, plan, reportFileToMarkdown, reportToMarkdown, restoreImageReferences, runConnectivityTest, splitReportFile, splitReportHtmlByExecution, unescapeContent, verifyElementByServiceLocate, verifyElementDescriptionAtPoint, verifyLocator, z };
 
 //# sourceMappingURL=index.mjs.map

package/dist/es/index.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"index.mjs","sources":["../../src/index.ts"],"sourcesContent":["import { z } from 'zod';\nimport Service from './service/index';\nimport { TaskRunner } from './task-runner';\nimport { getVersion } from './utils';\n\nexport {\n  plan,\n  AiLocateElement,\n  runConnectivityTest,\n  getMidsceneLocationSchema,\n  PointSchema,\n  SizeSchema,\n  RectSchema,\n  TMultimodalPromptSchema,\n  TUserPromptSchema,\n  type TMultimodalPrompt,\n  type TUserPrompt,\n  type ConnectivityCheckResultItem,\n  type ConnectivityTestConfig,\n  type ConnectivityTestResult,\n} from './ai-model/index';\n\nexport {\n  MIDSCENE_MODEL_NAME,\n  type CreateOpenAIClientFn,\n} from '@midscene/shared/env';\n\nexport type * from './types';\nexport {\n  ServiceError,\n  ExecutionDump,\n  ReportActionDump,\n  GroupedActionDump,\n  type IExecutionDump,\n  type IReportActionDump,\n  type IGroupedActionDump,\n  type ReportMeta,\n  type GroupMeta,\n} from './types';\n\nexport { z };\n\nexport default Service;\nexport { TaskRunner, Service, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n  MidsceneYamlConfigResult,\n  MidsceneYamlConfig,\n  MidsceneYamlScriptWebEnv,\n  MidsceneYamlScriptAndroidEnv,\n  MidsceneYamlScriptIOSEnv,\n  MidsceneYamlScriptEnv,\n  LocateOption,\n  DetailedLocateParam,\n} from './yaml';\n\nexport {\n  Agent,\n  type AgentOpt,\n  type AiActOptions,\n  createAgent,\n} from './agent';\n\n// Dump utilities\nexport {\n  restoreImageReferences,\n  escapeContent,\n  unescapeContent,\n  parseImageScripts,\n  parseDumpScript,\n  parseDumpScriptAttributes,\n  generateImageScriptTag,\n  generateDumpScriptTag,\n} from './dump';\nexport {\n  getTaskSearchArea,\n  getTaskServiceDump,\n} from './dump/task-service-dump';\n\n// Report generator\nexport type { IReportGenerator } from './report-generator';\nexport { ReportGenerator, nullReportGenerator } from './report-generator';\nexport {\n  collectDedupedExecutions,\n  ReportMergingTool,\n  dedupeExecutionsKeepLatest,\n  splitReportHtmlByExecution,\n} from './report';\nexport {\n  createReportCliCommands,\n  reportFileToMarkdown,\n  splitReportFile,\n  mergeReportFiles,\n  type ConsumeReportFileAction,\n  type ReportFileToMarkdownOptions,\n  type ReportCliCommandDefinition,\n  type ReportCliCommandEntry,\n  type SplitReportFileOptions,\n  type MergeReportFilesOptions,\n  type MergeReportFilesResult,\n} from './report-cli';\n\n// ScreenshotItem\nexport { ScreenshotItem } from './screenshot-item';\nexport { ScreenshotStore, type ScreenshotRef } from './dump/screenshot-store';\n\nexport {\n  executionToMarkdown,\n  reportToMarkdown,\n  type ExecutionMarkdownOptions,\n  type ExecutionMarkdownResult,\n  type ReportMarkdownResult,\n  type MarkdownAttachment,\n} from './report-markdown';\n"],"names":["Service"],"mappings":";;;;;;;;;;;;;;;;AA0CA,YAAeA"}
+{"version":3,"file":"index.mjs","sources":["../../src/index.ts"],"sourcesContent":["import { z } from 'zod';\nimport Service from './service/index';\nimport { TaskRunner } from './task-runner';\nimport { getVersion } from './utils';\n\nexport {\n  plan,\n  AiLocateElement,\n  runConnectivityTest,\n  getMidsceneLocationSchema,\n  PointSchema,\n  SizeSchema,\n  RectSchema,\n  TMultimodalPromptSchema,\n  TUserPromptSchema,\n  type TMultimodalPrompt,\n  type TUserPrompt,\n  type ConnectivityCheckResultItem,\n  type ConnectivityTestConfig,\n  type ConnectivityTestResult,\n} from './ai-model/index';\n\nexport {\n  MIDSCENE_MODEL_NAME,\n  type CreateOpenAIClientFn,\n} from '@midscene/shared/env';\n\nexport type * from './types';\nexport {\n  ServiceError,\n  ExecutionDump,\n  ReportActionDump,\n  GroupedActionDump,\n  type IExecutionDump,\n  type IReportActionDump,\n  type IGroupedActionDump,\n  type ReportMeta,\n  type GroupMeta,\n} from './types';\n\nexport { z };\n\nexport default Service;\nexport { TaskRunner, Service, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n  MidsceneYamlConfigResult,\n  MidsceneYamlConfig,\n  MidsceneYamlScriptWebEnv,\n  MidsceneYamlScriptAndroidEnv,\n  MidsceneYamlScriptIOSEnv,\n  MidsceneYamlScriptEnv,\n  LocateOption,\n  DetailedLocateParam,\n} from './yaml';\n\nexport {\n  Agent,\n  type AgentOpt,\n  type AiActOptions,\n  type GherkinStepKeyword,\n  type RunGherkinScenarioOptions,\n  createAgent,\n} from './agent';\nexport {\n  describeElementAtPoint,\n  verifyElementDescriptionAtPoint,\n  verifyElementByServiceLocate,\n  verifyLocator,\n  type DescribeElementAtPointOptions,\n  type DescribeElementCoordinateSpace,\n  type ElementDescriberRuntime,\n  type VerifyElementDescriptionAtPointOptions,\n  type VerifyElementByServiceLocateOptions,\n} from './element-describer';\n\n// Dump utilities\nexport {\n  restoreImageReferences,\n  escapeContent,\n  unescapeContent,\n  parseImageScripts,\n  parseDumpScript,\n  parseDumpScriptAttributes,\n  generateImageScriptTag,\n  generateDumpScriptTag,\n} from './dump';\nexport {\n  getTaskSearchArea,\n  getTaskServiceDump,\n} from './dump/task-service-dump';\n\n// Report generator\nexport type { IReportGenerator } from './report-generator';\nexport { ReportGenerator, nullReportGenerator } from './report-generator';\nexport {\n  collectDedupedExecutions,\n  ReportMergingTool,\n  dedupeExecutionsKeepLatest,\n  splitReportHtmlByExecution,\n} from './report';\nexport {\n  createReportCliCommands,\n  reportFileToMarkdown,\n  splitReportFile,\n  mergeReportFiles,\n  type ConsumeReportFileAction,\n  type ReportFileToMarkdownOptions,\n  type ReportCliCommandDefinition,\n  type ReportCliCommandEntry,\n  type SplitReportFileOptions,\n  type MergeReportFilesOptions,\n  type MergeReportFilesResult,\n} from './report-cli';\n\n// ScreenshotItem\nexport { ScreenshotItem } from './screenshot-item';\nexport { ScreenshotStore, type ScreenshotRef } from './dump/screenshot-store';\n\nexport {\n  executionToMarkdown,\n  reportToMarkdown,\n  type ExecutionMarkdownOptions,\n  type ExecutionMarkdownResult,\n  type ReportMarkdownResult,\n  type MarkdownAttachment,\n} from './report-markdown';\n"],"names":["Service"],"mappings":";;;;;;;;;;;;;;;;;AA0CA,YAAeA"}

package/dist/es/report-generator.mjs

@@ -61,16 +61,15 @@ class ReportGenerator {
         await this.flush();
         this.destroyed = true;
         if (!this.initialized) return;
-        this.printReportPath('finalized');
         return this.reportPath;
     }
     getReportPath() {
         return this.reportPath;
     }
-    printReportPath(verb) {
+    printReportPath() {
         if (!this.autoPrint || !this.reportPath) return;
         if (globalConfigManager.getEnvConfigInBoolean(MIDSCENE_REPORT_QUIET)) return;
-        'directory' === this.screenshotMode ? logMsg(`Midscene - report ${verb}: npx serve ${dirname(this.reportPath)}`) : logMsg(`Midscene - report ${verb}: ${this.reportPath}`);
+        'directory' === this.screenshotMode ? logMsg(`Midscene - report file updated: npx serve ${dirname(this.reportPath)}`) : logMsg(`Midscene - report file updated: ${this.reportPath}`);
     }
     async doWriteExecution(execution, reportMeta) {
         const singleDump = this.wrapAsReportDump(execution, reportMeta);
@@ -79,7 +78,7 @@ class ReportGenerator {
         if (this.shouldPersistExecutionDump) await this.persistExecutionDumpToFile(execution, singleDump);
         if (!this.firstWriteDone) {
             this.firstWriteDone = true;
-            this.printReportPath('generated');
+            this.printReportPath();
         }
     }
     mergeReportAttributes(attributes) {
@@ -189,7 +188,6 @@ class ReportGenerator {
             alsoWriteFileCopy: this.shouldPersistExecutionDump
         });
         if (options.reuseExistingReport) this.hydrateStateFromExistingReport();
-        this.printReportPath('will be generated at');
     }
 }
 function ensureHtmlFileName(reportFileName) {

package/dist/es/report-generator.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"report-generator.mjs","sources":["../../src/report-generator.ts"],"sourcesContent":["/*\n * PERF INVARIANT — DO NOT reintroduce sync fs APIs (writeFileSync /\n * appendFileSync) in this file's write paths. `ReportGenerator` runs on\n * the Electron main event loop during agent execution, and a single\n * progress tick appends a multi-MB ExecutionDump payload. Sync I/O here\n * blocked the loop for 20+ seconds per run, freezing IPC, scrcpy and\n * every renderer round-trip. Always use `fs/promises`. See commit\n * 6a25e05c and `report-generator-async-contract.test.ts`.\n */\nimport { existsSync, mkdirSync, readdirSync } from 'node:fs';\nimport {\n  appendFile as appendFileAsync,\n  writeFile as writeFileAsync,\n} from 'node:fs/promises';\nimport { dirname, join } from 'node:path';\nimport { getMidsceneRunSubDir } from '@midscene/shared/common';\nimport {\n  MIDSCENE_REPORT_QUIET,\n  globalConfigManager,\n} from '@midscene/shared/env';\nimport { ifInBrowser, logMsg, uuid } from '@midscene/shared/utils';\nimport {\n  generateDumpScriptTag,\n  generateImageScriptTag,\n  getBaseUrlFixScript,\n} from './dump/html-utils';\nimport { ScreenshotStore } from './dump/screenshot-store';\nimport {\n  type ExecutionDump,\n  ReportActionDump,\n  type ReportAttributes,\n  type ReportMeta,\n} from './types';\nimport { getReportTpl } from './utils';\n\nexport interface IReportGenerator {\n  /**\n   * Write or update a single execution.\n   * Each call appends a new dump script tag. The frontend deduplicates\n   * executions with the same id/name, keeping only the last one.\n   *\n   * @param execution  Current execution's full data\n   * @param reportMeta  Report-level metadata (groupName, sdkVersion, etc.)\n   */\n  onExecutionUpdate(\n    execution: ExecutionDump,\n    reportMeta: ReportMeta,\n    attributes?: ReportAttributes,\n  ): void;\n\n  /**\n   * @deprecated Use onExecutionUpdate instead. Kept for backward compatibility.\n   */\n  onDumpUpdate?(dump: ReportActionDump): void;\n\n  /**\n   * Wait for all queued write operations to complete.\n   */\n  flush(): Promise<void>;\n\n  /**\n   * Finalize the report. Calls flush() internally.\n   */\n  finalize(): Promise<string | undefined>;\n\n  getReportPath(): string | undefined;\n}\n\nexport const nullReportGenerator: IReportGenerator = {\n  onExecutionUpdate: () => {},\n  flush: async () => {},\n  finalize: async () => undefined,\n  getReportPath: () => undefined,\n};\n\nexport function assertReportGenerationOptions(opts: {\n  generateReport?: boolean;\n  persistExecutionDump?: boolean;\n}): void {\n  if (opts.generateReport === false && opts.persistExecutionDump === true) {\n    throw new Error(\n      'persistExecutionDump cannot be true when generateReport is false',\n    );\n  }\n}\n\nexport class ReportGenerator implements IReportGenerator {\n  private reportPath: string;\n  private screenshotMode: 'inline' | 'directory';\n  private shouldPersistExecutionDump: boolean;\n  private autoPrint: boolean;\n  private firstWriteDone = false;\n  private executionLogIndex = 0;\n  private executionLogFileIndexByExecutionKey = new Map<string, number>();\n\n  // Unique identifier for this report stream — used as data-group-id\n  private readonly reportStreamId: string;\n\n  // Tracks screenshots already written to disk (by id) to avoid duplicates\n  private screenshotStore: ScreenshotStore;\n  private initialized = false;\n\n  // Tracks the last execution + groupMeta for re-writing on finalize\n  private lastExecution?: ExecutionDump;\n  private lastReportMeta?: ReportMeta;\n  private reportAttributes: Record<string, string> = {};\n\n  // write queue for serial execution\n  private writeQueue: Promise<void> = Promise.resolve();\n  private destroyed = false;\n\n  constructor(options: {\n    reportPath: string;\n    screenshotMode: 'inline' | 'directory';\n    persistExecutionDump?: boolean;\n    autoPrint?: boolean;\n    reuseExistingReport?: boolean;\n  }) {\n    this.reportPath = options.reportPath;\n    this.screenshotMode = options.screenshotMode;\n    this.shouldPersistExecutionDump = options.persistExecutionDump ?? false;\n    this.autoPrint = options.autoPrint ?? true;\n    this.reportStreamId = uuid();\n    this.screenshotStore = new ScreenshotStore({\n      mode: this.screenshotMode === 'inline' ? 'inline' : 'directory',\n      reportPath: this.reportPath,\n      screenshotsDir: join(dirname(this.reportPath), 'screenshots'),\n      writeInlineImage: async (id, base64) => {\n        await appendFileAsync(\n          this.reportPath,\n          `\\n${generateImageScriptTag(id, base64)}`,\n        );\n      },\n      alsoWriteFileCopy: this.shouldPersistExecutionDump,\n    });\n    if (options.reuseExistingReport) {\n      this.hydrateStateFromExistingReport();\n    }\n    this.printReportPath('will be generated at');\n  }\n\n  static create(\n    reportFileName: string,\n    opts: {\n      generateReport?: boolean;\n      persistExecutionDump?: boolean;\n      outputFormat?: 'single-html' | 'html-and-external-assets';\n      autoPrintReportMsg?: boolean;\n      reuseExistingReport?: boolean;\n    },\n  ): IReportGenerator {\n    assertReportGenerationOptions(opts);\n    validateReportFileName(reportFileName);\n    if (opts.generateReport === false) return nullReportGenerator;\n\n    // In browser environment, file system is not available\n    if (ifInBrowser) return nullReportGenerator;\n\n    const reportRootDir = getMidsceneRunSubDir('report');\n    const outputDir = join(reportRootDir, reportFileName);\n    const reportPath =\n      opts.outputFormat === 'html-and-external-assets'\n        ? join(outputDir, 'index.html')\n        : join(reportRootDir, ensureHtmlFileName(reportFileName));\n    return new ReportGenerator({\n      reportPath,\n      screenshotMode:\n        opts.outputFormat === 'html-and-external-assets'\n          ? 'directory'\n          : 'inline',\n      persistExecutionDump: opts.persistExecutionDump,\n      autoPrint: opts.autoPrintReportMsg,\n      reuseExistingReport: opts.reuseExistingReport,\n    });\n  }\n\n  onExecutionUpdate(\n    execution: ExecutionDump,\n    reportMeta: ReportMeta,\n    attributes?: ReportAttributes,\n  ): void {\n    this.lastExecution = execution;\n    this.lastReportMeta = reportMeta;\n    this.mergeReportAttributes(attributes);\n    this.writeQueue = this.writeQueue.then(async () => {\n      if (this.destroyed) return;\n      await this.doWriteExecution(execution, reportMeta);\n    });\n  }\n\n  async flush(): Promise<void> {\n    await this.writeQueue;\n  }\n\n  async finalize(): Promise<string | undefined> {\n    // Re-write the last execution to capture any final state changes\n    if (this.lastExecution && this.lastReportMeta) {\n      this.onExecutionUpdate(this.lastExecution, this.lastReportMeta);\n    }\n    await this.flush();\n    this.destroyed = true;\n\n    if (!this.initialized) {\n      // No executions were ever written — no file exists\n      return undefined;\n    }\n\n    this.printReportPath('finalized');\n    return this.reportPath;\n  }\n\n  getReportPath(): string | undefined {\n    return this.reportPath;\n  }\n\n  private printReportPath(verb: string): void {\n    if (!this.autoPrint || !this.reportPath) return;\n    if (globalConfigManager.getEnvConfigInBoolean(MIDSCENE_REPORT_QUIET))\n      return;\n\n    if (this.screenshotMode === 'directory') {\n      logMsg(\n        `Midscene - report ${verb}: npx serve ${dirname(this.reportPath)}`,\n      );\n    } else {\n      logMsg(`Midscene - report ${verb}: ${this.reportPath}`);\n    }\n  }\n\n  private async doWriteExecution(\n    execution: ExecutionDump,\n    reportMeta: ReportMeta,\n  ): Promise<void> {\n    const singleDump = this.wrapAsReportDump(execution, reportMeta);\n\n    if (this.screenshotMode === 'inline') {\n      await this.writeInlineExecution(execution, singleDump);\n    } else {\n      await this.writeDirectoryExecution(execution, singleDump);\n    }\n\n    if (this.shouldPersistExecutionDump) {\n      await this.persistExecutionDumpToFile(execution, singleDump);\n    }\n\n    if (!this.firstWriteDone) {\n      this.firstWriteDone = true;\n      this.printReportPath('generated');\n    }\n  }\n\n  private mergeReportAttributes(attributes?: ReportAttributes): void {\n    if (!attributes) {\n      return;\n    }\n\n    for (const [key, value] of Object.entries(attributes)) {\n      if (value === undefined || value === null) {\n        continue;\n      }\n      this.reportAttributes[key] = String(value);\n    }\n  }\n\n  private hydrateStateFromExistingReport(): void {\n    if (!existsSync(this.reportPath)) {\n      return;\n    }\n\n    // Reuse existing report file and append new updates instead of rewriting.\n    this.initialized = true;\n\n    if (!this.shouldPersistExecutionDump) {\n      return;\n    }\n\n    const reportDir = dirname(this.reportPath);\n    const existingExecutionIndices = readdirSync(reportDir)\n      .map((name) => /^(\\d+)\\.execution\\.json$/.exec(name)?.[1])\n      .filter((index): index is string => Boolean(index))\n      .map((index) => Number.parseInt(index, 10))\n      .filter((index) => Number.isFinite(index));\n\n    if (existingExecutionIndices.length > 0) {\n      this.executionLogIndex = Math.max(...existingExecutionIndices);\n    }\n  }\n\n  private getDumpScriptAttributes(): Record<string, string> {\n    return {\n      'data-group-id': this.reportStreamId,\n      ...this.reportAttributes,\n    };\n  }\n\n  /**\n   * Wrap an ExecutionDump + ReportMeta into a single-execution ReportActionDump.\n   */\n  private wrapAsReportDump(\n    execution: ExecutionDump,\n    reportMeta: ReportMeta,\n  ): ReportActionDump {\n    return new ReportActionDump({\n      sdkVersion: reportMeta.sdkVersion,\n      groupName: reportMeta.groupName,\n      groupDescription: reportMeta.groupDescription,\n      modelBriefs: reportMeta.modelBriefs,\n      deviceType: reportMeta.deviceType,\n      executions: [execution],\n    });\n  }\n\n  /**\n   * Append-only inline mode: write new screenshots and a dump tag on every call.\n   * The frontend deduplicates executions with the same id/name (keeps last).\n   * Duplicate dump JSON is acceptable; only screenshots are deduplicated.\n   *\n   * All writes go through `fs/promises` so they run on libuv's thread pool\n   * rather than blocking the Node event loop. A long agent run previously\n   * appended multi-MB dumps (screenshots + serialized tasks) per progress\n   * tick on the main thread, starving IPC and UI for >10s per stall.\n   */\n  private async writeInlineExecution(\n    execution: ExecutionDump,\n    singleDump: ReportActionDump,\n  ): Promise<void> {\n    const dir = dirname(this.reportPath);\n    if (!existsSync(dir)) {\n      mkdirSync(dir, { recursive: true });\n    }\n\n    // Initialize: write HTML template once\n    if (!this.initialized) {\n      await writeFileAsync(this.reportPath, getReportTpl());\n      this.initialized = true;\n    }\n\n    // Append new screenshots (skip already-written ones)\n    for (const screenshot of execution.collectScreenshots()) {\n      await this.screenshotStore.persist(screenshot);\n    }\n\n    // Append dump tag (always — frontend keeps only last per execution id)\n    const serialized = singleDump.serialize();\n    await appendFileAsync(\n      this.reportPath,\n      `\\n${generateDumpScriptTag(serialized, this.getDumpScriptAttributes())}`,\n    );\n  }\n\n  private async writeDirectoryExecution(\n    execution: ExecutionDump,\n    singleDump: ReportActionDump,\n  ): Promise<void> {\n    const dir = dirname(this.reportPath);\n    if (!existsSync(dir)) {\n      mkdirSync(dir, { recursive: true });\n    }\n\n    for (const screenshot of execution.collectScreenshots()) {\n      await this.screenshotStore.persist(screenshot);\n    }\n\n    // 2. Append dump tag (always — frontend keeps only last per execution id)\n    const serialized = singleDump.serialize();\n\n    if (!this.initialized) {\n      await writeFileAsync(\n        this.reportPath,\n        `${getReportTpl()}${getBaseUrlFixScript()}`,\n      );\n      this.initialized = true;\n    }\n\n    await appendFileAsync(\n      this.reportPath,\n      `\\n${generateDumpScriptTag(serialized, this.getDumpScriptAttributes())}`,\n    );\n  }\n\n  private getExecutionLogKey(execution: ExecutionDump): string {\n    if (!execution.id) {\n      throw new Error(\n        'ReportGenerator: execution.id is required for persisting execution dumps',\n      );\n    }\n    return `id:${execution.id}`;\n  }\n\n  private async persistExecutionDumpToFile(\n    execution: ExecutionDump,\n    singleDump: ReportActionDump,\n  ): Promise<void> {\n    const dir = dirname(this.reportPath);\n    if (!existsSync(dir)) {\n      mkdirSync(dir, { recursive: true });\n    }\n\n    const executionLogKey = this.getExecutionLogKey(execution);\n    let fileIndex =\n      this.executionLogFileIndexByExecutionKey.get(executionLogKey);\n    if (!fileIndex) {\n      this.executionLogIndex += 1;\n      fileIndex = this.executionLogIndex;\n      this.executionLogFileIndexByExecutionKey.set(executionLogKey, fileIndex);\n    }\n\n    const fileName = `${fileIndex}.execution.json`;\n    const filePath = join(dirname(this.reportPath), fileName);\n    await writeFileAsync(filePath, singleDump.serialize(2), 'utf-8');\n  }\n}\n\nfunction ensureHtmlFileName(reportFileName: string): string {\n  return reportFileName.endsWith('.html')\n    ? reportFileName\n    : `${reportFileName}.html`;\n}\n\nfunction validateReportFileName(reportFileName: string): void {\n  if (!reportFileName?.trim()) {\n    throw new Error('reportFileName must be a non-empty string');\n  }\n\n  if (/[\\\\/]/.test(reportFileName)) {\n    throw new Error(\n      'reportFileName must not contain path separators (`/` or `\\\\\\\\`)',\n    );\n  }\n\n  if (/[:*?\"<>|]/.test(reportFileName)) {\n    throw new Error(\n      'reportFileName contains illegal filename characters: : * ? \" < > |',\n    );\n  }\n}\n"],"names":["nullReportGenerator","undefined","assertReportGenerationOptions","opts","Error","ReportGenerator","reportFileName","validateReportFileName","ifInBrowser","reportRootDir","getMidsceneRunSubDir","outputDir","join","reportPath","ensureHtmlFileName","execution","reportMeta","attributes","verb","globalConfigManager","MIDSCENE_REPORT_QUIET","logMsg","dirname","singleDump","key","value","Object","String","existsSync","reportDir","existingExecutionIndices","readdirSync","name","index","Boolean","Number","Math","ReportActionDump","dir","mkdirSync","writeFileAsync","getReportTpl","screenshot","serialized","appendFileAsync","generateDumpScriptTag","getBaseUrlFixScript","executionLogKey","fileIndex","fileName","filePath","options","Map","Promise","uuid","ScreenshotStore","id","base64","generateImageScriptTag"],"mappings":";;;;;;;;;;AAQC;;;;;;;;;;AA4DM,MAAMA,sBAAwC;IACnD,mBAAmB,KAAO;IAC1B,OAAO,WAAa;IACpB,UAAU,UAAYC;IACtB,eAAe,IAAMA;AACvB;AAEO,SAASC,8BAA8BC,IAG7C;IACC,IAAIA,AAAwB,UAAxBA,KAAK,cAAc,IAAcA,AAA8B,SAA9BA,KAAK,oBAAoB,EAC5D,MAAM,IAAIC,MACR;AAGN;AAEO,MAAMC;IAuDX,OAAO,OACLC,cAAsB,EACtBH,IAMC,EACiB;QAClBD,8BAA8BC;QAC9BI,uBAAuBD;QACvB,IAAIH,AAAwB,UAAxBA,KAAK,cAAc,EAAY,OAAOH;QAG1C,IAAIQ,aAAa,OAAOR;QAExB,MAAMS,gBAAgBC,qBAAqB;QAC3C,MAAMC,YAAYC,KAAKH,eAAeH;QACtC,MAAMO,aACJV,AAAsB,+BAAtBA,KAAK,YAAY,GACbS,KAAKD,WAAW,gBAChBC,KAAKH,eAAeK,mBAAmBR;QAC7C,OAAO,IAAID,gBAAgB;YACzBQ;YACA,gBACEV,AAAsB,+BAAtBA,KAAK,YAAY,GACb,cACA;YACN,sBAAsBA,KAAK,oBAAoB;YAC/C,WAAWA,KAAK,kBAAkB;YAClC,qBAAqBA,KAAK,mBAAmB;QAC/C;IACF;IAEA,kBACEY,SAAwB,EACxBC,UAAsB,EACtBC,UAA6B,EACvB;QACN,IAAI,CAAC,aAAa,GAAGF;QACrB,IAAI,CAAC,cAAc,GAAGC;QACtB,IAAI,CAAC,qBAAqB,CAACC;QAC3B,IAAI,CAAC,UAAU,GAAG,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC;YACrC,IAAI,IAAI,CAAC,SAAS,EAAE;YACpB,MAAM,IAAI,CAAC,gBAAgB,CAACF,WAAWC;QACzC;IACF;IAEA,MAAM,QAAuB;QAC3B,MAAM,IAAI,CAAC,UAAU;IACvB;IAEA,MAAM,WAAwC;QAE5C,IAAI,IAAI,CAAC,aAAa,IAAI,IAAI,CAAC,cAAc,EAC3C,IAAI,CAAC,iBAAiB,CAAC,IAAI,CAAC,aAAa,EAAE,IAAI,CAAC,cAAc;QAEhE,MAAM,IAAI,CAAC,KAAK;QAChB,IAAI,CAAC,SAAS,GAAG;QAEjB,IAAI,CAAC,IAAI,CAAC,WAAW,EAEnB;QAGF,IAAI,CAAC,eAAe,CAAC;QACrB,OAAO,IAAI,CAAC,UAAU;IACxB;IAEA,gBAAoC;QAClC,OAAO,IAAI,CAAC,UAAU;IACxB;IAEQ,gBAAgBE,IAAY,EAAQ;QAC1C,IAAI,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,IAAI,CAAC,UAAU,EAAE;QACzC,IAAIC,oBAAoB,qBAAqB,CAACC,wBAC5C;QAE0B,gBAAxB,IAAI,CAAC,cAAc,GACrBC,OACE,CAAC,kBAAkB,EAAEH,KAAK,YAAY,EAAEI,QAAQ,IAAI,CAAC,UAAU,GAAG,IAGpED,OAAO,CAAC,kBAAkB,EAAEH,KAAK,EAAE,EAAE,IAAI,CAAC,UAAU,EAAE;IAE1D;IAEA,MAAc,iBACZH,SAAwB,EACxBC,UAAsB,EACP;QACf,MAAMO,aAAa,IAAI,CAAC,gBAAgB,CAACR,WAAWC;QAEpD,IAAI,AAAwB,aAAxB,IAAI,CAAC,cAAc,EACrB,MAAM,IAAI,CAAC,oBAAoB,CAACD,WAAWQ;aAE3C,MAAM,IAAI,CAAC,uBAAuB,CAACR,WAAWQ;QAGhD,IAAI,IAAI,CAAC,0BAA0B,EACjC,MAAM,IAAI,CAAC,0BAA0B,CAACR,WAAWQ;QAGnD,IAAI,CAAC,IAAI,CAAC,cAAc,EAAE;YACxB,IAAI,CAAC,cAAc,GAAG;YACtB,IAAI,CAAC,eAAe,CAAC;QACvB;IACF;IAEQ,sBAAsBN,UAA6B,EAAQ;QACjE,IAAI,CAACA,YACH;QAGF,KAAK,MAAM,CAACO,KAAKC,MAAM,IAAIC,OAAO,OAAO,CAACT,YACxC,IAAIQ,QAAAA,OAGJ,IAAI,CAAC,gBAAgB,CAACD,IAAI,GAAGG,OAAOF;IAExC;IAEQ,iCAAuC;QAC7C,IAAI,CAACG,WAAW,IAAI,CAAC,UAAU,GAC7B;QAIF,IAAI,CAAC,WAAW,GAAG;QAEnB,IAAI,CAAC,IAAI,CAAC,0BAA0B,EAClC;QAGF,MAAMC,YAAYP,QAAQ,IAAI,CAAC,UAAU;QACzC,MAAMQ,2BAA2BC,YAAYF,WAC1C,GAAG,CAAC,CAACG,OAAS,2BAA2B,IAAI,CAACA,OAAO,CAAC,EAAE,EACxD,MAAM,CAAC,CAACC,QAA2BC,QAAQD,QAC3C,GAAG,CAAC,CAACA,QAAUE,OAAO,QAAQ,CAACF,OAAO,KACtC,MAAM,CAAC,CAACA,QAAUE,OAAO,QAAQ,CAACF;QAErC,IAAIH,yBAAyB,MAAM,GAAG,GACpC,IAAI,CAAC,iBAAiB,GAAGM,KAAK,GAAG,IAAIN;IAEzC;IAEQ,0BAAkD;QACxD,OAAO;YACL,iBAAiB,IAAI,CAAC,cAAc;YACpC,GAAG,IAAI,CAAC,gBAAgB;QAC1B;IACF;IAKQ,iBACNf,SAAwB,EACxBC,UAAsB,EACJ;QAClB,OAAO,IAAIqB,iBAAiB;YAC1B,YAAYrB,WAAW,UAAU;YACjC,WAAWA,WAAW,SAAS;YAC/B,kBAAkBA,WAAW,gBAAgB;YAC7C,aAAaA,WAAW,WAAW;YACnC,YAAYA,WAAW,UAAU;YACjC,YAAY;gBAACD;aAAU;QACzB;IACF;IAYA,MAAc,qBACZA,SAAwB,EACxBQ,UAA4B,EACb;QACf,MAAMe,MAAMhB,QAAQ,IAAI,CAAC,UAAU;QACnC,IAAI,CAACM,WAAWU,MACdC,UAAUD,KAAK;YAAE,WAAW;QAAK;QAInC,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE;YACrB,MAAME,UAAe,IAAI,CAAC,UAAU,EAAEC;YACtC,IAAI,CAAC,WAAW,GAAG;QACrB;QAGA,KAAK,MAAMC,cAAc3B,UAAU,kBAAkB,GACnD,MAAM,IAAI,CAAC,eAAe,CAAC,OAAO,CAAC2B;QAIrC,MAAMC,aAAapB,WAAW,SAAS;QACvC,MAAMqB,WACJ,IAAI,CAAC,UAAU,EACf,CAAC,EAAE,EAAEC,sBAAsBF,YAAY,IAAI,CAAC,uBAAuB,KAAK;IAE5E;IAEA,MAAc,wBACZ5B,SAAwB,EACxBQ,UAA4B,EACb;QACf,MAAMe,MAAMhB,QAAQ,IAAI,CAAC,UAAU;QACnC,IAAI,CAACM,WAAWU,MACdC,UAAUD,KAAK;YAAE,WAAW;QAAK;QAGnC,KAAK,MAAMI,cAAc3B,UAAU,kBAAkB,GACnD,MAAM,IAAI,CAAC,eAAe,CAAC,OAAO,CAAC2B;QAIrC,MAAMC,aAAapB,WAAW,SAAS;QAEvC,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE;YACrB,MAAMiB,UACJ,IAAI,CAAC,UAAU,EACf,GAAGC,iBAAiBK,uBAAuB;YAE7C,IAAI,CAAC,WAAW,GAAG;QACrB;QAEA,MAAMF,WACJ,IAAI,CAAC,UAAU,EACf,CAAC,EAAE,EAAEC,sBAAsBF,YAAY,IAAI,CAAC,uBAAuB,KAAK;IAE5E;IAEQ,mBAAmB5B,SAAwB,EAAU;QAC3D,IAAI,CAACA,UAAU,EAAE,EACf,MAAM,IAAIX,MACR;QAGJ,OAAO,CAAC,GAAG,EAAEW,UAAU,EAAE,EAAE;IAC7B;IAEA,MAAc,2BACZA,SAAwB,EACxBQ,UAA4B,EACb;QACf,MAAMe,MAAMhB,QAAQ,IAAI,CAAC,UAAU;QACnC,IAAI,CAACM,WAAWU,MACdC,UAAUD,KAAK;YAAE,WAAW;QAAK;QAGnC,MAAMS,kBAAkB,IAAI,CAAC,kBAAkB,CAAChC;QAChD,IAAIiC,YACF,IAAI,CAAC,mCAAmC,CAAC,GAAG,CAACD;QAC/C,IAAI,CAACC,WAAW;YACd,IAAI,CAAC,iBAAiB,IAAI;YAC1BA,YAAY,IAAI,CAAC,iBAAiB;YAClC,IAAI,CAAC,mCAAmC,CAAC,GAAG,CAACD,iBAAiBC;QAChE;QAEA,MAAMC,WAAW,GAAGD,UAAU,eAAe,CAAC;QAC9C,MAAME,WAAWtC,KAAKU,QAAQ,IAAI,CAAC,UAAU,GAAG2B;QAChD,MAAMT,UAAeU,UAAU3B,WAAW,SAAS,CAAC,IAAI;IAC1D;IA3SA,YAAY4B,OAMX,CAAE;QA9BH,uBAAQ,cAAR;QACA,uBAAQ,kBAAR;QACA,uBAAQ,8BAAR;QACA,uBAAQ,aAAR;QACA,uBAAQ,kBAAiB;QACzB,uBAAQ,qBAAoB;QAC5B,uBAAQ,uCAAsC,IAAIC;QAGlD,uBAAiB,kBAAjB;QAGA,uBAAQ,mBAAR;QACA,uBAAQ,eAAc;QAGtB,uBAAQ,iBAAR;QACA,uBAAQ,kBAAR;QACA,uBAAQ,oBAA2C,CAAC;QAGpD,uBAAQ,cAA4BC,QAAQ,OAAO;QACnD,uBAAQ,aAAY;QASlB,IAAI,CAAC,UAAU,GAAGF,QAAQ,UAAU;QACpC,IAAI,CAAC,cAAc,GAAGA,QAAQ,cAAc;QAC5C,IAAI,CAAC,0BAA0B,GAAGA,QAAQ,oBAAoB,IAAI;QAClE,IAAI,CAAC,SAAS,GAAGA,QAAQ,SAAS,IAAI;QACtC,IAAI,CAAC,cAAc,GAAGG;QACtB,IAAI,CAAC,eAAe,GAAG,IAAIC,gBAAgB;YACzC,MAAM,AAAwB,aAAxB,IAAI,CAAC,cAAc,GAAgB,WAAW;YACpD,YAAY,IAAI,CAAC,UAAU;YAC3B,gBAAgB3C,KAAKU,QAAQ,IAAI,CAAC,UAAU,GAAG;YAC/C,kBAAkB,OAAOkC,IAAIC;gBAC3B,MAAMb,WACJ,IAAI,CAAC,UAAU,EACf,CAAC,EAAE,EAAEc,uBAAuBF,IAAIC,SAAS;YAE7C;YACA,mBAAmB,IAAI,CAAC,0BAA0B;QACpD;QACA,IAAIN,QAAQ,mBAAmB,EAC7B,IAAI,CAAC,8BAA8B;QAErC,IAAI,CAAC,eAAe,CAAC;IACvB;AAgRF;AAEA,SAASrC,mBAAmBR,cAAsB;IAChD,OAAOA,eAAe,QAAQ,CAAC,WAC3BA,iBACA,GAAGA,eAAe,KAAK,CAAC;AAC9B;AAEA,SAASC,uBAAuBD,cAAsB;IACpD,IAAI,CAACA,gBAAgB,QACnB,MAAM,IAAIF,MAAM;IAGlB,IAAI,QAAQ,IAAI,CAACE,iBACf,MAAM,IAAIF,MACR;IAIJ,IAAI,YAAY,IAAI,CAACE,iBACnB,MAAM,IAAIF,MACR;AAGN"}
+{"version":3,"file":"report-generator.mjs","sources":["../../src/report-generator.ts"],"sourcesContent":["/*\n * PERF INVARIANT — DO NOT reintroduce sync fs APIs (writeFileSync /\n * appendFileSync) in this file's write paths. `ReportGenerator` runs on\n * the Electron main event loop during agent execution, and a single\n * progress tick appends a multi-MB ExecutionDump payload. Sync I/O here\n * blocked the loop for 20+ seconds per run, freezing IPC, scrcpy and\n * every renderer round-trip. Always use `fs/promises`. See commit\n * 6a25e05c and `report-generator-async-contract.test.ts`.\n */\nimport { existsSync, mkdirSync, readdirSync } from 'node:fs';\nimport {\n  appendFile as appendFileAsync,\n  writeFile as writeFileAsync,\n} from 'node:fs/promises';\nimport { dirname, join } from 'node:path';\nimport { getMidsceneRunSubDir } from '@midscene/shared/common';\nimport {\n  MIDSCENE_REPORT_QUIET,\n  globalConfigManager,\n} from '@midscene/shared/env';\nimport { ifInBrowser, logMsg, uuid } from '@midscene/shared/utils';\nimport {\n  generateDumpScriptTag,\n  generateImageScriptTag,\n  getBaseUrlFixScript,\n} from './dump/html-utils';\nimport { ScreenshotStore } from './dump/screenshot-store';\nimport {\n  type ExecutionDump,\n  ReportActionDump,\n  type ReportAttributes,\n  type ReportMeta,\n} from './types';\nimport { getReportTpl } from './utils';\n\nexport interface IReportGenerator {\n  /**\n   * Write or update a single execution.\n   * Each call appends a new dump script tag. The frontend deduplicates\n   * executions with the same id/name, keeping only the last one.\n   *\n   * @param execution  Current execution's full data\n   * @param reportMeta  Report-level metadata (groupName, sdkVersion, etc.)\n   */\n  onExecutionUpdate(\n    execution: ExecutionDump,\n    reportMeta: ReportMeta,\n    attributes?: ReportAttributes,\n  ): void;\n\n  /**\n   * @deprecated Use onExecutionUpdate instead. Kept for backward compatibility.\n   */\n  onDumpUpdate?(dump: ReportActionDump): void;\n\n  /**\n   * Wait for all queued write operations to complete.\n   */\n  flush(): Promise<void>;\n\n  /**\n   * Finalize the report. Calls flush() internally.\n   */\n  finalize(): Promise<string | undefined>;\n\n  getReportPath(): string | undefined;\n}\n\nexport const nullReportGenerator: IReportGenerator = {\n  onExecutionUpdate: () => {},\n  flush: async () => {},\n  finalize: async () => undefined,\n  getReportPath: () => undefined,\n};\n\nexport function assertReportGenerationOptions(opts: {\n  generateReport?: boolean;\n  persistExecutionDump?: boolean;\n}): void {\n  if (opts.generateReport === false && opts.persistExecutionDump === true) {\n    throw new Error(\n      'persistExecutionDump cannot be true when generateReport is false',\n    );\n  }\n}\n\nexport class ReportGenerator implements IReportGenerator {\n  private reportPath: string;\n  private screenshotMode: 'inline' | 'directory';\n  private shouldPersistExecutionDump: boolean;\n  private autoPrint: boolean;\n  private firstWriteDone = false;\n  private executionLogIndex = 0;\n  private executionLogFileIndexByExecutionKey = new Map<string, number>();\n\n  // Unique identifier for this report stream — used as data-group-id\n  private readonly reportStreamId: string;\n\n  // Tracks screenshots already written to disk (by id) to avoid duplicates\n  private screenshotStore: ScreenshotStore;\n  private initialized = false;\n\n  // Tracks the last execution + groupMeta for re-writing on finalize\n  private lastExecution?: ExecutionDump;\n  private lastReportMeta?: ReportMeta;\n  private reportAttributes: Record<string, string> = {};\n\n  // write queue for serial execution\n  private writeQueue: Promise<void> = Promise.resolve();\n  private destroyed = false;\n\n  constructor(options: {\n    reportPath: string;\n    screenshotMode: 'inline' | 'directory';\n    persistExecutionDump?: boolean;\n    autoPrint?: boolean;\n    reuseExistingReport?: boolean;\n  }) {\n    this.reportPath = options.reportPath;\n    this.screenshotMode = options.screenshotMode;\n    this.shouldPersistExecutionDump = options.persistExecutionDump ?? false;\n    this.autoPrint = options.autoPrint ?? true;\n    this.reportStreamId = uuid();\n    this.screenshotStore = new ScreenshotStore({\n      mode: this.screenshotMode === 'inline' ? 'inline' : 'directory',\n      reportPath: this.reportPath,\n      screenshotsDir: join(dirname(this.reportPath), 'screenshots'),\n      writeInlineImage: async (id, base64) => {\n        await appendFileAsync(\n          this.reportPath,\n          `\\n${generateImageScriptTag(id, base64)}`,\n        );\n      },\n      alsoWriteFileCopy: this.shouldPersistExecutionDump,\n    });\n    if (options.reuseExistingReport) {\n      this.hydrateStateFromExistingReport();\n    }\n  }\n\n  static create(\n    reportFileName: string,\n    opts: {\n      generateReport?: boolean;\n      persistExecutionDump?: boolean;\n      outputFormat?: 'single-html' | 'html-and-external-assets';\n      autoPrintReportMsg?: boolean;\n      reuseExistingReport?: boolean;\n    },\n  ): IReportGenerator {\n    assertReportGenerationOptions(opts);\n    validateReportFileName(reportFileName);\n    if (opts.generateReport === false) return nullReportGenerator;\n\n    // In browser environment, file system is not available\n    if (ifInBrowser) return nullReportGenerator;\n\n    const reportRootDir = getMidsceneRunSubDir('report');\n    const outputDir = join(reportRootDir, reportFileName);\n    const reportPath =\n      opts.outputFormat === 'html-and-external-assets'\n        ? join(outputDir, 'index.html')\n        : join(reportRootDir, ensureHtmlFileName(reportFileName));\n    return new ReportGenerator({\n      reportPath,\n      screenshotMode:\n        opts.outputFormat === 'html-and-external-assets'\n          ? 'directory'\n          : 'inline',\n      persistExecutionDump: opts.persistExecutionDump,\n      autoPrint: opts.autoPrintReportMsg,\n      reuseExistingReport: opts.reuseExistingReport,\n    });\n  }\n\n  onExecutionUpdate(\n    execution: ExecutionDump,\n    reportMeta: ReportMeta,\n    attributes?: ReportAttributes,\n  ): void {\n    this.lastExecution = execution;\n    this.lastReportMeta = reportMeta;\n    this.mergeReportAttributes(attributes);\n    this.writeQueue = this.writeQueue.then(async () => {\n      if (this.destroyed) return;\n      await this.doWriteExecution(execution, reportMeta);\n    });\n  }\n\n  async flush(): Promise<void> {\n    await this.writeQueue;\n  }\n\n  async finalize(): Promise<string | undefined> {\n    // Re-write the last execution to capture any final state changes\n    if (this.lastExecution && this.lastReportMeta) {\n      this.onExecutionUpdate(this.lastExecution, this.lastReportMeta);\n    }\n    await this.flush();\n    this.destroyed = true;\n\n    if (!this.initialized) {\n      // No executions were ever written — no file exists\n      return undefined;\n    }\n\n    return this.reportPath;\n  }\n\n  getReportPath(): string | undefined {\n    return this.reportPath;\n  }\n\n  private printReportPath(): void {\n    if (!this.autoPrint || !this.reportPath) return;\n    if (globalConfigManager.getEnvConfigInBoolean(MIDSCENE_REPORT_QUIET))\n      return;\n\n    if (this.screenshotMode === 'directory') {\n      logMsg(\n        `Midscene - report file updated: npx serve ${dirname(this.reportPath)}`,\n      );\n    } else {\n      logMsg(`Midscene - report file updated: ${this.reportPath}`);\n    }\n  }\n\n  private async doWriteExecution(\n    execution: ExecutionDump,\n    reportMeta: ReportMeta,\n  ): Promise<void> {\n    const singleDump = this.wrapAsReportDump(execution, reportMeta);\n\n    if (this.screenshotMode === 'inline') {\n      await this.writeInlineExecution(execution, singleDump);\n    } else {\n      await this.writeDirectoryExecution(execution, singleDump);\n    }\n\n    if (this.shouldPersistExecutionDump) {\n      await this.persistExecutionDumpToFile(execution, singleDump);\n    }\n\n    if (!this.firstWriteDone) {\n      this.firstWriteDone = true;\n      this.printReportPath();\n    }\n  }\n\n  private mergeReportAttributes(attributes?: ReportAttributes): void {\n    if (!attributes) {\n      return;\n    }\n\n    for (const [key, value] of Object.entries(attributes)) {\n      if (value === undefined || value === null) {\n        continue;\n      }\n      this.reportAttributes[key] = String(value);\n    }\n  }\n\n  private hydrateStateFromExistingReport(): void {\n    if (!existsSync(this.reportPath)) {\n      return;\n    }\n\n    // Reuse existing report file and append new updates instead of rewriting.\n    this.initialized = true;\n\n    if (!this.shouldPersistExecutionDump) {\n      return;\n    }\n\n    const reportDir = dirname(this.reportPath);\n    const existingExecutionIndices = readdirSync(reportDir)\n      .map((name) => /^(\\d+)\\.execution\\.json$/.exec(name)?.[1])\n      .filter((index): index is string => Boolean(index))\n      .map((index) => Number.parseInt(index, 10))\n      .filter((index) => Number.isFinite(index));\n\n    if (existingExecutionIndices.length > 0) {\n      this.executionLogIndex = Math.max(...existingExecutionIndices);\n    }\n  }\n\n  private getDumpScriptAttributes(): Record<string, string> {\n    return {\n      'data-group-id': this.reportStreamId,\n      ...this.reportAttributes,\n    };\n  }\n\n  /**\n   * Wrap an ExecutionDump + ReportMeta into a single-execution ReportActionDump.\n   */\n  private wrapAsReportDump(\n    execution: ExecutionDump,\n    reportMeta: ReportMeta,\n  ): ReportActionDump {\n    return new ReportActionDump({\n      sdkVersion: reportMeta.sdkVersion,\n      groupName: reportMeta.groupName,\n      groupDescription: reportMeta.groupDescription,\n      modelBriefs: reportMeta.modelBriefs,\n      deviceType: reportMeta.deviceType,\n      executions: [execution],\n    });\n  }\n\n  /**\n   * Append-only inline mode: write new screenshots and a dump tag on every call.\n   * The frontend deduplicates executions with the same id/name (keeps last).\n   * Duplicate dump JSON is acceptable; only screenshots are deduplicated.\n   *\n   * All writes go through `fs/promises` so they run on libuv's thread pool\n   * rather than blocking the Node event loop. A long agent run previously\n   * appended multi-MB dumps (screenshots + serialized tasks) per progress\n   * tick on the main thread, starving IPC and UI for >10s per stall.\n   */\n  private async writeInlineExecution(\n    execution: ExecutionDump,\n    singleDump: ReportActionDump,\n  ): Promise<void> {\n    const dir = dirname(this.reportPath);\n    if (!existsSync(dir)) {\n      mkdirSync(dir, { recursive: true });\n    }\n\n    // Initialize: write HTML template once\n    if (!this.initialized) {\n      await writeFileAsync(this.reportPath, getReportTpl());\n      this.initialized = true;\n    }\n\n    // Append new screenshots (skip already-written ones)\n    for (const screenshot of execution.collectScreenshots()) {\n      await this.screenshotStore.persist(screenshot);\n    }\n\n    // Append dump tag (always — frontend keeps only last per execution id)\n    const serialized = singleDump.serialize();\n    await appendFileAsync(\n      this.reportPath,\n      `\\n${generateDumpScriptTag(serialized, this.getDumpScriptAttributes())}`,\n    );\n  }\n\n  private async writeDirectoryExecution(\n    execution: ExecutionDump,\n    singleDump: ReportActionDump,\n  ): Promise<void> {\n    const dir = dirname(this.reportPath);\n    if (!existsSync(dir)) {\n      mkdirSync(dir, { recursive: true });\n    }\n\n    for (const screenshot of execution.collectScreenshots()) {\n      await this.screenshotStore.persist(screenshot);\n    }\n\n    // 2. Append dump tag (always — frontend keeps only last per execution id)\n    const serialized = singleDump.serialize();\n\n    if (!this.initialized) {\n      await writeFileAsync(\n        this.reportPath,\n        `${getReportTpl()}${getBaseUrlFixScript()}`,\n      );\n      this.initialized = true;\n    }\n\n    await appendFileAsync(\n      this.reportPath,\n      `\\n${generateDumpScriptTag(serialized, this.getDumpScriptAttributes())}`,\n    );\n  }\n\n  private getExecutionLogKey(execution: ExecutionDump): string {\n    if (!execution.id) {\n      throw new Error(\n        'ReportGenerator: execution.id is required for persisting execution dumps',\n      );\n    }\n    return `id:${execution.id}`;\n  }\n\n  private async persistExecutionDumpToFile(\n    execution: ExecutionDump,\n    singleDump: ReportActionDump,\n  ): Promise<void> {\n    const dir = dirname(this.reportPath);\n    if (!existsSync(dir)) {\n      mkdirSync(dir, { recursive: true });\n    }\n\n    const executionLogKey = this.getExecutionLogKey(execution);\n    let fileIndex =\n      this.executionLogFileIndexByExecutionKey.get(executionLogKey);\n    if (!fileIndex) {\n      this.executionLogIndex += 1;\n      fileIndex = this.executionLogIndex;\n      this.executionLogFileIndexByExecutionKey.set(executionLogKey, fileIndex);\n    }\n\n    const fileName = `${fileIndex}.execution.json`;\n    const filePath = join(dirname(this.reportPath), fileName);\n    await writeFileAsync(filePath, singleDump.serialize(2), 'utf-8');\n  }\n}\n\nfunction ensureHtmlFileName(reportFileName: string): string {\n  return reportFileName.endsWith('.html')\n    ? reportFileName\n    : `${reportFileName}.html`;\n}\n\nfunction validateReportFileName(reportFileName: string): void {\n  if (!reportFileName?.trim()) {\n    throw new Error('reportFileName must be a non-empty string');\n  }\n\n  if (/[\\\\/]/.test(reportFileName)) {\n    throw new Error(\n      'reportFileName must not contain path separators (`/` or `\\\\\\\\`)',\n    );\n  }\n\n  if (/[:*?\"<>|]/.test(reportFileName)) {\n    throw new Error(\n      'reportFileName contains illegal filename characters: : * ? \" < > |',\n    );\n  }\n}\n"],"names":["nullReportGenerator","undefined","assertReportGenerationOptions","opts","Error","ReportGenerator","reportFileName","validateReportFileName","ifInBrowser","reportRootDir","getMidsceneRunSubDir","outputDir","join","reportPath","ensureHtmlFileName","execution","reportMeta","attributes","globalConfigManager","MIDSCENE_REPORT_QUIET","logMsg","dirname","singleDump","key","value","Object","String","existsSync","reportDir","existingExecutionIndices","readdirSync","name","index","Boolean","Number","Math","ReportActionDump","dir","mkdirSync","writeFileAsync","getReportTpl","screenshot","serialized","appendFileAsync","generateDumpScriptTag","getBaseUrlFixScript","executionLogKey","fileIndex","fileName","filePath","options","Map","Promise","uuid","ScreenshotStore","id","base64","generateImageScriptTag"],"mappings":";;;;;;;;;;AAQC;;;;;;;;;;AA4DM,MAAMA,sBAAwC;IACnD,mBAAmB,KAAO;IAC1B,OAAO,WAAa;IACpB,UAAU,UAAYC;IACtB,eAAe,IAAMA;AACvB;AAEO,SAASC,8BAA8BC,IAG7C;IACC,IAAIA,AAAwB,UAAxBA,KAAK,cAAc,IAAcA,AAA8B,SAA9BA,KAAK,oBAAoB,EAC5D,MAAM,IAAIC,MACR;AAGN;AAEO,MAAMC;IAsDX,OAAO,OACLC,cAAsB,EACtBH,IAMC,EACiB;QAClBD,8BAA8BC;QAC9BI,uBAAuBD;QACvB,IAAIH,AAAwB,UAAxBA,KAAK,cAAc,EAAY,OAAOH;QAG1C,IAAIQ,aAAa,OAAOR;QAExB,MAAMS,gBAAgBC,qBAAqB;QAC3C,MAAMC,YAAYC,KAAKH,eAAeH;QACtC,MAAMO,aACJV,AAAsB,+BAAtBA,KAAK,YAAY,GACbS,KAAKD,WAAW,gBAChBC,KAAKH,eAAeK,mBAAmBR;QAC7C,OAAO,IAAID,gBAAgB;YACzBQ;YACA,gBACEV,AAAsB,+BAAtBA,KAAK,YAAY,GACb,cACA;YACN,sBAAsBA,KAAK,oBAAoB;YAC/C,WAAWA,KAAK,kBAAkB;YAClC,qBAAqBA,KAAK,mBAAmB;QAC/C;IACF;IAEA,kBACEY,SAAwB,EACxBC,UAAsB,EACtBC,UAA6B,EACvB;QACN,IAAI,CAAC,aAAa,GAAGF;QACrB,IAAI,CAAC,cAAc,GAAGC;QACtB,IAAI,CAAC,qBAAqB,CAACC;QAC3B,IAAI,CAAC,UAAU,GAAG,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC;YACrC,IAAI,IAAI,CAAC,SAAS,EAAE;YACpB,MAAM,IAAI,CAAC,gBAAgB,CAACF,WAAWC;QACzC;IACF;IAEA,MAAM,QAAuB;QAC3B,MAAM,IAAI,CAAC,UAAU;IACvB;IAEA,MAAM,WAAwC;QAE5C,IAAI,IAAI,CAAC,aAAa,IAAI,IAAI,CAAC,cAAc,EAC3C,IAAI,CAAC,iBAAiB,CAAC,IAAI,CAAC,aAAa,EAAE,IAAI,CAAC,cAAc;QAEhE,MAAM,IAAI,CAAC,KAAK;QAChB,IAAI,CAAC,SAAS,GAAG;QAEjB,IAAI,CAAC,IAAI,CAAC,WAAW,EAEnB;QAGF,OAAO,IAAI,CAAC,UAAU;IACxB;IAEA,gBAAoC;QAClC,OAAO,IAAI,CAAC,UAAU;IACxB;IAEQ,kBAAwB;QAC9B,IAAI,CAAC,IAAI,CAAC,SAAS,IAAI,CAAC,IAAI,CAAC,UAAU,EAAE;QACzC,IAAIE,oBAAoB,qBAAqB,CAACC,wBAC5C;QAE0B,gBAAxB,IAAI,CAAC,cAAc,GACrBC,OACE,CAAC,0CAA0C,EAAEC,QAAQ,IAAI,CAAC,UAAU,GAAG,IAGzED,OAAO,CAAC,gCAAgC,EAAE,IAAI,CAAC,UAAU,EAAE;IAE/D;IAEA,MAAc,iBACZL,SAAwB,EACxBC,UAAsB,EACP;QACf,MAAMM,aAAa,IAAI,CAAC,gBAAgB,CAACP,WAAWC;QAEpD,IAAI,AAAwB,aAAxB,IAAI,CAAC,cAAc,EACrB,MAAM,IAAI,CAAC,oBAAoB,CAACD,WAAWO;aAE3C,MAAM,IAAI,CAAC,uBAAuB,CAACP,WAAWO;QAGhD,IAAI,IAAI,CAAC,0BAA0B,EACjC,MAAM,IAAI,CAAC,0BAA0B,CAACP,WAAWO;QAGnD,IAAI,CAAC,IAAI,CAAC,cAAc,EAAE;YACxB,IAAI,CAAC,cAAc,GAAG;YACtB,IAAI,CAAC,eAAe;QACtB;IACF;IAEQ,sBAAsBL,UAA6B,EAAQ;QACjE,IAAI,CAACA,YACH;QAGF,KAAK,MAAM,CAACM,KAAKC,MAAM,IAAIC,OAAO,OAAO,CAACR,YACxC,IAAIO,QAAAA,OAGJ,IAAI,CAAC,gBAAgB,CAACD,IAAI,GAAGG,OAAOF;IAExC;IAEQ,iCAAuC;QAC7C,IAAI,CAACG,WAAW,IAAI,CAAC,UAAU,GAC7B;QAIF,IAAI,CAAC,WAAW,GAAG;QAEnB,IAAI,CAAC,IAAI,CAAC,0BAA0B,EAClC;QAGF,MAAMC,YAAYP,QAAQ,IAAI,CAAC,UAAU;QACzC,MAAMQ,2BAA2BC,YAAYF,WAC1C,GAAG,CAAC,CAACG,OAAS,2BAA2B,IAAI,CAACA,OAAO,CAAC,EAAE,EACxD,MAAM,CAAC,CAACC,QAA2BC,QAAQD,QAC3C,GAAG,CAAC,CAACA,QAAUE,OAAO,QAAQ,CAACF,OAAO,KACtC,MAAM,CAAC,CAACA,QAAUE,OAAO,QAAQ,CAACF;QAErC,IAAIH,yBAAyB,MAAM,GAAG,GACpC,IAAI,CAAC,iBAAiB,GAAGM,KAAK,GAAG,IAAIN;IAEzC;IAEQ,0BAAkD;QACxD,OAAO;YACL,iBAAiB,IAAI,CAAC,cAAc;YACpC,GAAG,IAAI,CAAC,gBAAgB;QAC1B;IACF;IAKQ,iBACNd,SAAwB,EACxBC,UAAsB,EACJ;QAClB,OAAO,IAAIoB,iBAAiB;YAC1B,YAAYpB,WAAW,UAAU;YACjC,WAAWA,WAAW,SAAS;YAC/B,kBAAkBA,WAAW,gBAAgB;YAC7C,aAAaA,WAAW,WAAW;YACnC,YAAYA,WAAW,UAAU;YACjC,YAAY;gBAACD;aAAU;QACzB;IACF;IAYA,MAAc,qBACZA,SAAwB,EACxBO,UAA4B,EACb;QACf,MAAMe,MAAMhB,QAAQ,IAAI,CAAC,UAAU;QACnC,IAAI,CAACM,WAAWU,MACdC,UAAUD,KAAK;YAAE,WAAW;QAAK;QAInC,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE;YACrB,MAAME,UAAe,IAAI,CAAC,UAAU,EAAEC;YACtC,IAAI,CAAC,WAAW,GAAG;QACrB;QAGA,KAAK,MAAMC,cAAc1B,UAAU,kBAAkB,GACnD,MAAM,IAAI,CAAC,eAAe,CAAC,OAAO,CAAC0B;QAIrC,MAAMC,aAAapB,WAAW,SAAS;QACvC,MAAMqB,WACJ,IAAI,CAAC,UAAU,EACf,CAAC,EAAE,EAAEC,sBAAsBF,YAAY,IAAI,CAAC,uBAAuB,KAAK;IAE5E;IAEA,MAAc,wBACZ3B,SAAwB,EACxBO,UAA4B,EACb;QACf,MAAMe,MAAMhB,QAAQ,IAAI,CAAC,UAAU;QACnC,IAAI,CAACM,WAAWU,MACdC,UAAUD,KAAK;YAAE,WAAW;QAAK;QAGnC,KAAK,MAAMI,cAAc1B,UAAU,kBAAkB,GACnD,MAAM,IAAI,CAAC,eAAe,CAAC,OAAO,CAAC0B;QAIrC,MAAMC,aAAapB,WAAW,SAAS;QAEvC,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE;YACrB,MAAMiB,UACJ,IAAI,CAAC,UAAU,EACf,GAAGC,iBAAiBK,uBAAuB;YAE7C,IAAI,CAAC,WAAW,GAAG;QACrB;QAEA,MAAMF,WACJ,IAAI,CAAC,UAAU,EACf,CAAC,EAAE,EAAEC,sBAAsBF,YAAY,IAAI,CAAC,uBAAuB,KAAK;IAE5E;IAEQ,mBAAmB3B,SAAwB,EAAU;QAC3D,IAAI,CAACA,UAAU,EAAE,EACf,MAAM,IAAIX,MACR;QAGJ,OAAO,CAAC,GAAG,EAAEW,UAAU,EAAE,EAAE;IAC7B;IAEA,MAAc,2BACZA,SAAwB,EACxBO,UAA4B,EACb;QACf,MAAMe,MAAMhB,QAAQ,IAAI,CAAC,UAAU;QACnC,IAAI,CAACM,WAAWU,MACdC,UAAUD,KAAK;YAAE,WAAW;QAAK;QAGnC,MAAMS,kBAAkB,IAAI,CAAC,kBAAkB,CAAC/B;QAChD,IAAIgC,YACF,IAAI,CAAC,mCAAmC,CAAC,GAAG,CAACD;QAC/C,IAAI,CAACC,WAAW;YACd,IAAI,CAAC,iBAAiB,IAAI;YAC1BA,YAAY,IAAI,CAAC,iBAAiB;YAClC,IAAI,CAAC,mCAAmC,CAAC,GAAG,CAACD,iBAAiBC;QAChE;QAEA,MAAMC,WAAW,GAAGD,UAAU,eAAe,CAAC;QAC9C,MAAME,WAAWrC,KAAKS,QAAQ,IAAI,CAAC,UAAU,GAAG2B;QAChD,MAAMT,UAAeU,UAAU3B,WAAW,SAAS,CAAC,IAAI;IAC1D;IAzSA,YAAY4B,OAMX,CAAE;QA9BH,uBAAQ,cAAR;QACA,uBAAQ,kBAAR;QACA,uBAAQ,8BAAR;QACA,uBAAQ,aAAR;QACA,uBAAQ,kBAAiB;QACzB,uBAAQ,qBAAoB;QAC5B,uBAAQ,uCAAsC,IAAIC;QAGlD,uBAAiB,kBAAjB;QAGA,uBAAQ,mBAAR;QACA,uBAAQ,eAAc;QAGtB,uBAAQ,iBAAR;QACA,uBAAQ,kBAAR;QACA,uBAAQ,oBAA2C,CAAC;QAGpD,uBAAQ,cAA4BC,QAAQ,OAAO;QACnD,uBAAQ,aAAY;QASlB,IAAI,CAAC,UAAU,GAAGF,QAAQ,UAAU;QACpC,IAAI,CAAC,cAAc,GAAGA,QAAQ,cAAc;QAC5C,IAAI,CAAC,0BAA0B,GAAGA,QAAQ,oBAAoB,IAAI;QAClE,IAAI,CAAC,SAAS,GAAGA,QAAQ,SAAS,IAAI;QACtC,IAAI,CAAC,cAAc,GAAGG;QACtB,IAAI,CAAC,eAAe,GAAG,IAAIC,gBAAgB;YACzC,MAAM,AAAwB,aAAxB,IAAI,CAAC,cAAc,GAAgB,WAAW;YACpD,YAAY,IAAI,CAAC,UAAU;YAC3B,gBAAgB1C,KAAKS,QAAQ,IAAI,CAAC,UAAU,GAAG;YAC/C,kBAAkB,OAAOkC,IAAIC;gBAC3B,MAAMb,WACJ,IAAI,CAAC,UAAU,EACf,CAAC,EAAE,EAAEc,uBAAuBF,IAAIC,SAAS;YAE7C;YACA,mBAAmB,IAAI,CAAC,0BAA0B;QACpD;QACA,IAAIN,QAAQ,mBAAmB,EAC7B,IAAI,CAAC,8BAA8B;IAEvC;AA+QF;AAEA,SAASpC,mBAAmBR,cAAsB;IAChD,OAAOA,eAAe,QAAQ,CAAC,WAC3BA,iBACA,GAAGA,eAAe,KAAK,CAAC;AAC9B;AAEA,SAASC,uBAAuBD,cAAsB;IACpD,IAAI,CAACA,gBAAgB,QACnB,MAAM,IAAIF,MAAM;IAGlB,IAAI,QAAQ,IAAI,CAACE,iBACf,MAAM,IAAIF,MACR;IAIJ,IAAI,YAAY,IAAI,CAACE,iBACnB,MAAM,IAAIF,MACR;AAGN"}

package/dist/es/service/index.mjs

@@ -160,7 +160,7 @@ class Service {
             }
         };
     }
-    async extract(dataDemand, modelRuntime, opt, pageDescription, multimodalPrompt, context) {
+    async extract(dataDemand, modelRuntime, opt, pageDescription, multimodalPrompt, context, executionOptions) {
         assert(context, 'context is required for extract');
         assert('object' == typeof dataDemand || 'string' == typeof dataDemand, `dataDemand should be object or string, but get ${typeof dataDemand}`);
         const startTime = Date.now();
@@ -176,7 +176,8 @@ class Service {
                 multimodalPrompt,
                 extractOption: opt,
                 modelRuntime,
-                pageDescription
+                pageDescription,
+                abortSignal: executionOptions?.abortSignal
             });
             parseResult = result.parseResult;
             rawResponse = result.rawResponse;

package/dist/es/service/index.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"service/index.mjs","sources":["../../../src/service/index.ts"],"sourcesContent":["import { defaultModelFamilyRequiredForLocateMessage } from '@/ai-model/errors';\nimport {\n  AiExtractElementInfo,\n  AiLocateElement,\n  AiLocateSection,\n  buildSearchAreaConfig,\n} from '@/ai-model/inspect';\nimport type { ModelRuntime } from '@/ai-model/models';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport {\n  AIResponseParseError,\n  callAIWithObjectResponse,\n} from '@/ai-model/service-caller';\nimport type { AIArgs } from '@/ai-model/types';\nimport type { SearchAreaConfig } from '@/ai-model/workflows/inspect/types';\nimport { expandSearchArea } from '@/common';\nimport type {\n  AIDescribeElementResponse,\n  AIUsageInfo,\n  DetailedLocateParam,\n  LocateResultElement,\n  LocateResultWithDump,\n  PartialServiceDumpFromSDK,\n  PlanningLocateParam,\n  Rect,\n  ServiceExtractOption,\n  ServiceExtractParam,\n  ServiceExtractResult,\n  ServiceTaskInfo,\n  UIContext,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TMultimodalPrompt, TUserPrompt } from '../common';\nimport {\n  createServiceDump,\n  recoverDescribeResponseFromParseError,\n} from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext;\n  planLocatedElement?: LocateResultElement;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\ninterface ServiceOptions {\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n}\n\ninterface LocateSearchAreaResult {\n  config?: SearchAreaConfig;\n  trace: {\n    sourceRect?: Rect;\n    rawResponse?: string;\n    rawChoiceMessage?: unknown;\n    usage?: AIUsageInfo;\n  };\n}\n\nconst debug = getDebug('ai:service');\nexport default class Service {\n  contextRetrieverFn: () => Promise<UIContext> | UIContext;\n\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n\n  constructor(\n    context: UIContext | (() => Promise<UIContext> | UIContext),\n    opt?: ServiceOptions,\n  ) {\n    assert(context, 'context is required for Service');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    if (typeof opt?.taskInfo !== 'undefined') {\n      this.taskInfo = opt.taskInfo;\n    }\n  }\n\n  async locate(\n    query: PlanningLocateParam,\n    opt: LocateOpts,\n    modelRuntime: ModelRuntime,\n    abortSignal?: AbortSignal,\n  ): Promise<LocateResultWithDump> {\n    const { config: modelConfig } = modelRuntime;\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    if (!modelConfig.modelFamily) {\n      throw new Error(defaultModelFamilyRequiredForLocateMessage);\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn());\n\n    const searchArea = await this.resolveLocateSearchArea({\n      query,\n      queryPrompt,\n      opt,\n      context,\n      modelRuntime,\n      abortSignal,\n    });\n\n    const startTime = Date.now();\n    const {\n      parseResult,\n      rect,\n      rawResponse,\n      rawChoiceMessage,\n      usage,\n      reasoning_content,\n    } = await AiLocateElement({\n      context,\n      targetElementDescription: queryPrompt,\n      searchConfig: searchArea.config,\n      modelRuntime,\n      abortSignal,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      rawChoiceMessage,\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea: searchArea.trace.sourceRect,\n      searchAreaRawResponse: searchArea.trace.rawResponse,\n      searchAreaRawChoiceMessage: searchArea.trace.rawChoiceMessage,\n      searchAreaUsage: searchArea.trace.usage,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `failed to locate element: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedRect: rect,\n      data: null,\n      taskInfo,\n      deepLocate: !!searchArea.trace.sourceRect,\n      error: errorLog,\n    };\n\n    const element = parseResult.element;\n\n    const dump = createServiceDump({\n      ...dumpData,\n      matchedElement: element ? [element] : [],\n    });\n\n    if (errorLog) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    if (element) {\n      return {\n        element: {\n          center: element.center,\n          rect: element.rect,\n          description: element.description,\n        },\n        rect,\n        dump,\n      };\n    }\n\n    return {\n      element: null,\n      rect,\n      dump,\n    };\n  }\n\n  private async resolveLocateSearchArea(options: {\n    query: PlanningLocateParam;\n    queryPrompt: TUserPrompt;\n    opt: LocateOpts;\n    context: UIContext;\n    modelRuntime: ModelRuntime;\n    abortSignal?: AbortSignal;\n  }): Promise<LocateSearchAreaResult> {\n    const { query, queryPrompt, opt, context, modelRuntime, abortSignal } =\n      options;\n    const { adapter } = modelRuntime;\n    const hasPlanLocatedElement = !!opt?.planLocatedElement?.rect;\n\n    if (!query.deepLocate) {\n      return { trace: {} };\n    }\n\n    if (hasPlanLocatedElement) {\n      const config = await buildSearchAreaConfig({\n        context,\n        baseRect: opt.planLocatedElement!.rect,\n      });\n\n      return {\n        config,\n        trace: {\n          sourceRect: config.sourceRect,\n          rawResponse: JSON.stringify({\n            source: 'plan-located-element',\n            rect: opt.planLocatedElement!.rect,\n          }),\n        },\n      };\n    }\n\n    if (adapter.locate.supportsSearchArea) {\n      const searchAreaResponse = await AiLocateSection({\n        context,\n        sectionDescription: queryPrompt,\n        modelRuntime,\n        abortSignal,\n      });\n      const { searchAreaConfig } = searchAreaResponse;\n      assert(\n        searchAreaConfig,\n        `cannot find search area for \"${queryPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n\n      return {\n        config: searchAreaConfig,\n        trace: {\n          sourceRect: searchAreaConfig.sourceRect,\n          rawResponse: searchAreaResponse.rawResponse,\n          rawChoiceMessage: searchAreaResponse.rawChoiceMessage,\n          usage: searchAreaResponse.usage,\n        },\n      };\n    }\n\n    const firstPassLocateResult = await AiLocateElement({\n      context,\n      targetElementDescription: queryPrompt,\n      modelRuntime,\n      abortSignal,\n    });\n    assert(\n      firstPassLocateResult.rect,\n      `cannot find search area for \"${queryPrompt}\"${\n        firstPassLocateResult.parseResult.errors?.length\n          ? `: ${firstPassLocateResult.parseResult.errors.join('\\n')}`\n          : ''\n      }`,\n    );\n\n    const config = await buildSearchAreaConfig({\n      context,\n      baseRect: firstPassLocateResult.rect,\n    });\n\n    return {\n      config,\n      trace: {\n        sourceRect: config.sourceRect,\n        rawResponse: JSON.stringify({\n          source: 'deep-locate-first-pass',\n          rect: firstPassLocateResult.rect,\n          rawResponse: firstPassLocateResult.rawResponse,\n        }),\n        rawChoiceMessage: firstPassLocateResult.rawChoiceMessage,\n        usage: firstPassLocateResult.usage,\n      },\n    };\n  }\n\n  async extract<T>(\n    dataDemand: ServiceExtractParam,\n    modelRuntime: ModelRuntime,\n    opt?: ServiceExtractOption,\n    pageDescription?: string,\n    multimodalPrompt?: TMultimodalPrompt,\n    context?: UIContext,\n  ): Promise<ServiceExtractResult<T>> {\n    assert(context, 'context is required for extract');\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n\n    const startTime = Date.now();\n\n    let parseResult: Awaited<\n      ReturnType<typeof AiExtractElementInfo<T>>\n    >['parseResult'];\n    let rawResponse: string;\n    let rawChoiceMessage: unknown;\n    let usage: Awaited<ReturnType<typeof AiExtractElementInfo<T>>>['usage'];\n    let reasoning_content: string | undefined;\n\n    try {\n      const result = await AiExtractElementInfo<T>({\n        context,\n        dataQuery: dataDemand,\n        multimodalPrompt,\n        extractOption: opt,\n        modelRuntime,\n        pageDescription,\n      });\n      parseResult = result.parseResult;\n      rawResponse = result.rawResponse;\n      rawChoiceMessage = result.rawChoiceMessage;\n      usage = result.usage;\n      reasoning_content = result.reasoning_content;\n    } catch (error) {\n      if (error instanceof AIResponseParseError) {\n        // Create dump with usage and rawResponse from the error\n        const timeCost = Date.now() - startTime;\n        const taskInfo: ServiceTaskInfo = {\n          ...(this.taskInfo ? this.taskInfo : {}),\n          durationMs: timeCost,\n          rawResponse: error.rawResponse,\n          rawChoiceMessage: error.rawChoiceMessage,\n          usage: error.usage,\n        };\n        const dump = createServiceDump({\n          type: 'extract',\n          userQuery: { dataDemand },\n          data: null,\n          taskInfo,\n          error: error.message,\n        });\n        throw new ServiceError(error.message, dump);\n      }\n      throw error;\n    }\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse,\n      rawChoiceMessage,\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data, thought } = parseResult || {};\n\n    const dump = createServiceDump({\n      ...dumpData,\n      data,\n    });\n\n    if (errorLog && !data) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    return {\n      data,\n      thought,\n      usage,\n      reasoning_content,\n      dump,\n    };\n  }\n\n  async describe(\n    target: Rect | [number, number],\n    modelRuntime: ModelRuntime,\n    opt?: {\n      deepLocate?: boolean;\n      context?: UIContext;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for service.describe');\n    const context = opt?.context || (await this.contextRetrieverFn());\n    const { shotSize } = context;\n    const screenshotBase64 = context.screenshot.base64;\n    assert(screenshotBase64, 'screenshot is required for service.describe');\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      size: shotSize,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepLocate) {\n      const searchArea = expandSearchArea(targetRect, shotSize);\n      // Always crop in describe mode. Unlike locate's deepLocate (where\n      // cropping too small loses context for finding elements), describe's\n      // deepLocate intentionally zooms in so the model produces a more\n      // precise description from a focused view. expandSearchArea already\n      // guarantees a minimum 400x400 area with surrounding context.\n      // Describe is not a coordinate-parsing flow, so it does not need image\n      // padding for bbox normalization.\n      debug('describe: cropping to searchArea', searchArea);\n      const croppedResult = await cropByRect(imagePayload, searchArea);\n      imagePayload = croppedResult.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      },\n    ];\n\n    let res: Awaited<\n      ReturnType<typeof callAIWithObjectResponse<AIDescribeElementResponse>>\n    >;\n    try {\n      res = await callAIWithObjectResponse<AIDescribeElementResponse>(\n        msgs,\n        modelRuntime,\n      );\n    } catch (error) {\n      const recoveredResponse = recoverDescribeResponseFromParseError(error);\n      if (!recoveredResponse) {\n        throw error;\n      }\n      debug('describe: recovered malformed description JSON response');\n      return recoveredResponse;\n    }\n\n    const { content } = res;\n    assert(!content.error, `describe failed: ${content.error}`);\n    assert(content.description, 'failed to describe the element');\n    return content;\n  }\n}\n"],"names":["debug","getDebug","Service","query","opt","modelRuntime","abortSignal","modelConfig","queryPrompt","assert","Error","defaultModelFamilyRequiredForLocateMessage","context","searchArea","startTime","Date","parseResult","rect","rawResponse","rawChoiceMessage","usage","reasoning_content","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","element","dump","createServiceDump","ServiceError","options","adapter","hasPlanLocatedElement","config","buildSearchAreaConfig","searchAreaResponse","AiLocateSection","searchAreaConfig","firstPassLocateResult","dataDemand","pageDescription","multimodalPrompt","result","AiExtractElementInfo","error","AIResponseParseError","data","thought","target","shotSize","screenshotBase64","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","croppedResult","cropByRect","msgs","res","callAIWithObjectResponse","recoveredResponse","recoverDescribeResponseFromParseError","content","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;;AAgEA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IAqBnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,YAA0B,EAC1BC,WAAyB,EACM;QAC/B,MAAM,EAAE,QAAQC,WAAW,EAAE,GAAGF;QAChC,MAAMG,cAAc,AAAiB,YAAjB,OAAOL,QAAqBA,QAAQA,MAAM,MAAM;QACpEM,OAAOD,aAAa;QAEpBC,OAAO,AAAiB,YAAjB,OAAON,OAAoB;QAElC,IAAI,CAACI,YAAY,WAAW,EAC1B,MAAM,IAAIG,MAAMC;QAGlB,MAAMC,UAAUR,KAAK,WAAY,MAAM,IAAI,CAAC,kBAAkB;QAE9D,MAAMS,aAAa,MAAM,IAAI,CAAC,uBAAuB,CAAC;YACpDV;YACAK;YACAJ;YACAQ;YACAP;YACAC;QACF;QAEA,MAAMQ,YAAYC,KAAK,GAAG;QAC1B,MAAM,EACJC,WAAW,EACXC,IAAI,EACJC,WAAW,EACXC,gBAAgB,EAChBC,KAAK,EACLC,iBAAiB,EAClB,GAAG,MAAMC,gBAAgB;YACxBV;YACA,0BAA0BJ;YAC1B,cAAcK,WAAW,MAAM;YAC/BR;YACAC;QACF;QAEA,MAAMiB,WAAWR,KAAK,GAAG,KAAKD;QAC9B,MAAMU,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACP;YAC5BC;YACA,gBAAgBM,KAAK,SAAS,CAACT;YAC/BI;YACA,YAAYP,WAAW,KAAK,CAAC,UAAU;YACvC,uBAAuBA,WAAW,KAAK,CAAC,WAAW;YACnD,4BAA4BA,WAAW,KAAK,CAAC,gBAAgB;YAC7D,iBAAiBA,WAAW,KAAK,CAAC,KAAK;YACvCQ;QACF;QAEA,IAAIK;QACJ,IAAIV,YAAY,MAAM,EAAE,QACtBU,WAAW,CAAC,4BAA4B,EAAEV,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG3E,MAAMW,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAASnB;YACX;YACA,aAAaS;YACb,MAAM;YACNO;YACA,YAAY,CAAC,CAACX,WAAW,KAAK,CAAC,UAAU;YACzC,OAAOa;QACT;QAEA,MAAME,UAAUZ,YAAY,OAAO;QAEnC,MAAMa,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACX,gBAAgBC,UAAU;gBAACA;aAAQ,GAAG,EAAE;QAC1C;QAEA,IAAIF,UACF,MAAM,IAAIK,aAAaL,UAAUG;QAGnC,IAAID,SACF,OAAO;YACL,SAAS;gBACP,QAAQA,QAAQ,MAAM;gBACtB,MAAMA,QAAQ,IAAI;gBAClB,aAAaA,QAAQ,WAAW;YAClC;YACAX;YACAY;QACF;QAGF,OAAO;YACL,SAAS;YACTZ;YACAY;QACF;IACF;IAEA,MAAc,wBAAwBG,OAOrC,EAAmC;QAClC,MAAM,EAAE7B,KAAK,EAAEK,WAAW,EAAEJ,GAAG,EAAEQ,OAAO,EAAEP,YAAY,EAAEC,WAAW,EAAE,GACnE0B;QACF,MAAM,EAAEC,OAAO,EAAE,GAAG5B;QACpB,MAAM6B,wBAAwB,CAAC,CAAC9B,KAAK,oBAAoB;QAEzD,IAAI,CAACD,MAAM,UAAU,EACnB,OAAO;YAAE,OAAO,CAAC;QAAE;QAGrB,IAAI+B,uBAAuB;YACzB,MAAMC,SAAS,MAAMC,sBAAsB;gBACzCxB;gBACA,UAAUR,IAAI,kBAAkB,CAAE,IAAI;YACxC;YAEA,OAAO;gBACL+B;gBACA,OAAO;oBACL,YAAYA,OAAO,UAAU;oBAC7B,aAAaV,KAAK,SAAS,CAAC;wBAC1B,QAAQ;wBACR,MAAMrB,IAAI,kBAAkB,CAAE,IAAI;oBACpC;gBACF;YACF;QACF;QAEA,IAAI6B,QAAQ,MAAM,CAAC,kBAAkB,EAAE;YACrC,MAAMI,qBAAqB,MAAMC,gBAAgB;gBAC/C1B;gBACA,oBAAoBJ;gBACpBH;gBACAC;YACF;YACA,MAAM,EAAEiC,gBAAgB,EAAE,GAAGF;YAC7B5B,OACE8B,kBACA,CAAC,6BAA6B,EAAE/B,YAAY,CAAC,EAC3C6B,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAGJ,OAAO;gBACL,QAAQE;gBACR,OAAO;oBACL,YAAYA,iBAAiB,UAAU;oBACvC,aAAaF,mBAAmB,WAAW;oBAC3C,kBAAkBA,mBAAmB,gBAAgB;oBACrD,OAAOA,mBAAmB,KAAK;gBACjC;YACF;QACF;QAEA,MAAMG,wBAAwB,MAAMlB,gBAAgB;YAClDV;YACA,0BAA0BJ;YAC1BH;YACAC;QACF;QACAG,OACE+B,sBAAsB,IAAI,EAC1B,CAAC,6BAA6B,EAAEhC,YAAY,CAAC,EAC3CgC,sBAAsB,WAAW,CAAC,MAAM,EAAE,SACtC,CAAC,EAAE,EAAEA,sBAAsB,WAAW,CAAC,MAAM,CAAC,IAAI,CAAC,OAAO,GAC1D,IACJ;QAGJ,MAAML,SAAS,MAAMC,sBAAsB;YACzCxB;YACA,UAAU4B,sBAAsB,IAAI;QACtC;QAEA,OAAO;YACLL;YACA,OAAO;gBACL,YAAYA,OAAO,UAAU;gBAC7B,aAAaV,KAAK,SAAS,CAAC;oBAC1B,QAAQ;oBACR,MAAMe,sBAAsB,IAAI;oBAChC,aAAaA,sBAAsB,WAAW;gBAChD;gBACA,kBAAkBA,sBAAsB,gBAAgB;gBACxD,OAAOA,sBAAsB,KAAK;YACpC;QACF;IACF;IAEA,MAAM,QACJC,UAA+B,EAC/BpC,YAA0B,EAC1BD,GAA0B,EAC1BsC,eAAwB,EACxBC,gBAAoC,EACpC/B,OAAmB,EACe;QAClCH,OAAOG,SAAS;QAChBH,OACE,AAAsB,YAAtB,OAAOgC,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAGvE,MAAM3B,YAAYC,KAAK,GAAG;QAE1B,IAAIC;QAGJ,IAAIE;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAEJ,IAAI;YACF,MAAMuB,SAAS,MAAMC,qBAAwB;gBAC3CjC;gBACA,WAAW6B;gBACXE;gBACA,eAAevC;gBACfC;gBACAqC;YACF;YACA1B,cAAc4B,OAAO,WAAW;YAChC1B,cAAc0B,OAAO,WAAW;YAChCzB,mBAAmByB,OAAO,gBAAgB;YAC1CxB,QAAQwB,OAAO,KAAK;YACpBvB,oBAAoBuB,OAAO,iBAAiB;QAC9C,EAAE,OAAOE,OAAO;YACd,IAAIA,iBAAiBC,sBAAsB;gBAEzC,MAAMxB,WAAWR,KAAK,GAAG,KAAKD;gBAC9B,MAAMU,WAA4B;oBAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;oBACtC,YAAYD;oBACZ,aAAauB,MAAM,WAAW;oBAC9B,kBAAkBA,MAAM,gBAAgB;oBACxC,OAAOA,MAAM,KAAK;gBACpB;gBACA,MAAMjB,OAAOC,kBAAkB;oBAC7B,MAAM;oBACN,WAAW;wBAAEW;oBAAW;oBACxB,MAAM;oBACNjB;oBACA,OAAOsB,MAAM,OAAO;gBACtB;gBACA,MAAM,IAAIf,aAAae,MAAM,OAAO,EAAEjB;YACxC;YACA,MAAMiB;QACR;QAEA,MAAMvB,WAAWR,KAAK,GAAG,KAAKD;QAC9B,MAAMU,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZL;YACAC;YACA,gBAAgBM,KAAK,SAAS,CAACT;YAC/BI;YACAC;QACF;QAEA,IAAIK;QACJ,IAAIV,YAAY,MAAM,EAAE,QACtBU,WAAW,CAAC,qBAAqB,EAAEV,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMW,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTc;YACF;YACA,MAAM;YACNjB;YACA,OAAOE;QACT;QAEA,MAAM,EAAEsB,IAAI,EAAEC,OAAO,EAAE,GAAGjC,eAAe,CAAC;QAE1C,MAAMa,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACXqB;QACF;QAEA,IAAItB,YAAY,CAACsB,MACf,MAAM,IAAIjB,aAAaL,UAAUG;QAGnC,OAAO;YACLmB;YACAC;YACA7B;YACAC;YACAQ;QACF;IACF;IAEA,MAAM,SACJqB,MAA+B,EAC/B7C,YAA0B,EAC1BD,GAGC,EACwD;QACzDK,OAAOyC,QAAQ;QACf,MAAMtC,UAAUR,KAAK,WAAY,MAAM,IAAI,CAAC,kBAAkB;QAC9D,MAAM,EAAE+C,QAAQ,EAAE,GAAGvC;QACrB,MAAMwC,mBAAmBxC,QAAQ,UAAU,CAAC,MAAM;QAClDH,OAAO2C,kBAAkB;QACzB,MAAMC,eAAeC;QAGrB,MAAMC,kBAAkB;QACxB,MAAMC,aAAmBC,MAAM,OAAO,CAACP,UACnC;YACE,MAAMQ,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC/C,KAAKG,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC9C,OAAOA;YACP,QAAQA;QACV,IACAL;QAEJ,IAAIS,eAAe,MAAMC,wBAAwB;YAC/C,gBAAgBR;YAChB,MAAMD;YACN,sBAAsB;gBACpB;oBACE,MAAMK;gBACR;aACD;YACD,iBAAiB;QACnB;QAEA,IAAIpD,KAAK,YAAY;YACnB,MAAMS,aAAagD,iBAAiBL,YAAYL;YAQhDnD,MAAM,oCAAoCa;YAC1C,MAAMiD,gBAAgB,MAAMC,WAAWJ,cAAc9C;YACrD8C,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,IAAIM;QAGJ,IAAI;YACFA,MAAM,MAAMC,yBACVF,MACA3D;QAEJ,EAAE,OAAOyC,OAAO;YACd,MAAMqB,oBAAoBC,sCAAsCtB;YAChE,IAAI,CAACqB,mBACH,MAAMrB;YAER9C,MAAM;YACN,OAAOmE;QACT;QAEA,MAAM,EAAEE,OAAO,EAAE,GAAGJ;QACpBxD,OAAO,CAAC4D,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1D5D,OAAO4D,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IA5ZA,YACEzD,OAA2D,EAC3DR,GAAoB,CACpB;QAPF;QAEA;QAMEK,OAAOG,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAM0D,QAAQ,OAAO,CAAC1D;QAGlD,IAAI,AAAyB,WAAlBR,KAAK,UACd,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AA+YF"}
+{"version":3,"file":"service/index.mjs","sources":["../../../src/service/index.ts"],"sourcesContent":["import { defaultModelFamilyRequiredForLocateMessage } from '@/ai-model/errors';\nimport {\n  AiExtractElementInfo,\n  AiLocateElement,\n  AiLocateSection,\n  buildSearchAreaConfig,\n} from '@/ai-model/inspect';\nimport type { ModelRuntime } from '@/ai-model/models';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport {\n  AIResponseParseError,\n  callAIWithObjectResponse,\n} from '@/ai-model/service-caller';\nimport type { AIArgs } from '@/ai-model/types';\nimport type { SearchAreaConfig } from '@/ai-model/workflows/inspect/types';\nimport { expandSearchArea } from '@/common';\nimport type {\n  AIDescribeElementResponse,\n  AIUsageInfo,\n  DetailedLocateParam,\n  LocateResultElement,\n  LocateResultWithDump,\n  PartialServiceDumpFromSDK,\n  PlanningLocateParam,\n  Rect,\n  ServiceExtractOption,\n  ServiceExtractParam,\n  ServiceExtractResult,\n  ServiceTaskInfo,\n  UIContext,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TMultimodalPrompt, TUserPrompt } from '../common';\nimport {\n  createServiceDump,\n  recoverDescribeResponseFromParseError,\n} from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext;\n  planLocatedElement?: LocateResultElement;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\ninterface ServiceOptions {\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n}\n\ninterface LocateSearchAreaResult {\n  config?: SearchAreaConfig;\n  trace: {\n    sourceRect?: Rect;\n    rawResponse?: string;\n    rawChoiceMessage?: unknown;\n    usage?: AIUsageInfo;\n  };\n}\n\nconst debug = getDebug('ai:service');\nexport default class Service {\n  contextRetrieverFn: () => Promise<UIContext> | UIContext;\n\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n\n  constructor(\n    context: UIContext | (() => Promise<UIContext> | UIContext),\n    opt?: ServiceOptions,\n  ) {\n    assert(context, 'context is required for Service');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    if (typeof opt?.taskInfo !== 'undefined') {\n      this.taskInfo = opt.taskInfo;\n    }\n  }\n\n  async locate(\n    query: PlanningLocateParam,\n    opt: LocateOpts,\n    modelRuntime: ModelRuntime,\n    abortSignal?: AbortSignal,\n  ): Promise<LocateResultWithDump> {\n    const { config: modelConfig } = modelRuntime;\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    if (!modelConfig.modelFamily) {\n      throw new Error(defaultModelFamilyRequiredForLocateMessage);\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn());\n\n    const searchArea = await this.resolveLocateSearchArea({\n      query,\n      queryPrompt,\n      opt,\n      context,\n      modelRuntime,\n      abortSignal,\n    });\n\n    const startTime = Date.now();\n    const {\n      parseResult,\n      rect,\n      rawResponse,\n      rawChoiceMessage,\n      usage,\n      reasoning_content,\n    } = await AiLocateElement({\n      context,\n      targetElementDescription: queryPrompt,\n      searchConfig: searchArea.config,\n      modelRuntime,\n      abortSignal,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      rawChoiceMessage,\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea: searchArea.trace.sourceRect,\n      searchAreaRawResponse: searchArea.trace.rawResponse,\n      searchAreaRawChoiceMessage: searchArea.trace.rawChoiceMessage,\n      searchAreaUsage: searchArea.trace.usage,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `failed to locate element: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedRect: rect,\n      data: null,\n      taskInfo,\n      deepLocate: !!searchArea.trace.sourceRect,\n      error: errorLog,\n    };\n\n    const element = parseResult.element;\n\n    const dump = createServiceDump({\n      ...dumpData,\n      matchedElement: element ? [element] : [],\n    });\n\n    if (errorLog) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    if (element) {\n      return {\n        element: {\n          center: element.center,\n          rect: element.rect,\n          description: element.description,\n        },\n        rect,\n        dump,\n      };\n    }\n\n    return {\n      element: null,\n      rect,\n      dump,\n    };\n  }\n\n  private async resolveLocateSearchArea(options: {\n    query: PlanningLocateParam;\n    queryPrompt: TUserPrompt;\n    opt: LocateOpts;\n    context: UIContext;\n    modelRuntime: ModelRuntime;\n    abortSignal?: AbortSignal;\n  }): Promise<LocateSearchAreaResult> {\n    const { query, queryPrompt, opt, context, modelRuntime, abortSignal } =\n      options;\n    const { adapter } = modelRuntime;\n    const hasPlanLocatedElement = !!opt?.planLocatedElement?.rect;\n\n    if (!query.deepLocate) {\n      return { trace: {} };\n    }\n\n    if (hasPlanLocatedElement) {\n      const config = await buildSearchAreaConfig({\n        context,\n        baseRect: opt.planLocatedElement!.rect,\n      });\n\n      return {\n        config,\n        trace: {\n          sourceRect: config.sourceRect,\n          rawResponse: JSON.stringify({\n            source: 'plan-located-element',\n            rect: opt.planLocatedElement!.rect,\n          }),\n        },\n      };\n    }\n\n    if (adapter.locate.supportsSearchArea) {\n      const searchAreaResponse = await AiLocateSection({\n        context,\n        sectionDescription: queryPrompt,\n        modelRuntime,\n        abortSignal,\n      });\n      const { searchAreaConfig } = searchAreaResponse;\n      assert(\n        searchAreaConfig,\n        `cannot find search area for \"${queryPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n\n      return {\n        config: searchAreaConfig,\n        trace: {\n          sourceRect: searchAreaConfig.sourceRect,\n          rawResponse: searchAreaResponse.rawResponse,\n          rawChoiceMessage: searchAreaResponse.rawChoiceMessage,\n          usage: searchAreaResponse.usage,\n        },\n      };\n    }\n\n    const firstPassLocateResult = await AiLocateElement({\n      context,\n      targetElementDescription: queryPrompt,\n      modelRuntime,\n      abortSignal,\n    });\n    assert(\n      firstPassLocateResult.rect,\n      `cannot find search area for \"${queryPrompt}\"${\n        firstPassLocateResult.parseResult.errors?.length\n          ? `: ${firstPassLocateResult.parseResult.errors.join('\\n')}`\n          : ''\n      }`,\n    );\n\n    const config = await buildSearchAreaConfig({\n      context,\n      baseRect: firstPassLocateResult.rect,\n    });\n\n    return {\n      config,\n      trace: {\n        sourceRect: config.sourceRect,\n        rawResponse: JSON.stringify({\n          source: 'deep-locate-first-pass',\n          rect: firstPassLocateResult.rect,\n          rawResponse: firstPassLocateResult.rawResponse,\n        }),\n        rawChoiceMessage: firstPassLocateResult.rawChoiceMessage,\n        usage: firstPassLocateResult.usage,\n      },\n    };\n  }\n\n  async extract<T>(\n    dataDemand: ServiceExtractParam,\n    modelRuntime: ModelRuntime,\n    opt?: ServiceExtractOption,\n    pageDescription?: string,\n    multimodalPrompt?: TMultimodalPrompt,\n    context?: UIContext,\n    executionOptions?: {\n      abortSignal?: AbortSignal;\n    },\n  ): Promise<ServiceExtractResult<T>> {\n    assert(context, 'context is required for extract');\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n\n    const startTime = Date.now();\n\n    let parseResult: Awaited<\n      ReturnType<typeof AiExtractElementInfo<T>>\n    >['parseResult'];\n    let rawResponse: string;\n    let rawChoiceMessage: unknown;\n    let usage: Awaited<ReturnType<typeof AiExtractElementInfo<T>>>['usage'];\n    let reasoning_content: string | undefined;\n\n    try {\n      const result = await AiExtractElementInfo<T>({\n        context,\n        dataQuery: dataDemand,\n        multimodalPrompt,\n        extractOption: opt,\n        modelRuntime,\n        pageDescription,\n        abortSignal: executionOptions?.abortSignal,\n      });\n      parseResult = result.parseResult;\n      rawResponse = result.rawResponse;\n      rawChoiceMessage = result.rawChoiceMessage;\n      usage = result.usage;\n      reasoning_content = result.reasoning_content;\n    } catch (error) {\n      if (error instanceof AIResponseParseError) {\n        // Create dump with usage and rawResponse from the error\n        const timeCost = Date.now() - startTime;\n        const taskInfo: ServiceTaskInfo = {\n          ...(this.taskInfo ? this.taskInfo : {}),\n          durationMs: timeCost,\n          rawResponse: error.rawResponse,\n          rawChoiceMessage: error.rawChoiceMessage,\n          usage: error.usage,\n        };\n        const dump = createServiceDump({\n          type: 'extract',\n          userQuery: { dataDemand },\n          data: null,\n          taskInfo,\n          error: error.message,\n        });\n        throw new ServiceError(error.message, dump);\n      }\n      throw error;\n    }\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse,\n      rawChoiceMessage,\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data, thought } = parseResult || {};\n\n    const dump = createServiceDump({\n      ...dumpData,\n      data,\n    });\n\n    if (errorLog && !data) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    return {\n      data,\n      thought,\n      usage,\n      reasoning_content,\n      dump,\n    };\n  }\n\n  async describe(\n    target: Rect | [number, number],\n    modelRuntime: ModelRuntime,\n    opt?: {\n      deepLocate?: boolean;\n      context?: UIContext;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for service.describe');\n    const context = opt?.context || (await this.contextRetrieverFn());\n    const { shotSize } = context;\n    const screenshotBase64 = context.screenshot.base64;\n    assert(screenshotBase64, 'screenshot is required for service.describe');\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      size: shotSize,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepLocate) {\n      const searchArea = expandSearchArea(targetRect, shotSize);\n      // Always crop in describe mode. Unlike locate's deepLocate (where\n      // cropping too small loses context for finding elements), describe's\n      // deepLocate intentionally zooms in so the model produces a more\n      // precise description from a focused view. expandSearchArea already\n      // guarantees a minimum 400x400 area with surrounding context.\n      // Describe is not a coordinate-parsing flow, so it does not need image\n      // padding for bbox normalization.\n      debug('describe: cropping to searchArea', searchArea);\n      const croppedResult = await cropByRect(imagePayload, searchArea);\n      imagePayload = croppedResult.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      },\n    ];\n\n    let res: Awaited<\n      ReturnType<typeof callAIWithObjectResponse<AIDescribeElementResponse>>\n    >;\n    try {\n      res = await callAIWithObjectResponse<AIDescribeElementResponse>(\n        msgs,\n        modelRuntime,\n      );\n    } catch (error) {\n      const recoveredResponse = recoverDescribeResponseFromParseError(error);\n      if (!recoveredResponse) {\n        throw error;\n      }\n      debug('describe: recovered malformed description JSON response');\n      return recoveredResponse;\n    }\n\n    const { content } = res;\n    assert(!content.error, `describe failed: ${content.error}`);\n    assert(content.description, 'failed to describe the element');\n    return content;\n  }\n}\n"],"names":["debug","getDebug","Service","query","opt","modelRuntime","abortSignal","modelConfig","queryPrompt","assert","Error","defaultModelFamilyRequiredForLocateMessage","context","searchArea","startTime","Date","parseResult","rect","rawResponse","rawChoiceMessage","usage","reasoning_content","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","element","dump","createServiceDump","ServiceError","options","adapter","hasPlanLocatedElement","config","buildSearchAreaConfig","searchAreaResponse","AiLocateSection","searchAreaConfig","firstPassLocateResult","dataDemand","pageDescription","multimodalPrompt","executionOptions","result","AiExtractElementInfo","error","AIResponseParseError","data","thought","target","shotSize","screenshotBase64","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","croppedResult","cropByRect","msgs","res","callAIWithObjectResponse","recoveredResponse","recoverDescribeResponseFromParseError","content","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;;AAgEA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IAqBnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,YAA0B,EAC1BC,WAAyB,EACM;QAC/B,MAAM,EAAE,QAAQC,WAAW,EAAE,GAAGF;QAChC,MAAMG,cAAc,AAAiB,YAAjB,OAAOL,QAAqBA,QAAQA,MAAM,MAAM;QACpEM,OAAOD,aAAa;QAEpBC,OAAO,AAAiB,YAAjB,OAAON,OAAoB;QAElC,IAAI,CAACI,YAAY,WAAW,EAC1B,MAAM,IAAIG,MAAMC;QAGlB,MAAMC,UAAUR,KAAK,WAAY,MAAM,IAAI,CAAC,kBAAkB;QAE9D,MAAMS,aAAa,MAAM,IAAI,CAAC,uBAAuB,CAAC;YACpDV;YACAK;YACAJ;YACAQ;YACAP;YACAC;QACF;QAEA,MAAMQ,YAAYC,KAAK,GAAG;QAC1B,MAAM,EACJC,WAAW,EACXC,IAAI,EACJC,WAAW,EACXC,gBAAgB,EAChBC,KAAK,EACLC,iBAAiB,EAClB,GAAG,MAAMC,gBAAgB;YACxBV;YACA,0BAA0BJ;YAC1B,cAAcK,WAAW,MAAM;YAC/BR;YACAC;QACF;QAEA,MAAMiB,WAAWR,KAAK,GAAG,KAAKD;QAC9B,MAAMU,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACP;YAC5BC;YACA,gBAAgBM,KAAK,SAAS,CAACT;YAC/BI;YACA,YAAYP,WAAW,KAAK,CAAC,UAAU;YACvC,uBAAuBA,WAAW,KAAK,CAAC,WAAW;YACnD,4BAA4BA,WAAW,KAAK,CAAC,gBAAgB;YAC7D,iBAAiBA,WAAW,KAAK,CAAC,KAAK;YACvCQ;QACF;QAEA,IAAIK;QACJ,IAAIV,YAAY,MAAM,EAAE,QACtBU,WAAW,CAAC,4BAA4B,EAAEV,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG3E,MAAMW,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAASnB;YACX;YACA,aAAaS;YACb,MAAM;YACNO;YACA,YAAY,CAAC,CAACX,WAAW,KAAK,CAAC,UAAU;YACzC,OAAOa;QACT;QAEA,MAAME,UAAUZ,YAAY,OAAO;QAEnC,MAAMa,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACX,gBAAgBC,UAAU;gBAACA;aAAQ,GAAG,EAAE;QAC1C;QAEA,IAAIF,UACF,MAAM,IAAIK,aAAaL,UAAUG;QAGnC,IAAID,SACF,OAAO;YACL,SAAS;gBACP,QAAQA,QAAQ,MAAM;gBACtB,MAAMA,QAAQ,IAAI;gBAClB,aAAaA,QAAQ,WAAW;YAClC;YACAX;YACAY;QACF;QAGF,OAAO;YACL,SAAS;YACTZ;YACAY;QACF;IACF;IAEA,MAAc,wBAAwBG,OAOrC,EAAmC;QAClC,MAAM,EAAE7B,KAAK,EAAEK,WAAW,EAAEJ,GAAG,EAAEQ,OAAO,EAAEP,YAAY,EAAEC,WAAW,EAAE,GACnE0B;QACF,MAAM,EAAEC,OAAO,EAAE,GAAG5B;QACpB,MAAM6B,wBAAwB,CAAC,CAAC9B,KAAK,oBAAoB;QAEzD,IAAI,CAACD,MAAM,UAAU,EACnB,OAAO;YAAE,OAAO,CAAC;QAAE;QAGrB,IAAI+B,uBAAuB;YACzB,MAAMC,SAAS,MAAMC,sBAAsB;gBACzCxB;gBACA,UAAUR,IAAI,kBAAkB,CAAE,IAAI;YACxC;YAEA,OAAO;gBACL+B;gBACA,OAAO;oBACL,YAAYA,OAAO,UAAU;oBAC7B,aAAaV,KAAK,SAAS,CAAC;wBAC1B,QAAQ;wBACR,MAAMrB,IAAI,kBAAkB,CAAE,IAAI;oBACpC;gBACF;YACF;QACF;QAEA,IAAI6B,QAAQ,MAAM,CAAC,kBAAkB,EAAE;YACrC,MAAMI,qBAAqB,MAAMC,gBAAgB;gBAC/C1B;gBACA,oBAAoBJ;gBACpBH;gBACAC;YACF;YACA,MAAM,EAAEiC,gBAAgB,EAAE,GAAGF;YAC7B5B,OACE8B,kBACA,CAAC,6BAA6B,EAAE/B,YAAY,CAAC,EAC3C6B,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAGJ,OAAO;gBACL,QAAQE;gBACR,OAAO;oBACL,YAAYA,iBAAiB,UAAU;oBACvC,aAAaF,mBAAmB,WAAW;oBAC3C,kBAAkBA,mBAAmB,gBAAgB;oBACrD,OAAOA,mBAAmB,KAAK;gBACjC;YACF;QACF;QAEA,MAAMG,wBAAwB,MAAMlB,gBAAgB;YAClDV;YACA,0BAA0BJ;YAC1BH;YACAC;QACF;QACAG,OACE+B,sBAAsB,IAAI,EAC1B,CAAC,6BAA6B,EAAEhC,YAAY,CAAC,EAC3CgC,sBAAsB,WAAW,CAAC,MAAM,EAAE,SACtC,CAAC,EAAE,EAAEA,sBAAsB,WAAW,CAAC,MAAM,CAAC,IAAI,CAAC,OAAO,GAC1D,IACJ;QAGJ,MAAML,SAAS,MAAMC,sBAAsB;YACzCxB;YACA,UAAU4B,sBAAsB,IAAI;QACtC;QAEA,OAAO;YACLL;YACA,OAAO;gBACL,YAAYA,OAAO,UAAU;gBAC7B,aAAaV,KAAK,SAAS,CAAC;oBAC1B,QAAQ;oBACR,MAAMe,sBAAsB,IAAI;oBAChC,aAAaA,sBAAsB,WAAW;gBAChD;gBACA,kBAAkBA,sBAAsB,gBAAgB;gBACxD,OAAOA,sBAAsB,KAAK;YACpC;QACF;IACF;IAEA,MAAM,QACJC,UAA+B,EAC/BpC,YAA0B,EAC1BD,GAA0B,EAC1BsC,eAAwB,EACxBC,gBAAoC,EACpC/B,OAAmB,EACnBgC,gBAEC,EACiC;QAClCnC,OAAOG,SAAS;QAChBH,OACE,AAAsB,YAAtB,OAAOgC,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAGvE,MAAM3B,YAAYC,KAAK,GAAG;QAE1B,IAAIC;QAGJ,IAAIE;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAEJ,IAAI;YACF,MAAMwB,SAAS,MAAMC,qBAAwB;gBAC3ClC;gBACA,WAAW6B;gBACXE;gBACA,eAAevC;gBACfC;gBACAqC;gBACA,aAAaE,kBAAkB;YACjC;YACA5B,cAAc6B,OAAO,WAAW;YAChC3B,cAAc2B,OAAO,WAAW;YAChC1B,mBAAmB0B,OAAO,gBAAgB;YAC1CzB,QAAQyB,OAAO,KAAK;YACpBxB,oBAAoBwB,OAAO,iBAAiB;QAC9C,EAAE,OAAOE,OAAO;YACd,IAAIA,iBAAiBC,sBAAsB;gBAEzC,MAAMzB,WAAWR,KAAK,GAAG,KAAKD;gBAC9B,MAAMU,WAA4B;oBAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;oBACtC,YAAYD;oBACZ,aAAawB,MAAM,WAAW;oBAC9B,kBAAkBA,MAAM,gBAAgB;oBACxC,OAAOA,MAAM,KAAK;gBACpB;gBACA,MAAMlB,OAAOC,kBAAkB;oBAC7B,MAAM;oBACN,WAAW;wBAAEW;oBAAW;oBACxB,MAAM;oBACNjB;oBACA,OAAOuB,MAAM,OAAO;gBACtB;gBACA,MAAM,IAAIhB,aAAagB,MAAM,OAAO,EAAElB;YACxC;YACA,MAAMkB;QACR;QAEA,MAAMxB,WAAWR,KAAK,GAAG,KAAKD;QAC9B,MAAMU,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZL;YACAC;YACA,gBAAgBM,KAAK,SAAS,CAACT;YAC/BI;YACAC;QACF;QAEA,IAAIK;QACJ,IAAIV,YAAY,MAAM,EAAE,QACtBU,WAAW,CAAC,qBAAqB,EAAEV,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMW,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTc;YACF;YACA,MAAM;YACNjB;YACA,OAAOE;QACT;QAEA,MAAM,EAAEuB,IAAI,EAAEC,OAAO,EAAE,GAAGlC,eAAe,CAAC;QAE1C,MAAMa,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACXsB;QACF;QAEA,IAAIvB,YAAY,CAACuB,MACf,MAAM,IAAIlB,aAAaL,UAAUG;QAGnC,OAAO;YACLoB;YACAC;YACA9B;YACAC;YACAQ;QACF;IACF;IAEA,MAAM,SACJsB,MAA+B,EAC/B9C,YAA0B,EAC1BD,GAGC,EACwD;QACzDK,OAAO0C,QAAQ;QACf,MAAMvC,UAAUR,KAAK,WAAY,MAAM,IAAI,CAAC,kBAAkB;QAC9D,MAAM,EAAEgD,QAAQ,EAAE,GAAGxC;QACrB,MAAMyC,mBAAmBzC,QAAQ,UAAU,CAAC,MAAM;QAClDH,OAAO4C,kBAAkB;QACzB,MAAMC,eAAeC;QAGrB,MAAMC,kBAAkB;QACxB,MAAMC,aAAmBC,MAAM,OAAO,CAACP,UACnC;YACE,MAAMQ,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC/C,KAAKG,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC9C,OAAOA;YACP,QAAQA;QACV,IACAL;QAEJ,IAAIS,eAAe,MAAMC,wBAAwB;YAC/C,gBAAgBR;YAChB,MAAMD;YACN,sBAAsB;gBACpB;oBACE,MAAMK;gBACR;aACD;YACD,iBAAiB;QACnB;QAEA,IAAIrD,KAAK,YAAY;YACnB,MAAMS,aAAaiD,iBAAiBL,YAAYL;YAQhDpD,MAAM,oCAAoCa;YAC1C,MAAMkD,gBAAgB,MAAMC,WAAWJ,cAAc/C;YACrD+C,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,IAAIM;QAGJ,IAAI;YACFA,MAAM,MAAMC,yBACVF,MACA5D;QAEJ,EAAE,OAAO0C,OAAO;YACd,MAAMqB,oBAAoBC,sCAAsCtB;YAChE,IAAI,CAACqB,mBACH,MAAMrB;YAER/C,MAAM;YACN,OAAOoE;QACT;QAEA,MAAM,EAAEE,OAAO,EAAE,GAAGJ;QACpBzD,OAAO,CAAC6D,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1D7D,OAAO6D,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IAhaA,YACE1D,OAA2D,EAC3DR,GAAoB,CACpB;QAPF;QAEA;QAMEK,OAAOG,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAM2D,QAAQ,OAAO,CAAC3D;QAGlD,IAAI,AAAyB,WAAlBR,KAAK,UACd,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAmZF"}

package/dist/es/skill/index.mjs

@@ -1,5 +1,5 @@
+import { BaseMidsceneTools } from "@midscene/shared/agent-tools/base-tools";
 import { reportCLIError, runToolsCLI } from "@midscene/shared/cli";
-import { BaseMidsceneTools } from "@midscene/shared/mcp/base-tools";
 import { Agent } from "../agent/agent.mjs";
 function _define_property(obj, key, value) {
     if (key in obj) Object.defineProperty(obj, key, {

package/dist/es/skill/index.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"skill/index.mjs","sources":["../../../src/skill/index.ts"],"sourcesContent":["import { reportCLIError, runToolsCLI } from '@midscene/shared/cli';\nimport { BaseMidsceneTools } from '@midscene/shared/mcp/base-tools';\nimport type { BaseAgent, BaseDevice } from '@midscene/shared/mcp/types';\nimport { Agent } from '../agent/agent';\nimport type { AbstractInterface } from '../device';\n\ntype DeviceClass = new (...args: any[]) => AbstractInterface;\n\n/**\n * Skill tools manager that lazily creates Agent from a Device class.\n * Used by runSkillCLI for CLI / Agent Skills scenarios where no agent exists at startup.\n */\nclass SkillMidsceneTools extends BaseMidsceneTools<BaseAgent> {\n  constructor(private DeviceClass: DeviceClass) {\n    super();\n  }\n\n  protected createTemporaryDevice(): BaseDevice {\n    return new this.DeviceClass() as unknown as BaseDevice;\n  }\n\n  protected async ensureAgent(): Promise<BaseAgent> {\n    if (!this.agent) {\n      const device = new this.DeviceClass();\n      this.agent = new Agent(device) as unknown as BaseAgent;\n    }\n    return this.agent;\n  }\n}\n\nexport interface SkillCLIOptions {\n  scriptName: string;\n  DeviceClass: DeviceClass;\n}\n\n/**\n * Launch a Skill CLI for a custom interface Device class.\n * This enables AI coding assistants (Claude Code, Cline, etc.) to control\n * your custom interface through CLI commands.\n *\n * @example\n * ```typescript\n * #!/usr/bin/env node\n * import { runSkillCLI } from '@midscene/core/skill';\n * import { SampleDevice } from './sample-device';\n *\n * runSkillCLI({\n *   DeviceClass: SampleDevice,\n *   scriptName: 'my-device',\n * });\n * ```\n */\nexport function runSkillCLI(options: SkillCLIOptions): Promise<void> {\n  const tools = new SkillMidsceneTools(options.DeviceClass);\n  return runToolsCLI(tools, options.scriptName).catch((e) => {\n    process.exit(reportCLIError(e));\n  });\n}\n"],"names":["SkillMidsceneTools","BaseMidsceneTools","device","Agent","DeviceClass","runSkillCLI","options","tools","runToolsCLI","e","process","reportCLIError"],"mappings":";;;;;;;;;;;;;AAYA,MAAMA,2BAA2BC;IAKrB,wBAAoC;QAC5C,OAAO,IAAI,IAAI,CAAC,WAAW;IAC7B;IAEA,MAAgB,cAAkC;QAChD,IAAI,CAAC,IAAI,CAAC,KAAK,EAAE;YACf,MAAMC,SAAS,IAAI,IAAI,CAAC,WAAW;YACnC,IAAI,CAAC,KAAK,GAAG,IAAIC,MAAMD;QACzB;QACA,OAAO,IAAI,CAAC,KAAK;IACnB;IAdA,YAAoBE,WAAwB,CAAE;QAC5C,KAAK,wDADaA,WAAW,GAAXA;IAEpB;AAaF;AAwBO,SAASC,YAAYC,OAAwB;IAClD,MAAMC,QAAQ,IAAIP,mBAAmBM,QAAQ,WAAW;IACxD,OAAOE,YAAYD,OAAOD,QAAQ,UAAU,EAAE,KAAK,CAAC,CAACG;QACnDC,QAAQ,IAAI,CAACC,eAAeF;IAC9B;AACF"}
+{"version":3,"file":"skill/index.mjs","sources":["../../../src/skill/index.ts"],"sourcesContent":["import { BaseMidsceneTools } from '@midscene/shared/agent-tools/base-tools';\nimport type { BaseAgent, BaseDevice } from '@midscene/shared/agent-tools/types';\nimport { reportCLIError, runToolsCLI } from '@midscene/shared/cli';\nimport { Agent } from '../agent/agent';\nimport type { AbstractInterface } from '../device';\n\ntype DeviceClass = new (...args: any[]) => AbstractInterface;\n\n/**\n * Skill tools manager that lazily creates Agent from a Device class.\n * Used by runSkillCLI for CLI / Agent Skills scenarios where no agent exists at startup.\n */\nclass SkillMidsceneTools extends BaseMidsceneTools<BaseAgent> {\n  constructor(private DeviceClass: DeviceClass) {\n    super();\n  }\n\n  protected createTemporaryDevice(): BaseDevice {\n    return new this.DeviceClass() as unknown as BaseDevice;\n  }\n\n  protected async ensureAgent(): Promise<BaseAgent> {\n    if (!this.agent) {\n      const device = new this.DeviceClass();\n      this.agent = new Agent(device) as unknown as BaseAgent;\n    }\n    return this.agent;\n  }\n}\n\nexport interface SkillCLIOptions {\n  scriptName: string;\n  DeviceClass: DeviceClass;\n}\n\n/**\n * Launch a Skill CLI for a custom interface Device class.\n * This enables AI coding assistants (Claude Code, Cline, etc.) to control\n * your custom interface through CLI commands.\n *\n * @example\n * ```typescript\n * #!/usr/bin/env node\n * import { runSkillCLI } from '@midscene/core/skill';\n * import { SampleDevice } from './sample-device';\n *\n * runSkillCLI({\n *   DeviceClass: SampleDevice,\n *   scriptName: 'my-device',\n * });\n * ```\n */\nexport function runSkillCLI(options: SkillCLIOptions): Promise<void> {\n  const tools = new SkillMidsceneTools(options.DeviceClass);\n  return runToolsCLI(tools, options.scriptName).catch((e) => {\n    process.exit(reportCLIError(e));\n  });\n}\n"],"names":["SkillMidsceneTools","BaseMidsceneTools","device","Agent","DeviceClass","runSkillCLI","options","tools","runToolsCLI","e","process","reportCLIError"],"mappings":";;;;;;;;;;;;;AAYA,MAAMA,2BAA2BC;IAKrB,wBAAoC;QAC5C,OAAO,IAAI,IAAI,CAAC,WAAW;IAC7B;IAEA,MAAgB,cAAkC;QAChD,IAAI,CAAC,IAAI,CAAC,KAAK,EAAE;YACf,MAAMC,SAAS,IAAI,IAAI,CAAC,WAAW;YACnC,IAAI,CAAC,KAAK,GAAG,IAAIC,MAAMD;QACzB;QACA,OAAO,IAAI,CAAC,KAAK;IACnB;IAdA,YAAoBE,WAAwB,CAAE;QAC5C,KAAK,wDADaA,WAAW,GAAXA;IAEpB;AAaF;AAwBO,SAASC,YAAYC,OAAwB;IAClD,MAAMC,QAAQ,IAAIP,mBAAmBM,QAAQ,WAAW;IACxD,OAAOE,YAAYD,OAAOD,QAAQ,UAAU,EAAE,KAAK,CAAC,CAACG;QACnDC,QAAQ,IAAI,CAACC,eAAeF;IAC9B;AACF"}