@midscene/core 0.17.3 → 0.17.4-beta-20250528162713.0

package/dist/lib/chunk-CDRBBE7D.js.map

@@ -0,0 +1 @@
+{"version":3,"mappings":";AAAA,SAAS,gBAAgB;AACzB,SAAS,YAAY,WAAW,cAAc,qBAAqB;AACnE,SAAS,cAAc;AACvB,YAAY,UAAU;AACtB,SAAS,eAAe;AACxB;AAAA,EACE;AAAA,EACA;AAAA,OACK;AACP;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,yBAAyB;AAClC,SAAS,cAAc;AACvB,SAAS,YAAY,aAAa,YAAY;AAC9C,OAAO,SAAS;AAGhB,IAAI,cAAc;AAElB,IAAM,aAAa;AAAA,EACjB;AACF;AACO,IAAM,2BAA2B;AAExC,IAAM,YAAY;AAElB,SAAS,eAAe;AACtB,SAAO;AACT;AAEO,SAAS,iCACd,KACA,QACA,aACA;AACA,QAAM,QAAQ,IAAI,QAAQ,MAAM;AAChC,SAAO,IAAI,MAAM,GAAG,KAAK,IAAI,cAAc,IAAI,MAAM,QAAQ,OAAO,MAAM;AAC5E;AAEO,SAAS,kBACd,UACA,YACQ;AACR,QAAM,MAAM,aAAa;AACzB,MAAI,CAAC,KAAK;AACR,YAAQ,KAAK,6CAA6C;AAC1D,WAAO;AAAA,EACT;AAEA,QAAM,kBAAkB;AAGxB,MAAI,CAAC,IAAI,SAAS,eAAe,GAAG;AAClC,YAAQ;AAAA,MACN;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAGA,QAAM,mBAAmB,IAAI,QAAQ,eAAe;AAGpD,QAAM,YAAY,IAAI,UAAU,GAAG,gBAAgB;AACnD,QAAM,aAAa,IAAI,UAAU,mBAAmB,gBAAgB,MAAM;AAG1E,QAAM,cAAc,cAAc,CAAC;AACnC,MAAI,gBAAgB;AAGpB,QAAM,gBAAgB,CAAC,YAA0B;AAC/C,QAAI,aAAa;AACf,oBAAc,YAAa,GAAG,OAAO;AAAA,GAAM;AAAA,QACzC,MAAM;AAAA,MACR,CAAC;AAAA,IACH,OAAO;AACL,uBAAiB,GAAG,OAAO;AAAA;AAAA,IAC7B;AAAA,EACF;AAGA,MAAI,aAAa;AACf,kBAAc,YAAa,WAAW,EAAE,MAAM,IAAI,CAAC;AAAA,EACrD,OAAO;AACL,oBAAgB;AAAA,EAClB;AAIA,MACG,MAAM,QAAQ,QAAQ,KAAK,SAAS,WAAW,KAChD,OAAO,aAAa,aACpB;AACA,UAAM,cACJ;AACF,kBAAc,WAAW;AAAA,EAC3B,WAES,OAAO,aAAa,UAAU;AACrC,UAAM;AAAA;AAAA,MAEJ,gEACA,IAAI,UAAU,UAAU,IACxB;AAAA;AACF,kBAAc,WAAW;AAAA,EAC3B,OAEK;AAEH,aAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,YAAM,EAAE,YAAY,WAAW,IAAI,SAAS,CAAC;AAC7C,YAAM,gBAAgB,OAAO,KAAK,cAAc,CAAC,CAAC,EAAE,IAAI,CAAC,QAAQ;AAC/D,eAAO,GAAG,GAAG,KAAK,mBAAmB,WAAY,GAAG,CAAC,CAAC;AAAA,MACxD,CAAC;AAED,YAAM;AAAA;AAAA,QAEJ,8DACA,cAAc,KAAK,GAAG,IACtB,QACA,IAAI,YAAY,UAAU,IAC1B;AAAA;AACF,oBAAc,WAAW;AAAA,IAC3B;AAAA,EACF;AAGA,MAAI,aAAa;AACf,kBAAc,YAAa,YAAY,EAAE,MAAM,IAAI,CAAC;AACpD,WAAO;AAAA,EACT;AAEA,mBAAiB;AACjB,SAAO;AACT;AAEO,SAAS,gBACd,UACA,UACe;AACf,MAAI,aAAa;AACf,YAAQ,IAAI,kCAAkC;AAC9C,WAAO;AAAA,EACT;AAEA,QAAM,YAAY,QAAQ,UAAU;AACpC,QAAM,kBAAkB,kBAAkB,SAAS;AACnD,MAAI,CAAC,iBAAiB;AACpB,YAAQ,KAAK,kDAAkD;AAC/D,WAAO;AAAA,EACT;AAEA,QAAM,aAAkB;AAAA,IACtB,qBAAqB,QAAQ;AAAA,IAC7B,GAAG,QAAQ;AAAA,EACb;AAEA,oBAAkB,UAAU,UAAU;AAEtC,MAAI,QAAQ,IAAI,yBAAyB;AACvC;AAAA,MACE,GAAG,UAAU;AAAA,MACb,OAAO,aAAa,WAChB,WACA,KAAK,UAAU,UAAU,MAAM,CAAC;AAAA,IACtC;AAAA,EACF;AAEA,SAAO;AACT;AAEO,SAAS,aAAa,MAM1B;AACD,MAAI,aAAa;AACf,WAAO;AAAA,EACT;AACA,QAAM,EAAE,UAAU,SAAS,aAAa,OAAO,OAAO,IAAI;AAC1D,QAAM,YAAY,qBAAqB,IAAI;AAE3C,MAAI,CAAC,aAAa;AAChB,WAAO,WAAW,+CAA+C;AAGjE,UAAM,gBAAqB,UAAK,WAAW,kBAAkB;AAC7D,UAAM,UAAe,UAAK,WAAW,YAAY;AACjD,QAAI,mBAAmB;AAEvB,QAAI,WAAW,OAAO,GAAG;AAEvB,UAAI,WAAW,aAAa,GAAG;AAC7B,2BAAmB,aAAa,eAAe,OAAO;AAAA,MACxD;AAGA,UAAI,CAAC,iBAAiB,SAAS,GAAG,iBAAiB,GAAG,GAAG;AACvD;AAAA,UACE;AAAA,UACA,GAAG,gBAAgB;AAAA;AAAA,EAA+B,iBAAiB;AAAA,EAAU,iBAAiB;AAAA,EAAY,iBAAiB;AAAA,EAAS,iBAAiB;AAAA;AAAA,UACrJ;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,kBAAc;AAAA,EAChB;AAEA,QAAM,WAAgB,UAAK,WAAW,GAAG,QAAQ,IAAI,OAAO,EAAE;AAE9D,MAAI,SAAS,QAAQ;AAEnB,kBAAc,UAAU,WAAW;AAAA,EACrC;AAEA,MAAI,MAAM,gBAAgB;AACxB,WAAO,gBAAgB,UAAU,WAAW;AAAA,EAC9C;AAEA,SAAO;AACT;AAEO,SAAS,YAA2B;AACzC,MAAI;AACF,UAAM,iBAAiB,kBAAkB;AACzC,QAAI,CAAC,gBAAgB;AACnB,aAAO;AAAA,IACT;AACA,UAAM,EAAE,KAAK,IAAI;AACjB,UAAM,UAAe,UAAK,OAAO,GAAG,IAAI;AACxC,cAAU,SAAS,EAAE,WAAW,KAAK,CAAC;AACtC,WAAO;AAAA,EACT,SAAS,GAAG;AACV,WAAO;AAAA,EACT;AACF;AAEO,SAAS,WAAW,mBAA0C;AACnE,MAAI,aAAa;AACf,WAAO;AAAA,EACT;AACA,QAAM,SAAS,UAAU;AACzB,QAAM,WAAW,GAAG,KAAK,CAAC,IAAI,iBAAiB;AAC/C,SAAY,UAAK,QAAS,QAAQ;AACpC;AAEO,SAAS,WAAW,WAAiB,QAAc;AAExD,SACE,UAAU,OAAO,OAAO,OAAO,OAAO,SACtC,UAAU,OAAO,UAAU,QAAQ,OAAO,QAC1C,UAAU,MAAM,OAAO,MAAM,OAAO,UACpC,UAAU,MAAM,UAAU,SAAS,OAAO;AAE9C;AAEA,eAAsB,MAAM,IAAY;AACtC,SAAO,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,EAAE,CAAC;AACzD;AAEO,SAAS,sBAAsB,KAAa,OAAY;AAC7D,MAAI,SAAS,MAAM,aAAa,SAAS,QAAQ;AAC/C,WAAO;AAAA,EACT;AACA,MAAI,SAAS,MAAM,aAAa,SAAS,WAAW;AAClD,WAAO;AAAA,EACT;AACA,SAAO;AACT;AAEO,SAAS,kBAAkB,MAAW,SAAkB;AAC7D,SAAO,KAAK,UAAU,MAAM,uBAAuB,OAAO;AAC5D;AAIO,SAAS,aAAa;AAC3B,SAAO;AACT;AAEA,SAAS,YAAY,SAAgB;AACnC,QAAM,YAAY,YAAY,mBAAmB;AACjD,MAAI,WAAW;AACb,YAAQ,IAAI,cAAc,GAAG,OAAO;AAAA,EACtC;AACF;AAEA,IAAI,sBAAsB;AACnB,SAAS,uBAAuB,EAAE,QAAQ,GAAwB;AACvE,MAAI,UAAU;AACd,MAAI,YAAY;AAEhB,QAAM,cAAc,kBAAkB,gCAAgC;AACtE,QAAM,YAAY,aAAa;AAE/B,MAAI;AACF,cAAU,SAAS,oCAAoC,EAAE,SAAS,EAAE,KAAK;AACzE,gBAAY,SAAS,6BAA6B,EAAE,SAAS,EAAE,KAAK;AAAA,EACtE,SAAS,OAAO;AACd,aAAS,2BAA2B,KAAK;AAAA,EAC3C;AAOA,MACE,cACE,WAAW,YAAY,uBAAyB,CAAC,WAAW,UAC9D;AACA,aAAS,iCAAiC;AAAA,MACxC;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AAED,UAAM,WAAW;AAAA,MACf,QAAQ;AAAA,MACR,SAAS;AAAA,QACP,gBAAgB;AAAA,MAClB;AAAA,MACA,MAAM,KAAK,UAAU;AAAA,QACnB,UAAU;AAAA,QACV,UAAU;AAAA,QACV,YAAY;AAAA,MACd,CAAC;AAAA,IACH,CAAC,EACE,KAAK,CAAC,aAAa,SAAS,KAAK,CAAC,EAClC,KAAK,CAAC,SAAS;AACd,eAAS,8CAA8C,IAAI;AAAA,IAC7D,CAAC,EACA;AAAA,MAAM,CAAC,UACN,SAAS,yCAAyC,KAAK;AAAA,IACzD;AACF,0BAAsB;AAAA,EACxB;AACF","names":[],"ignoreList":[],"sources":["../../src/utils.ts"],"sourcesContent":["import { execSync } from 'node:child_process';\nimport { existsSync, mkdirSync, readFileSync, writeFileSync } from 'node:fs';\nimport { tmpdir } from 'node:os';\nimport * as path from 'node:path';\nimport { dirname } from 'node:path';\nimport {\n  defaultRunDirName,\n  getMidsceneRunSubDir,\n} from '@midscene/shared/common';\nimport {\n  MIDSCENE_DEBUG_MODE,\n  MIDSCENE_OPENAI_INIT_CONFIG_JSON,\n  getAIConfig,\n  getAIConfigInJson,\n} from '@midscene/shared/env';\nimport { getRunningPkgInfo } from '@midscene/shared/fs';\nimport { assert } from '@midscene/shared/utils';\nimport { escapeHtml, ifInBrowser, uuid } from '@midscene/shared/utils';\nimport xss from 'xss';\nimport type { Rect, ReportDumpWithAttributes } from './types';\n\nlet logEnvReady = false;\n\nconst xssOptions = {\n  escapeHtml,\n};\nexport const groupedActionDumpFileExt = 'web-dump.json';\n\nconst reportTpl = 'REPLACE_ME_WITH_REPORT_HTML';\n\nfunction getReportTpl() {\n  return reportTpl;\n}\n\nexport function replaceStringWithFirstAppearance(\n  str: string,\n  target: string,\n  replacement: string,\n) {\n  const index = str.indexOf(target);\n  return str.slice(0, index) + replacement + str.slice(index + target.length);\n}\n\nexport function reportHTMLContent(\n  dumpData: string | ReportDumpWithAttributes[],\n  reportPath?: string,\n): string {\n  const tpl = getReportTpl();\n  if (!tpl) {\n    console.warn('reportTpl is not set, will not write report');\n    return '';\n  }\n\n  const dumpPlaceholder = '{{dump}}';\n\n  // verify the template contains the placeholder\n  if (!tpl.includes(dumpPlaceholder)) {\n    console.warn(\n      'Failed to get the Midscene report template due to the lack of the {{dump}} placeholder. If you are building Midscene.js by yourself, please refer to the contribution guide for more information: https://github.com/web-infra-dev/midscene/blob/main/CONTRIBUTING.md#FAQ',\n    );\n    return '';\n  }\n\n  // find the first placeholder position\n  const placeholderIndex = tpl.indexOf(dumpPlaceholder);\n\n  // split the template into two parts before and after the placeholder\n  const firstPart = tpl.substring(0, placeholderIndex);\n  const secondPart = tpl.substring(placeholderIndex + dumpPlaceholder.length);\n\n  // if reportPath is set, it means we are in write to file mode\n  const writeToFile = reportPath && !ifInBrowser;\n  let resultContent = '';\n\n  // helper function: decide to write to file or append to resultContent\n  const appendOrWrite = (content: string): void => {\n    if (writeToFile) {\n      writeFileSync(reportPath!, `${content}\\n`, {\n        flag: 'a',\n      });\n    } else {\n      resultContent += `${content}\\n`;\n    }\n  };\n\n  // if writeToFile is true, write the first part to file, otherwise set the first part to the initial value of resultContent\n  if (writeToFile) {\n    writeFileSync(reportPath!, firstPart, { flag: 'w' }); // use 'w' flag to overwrite the existing file\n  } else {\n    resultContent = firstPart;\n  }\n\n  // generate dump content\n  // handle empty data or undefined\n  if (\n    (Array.isArray(dumpData) && dumpData.length === 0) ||\n    typeof dumpData === 'undefined'\n  ) {\n    const dumpContent =\n      '<script type=\"midscene_web_dump\" type=\"application/json\"></script>';\n    appendOrWrite(dumpContent);\n  }\n  // handle string type dumpData\n  else if (typeof dumpData === 'string') {\n    const dumpContent =\n      // biome-ignore lint/style/useTemplate: <explanation> do not use template string here, will cause bundle error\n      '<script type=\"midscene_web_dump\" type=\"application/json\">\\n' +\n      xss(dumpData, xssOptions) +\n      '\\n</script>';\n    appendOrWrite(dumpContent);\n  }\n  // handle array type dumpData\n  else {\n    // for array, handle each item\n    for (let i = 0; i < dumpData.length; i++) {\n      const { dumpString, attributes } = dumpData[i];\n      const attributesArr = Object.keys(attributes || {}).map((key) => {\n        return `${key}=\"${encodeURIComponent(attributes![key])}\"`;\n      });\n\n      const dumpContent =\n        // biome-ignore lint/style/useTemplate: <explanation> do not use template string here, will cause bundle error\n        '<script type=\"midscene_web_dump\" type=\"application/json\" ' +\n        attributesArr.join(' ') +\n        '>\\n' +\n        xss(dumpString, xssOptions) +\n        '\\n</script>';\n      appendOrWrite(dumpContent);\n    }\n  }\n\n  // add the second part\n  if (writeToFile) {\n    writeFileSync(reportPath!, secondPart, { flag: 'a' });\n    return reportPath!;\n  }\n\n  resultContent += secondPart;\n  return resultContent;\n}\n\nexport function writeDumpReport(\n  fileName: string,\n  dumpData: string | ReportDumpWithAttributes[],\n): string | null {\n  if (ifInBrowser) {\n    console.log('will not write report in browser');\n    return null;\n  }\n\n  const __dirname = dirname(__filename);\n  const midscenePkgInfo = getRunningPkgInfo(__dirname);\n  if (!midscenePkgInfo) {\n    console.warn('midscenePkgInfo not found, will not write report');\n    return null;\n  }\n\n  const reportPath = path.join(\n    getMidsceneRunSubDir('report'),\n    `${fileName}.html`,\n  );\n\n  reportHTMLContent(dumpData, reportPath);\n\n  if (process.env.MIDSCENE_DEBUG_LOG_JSON) {\n    writeFileSync(\n      `${reportPath}.json`,\n      typeof dumpData === 'string'\n        ? dumpData\n        : JSON.stringify(dumpData, null, 2),\n    );\n  }\n\n  return reportPath;\n}\n\nexport function writeLogFile(opts: {\n  fileName: string;\n  fileExt: string;\n  fileContent: string;\n  type: 'dump' | 'cache' | 'report' | 'tmp';\n  generateReport?: boolean;\n}) {\n  if (ifInBrowser) {\n    return '/mock/report.html';\n  }\n  const { fileName, fileExt, fileContent, type = 'dump' } = opts;\n  const targetDir = getMidsceneRunSubDir(type);\n  // Ensure directory exists\n  if (!logEnvReady) {\n    assert(targetDir, 'logDir should be set before writing dump file');\n\n    // gitIgnore in the parent directory\n    const gitIgnorePath = path.join(targetDir, '../../.gitignore');\n    const gitPath = path.join(targetDir, '../../.git');\n    let gitIgnoreContent = '';\n\n    if (existsSync(gitPath)) {\n      // if the git path exists, we need to add the log folder to the git ignore file\n      if (existsSync(gitIgnorePath)) {\n        gitIgnoreContent = readFileSync(gitIgnorePath, 'utf-8');\n      }\n\n      // ignore the log folder\n      if (!gitIgnoreContent.includes(`${defaultRunDirName}/`)) {\n        writeFileSync(\n          gitIgnorePath,\n          `${gitIgnoreContent}\\n# Midscene.js dump files\\n${defaultRunDirName}/dump\\n${defaultRunDirName}/report\\n${defaultRunDirName}/tmp\\n${defaultRunDirName}/log\\n`,\n          'utf-8',\n        );\n      }\n    }\n\n    logEnvReady = true;\n  }\n\n  const filePath = path.join(targetDir, `${fileName}.${fileExt}`);\n\n  if (type !== 'dump') {\n    // do not write dump file any more\n    writeFileSync(filePath, fileContent);\n  }\n\n  if (opts?.generateReport) {\n    return writeDumpReport(fileName, fileContent);\n  }\n\n  return filePath;\n}\n\nexport function getTmpDir(): string | null {\n  try {\n    const runningPkgInfo = getRunningPkgInfo();\n    if (!runningPkgInfo) {\n      return null;\n    }\n    const { name } = runningPkgInfo;\n    const tmpPath = path.join(tmpdir(), name);\n    mkdirSync(tmpPath, { recursive: true });\n    return tmpPath;\n  } catch (e) {\n    return null;\n  }\n}\n\nexport function getTmpFile(fileExtWithoutDot: string): string | null {\n  if (ifInBrowser) {\n    return null;\n  }\n  const tmpDir = getTmpDir();\n  const filename = `${uuid()}.${fileExtWithoutDot}`;\n  return path.join(tmpDir!, filename);\n}\n\nexport function overlapped(container: Rect, target: Rect) {\n  // container and the target have some part overlapped\n  return (\n    container.left < target.left + target.width &&\n    container.left + container.width > target.left &&\n    container.top < target.top + target.height &&\n    container.top + container.height > target.top\n  );\n}\n\nexport async function sleep(ms: number) {\n  return new Promise((resolve) => setTimeout(resolve, ms));\n}\n\nexport function replacerForPageObject(key: string, value: any) {\n  if (value && value.constructor?.name === 'Page') {\n    return '[Page object]';\n  }\n  if (value && value.constructor?.name === 'Browser') {\n    return '[Browser object]';\n  }\n  return value;\n}\n\nexport function stringifyDumpData(data: any, indents?: number) {\n  return JSON.stringify(data, replacerForPageObject, indents);\n}\n\ndeclare const __VERSION__: string;\n\nexport function getVersion() {\n  return __VERSION__;\n}\n\nfunction debugLog(...message: any[]) {\n  const debugMode = getAIConfig(MIDSCENE_DEBUG_MODE);\n  if (debugMode) {\n    console.log('[Midscene]', ...message);\n  }\n}\n\nlet lastReportedRepoUrl = '';\nexport function uploadTestInfoToServer({ testUrl }: { testUrl: string }) {\n  let repoUrl = '';\n  let userEmail = '';\n\n  const extraConfig = getAIConfigInJson(MIDSCENE_OPENAI_INIT_CONFIG_JSON);\n  const serverUrl = extraConfig?.REPORT_SERVER_URL;\n\n  try {\n    repoUrl = execSync('git config --get remote.origin.url').toString().trim();\n    userEmail = execSync('git config --get user.email').toString().trim();\n  } catch (error) {\n    debugLog('Failed to get git info:', error);\n  }\n\n  // Only upload test info if:\n  // 1. Server URL is configured AND\n  // 2. Either:\n  //    - We have a repo URL that's different from last reported one (to avoid duplicate reports)\n  //    - OR we don't have a repo URL but have a test URL (for non-git environments)\n  if (\n    serverUrl &&\n    ((repoUrl && repoUrl !== lastReportedRepoUrl) || (!repoUrl && testUrl))\n  ) {\n    debugLog('Uploading test info to server', {\n      serverUrl,\n      repoUrl,\n      testUrl,\n      userEmail,\n    });\n\n    fetch(serverUrl, {\n      method: 'POST',\n      headers: {\n        'Content-Type': 'application/json',\n      },\n      body: JSON.stringify({\n        repo_url: repoUrl,\n        test_url: testUrl,\n        user_email: userEmail,\n      }),\n    })\n      .then((response) => response.json())\n      .then((data) => {\n        debugLog('Successfully uploaded test info to server:', data);\n      })\n      .catch((error) =>\n        debugLog('Failed to upload test info to server:', error),\n      );\n    lastReportedRepoUrl = repoUrl;\n  }\n}\n"]}

package/dist/lib/index.d.ts

@@ -1,7 +1,7 @@
-import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, B as BaseElement, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightAssertionResponse, R as Rect, A as AIDescribeElementResponse } from './types-81bb2c02.js';
-export { u as AIAssertionResponse, s as AIDataExtractionResponse, q as AIElementCoordinatesResponse, p as AIElementLocatorResponse, r as AIElementResponse, l as AIResponseFormat, t as AISectionLocatorResponse, o as AISingleElementResponse, m as AISingleElementResponseById, n as AISingleElementResponseByPosition, k as AIUsageInfo, V as AgentAssertOpt, x as AgentDescribeElementAtPointResult, T as AgentWaitForOpt, a6 as BaseAgentParserOpt, C as CallAIFn, a5 as Color, G as DumpMeta, O as ElementById, j as ElementTreeNode, y as EnsureObject, a9 as ExecutionRecorderItem, ar as ExecutionTaskAction, aq as ExecutionTaskActionApply, ap as ExecutionTaskInsightAssertion, ao as ExecutionTaskInsightAssertionApply, an as ExecutionTaskInsightAssertionParam, ag as ExecutionTaskInsightDumpLog, ai as ExecutionTaskInsightLocate, ah as ExecutionTaskInsightLocateApply, af as ExecutionTaskInsightLocateOutput, ae as ExecutionTaskInsightLocateParam, am as ExecutionTaskInsightQuery, al as ExecutionTaskInsightQueryApply, ak as ExecutionTaskInsightQueryOutput, aj as ExecutionTaskInsightQueryParam, at as ExecutionTaskPlanning, as as ExecutionTaskPlanningApply, ad as ExecutionTaskReturn, aa as ExecutionTaskType, ab as ExecutorContext, aS as FreeFn, au as GroupedActionDump, J as InsightDump, z as InsightExtractParam, N as LiteUISection, aw as LocateOption, F as LocateResultElement, w as LocateValidatorResult, v as LocatorValidatorOption, i as MidsceneYamlFlowItem, aD as MidsceneYamlFlowItemAIAction, aE as MidsceneYamlFlowItemAIAssert, aI as MidsceneYamlFlowItemAIBoolean, aM as MidsceneYamlFlowItemAIHover, aN as MidsceneYamlFlowItemAIInput, aO as MidsceneYamlFlowItemAIKeyboardPress, aJ as MidsceneYamlFlowItemAILocate, aH as MidsceneYamlFlowItemAINString, aG as MidsceneYamlFlowItemAINumber, aF as MidsceneYamlFlowItemAIQuery, aP as MidsceneYamlFlowItemAIScroll, aL as MidsceneYamlFlowItemAITap, aK as MidsceneYamlFlowItemAIWaitFor, aQ as MidsceneYamlFlowItemEvaluateJavaScript, aR as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aB as MidsceneYamlScriptAndroidEnv, aC as MidsceneYamlScriptEnv, az as MidsceneYamlScriptEnvBase, aA as MidsceneYamlScriptWebEnv, h as MidsceneYamlTask, Q as OnTaskStartTip, av as PageType, K as PartialInsightDumpFromSDK, Y as PlanningAIResponse, X as PlanningAction, a1 as PlanningActionParamAssert, a3 as PlanningActionParamError, _ as PlanningActionParamHover, $ as PlanningActionParamInputOrKeyPress, a0 as PlanningActionParamScroll, a2 as PlanningActionParamSleep, Z as PlanningActionParamTap, a4 as PlanningActionParamWaitFor, W as PlanningLocateParam, a8 as PlaywrightParserOpt, P as Point, a7 as PuppeteerParserOpt, ax as ReferenceImage, H as ReportDumpWithAttributes, aU as ScriptPlayerStatusValue, aT as ScriptPlayerTaskStatus, S as Size, ac as TaskCacheInfo, ay as scrollParam } from './types-81bb2c02.js';
-import { c as callAiFn } from './llm-planning-9c97f5c2.js';
-export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-9c97f5c2.js';
+import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, B as BaseElement, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightAssertionResponse, R as Rect, A as AIDescribeElementResponse } from './types-c892e193.js';
+export { v as AIAssertionResponse, t as AIDataExtractionResponse, r as AIElementCoordinatesResponse, q as AIElementLocatorResponse, s as AIElementResponse, m as AIResponseFormat, u as AISectionLocatorResponse, p as AISingleElementResponse, n as AISingleElementResponseById, o as AISingleElementResponseByPosition, l as AIUsageInfo, W as AgentAssertOpt, y as AgentDescribeElementAtPointResult, V as AgentWaitForOpt, a8 as BaseAgentParserOpt, C as CallAIFn, a7 as Color, H as DumpMeta, Q as ElementById, k as ElementTreeNode, z as EnsureObject, ab as ExecutionRecorderItem, at as ExecutionTaskAction, as as ExecutionTaskActionApply, ar as ExecutionTaskInsightAssertion, aq as ExecutionTaskInsightAssertionApply, ap as ExecutionTaskInsightAssertionParam, ai as ExecutionTaskInsightDumpLog, ak as ExecutionTaskInsightLocate, aj as ExecutionTaskInsightLocateApply, ah as ExecutionTaskInsightLocateOutput, ag as ExecutionTaskInsightLocateParam, ao as ExecutionTaskInsightQuery, an as ExecutionTaskInsightQueryApply, am as ExecutionTaskInsightQueryOutput, al as ExecutionTaskInsightQueryParam, av as ExecutionTaskPlanning, au as ExecutionTaskPlanningApply, af as ExecutionTaskReturn, ac as ExecutionTaskType, ad as ExecutorContext, aU as FreeFn, aw as GroupedActionDump, K as InsightDump, F as InsightExtractParam, O as LiteUISection, ay as LocateOption, G as LocateResultElement, x as LocateValidatorResult, w as LocatorValidatorOption, i as MidsceneYamlFlowItem, aF as MidsceneYamlFlowItemAIAction, aG as MidsceneYamlFlowItemAIAssert, aK as MidsceneYamlFlowItemAIBoolean, aO as MidsceneYamlFlowItemAIHover, aP as MidsceneYamlFlowItemAIInput, aQ as MidsceneYamlFlowItemAIKeyboardPress, aL as MidsceneYamlFlowItemAILocate, aJ as MidsceneYamlFlowItemAINString, aI as MidsceneYamlFlowItemAINumber, aH as MidsceneYamlFlowItemAIQuery, j as MidsceneYamlFlowItemAIRightClick, aR as MidsceneYamlFlowItemAIScroll, aN as MidsceneYamlFlowItemAITap, aM as MidsceneYamlFlowItemAIWaitFor, aS as MidsceneYamlFlowItemEvaluateJavaScript, aT as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aD as MidsceneYamlScriptAndroidEnv, aE as MidsceneYamlScriptEnv, aB as MidsceneYamlScriptEnvBase, aC as MidsceneYamlScriptWebEnv, h as MidsceneYamlTask, T as OnTaskStartTip, ax as PageType, N as PartialInsightDumpFromSDK, Z as PlanningAIResponse, Y as PlanningAction, a3 as PlanningActionParamAssert, a5 as PlanningActionParamError, $ as PlanningActionParamHover, a1 as PlanningActionParamInputOrKeyPress, a0 as PlanningActionParamRightClick, a2 as PlanningActionParamScroll, a4 as PlanningActionParamSleep, _ as PlanningActionParamTap, a6 as PlanningActionParamWaitFor, X as PlanningLocateParam, aa as PlaywrightParserOpt, P as Point, a9 as PuppeteerParserOpt, az as ReferenceImage, J as ReportDumpWithAttributes, aW as ScriptPlayerStatusValue, aV as ScriptPlayerTaskStatus, S as Size, ae as TaskCacheInfo, aA as scrollParam } from './types-c892e193.js';
+import { c as callAiFn } from './llm-planning-9cfa38ad.js';
+export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-9cfa38ad.js';
 export { getVersion } from './utils.js';
 export { MIDSCENE_MODEL_NAME, getAIConfig } from '@midscene/shared/env';
 import '@midscene/shared/constants';

package/dist/lib/index.js

@@ -1,6 +1,6 @@
 "use strict";Object.defineProperty(exports, "__esModule", {value: true}); function _optionalChain(ops) { let lastAccessLHS = undefined; let value = ops[0]; let i = 1; while (i < ops.length) { const op = ops[i]; const fn = ops[i + 1]; i += 2; if ((op === 'optionalAccess' || op === 'optionalCall') && value == null) { return undefined; } if (op === 'access' || op === 'optionalAccess') { lastAccessLHS = value; value = fn(value); } else if (op === 'call' || op === 'optionalCall') { value = fn((...args) => value.call(lastAccessLHS, ...args)); lastAccessLHS = undefined; } } return value; }
 
-var _chunkJ2RYPDNYjs = require('./chunk-J2RYPDNY.js');
+var _chunkCDRBBE7Djs = require('./chunk-CDRBBE7D.js');
 
 
 
@@ -11,9 +11,14 @@ var _chunkJ2RYPDNYjs = require('./chunk-J2RYPDNY.js');
 
 
 
-var _chunkAV6BGC5Xjs = require('./chunk-AV6BGC5X.js');
+var _chunk5HHR4GGSjs = require('./chunk-5HHR4GGS.js');
 
 // src/ai-model/action-executor.ts
+
+
+
+
+
 var _env = require('@midscene/shared/env');
 var _utils = require('@midscene/shared/utils');
 var Executor = class {
@@ -157,9 +162,19 @@ ${_optionalChain([this, 'access', _7 => _7.latestErrorTask, 'call', _8 => _8(),
     return null;
   }
   dump() {
+    let modelDescription = "";
+    if (_env.vlLocateMode.call(void 0, )) {
+      const uiTarsModelVer = _env.uiTarsModelVersion.call(void 0, );
+      if (uiTarsModelVer) {
+        modelDescription = `UI-TARS=${uiTarsModelVer}`;
+      } else {
+        modelDescription = `${_env.vlLocateMode.call(void 0, )} mode`;
+      }
+    }
     const dumpData = {
-      sdkVersion: _chunkJ2RYPDNYjs.getVersion.call(void 0, ),
+      sdkVersion: _chunkCDRBBE7Djs.getVersion.call(void 0, ),
       model_name: _env.getAIConfig.call(void 0, _env.MIDSCENE_MODEL_NAME) || "",
+      model_description: modelDescription,
       logTime: Date.now(),
       name: this.name,
       tasks: this.tasks
@@ -210,23 +225,11 @@ var _logger = require('@midscene/shared/logger');
 
 
 
-
-
 function emitInsightDump(data, dumpSubscriber) {
-  let modelDescription = "";
-  if (_env.vlLocateMode.call(void 0, )) {
-    const uiTarsModelVer = _env.uiTarsModelVersion.call(void 0, );
-    if (uiTarsModelVer) {
-      modelDescription = `UI-TARS=${uiTarsModelVer}`;
-    } else {
-      modelDescription = `${_env.vlLocateMode.call(void 0, )} mode`;
-    }
-  }
   const baseData = {
-    sdkVersion: _chunkJ2RYPDNYjs.getVersion.call(void 0, ),
+    sdkVersion: _chunkCDRBBE7Djs.getVersion.call(void 0, ),
     logTime: Date.now(),
-    model_name: _env.getAIConfig.call(void 0, _env.MIDSCENE_MODEL_NAME) || "",
-    model_description: modelDescription
+    model_name: _env.getAIConfig.call(void 0, _env.MIDSCENE_MODEL_NAME) || ""
   };
   const finalData = {
     logId: _utils.uuid.call(void 0, ),
@@ -240,7 +243,7 @@ function emitInsightDump(data, dumpSubscriber) {
 var debug = _logger.getDebug.call(void 0, "ai:insight");
 var Insight = class {
   constructor(context, opt) {
-    this.aiVendorFn = _chunkAV6BGC5Xjs.callAiFn;
+    this.aiVendorFn = _chunk5HHR4GGSjs.callAiFn;
     _utils.assert.call(void 0, context, "context is required for Insight");
     if (typeof context === "function") {
       this.contextRetrieverFn = context;
@@ -283,7 +286,7 @@ var Insight = class {
     let searchAreaUsage = void 0;
     let searchAreaResponse = void 0;
     if (searchAreaPrompt) {
-      searchAreaResponse = await _chunkAV6BGC5Xjs.AiLocateSection.call(void 0, {
+      searchAreaResponse = await _chunk5HHR4GGSjs.AiLocateSection.call(void 0, {
         context,
         sectionDescription: searchAreaPrompt
       });
@@ -296,7 +299,7 @@ var Insight = class {
       searchArea = searchAreaResponse.rect;
     }
     const startTime = Date.now();
-    const { parseResult, rect, elementById, rawResponse, usage } = await _chunkAV6BGC5Xjs.AiLocateElement.call(void 0, {
+    const { parseResult, rect, elementById, rawResponse, usage } = await _chunk5HHR4GGSjs.AiLocateElement.call(void 0, {
       callAI: callAI || this.aiVendorFn,
       context,
       targetElementDescription: queryPrompt,
@@ -384,7 +387,7 @@ ${parseResult.errors.join("\n")}`;
     this.onceDumpUpdatedFn = void 0;
     const context = await this.contextRetrieverFn("extract");
     const startTime = Date.now();
-    const { parseResult, usage } = await _chunkAV6BGC5Xjs.AiExtractElementInfo.call(void 0, {
+    const { parseResult, usage } = await _chunk5HHR4GGSjs.AiExtractElementInfo.call(void 0, {
       context,
       dataQuery: dataDemand
     });
@@ -435,7 +438,7 @@ ${parseResult.errors.join("\n")}`;
     this.onceDumpUpdatedFn = void 0;
     const context = await this.contextRetrieverFn("assert");
     const startTime = Date.now();
-    const assertResult = await _chunkAV6BGC5Xjs.AiAssert.call(void 0, {
+    const assertResult = await _chunk5HHR4GGSjs.AiAssert.call(void 0, {
       assertion,
       context
     });
@@ -488,7 +491,7 @@ ${parseResult.errors.join("\n")}`;
       borderThickness: 3
     });
     if (_optionalChain([opt, 'optionalAccess', _23 => _23.deepThink])) {
-      const searchArea = _chunkAV6BGC5Xjs.expandSearchArea.call(void 0, targetRect, context.size);
+      const searchArea = _chunk5HHR4GGSjs.expandSearchArea.call(void 0, targetRect, context.size);
       debug("describe: set searchArea", searchArea);
       imagePayload = await _img.cropByRect.call(void 0, 
         imagePayload,
@@ -511,7 +514,7 @@ ${parseResult.errors.join("\n")}`;
         ]
       }
     ];
-    const callAIFn = this.aiVendorFn || _chunkAV6BGC5Xjs.callToGetJSONObject;
+    const callAIFn = this.aiVendorFn || _chunk5HHR4GGSjs.callToGetJSONObject;
     const res = await callAIFn(msgs, 4 /* DESCRIBE_ELEMENT */);
     const { content } = res;
     _utils.assert.call(void 0, !content.error, `describe failed: ${content.error}`);
@@ -534,6 +537,6 @@ var src_default = Insight;
 
 
 
-exports.AiAssert = _chunkAV6BGC5Xjs.AiAssert; exports.AiLocateElement = _chunkAV6BGC5Xjs.AiLocateElement; exports.Executor = Executor; exports.Insight = Insight; exports.MIDSCENE_MODEL_NAME = _env.MIDSCENE_MODEL_NAME; exports.default = src_default; exports.describeUserPage = _chunkAV6BGC5Xjs.describeUserPage; exports.getAIConfig = _env.getAIConfig; exports.getVersion = _chunkJ2RYPDNYjs.getVersion; exports.plan = _chunkAV6BGC5Xjs.plan;
+exports.AiAssert = _chunk5HHR4GGSjs.AiAssert; exports.AiLocateElement = _chunk5HHR4GGSjs.AiLocateElement; exports.Executor = Executor; exports.Insight = Insight; exports.MIDSCENE_MODEL_NAME = _env.MIDSCENE_MODEL_NAME; exports.default = src_default; exports.describeUserPage = _chunk5HHR4GGSjs.describeUserPage; exports.getAIConfig = _env.getAIConfig; exports.getVersion = _chunkCDRBBE7Djs.getVersion; exports.plan = _chunk5HHR4GGSjs.plan;
 
 //# sourceMappingURL=index.js.map

package/dist/lib/index.js.map

@@ -1 +1 @@
-{"version":3,"mappings":";;;;;;;;;;;;;;;;AAUA,SAAS,qBAAqB,mBAAmB;AACjD,SAAS,cAAc;AAEhB,IAAM,WAAN,MAAe;AAAA,EAUpB,YACE,MACA,SAGA;AACA,SAAK,SACH,SAAS,SAAS,QAAQ,MAAM,SAAS,IAAI,YAAY;AAC3D,SAAK,OAAO;AACZ,SAAK,SAAS,SAAS,SAAS,CAAC,GAAG;AAAA,MAAI,CAAC,SACvC,KAAK,kBAAkB,IAAI;AAAA,IAC7B;AACA,SAAK,cAAc,SAAS;AAAA,EAC9B;AAAA,EAEQ,kBAAkB,MAAyC;AACjE,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,MAAgE;AAC3E;AAAA,MACE,KAAK,WAAW;AAAA,MAChB;AAAA,QAAyD,KAAK,gBAAgB,GAAG,KAAK;AAAA,EAAK,KAAK,gBAAgB,GAAG,UAAU;AAAA,IAC/H;AACA,QAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,WAAK,MAAM,KAAK,GAAG,KAAK,IAAI,CAAC,SAAS,KAAK,kBAAkB,IAAI,CAAC,CAAC;AAAA,IACrE,OAAO;AACL,WAAK,MAAM,KAAK,KAAK,kBAAkB,IAAI,CAAC;AAAA,IAC9C;AACA,QAAI,KAAK,WAAW,WAAW;AAC7B,WAAK,SAAS;AAAA,IAChB;AAAA,EACF;AAAA,EAEA,MAAM,QAAsB;AAC1B,QAAI,KAAK,WAAW,UAAU,KAAK,MAAM,SAAS,GAAG;AACnD,cAAQ;AAAA,QACN;AAAA,MACF;AAAA,IACF;AAEA,WAAO,KAAK,WAAW,WAAW,6BAA6B;AAC/D,WAAO,KAAK,WAAW,aAAa,+BAA+B;AACnE,WAAO,KAAK,WAAW,SAAS,4BAA4B;AAE5D,UAAM,mBAAmB,KAAK,MAAM;AAAA,MAClC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,mBAAmB,GAAG;AAExB;AAAA,IACF;AAEA,SAAK,SAAS;AACd,QAAI,YAAY;AAChB,QAAI,wBAAwB;AAE5B,QAAI;AAEJ,WAAO,YAAY,KAAK,MAAM,QAAQ;AACpC,YAAM,OAAO,KAAK,MAAM,SAAS;AACjC;AAAA,QACE,KAAK,WAAW;AAAA,QAChB,2CAA2C,KAAK,MAAM;AAAA,MACxD;AACA,WAAK,SAAS;AAAA,QACZ,OAAO,KAAK,IAAI;AAAA,MAClB;AACA,UAAI;AACF,aAAK,SAAS;AACd,YAAI;AACF,cAAI,KAAK,aAAa;AACpB,kBAAM,KAAK,YAAY,IAAI;AAAA,UAC7B;AAAA,QACF,SAAS,GAAG;AACV,kBAAQ,MAAM,wBAAwB,CAAC;AAAA,QACzC;AACA;AAAA,UACE,CAAC,WAAW,UAAU,UAAU,EAAE,QAAQ,KAAK,IAAI,KAAK;AAAA,UACxD,0BAA0B,KAAK,IAAI;AAAA,QACrC;AAEA,cAAM,EAAE,UAAU,MAAM,IAAI;AAC5B,eAAO,UAAU,uCAAuC,KAAK,IAAI,EAAE;AAEnE,YAAI;AACJ,cAAM,kBAAmC;AAAA,UACvC;AAAA,UACA,SAAS,oBAAoB;AAAA,QAC/B;AAEA,YAAI,KAAK,SAAS,WAAW;AAC3B;AAAA,YACE,KAAK,YAAY,YACf,KAAK,YAAY,WACjB,KAAK,YAAY,YACjB,KAAK,YAAY,aACjB,KAAK,YAAY,YACjB,KAAK,YAAY;AAAA,YACnB,gCAAgC,KAAK,OAAO;AAAA,UAC9C;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AACxD,cAAI,KAAK,YAAY,UAAU;AAC7B,iCACE,aACC;AAAA,UACL;AAAA,QACF,WAAW,KAAK,SAAS,YAAY,KAAK,SAAS,YAAY;AAC7D,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D,OAAO;AACL,kBAAQ;AAAA,YACN,0BAA0B,KAAK,IAAI;AAAA,UACrC;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D;AAEA,eAAO,OAAO,MAAM,WAAW;AAC/B,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD,aAAK,OAAO,SAAU,aAAqB,UAAU;AACrD;AAAA,MACF,SAAS,GAAQ;AACf,gCAAwB;AACxB,aAAK,QACH,GAAG,YAAY,OAAO,MAAM,WAAW,IAAI;AAC7C,aAAK,aAAa,EAAE;AAEpB,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF;AAAA,IACF;AAGA,aAAS,IAAI,YAAY,GAAG,IAAI,KAAK,MAAM,QAAQ,KAAK;AACtD,WAAK,MAAM,CAAC,EAAE,SAAS;AAAA,IACzB;AAEA,QAAI,uBAAuB;AACzB,WAAK,SAAS;AAAA,IAChB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAEA,QAAI,KAAK,MAAM,QAAQ;AAErB,YAAM,cAAc,KAAK,IAAI,WAAW,KAAK,MAAM,SAAS,CAAC;AAC7D,aAAO,KAAK,MAAM,WAAW,EAAE;AAAA,IACjC;AAAA,EACF;AAAA,EAEA,iBAA0B;AACxB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA,EAEA,kBAAwC;AACtC,QAAI,KAAK,WAAW,SAAS;AAC3B,aAAO;AAAA,IACT;AACA,UAAM,iBAAiB,KAAK,MAAM;AAAA,MAChC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,kBAAkB,GAAG;AACvB,aAAO,KAAK,MAAM,cAAc;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EAEA,OAAsB;AACpB,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC9MA,SAAS,4BAA4B;AAE9B,IAAM,8BAA8B,MAAM;AAC/C,SAAO,mNAAmN,qBAAqB,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAoBlP;;;ACQA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA,gBAAAA;AAAA,OACK;AACP,SAAS,yBAAyB,kBAAkB;AACpD,SAAS,gBAAgB;AACzB,SAAS,UAAAC,eAAc;;;AChCvB;AAAA,EACE,uBAAAC;AAAA,EACA,eAAAC;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,YAAY;AAEd,SAAS,gBACd,MACA,gBACA;AACA,MAAI,mBAAmB;AAEvB,MAAI,aAAa,GAAG;AAClB,UAAM,iBAAiB,mBAAmB;AAC1C,QAAI,gBAAgB;AAClB,yBAAmB,WAAW,cAAc;AAAA,IAC9C,OAAO;AACL,yBAAmB,GAAG,aAAa,CAAC;AAAA,IACtC;AAAA,EACF;AAEA,QAAM,WAAqB;AAAA,IACzB,YAAY,WAAW;AAAA,IACvB,SAAS,KAAK,IAAI;AAAA,IAClB,YAAYA,aAAYD,oBAAmB,KAAK;AAAA,IAChD,mBAAmB;AAAA,EACrB;AACA,QAAM,YAAyB;AAAA,IAC7B,OAAO,KAAK;AAAA,IACZ,GAAG;AAAA,IACH,GAAG;AAAA,EACL;AAEA,mBAAiB,SAAS;AAC5B;;;ADQA,IAAM,QAAQ,SAAS,YAAY;AACnC,IAAqB,UAArB,MAGE;AAAA,EAWA,YACE,SAGA,KACA;AAXF,sBAAoD;AAYlD,IAAAD,QAAO,SAAS,iCAAiC;AACjD,QAAI,OAAO,YAAY,YAAY;AACjC,WAAK,qBAAqB;AAAA,IAC5B,OAAO;AACL,WAAK,qBAAqB,MAAM,QAAQ,QAAQ,OAAO;AAAA,IACzD;AAEA,QAAI,OAAO,KAAK,eAAe,aAAa;AAC1C,WAAK,aAAa,IAAI;AAAA,IACxB;AACA,QAAI,OAAO,KAAK,aAAa,aAAa;AACxC,WAAK,WAAW,IAAI;AAAA,IACtB;AAAA,EACF;AAAA,EAEA,MAAM,OACJ,OACA,KACuB;AACvB,UAAM,EAAE,OAAO,IAAI,OAAO,CAAC;AAC3B,UAAM,cAAc,OAAO,UAAU,WAAW,QAAQ,MAAM;AAC9D,IAAAA,QAAO,aAAa,8BAA8B;AAClD,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,IAAAA,QAAO,OAAO,UAAU,UAAU,sCAAsC;AAExE,UAAM,wBAAwB;AAAA,MAC5B;AAAA,IACF;AACA,QAAI,uBAAuB;AACzB,YAAM,yBAAyB,qBAAqB;AAAA,IACtD;AACA,QAAI;AACJ,QAAI,MAAM,aAAa,uBAAuB;AAC5C,yBAAmB,MAAM;AAAA,IAC3B;AAEA,QAAI,oBAAoB,CAACD,cAAa,GAAG;AACvC,cAAQ;AAAA,QACN;AAAA,MACF;AACA,yBAAmB;AAAA,IACrB;AAEA,UAAM,UAAU,KAAK,WAAY,MAAM,KAAK,mBAAmB,QAAQ;AAEvE,QAAI,aAA+B;AACnC,QAAI,wBAA4C;AAChD,QAAI,kBAA2C;AAC/C,QAAI,qBAEY;AAChB,QAAI,kBAAkB;AACpB,2BAAqB,MAAM,gBAAgB;AAAA,QACzC;AAAA,QACA,oBAAoB;AAAA,MACtB,CAAC;AACD,MAAAC;AAAA,QACE,mBAAmB;AAAA,QACnB,gCAAgC,gBAAgB,IAC9C,mBAAmB,QAAQ,KAAK,mBAAmB,KAAK,KAAK,EAC/D;AAAA,MACF;AACA,8BAAwB,mBAAmB;AAC3C,wBAAkB,mBAAmB;AACrC,mBAAa,mBAAmB;AAAA,IAClC;AAEA,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,aAAa,aAAa,MAAM,IACzD,MAAM,gBAAgB;AAAA,MACpB,QAAQ,UAAU,KAAK;AAAA,MACvB;AAAA,MACA,0BAA0B;AAAA,MAC1B,cAAc;AAAA,IAChB,CAAC;AAEH,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,MACvC,gBAAgB,KAAK,UAAU,WAAW;AAAA,MAC1C;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAgC,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAC1E;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT,SAAS;AAAA,MACX;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,aAAa;AAAA,MACb,MAAM;AAAA,MACN;AAAA,MACA,WAAW,CAAC,CAAC;AAAA,MACb,OAAO;AAAA,IACT;AAEA,UAAM,WAA0B,CAAC;AACjC,KAAC,YAAY,YAAY,CAAC,GAAG,QAAQ,CAAC,SAAS;AAC7C,UAAI,QAAQ,MAAM;AAChB,cAAM,UAAU,YAAY,MAAM,EAAE;AAEpC,YAAI,CAAC,SAAS;AACZ,kBAAQ;AAAA,YACN,kCAAkC,KAAK,EAAE;AAAA,UAC3C;AACA;AAAA,QACF;AACA,iBAAS,KAAK,OAAO;AAAA,MACvB;AAAA,IACF,CAAC;AAED;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,IACF;AAEA,QAAI,UAAU;AACZ,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,IAAAA;AAAA,MACE,SAAS,UAAU;AAAA,MACnB,6CAA6C,SAAS,MAAM;AAAA,IAC9D;AAEA,QAAI,SAAS,WAAW,GAAG;AACzB,aAAO;AAAA,QACL,SAAS;AAAA,UACP,IAAI,SAAS,CAAC,EAAG;AAAA,UACjB,SAAS,SAAS,CAAC,EAAG;AAAA,UACtB,QAAQ,SAAS,CAAC,EAAG;AAAA,UACrB,MAAM,SAAS,CAAC,EAAG;AAAA,UACnB,QAAQ,SAAS,CAAC,EAAG,UAAU,CAAC;AAAA,UAChC,YAAY,SAAS,CAAC,EAAG;AAAA,QAC3B;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,WAAO;AAAA,MACL,SAAS;AAAA,MACT;AAAA,IACF;AAAA,EACF;AAAA,EAQA,MAAM,QAAW,YAA+C;AAC9D,IAAAA;AAAA,MACE,OAAO,eAAe,YAAY,OAAO,eAAe;AAAA,MACxD,kDAAkD,OAAO,UAAU;AAAA,IACrE;AACA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,SAAS;AAEvD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,IAAI,MAAM,qBAAwB;AAAA,MAC3D;AAAA,MACA,WAAW;AAAA,IACb,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,IACzC;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAwB,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAClE;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,OAAO;AAAA,IACT;AAEA,UAAM,EAAE,KAAK,IAAI,eAAe,CAAC;AAGjC;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH;AAAA,MACF;AAAA,MACA;AAAA,IACF;AAEA,QAAI,YAAY,CAAC,MAAM;AACrB,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,WAAsD;AACjE,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,QAAQ;AACtD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,eAAe,MAAM,SAAS;AAAA,MAClC;AAAA,MACA;AAAA,IACF,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,aAAa,OAAO;AAAA,IAClD;AAEA,UAAM,EAAE,SAAS,KAAK,IAAI,aAAa;AACvC,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,eAAe;AAAA,MACf,kBAAkB;AAAA,MAClB,OAAO,OAAO,SAAY;AAAA,IAC5B;AACA,oBAAgB,UAAU,cAAc;AAExC,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA,OAAO,aAAa;AAAA,IACtB;AAAA,EACF;AAAA,EACA,MAAM,SACJ,QACA,KAGyD;AACzD,IAAAA,QAAO,QAAQ,yCAAyC;AACxD,UAAM,UAAU,MAAM,KAAK,mBAAmB,UAAU;AACxD,UAAM,EAAE,iBAAiB,IAAI;AAC7B,IAAAA,QAAO,kBAAkB,6CAA6C;AAEtE,UAAM,eAAe,4BAA4B;AAGjD,UAAM,kBAAkB;AACxB,UAAM,aAAmB,MAAM,QAAQ,MAAM,IACzC;AAAA,MACE,MAAM,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAChD,KAAK,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAC/C,OAAO;AAAA,MACP,QAAQ;AAAA,IACV,IACA;AAEJ,QAAI,eAAe,MAAM,wBAAwB;AAAA,MAC/C,gBAAgB;AAAA,MAChB,sBAAsB;AAAA,QACpB;AAAA,UACE,MAAM;AAAA,QACR;AAAA,MACF;AAAA,MACA,iBAAiB;AAAA,IACnB,CAAC;AAED,QAAI,KAAK,WAAW;AAClB,YAAM,aAAa,iBAAiB,YAAY,QAAQ,IAAI;AAC5D,YAAM,4BAA4B,UAAU;AAC5C,qBAAe,MAAM;AAAA,QACnB;AAAA,QACA;AAAA,QACA,qBAAqB,oBAAoB;AAAA,MAC3C;AAAA,IACF;AAEA,UAAM,OAAe;AAAA,MACnB,EAAE,MAAM,UAAU,SAAS,aAAa;AAAA,MACxC;AAAA,QACE,MAAM;AAAA,QACN,SAAS;AAAA,UACP;AAAA,YACE,MAAM;AAAA,YACN,WAAW;AAAA,cACT,KAAK;AAAA,cACL,QAAQ;AAAA,YACV;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,WACJ,KAAK,cAAc;AAErB,UAAM,MAAM,MAAM,SAAS,8BAAmC;AAE9D,UAAM,EAAE,QAAQ,IAAI;AACpB,IAAAA,QAAO,CAAC,QAAQ,OAAO,oBAAoB,QAAQ,KAAK,EAAE;AAC1D,IAAAA,QAAO,QAAQ,aAAa,gCAAgC;AAC5D,WAAO;AAAA,EACT;AACF;;;AEjZA,SAAS,eAAAE,cAAa,uBAAAD,4BAA2B;AAGjD,IAAO,cAAQ","names":["vlLocateMode","assert","MIDSCENE_MODEL_NAME","getAIConfig"],"ignoreList":[],"sources":["../../src/ai-model/action-executor.ts","../../src/ai-model/prompt/describe.ts","../../src/insight/index.ts","../../src/insight/utils.ts","../../src/index.ts"],"sourcesContent":["import type {\n  ExecutionDump,\n  ExecutionTask,\n  ExecutionTaskApply,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskProgressOptions,\n  ExecutionTaskReturn,\n  ExecutorContext,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport { MIDSCENE_MODEL_NAME, getAIConfig } from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\n\nexport class Executor {\n  name: string;\n\n  tasks: ExecutionTask[];\n\n  // status of executor\n  status: 'init' | 'pending' | 'running' | 'completed' | 'error';\n\n  onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n\n  constructor(\n    name: string,\n    options?: ExecutionTaskProgressOptions & {\n      tasks?: ExecutionTaskApply[];\n    },\n  ) {\n    this.status =\n      options?.tasks && options.tasks.length > 0 ? 'pending' : 'init';\n    this.name = name;\n    this.tasks = (options?.tasks || []).map((item) =>\n      this.markTaskAsPending(item),\n    );\n    this.onTaskStart = options?.onTaskStart;\n  }\n\n  private markTaskAsPending(task: ExecutionTaskApply): ExecutionTask {\n    return {\n      status: 'pending',\n      ...task,\n    };\n  }\n\n  async append(task: ExecutionTaskApply[] | ExecutionTaskApply): Promise<void> {\n    assert(\n      this.status !== 'error',\n      `executor is in error state, cannot append task\\nerror=${this.latestErrorTask()?.error}\\n${this.latestErrorTask()?.errorStack}`,\n    );\n    if (Array.isArray(task)) {\n      this.tasks.push(...task.map((item) => this.markTaskAsPending(item)));\n    } else {\n      this.tasks.push(this.markTaskAsPending(task));\n    }\n    if (this.status !== 'running') {\n      this.status = 'pending';\n    }\n  }\n\n  async flush(): Promise<any> {\n    if (this.status === 'init' && this.tasks.length > 0) {\n      console.warn(\n        'illegal state for executor, status is init but tasks are not empty',\n      );\n    }\n\n    assert(this.status !== 'running', 'executor is already running');\n    assert(this.status !== 'completed', 'executor is already completed');\n    assert(this.status !== 'error', 'executor is in error state');\n\n    const nextPendingIndex = this.tasks.findIndex(\n      (task) => task.status === 'pending',\n    );\n    if (nextPendingIndex < 0) {\n      // all tasks are completed\n      return;\n    }\n\n    this.status = 'running';\n    let taskIndex = nextPendingIndex;\n    let successfullyCompleted = true;\n\n    let previousFindOutput: ExecutionTaskInsightLocateOutput | undefined;\n\n    while (taskIndex < this.tasks.length) {\n      const task = this.tasks[taskIndex];\n      assert(\n        task.status === 'pending',\n        `task status should be pending, but got: ${task.status}`,\n      );\n      task.timing = {\n        start: Date.now(),\n      };\n      try {\n        task.status = 'running';\n        try {\n          if (this.onTaskStart) {\n            await this.onTaskStart(task);\n          }\n        } catch (e) {\n          console.error('error in onTaskStart', e);\n        }\n        assert(\n          ['Insight', 'Action', 'Planning'].indexOf(task.type) >= 0,\n          `unsupported task type: ${task.type}`,\n        );\n\n        const { executor, param } = task;\n        assert(executor, `executor is required for task type: ${task.type}`);\n\n        let returnValue;\n        const executorContext: ExecutorContext = {\n          task,\n          element: previousFindOutput?.element,\n        };\n\n        if (task.type === 'Insight') {\n          assert(\n            task.subType === 'Locate' ||\n              task.subType === 'Query' ||\n              task.subType === 'Assert' ||\n              task.subType === 'Boolean' ||\n              task.subType === 'Number' ||\n              task.subType === 'String',\n            `unsupported insight subType: ${task.subType}`,\n          );\n          returnValue = await task.executor(param, executorContext);\n          if (task.subType === 'Locate') {\n            previousFindOutput = (\n              returnValue as ExecutionTaskReturn<ExecutionTaskInsightLocateOutput>\n            )?.output;\n          }\n        } else if (task.type === 'Action' || task.type === 'Planning') {\n          returnValue = await task.executor(param, executorContext);\n        } else {\n          console.warn(\n            `unsupported task type: ${task.type}, will try to execute it directly`,\n          );\n          returnValue = await task.executor(param, executorContext);\n        }\n\n        Object.assign(task, returnValue);\n        task.status = 'finished';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        task.timing.aiCost = (returnValue as any)?.aiCost || 0;\n        taskIndex++;\n      } catch (e: any) {\n        successfullyCompleted = false;\n        task.error =\n          e?.message || (typeof e === 'string' ? e : 'error-without-message');\n        task.errorStack = e.stack;\n\n        task.status = 'failed';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        break;\n      }\n    }\n\n    // set all remaining tasks as cancelled\n    for (let i = taskIndex + 1; i < this.tasks.length; i++) {\n      this.tasks[i].status = 'cancelled';\n    }\n\n    if (successfullyCompleted) {\n      this.status = 'completed';\n    } else {\n      this.status = 'error';\n    }\n\n    if (this.tasks.length) {\n      // return the last output\n      const outputIndex = Math.min(taskIndex, this.tasks.length - 1);\n      return this.tasks[outputIndex].output;\n    }\n  }\n\n  isInErrorState(): boolean {\n    return this.status === 'error';\n  }\n\n  latestErrorTask(): ExecutionTask | null {\n    if (this.status !== 'error') {\n      return null;\n    }\n    const errorTaskIndex = this.tasks.findIndex(\n      (task) => task.status === 'failed',\n    );\n    if (errorTaskIndex >= 0) {\n      return this.tasks[errorTaskIndex];\n    }\n    return null;\n  }\n\n  dump(): ExecutionDump {\n    const dumpData: ExecutionDump = {\n      sdkVersion: getVersion(),\n      model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n      logTime: Date.now(),\n      name: this.name,\n      tasks: this.tasks,\n    };\n    return dumpData;\n  }\n}\n","import { getPreferredLanguage } from '@midscene/shared/env';\n\nexport const elementDescriberInstruction = () => {\n  return `Tell what is the content of the element wrapped by the read rectangle in the screenshot. Your description is expected to be used to precisely locate the element from other similar elements on screenshot. Use ${getPreferredLanguage()} in the description.\n\nPlease follow the following rules:\n1. The description should be start with a brief description, like \"a button for confirming the action\".\n\n2. Include these information in the description to distinguish the element from its siblings and other similar elements, as much as possible:\n- The text of the element, like \"with text 'Confirm'\"\n- What the element looks like if it's an image, like \"with image '...'\"\n- The relative position of the element, like \"on the left of ..., around ...\"\n- How to distinguish the element from its siblings elements, like \"it is the icon instead of the text\"\n\n3. Do NOT mention the red rectangle in the description.\n\n4. Use the error field to describe the unexpected situations, if any. If not, put null.\n\nReturn in JSON:\n{\n  \"description\": \"[{brief description}]: {text of the element} {image of the element} {relative position of the element} ... \",\n  \"error\"?: \"...\"\n}`;\n};\n","import {\n  AIActionType,\n  type AIArgs,\n  callAiFn,\n  expandSearchArea,\n} from '@/ai-model/common';\nimport {\n  AiExtractElementInfo,\n  AiLocateElement,\n  callToGetJSONObject,\n} from '@/ai-model/index';\nimport { AiAssert, AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n  AIDescribeElementResponse,\n  AIElementResponse,\n  AISingleElementResponse,\n  AIUsageInfo,\n  BaseElement,\n  DetailedLocateParam,\n  DumpSubscriber,\n  InsightAction,\n  InsightAssertionResponse,\n  InsightExtractParam,\n  InsightOptions,\n  InsightTaskInfo,\n  LocateResult,\n  PartialInsightDumpFromSDK,\n  Rect,\n  UIContext,\n} from '@/types';\nimport {\n  MIDSCENE_FORCE_DEEP_THINK,\n  MIDSCENE_USE_QWEN_VL,\n  getAIConfigInBoolean,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext<BaseElement>;\n  callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n  ElementType extends BaseElement = BaseElement,\n  ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n  contextRetrieverFn: (\n    action: InsightAction,\n  ) => Promise<ContextType> | ContextType;\n\n  aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n  onceDumpUpdatedFn?: DumpSubscriber;\n\n  taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n  constructor(\n    context:\n      | ContextType\n      | ((action: InsightAction) => Promise<ContextType> | ContextType),\n    opt?: InsightOptions,\n  ) {\n    assert(context, 'context is required for Insight');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    if (typeof opt?.aiVendorFn !== 'undefined') {\n      this.aiVendorFn = opt.aiVendorFn;\n    }\n    if (typeof opt?.taskInfo !== 'undefined') {\n      this.taskInfo = opt.taskInfo;\n    }\n  }\n\n  async locate(\n    query: DetailedLocateParam,\n    opt?: LocateOpts,\n  ): Promise<LocateResult> {\n    const { callAI } = opt || {};\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    const globalDeepThinkSwitch = getAIConfigInBoolean(\n      MIDSCENE_FORCE_DEEP_THINK,\n    );\n    if (globalDeepThinkSwitch) {\n      debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n    }\n    let searchAreaPrompt;\n    if (query.deepThink || globalDeepThinkSwitch) {\n      searchAreaPrompt = query.prompt;\n    }\n\n    if (searchAreaPrompt && !vlLocateMode()) {\n      console.warn(\n        'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n      );\n      searchAreaPrompt = undefined;\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n    let searchArea: Rect | undefined = undefined;\n    let searchAreaRawResponse: string | undefined = undefined;\n    let searchAreaUsage: AIUsageInfo | undefined = undefined;\n    let searchAreaResponse:\n      | Awaited<ReturnType<typeof AiLocateSection>>\n      | undefined = undefined;\n    if (searchAreaPrompt) {\n      searchAreaResponse = await AiLocateSection({\n        context,\n        sectionDescription: searchAreaPrompt,\n      });\n      assert(\n        searchAreaResponse.rect,\n        `cannot find search area for \"${searchAreaPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const { parseResult, rect, elementById, rawResponse, usage } =\n      await AiLocateElement({\n        callAI: callAI || this.aiVendorFn,\n        context,\n        targetElementDescription: queryPrompt,\n        searchConfig: searchAreaResponse,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedElement: [],\n      matchedRect: rect,\n      data: null,\n      taskInfo,\n      deepThink: !!searchArea,\n      error: errorLog,\n    };\n\n    const elements: BaseElement[] = [];\n    (parseResult.elements || []).forEach((item) => {\n      if ('id' in item) {\n        const element = elementById(item?.id);\n\n        if (!element) {\n          console.warn(\n            `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n          );\n          return;\n        }\n        elements.push(element);\n      }\n    });\n\n    emitInsightDump(\n      {\n        ...dumpData,\n        matchedElement: elements,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog) {\n      throw new Error(errorLog);\n    }\n\n    assert(\n      elements.length <= 1,\n      `locate: multiple elements found, length = ${elements.length}`,\n    );\n\n    if (elements.length === 1) {\n      return {\n        element: {\n          id: elements[0]!.id,\n          indexId: elements[0]!.indexId,\n          center: elements[0]!.center,\n          rect: elements[0]!.rect,\n          xpaths: elements[0]!.xpaths || [],\n          attributes: elements[0]!.attributes,\n        },\n        rect,\n      };\n    }\n    return {\n      element: null,\n      rect,\n    };\n  }\n\n  async extract<T = any>(input: string): Promise<T>;\n  async extract<T extends Record<string, string>>(\n    input: T,\n  ): Promise<Record<keyof T, any>>;\n  async extract<T extends object>(input: Record<keyof T, string>): Promise<T>;\n\n  async extract<T>(dataDemand: InsightExtractParam): Promise<any> {\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const context = await this.contextRetrieverFn('extract');\n\n    const startTime = Date.now();\n    const { parseResult, usage } = await AiExtractElementInfo<T>({\n      context,\n      dataQuery: dataDemand,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(parseResult),\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data } = parseResult || {};\n\n    // 4\n    emitInsightDump(\n      {\n        ...dumpData,\n        data,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog && !data) {\n      throw new Error(errorLog);\n    }\n\n    return {\n      data,\n      usage,\n    };\n  }\n\n  async assert(assertion: string): Promise<InsightAssertionResponse> {\n    if (typeof assertion !== 'string') {\n      throw new Error(\n        'This is the assert method for Midscene, the first argument should be a string. If you want to use the assert method from Node.js, please import it from the Node.js assert module.',\n      );\n    }\n\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const context = await this.contextRetrieverFn('assert');\n    const startTime = Date.now();\n    const assertResult = await AiAssert({\n      assertion,\n      context,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(assertResult.content),\n    };\n\n    const { thought, pass } = assertResult.content;\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'assert',\n      userQuery: {\n        assertion,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      assertionPass: pass,\n      assertionThought: thought,\n      error: pass ? undefined : thought,\n    };\n    emitInsightDump(dumpData, dumpSubscriber);\n\n    return {\n      pass,\n      thought,\n      usage: assertResult.usage,\n    };\n  }\n  async describe(\n    target: Rect | [number, number],\n    opt?: {\n      deepThink?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for insight.describe');\n    const context = await this.contextRetrieverFn('describe');\n    const { screenshotBase64 } = context;\n    assert(screenshotBase64, 'screenshot is required for insight.describe');\n\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepThink) {\n      const searchArea = expandSearchArea(targetRect, context.size);\n      debug('describe: set searchArea', searchArea);\n      imagePayload = await cropByRect(\n        imagePayload,\n        searchArea,\n        getAIConfigInBoolean(MIDSCENE_USE_QWEN_VL),\n      );\n    }\n\n    const msgs: AIArgs = [\n      { role: 'system', content: systemPrompt },\n      {\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: imagePayload,\n              detail: 'high',\n            },\n          },\n        ],\n      },\n    ];\n\n    const callAIFn =\n      this.aiVendorFn || callToGetJSONObject<AIDescribeElementResponse>;\n\n    const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT);\n\n    const { content } = res;\n    assert(!content.error, `describe failed: ${content.error}`);\n    assert(content.description, 'failed to describe the element');\n    return content;\n  }\n}\n","import type {\n  DumpMeta,\n  DumpSubscriber,\n  InsightDump,\n  PartialInsightDumpFromSDK,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n  MIDSCENE_MODEL_NAME,\n  getAIConfig,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { uuid } from '@midscene/shared/utils';\n\nexport function emitInsightDump(\n  data: PartialInsightDumpFromSDK,\n  dumpSubscriber?: DumpSubscriber,\n) {\n  let modelDescription = '';\n\n  if (vlLocateMode()) {\n    const uiTarsModelVer = uiTarsModelVersion();\n    if (uiTarsModelVer) {\n      modelDescription = `UI-TARS=${uiTarsModelVer}`;\n    } else {\n      modelDescription = `${vlLocateMode()} mode`;\n    }\n  }\n\n  const baseData: DumpMeta = {\n    sdkVersion: getVersion(),\n    logTime: Date.now(),\n    model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n    model_description: modelDescription,\n  };\n  const finalData: InsightDump = {\n    logId: uuid(),\n    ...baseData,\n    ...data,\n  };\n\n  dumpSubscriber?.(finalData);\n}\n","import { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n  plan,\n  describeUserPage,\n  AiLocateElement,\n  AiAssert,\n} from './ai-model/index';\n\nexport { getAIConfig, MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n} from './yaml';\n"]}
+{"version":3,"mappings":";;;;;;;;;;;;;;;;AAUA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,cAAc;AAEhB,IAAM,WAAN,MAAe;AAAA,EAUpB,YACE,MACA,SAGA;AACA,SAAK,SACH,SAAS,SAAS,QAAQ,MAAM,SAAS,IAAI,YAAY;AAC3D,SAAK,OAAO;AACZ,SAAK,SAAS,SAAS,SAAS,CAAC,GAAG;AAAA,MAAI,CAAC,SACvC,KAAK,kBAAkB,IAAI;AAAA,IAC7B;AACA,SAAK,cAAc,SAAS;AAAA,EAC9B;AAAA,EAEQ,kBAAkB,MAAyC;AACjE,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,MAAgE;AAC3E;AAAA,MACE,KAAK,WAAW;AAAA,MAChB;AAAA,QAAyD,KAAK,gBAAgB,GAAG,KAAK;AAAA,EAAK,KAAK,gBAAgB,GAAG,UAAU;AAAA,IAC/H;AACA,QAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,WAAK,MAAM,KAAK,GAAG,KAAK,IAAI,CAAC,SAAS,KAAK,kBAAkB,IAAI,CAAC,CAAC;AAAA,IACrE,OAAO;AACL,WAAK,MAAM,KAAK,KAAK,kBAAkB,IAAI,CAAC;AAAA,IAC9C;AACA,QAAI,KAAK,WAAW,WAAW;AAC7B,WAAK,SAAS;AAAA,IAChB;AAAA,EACF;AAAA,EAEA,MAAM,QAAsB;AAC1B,QAAI,KAAK,WAAW,UAAU,KAAK,MAAM,SAAS,GAAG;AACnD,cAAQ;AAAA,QACN;AAAA,MACF;AAAA,IACF;AAEA,WAAO,KAAK,WAAW,WAAW,6BAA6B;AAC/D,WAAO,KAAK,WAAW,aAAa,+BAA+B;AACnE,WAAO,KAAK,WAAW,SAAS,4BAA4B;AAE5D,UAAM,mBAAmB,KAAK,MAAM;AAAA,MAClC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,mBAAmB,GAAG;AAExB;AAAA,IACF;AAEA,SAAK,SAAS;AACd,QAAI,YAAY;AAChB,QAAI,wBAAwB;AAE5B,QAAI;AAEJ,WAAO,YAAY,KAAK,MAAM,QAAQ;AACpC,YAAM,OAAO,KAAK,MAAM,SAAS;AACjC;AAAA,QACE,KAAK,WAAW;AAAA,QAChB,2CAA2C,KAAK,MAAM;AAAA,MACxD;AACA,WAAK,SAAS;AAAA,QACZ,OAAO,KAAK,IAAI;AAAA,MAClB;AACA,UAAI;AACF,aAAK,SAAS;AACd,YAAI;AACF,cAAI,KAAK,aAAa;AACpB,kBAAM,KAAK,YAAY,IAAI;AAAA,UAC7B;AAAA,QACF,SAAS,GAAG;AACV,kBAAQ,MAAM,wBAAwB,CAAC;AAAA,QACzC;AACA;AAAA,UACE,CAAC,WAAW,UAAU,UAAU,EAAE,QAAQ,KAAK,IAAI,KAAK;AAAA,UACxD,0BAA0B,KAAK,IAAI;AAAA,QACrC;AAEA,cAAM,EAAE,UAAU,MAAM,IAAI;AAC5B,eAAO,UAAU,uCAAuC,KAAK,IAAI,EAAE;AAEnE,YAAI;AACJ,cAAM,kBAAmC;AAAA,UACvC;AAAA,UACA,SAAS,oBAAoB;AAAA,QAC/B;AAEA,YAAI,KAAK,SAAS,WAAW;AAC3B;AAAA,YACE,KAAK,YAAY,YACf,KAAK,YAAY,WACjB,KAAK,YAAY,YACjB,KAAK,YAAY,aACjB,KAAK,YAAY,YACjB,KAAK,YAAY;AAAA,YACnB,gCAAgC,KAAK,OAAO;AAAA,UAC9C;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AACxD,cAAI,KAAK,YAAY,UAAU;AAC7B,iCACE,aACC;AAAA,UACL;AAAA,QACF,WAAW,KAAK,SAAS,YAAY,KAAK,SAAS,YAAY;AAC7D,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D,OAAO;AACL,kBAAQ;AAAA,YACN,0BAA0B,KAAK,IAAI;AAAA,UACrC;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D;AAEA,eAAO,OAAO,MAAM,WAAW;AAC/B,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD,aAAK,OAAO,SAAU,aAAqB,UAAU;AACrD;AAAA,MACF,SAAS,GAAQ;AACf,gCAAwB;AACxB,aAAK,QACH,GAAG,YAAY,OAAO,MAAM,WAAW,IAAI;AAC7C,aAAK,aAAa,EAAE;AAEpB,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF;AAAA,IACF;AAGA,aAAS,IAAI,YAAY,GAAG,IAAI,KAAK,MAAM,QAAQ,KAAK;AACtD,WAAK,MAAM,CAAC,EAAE,SAAS;AAAA,IACzB;AAEA,QAAI,uBAAuB;AACzB,WAAK,SAAS;AAAA,IAChB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAEA,QAAI,KAAK,MAAM,QAAQ;AAErB,YAAM,cAAc,KAAK,IAAI,WAAW,KAAK,MAAM,SAAS,CAAC;AAC7D,aAAO,KAAK,MAAM,WAAW,EAAE;AAAA,IACjC;AAAA,EACF;AAAA,EAEA,iBAA0B;AACxB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA,EAEA,kBAAwC;AACtC,QAAI,KAAK,WAAW,SAAS;AAC3B,aAAO;AAAA,IACT;AACA,UAAM,iBAAiB,KAAK,MAAM;AAAA,MAChC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,kBAAkB,GAAG;AACvB,aAAO,KAAK,MAAM,cAAc;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EAEA,OAAsB;AACpB,QAAI,mBAAmB;AAEvB,QAAI,aAAa,GAAG;AAClB,YAAM,iBAAiB,mBAAmB;AAC1C,UAAI,gBAAgB;AAClB,2BAAmB,WAAW,cAAc;AAAA,MAC9C,OAAO;AACL,2BAAmB,GAAG,aAAa,CAAC;AAAA,MACtC;AAAA,IACF;AACA,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,mBAAmB;AAAA,MACnB,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC9NA,SAAS,4BAA4B;AAE9B,IAAM,8BAA8B,MAAM;AAC/C,SAAO,mNAAmN,qBAAqB,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAoBlP;;;ACOA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA,gBAAAA;AAAA,OACK;AACP,SAAS,yBAAyB,kBAAkB;AACpD,SAAS,gBAAgB;AACzB,SAAS,UAAAC,eAAc;;;AC/BvB;AAAA,EACE,uBAAAC;AAAA,EACA,eAAAC;AAAA,OAGK;AACP,SAAS,YAAY;AAEd,SAAS,gBACd,MACA,gBACA;AACA,QAAM,WAAqB;AAAA,IACzB,YAAY,WAAW;AAAA,IACvB,SAAS,KAAK,IAAI;AAAA,IAClB,YAAYA,aAAYD,oBAAmB,KAAK;AAAA,EAClD;AACA,QAAM,YAAyB;AAAA,IAC7B,OAAO,KAAK;AAAA,IACZ,GAAG;AAAA,IACH,GAAG;AAAA,EACL;AAEA,mBAAiB,SAAS;AAC5B;;;ADmBA,IAAM,QAAQ,SAAS,YAAY;AACnC,IAAqB,UAArB,MAGE;AAAA,EAWA,YACE,SAGA,KACA;AAXF,sBAAoD;AAYlD,IAAAD,QAAO,SAAS,iCAAiC;AACjD,QAAI,OAAO,YAAY,YAAY;AACjC,WAAK,qBAAqB;AAAA,IAC5B,OAAO;AACL,WAAK,qBAAqB,MAAM,QAAQ,QAAQ,OAAO;AAAA,IACzD;AAEA,QAAI,OAAO,KAAK,eAAe,aAAa;AAC1C,WAAK,aAAa,IAAI;AAAA,IACxB;AACA,QAAI,OAAO,KAAK,aAAa,aAAa;AACxC,WAAK,WAAW,IAAI;AAAA,IACtB;AAAA,EACF;AAAA,EAEA,MAAM,OACJ,OACA,KACuB;AACvB,UAAM,EAAE,OAAO,IAAI,OAAO,CAAC;AAC3B,UAAM,cAAc,OAAO,UAAU,WAAW,QAAQ,MAAM;AAC9D,IAAAA,QAAO,aAAa,8BAA8B;AAClD,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,IAAAA,QAAO,OAAO,UAAU,UAAU,sCAAsC;AAExE,UAAM,wBAAwB;AAAA,MAC5B;AAAA,IACF;AACA,QAAI,uBAAuB;AACzB,YAAM,yBAAyB,qBAAqB;AAAA,IACtD;AACA,QAAI;AACJ,QAAI,MAAM,aAAa,uBAAuB;AAC5C,yBAAmB,MAAM;AAAA,IAC3B;AAEA,QAAI,oBAAoB,CAACD,cAAa,GAAG;AACvC,cAAQ;AAAA,QACN;AAAA,MACF;AACA,yBAAmB;AAAA,IACrB;AAEA,UAAM,UAAU,KAAK,WAAY,MAAM,KAAK,mBAAmB,QAAQ;AAEvE,QAAI,aAA+B;AACnC,QAAI,wBAA4C;AAChD,QAAI,kBAA2C;AAC/C,QAAI,qBAEY;AAChB,QAAI,kBAAkB;AACpB,2BAAqB,MAAM,gBAAgB;AAAA,QACzC;AAAA,QACA,oBAAoB;AAAA,MACtB,CAAC;AACD,MAAAC;AAAA,QACE,mBAAmB;AAAA,QACnB,gCAAgC,gBAAgB,IAC9C,mBAAmB,QAAQ,KAAK,mBAAmB,KAAK,KAAK,EAC/D;AAAA,MACF;AACA,8BAAwB,mBAAmB;AAC3C,wBAAkB,mBAAmB;AACrC,mBAAa,mBAAmB;AAAA,IAClC;AAEA,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,aAAa,aAAa,MAAM,IACzD,MAAM,gBAAgB;AAAA,MACpB,QAAQ,UAAU,KAAK;AAAA,MACvB;AAAA,MACA,0BAA0B;AAAA,MAC1B,cAAc;AAAA,IAChB,CAAC;AAEH,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,MACvC,gBAAgB,KAAK,UAAU,WAAW;AAAA,MAC1C;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAgC,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAC1E;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT,SAAS;AAAA,MACX;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,aAAa;AAAA,MACb,MAAM;AAAA,MACN;AAAA,MACA,WAAW,CAAC,CAAC;AAAA,MACb,OAAO;AAAA,IACT;AAEA,UAAM,WAA0B,CAAC;AACjC,KAAC,YAAY,YAAY,CAAC,GAAG,QAAQ,CAAC,SAAS;AAC7C,UAAI,QAAQ,MAAM;AAChB,cAAM,UAAU,YAAY,MAAM,EAAE;AAEpC,YAAI,CAAC,SAAS;AACZ,kBAAQ;AAAA,YACN,kCAAkC,KAAK,EAAE;AAAA,UAC3C;AACA;AAAA,QACF;AACA,iBAAS,KAAK,OAAO;AAAA,MACvB;AAAA,IACF,CAAC;AAED;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,IACF;AAEA,QAAI,UAAU;AACZ,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,IAAAA;AAAA,MACE,SAAS,UAAU;AAAA,MACnB,6CAA6C,SAAS,MAAM;AAAA,IAC9D;AAEA,QAAI,SAAS,WAAW,GAAG;AACzB,aAAO;AAAA,QACL,SAAS;AAAA,UACP,IAAI,SAAS,CAAC,EAAG;AAAA,UACjB,SAAS,SAAS,CAAC,EAAG;AAAA,UACtB,QAAQ,SAAS,CAAC,EAAG;AAAA,UACrB,MAAM,SAAS,CAAC,EAAG;AAAA,UACnB,QAAQ,SAAS,CAAC,EAAG,UAAU,CAAC;AAAA,UAChC,YAAY,SAAS,CAAC,EAAG;AAAA,QAC3B;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,WAAO;AAAA,MACL,SAAS;AAAA,MACT;AAAA,IACF;AAAA,EACF;AAAA,EAQA,MAAM,QAAW,YAA+C;AAC9D,IAAAA;AAAA,MACE,OAAO,eAAe,YAAY,OAAO,eAAe;AAAA,MACxD,kDAAkD,OAAO,UAAU;AAAA,IACrE;AACA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,SAAS;AAEvD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,IAAI,MAAM,qBAAwB;AAAA,MAC3D;AAAA,MACA,WAAW;AAAA,IACb,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,IACzC;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAwB,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAClE;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,OAAO;AAAA,IACT;AAEA,UAAM,EAAE,KAAK,IAAI,eAAe,CAAC;AAGjC;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH;AAAA,MACF;AAAA,MACA;AAAA,IACF;AAEA,QAAI,YAAY,CAAC,MAAM;AACrB,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,WAAsD;AACjE,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,QAAQ;AACtD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,eAAe,MAAM,SAAS;AAAA,MAClC;AAAA,MACA;AAAA,IACF,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,aAAa,OAAO;AAAA,IAClD;AAEA,UAAM,EAAE,SAAS,KAAK,IAAI,aAAa;AACvC,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,eAAe;AAAA,MACf,kBAAkB;AAAA,MAClB,OAAO,OAAO,SAAY;AAAA,IAC5B;AACA,oBAAgB,UAAU,cAAc;AAExC,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA,OAAO,aAAa;AAAA,IACtB;AAAA,EACF;AAAA,EACA,MAAM,SACJ,QACA,KAGyD;AACzD,IAAAA,QAAO,QAAQ,yCAAyC;AACxD,UAAM,UAAU,MAAM,KAAK,mBAAmB,UAAU;AACxD,UAAM,EAAE,iBAAiB,IAAI;AAC7B,IAAAA,QAAO,kBAAkB,6CAA6C;AAEtE,UAAM,eAAe,4BAA4B;AAGjD,UAAM,kBAAkB;AACxB,UAAM,aAAmB,MAAM,QAAQ,MAAM,IACzC;AAAA,MACE,MAAM,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAChD,KAAK,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAC/C,OAAO;AAAA,MACP,QAAQ;AAAA,IACV,IACA;AAEJ,QAAI,eAAe,MAAM,wBAAwB;AAAA,MAC/C,gBAAgB;AAAA,MAChB,sBAAsB;AAAA,QACpB;AAAA,UACE,MAAM;AAAA,QACR;AAAA,MACF;AAAA,MACA,iBAAiB;AAAA,IACnB,CAAC;AAED,QAAI,KAAK,WAAW;AAClB,YAAM,aAAa,iBAAiB,YAAY,QAAQ,IAAI;AAC5D,YAAM,4BAA4B,UAAU;AAC5C,qBAAe,MAAM;AAAA,QACnB;AAAA,QACA;AAAA,QACA,qBAAqB,oBAAoB;AAAA,MAC3C;AAAA,IACF;AAEA,UAAM,OAAe;AAAA,MACnB,EAAE,MAAM,UAAU,SAAS,aAAa;AAAA,MACxC;AAAA,QACE,MAAM;AAAA,QACN,SAAS;AAAA,UACP;AAAA,YACE,MAAM;AAAA,YACN,WAAW;AAAA,cACT,KAAK;AAAA,cACL,QAAQ;AAAA,YACV;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,WACJ,KAAK,cAAc;AAErB,UAAM,MAAM,MAAM,SAAS,8BAAmC;AAE9D,UAAM,EAAE,QAAQ,IAAI;AACpB,IAAAA,QAAO,CAAC,QAAQ,OAAO,oBAAoB,QAAQ,KAAK,EAAE;AAC1D,IAAAA,QAAO,QAAQ,aAAa,gCAAgC;AAC5D,WAAO;AAAA,EACT;AACF;;;AEhZA,SAAS,eAAAE,cAAa,uBAAAD,4BAA2B;AAGjD,IAAO,cAAQ","names":["vlLocateMode","assert","MIDSCENE_MODEL_NAME","getAIConfig"],"ignoreList":[],"sources":["../../src/ai-model/action-executor.ts","../../src/ai-model/prompt/describe.ts","../../src/insight/index.ts","../../src/insight/utils.ts","../../src/index.ts"],"sourcesContent":["import type {\n  ExecutionDump,\n  ExecutionTask,\n  ExecutionTaskApply,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskProgressOptions,\n  ExecutionTaskReturn,\n  ExecutorContext,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n  MIDSCENE_MODEL_NAME,\n  getAIConfig,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\n\nexport class Executor {\n  name: string;\n\n  tasks: ExecutionTask[];\n\n  // status of executor\n  status: 'init' | 'pending' | 'running' | 'completed' | 'error';\n\n  onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n\n  constructor(\n    name: string,\n    options?: ExecutionTaskProgressOptions & {\n      tasks?: ExecutionTaskApply[];\n    },\n  ) {\n    this.status =\n      options?.tasks && options.tasks.length > 0 ? 'pending' : 'init';\n    this.name = name;\n    this.tasks = (options?.tasks || []).map((item) =>\n      this.markTaskAsPending(item),\n    );\n    this.onTaskStart = options?.onTaskStart;\n  }\n\n  private markTaskAsPending(task: ExecutionTaskApply): ExecutionTask {\n    return {\n      status: 'pending',\n      ...task,\n    };\n  }\n\n  async append(task: ExecutionTaskApply[] | ExecutionTaskApply): Promise<void> {\n    assert(\n      this.status !== 'error',\n      `executor is in error state, cannot append task\\nerror=${this.latestErrorTask()?.error}\\n${this.latestErrorTask()?.errorStack}`,\n    );\n    if (Array.isArray(task)) {\n      this.tasks.push(...task.map((item) => this.markTaskAsPending(item)));\n    } else {\n      this.tasks.push(this.markTaskAsPending(task));\n    }\n    if (this.status !== 'running') {\n      this.status = 'pending';\n    }\n  }\n\n  async flush(): Promise<any> {\n    if (this.status === 'init' && this.tasks.length > 0) {\n      console.warn(\n        'illegal state for executor, status is init but tasks are not empty',\n      );\n    }\n\n    assert(this.status !== 'running', 'executor is already running');\n    assert(this.status !== 'completed', 'executor is already completed');\n    assert(this.status !== 'error', 'executor is in error state');\n\n    const nextPendingIndex = this.tasks.findIndex(\n      (task) => task.status === 'pending',\n    );\n    if (nextPendingIndex < 0) {\n      // all tasks are completed\n      return;\n    }\n\n    this.status = 'running';\n    let taskIndex = nextPendingIndex;\n    let successfullyCompleted = true;\n\n    let previousFindOutput: ExecutionTaskInsightLocateOutput | undefined;\n\n    while (taskIndex < this.tasks.length) {\n      const task = this.tasks[taskIndex];\n      assert(\n        task.status === 'pending',\n        `task status should be pending, but got: ${task.status}`,\n      );\n      task.timing = {\n        start: Date.now(),\n      };\n      try {\n        task.status = 'running';\n        try {\n          if (this.onTaskStart) {\n            await this.onTaskStart(task);\n          }\n        } catch (e) {\n          console.error('error in onTaskStart', e);\n        }\n        assert(\n          ['Insight', 'Action', 'Planning'].indexOf(task.type) >= 0,\n          `unsupported task type: ${task.type}`,\n        );\n\n        const { executor, param } = task;\n        assert(executor, `executor is required for task type: ${task.type}`);\n\n        let returnValue;\n        const executorContext: ExecutorContext = {\n          task,\n          element: previousFindOutput?.element,\n        };\n\n        if (task.type === 'Insight') {\n          assert(\n            task.subType === 'Locate' ||\n              task.subType === 'Query' ||\n              task.subType === 'Assert' ||\n              task.subType === 'Boolean' ||\n              task.subType === 'Number' ||\n              task.subType === 'String',\n            `unsupported insight subType: ${task.subType}`,\n          );\n          returnValue = await task.executor(param, executorContext);\n          if (task.subType === 'Locate') {\n            previousFindOutput = (\n              returnValue as ExecutionTaskReturn<ExecutionTaskInsightLocateOutput>\n            )?.output;\n          }\n        } else if (task.type === 'Action' || task.type === 'Planning') {\n          returnValue = await task.executor(param, executorContext);\n        } else {\n          console.warn(\n            `unsupported task type: ${task.type}, will try to execute it directly`,\n          );\n          returnValue = await task.executor(param, executorContext);\n        }\n\n        Object.assign(task, returnValue);\n        task.status = 'finished';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        task.timing.aiCost = (returnValue as any)?.aiCost || 0;\n        taskIndex++;\n      } catch (e: any) {\n        successfullyCompleted = false;\n        task.error =\n          e?.message || (typeof e === 'string' ? e : 'error-without-message');\n        task.errorStack = e.stack;\n\n        task.status = 'failed';\n        task.timing.end = Date.now();\n        task.timing.cost = task.timing.end - task.timing.start;\n        break;\n      }\n    }\n\n    // set all remaining tasks as cancelled\n    for (let i = taskIndex + 1; i < this.tasks.length; i++) {\n      this.tasks[i].status = 'cancelled';\n    }\n\n    if (successfullyCompleted) {\n      this.status = 'completed';\n    } else {\n      this.status = 'error';\n    }\n\n    if (this.tasks.length) {\n      // return the last output\n      const outputIndex = Math.min(taskIndex, this.tasks.length - 1);\n      return this.tasks[outputIndex].output;\n    }\n  }\n\n  isInErrorState(): boolean {\n    return this.status === 'error';\n  }\n\n  latestErrorTask(): ExecutionTask | null {\n    if (this.status !== 'error') {\n      return null;\n    }\n    const errorTaskIndex = this.tasks.findIndex(\n      (task) => task.status === 'failed',\n    );\n    if (errorTaskIndex >= 0) {\n      return this.tasks[errorTaskIndex];\n    }\n    return null;\n  }\n\n  dump(): ExecutionDump {\n    let modelDescription = '';\n\n    if (vlLocateMode()) {\n      const uiTarsModelVer = uiTarsModelVersion();\n      if (uiTarsModelVer) {\n        modelDescription = `UI-TARS=${uiTarsModelVer}`;\n      } else {\n        modelDescription = `${vlLocateMode()} mode`;\n      }\n    }\n    const dumpData: ExecutionDump = {\n      sdkVersion: getVersion(),\n      model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n      model_description: modelDescription,\n      logTime: Date.now(),\n      name: this.name,\n      tasks: this.tasks,\n    };\n    return dumpData;\n  }\n}\n","import { getPreferredLanguage } from '@midscene/shared/env';\n\nexport const elementDescriberInstruction = () => {\n  return `Tell what is the content of the element wrapped by the read rectangle in the screenshot. Your description is expected to be used to precisely locate the element from other similar elements on screenshot. Use ${getPreferredLanguage()} in the description.\n\nPlease follow the following rules:\n1. The description should be start with a brief description, like \"a button for confirming the action\".\n\n2. Include these information in the description to distinguish the element from its siblings and other similar elements, as much as possible:\n- The text of the element, like \"with text 'Confirm'\"\n- What the element looks like if it's an image, like \"with image '...'\"\n- The relative position of the element, like \"on the left of ..., around ...\"\n- How to distinguish the element from its siblings elements, like \"it is the icon instead of the text\"\n\n3. Do NOT mention the red rectangle in the description.\n\n4. Use the error field to describe the unexpected situations, if any. If not, put null.\n\nReturn in JSON:\n{\n  \"description\": \"[{brief description}]: {text of the element} {image of the element} {relative position of the element} ... \",\n  \"error\"?: \"...\"\n}`;\n};\n","import {\n  AIActionType,\n  type AIArgs,\n  callAiFn,\n  expandSearchArea,\n} from '@/ai-model/common';\nimport {\n  AiExtractElementInfo,\n  AiLocateElement,\n  callToGetJSONObject,\n} from '@/ai-model/index';\nimport { AiAssert, AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n  AIDescribeElementResponse,\n  AIElementResponse,\n  AIUsageInfo,\n  BaseElement,\n  DetailedLocateParam,\n  DumpSubscriber,\n  InsightAction,\n  InsightAssertionResponse,\n  InsightExtractParam,\n  InsightOptions,\n  InsightTaskInfo,\n  LocateResult,\n  PartialInsightDumpFromSDK,\n  Rect,\n  UIContext,\n} from '@/types';\nimport {\n  MIDSCENE_FORCE_DEEP_THINK,\n  MIDSCENE_USE_QWEN_VL,\n  getAIConfigInBoolean,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext<BaseElement>;\n  callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n  ElementType extends BaseElement = BaseElement,\n  ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n  contextRetrieverFn: (\n    action: InsightAction,\n  ) => Promise<ContextType> | ContextType;\n\n  aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n  onceDumpUpdatedFn?: DumpSubscriber;\n\n  taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n  constructor(\n    context:\n      | ContextType\n      | ((action: InsightAction) => Promise<ContextType> | ContextType),\n    opt?: InsightOptions,\n  ) {\n    assert(context, 'context is required for Insight');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    if (typeof opt?.aiVendorFn !== 'undefined') {\n      this.aiVendorFn = opt.aiVendorFn;\n    }\n    if (typeof opt?.taskInfo !== 'undefined') {\n      this.taskInfo = opt.taskInfo;\n    }\n  }\n\n  async locate(\n    query: DetailedLocateParam,\n    opt?: LocateOpts,\n  ): Promise<LocateResult> {\n    const { callAI } = opt || {};\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    const globalDeepThinkSwitch = getAIConfigInBoolean(\n      MIDSCENE_FORCE_DEEP_THINK,\n    );\n    if (globalDeepThinkSwitch) {\n      debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n    }\n    let searchAreaPrompt;\n    if (query.deepThink || globalDeepThinkSwitch) {\n      searchAreaPrompt = query.prompt;\n    }\n\n    if (searchAreaPrompt && !vlLocateMode()) {\n      console.warn(\n        'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n      );\n      searchAreaPrompt = undefined;\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n    let searchArea: Rect | undefined = undefined;\n    let searchAreaRawResponse: string | undefined = undefined;\n    let searchAreaUsage: AIUsageInfo | undefined = undefined;\n    let searchAreaResponse:\n      | Awaited<ReturnType<typeof AiLocateSection>>\n      | undefined = undefined;\n    if (searchAreaPrompt) {\n      searchAreaResponse = await AiLocateSection({\n        context,\n        sectionDescription: searchAreaPrompt,\n      });\n      assert(\n        searchAreaResponse.rect,\n        `cannot find search area for \"${searchAreaPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const { parseResult, rect, elementById, rawResponse, usage } =\n      await AiLocateElement({\n        callAI: callAI || this.aiVendorFn,\n        context,\n        targetElementDescription: queryPrompt,\n        searchConfig: searchAreaResponse,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedElement: [],\n      matchedRect: rect,\n      data: null,\n      taskInfo,\n      deepThink: !!searchArea,\n      error: errorLog,\n    };\n\n    const elements: BaseElement[] = [];\n    (parseResult.elements || []).forEach((item) => {\n      if ('id' in item) {\n        const element = elementById(item?.id);\n\n        if (!element) {\n          console.warn(\n            `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n          );\n          return;\n        }\n        elements.push(element);\n      }\n    });\n\n    emitInsightDump(\n      {\n        ...dumpData,\n        matchedElement: elements,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog) {\n      throw new Error(errorLog);\n    }\n\n    assert(\n      elements.length <= 1,\n      `locate: multiple elements found, length = ${elements.length}`,\n    );\n\n    if (elements.length === 1) {\n      return {\n        element: {\n          id: elements[0]!.id,\n          indexId: elements[0]!.indexId,\n          center: elements[0]!.center,\n          rect: elements[0]!.rect,\n          xpaths: elements[0]!.xpaths || [],\n          attributes: elements[0]!.attributes,\n        },\n        rect,\n      };\n    }\n    return {\n      element: null,\n      rect,\n    };\n  }\n\n  async extract<T = any>(input: string): Promise<T>;\n  async extract<T extends Record<string, string>>(\n    input: T,\n  ): Promise<Record<keyof T, any>>;\n  async extract<T extends object>(input: Record<keyof T, string>): Promise<T>;\n\n  async extract<T>(dataDemand: InsightExtractParam): Promise<any> {\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const context = await this.contextRetrieverFn('extract');\n\n    const startTime = Date.now();\n    const { parseResult, usage } = await AiExtractElementInfo<T>({\n      context,\n      dataQuery: dataDemand,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(parseResult),\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data } = parseResult || {};\n\n    // 4\n    emitInsightDump(\n      {\n        ...dumpData,\n        data,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog && !data) {\n      throw new Error(errorLog);\n    }\n\n    return {\n      data,\n      usage,\n    };\n  }\n\n  async assert(assertion: string): Promise<InsightAssertionResponse> {\n    if (typeof assertion !== 'string') {\n      throw new Error(\n        'This is the assert method for Midscene, the first argument should be a string. If you want to use the assert method from Node.js, please import it from the Node.js assert module.',\n      );\n    }\n\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const context = await this.contextRetrieverFn('assert');\n    const startTime = Date.now();\n    const assertResult = await AiAssert({\n      assertion,\n      context,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(assertResult.content),\n    };\n\n    const { thought, pass } = assertResult.content;\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'assert',\n      userQuery: {\n        assertion,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      assertionPass: pass,\n      assertionThought: thought,\n      error: pass ? undefined : thought,\n    };\n    emitInsightDump(dumpData, dumpSubscriber);\n\n    return {\n      pass,\n      thought,\n      usage: assertResult.usage,\n    };\n  }\n  async describe(\n    target: Rect | [number, number],\n    opt?: {\n      deepThink?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for insight.describe');\n    const context = await this.contextRetrieverFn('describe');\n    const { screenshotBase64 } = context;\n    assert(screenshotBase64, 'screenshot is required for insight.describe');\n\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepThink) {\n      const searchArea = expandSearchArea(targetRect, context.size);\n      debug('describe: set searchArea', searchArea);\n      imagePayload = await cropByRect(\n        imagePayload,\n        searchArea,\n        getAIConfigInBoolean(MIDSCENE_USE_QWEN_VL),\n      );\n    }\n\n    const msgs: AIArgs = [\n      { role: 'system', content: systemPrompt },\n      {\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: imagePayload,\n              detail: 'high',\n            },\n          },\n        ],\n      },\n    ];\n\n    const callAIFn =\n      this.aiVendorFn || callToGetJSONObject<AIDescribeElementResponse>;\n\n    const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT);\n\n    const { content } = res;\n    assert(!content.error, `describe failed: ${content.error}`);\n    assert(content.description, 'failed to describe the element');\n    return content;\n  }\n}\n","import type {\n  DumpMeta,\n  DumpSubscriber,\n  InsightDump,\n  PartialInsightDumpFromSDK,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n  MIDSCENE_MODEL_NAME,\n  getAIConfig,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { uuid } from '@midscene/shared/utils';\n\nexport function emitInsightDump(\n  data: PartialInsightDumpFromSDK,\n  dumpSubscriber?: DumpSubscriber,\n) {\n  const baseData: DumpMeta = {\n    sdkVersion: getVersion(),\n    logTime: Date.now(),\n    model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n  };\n  const finalData: InsightDump = {\n    logId: uuid(),\n    ...baseData,\n    ...data,\n  };\n\n  dumpSubscriber?.(finalData);\n}\n","import { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n  plan,\n  describeUserPage,\n  AiLocateElement,\n  AiAssert,\n} from './ai-model/index';\n\nexport { getAIConfig, MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n  MidsceneYamlFlowItemAIRightClick,\n} from './yaml';\n"]}

package/dist/lib/{llm-planning-9c97f5c2.d.ts → llm-planning-9cfa38ad.d.ts}

@@ -1,4 +1,4 @@
-import { k as AIUsageInfo, R as Rect, j as ElementTreeNode, B as BaseElement, U as UIContext, ax as ReferenceImage, p as AIElementLocatorResponse, O as ElementById, s as AIDataExtractionResponse, u as AIAssertionResponse, av as PageType, Y as PlanningAIResponse } from './types-81bb2c02.js';
+import { l as AIUsageInfo, R as Rect, k as ElementTreeNode, B as BaseElement, U as UIContext, az as ReferenceImage, q as AIElementLocatorResponse, Q as ElementById, t as AIDataExtractionResponse, v as AIAssertionResponse, ax as PageType, Z as PlanningAIResponse } from './types-c892e193.js';
 import { ChatCompletionSystemMessageParam, ChatCompletionUserMessageParam } from 'openai/resources';
 
 type AIArgs = [

package/dist/lib/tree.d.ts

@@ -1,5 +1,5 @@
 import * as _midscene_shared_constants from '@midscene/shared/constants';
-import { B as BaseElement, j as ElementTreeNode } from './types-81bb2c02.js';
+import { B as BaseElement, k as ElementTreeNode } from './types-c892e193.js';
 import 'openai/resources';
 
 declare function truncateText(text: string | number | object | undefined, maxLength?: number): string;

package/dist/{types/types-81bb2c02.d.ts → lib/types-c892e193.d.ts}

@@ -64,6 +64,7 @@ interface MidsceneYamlFlowItemAIAction {
 }
 interface MidsceneYamlFlowItemAIAssert {
     aiAssert: string;
+    errorMessage?: string;
 }
 interface MidsceneYamlFlowItemAIQuery {
     aiQuery: string;
@@ -92,6 +93,9 @@ interface MidsceneYamlFlowItemAIWaitFor {
 interface MidsceneYamlFlowItemAITap extends LocateOption {
     aiTap: string;
 }
+interface MidsceneYamlFlowItemAIRightClick extends LocateOption {
+    aiRightClick: string;
+}
 interface MidsceneYamlFlowItemAIHover extends LocateOption {
     aiHover: string;
 }
@@ -114,7 +118,7 @@ interface MidsceneYamlFlowItemEvaluateJavaScript {
 interface MidsceneYamlFlowItemSleep {
     sleep: number;
 }
-type MidsceneYamlFlowItem = MidsceneYamlFlowItemAIAction | MidsceneYamlFlowItemAIAssert | MidsceneYamlFlowItemAIQuery | MidsceneYamlFlowItemAIWaitFor | MidsceneYamlFlowItemAITap | MidsceneYamlFlowItemAIHover | MidsceneYamlFlowItemAIInput | MidsceneYamlFlowItemAIKeyboardPress | MidsceneYamlFlowItemAIScroll | MidsceneYamlFlowItemSleep;
+type MidsceneYamlFlowItem = MidsceneYamlFlowItemAIAction | MidsceneYamlFlowItemAIAssert | MidsceneYamlFlowItemAIQuery | MidsceneYamlFlowItemAIWaitFor | MidsceneYamlFlowItemAITap | MidsceneYamlFlowItemAIRightClick | MidsceneYamlFlowItemAIHover | MidsceneYamlFlowItemAIInput | MidsceneYamlFlowItemAIKeyboardPress | MidsceneYamlFlowItemAIScroll | MidsceneYamlFlowItemSleep;
 interface FreeFn {
     name: string;
     fn: () => void;
@@ -338,7 +342,7 @@ interface PlanningLocateParam extends DetailedLocateParam {
 }
 interface PlanningAction<ParamType = any> {
     thought?: string;
-    type: 'Locate' | 'Tap' | 'Drag' | 'Hover' | 'Input' | 'KeyboardPress' | 'Scroll' | 'Error' | 'ExpectedFalsyCondition' | 'Assert' | 'AssertWithoutThrow' | 'Sleep' | 'Finished' | 'AndroidBackButton' | 'AndroidHomeButton' | 'AndroidRecentAppsButton';
+    type: 'Locate' | 'Tap' | 'RightClick' | 'Hover' | 'Drag' | 'Input' | 'KeyboardPress' | 'Scroll' | 'Error' | 'ExpectedFalsyCondition' | 'Assert' | 'AssertWithoutThrow' | 'Sleep' | 'Finished' | 'AndroidBackButton' | 'AndroidHomeButton' | 'AndroidRecentAppsButton';
     param: ParamType;
     locate?: PlanningLocateParam | null;
 }
@@ -356,6 +360,7 @@ interface PlanningAIResponse {
 }
 type PlanningActionParamTap = null;
 type PlanningActionParamHover = null;
+type PlanningActionParamRightClick = null;
 interface PlanningActionParamInputOrKeyPress {
     value: string;
 }
@@ -472,4 +477,4 @@ interface GroupedActionDump {
 }
 type PageType = 'puppeteer' | 'playwright' | 'static' | 'chrome-extension-proxy' | 'android';
 
-export { type PlanningActionParamInputOrKeyPress as $, type AIDescribeElementResponse as A, BaseElement as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type LocateResultElement as F, type DumpMeta as G, type ReportDumpWithAttributes as H, type InsightAction as I, type InsightDump as J, type PartialInsightDumpFromSDK as K, type LocateResult as L, type MidsceneYamlScript as M, type LiteUISection as N, type ElementById as O, type Point as P, type OnTaskStartTip as Q, type Rect as R, type Size as S, type AgentWaitForOpt as T, UIContext as U, type AgentAssertOpt as V, type PlanningLocateParam as W, type PlanningAction as X, type PlanningAIResponse as Y, type PlanningActionParamTap as Z, type PlanningActionParamHover as _, type ExecutionTaskProgressOptions as a, type PlanningActionParamScroll as a0, type PlanningActionParamAssert as a1, type PlanningActionParamSleep as a2, type PlanningActionParamError as a3, type PlanningActionParamWaitFor as a4, type Color as a5, type BaseAgentParserOpt as a6, type PuppeteerParserOpt as a7, type PlaywrightParserOpt as a8, type ExecutionRecorderItem as a9, type MidsceneYamlScriptWebEnv as aA, type MidsceneYamlScriptAndroidEnv as aB, type MidsceneYamlScriptEnv as aC, type MidsceneYamlFlowItemAIAction as aD, type MidsceneYamlFlowItemAIAssert as aE, type MidsceneYamlFlowItemAIQuery as aF, type MidsceneYamlFlowItemAINumber as aG, type MidsceneYamlFlowItemAINString as aH, type MidsceneYamlFlowItemAIBoolean as aI, type MidsceneYamlFlowItemAILocate as aJ, type MidsceneYamlFlowItemAIWaitFor as aK, type MidsceneYamlFlowItemAITap as aL, type MidsceneYamlFlowItemAIHover as aM, type MidsceneYamlFlowItemAIInput as aN, type MidsceneYamlFlowItemAIKeyboardPress as aO, type MidsceneYamlFlowItemAIScroll as aP, type MidsceneYamlFlowItemEvaluateJavaScript as aQ, type MidsceneYamlFlowItemSleep as aR, type FreeFn as aS, type ScriptPlayerTaskStatus as aT, type ScriptPlayerStatusValue as aU, type ExecutionTaskType as aa, type ExecutorContext as ab, type TaskCacheInfo as ac, type ExecutionTaskReturn as ad, type ExecutionTaskInsightLocateParam as ae, type ExecutionTaskInsightLocateOutput as af, type ExecutionTaskInsightDumpLog as ag, type ExecutionTaskInsightLocateApply as ah, type ExecutionTaskInsightLocate as ai, type ExecutionTaskInsightQueryParam as aj, type ExecutionTaskInsightQueryOutput as ak, type ExecutionTaskInsightQueryApply as al, type ExecutionTaskInsightQuery as am, type ExecutionTaskInsightAssertionParam as an, type ExecutionTaskInsightAssertionApply as ao, type ExecutionTaskInsightAssertion as ap, type ExecutionTaskActionApply as aq, type ExecutionTaskAction as ar, type ExecutionTaskPlanningApply as as, type ExecutionTaskPlanning as at, type GroupedActionDump as au, type PageType as av, type LocateOption as aw, type ReferenceImage as ax, type scrollParam as ay, type MidsceneYamlScriptEnvBase as az, type ExecutionTaskApply as b, type ExecutionDump as c, type InsightTaskInfo as d, type InsightOptions as e, type DetailedLocateParam as f, type InsightAssertionResponse as g, type MidsceneYamlTask as h, type MidsceneYamlFlowItem as i, type ElementTreeNode as j, type AIUsageInfo as k, AIResponseFormat as l, type AISingleElementResponseById as m, type AISingleElementResponseByPosition as n, type AISingleElementResponse as o, type AIElementLocatorResponse as p, type AIElementCoordinatesResponse as q, type AIElementResponse as r, type AIDataExtractionResponse as s, type AISectionLocatorResponse as t, type AIAssertionResponse as u, type LocatorValidatorOption as v, type LocateValidatorResult as w, type AgentDescribeElementAtPointResult as x, type EnsureObject as y, type InsightExtractParam as z };
+export { type PlanningActionParamHover as $, type AIDescribeElementResponse as A, BaseElement as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type InsightExtractParam as F, type LocateResultElement as G, type DumpMeta as H, type InsightAction as I, type ReportDumpWithAttributes as J, type InsightDump as K, type LocateResult as L, type MidsceneYamlScript as M, type PartialInsightDumpFromSDK as N, type LiteUISection as O, type Point as P, type ElementById as Q, type Rect as R, type Size as S, type OnTaskStartTip as T, UIContext as U, type AgentWaitForOpt as V, type AgentAssertOpt as W, type PlanningLocateParam as X, type PlanningAction as Y, type PlanningAIResponse as Z, type PlanningActionParamTap as _, type ExecutionTaskProgressOptions as a, type PlanningActionParamRightClick as a0, type PlanningActionParamInputOrKeyPress as a1, type PlanningActionParamScroll as a2, type PlanningActionParamAssert as a3, type PlanningActionParamSleep as a4, type PlanningActionParamError as a5, type PlanningActionParamWaitFor as a6, type Color as a7, type BaseAgentParserOpt as a8, type PuppeteerParserOpt as a9, type scrollParam as aA, type MidsceneYamlScriptEnvBase as aB, type MidsceneYamlScriptWebEnv as aC, type MidsceneYamlScriptAndroidEnv as aD, type MidsceneYamlScriptEnv as aE, type MidsceneYamlFlowItemAIAction as aF, type MidsceneYamlFlowItemAIAssert as aG, type MidsceneYamlFlowItemAIQuery as aH, type MidsceneYamlFlowItemAINumber as aI, type MidsceneYamlFlowItemAINString as aJ, type MidsceneYamlFlowItemAIBoolean as aK, type MidsceneYamlFlowItemAILocate as aL, type MidsceneYamlFlowItemAIWaitFor as aM, type MidsceneYamlFlowItemAITap as aN, type MidsceneYamlFlowItemAIHover as aO, type MidsceneYamlFlowItemAIInput as aP, type MidsceneYamlFlowItemAIKeyboardPress as aQ, type MidsceneYamlFlowItemAIScroll as aR, type MidsceneYamlFlowItemEvaluateJavaScript as aS, type MidsceneYamlFlowItemSleep as aT, type FreeFn as aU, type ScriptPlayerTaskStatus as aV, type ScriptPlayerStatusValue as aW, type PlaywrightParserOpt as aa, type ExecutionRecorderItem as ab, type ExecutionTaskType as ac, type ExecutorContext as ad, type TaskCacheInfo as ae, type ExecutionTaskReturn as af, type ExecutionTaskInsightLocateParam as ag, type ExecutionTaskInsightLocateOutput as ah, type ExecutionTaskInsightDumpLog as ai, type ExecutionTaskInsightLocateApply as aj, type ExecutionTaskInsightLocate as ak, type ExecutionTaskInsightQueryParam as al, type ExecutionTaskInsightQueryOutput as am, type ExecutionTaskInsightQueryApply as an, type ExecutionTaskInsightQuery as ao, type ExecutionTaskInsightAssertionParam as ap, type ExecutionTaskInsightAssertionApply as aq, type ExecutionTaskInsightAssertion as ar, type ExecutionTaskActionApply as as, type ExecutionTaskAction as at, type ExecutionTaskPlanningApply as au, type ExecutionTaskPlanning as av, type GroupedActionDump as aw, type PageType as ax, type LocateOption as ay, type ReferenceImage as az, type ExecutionTaskApply as b, type ExecutionDump as c, type InsightTaskInfo as d, type InsightOptions as e, type DetailedLocateParam as f, type InsightAssertionResponse as g, type MidsceneYamlTask as h, type MidsceneYamlFlowItem as i, type MidsceneYamlFlowItemAIRightClick as j, type ElementTreeNode as k, type AIUsageInfo as l, AIResponseFormat as m, type AISingleElementResponseById as n, type AISingleElementResponseByPosition as o, type AISingleElementResponse as p, type AIElementLocatorResponse as q, type AIElementCoordinatesResponse as r, type AIElementResponse as s, type AIDataExtractionResponse as t, type AISectionLocatorResponse as u, type AIAssertionResponse as v, type LocatorValidatorOption as w, type LocateValidatorResult as x, type AgentDescribeElementAtPointResult as y, type EnsureObject as z };

package/dist/lib/utils.d.ts

@@ -1,4 +1,4 @@
-import { H as ReportDumpWithAttributes, R as Rect } from './types-81bb2c02.js';
+import { J as ReportDumpWithAttributes, R as Rect } from './types-c892e193.js';
 import '@midscene/shared/constants';
 import 'openai/resources';
 

package/dist/lib/utils.js

@@ -12,7 +12,7 @@
 
 
 
-var _chunkJ2RYPDNYjs = require('./chunk-J2RYPDNY.js');
+var _chunkCDRBBE7Djs = require('./chunk-CDRBBE7D.js');
 
 
 
@@ -27,4 +27,4 @@ var _chunkJ2RYPDNYjs = require('./chunk-J2RYPDNY.js');
 
 
 
-exports.getTmpDir = _chunkJ2RYPDNYjs.getTmpDir; exports.getTmpFile = _chunkJ2RYPDNYjs.getTmpFile; exports.getVersion = _chunkJ2RYPDNYjs.getVersion; exports.groupedActionDumpFileExt = _chunkJ2RYPDNYjs.groupedActionDumpFileExt; exports.overlapped = _chunkJ2RYPDNYjs.overlapped; exports.replaceStringWithFirstAppearance = _chunkJ2RYPDNYjs.replaceStringWithFirstAppearance; exports.replacerForPageObject = _chunkJ2RYPDNYjs.replacerForPageObject; exports.reportHTMLContent = _chunkJ2RYPDNYjs.reportHTMLContent; exports.sleep = _chunkJ2RYPDNYjs.sleep; exports.stringifyDumpData = _chunkJ2RYPDNYjs.stringifyDumpData; exports.uploadTestInfoToServer = _chunkJ2RYPDNYjs.uploadTestInfoToServer; exports.writeDumpReport = _chunkJ2RYPDNYjs.writeDumpReport; exports.writeLogFile = _chunkJ2RYPDNYjs.writeLogFile;
+exports.getTmpDir = _chunkCDRBBE7Djs.getTmpDir; exports.getTmpFile = _chunkCDRBBE7Djs.getTmpFile; exports.getVersion = _chunkCDRBBE7Djs.getVersion; exports.groupedActionDumpFileExt = _chunkCDRBBE7Djs.groupedActionDumpFileExt; exports.overlapped = _chunkCDRBBE7Djs.overlapped; exports.replaceStringWithFirstAppearance = _chunkCDRBBE7Djs.replaceStringWithFirstAppearance; exports.replacerForPageObject = _chunkCDRBBE7Djs.replacerForPageObject; exports.reportHTMLContent = _chunkCDRBBE7Djs.reportHTMLContent; exports.sleep = _chunkCDRBBE7Djs.sleep; exports.stringifyDumpData = _chunkCDRBBE7Djs.stringifyDumpData; exports.uploadTestInfoToServer = _chunkCDRBBE7Djs.uploadTestInfoToServer; exports.writeDumpReport = _chunkCDRBBE7Djs.writeDumpReport; exports.writeLogFile = _chunkCDRBBE7Djs.writeLogFile;

package/dist/types/ai-model.d.ts

@@ -1,8 +1,8 @@
-import { k as AIUsageInfo, X as PlanningAction, i as MidsceneYamlFlowItem, S as Size } from './types-81bb2c02.js';
+import { l as AIUsageInfo, Y as PlanningAction, i as MidsceneYamlFlowItem, S as Size } from './types-c892e193.js';
 import { ChatCompletionMessageParam } from 'openai/resources';
 export { ChatCompletionMessageParam } from 'openai/resources';
-import { b as AIActionType } from './llm-planning-9c97f5c2.js';
-export { a as AiAssert, f as AiExtractElementInfo, A as AiLocateElement, g as AiLocateSection, h as adaptBboxToRect, c as callAiFn, d as describeUserPage, e as elementByPositionWithElementInfo, p as plan } from './llm-planning-9c97f5c2.js';
+import { b as AIActionType } from './llm-planning-9cfa38ad.js';
+export { a as AiAssert, f as AiExtractElementInfo, A as AiLocateElement, g as AiLocateSection, h as adaptBboxToRect, c as callAiFn, d as describeUserPage, e as elementByPositionWithElementInfo, p as plan } from './llm-planning-9cfa38ad.js';
 import { vlLocateMode } from '@midscene/shared/env';
 import { actionParser } from '@ui-tars/action-parser';
 import '@midscene/shared/constants';

package/dist/types/index.d.ts

@@ -1,7 +1,7 @@
-import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, B as BaseElement, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightAssertionResponse, R as Rect, A as AIDescribeElementResponse } from './types-81bb2c02.js';
-export { u as AIAssertionResponse, s as AIDataExtractionResponse, q as AIElementCoordinatesResponse, p as AIElementLocatorResponse, r as AIElementResponse, l as AIResponseFormat, t as AISectionLocatorResponse, o as AISingleElementResponse, m as AISingleElementResponseById, n as AISingleElementResponseByPosition, k as AIUsageInfo, V as AgentAssertOpt, x as AgentDescribeElementAtPointResult, T as AgentWaitForOpt, a6 as BaseAgentParserOpt, C as CallAIFn, a5 as Color, G as DumpMeta, O as ElementById, j as ElementTreeNode, y as EnsureObject, a9 as ExecutionRecorderItem, ar as ExecutionTaskAction, aq as ExecutionTaskActionApply, ap as ExecutionTaskInsightAssertion, ao as ExecutionTaskInsightAssertionApply, an as ExecutionTaskInsightAssertionParam, ag as ExecutionTaskInsightDumpLog, ai as ExecutionTaskInsightLocate, ah as ExecutionTaskInsightLocateApply, af as ExecutionTaskInsightLocateOutput, ae as ExecutionTaskInsightLocateParam, am as ExecutionTaskInsightQuery, al as ExecutionTaskInsightQueryApply, ak as ExecutionTaskInsightQueryOutput, aj as ExecutionTaskInsightQueryParam, at as ExecutionTaskPlanning, as as ExecutionTaskPlanningApply, ad as ExecutionTaskReturn, aa as ExecutionTaskType, ab as ExecutorContext, aS as FreeFn, au as GroupedActionDump, J as InsightDump, z as InsightExtractParam, N as LiteUISection, aw as LocateOption, F as LocateResultElement, w as LocateValidatorResult, v as LocatorValidatorOption, i as MidsceneYamlFlowItem, aD as MidsceneYamlFlowItemAIAction, aE as MidsceneYamlFlowItemAIAssert, aI as MidsceneYamlFlowItemAIBoolean, aM as MidsceneYamlFlowItemAIHover, aN as MidsceneYamlFlowItemAIInput, aO as MidsceneYamlFlowItemAIKeyboardPress, aJ as MidsceneYamlFlowItemAILocate, aH as MidsceneYamlFlowItemAINString, aG as MidsceneYamlFlowItemAINumber, aF as MidsceneYamlFlowItemAIQuery, aP as MidsceneYamlFlowItemAIScroll, aL as MidsceneYamlFlowItemAITap, aK as MidsceneYamlFlowItemAIWaitFor, aQ as MidsceneYamlFlowItemEvaluateJavaScript, aR as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aB as MidsceneYamlScriptAndroidEnv, aC as MidsceneYamlScriptEnv, az as MidsceneYamlScriptEnvBase, aA as MidsceneYamlScriptWebEnv, h as MidsceneYamlTask, Q as OnTaskStartTip, av as PageType, K as PartialInsightDumpFromSDK, Y as PlanningAIResponse, X as PlanningAction, a1 as PlanningActionParamAssert, a3 as PlanningActionParamError, _ as PlanningActionParamHover, $ as PlanningActionParamInputOrKeyPress, a0 as PlanningActionParamScroll, a2 as PlanningActionParamSleep, Z as PlanningActionParamTap, a4 as PlanningActionParamWaitFor, W as PlanningLocateParam, a8 as PlaywrightParserOpt, P as Point, a7 as PuppeteerParserOpt, ax as ReferenceImage, H as ReportDumpWithAttributes, aU as ScriptPlayerStatusValue, aT as ScriptPlayerTaskStatus, S as Size, ac as TaskCacheInfo, ay as scrollParam } from './types-81bb2c02.js';
-import { c as callAiFn } from './llm-planning-9c97f5c2.js';
-export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-9c97f5c2.js';
+import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, B as BaseElement, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightAssertionResponse, R as Rect, A as AIDescribeElementResponse } from './types-c892e193.js';
+export { v as AIAssertionResponse, t as AIDataExtractionResponse, r as AIElementCoordinatesResponse, q as AIElementLocatorResponse, s as AIElementResponse, m as AIResponseFormat, u as AISectionLocatorResponse, p as AISingleElementResponse, n as AISingleElementResponseById, o as AISingleElementResponseByPosition, l as AIUsageInfo, W as AgentAssertOpt, y as AgentDescribeElementAtPointResult, V as AgentWaitForOpt, a8 as BaseAgentParserOpt, C as CallAIFn, a7 as Color, H as DumpMeta, Q as ElementById, k as ElementTreeNode, z as EnsureObject, ab as ExecutionRecorderItem, at as ExecutionTaskAction, as as ExecutionTaskActionApply, ar as ExecutionTaskInsightAssertion, aq as ExecutionTaskInsightAssertionApply, ap as ExecutionTaskInsightAssertionParam, ai as ExecutionTaskInsightDumpLog, ak as ExecutionTaskInsightLocate, aj as ExecutionTaskInsightLocateApply, ah as ExecutionTaskInsightLocateOutput, ag as ExecutionTaskInsightLocateParam, ao as ExecutionTaskInsightQuery, an as ExecutionTaskInsightQueryApply, am as ExecutionTaskInsightQueryOutput, al as ExecutionTaskInsightQueryParam, av as ExecutionTaskPlanning, au as ExecutionTaskPlanningApply, af as ExecutionTaskReturn, ac as ExecutionTaskType, ad as ExecutorContext, aU as FreeFn, aw as GroupedActionDump, K as InsightDump, F as InsightExtractParam, O as LiteUISection, ay as LocateOption, G as LocateResultElement, x as LocateValidatorResult, w as LocatorValidatorOption, i as MidsceneYamlFlowItem, aF as MidsceneYamlFlowItemAIAction, aG as MidsceneYamlFlowItemAIAssert, aK as MidsceneYamlFlowItemAIBoolean, aO as MidsceneYamlFlowItemAIHover, aP as MidsceneYamlFlowItemAIInput, aQ as MidsceneYamlFlowItemAIKeyboardPress, aL as MidsceneYamlFlowItemAILocate, aJ as MidsceneYamlFlowItemAINString, aI as MidsceneYamlFlowItemAINumber, aH as MidsceneYamlFlowItemAIQuery, j as MidsceneYamlFlowItemAIRightClick, aR as MidsceneYamlFlowItemAIScroll, aN as MidsceneYamlFlowItemAITap, aM as MidsceneYamlFlowItemAIWaitFor, aS as MidsceneYamlFlowItemEvaluateJavaScript, aT as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aD as MidsceneYamlScriptAndroidEnv, aE as MidsceneYamlScriptEnv, aB as MidsceneYamlScriptEnvBase, aC as MidsceneYamlScriptWebEnv, h as MidsceneYamlTask, T as OnTaskStartTip, ax as PageType, N as PartialInsightDumpFromSDK, Z as PlanningAIResponse, Y as PlanningAction, a3 as PlanningActionParamAssert, a5 as PlanningActionParamError, $ as PlanningActionParamHover, a1 as PlanningActionParamInputOrKeyPress, a0 as PlanningActionParamRightClick, a2 as PlanningActionParamScroll, a4 as PlanningActionParamSleep, _ as PlanningActionParamTap, a6 as PlanningActionParamWaitFor, X as PlanningLocateParam, aa as PlaywrightParserOpt, P as Point, a9 as PuppeteerParserOpt, az as ReferenceImage, J as ReportDumpWithAttributes, aW as ScriptPlayerStatusValue, aV as ScriptPlayerTaskStatus, S as Size, ae as TaskCacheInfo, aA as scrollParam } from './types-c892e193.js';
+import { c as callAiFn } from './llm-planning-9cfa38ad.js';
+export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-9cfa38ad.js';
 export { getVersion } from './utils.js';
 export { MIDSCENE_MODEL_NAME, getAIConfig } from '@midscene/shared/env';
 import '@midscene/shared/constants';

package/dist/types/{llm-planning-9c97f5c2.d.ts → llm-planning-9cfa38ad.d.ts}

@@ -1,4 +1,4 @@
-import { k as AIUsageInfo, R as Rect, j as ElementTreeNode, B as BaseElement, U as UIContext, ax as ReferenceImage, p as AIElementLocatorResponse, O as ElementById, s as AIDataExtractionResponse, u as AIAssertionResponse, av as PageType, Y as PlanningAIResponse } from './types-81bb2c02.js';
+import { l as AIUsageInfo, R as Rect, k as ElementTreeNode, B as BaseElement, U as UIContext, az as ReferenceImage, q as AIElementLocatorResponse, Q as ElementById, t as AIDataExtractionResponse, v as AIAssertionResponse, ax as PageType, Z as PlanningAIResponse } from './types-c892e193.js';
 import { ChatCompletionSystemMessageParam, ChatCompletionUserMessageParam } from 'openai/resources';
 
 type AIArgs = [

package/dist/types/tree.d.ts

@@ -1,5 +1,5 @@
 import * as _midscene_shared_constants from '@midscene/shared/constants';
-import { B as BaseElement, j as ElementTreeNode } from './types-81bb2c02.js';
+import { B as BaseElement, k as ElementTreeNode } from './types-c892e193.js';
 import 'openai/resources';
 
 declare function truncateText(text: string | number | object | undefined, maxLength?: number): string;

package/dist/{lib/types-81bb2c02.d.ts → types/types-c892e193.d.ts}

@@ -64,6 +64,7 @@ interface MidsceneYamlFlowItemAIAction {
 }
 interface MidsceneYamlFlowItemAIAssert {
     aiAssert: string;
+    errorMessage?: string;
 }
 interface MidsceneYamlFlowItemAIQuery {
     aiQuery: string;
@@ -92,6 +93,9 @@ interface MidsceneYamlFlowItemAIWaitFor {
 interface MidsceneYamlFlowItemAITap extends LocateOption {
     aiTap: string;
 }
+interface MidsceneYamlFlowItemAIRightClick extends LocateOption {
+    aiRightClick: string;
+}
 interface MidsceneYamlFlowItemAIHover extends LocateOption {
     aiHover: string;
 }
@@ -114,7 +118,7 @@ interface MidsceneYamlFlowItemEvaluateJavaScript {
 interface MidsceneYamlFlowItemSleep {
     sleep: number;
 }
-type MidsceneYamlFlowItem = MidsceneYamlFlowItemAIAction | MidsceneYamlFlowItemAIAssert | MidsceneYamlFlowItemAIQuery | MidsceneYamlFlowItemAIWaitFor | MidsceneYamlFlowItemAITap | MidsceneYamlFlowItemAIHover | MidsceneYamlFlowItemAIInput | MidsceneYamlFlowItemAIKeyboardPress | MidsceneYamlFlowItemAIScroll | MidsceneYamlFlowItemSleep;
+type MidsceneYamlFlowItem = MidsceneYamlFlowItemAIAction | MidsceneYamlFlowItemAIAssert | MidsceneYamlFlowItemAIQuery | MidsceneYamlFlowItemAIWaitFor | MidsceneYamlFlowItemAITap | MidsceneYamlFlowItemAIRightClick | MidsceneYamlFlowItemAIHover | MidsceneYamlFlowItemAIInput | MidsceneYamlFlowItemAIKeyboardPress | MidsceneYamlFlowItemAIScroll | MidsceneYamlFlowItemSleep;
 interface FreeFn {
     name: string;
     fn: () => void;
@@ -338,7 +342,7 @@ interface PlanningLocateParam extends DetailedLocateParam {
 }
 interface PlanningAction<ParamType = any> {
     thought?: string;
-    type: 'Locate' | 'Tap' | 'Drag' | 'Hover' | 'Input' | 'KeyboardPress' | 'Scroll' | 'Error' | 'ExpectedFalsyCondition' | 'Assert' | 'AssertWithoutThrow' | 'Sleep' | 'Finished' | 'AndroidBackButton' | 'AndroidHomeButton' | 'AndroidRecentAppsButton';
+    type: 'Locate' | 'Tap' | 'RightClick' | 'Hover' | 'Drag' | 'Input' | 'KeyboardPress' | 'Scroll' | 'Error' | 'ExpectedFalsyCondition' | 'Assert' | 'AssertWithoutThrow' | 'Sleep' | 'Finished' | 'AndroidBackButton' | 'AndroidHomeButton' | 'AndroidRecentAppsButton';
     param: ParamType;
     locate?: PlanningLocateParam | null;
 }
@@ -356,6 +360,7 @@ interface PlanningAIResponse {
 }
 type PlanningActionParamTap = null;
 type PlanningActionParamHover = null;
+type PlanningActionParamRightClick = null;
 interface PlanningActionParamInputOrKeyPress {
     value: string;
 }
@@ -472,4 +477,4 @@ interface GroupedActionDump {
 }
 type PageType = 'puppeteer' | 'playwright' | 'static' | 'chrome-extension-proxy' | 'android';
 
-export { type PlanningActionParamInputOrKeyPress as $, type AIDescribeElementResponse as A, BaseElement as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type LocateResultElement as F, type DumpMeta as G, type ReportDumpWithAttributes as H, type InsightAction as I, type InsightDump as J, type PartialInsightDumpFromSDK as K, type LocateResult as L, type MidsceneYamlScript as M, type LiteUISection as N, type ElementById as O, type Point as P, type OnTaskStartTip as Q, type Rect as R, type Size as S, type AgentWaitForOpt as T, UIContext as U, type AgentAssertOpt as V, type PlanningLocateParam as W, type PlanningAction as X, type PlanningAIResponse as Y, type PlanningActionParamTap as Z, type PlanningActionParamHover as _, type ExecutionTaskProgressOptions as a, type PlanningActionParamScroll as a0, type PlanningActionParamAssert as a1, type PlanningActionParamSleep as a2, type PlanningActionParamError as a3, type PlanningActionParamWaitFor as a4, type Color as a5, type BaseAgentParserOpt as a6, type PuppeteerParserOpt as a7, type PlaywrightParserOpt as a8, type ExecutionRecorderItem as a9, type MidsceneYamlScriptWebEnv as aA, type MidsceneYamlScriptAndroidEnv as aB, type MidsceneYamlScriptEnv as aC, type MidsceneYamlFlowItemAIAction as aD, type MidsceneYamlFlowItemAIAssert as aE, type MidsceneYamlFlowItemAIQuery as aF, type MidsceneYamlFlowItemAINumber as aG, type MidsceneYamlFlowItemAINString as aH, type MidsceneYamlFlowItemAIBoolean as aI, type MidsceneYamlFlowItemAILocate as aJ, type MidsceneYamlFlowItemAIWaitFor as aK, type MidsceneYamlFlowItemAITap as aL, type MidsceneYamlFlowItemAIHover as aM, type MidsceneYamlFlowItemAIInput as aN, type MidsceneYamlFlowItemAIKeyboardPress as aO, type MidsceneYamlFlowItemAIScroll as aP, type MidsceneYamlFlowItemEvaluateJavaScript as aQ, type MidsceneYamlFlowItemSleep as aR, type FreeFn as aS, type ScriptPlayerTaskStatus as aT, type ScriptPlayerStatusValue as aU, type ExecutionTaskType as aa, type ExecutorContext as ab, type TaskCacheInfo as ac, type ExecutionTaskReturn as ad, type ExecutionTaskInsightLocateParam as ae, type ExecutionTaskInsightLocateOutput as af, type ExecutionTaskInsightDumpLog as ag, type ExecutionTaskInsightLocateApply as ah, type ExecutionTaskInsightLocate as ai, type ExecutionTaskInsightQueryParam as aj, type ExecutionTaskInsightQueryOutput as ak, type ExecutionTaskInsightQueryApply as al, type ExecutionTaskInsightQuery as am, type ExecutionTaskInsightAssertionParam as an, type ExecutionTaskInsightAssertionApply as ao, type ExecutionTaskInsightAssertion as ap, type ExecutionTaskActionApply as aq, type ExecutionTaskAction as ar, type ExecutionTaskPlanningApply as as, type ExecutionTaskPlanning as at, type GroupedActionDump as au, type PageType as av, type LocateOption as aw, type ReferenceImage as ax, type scrollParam as ay, type MidsceneYamlScriptEnvBase as az, type ExecutionTaskApply as b, type ExecutionDump as c, type InsightTaskInfo as d, type InsightOptions as e, type DetailedLocateParam as f, type InsightAssertionResponse as g, type MidsceneYamlTask as h, type MidsceneYamlFlowItem as i, type ElementTreeNode as j, type AIUsageInfo as k, AIResponseFormat as l, type AISingleElementResponseById as m, type AISingleElementResponseByPosition as n, type AISingleElementResponse as o, type AIElementLocatorResponse as p, type AIElementCoordinatesResponse as q, type AIElementResponse as r, type AIDataExtractionResponse as s, type AISectionLocatorResponse as t, type AIAssertionResponse as u, type LocatorValidatorOption as v, type LocateValidatorResult as w, type AgentDescribeElementAtPointResult as x, type EnsureObject as y, type InsightExtractParam as z };
+export { type PlanningActionParamHover as $, type AIDescribeElementResponse as A, BaseElement as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type InsightExtractParam as F, type LocateResultElement as G, type DumpMeta as H, type InsightAction as I, type ReportDumpWithAttributes as J, type InsightDump as K, type LocateResult as L, type MidsceneYamlScript as M, type PartialInsightDumpFromSDK as N, type LiteUISection as O, type Point as P, type ElementById as Q, type Rect as R, type Size as S, type OnTaskStartTip as T, UIContext as U, type AgentWaitForOpt as V, type AgentAssertOpt as W, type PlanningLocateParam as X, type PlanningAction as Y, type PlanningAIResponse as Z, type PlanningActionParamTap as _, type ExecutionTaskProgressOptions as a, type PlanningActionParamRightClick as a0, type PlanningActionParamInputOrKeyPress as a1, type PlanningActionParamScroll as a2, type PlanningActionParamAssert as a3, type PlanningActionParamSleep as a4, type PlanningActionParamError as a5, type PlanningActionParamWaitFor as a6, type Color as a7, type BaseAgentParserOpt as a8, type PuppeteerParserOpt as a9, type scrollParam as aA, type MidsceneYamlScriptEnvBase as aB, type MidsceneYamlScriptWebEnv as aC, type MidsceneYamlScriptAndroidEnv as aD, type MidsceneYamlScriptEnv as aE, type MidsceneYamlFlowItemAIAction as aF, type MidsceneYamlFlowItemAIAssert as aG, type MidsceneYamlFlowItemAIQuery as aH, type MidsceneYamlFlowItemAINumber as aI, type MidsceneYamlFlowItemAINString as aJ, type MidsceneYamlFlowItemAIBoolean as aK, type MidsceneYamlFlowItemAILocate as aL, type MidsceneYamlFlowItemAIWaitFor as aM, type MidsceneYamlFlowItemAITap as aN, type MidsceneYamlFlowItemAIHover as aO, type MidsceneYamlFlowItemAIInput as aP, type MidsceneYamlFlowItemAIKeyboardPress as aQ, type MidsceneYamlFlowItemAIScroll as aR, type MidsceneYamlFlowItemEvaluateJavaScript as aS, type MidsceneYamlFlowItemSleep as aT, type FreeFn as aU, type ScriptPlayerTaskStatus as aV, type ScriptPlayerStatusValue as aW, type PlaywrightParserOpt as aa, type ExecutionRecorderItem as ab, type ExecutionTaskType as ac, type ExecutorContext as ad, type TaskCacheInfo as ae, type ExecutionTaskReturn as af, type ExecutionTaskInsightLocateParam as ag, type ExecutionTaskInsightLocateOutput as ah, type ExecutionTaskInsightDumpLog as ai, type ExecutionTaskInsightLocateApply as aj, type ExecutionTaskInsightLocate as ak, type ExecutionTaskInsightQueryParam as al, type ExecutionTaskInsightQueryOutput as am, type ExecutionTaskInsightQueryApply as an, type ExecutionTaskInsightQuery as ao, type ExecutionTaskInsightAssertionParam as ap, type ExecutionTaskInsightAssertionApply as aq, type ExecutionTaskInsightAssertion as ar, type ExecutionTaskActionApply as as, type ExecutionTaskAction as at, type ExecutionTaskPlanningApply as au, type ExecutionTaskPlanning as av, type GroupedActionDump as aw, type PageType as ax, type LocateOption as ay, type ReferenceImage as az, type ExecutionTaskApply as b, type ExecutionDump as c, type InsightTaskInfo as d, type InsightOptions as e, type DetailedLocateParam as f, type InsightAssertionResponse as g, type MidsceneYamlTask as h, type MidsceneYamlFlowItem as i, type MidsceneYamlFlowItemAIRightClick as j, type ElementTreeNode as k, type AIUsageInfo as l, AIResponseFormat as m, type AISingleElementResponseById as n, type AISingleElementResponseByPosition as o, type AISingleElementResponse as p, type AIElementLocatorResponse as q, type AIElementCoordinatesResponse as r, type AIElementResponse as s, type AIDataExtractionResponse as t, type AISectionLocatorResponse as u, type AIAssertionResponse as v, type LocatorValidatorOption as w, type LocateValidatorResult as x, type AgentDescribeElementAtPointResult as y, type EnsureObject as z };