@midscene/core 0.18.0 → 0.18.1-beta-20250611081529.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (36) hide show
  1. package/dist/es/ai-model.d.ts +3 -3
  2. package/dist/es/ai-model.js +1 -1
  3. package/dist/es/{chunk-7MNTWX2A.js → chunk-CLWM3F4J.js} +15 -2
  4. package/dist/es/chunk-CLWM3F4J.js.map +1 -0
  5. package/dist/es/{chunk-YGJNESNY.js → chunk-TO33YH6H.js} +19 -6
  6. package/dist/es/chunk-TO33YH6H.js.map +1 -0
  7. package/dist/es/index.d.ts +4 -4
  8. package/dist/es/index.js +5 -9
  9. package/dist/es/index.js.map +1 -1
  10. package/dist/es/{llm-planning-453ffce5.d.ts → llm-planning-45954424.d.ts} +1 -1
  11. package/dist/es/{types-00c9086f.d.ts → types-e7be1eb0.d.ts} +5 -4
  12. package/dist/es/utils.d.ts +1 -1
  13. package/dist/es/utils.js +1 -1
  14. package/dist/lib/ai-model.d.ts +3 -3
  15. package/dist/lib/ai-model.js +2 -2
  16. package/dist/lib/{chunk-7MNTWX2A.js → chunk-CLWM3F4J.js} +27 -14
  17. package/dist/lib/chunk-CLWM3F4J.js.map +1 -0
  18. package/dist/lib/{chunk-YGJNESNY.js → chunk-TO33YH6H.js} +18 -5
  19. package/dist/lib/chunk-TO33YH6H.js.map +1 -0
  20. package/dist/lib/index.d.ts +4 -4
  21. package/dist/lib/index.js +21 -25
  22. package/dist/lib/index.js.map +1 -1
  23. package/dist/lib/{llm-planning-453ffce5.d.ts → llm-planning-45954424.d.ts} +1 -1
  24. package/dist/{types/types-00c9086f.d.ts → lib/types-e7be1eb0.d.ts} +5 -4
  25. package/dist/lib/utils.d.ts +1 -1
  26. package/dist/lib/utils.js +2 -2
  27. package/dist/types/ai-model.d.ts +3 -3
  28. package/dist/types/index.d.ts +4 -4
  29. package/dist/types/{llm-planning-453ffce5.d.ts → llm-planning-45954424.d.ts} +1 -1
  30. package/dist/{lib/types-00c9086f.d.ts → types/types-e7be1eb0.d.ts} +5 -4
  31. package/dist/types/utils.d.ts +1 -1
  32. package/package.json +2 -2
  33. package/dist/es/chunk-7MNTWX2A.js.map +0 -1
  34. package/dist/es/chunk-YGJNESNY.js.map +0 -1
  35. package/dist/lib/chunk-7MNTWX2A.js.map +0 -1
  36. package/dist/lib/chunk-YGJNESNY.js.map +0 -1
@@ -0,0 +1 @@
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,QAAQ,cAAc;AAC/B,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,UAAM,WAAW,GAAG,UAAU;AAC9B,QAAI,OAAO;AAEX,QAAI,OAAO,aAAa,UAAU;AAChC,aAAO,KAAK,MAAM,QAAQ;AAAA,IAC5B;AAEA;AAAA,MACE;AAAA,MACA,KAAK;AAAA,QACH,KAAK,IAAI,CAAC,UAAU;AAAA,UAClB,YAAY,KAAK,MAAM,KAAK,UAAU;AAAA,UACtC,YAAY,KAAK;AAAA,QACnB,EAAE;AAAA,QACF;AAAA,QACA;AAAA,MACF;AAAA,IACF;AAEA,WAAO,iCAAiC,QAAQ,EAAE;AAAA,EACpD;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, logMsg } 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 `${fileName}.html`,\n );\n\n reportHTMLContent(dumpData, reportPath);\n\n if (process.env.MIDSCENE_DEBUG_LOG_JSON) {\n const jsonPath = `${reportPath}.json`;\n let data = dumpData as ReportDumpWithAttributes[];\n\n if (typeof dumpData === 'string') {\n data = JSON.parse(dumpData) as ReportDumpWithAttributes[];\n }\n\n writeFileSync(\n jsonPath,\n JSON.stringify(\n data.map((item) => ({\n dumpString: JSON.parse(item.dumpString),\n attributes: item.attributes,\n })),\n null,\n 2,\n ),\n );\n\n logMsg(`Midscene - dump file written: ${jsonPath}`);\n }\n\n return reportPath;\n}\n\nexport function writeLogFile(opts: {\n fileName: string;\n fileExt: string;\n fileContent: string;\n type: 'dump' | 'cache' | 'report' | 'tmp';\n generateReport?: boolean;\n}) {\n if (ifInBrowser) {\n return '/mock/report.html';\n }\n const { fileName, fileExt, fileContent, type = 'dump' } = opts;\n const targetDir = getMidsceneRunSubDir(type);\n // Ensure directory exists\n if (!logEnvReady) {\n assert(targetDir, 'logDir should be set before writing dump file');\n\n // gitIgnore in the parent directory\n const gitIgnorePath = path.join(targetDir, '../../.gitignore');\n const gitPath = path.join(targetDir, '../../.git');\n let gitIgnoreContent = '';\n\n if (existsSync(gitPath)) {\n // if the git path exists, we need to add the log folder to the git ignore file\n if (existsSync(gitIgnorePath)) {\n gitIgnoreContent = readFileSync(gitIgnorePath, 'utf-8');\n }\n\n // ignore the log folder\n if (!gitIgnoreContent.includes(`${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"]}
@@ -1,7 +1,7 @@
1
- import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightExtractOption, h as InsightAssertionResponse, A as AIDescribeElementResponse } from './types-00c9086f.js';
2
- export { v as AIAssertionResponse, t as AIDataExtractionResponse, r as AIElementCoordinatesResponse, q as AIElementLocatorResponse, s as AIElementResponse, m as AIResponseFormat, u as AISectionLocatorResponse, p as AISingleElementResponse, n as AISingleElementResponseById, o as AISingleElementResponseByPosition, l as AIUsageInfo, Q as AgentAssertOpt, y as AgentDescribeElementAtPointResult, N as AgentWaitForOpt, a4 as BaseAgentParserOpt, C as CallAIFn, a3 as Color, G as DumpMeta, K as ElementById, z as EnsureObject, a7 as ExecutionRecorderItem, ap as ExecutionTaskAction, ao as ExecutionTaskActionApply, an as ExecutionTaskInsightAssertion, am as ExecutionTaskInsightAssertionApply, al as ExecutionTaskInsightAssertionParam, ae as ExecutionTaskInsightDumpLog, ag as ExecutionTaskInsightLocate, af as ExecutionTaskInsightLocateApply, ad as ExecutionTaskInsightLocateOutput, ac as ExecutionTaskInsightLocateParam, ak as ExecutionTaskInsightQuery, aj as ExecutionTaskInsightQueryApply, ai as ExecutionTaskInsightQueryOutput, ah as ExecutionTaskInsightQueryParam, ar as ExecutionTaskLog, aq as ExecutionTaskLogApply, at as ExecutionTaskPlanning, as as ExecutionTaskPlanningApply, ab as ExecutionTaskReturn, a8 as ExecutionTaskType, a9 as ExecutorContext, aT as FreeFn, au as GroupedActionDump, H as InsightDump, B as InsightExtractParam, J as LiteUISection, aw as LocateOption, F as LocateResultElement, x as LocateValidatorResult, w as LocatorValidatorOption, j as MidsceneYamlFlowItem, aD as MidsceneYamlFlowItemAIAction, aE as MidsceneYamlFlowItemAIAssert, aI as MidsceneYamlFlowItemAIBoolean, aM as MidsceneYamlFlowItemAIHover, aN as MidsceneYamlFlowItemAIInput, aO as MidsceneYamlFlowItemAIKeyboardPress, aJ as MidsceneYamlFlowItemAILocate, aH as MidsceneYamlFlowItemAINString, aG as MidsceneYamlFlowItemAINumber, aF as MidsceneYamlFlowItemAIQuery, k as MidsceneYamlFlowItemAIRightClick, aP as MidsceneYamlFlowItemAIScroll, aL as MidsceneYamlFlowItemAITap, aK as MidsceneYamlFlowItemAIWaitFor, aQ as MidsceneYamlFlowItemEvaluateJavaScript, aS as MidsceneYamlFlowItemLogScreenshot, aR as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aB as MidsceneYamlScriptAndroidEnv, aC as MidsceneYamlScriptEnv, az as MidsceneYamlScriptEnvBase, aA as MidsceneYamlScriptWebEnv, i as MidsceneYamlTask, O as OnTaskStartTip, av as PageType, P as PartialInsightDumpFromSDK, V as PlanningAIResponse, T as PlanningAction, $ as PlanningActionParamAssert, a1 as PlanningActionParamError, X as PlanningActionParamHover, Z as PlanningActionParamInputOrKeyPress, Y as PlanningActionParamRightClick, _ as PlanningActionParamScroll, a0 as PlanningActionParamSleep, W as PlanningActionParamTap, a2 as PlanningActionParamWaitFor, S as PlanningLocateParam, a6 as PlaywrightParserOpt, a5 as PuppeteerParserOpt, ax as ReferenceImage, R as ReportDumpWithAttributes, aV as ScriptPlayerStatusValue, aU as ScriptPlayerTaskStatus, aa as TaskCacheInfo, ay as scrollParam } from './types-00c9086f.js';
3
- import { c as callAiFn } from './llm-planning-453ffce5.js';
4
- export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-453ffce5.js';
1
+ import { E as ExecutionTask, a as ExecutionTaskProgressOptions, b as ExecutionTaskApply, c as ExecutionDump, U as UIContext, I as InsightAction, D as DumpSubscriber, d as InsightTaskInfo, e as InsightOptions, f as DetailedLocateParam, L as LocateResult, g as InsightExtractOption, h as InsightAssertionResponse, A as AIDescribeElementResponse } from './types-e7be1eb0.js';
2
+ export { v as AIAssertionResponse, t as AIDataExtractionResponse, r as AIElementCoordinatesResponse, q as AIElementLocatorResponse, s as AIElementResponse, m as AIResponseFormat, u as AISectionLocatorResponse, p as AISingleElementResponse, n as AISingleElementResponseById, o as AISingleElementResponseByPosition, l as AIUsageInfo, Q as AgentAssertOpt, y as AgentDescribeElementAtPointResult, N as AgentWaitForOpt, a4 as BaseAgentParserOpt, C as CallAIFn, a3 as Color, G as DumpMeta, K as ElementById, z as EnsureObject, a7 as ExecutionRecorderItem, ap as ExecutionTaskAction, ao as ExecutionTaskActionApply, an as ExecutionTaskInsightAssertion, am as ExecutionTaskInsightAssertionApply, al as ExecutionTaskInsightAssertionParam, ae as ExecutionTaskInsightDumpLog, ag as ExecutionTaskInsightLocate, af as ExecutionTaskInsightLocateApply, ad as ExecutionTaskInsightLocateOutput, ac as ExecutionTaskInsightLocateParam, ak as ExecutionTaskInsightQuery, aj as ExecutionTaskInsightQueryApply, ai as ExecutionTaskInsightQueryOutput, ah as ExecutionTaskInsightQueryParam, ar as ExecutionTaskLog, aq as ExecutionTaskLogApply, at as ExecutionTaskPlanning, as as ExecutionTaskPlanningApply, ab as ExecutionTaskReturn, a8 as ExecutionTaskType, a9 as ExecutorContext, aT as FreeFn, au as GroupedActionDump, H as InsightDump, B as InsightExtractParam, J as LiteUISection, aw as LocateOption, F as LocateResultElement, x as LocateValidatorResult, w as LocatorValidatorOption, j as MidsceneYamlFlowItem, aD as MidsceneYamlFlowItemAIAction, aE as MidsceneYamlFlowItemAIAssert, aI as MidsceneYamlFlowItemAIBoolean, aM as MidsceneYamlFlowItemAIHover, aN as MidsceneYamlFlowItemAIInput, aO as MidsceneYamlFlowItemAIKeyboardPress, aJ as MidsceneYamlFlowItemAILocate, aH as MidsceneYamlFlowItemAINString, aG as MidsceneYamlFlowItemAINumber, aF as MidsceneYamlFlowItemAIQuery, k as MidsceneYamlFlowItemAIRightClick, aP as MidsceneYamlFlowItemAIScroll, aL as MidsceneYamlFlowItemAITap, aK as MidsceneYamlFlowItemAIWaitFor, aQ as MidsceneYamlFlowItemEvaluateJavaScript, aS as MidsceneYamlFlowItemLogScreenshot, aR as MidsceneYamlFlowItemSleep, M as MidsceneYamlScript, aB as MidsceneYamlScriptAndroidEnv, aC as MidsceneYamlScriptEnv, az as MidsceneYamlScriptEnvBase, aA as MidsceneYamlScriptWebEnv, i as MidsceneYamlTask, O as OnTaskStartTip, av as PageType, P as PartialInsightDumpFromSDK, V as PlanningAIResponse, T as PlanningAction, $ as PlanningActionParamAssert, a1 as PlanningActionParamError, X as PlanningActionParamHover, Z as PlanningActionParamInputOrKeyPress, Y as PlanningActionParamRightClick, _ as PlanningActionParamScroll, a0 as PlanningActionParamSleep, W as PlanningActionParamTap, a2 as PlanningActionParamWaitFor, S as PlanningLocateParam, a6 as PlaywrightParserOpt, a5 as PuppeteerParserOpt, ax as ReferenceImage, R as ReportDumpWithAttributes, aV as ScriptPlayerStatusValue, aU as ScriptPlayerTaskStatus, aa as TaskCacheInfo, ay as scrollParam } from './types-e7be1eb0.js';
3
+ import { c as callAiFn } from './llm-planning-45954424.js';
4
+ export { a as AiAssert, A as AiLocateElement, d as describeUserPage, p as plan } from './llm-planning-45954424.js';
5
5
  import { BaseElement, Rect } from '@midscene/shared/types';
6
6
  export { BaseElement, ElementTreeNode, Point, Rect, Size } from '@midscene/shared/types';
7
7
  export { getVersion } from './utils.js';
package/dist/es/index.js CHANGED
@@ -1,6 +1,6 @@
1
1
  import {
2
2
  getVersion
3
- } from "./chunk-YGJNESNY.js";
3
+ } from "./chunk-TO33YH6H.js";
4
4
  import {
5
5
  AiAssert,
6
6
  AiExtractElementInfo,
@@ -11,7 +11,7 @@ import {
11
11
  describeUserPage,
12
12
  expandSearchArea,
13
13
  plan
14
- } from "./chunk-7MNTWX2A.js";
14
+ } from "./chunk-CLWM3F4J.js";
15
15
 
16
16
  // src/ai-model/action-executor.ts
17
17
  import {
@@ -121,7 +121,6 @@ ${this.latestErrorTask()?.errorStack}`
121
121
  task.status = "finished";
122
122
  task.timing.end = Date.now();
123
123
  task.timing.cost = task.timing.end - task.timing.start;
124
- task.timing.aiCost = returnValue?.aiCost || 0;
125
124
  taskIndex++;
126
125
  } catch (e) {
127
126
  successfullyCompleted = false;
@@ -213,17 +212,14 @@ import {
213
212
  MIDSCENE_FORCE_DEEP_THINK,
214
213
  MIDSCENE_USE_QWEN_VL,
215
214
  getAIConfigInBoolean,
216
- vlLocateMode as vlLocateMode3
215
+ vlLocateMode as vlLocateMode2
217
216
  } from "@midscene/shared/env";
218
217
  import { compositeElementInfoImg, cropByRect } from "@midscene/shared/img";
219
218
  import { getDebug } from "@midscene/shared/logger";
220
219
  import { assert as assert2 } from "@midscene/shared/utils";
221
220
 
222
221
  // src/insight/utils.ts
223
- import {
224
- MIDSCENE_MODEL_NAME as MIDSCENE_MODEL_NAME2,
225
- getAIConfig as getAIConfig2
226
- } from "@midscene/shared/env";
222
+ import { MIDSCENE_MODEL_NAME as MIDSCENE_MODEL_NAME2, getAIConfig as getAIConfig2 } from "@midscene/shared/env";
227
223
  import { uuid } from "@midscene/shared/utils";
228
224
  function emitInsightDump(data, dumpSubscriber) {
229
225
  const baseData = {
@@ -274,7 +270,7 @@ var Insight = class {
274
270
  if (query.deepThink || globalDeepThinkSwitch) {
275
271
  searchAreaPrompt = query.prompt;
276
272
  }
277
- if (searchAreaPrompt && !vlLocateMode3()) {
273
+ if (searchAreaPrompt && !vlLocateMode2()) {
278
274
  console.warn(
279
275
  'The "deepThink" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model'
280
276
  );
@@ -1 +1 @@
1
- {"version":3,"mappings":";;;;;;;;;;;;;;;;AAUA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,cAAc;AAEhB,IAAM,WAAN,MAAe;AAAA,EAUpB,YACE,MACA,SAGA;AACA,SAAK,SACH,SAAS,SAAS,QAAQ,MAAM,SAAS,IAAI,YAAY;AAC3D,SAAK,OAAO;AACZ,SAAK,SAAS,SAAS,SAAS,CAAC,GAAG;AAAA,MAAI,CAAC,SACvC,KAAK,kBAAkB,IAAI;AAAA,IAC7B;AACA,SAAK,cAAc,SAAS;AAAA,EAC9B;AAAA,EAEQ,kBAAkB,MAAyC;AACjE,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,MAAgE;AAC3E;AAAA,MACE,KAAK,WAAW;AAAA,MAChB;AAAA,QAAyD,KAAK,gBAAgB,GAAG,KAAK;AAAA,EAAK,KAAK,gBAAgB,GAAG,UAAU;AAAA,IAC/H;AACA,QAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,WAAK,MAAM,KAAK,GAAG,KAAK,IAAI,CAAC,SAAS,KAAK,kBAAkB,IAAI,CAAC,CAAC;AAAA,IACrE,OAAO;AACL,WAAK,MAAM,KAAK,KAAK,kBAAkB,IAAI,CAAC;AAAA,IAC9C;AACA,QAAI,KAAK,WAAW,WAAW;AAC7B,WAAK,SAAS;AAAA,IAChB;AAAA,EACF;AAAA,EAEA,MAAM,QAAsB;AAC1B,QAAI,KAAK,WAAW,UAAU,KAAK,MAAM,SAAS,GAAG;AACnD,cAAQ;AAAA,QACN;AAAA,MACF;AAAA,IACF;AAEA,WAAO,KAAK,WAAW,WAAW,6BAA6B;AAC/D,WAAO,KAAK,WAAW,aAAa,+BAA+B;AACnE,WAAO,KAAK,WAAW,SAAS,4BAA4B;AAE5D,UAAM,mBAAmB,KAAK,MAAM;AAAA,MAClC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,mBAAmB,GAAG;AAExB;AAAA,IACF;AAEA,SAAK,SAAS;AACd,QAAI,YAAY;AAChB,QAAI,wBAAwB;AAE5B,QAAI;AAEJ,WAAO,YAAY,KAAK,MAAM,QAAQ;AACpC,YAAM,OAAO,KAAK,MAAM,SAAS;AACjC;AAAA,QACE,KAAK,WAAW;AAAA,QAChB,2CAA2C,KAAK,MAAM;AAAA,MACxD;AACA,WAAK,SAAS;AAAA,QACZ,OAAO,KAAK,IAAI;AAAA,MAClB;AACA,UAAI;AACF,aAAK,SAAS;AACd,YAAI;AACF,cAAI,KAAK,aAAa;AACpB,kBAAM,KAAK,YAAY,IAAI;AAAA,UAC7B;AAAA,QACF,SAAS,GAAG;AACV,kBAAQ,MAAM,wBAAwB,CAAC;AAAA,QACzC;AACA;AAAA,UACE,CAAC,WAAW,UAAU,UAAU,EAAE,QAAQ,KAAK,IAAI,KAAK;AAAA,UACxD,0BAA0B,KAAK,IAAI;AAAA,QACrC;AAEA,cAAM,EAAE,UAAU,MAAM,IAAI;AAC5B,eAAO,UAAU,uCAAuC,KAAK,IAAI,EAAE;AAEnE,YAAI;AACJ,cAAM,kBAAmC;AAAA,UACvC;AAAA,UACA,SAAS,oBAAoB;AAAA,QAC/B;AAEA,YAAI,KAAK,SAAS,WAAW;AAC3B;AAAA,YACE,KAAK,YAAY,YACf,KAAK,YAAY,WACjB,KAAK,YAAY,YACjB,KAAK,YAAY,aACjB,KAAK,YAAY,YACjB,KAAK,YAAY;AAAA,YACnB,gCAAgC,KAAK,OAAO;AAAA,UAC9C;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AACxD,cAAI,KAAK,YAAY,UAAU;AAC7B,iCACE,aACC;AAAA,UACL;AAAA,QACF,WAAW,KAAK,SAAS,YAAY,KAAK,SAAS,YAAY;AAC7D,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D,OAAO;AACL,kBAAQ;AAAA,YACN,0BAA0B,KAAK,IAAI;AAAA,UACrC;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D;AAEA,eAAO,OAAO,MAAM,WAAW;AAC/B,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD,aAAK,OAAO,SAAU,aAAqB,UAAU;AACrD;AAAA,MACF,SAAS,GAAQ;AACf,gCAAwB;AACxB,aAAK,QACH,GAAG,YAAY,OAAO,MAAM,WAAW,IAAI;AAC7C,aAAK,aAAa,EAAE;AAEpB,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF;AAAA,IACF;AAGA,aAAS,IAAI,YAAY,GAAG,IAAI,KAAK,MAAM,QAAQ,KAAK;AACtD,WAAK,MAAM,CAAC,EAAE,SAAS;AAAA,IACzB;AAEA,QAAI,uBAAuB;AACzB,WAAK,SAAS;AAAA,IAChB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAEA,QAAI,KAAK,MAAM,QAAQ;AAErB,YAAM,cAAc,KAAK,IAAI,WAAW,KAAK,MAAM,SAAS,CAAC;AAC7D,aAAO,KAAK,MAAM,WAAW,EAAE;AAAA,IACjC;AAAA,EACF;AAAA,EAEA,iBAA0B;AACxB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA,EAEA,kBAAwC;AACtC,QAAI,KAAK,WAAW,SAAS;AAC3B,aAAO;AAAA,IACT;AACA,UAAM,iBAAiB,KAAK,MAAM;AAAA,MAChC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,kBAAkB,GAAG;AACvB,aAAO,KAAK,MAAM,cAAc;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EAEA,OAAsB;AACpB,QAAI,mBAAmB;AAEvB,QAAI,aAAa,GAAG;AAClB,YAAM,iBAAiB,mBAAmB;AAC1C,UAAI,gBAAgB;AAClB,2BAAmB,WAAW,cAAc;AAAA,MAC9C,OAAO;AACL,2BAAmB,GAAG,aAAa,CAAC;AAAA,MACtC;AAAA,IACF;AACA,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,mBAAmB;AAAA,MACnB,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC9NA,SAAS,4BAA4B;AAE9B,IAAM,8BAA8B,MAAM;AAC/C,SAAO,mNAAmN,qBAAqB,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAoBlP;;;ACQA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA,gBAAAA;AAAA,OACK;AACP,SAAS,yBAAyB,kBAAkB;AACpD,SAAS,gBAAgB;AACzB,SAAS,UAAAC,eAAc;;;AChCvB;AAAA,EACE,uBAAAC;AAAA,EACA,eAAAC;AAAA,OAGK;AACP,SAAS,YAAY;AAEd,SAAS,gBACd,MACA,gBACA;AACA,QAAM,WAAqB;AAAA,IACzB,YAAY,WAAW;AAAA,IACvB,SAAS,KAAK,IAAI;AAAA,IAClB,YAAYA,aAAYD,oBAAmB,KAAK;AAAA,EAClD;AACA,QAAM,YAAyB;AAAA,IAC7B,OAAO,KAAK;AAAA,IACZ,GAAG;AAAA,IACH,GAAG;AAAA,EACL;AAEA,mBAAiB,SAAS;AAC5B;;;ADoBA,IAAM,QAAQ,SAAS,YAAY;AACnC,IAAqB,UAArB,MAGE;AAAA,EAWA,YACE,SAGA,KACA;AAXF,sBAAoD;AAYlD,IAAAD,QAAO,SAAS,iCAAiC;AACjD,QAAI,OAAO,YAAY,YAAY;AACjC,WAAK,qBAAqB;AAAA,IAC5B,OAAO;AACL,WAAK,qBAAqB,MAAM,QAAQ,QAAQ,OAAO;AAAA,IACzD;AAEA,QAAI,OAAO,KAAK,eAAe,aAAa;AAC1C,WAAK,aAAa,IAAI;AAAA,IACxB;AACA,QAAI,OAAO,KAAK,aAAa,aAAa;AACxC,WAAK,WAAW,IAAI;AAAA,IACtB;AAAA,EACF;AAAA,EAEA,MAAM,OACJ,OACA,KACuB;AACvB,UAAM,EAAE,OAAO,IAAI,OAAO,CAAC;AAC3B,UAAM,cAAc,OAAO,UAAU,WAAW,QAAQ,MAAM;AAC9D,IAAAA,QAAO,aAAa,8BAA8B;AAClD,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,IAAAA,QAAO,OAAO,UAAU,UAAU,sCAAsC;AAExE,UAAM,wBAAwB;AAAA,MAC5B;AAAA,IACF;AACA,QAAI,uBAAuB;AACzB,YAAM,yBAAyB,qBAAqB;AAAA,IACtD;AACA,QAAI;AACJ,QAAI,MAAM,aAAa,uBAAuB;AAC5C,yBAAmB,MAAM;AAAA,IAC3B;AAEA,QAAI,oBAAoB,CAACD,cAAa,GAAG;AACvC,cAAQ;AAAA,QACN;AAAA,MACF;AACA,yBAAmB;AAAA,IACrB;AAEA,UAAM,UAAU,KAAK,WAAY,MAAM,KAAK,mBAAmB,QAAQ;AAEvE,QAAI,aAA+B;AACnC,QAAI,wBAA4C;AAChD,QAAI,kBAA2C;AAC/C,QAAI,qBAEY;AAChB,QAAI,kBAAkB;AACpB,2BAAqB,MAAM,gBAAgB;AAAA,QACzC;AAAA,QACA,oBAAoB;AAAA,MACtB,CAAC;AACD,MAAAC;AAAA,QACE,mBAAmB;AAAA,QACnB,gCAAgC,gBAAgB,IAC9C,mBAAmB,QAAQ,KAAK,mBAAmB,KAAK,KAAK,EAC/D;AAAA,MACF;AACA,8BAAwB,mBAAmB;AAC3C,wBAAkB,mBAAmB;AACrC,mBAAa,mBAAmB;AAAA,IAClC;AAEA,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,aAAa,aAAa,MAAM,IACzD,MAAM,gBAAgB;AAAA,MACpB,QAAQ,UAAU,KAAK;AAAA,MACvB;AAAA,MACA,0BAA0B;AAAA,MAC1B,cAAc;AAAA,IAChB,CAAC;AAEH,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,MACvC,gBAAgB,KAAK,UAAU,WAAW;AAAA,MAC1C;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAgC,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAC1E;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT,SAAS;AAAA,MACX;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,aAAa;AAAA,MACb,MAAM;AAAA,MACN;AAAA,MACA,WAAW,CAAC,CAAC;AAAA,MACb,OAAO;AAAA,IACT;AAEA,UAAM,WAA0B,CAAC;AACjC,KAAC,YAAY,YAAY,CAAC,GAAG,QAAQ,CAAC,SAAS;AAC7C,UAAI,QAAQ,MAAM;AAChB,cAAM,UAAU,YAAY,MAAM,EAAE;AAEpC,YAAI,CAAC,SAAS;AACZ,kBAAQ;AAAA,YACN,kCAAkC,KAAK,EAAE;AAAA,UAC3C;AACA;AAAA,QACF;AACA,iBAAS,KAAK,OAAO;AAAA,MACvB;AAAA,IACF,CAAC;AAED;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,IACF;AAEA,QAAI,UAAU;AACZ,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,IAAAA;AAAA,MACE,SAAS,UAAU;AAAA,MACnB,6CAA6C,SAAS,MAAM;AAAA,IAC9D;AAEA,QAAI,SAAS,WAAW,GAAG;AACzB,aAAO;AAAA,QACL,SAAS;AAAA,UACP,IAAI,SAAS,CAAC,EAAG;AAAA,UACjB,SAAS,SAAS,CAAC,EAAG;AAAA,UACtB,QAAQ,SAAS,CAAC,EAAG;AAAA,UACrB,MAAM,SAAS,CAAC,EAAG;AAAA,UACnB,QAAQ,SAAS,CAAC,EAAG,UAAU,CAAC;AAAA,UAChC,YAAY,SAAS,CAAC,EAAG;AAAA,QAC3B;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,WAAO;AAAA,MACL,SAAS;AAAA,MACT;AAAA,IACF;AAAA,EACF;AAAA,EAYA,MAAM,QACJ,YACA,KACc;AACd,IAAAA;AAAA,MACE,OAAO,eAAe,YAAY,OAAO,eAAe;AAAA,MACxD,kDAAkD,OAAO,UAAU;AAAA,IACrE;AACA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,SAAS;AAEvD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,IAAI,MAAM,qBAAwB;AAAA,MAC3D;AAAA,MACA,WAAW;AAAA,MACX,eAAe;AAAA,IACjB,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,IACzC;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAwB,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAClE;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,OAAO;AAAA,IACT;AAEA,UAAM,EAAE,KAAK,IAAI,eAAe,CAAC;AAGjC;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH;AAAA,MACF;AAAA,MACA;AAAA,IACF;AAEA,QAAI,YAAY,CAAC,MAAM;AACrB,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,WAAsD;AACjE,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,QAAQ;AACtD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,eAAe,MAAM,SAAS;AAAA,MAClC;AAAA,MACA;AAAA,IACF,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,aAAa,OAAO;AAAA,IAClD;AAEA,UAAM,EAAE,SAAS,KAAK,IAAI,aAAa;AACvC,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,eAAe;AAAA,MACf,kBAAkB;AAAA,MAClB,OAAO,OAAO,SAAY;AAAA,IAC5B;AACA,oBAAgB,UAAU,cAAc;AAExC,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA,OAAO,aAAa;AAAA,IACtB;AAAA,EACF;AAAA,EACA,MAAM,SACJ,QACA,KAGyD;AACzD,IAAAA,QAAO,QAAQ,yCAAyC;AACxD,UAAM,UAAU,MAAM,KAAK,mBAAmB,UAAU;AACxD,UAAM,EAAE,iBAAiB,IAAI;AAC7B,IAAAA,QAAO,kBAAkB,6CAA6C;AAEtE,UAAM,eAAe,4BAA4B;AAGjD,UAAM,kBAAkB;AACxB,UAAM,aAAmB,MAAM,QAAQ,MAAM,IACzC;AAAA,MACE,MAAM,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAChD,KAAK,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAC/C,OAAO;AAAA,MACP,QAAQ;AAAA,IACV,IACA;AAEJ,QAAI,eAAe,MAAM,wBAAwB;AAAA,MAC/C,gBAAgB;AAAA,MAChB,sBAAsB;AAAA,QACpB;AAAA,UACE,MAAM;AAAA,QACR;AAAA,MACF;AAAA,MACA,iBAAiB;AAAA,IACnB,CAAC;AAED,QAAI,KAAK,WAAW;AAClB,YAAM,aAAa,iBAAiB,YAAY,QAAQ,IAAI;AAC5D,YAAM,4BAA4B,UAAU;AAC5C,qBAAe,MAAM;AAAA,QACnB;AAAA,QACA;AAAA,QACA,qBAAqB,oBAAoB;AAAA,MAC3C;AAAA,IACF;AAEA,UAAM,OAAe;AAAA,MACnB,EAAE,MAAM,UAAU,SAAS,aAAa;AAAA,MACxC;AAAA,QACE,MAAM;AAAA,QACN,SAAS;AAAA,UACP;AAAA,YACE,MAAM;AAAA,YACN,WAAW;AAAA,cACT,KAAK;AAAA,cACL,QAAQ;AAAA,YACV;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,WACJ,KAAK,cAAc;AAErB,UAAM,MAAM,MAAM,SAAS,8BAAmC;AAE9D,UAAM,EAAE,QAAQ,IAAI;AACpB,IAAAA,QAAO,CAAC,QAAQ,OAAO,oBAAoB,QAAQ,KAAK,EAAE;AAC1D,IAAAA,QAAO,QAAQ,aAAa,gCAAgC;AAC5D,WAAO;AAAA,EACT;AACF;;;AEzZA,SAAS,eAAAE,cAAa,uBAAAD,4BAA2B;AAGjD,IAAO,cAAQ","names":["vlLocateMode","assert","MIDSCENE_MODEL_NAME","getAIConfig"],"ignoreList":[],"sources":["../../src/ai-model/action-executor.ts","../../src/ai-model/prompt/describe.ts","../../src/insight/index.ts","../../src/insight/utils.ts","../../src/index.ts"],"sourcesContent":["import type {\n ExecutionDump,\n ExecutionTask,\n ExecutionTaskApply,\n ExecutionTaskInsightLocateOutput,\n ExecutionTaskProgressOptions,\n ExecutionTaskReturn,\n ExecutorContext,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n MIDSCENE_MODEL_NAME,\n getAIConfig,\n uiTarsModelVersion,\n vlLocateMode,\n} from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\n\nexport class Executor {\n name: string;\n\n tasks: ExecutionTask[];\n\n // status of executor\n status: 'init' | 'pending' | 'running' | 'completed' | 'error';\n\n onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n\n constructor(\n name: string,\n options?: ExecutionTaskProgressOptions & {\n tasks?: ExecutionTaskApply[];\n },\n ) {\n this.status =\n options?.tasks && options.tasks.length > 0 ? 'pending' : 'init';\n this.name = name;\n this.tasks = (options?.tasks || []).map((item) =>\n this.markTaskAsPending(item),\n );\n this.onTaskStart = options?.onTaskStart;\n }\n\n private markTaskAsPending(task: ExecutionTaskApply): ExecutionTask {\n return {\n status: 'pending',\n ...task,\n };\n }\n\n async append(task: ExecutionTaskApply[] | ExecutionTaskApply): Promise<void> {\n assert(\n this.status !== 'error',\n `executor is in error state, cannot append task\\nerror=${this.latestErrorTask()?.error}\\n${this.latestErrorTask()?.errorStack}`,\n );\n if (Array.isArray(task)) {\n this.tasks.push(...task.map((item) => this.markTaskAsPending(item)));\n } else {\n this.tasks.push(this.markTaskAsPending(task));\n }\n if (this.status !== 'running') {\n this.status = 'pending';\n }\n }\n\n async flush(): Promise<any> {\n if (this.status === 'init' && this.tasks.length > 0) {\n console.warn(\n 'illegal state for executor, status is init but tasks are not empty',\n );\n }\n\n assert(this.status !== 'running', 'executor is already running');\n assert(this.status !== 'completed', 'executor is already completed');\n assert(this.status !== 'error', 'executor is in error state');\n\n const nextPendingIndex = this.tasks.findIndex(\n (task) => task.status === 'pending',\n );\n if (nextPendingIndex < 0) {\n // all tasks are completed\n return;\n }\n\n this.status = 'running';\n let taskIndex = nextPendingIndex;\n let successfullyCompleted = true;\n\n let previousFindOutput: ExecutionTaskInsightLocateOutput | undefined;\n\n while (taskIndex < this.tasks.length) {\n const task = this.tasks[taskIndex];\n assert(\n task.status === 'pending',\n `task status should be pending, but got: ${task.status}`,\n );\n task.timing = {\n start: Date.now(),\n };\n try {\n task.status = 'running';\n try {\n if (this.onTaskStart) {\n await this.onTaskStart(task);\n }\n } catch (e) {\n console.error('error in onTaskStart', e);\n }\n assert(\n ['Insight', 'Action', 'Planning'].indexOf(task.type) >= 0,\n `unsupported task type: ${task.type}`,\n );\n\n const { executor, param } = task;\n assert(executor, `executor is required for task type: ${task.type}`);\n\n let returnValue;\n const executorContext: ExecutorContext = {\n task,\n element: previousFindOutput?.element,\n };\n\n if (task.type === 'Insight') {\n assert(\n task.subType === 'Locate' ||\n task.subType === 'Query' ||\n task.subType === 'Assert' ||\n task.subType === 'Boolean' ||\n task.subType === 'Number' ||\n task.subType === 'String',\n `unsupported insight subType: ${task.subType}`,\n );\n returnValue = await task.executor(param, executorContext);\n if (task.subType === 'Locate') {\n previousFindOutput = (\n returnValue as ExecutionTaskReturn<ExecutionTaskInsightLocateOutput>\n )?.output;\n }\n } else if (task.type === 'Action' || task.type === 'Planning') {\n returnValue = await task.executor(param, executorContext);\n } else {\n console.warn(\n `unsupported task type: ${task.type}, will try to execute it directly`,\n );\n returnValue = await task.executor(param, executorContext);\n }\n\n Object.assign(task, returnValue);\n task.status = 'finished';\n task.timing.end = Date.now();\n task.timing.cost = task.timing.end - task.timing.start;\n task.timing.aiCost = (returnValue as any)?.aiCost || 0;\n taskIndex++;\n } catch (e: any) {\n successfullyCompleted = false;\n task.error =\n e?.message || (typeof e === 'string' ? e : 'error-without-message');\n task.errorStack = e.stack;\n\n task.status = 'failed';\n task.timing.end = Date.now();\n task.timing.cost = task.timing.end - task.timing.start;\n break;\n }\n }\n\n // set all remaining tasks as cancelled\n for (let i = taskIndex + 1; i < this.tasks.length; i++) {\n this.tasks[i].status = 'cancelled';\n }\n\n if (successfullyCompleted) {\n this.status = 'completed';\n } else {\n this.status = 'error';\n }\n\n if (this.tasks.length) {\n // return the last output\n const outputIndex = Math.min(taskIndex, this.tasks.length - 1);\n return this.tasks[outputIndex].output;\n }\n }\n\n isInErrorState(): boolean {\n return this.status === 'error';\n }\n\n latestErrorTask(): ExecutionTask | null {\n if (this.status !== 'error') {\n return null;\n }\n const errorTaskIndex = this.tasks.findIndex(\n (task) => task.status === 'failed',\n );\n if (errorTaskIndex >= 0) {\n return this.tasks[errorTaskIndex];\n }\n return null;\n }\n\n dump(): ExecutionDump {\n let modelDescription = '';\n\n if (vlLocateMode()) {\n const uiTarsModelVer = uiTarsModelVersion();\n if (uiTarsModelVer) {\n modelDescription = `UI-TARS=${uiTarsModelVer}`;\n } else {\n modelDescription = `${vlLocateMode()} mode`;\n }\n }\n const dumpData: ExecutionDump = {\n sdkVersion: getVersion(),\n model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n model_description: modelDescription,\n logTime: Date.now(),\n name: this.name,\n tasks: this.tasks,\n };\n return dumpData;\n }\n}\n","import { getPreferredLanguage } from '@midscene/shared/env';\n\nexport const elementDescriberInstruction = () => {\n return `Tell what is the content of the element wrapped by the read rectangle in the screenshot. Your description is expected to be used to precisely locate the element from other similar elements on screenshot. Use ${getPreferredLanguage()} in the description.\n\nPlease follow the following rules:\n1. The description should be start with a brief description, like \"a button for confirming the action\".\n\n2. Include these information in the description to distinguish the element from its siblings and other similar elements, as much as possible:\n- The text of the element, like \"with text 'Confirm'\"\n- What the element looks like if it's an image, like \"with image '...'\"\n- The relative position of the element, like \"on the left of ..., around ...\"\n- How to distinguish the element from its siblings elements, like \"it is the icon instead of the text\"\n\n3. Do NOT mention the red rectangle in the description.\n\n4. Use the error field to describe the unexpected situations, if any. If not, put null.\n\nReturn in JSON:\n{\n \"description\": \"[{brief description}]: {text of the element} {image of the element} {relative position of the element} ... \",\n \"error\"?: \"...\"\n}`;\n};\n","import {\n AIActionType,\n type AIArgs,\n callAiFn,\n expandSearchArea,\n} from '@/ai-model/common';\nimport {\n AiExtractElementInfo,\n AiLocateElement,\n callToGetJSONObject,\n} from '@/ai-model/index';\nimport { AiAssert, AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n AIDescribeElementResponse,\n AIElementResponse,\n AIUsageInfo,\n BaseElement,\n DetailedLocateParam,\n DumpSubscriber,\n InsightAction,\n InsightAssertionResponse,\n InsightExtractOption,\n InsightExtractParam,\n InsightOptions,\n InsightTaskInfo,\n LocateResult,\n PartialInsightDumpFromSDK,\n Rect,\n UIContext,\n} from '@/types';\nimport {\n MIDSCENE_FORCE_DEEP_THINK,\n MIDSCENE_USE_QWEN_VL,\n getAIConfigInBoolean,\n vlLocateMode,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n context?: UIContext<BaseElement>;\n callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n ElementType extends BaseElement = BaseElement,\n ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n contextRetrieverFn: (\n action: InsightAction,\n ) => Promise<ContextType> | ContextType;\n\n aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n onceDumpUpdatedFn?: DumpSubscriber;\n\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n constructor(\n context:\n | ContextType\n | ((action: InsightAction) => Promise<ContextType> | ContextType),\n opt?: InsightOptions,\n ) {\n assert(context, 'context is required for Insight');\n if (typeof context === 'function') {\n this.contextRetrieverFn = context;\n } else {\n this.contextRetrieverFn = () => Promise.resolve(context);\n }\n\n if (typeof opt?.aiVendorFn !== 'undefined') {\n this.aiVendorFn = opt.aiVendorFn;\n }\n if (typeof opt?.taskInfo !== 'undefined') {\n this.taskInfo = opt.taskInfo;\n }\n }\n\n async locate(\n query: DetailedLocateParam,\n opt?: LocateOpts,\n ): Promise<LocateResult> {\n const { callAI } = opt || {};\n const queryPrompt = typeof query === 'string' ? query : query.prompt;\n assert(queryPrompt, 'query is required for locate');\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n assert(typeof query === 'object', 'query should be an object for locate');\n\n const globalDeepThinkSwitch = getAIConfigInBoolean(\n MIDSCENE_FORCE_DEEP_THINK,\n );\n if (globalDeepThinkSwitch) {\n debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n }\n let searchAreaPrompt;\n if (query.deepThink || globalDeepThinkSwitch) {\n searchAreaPrompt = query.prompt;\n }\n\n if (searchAreaPrompt && !vlLocateMode()) {\n console.warn(\n 'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n );\n searchAreaPrompt = undefined;\n }\n\n const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n let searchArea: Rect | undefined = undefined;\n let searchAreaRawResponse: string | undefined = undefined;\n let searchAreaUsage: AIUsageInfo | undefined = undefined;\n let searchAreaResponse:\n | Awaited<ReturnType<typeof AiLocateSection>>\n | undefined = undefined;\n if (searchAreaPrompt) {\n searchAreaResponse = await AiLocateSection({\n context,\n sectionDescription: searchAreaPrompt,\n });\n assert(\n searchAreaResponse.rect,\n `cannot find search area for \"${searchAreaPrompt}\"${\n searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n }`,\n );\n searchAreaRawResponse = searchAreaResponse.rawResponse;\n searchAreaUsage = searchAreaResponse.usage;\n searchArea = searchAreaResponse.rect;\n }\n\n const startTime = Date.now();\n const { parseResult, rect, elementById, rawResponse, usage } =\n await AiLocateElement({\n callAI: callAI || this.aiVendorFn,\n context,\n targetElementDescription: queryPrompt,\n searchConfig: searchAreaResponse,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(rawResponse),\n formatResponse: JSON.stringify(parseResult),\n usage,\n searchArea,\n searchAreaRawResponse,\n searchAreaUsage,\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'locate',\n userQuery: {\n element: queryPrompt,\n },\n matchedElement: [],\n matchedRect: rect,\n data: null,\n taskInfo,\n deepThink: !!searchArea,\n error: errorLog,\n };\n\n const elements: BaseElement[] = [];\n (parseResult.elements || []).forEach((item) => {\n if ('id' in item) {\n const element = elementById(item?.id);\n\n if (!element) {\n console.warn(\n `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n );\n return;\n }\n elements.push(element);\n }\n });\n\n emitInsightDump(\n {\n ...dumpData,\n matchedElement: elements,\n },\n dumpSubscriber,\n );\n\n if (errorLog) {\n throw new Error(errorLog);\n }\n\n assert(\n elements.length <= 1,\n `locate: multiple elements found, length = ${elements.length}`,\n );\n\n if (elements.length === 1) {\n return {\n element: {\n id: elements[0]!.id,\n indexId: elements[0]!.indexId,\n center: elements[0]!.center,\n rect: elements[0]!.rect,\n xpaths: elements[0]!.xpaths || [],\n attributes: elements[0]!.attributes,\n },\n rect,\n };\n }\n return {\n element: null,\n rect,\n };\n }\n\n async extract<T = any>(input: string, opt?: InsightExtractOption): Promise<T>;\n async extract<T extends Record<string, string>>(\n input: T,\n opt?: InsightExtractOption,\n ): Promise<Record<keyof T, any>>;\n async extract<T extends object>(\n input: Record<keyof T, string>,\n opt?: InsightExtractOption,\n ): Promise<T>;\n\n async extract<T>(\n dataDemand: InsightExtractParam,\n opt?: InsightExtractOption,\n ): Promise<any> {\n assert(\n typeof dataDemand === 'object' || typeof dataDemand === 'string',\n `dataDemand should be object or string, but get ${typeof dataDemand}`,\n );\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n const context = await this.contextRetrieverFn('extract');\n\n const startTime = Date.now();\n const { parseResult, usage } = await AiExtractElementInfo<T>({\n context,\n dataQuery: dataDemand,\n extractOption: opt,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(parseResult),\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'extract',\n userQuery: {\n dataDemand,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n error: errorLog,\n };\n\n const { data } = parseResult || {};\n\n // 4\n emitInsightDump(\n {\n ...dumpData,\n data,\n },\n dumpSubscriber,\n );\n\n if (errorLog && !data) {\n throw new Error(errorLog);\n }\n\n return {\n data,\n usage,\n };\n }\n\n async assert(assertion: string): Promise<InsightAssertionResponse> {\n if (typeof assertion !== 'string') {\n throw new Error(\n 'This is the assert method for Midscene, the first argument should be a string. If you want to use the assert method from Node.js, please import it from the Node.js assert module.',\n );\n }\n\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n const context = await this.contextRetrieverFn('assert');\n const startTime = Date.now();\n const assertResult = await AiAssert({\n assertion,\n context,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(assertResult.content),\n };\n\n const { thought, pass } = assertResult.content;\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'assert',\n userQuery: {\n assertion,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n assertionPass: pass,\n assertionThought: thought,\n error: pass ? undefined : thought,\n };\n emitInsightDump(dumpData, dumpSubscriber);\n\n return {\n pass,\n thought,\n usage: assertResult.usage,\n };\n }\n async describe(\n target: Rect | [number, number],\n opt?: {\n deepThink?: boolean;\n },\n ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n assert(target, 'target is required for insight.describe');\n const context = await this.contextRetrieverFn('describe');\n const { screenshotBase64 } = context;\n assert(screenshotBase64, 'screenshot is required for insight.describe');\n\n const systemPrompt = elementDescriberInstruction();\n\n // Convert [x,y] center point to Rect if needed\n const defaultRectSize = 30;\n const targetRect: Rect = Array.isArray(target)\n ? {\n left: Math.floor(target[0] - defaultRectSize / 2),\n top: Math.floor(target[1] - defaultRectSize / 2),\n width: defaultRectSize,\n height: defaultRectSize,\n }\n : target;\n\n let imagePayload = await compositeElementInfoImg({\n inputImgBase64: screenshotBase64,\n elementsPositionInfo: [\n {\n rect: targetRect,\n },\n ],\n borderThickness: 3,\n });\n\n if (opt?.deepThink) {\n const searchArea = expandSearchArea(targetRect, context.size);\n debug('describe: set searchArea', searchArea);\n imagePayload = await cropByRect(\n imagePayload,\n searchArea,\n getAIConfigInBoolean(MIDSCENE_USE_QWEN_VL),\n );\n }\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n detail: 'high',\n },\n },\n ],\n },\n ];\n\n const callAIFn =\n this.aiVendorFn || callToGetJSONObject<AIDescribeElementResponse>;\n\n const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT);\n\n const { content } = res;\n assert(!content.error, `describe failed: ${content.error}`);\n assert(content.description, 'failed to describe the element');\n return content;\n }\n}\n","import type {\n DumpMeta,\n DumpSubscriber,\n InsightDump,\n PartialInsightDumpFromSDK,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n MIDSCENE_MODEL_NAME,\n getAIConfig,\n uiTarsModelVersion,\n vlLocateMode,\n} from '@midscene/shared/env';\nimport { uuid } from '@midscene/shared/utils';\n\nexport function emitInsightDump(\n data: PartialInsightDumpFromSDK,\n dumpSubscriber?: DumpSubscriber,\n) {\n const baseData: DumpMeta = {\n sdkVersion: getVersion(),\n logTime: Date.now(),\n model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n };\n const finalData: InsightDump = {\n logId: uuid(),\n ...baseData,\n ...data,\n };\n\n dumpSubscriber?.(finalData);\n}\n","import { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n plan,\n describeUserPage,\n AiLocateElement,\n AiAssert,\n} from './ai-model/index';\n\nexport { getAIConfig, MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n MidsceneYamlScript,\n MidsceneYamlTask,\n MidsceneYamlFlowItem,\n MidsceneYamlFlowItemAIRightClick,\n} from './yaml';\n"]}
1
+ {"version":3,"mappings":";;;;;;;;;;;;;;;;AAUA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,cAAc;AAEhB,IAAM,WAAN,MAAe;AAAA,EAUpB,YACE,MACA,SAGA;AACA,SAAK,SACH,SAAS,SAAS,QAAQ,MAAM,SAAS,IAAI,YAAY;AAC3D,SAAK,OAAO;AACZ,SAAK,SAAS,SAAS,SAAS,CAAC,GAAG;AAAA,MAAI,CAAC,SACvC,KAAK,kBAAkB,IAAI;AAAA,IAC7B;AACA,SAAK,cAAc,SAAS;AAAA,EAC9B;AAAA,EAEQ,kBAAkB,MAAyC;AACjE,WAAO;AAAA,MACL,QAAQ;AAAA,MACR,GAAG;AAAA,IACL;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,MAAgE;AAC3E;AAAA,MACE,KAAK,WAAW;AAAA,MAChB;AAAA,QAAyD,KAAK,gBAAgB,GAAG,KAAK;AAAA,EAAK,KAAK,gBAAgB,GAAG,UAAU;AAAA,IAC/H;AACA,QAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,WAAK,MAAM,KAAK,GAAG,KAAK,IAAI,CAAC,SAAS,KAAK,kBAAkB,IAAI,CAAC,CAAC;AAAA,IACrE,OAAO;AACL,WAAK,MAAM,KAAK,KAAK,kBAAkB,IAAI,CAAC;AAAA,IAC9C;AACA,QAAI,KAAK,WAAW,WAAW;AAC7B,WAAK,SAAS;AAAA,IAChB;AAAA,EACF;AAAA,EAEA,MAAM,QAAsB;AAC1B,QAAI,KAAK,WAAW,UAAU,KAAK,MAAM,SAAS,GAAG;AACnD,cAAQ;AAAA,QACN;AAAA,MACF;AAAA,IACF;AAEA,WAAO,KAAK,WAAW,WAAW,6BAA6B;AAC/D,WAAO,KAAK,WAAW,aAAa,+BAA+B;AACnE,WAAO,KAAK,WAAW,SAAS,4BAA4B;AAE5D,UAAM,mBAAmB,KAAK,MAAM;AAAA,MAClC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,mBAAmB,GAAG;AAExB;AAAA,IACF;AAEA,SAAK,SAAS;AACd,QAAI,YAAY;AAChB,QAAI,wBAAwB;AAE5B,QAAI;AAEJ,WAAO,YAAY,KAAK,MAAM,QAAQ;AACpC,YAAM,OAAO,KAAK,MAAM,SAAS;AACjC;AAAA,QACE,KAAK,WAAW;AAAA,QAChB,2CAA2C,KAAK,MAAM;AAAA,MACxD;AACA,WAAK,SAAS;AAAA,QACZ,OAAO,KAAK,IAAI;AAAA,MAClB;AACA,UAAI;AACF,aAAK,SAAS;AACd,YAAI;AACF,cAAI,KAAK,aAAa;AACpB,kBAAM,KAAK,YAAY,IAAI;AAAA,UAC7B;AAAA,QACF,SAAS,GAAG;AACV,kBAAQ,MAAM,wBAAwB,CAAC;AAAA,QACzC;AACA;AAAA,UACE,CAAC,WAAW,UAAU,UAAU,EAAE,QAAQ,KAAK,IAAI,KAAK;AAAA,UACxD,0BAA0B,KAAK,IAAI;AAAA,QACrC;AAEA,cAAM,EAAE,UAAU,MAAM,IAAI;AAC5B,eAAO,UAAU,uCAAuC,KAAK,IAAI,EAAE;AAEnE,YAAI;AACJ,cAAM,kBAAmC;AAAA,UACvC;AAAA,UACA,SAAS,oBAAoB;AAAA,QAC/B;AAEA,YAAI,KAAK,SAAS,WAAW;AAC3B;AAAA,YACE,KAAK,YAAY,YACf,KAAK,YAAY,WACjB,KAAK,YAAY,YACjB,KAAK,YAAY,aACjB,KAAK,YAAY,YACjB,KAAK,YAAY;AAAA,YACnB,gCAAgC,KAAK,OAAO;AAAA,UAC9C;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AACxD,cAAI,KAAK,YAAY,UAAU;AAC7B,iCACE,aACC;AAAA,UACL;AAAA,QACF,WAAW,KAAK,SAAS,YAAY,KAAK,SAAS,YAAY;AAC7D,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D,OAAO;AACL,kBAAQ;AAAA,YACN,0BAA0B,KAAK,IAAI;AAAA,UACrC;AACA,wBAAc,MAAM,KAAK,SAAS,OAAO,eAAe;AAAA,QAC1D;AAEA,eAAO,OAAO,MAAM,WAAW;AAC/B,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF,SAAS,GAAQ;AACf,gCAAwB;AACxB,aAAK,QACH,GAAG,YAAY,OAAO,MAAM,WAAW,IAAI;AAC7C,aAAK,aAAa,EAAE;AAEpB,aAAK,SAAS;AACd,aAAK,OAAO,MAAM,KAAK,IAAI;AAC3B,aAAK,OAAO,OAAO,KAAK,OAAO,MAAM,KAAK,OAAO;AACjD;AAAA,MACF;AAAA,IACF;AAGA,aAAS,IAAI,YAAY,GAAG,IAAI,KAAK,MAAM,QAAQ,KAAK;AACtD,WAAK,MAAM,CAAC,EAAE,SAAS;AAAA,IACzB;AAEA,QAAI,uBAAuB;AACzB,WAAK,SAAS;AAAA,IAChB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAEA,QAAI,KAAK,MAAM,QAAQ;AAErB,YAAM,cAAc,KAAK,IAAI,WAAW,KAAK,MAAM,SAAS,CAAC;AAC7D,aAAO,KAAK,MAAM,WAAW,EAAE;AAAA,IACjC;AAAA,EACF;AAAA,EAEA,iBAA0B;AACxB,WAAO,KAAK,WAAW;AAAA,EACzB;AAAA,EAEA,kBAAwC;AACtC,QAAI,KAAK,WAAW,SAAS;AAC3B,aAAO;AAAA,IACT;AACA,UAAM,iBAAiB,KAAK,MAAM;AAAA,MAChC,CAAC,SAAS,KAAK,WAAW;AAAA,IAC5B;AACA,QAAI,kBAAkB,GAAG;AACvB,aAAO,KAAK,MAAM,cAAc;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EAEA,OAAsB;AACpB,QAAI,mBAAmB;AAEvB,QAAI,aAAa,GAAG;AAClB,YAAM,iBAAiB,mBAAmB;AAC1C,UAAI,gBAAgB;AAClB,2BAAmB,WAAW,cAAc;AAAA,MAC9C,OAAO;AACL,2BAAmB,GAAG,aAAa,CAAC;AAAA,MACtC;AAAA,IACF;AACA,UAAM,WAA0B;AAAA,MAC9B,YAAY,WAAW;AAAA,MACvB,YAAY,YAAY,mBAAmB,KAAK;AAAA,MAChD,mBAAmB;AAAA,MACnB,SAAS,KAAK,IAAI;AAAA,MAClB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACF;;;AC7NA,SAAS,4BAA4B;AAE9B,IAAM,8BAA8B,MAAM;AAC/C,SAAO,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,SAAS,uBAAAC,sBAAqB,eAAAC,oBAAmB;AACjD,SAAS,YAAY;AAEd,SAAS,gBACd,MACA,gBACA;AACA,QAAM,WAAqB;AAAA,IACzB,YAAY,WAAW;AAAA,IACvB,SAAS,KAAK,IAAI;AAAA,IAClB,YAAYA,aAAYD,oBAAmB,KAAK;AAAA,EAClD;AACA,QAAM,YAAyB;AAAA,IAC7B,OAAO,KAAK;AAAA,IACZ,GAAG;AAAA,IACH,GAAG;AAAA,EACL;AAEA,mBAAiB,SAAS;AAC5B;;;ADyBA,IAAM,QAAQ,SAAS,YAAY;AACnC,IAAqB,UAArB,MAGE;AAAA,EAWA,YACE,SAGA,KACA;AAXF,sBAAoD;AAYlD,IAAAD,QAAO,SAAS,iCAAiC;AACjD,QAAI,OAAO,YAAY,YAAY;AACjC,WAAK,qBAAqB;AAAA,IAC5B,OAAO;AACL,WAAK,qBAAqB,MAAM,QAAQ,QAAQ,OAAO;AAAA,IACzD;AAEA,QAAI,OAAO,KAAK,eAAe,aAAa;AAC1C,WAAK,aAAa,IAAI;AAAA,IACxB;AACA,QAAI,OAAO,KAAK,aAAa,aAAa;AACxC,WAAK,WAAW,IAAI;AAAA,IACtB;AAAA,EACF;AAAA,EAEA,MAAM,OACJ,OACA,KACuB;AACvB,UAAM,EAAE,OAAO,IAAI,OAAO,CAAC;AAC3B,UAAM,cAAc,OAAO,UAAU,WAAW,QAAQ,MAAM;AAC9D,IAAAA,QAAO,aAAa,8BAA8B;AAClD,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,IAAAA,QAAO,OAAO,UAAU,UAAU,sCAAsC;AAExE,UAAM,wBAAwB;AAAA,MAC5B;AAAA,IACF;AACA,QAAI,uBAAuB;AACzB,YAAM,yBAAyB,qBAAqB;AAAA,IACtD;AACA,QAAI;AACJ,QAAI,MAAM,aAAa,uBAAuB;AAC5C,yBAAmB,MAAM;AAAA,IAC3B;AAEA,QAAI,oBAAoB,CAACD,cAAa,GAAG;AACvC,cAAQ;AAAA,QACN;AAAA,MACF;AACA,yBAAmB;AAAA,IACrB;AAEA,UAAM,UAAU,KAAK,WAAY,MAAM,KAAK,mBAAmB,QAAQ;AAEvE,QAAI,aAA+B;AACnC,QAAI,wBAA4C;AAChD,QAAI,kBAA2C;AAC/C,QAAI,qBAEY;AAChB,QAAI,kBAAkB;AACpB,2BAAqB,MAAM,gBAAgB;AAAA,QACzC;AAAA,QACA,oBAAoB;AAAA,MACtB,CAAC;AACD,MAAAC;AAAA,QACE,mBAAmB;AAAA,QACnB,gCAAgC,gBAAgB,IAC9C,mBAAmB,QAAQ,KAAK,mBAAmB,KAAK,KAAK,EAC/D;AAAA,MACF;AACA,8BAAwB,mBAAmB;AAC3C,wBAAkB,mBAAmB;AACrC,mBAAa,mBAAmB;AAAA,IAClC;AAEA,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,aAAa,aAAa,MAAM,IACzD,MAAM,gBAAgB;AAAA,MACpB,QAAQ,UAAU,KAAK;AAAA,MACvB;AAAA,MACA,0BAA0B;AAAA,MAC1B,cAAc;AAAA,IAChB,CAAC;AAEH,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,MACvC,gBAAgB,KAAK,UAAU,WAAW;AAAA,MAC1C;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAgC,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAC1E;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT,SAAS;AAAA,MACX;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,aAAa;AAAA,MACb,MAAM;AAAA,MACN;AAAA,MACA,WAAW,CAAC,CAAC;AAAA,MACb,OAAO;AAAA,IACT;AAEA,UAAM,WAA0B,CAAC;AACjC,KAAC,YAAY,YAAY,CAAC,GAAG,QAAQ,CAAC,SAAS;AAC7C,UAAI,QAAQ,MAAM;AAChB,cAAM,UAAU,YAAY,MAAM,EAAE;AAEpC,YAAI,CAAC,SAAS;AACZ,kBAAQ;AAAA,YACN,kCAAkC,KAAK,EAAE;AAAA,UAC3C;AACA;AAAA,QACF;AACA,iBAAS,KAAK,OAAO;AAAA,MACvB;AAAA,IACF,CAAC;AAED;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,IACF;AAEA,QAAI,UAAU;AACZ,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,IAAAA;AAAA,MACE,SAAS,UAAU;AAAA,MACnB,6CAA6C,SAAS,MAAM;AAAA,IAC9D;AAEA,QAAI,SAAS,WAAW,GAAG;AACzB,aAAO;AAAA,QACL,SAAS;AAAA,UACP,IAAI,SAAS,CAAC,EAAG;AAAA,UACjB,SAAS,SAAS,CAAC,EAAG;AAAA,UACtB,QAAQ,SAAS,CAAC,EAAG;AAAA,UACrB,MAAM,SAAS,CAAC,EAAG;AAAA,UACnB,QAAQ,SAAS,CAAC,EAAG,UAAU,CAAC;AAAA,UAChC,YAAY,SAAS,CAAC,EAAG;AAAA,QAC3B;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,WAAO;AAAA,MACL,SAAS;AAAA,MACT;AAAA,IACF;AAAA,EACF;AAAA,EAYA,MAAM,QACJ,YACA,KACc;AACd,IAAAA;AAAA,MACE,OAAO,eAAe,YAAY,OAAO,eAAe;AAAA,MACxD,kDAAkD,OAAO,UAAU;AAAA,IACrE;AACA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,SAAS;AAEvD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,EAAE,aAAa,MAAM,IAAI,MAAM,qBAAwB;AAAA,MAC3D;AAAA,MACA,WAAW;AAAA,MACX,eAAe;AAAA,IACjB,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,WAAW;AAAA,IACzC;AAEA,QAAI;AACJ,QAAI,YAAY,QAAQ,QAAQ;AAC9B,iBAAW;AAAA,EAAwB,YAAY,OAAO,KAAK,IAAI,CAAC;AAAA,IAClE;AAEA,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,OAAO;AAAA,IACT;AAEA,UAAM,EAAE,KAAK,IAAI,eAAe,CAAC;AAGjC;AAAA,MACE;AAAA,QACE,GAAG;AAAA,QACH;AAAA,MACF;AAAA,MACA;AAAA,IACF;AAEA,QAAI,YAAY,CAAC,MAAM;AACrB,YAAM,IAAI,MAAM,QAAQ;AAAA,IAC1B;AAEA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,OAAO,WAAsD;AACjE,QAAI,OAAO,cAAc,UAAU;AACjC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,UAAM,iBAAiB,KAAK;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,UAAU,MAAM,KAAK,mBAAmB,QAAQ;AACtD,UAAM,YAAY,KAAK,IAAI;AAC3B,UAAM,eAAe,MAAM,SAAS;AAAA,MAClC;AAAA,MACA;AAAA,IACF,CAAC;AAED,UAAM,WAAW,KAAK,IAAI,IAAI;AAC9B,UAAM,WAA4B;AAAA,MAChC,GAAI,KAAK,WAAW,KAAK,WAAW,CAAC;AAAA,MACrC,YAAY;AAAA,MACZ,aAAa,KAAK,UAAU,aAAa,OAAO;AAAA,IAClD;AAEA,UAAM,EAAE,SAAS,KAAK,IAAI,aAAa;AACvC,UAAM,WAAsC;AAAA,MAC1C,MAAM;AAAA,MACN,WAAW;AAAA,QACT;AAAA,MACF;AAAA,MACA,gBAAgB,CAAC;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA,eAAe;AAAA,MACf,kBAAkB;AAAA,MAClB,OAAO,OAAO,SAAY;AAAA,IAC5B;AACA,oBAAgB,UAAU,cAAc;AAExC,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA,OAAO,aAAa;AAAA,IACtB;AAAA,EACF;AAAA,EACA,MAAM,SACJ,QACA,KAGyD;AACzD,IAAAA,QAAO,QAAQ,yCAAyC;AACxD,UAAM,UAAU,MAAM,KAAK,mBAAmB,UAAU;AACxD,UAAM,EAAE,iBAAiB,IAAI;AAC7B,IAAAA,QAAO,kBAAkB,6CAA6C;AAEtE,UAAM,eAAe,4BAA4B;AAGjD,UAAM,kBAAkB;AACxB,UAAM,aAAmB,MAAM,QAAQ,MAAM,IACzC;AAAA,MACE,MAAM,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAChD,KAAK,KAAK,MAAM,OAAO,CAAC,IAAI,kBAAkB,CAAC;AAAA,MAC/C,OAAO;AAAA,MACP,QAAQ;AAAA,IACV,IACA;AAEJ,QAAI,eAAe,MAAM,wBAAwB;AAAA,MAC/C,gBAAgB;AAAA,MAChB,sBAAsB;AAAA,QACpB;AAAA,UACE,MAAM;AAAA,QACR;AAAA,MACF;AAAA,MACA,iBAAiB;AAAA,IACnB,CAAC;AAED,QAAI,KAAK,WAAW;AAClB,YAAM,aAAa,iBAAiB,YAAY,QAAQ,IAAI;AAC5D,YAAM,4BAA4B,UAAU;AAC5C,qBAAe,MAAM;AAAA,QACnB;AAAA,QACA;AAAA,QACA,qBAAqB,oBAAoB;AAAA,MAC3C;AAAA,IACF;AAEA,UAAM,OAAe;AAAA,MACnB,EAAE,MAAM,UAAU,SAAS,aAAa;AAAA,MACxC;AAAA,QACE,MAAM;AAAA,QACN,SAAS;AAAA,UACP;AAAA,YACE,MAAM;AAAA,YACN,WAAW;AAAA,cACT,KAAK;AAAA,cACL,QAAQ;AAAA,YACV;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,WACJ,KAAK,cAAc;AAErB,UAAM,MAAM,MAAM,SAAS,8BAAmC;AAE9D,UAAM,EAAE,QAAQ,IAAI;AACpB,IAAAA,QAAO,CAAC,QAAQ,OAAO,oBAAoB,QAAQ,KAAK,EAAE;AAC1D,IAAAA,QAAO,QAAQ,aAAa,gCAAgC;AAC5D,WAAO;AAAA,EACT;AACF;;;AEzZA,SAAS,eAAAE,cAAa,uBAAAD,4BAA2B;AAGjD,IAAO,cAAQ","names":["vlLocateMode","assert","MIDSCENE_MODEL_NAME","getAIConfig"],"ignoreList":[],"sources":["../../src/ai-model/action-executor.ts","../../src/ai-model/prompt/describe.ts","../../src/insight/index.ts","../../src/insight/utils.ts","../../src/index.ts"],"sourcesContent":["import type {\n ExecutionDump,\n ExecutionTask,\n ExecutionTaskApply,\n ExecutionTaskInsightLocateOutput,\n ExecutionTaskProgressOptions,\n ExecutionTaskReturn,\n ExecutorContext,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport {\n MIDSCENE_MODEL_NAME,\n getAIConfig,\n uiTarsModelVersion,\n vlLocateMode,\n} from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\n\nexport class Executor {\n name: string;\n\n tasks: ExecutionTask[];\n\n // status of executor\n status: 'init' | 'pending' | 'running' | 'completed' | 'error';\n\n onTaskStart?: ExecutionTaskProgressOptions['onTaskStart'];\n\n constructor(\n name: string,\n options?: ExecutionTaskProgressOptions & {\n tasks?: ExecutionTaskApply[];\n },\n ) {\n this.status =\n options?.tasks && options.tasks.length > 0 ? 'pending' : 'init';\n this.name = name;\n this.tasks = (options?.tasks || []).map((item) =>\n this.markTaskAsPending(item),\n );\n this.onTaskStart = options?.onTaskStart;\n }\n\n private markTaskAsPending(task: ExecutionTaskApply): ExecutionTask {\n return {\n status: 'pending',\n ...task,\n };\n }\n\n async append(task: ExecutionTaskApply[] | ExecutionTaskApply): Promise<void> {\n assert(\n this.status !== 'error',\n `executor is in error state, cannot append task\\nerror=${this.latestErrorTask()?.error}\\n${this.latestErrorTask()?.errorStack}`,\n );\n if (Array.isArray(task)) {\n this.tasks.push(...task.map((item) => this.markTaskAsPending(item)));\n } else {\n this.tasks.push(this.markTaskAsPending(task));\n }\n if (this.status !== 'running') {\n this.status = 'pending';\n }\n }\n\n async flush(): Promise<any> {\n if (this.status === 'init' && this.tasks.length > 0) {\n console.warn(\n 'illegal state for executor, status is init but tasks are not empty',\n );\n }\n\n assert(this.status !== 'running', 'executor is already running');\n assert(this.status !== 'completed', 'executor is already completed');\n assert(this.status !== 'error', 'executor is in error state');\n\n const nextPendingIndex = this.tasks.findIndex(\n (task) => task.status === 'pending',\n );\n if (nextPendingIndex < 0) {\n // all tasks are completed\n return;\n }\n\n this.status = 'running';\n let taskIndex = nextPendingIndex;\n let successfullyCompleted = true;\n\n let previousFindOutput: ExecutionTaskInsightLocateOutput | undefined;\n\n while (taskIndex < this.tasks.length) {\n const task = this.tasks[taskIndex];\n assert(\n task.status === 'pending',\n `task status should be pending, but got: ${task.status}`,\n );\n task.timing = {\n start: Date.now(),\n };\n try {\n task.status = 'running';\n try {\n if (this.onTaskStart) {\n await this.onTaskStart(task);\n }\n } catch (e) {\n console.error('error in onTaskStart', e);\n }\n assert(\n ['Insight', 'Action', 'Planning'].indexOf(task.type) >= 0,\n `unsupported task type: ${task.type}`,\n );\n\n const { executor, param } = task;\n assert(executor, `executor is required for task type: ${task.type}`);\n\n let returnValue;\n const executorContext: ExecutorContext = {\n task,\n element: previousFindOutput?.element,\n };\n\n if (task.type === 'Insight') {\n assert(\n task.subType === 'Locate' ||\n task.subType === 'Query' ||\n task.subType === 'Assert' ||\n task.subType === 'Boolean' ||\n task.subType === 'Number' ||\n task.subType === 'String',\n `unsupported insight subType: ${task.subType}`,\n );\n returnValue = await task.executor(param, executorContext);\n if (task.subType === 'Locate') {\n previousFindOutput = (\n returnValue as ExecutionTaskReturn<ExecutionTaskInsightLocateOutput>\n )?.output;\n }\n } else if (task.type === 'Action' || task.type === 'Planning') {\n returnValue = await task.executor(param, executorContext);\n } else {\n console.warn(\n `unsupported task type: ${task.type}, will try to execute it directly`,\n );\n returnValue = await task.executor(param, executorContext);\n }\n\n Object.assign(task, returnValue);\n task.status = 'finished';\n task.timing.end = Date.now();\n task.timing.cost = task.timing.end - task.timing.start;\n taskIndex++;\n } catch (e: any) {\n successfullyCompleted = false;\n task.error =\n e?.message || (typeof e === 'string' ? e : 'error-without-message');\n task.errorStack = e.stack;\n\n task.status = 'failed';\n task.timing.end = Date.now();\n task.timing.cost = task.timing.end - task.timing.start;\n break;\n }\n }\n\n // set all remaining tasks as cancelled\n for (let i = taskIndex + 1; i < this.tasks.length; i++) {\n this.tasks[i].status = 'cancelled';\n }\n\n if (successfullyCompleted) {\n this.status = 'completed';\n } else {\n this.status = 'error';\n }\n\n if (this.tasks.length) {\n // return the last output\n const outputIndex = Math.min(taskIndex, this.tasks.length - 1);\n return this.tasks[outputIndex].output;\n }\n }\n\n isInErrorState(): boolean {\n return this.status === 'error';\n }\n\n latestErrorTask(): ExecutionTask | null {\n if (this.status !== 'error') {\n return null;\n }\n const errorTaskIndex = this.tasks.findIndex(\n (task) => task.status === 'failed',\n );\n if (errorTaskIndex >= 0) {\n return this.tasks[errorTaskIndex];\n }\n return null;\n }\n\n dump(): ExecutionDump {\n let modelDescription = '';\n\n if (vlLocateMode()) {\n const uiTarsModelVer = uiTarsModelVersion();\n if (uiTarsModelVer) {\n modelDescription = `UI-TARS=${uiTarsModelVer}`;\n } else {\n modelDescription = `${vlLocateMode()} mode`;\n }\n }\n const dumpData: ExecutionDump = {\n sdkVersion: getVersion(),\n model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n model_description: modelDescription,\n logTime: Date.now(),\n name: this.name,\n tasks: this.tasks,\n };\n return dumpData;\n }\n}\n","import { getPreferredLanguage } from '@midscene/shared/env';\n\nexport const elementDescriberInstruction = () => {\n return `Tell what is the content of the element wrapped by the read rectangle in the screenshot. Your description is expected to be used to precisely locate the element from other similar elements on screenshot. Use ${getPreferredLanguage()} in the description.\n\nPlease follow the following rules:\n1. The description should be start with a brief description, like \"a button for confirming the action\".\n\n2. Include these information in the description to distinguish the element from its siblings and other similar elements, as much as possible:\n- The text of the element, like \"with text 'Confirm'\"\n- What the element looks like if it's an image, like \"with image '...'\"\n- The relative position of the element, like \"on the left of ..., around ...\"\n- How to distinguish the element from its siblings elements, like \"it is the icon instead of the text\"\n\n3. Do NOT mention the red rectangle in the description.\n\n4. Use the error field to describe the unexpected situations, if any. If not, put null.\n\nReturn in JSON:\n{\n \"description\": \"[{brief description}]: {text of the element} {image of the element} {relative position of the element} ... \",\n \"error\"?: \"...\"\n}`;\n};\n","import {\n AIActionType,\n type AIArgs,\n callAiFn,\n expandSearchArea,\n} from '@/ai-model/common';\nimport {\n AiExtractElementInfo,\n AiLocateElement,\n callToGetJSONObject,\n} from '@/ai-model/index';\nimport { AiAssert, AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n AIDescribeElementResponse,\n AIElementResponse,\n AIUsageInfo,\n BaseElement,\n DetailedLocateParam,\n DumpSubscriber,\n InsightAction,\n InsightAssertionResponse,\n InsightExtractOption,\n InsightExtractParam,\n InsightOptions,\n InsightTaskInfo,\n LocateResult,\n PartialInsightDumpFromSDK,\n Rect,\n UIContext,\n} from '@/types';\nimport {\n MIDSCENE_FORCE_DEEP_THINK,\n MIDSCENE_USE_QWEN_VL,\n getAIConfigInBoolean,\n vlLocateMode,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n context?: UIContext<BaseElement>;\n callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n ElementType extends BaseElement = BaseElement,\n ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n contextRetrieverFn: (\n action: InsightAction,\n ) => Promise<ContextType> | ContextType;\n\n aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n onceDumpUpdatedFn?: DumpSubscriber;\n\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n constructor(\n context:\n | ContextType\n | ((action: InsightAction) => Promise<ContextType> | ContextType),\n opt?: InsightOptions,\n ) {\n assert(context, 'context is required for Insight');\n if (typeof context === 'function') {\n this.contextRetrieverFn = context;\n } else {\n this.contextRetrieverFn = () => Promise.resolve(context);\n }\n\n if (typeof opt?.aiVendorFn !== 'undefined') {\n this.aiVendorFn = opt.aiVendorFn;\n }\n if (typeof opt?.taskInfo !== 'undefined') {\n this.taskInfo = opt.taskInfo;\n }\n }\n\n async locate(\n query: DetailedLocateParam,\n opt?: LocateOpts,\n ): Promise<LocateResult> {\n const { callAI } = opt || {};\n const queryPrompt = typeof query === 'string' ? query : query.prompt;\n assert(queryPrompt, 'query is required for locate');\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n assert(typeof query === 'object', 'query should be an object for locate');\n\n const globalDeepThinkSwitch = getAIConfigInBoolean(\n MIDSCENE_FORCE_DEEP_THINK,\n );\n if (globalDeepThinkSwitch) {\n debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n }\n let searchAreaPrompt;\n if (query.deepThink || globalDeepThinkSwitch) {\n searchAreaPrompt = query.prompt;\n }\n\n if (searchAreaPrompt && !vlLocateMode()) {\n console.warn(\n 'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n );\n searchAreaPrompt = undefined;\n }\n\n const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n let searchArea: Rect | undefined = undefined;\n let searchAreaRawResponse: string | undefined = undefined;\n let searchAreaUsage: AIUsageInfo | undefined = undefined;\n let searchAreaResponse:\n | Awaited<ReturnType<typeof AiLocateSection>>\n | undefined = undefined;\n if (searchAreaPrompt) {\n searchAreaResponse = await AiLocateSection({\n context,\n sectionDescription: searchAreaPrompt,\n });\n assert(\n searchAreaResponse.rect,\n `cannot find search area for \"${searchAreaPrompt}\"${\n searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n }`,\n );\n searchAreaRawResponse = searchAreaResponse.rawResponse;\n searchAreaUsage = searchAreaResponse.usage;\n searchArea = searchAreaResponse.rect;\n }\n\n const startTime = Date.now();\n const { parseResult, rect, elementById, rawResponse, usage } =\n await AiLocateElement({\n callAI: callAI || this.aiVendorFn,\n context,\n targetElementDescription: queryPrompt,\n searchConfig: searchAreaResponse,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(rawResponse),\n formatResponse: JSON.stringify(parseResult),\n usage,\n searchArea,\n searchAreaRawResponse,\n searchAreaUsage,\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'locate',\n userQuery: {\n element: queryPrompt,\n },\n matchedElement: [],\n matchedRect: rect,\n data: null,\n taskInfo,\n deepThink: !!searchArea,\n error: errorLog,\n };\n\n const elements: BaseElement[] = [];\n (parseResult.elements || []).forEach((item) => {\n if ('id' in item) {\n const element = elementById(item?.id);\n\n if (!element) {\n console.warn(\n `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n );\n return;\n }\n elements.push(element);\n }\n });\n\n emitInsightDump(\n {\n ...dumpData,\n matchedElement: elements,\n },\n dumpSubscriber,\n );\n\n if (errorLog) {\n throw new Error(errorLog);\n }\n\n assert(\n elements.length <= 1,\n `locate: multiple elements found, length = ${elements.length}`,\n );\n\n if (elements.length === 1) {\n return {\n element: {\n id: elements[0]!.id,\n indexId: elements[0]!.indexId,\n center: elements[0]!.center,\n rect: elements[0]!.rect,\n xpaths: elements[0]!.xpaths || [],\n attributes: elements[0]!.attributes,\n },\n rect,\n };\n }\n return {\n element: null,\n rect,\n };\n }\n\n async extract<T = any>(input: string, opt?: InsightExtractOption): Promise<T>;\n async extract<T extends Record<string, string>>(\n input: T,\n opt?: InsightExtractOption,\n ): Promise<Record<keyof T, any>>;\n async extract<T extends object>(\n input: Record<keyof T, string>,\n opt?: InsightExtractOption,\n ): Promise<T>;\n\n async extract<T>(\n dataDemand: InsightExtractParam,\n opt?: InsightExtractOption,\n ): Promise<any> {\n assert(\n typeof dataDemand === 'object' || typeof dataDemand === 'string',\n `dataDemand should be object or string, but get ${typeof dataDemand}`,\n );\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n const context = await this.contextRetrieverFn('extract');\n\n const startTime = Date.now();\n const { parseResult, usage } = await AiExtractElementInfo<T>({\n context,\n dataQuery: dataDemand,\n extractOption: opt,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(parseResult),\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'extract',\n userQuery: {\n dataDemand,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n error: errorLog,\n };\n\n const { data } = parseResult || {};\n\n // 4\n emitInsightDump(\n {\n ...dumpData,\n data,\n },\n dumpSubscriber,\n );\n\n if (errorLog && !data) {\n throw new Error(errorLog);\n }\n\n return {\n data,\n usage,\n };\n }\n\n async assert(assertion: string): Promise<InsightAssertionResponse> {\n if (typeof assertion !== 'string') {\n throw new Error(\n 'This is the assert method for Midscene, the first argument should be a string. If you want to use the assert method from Node.js, please import it from the Node.js assert module.',\n );\n }\n\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n const context = await this.contextRetrieverFn('assert');\n const startTime = Date.now();\n const assertResult = await AiAssert({\n assertion,\n context,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(assertResult.content),\n };\n\n const { thought, pass } = assertResult.content;\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'assert',\n userQuery: {\n assertion,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n assertionPass: pass,\n assertionThought: thought,\n error: pass ? undefined : thought,\n };\n emitInsightDump(dumpData, dumpSubscriber);\n\n return {\n pass,\n thought,\n usage: assertResult.usage,\n };\n }\n async describe(\n target: Rect | [number, number],\n opt?: {\n deepThink?: boolean;\n },\n ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n assert(target, 'target is required for insight.describe');\n const context = await this.contextRetrieverFn('describe');\n const { screenshotBase64 } = context;\n assert(screenshotBase64, 'screenshot is required for insight.describe');\n\n const systemPrompt = elementDescriberInstruction();\n\n // Convert [x,y] center point to Rect if needed\n const defaultRectSize = 30;\n const targetRect: Rect = Array.isArray(target)\n ? {\n left: Math.floor(target[0] - defaultRectSize / 2),\n top: Math.floor(target[1] - defaultRectSize / 2),\n width: defaultRectSize,\n height: defaultRectSize,\n }\n : target;\n\n let imagePayload = await compositeElementInfoImg({\n inputImgBase64: screenshotBase64,\n elementsPositionInfo: [\n {\n rect: targetRect,\n },\n ],\n borderThickness: 3,\n });\n\n if (opt?.deepThink) {\n const searchArea = expandSearchArea(targetRect, context.size);\n debug('describe: set searchArea', searchArea);\n imagePayload = await cropByRect(\n imagePayload,\n searchArea,\n getAIConfigInBoolean(MIDSCENE_USE_QWEN_VL),\n );\n }\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n detail: 'high',\n },\n },\n ],\n },\n ];\n\n const callAIFn =\n this.aiVendorFn || callToGetJSONObject<AIDescribeElementResponse>;\n\n const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT);\n\n const { content } = res;\n assert(!content.error, `describe failed: ${content.error}`);\n assert(content.description, 'failed to describe the element');\n return content;\n }\n}\n","import type {\n DumpMeta,\n DumpSubscriber,\n InsightDump,\n PartialInsightDumpFromSDK,\n} from '@/types';\nimport { getVersion } from '@/utils';\nimport { MIDSCENE_MODEL_NAME, getAIConfig } from '@midscene/shared/env';\nimport { uuid } from '@midscene/shared/utils';\n\nexport function emitInsightDump(\n data: PartialInsightDumpFromSDK,\n dumpSubscriber?: DumpSubscriber,\n) {\n const baseData: DumpMeta = {\n sdkVersion: getVersion(),\n logTime: Date.now(),\n model_name: getAIConfig(MIDSCENE_MODEL_NAME) || '',\n };\n const finalData: InsightDump = {\n logId: uuid(),\n ...baseData,\n ...data,\n };\n\n dumpSubscriber?.(finalData);\n}\n","import { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n plan,\n describeUserPage,\n AiLocateElement,\n AiAssert,\n} from './ai-model/index';\n\nexport { getAIConfig, MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n MidsceneYamlScript,\n MidsceneYamlTask,\n MidsceneYamlFlowItem,\n MidsceneYamlFlowItemAIRightClick,\n} from './yaml';\n"]}
@@ -1,4 +1,4 @@
1
- import { l as AIUsageInfo, U as UIContext, ax as ReferenceImage, q as AIElementLocatorResponse, K as ElementById, g as InsightExtractOption, t as AIDataExtractionResponse, v as AIAssertionResponse, av as PageType, V as PlanningAIResponse } from './types-00c9086f.js';
1
+ import { l as AIUsageInfo, U as UIContext, ax as ReferenceImage, q as AIElementLocatorResponse, K as ElementById, g as InsightExtractOption, t as AIDataExtractionResponse, v as AIAssertionResponse, av as PageType, V as PlanningAIResponse } from './types-e7be1eb0.js';
2
2
  import { Rect, ElementTreeNode, BaseElement } from '@midscene/shared/types';
3
3
  import { ChatCompletionSystemMessageParam, ChatCompletionUserMessageParam } from 'openai/resources';
4
4
 
@@ -37,6 +37,7 @@ interface MidsceneYamlTask {
37
37
  }
38
38
  interface MidsceneYamlScriptEnvBase {
39
39
  output?: string;
40
+ unstableLogContent?: boolean | string;
40
41
  aiActionContext?: string;
41
42
  }
42
43
  interface MidsceneYamlScriptWebEnv extends MidsceneYamlScriptEnvBase {
@@ -141,9 +142,10 @@ interface ScriptPlayerTaskStatus extends MidsceneYamlTask {
141
142
  type ScriptPlayerStatusValue = 'init' | 'running' | 'done' | 'error';
142
143
 
143
144
  type AIUsageInfo = Record<string, any> & {
144
- prompt_tokens: number;
145
- completion_tokens: number;
146
- total_tokens: number;
145
+ prompt_tokens: number | undefined;
146
+ completion_tokens: number | undefined;
147
+ total_tokens: number | undefined;
148
+ time_cost: number | undefined;
147
149
  };
148
150
  /**
149
151
  * openai
@@ -413,7 +415,6 @@ type ExecutionTask<E extends ExecutionTaskApply<any, any, any> = ExecutionTaskAp
413
415
  start: number;
414
416
  end?: number;
415
417
  cost?: number;
416
- aiCost?: number;
417
418
  };
418
419
  usage?: AIUsageInfo;
419
420
  };
@@ -1,4 +1,4 @@
1
- import { R as ReportDumpWithAttributes } from './types-00c9086f.js';
1
+ import { R as ReportDumpWithAttributes } from './types-e7be1eb0.js';
2
2
  import { Rect } from '@midscene/shared/types';
3
3
  import '@midscene/shared/constants';
4
4
  import 'openai/resources';
package/dist/es/utils.js CHANGED
@@ -12,7 +12,7 @@ import {
12
12
  uploadTestInfoToServer,
13
13
  writeDumpReport,
14
14
  writeLogFile
15
- } from "./chunk-YGJNESNY.js";
15
+ } from "./chunk-TO33YH6H.js";
16
16
  export {
17
17
  getTmpDir,
18
18
  getTmpFile,
@@ -1,8 +1,8 @@
1
- import { l as AIUsageInfo, T as PlanningAction, j as MidsceneYamlFlowItem } from './types-00c9086f.js';
1
+ import { l as AIUsageInfo, T as PlanningAction, j as MidsceneYamlFlowItem } from './types-e7be1eb0.js';
2
2
  import { ChatCompletionMessageParam } from 'openai/resources';
3
3
  export { ChatCompletionMessageParam } from 'openai/resources';
4
- import { b as AIActionType } from './llm-planning-453ffce5.js';
5
- export { a as AiAssert, f as AiExtractElementInfo, A as AiLocateElement, g as AiLocateSection, h as adaptBboxToRect, c as callAiFn, d as describeUserPage, e as elementByPositionWithElementInfo, p as plan } from './llm-planning-453ffce5.js';
4
+ import { b as AIActionType } from './llm-planning-45954424.js';
5
+ export { a as AiAssert, f as AiExtractElementInfo, A as AiLocateElement, g as AiLocateSection, h as adaptBboxToRect, c as callAiFn, d as describeUserPage, e as elementByPositionWithElementInfo, p as plan } from './llm-planning-45954424.js';
6
6
  import { vlLocateMode } from '@midscene/shared/env';
7
7
  import { actionParser } from '@ui-tars/action-parser';
8
8
  import { Size } from '@midscene/shared/types';
@@ -12,7 +12,7 @@
12
12
 
13
13
 
14
14
 
15
- var _chunk7MNTWX2Ajs = require('./chunk-7MNTWX2A.js');
15
+ var _chunkCLWM3F4Jjs = require('./chunk-CLWM3F4J.js');
16
16
 
17
17
 
18
18
 
@@ -27,4 +27,4 @@ var _chunk7MNTWX2Ajs = require('./chunk-7MNTWX2A.js');
27
27
 
28
28
 
29
29
 
30
- exports.AiAssert = _chunk7MNTWX2Ajs.AiAssert; exports.AiExtractElementInfo = _chunk7MNTWX2Ajs.AiExtractElementInfo; exports.AiLocateElement = _chunk7MNTWX2Ajs.AiLocateElement; exports.AiLocateSection = _chunk7MNTWX2Ajs.AiLocateSection; exports.adaptBboxToRect = _chunk7MNTWX2Ajs.adaptBboxToRect; exports.callAiFn = _chunk7MNTWX2Ajs.callAiFn; exports.callToGetJSONObject = _chunk7MNTWX2Ajs.callToGetJSONObject; exports.describeUserPage = _chunk7MNTWX2Ajs.describeUserPage; exports.elementByPositionWithElementInfo = _chunk7MNTWX2Ajs.elementByPositionWithElementInfo; exports.plan = _chunk7MNTWX2Ajs.plan; exports.resizeImageForUiTars = _chunk7MNTWX2Ajs.resizeImageForUiTars; exports.systemPromptToLocateElement = _chunk7MNTWX2Ajs.systemPromptToLocateElement; exports.vlmPlanning = _chunk7MNTWX2Ajs.vlmPlanning;
30
+ exports.AiAssert = _chunkCLWM3F4Jjs.AiAssert; exports.AiExtractElementInfo = _chunkCLWM3F4Jjs.AiExtractElementInfo; exports.AiLocateElement = _chunkCLWM3F4Jjs.AiLocateElement; exports.AiLocateSection = _chunkCLWM3F4Jjs.AiLocateSection; exports.adaptBboxToRect = _chunkCLWM3F4Jjs.adaptBboxToRect; exports.callAiFn = _chunkCLWM3F4Jjs.callAiFn; exports.callToGetJSONObject = _chunkCLWM3F4Jjs.callToGetJSONObject; exports.describeUserPage = _chunkCLWM3F4Jjs.describeUserPage; exports.elementByPositionWithElementInfo = _chunkCLWM3F4Jjs.elementByPositionWithElementInfo; exports.plan = _chunkCLWM3F4Jjs.plan; exports.resizeImageForUiTars = _chunkCLWM3F4Jjs.resizeImageForUiTars; exports.systemPromptToLocateElement = _chunkCLWM3F4Jjs.systemPromptToLocateElement; exports.vlmPlanning = _chunkCLWM3F4Jjs.vlmPlanning;
@@ -1288,6 +1288,7 @@ async function call(messages, AIActionTypeValue, responseFormat) {
1288
1288
  const model = getModelName();
1289
1289
  let content;
1290
1290
  let usage;
1291
+ let timeCost;
1291
1292
  const commonConfig = {
1292
1293
  temperature: _env.vlLocateMode.call(void 0, ) === "vlm-ui-tars" ? 0 : 0.1,
1293
1294
  stream: false,
@@ -1300,12 +1301,14 @@ async function call(messages, AIActionTypeValue, responseFormat) {
1300
1301
  debugCall(`sending request to ${model}`);
1301
1302
  let result;
1302
1303
  try {
1304
+ const startTime2 = Date.now();
1303
1305
  result = await completion.create({
1304
1306
  model,
1305
1307
  messages,
1306
1308
  response_format: responseFormat,
1307
1309
  ...commonConfig
1308
1310
  });
1311
+ timeCost = Date.now() - startTime2;
1309
1312
  } catch (e) {
1310
1313
  const newError = new Error(
1311
1314
  `failed to call AI model service: ${e.message}. Trouble shooting: https://midscenejs.com/model-provider.html`,
@@ -1343,6 +1346,7 @@ async function call(messages, AIActionTypeValue, responseFormat) {
1343
1346
  }
1344
1347
  return content2;
1345
1348
  };
1349
+ const startTime2 = Date.now();
1346
1350
  const result = await completion.create({
1347
1351
  model,
1348
1352
  system: "You are a versatile professional in software UI automation",
@@ -1353,11 +1357,20 @@ async function call(messages, AIActionTypeValue, responseFormat) {
1353
1357
  response_format: responseFormat,
1354
1358
  ...commonConfig
1355
1359
  });
1360
+ timeCost = Date.now() - startTime2;
1356
1361
  content = result.content[0].text;
1357
1362
  _utils.assert.call(void 0, content, "empty content");
1358
1363
  usage = result.usage;
1359
1364
  }
1360
- return { content: content || "", usage };
1365
+ return {
1366
+ content: content || "",
1367
+ usage: {
1368
+ prompt_tokens: _nullishCoalesce(_optionalChain([usage, 'optionalAccess', _26 => _26.prompt_tokens]), () => ( 0)),
1369
+ completion_tokens: _nullishCoalesce(_optionalChain([usage, 'optionalAccess', _27 => _27.completion_tokens]), () => ( 0)),
1370
+ total_tokens: _nullishCoalesce(_optionalChain([usage, 'optionalAccess', _28 => _28.total_tokens]), () => ( 0)),
1371
+ time_cost: _nullishCoalesce(timeCost, () => ( 0))
1372
+ }
1373
+ };
1361
1374
  }
1362
1375
  async function callToGetJSONObject(messages, AIActionTypeValue) {
1363
1376
  let responseFormat;
@@ -1417,8 +1430,8 @@ function preprocessDoubaoBboxJson(input) {
1417
1430
  }
1418
1431
  function safeParseJson(input) {
1419
1432
  const cleanJsonString = extractJSONFromCodeBlock(input);
1420
- if (_optionalChain([cleanJsonString, 'optionalAccess', _26 => _26.match, 'call', _27 => _27(/\((\d+),(\d+)\)/)])) {
1421
- return _optionalChain([cleanJsonString, 'access', _28 => _28.match, 'call', _29 => _29(/\((\d+),(\d+)\)/), 'optionalAccess', _30 => _30.slice, 'call', _31 => _31(1), 'access', _32 => _32.map, 'call', _33 => _33(Number)]);
1433
+ if (_optionalChain([cleanJsonString, 'optionalAccess', _29 => _29.match, 'call', _30 => _30(/\((\d+),(\d+)\)/)])) {
1434
+ return _optionalChain([cleanJsonString, 'access', _31 => _31.match, 'call', _32 => _32(/\((\d+),(\d+)\)/), 'optionalAccess', _33 => _33.slice, 'call', _34 => _34(1), 'access', _35 => _35.map, 'call', _36 => _36(Number)]);
1422
1435
  }
1423
1436
  try {
1424
1437
  return JSON.parse(cleanJsonString);
@@ -1630,7 +1643,7 @@ async function AiLocateElement(options) {
1630
1643
  );
1631
1644
  }
1632
1645
  let referenceImagePayload;
1633
- if (_optionalChain([options, 'access', _34 => _34.referenceImage, 'optionalAccess', _35 => _35.rect]) && options.referenceImage.base64) {
1646
+ if (_optionalChain([options, 'access', _37 => _37.referenceImage, 'optionalAccess', _38 => _38.rect]) && options.referenceImage.base64) {
1634
1647
  referenceImagePayload = await _img.cropByRect.call(void 0,
1635
1648
  options.referenceImage.base64,
1636
1649
  options.referenceImage.rect,
@@ -1666,10 +1679,10 @@ async function AiLocateElement(options) {
1666
1679
  if ("bbox" in res.content && Array.isArray(res.content.bbox)) {
1667
1680
  resRect = adaptBboxToRect(
1668
1681
  res.content.bbox,
1669
- _optionalChain([options, 'access', _36 => _36.searchConfig, 'optionalAccess', _37 => _37.rect, 'optionalAccess', _38 => _38.width]) || context.size.width,
1670
- _optionalChain([options, 'access', _39 => _39.searchConfig, 'optionalAccess', _40 => _40.rect, 'optionalAccess', _41 => _41.height]) || context.size.height,
1671
- _optionalChain([options, 'access', _42 => _42.searchConfig, 'optionalAccess', _43 => _43.rect, 'optionalAccess', _44 => _44.left]),
1672
- _optionalChain([options, 'access', _45 => _45.searchConfig, 'optionalAccess', _46 => _46.rect, 'optionalAccess', _47 => _47.top])
1682
+ _optionalChain([options, 'access', _39 => _39.searchConfig, 'optionalAccess', _40 => _40.rect, 'optionalAccess', _41 => _41.width]) || context.size.width,
1683
+ _optionalChain([options, 'access', _42 => _42.searchConfig, 'optionalAccess', _43 => _43.rect, 'optionalAccess', _44 => _44.height]) || context.size.height,
1684
+ _optionalChain([options, 'access', _45 => _45.searchConfig, 'optionalAccess', _46 => _46.rect, 'optionalAccess', _47 => _47.left]),
1685
+ _optionalChain([options, 'access', _48 => _48.searchConfig, 'optionalAccess', _49 => _49.rect, 'optionalAccess', _50 => _50.top])
1673
1686
  );
1674
1687
  debugInspect("resRect", resRect);
1675
1688
  const rectCenter = {
@@ -1688,7 +1701,7 @@ async function AiLocateElement(options) {
1688
1701
  }
1689
1702
  } catch (e) {
1690
1703
  const msg = e instanceof Error ? `Failed to parse bbox: ${e.message}` : "unknown error in locate";
1691
- if (!errors || _optionalChain([errors, 'optionalAccess', _48 => _48.length]) === 0) {
1704
+ if (!errors || _optionalChain([errors, 'optionalAccess', _51 => _51.length]) === 0) {
1692
1705
  errors = [msg];
1693
1706
  } else {
1694
1707
  errors.push(`(${msg})`);
@@ -1779,14 +1792,14 @@ async function AiExtractElementInfo(options) {
1779
1792
  truncateTextLength: 200,
1780
1793
  filterNonTextContent: false,
1781
1794
  visibleOnly: false,
1782
- domIncluded: _optionalChain([extractOption, 'optionalAccess', _49 => _49.domIncluded])
1795
+ domIncluded: _optionalChain([extractOption, 'optionalAccess', _52 => _52.domIncluded])
1783
1796
  });
1784
1797
  const extractDataPromptText = await extractDataQueryPrompt(
1785
1798
  description,
1786
1799
  dataQuery
1787
1800
  );
1788
1801
  const userContent = [];
1789
- if (_optionalChain([extractOption, 'optionalAccess', _50 => _50.screenshotIncluded]) !== false) {
1802
+ if (_optionalChain([extractOption, 'optionalAccess', _53 => _53.screenshotIncluded]) !== false) {
1790
1803
  userContent.push({
1791
1804
  type: "image_url",
1792
1805
  image_url: {
@@ -1914,7 +1927,7 @@ async function plan(userInstruction, opts) {
1914
1927
  const { content, usage } = await call2(msgs, 3 /* PLAN */);
1915
1928
  const rawResponse = JSON.stringify(content, void 0, 2);
1916
1929
  const planFromAI = content;
1917
- const actions = (_optionalChain([planFromAI, 'access', _51 => _51.action, 'optionalAccess', _52 => _52.type]) ? [planFromAI.action] : planFromAI.actions) || [];
1930
+ const actions = (_optionalChain([planFromAI, 'access', _54 => _54.action, 'optionalAccess', _55 => _55.type]) ? [planFromAI.action] : planFromAI.actions) || [];
1918
1931
  const returnValue = {
1919
1932
  ...planFromAI,
1920
1933
  actions,
@@ -1941,7 +1954,7 @@ async function plan(userInstruction, opts) {
1941
1954
  _utils.assert.call(void 0, !planFromAI.error, `Failed to plan actions: ${planFromAI.error}`);
1942
1955
  } else {
1943
1956
  actions.forEach((action) => {
1944
- if (_optionalChain([action, 'access', _53 => _53.locate, 'optionalAccess', _54 => _54.id])) {
1957
+ if (_optionalChain([action, 'access', _56 => _56.locate, 'optionalAccess', _57 => _57.id])) {
1945
1958
  const element = elementById(action.locate.id);
1946
1959
  if (element) {
1947
1960
  action.locate.id = element.id;
@@ -2229,4 +2242,4 @@ async function resizeImageForUiTars(imageBase64, size) {
2229
2242
 
2230
2243
  exports.systemPromptToLocateElement = systemPromptToLocateElement; exports.elementByPositionWithElementInfo = elementByPositionWithElementInfo; exports.describeUserPage = describeUserPage; exports.callToGetJSONObject = callToGetJSONObject; exports.callAiFn = callAiFn; exports.adaptBboxToRect = adaptBboxToRect; exports.expandSearchArea = expandSearchArea; exports.AiLocateElement = AiLocateElement; exports.AiLocateSection = AiLocateSection; exports.AiExtractElementInfo = AiExtractElementInfo; exports.AiAssert = AiAssert; exports.plan = plan; exports.vlmPlanning = vlmPlanning; exports.resizeImageForUiTars = resizeImageForUiTars;
2231
2244
 
2232
- //# sourceMappingURL=chunk-7MNTWX2A.js.map
2245
+ //# sourceMappingURL=chunk-CLWM3F4J.js.map