@midscene/core 0.26.7-beta-20250818035341.0 → 0.26.7-beta-20250820105545.0

package/dist/es/ai-model/prompt/util.mjs

@@ -65,7 +65,7 @@ Description of all the elements in screenshot:
 </div>
 ====================
 `;
-async function describeUserPage(context, opt) {
+async function describeUserPage(context, modelPreferences, opt) {
     const { screenshotBase64 } = context;
     let width;
     let height;
@@ -84,7 +84,7 @@ async function describeUserPage(context, opt) {
     });
     let pageDescription = '';
     const visibleOnly = (null == opt ? void 0 : opt.visibleOnly) ?? (null == opt ? void 0 : opt.domIncluded) === 'visible-only';
-    if ((null == opt ? void 0 : opt.domIncluded) || !vlLocateMode()) {
+    if ((null == opt ? void 0 : opt.domIncluded) || !vlLocateMode(modelPreferences)) {
         const contentTree = await descriptionOfTree(treeRoot, null == opt ? void 0 : opt.truncateTextLength, null == opt ? void 0 : opt.filterNonTextContent, visibleOnly);
         const sizeDescription = describeSize({
             width,

package/dist/es/ai-model/prompt/util.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/prompt/util.mjs","sources":["webpack://@midscene/core/./src/ai-model/prompt/util.ts"],"sourcesContent":["import { imageInfoOfBase64 } from '@/image/index';\nimport type { BaseElement, ElementTreeNode, Size, UIContext } from '@/types';\nimport { NodeType } from '@midscene/shared/constants';\nimport { vlLocateMode } from '@midscene/shared/env';\nimport {\n  descriptionOfTree,\n  generateElementByPosition,\n  treeToList,\n} from '@midscene/shared/extractor';\nimport { assert } from '@midscene/shared/utils';\n\nexport function describeSize(size: Size) {\n  return `${size.width} x ${size.height}`;\n}\n\nexport function describeElement(\n  elements: (Pick<BaseElement, 'rect' | 'content'> & { id: string })[],\n) {\n  const sliceLength = 80;\n  return elements\n    .map((item) =>\n      [\n        item.id,\n        item.rect.left,\n        item.rect.top,\n        item.rect.left + item.rect.width,\n        item.rect.top + item.rect.height,\n        item.content.length > sliceLength\n          ? `${item.content.slice(0, sliceLength)}...`\n          : item.content,\n      ].join(', '),\n    )\n    .join('\\n');\n}\nexport const distanceThreshold = 16;\n\nexport function elementByPositionWithElementInfo(\n  treeRoot: ElementTreeNode<BaseElement>,\n  position: {\n    x: number;\n    y: number;\n  },\n  options?: {\n    requireStrictDistance?: boolean;\n    filterPositionElements?: boolean;\n  },\n) {\n  const requireStrictDistance = options?.requireStrictDistance ?? true;\n  const filterPositionElements = options?.filterPositionElements ?? false;\n\n  assert(typeof position !== 'undefined', 'position is required for query');\n\n  const matchingElements: BaseElement[] = [];\n\n  function dfs(node: ElementTreeNode<BaseElement>) {\n    if (node?.node) {\n      const item = node.node;\n      if (\n        item.rect.left <= position.x &&\n        position.x <= item.rect.left + item.rect.width &&\n        item.rect.top <= position.y &&\n        position.y <= item.rect.top + item.rect.height\n      ) {\n        if (\n          !(\n            filterPositionElements &&\n            item.attributes?.nodeType === NodeType.POSITION\n          ) &&\n          item.isVisible\n        ) {\n          matchingElements.push(item);\n        }\n      }\n    }\n\n    for (const child of node.children) {\n      dfs(child);\n    }\n  }\n\n  dfs(treeRoot);\n\n  if (matchingElements.length === 0) {\n    return undefined;\n  }\n\n  // Find the smallest element by area\n  const element = matchingElements.reduce((smallest, current) => {\n    const smallestArea = smallest.rect.width * smallest.rect.height;\n    const currentArea = current.rect.width * current.rect.height;\n    return currentArea < smallestArea ? current : smallest;\n  });\n\n  const distanceToCenter = distance(\n    { x: element.center[0], y: element.center[1] },\n    position,\n  );\n\n  if (requireStrictDistance) {\n    return distanceToCenter <= distanceThreshold ? element : undefined;\n  }\n\n  return element;\n}\n\nexport function distance(\n  point1: { x: number; y: number },\n  point2: { x: number; y: number },\n) {\n  return Math.sqrt((point1.x - point2.x) ** 2 + (point1.y - point2.y) ** 2);\n}\n\nexport const samplePageDescription = `\nAnd the page is described as follows:\n====================\nThe size of the page: 1280 x 720\nSome of the elements are marked with a rectangle in the screenshot corresponding to the markerId, some are not.\n\nDescription of all the elements in screenshot:\n<div id=\"969f1637\" markerId=\"1\" left=\"100\" top=\"100\" width=\"100\" height=\"100\"> // The markerId indicated by the rectangle label in the screenshot\n  <h4 id=\"b211ecb2\" markerId=\"5\" left=\"150\" top=\"150\" width=\"90\" height=\"60\">\n    The username is accepted\n  </h4>\n  ...many more\n</div>\n====================\n`;\n\nexport async function describeUserPage<\n  ElementType extends BaseElement = BaseElement,\n>(\n  context: Omit<UIContext<ElementType>, 'describer'>,\n  opt?: {\n    truncateTextLength?: number;\n    filterNonTextContent?: boolean;\n    domIncluded?: boolean | 'visible-only';\n    visibleOnly?: boolean;\n  },\n) {\n  const { screenshotBase64 } = context;\n  let width: number;\n  let height: number;\n\n  if (context.size) {\n    ({ width, height } = context.size);\n  } else {\n    const imgSize = await imageInfoOfBase64(screenshotBase64);\n    ({ width, height } = imgSize);\n  }\n\n  const treeRoot = context.tree;\n  // dfs tree, save the id and element info\n  const idElementMap: Record<string, ElementType> = {};\n  const flatElements: ElementType[] = treeToList(treeRoot);\n\n  if (opt?.domIncluded === true && flatElements.length >= 5000) {\n    console.warn(\n      'The number of elements is too large, it may cause the prompt to be too long, please use domIncluded: \"visible-only\" to reduce the number of elements',\n    );\n  }\n\n  flatElements.forEach((element) => {\n    idElementMap[element.id] = element;\n    if (typeof element.indexId !== 'undefined') {\n      idElementMap[`${element.indexId}`] = element;\n    }\n  });\n\n  let pageDescription = '';\n  const visibleOnly = opt?.visibleOnly ?? opt?.domIncluded === 'visible-only';\n  if (opt?.domIncluded || !vlLocateMode()) {\n    // non-vl mode must provide the page description\n    const contentTree = await descriptionOfTree(\n      treeRoot,\n      opt?.truncateTextLength,\n      opt?.filterNonTextContent,\n      visibleOnly,\n    );\n\n    // if match by position, don't need to provide the page description\n    const sizeDescription = describeSize({ width, height });\n    pageDescription = `The size of the page: ${sizeDescription} \\n The page elements tree:\\n${contentTree}`;\n  }\n\n  return {\n    description: pageDescription,\n    elementById(idOrIndexId: string) {\n      assert(typeof idOrIndexId !== 'undefined', 'id is required for query');\n      const item = idElementMap[`${idOrIndexId}`];\n      return item;\n    },\n    elementByPosition(\n      position: { x: number; y: number },\n      size: { width: number; height: number },\n    ) {\n      return elementByPositionWithElementInfo(treeRoot, position);\n    },\n    insertElementByPosition(position: { x: number; y: number }) {\n      const element = generateElementByPosition(position) as ElementType;\n\n      treeRoot.children.push({\n        node: element,\n        children: [],\n      });\n      flatElements.push(element);\n      idElementMap[element.id] = element;\n      return element;\n    },\n    size: { width, height },\n  };\n}\n"],"names":["describeSize","size","describeElement","elements","sliceLength","item","distanceThreshold","elementByPositionWithElementInfo","treeRoot","position","options","requireStrictDistance","filterPositionElements","assert","matchingElements","dfs","node","_item_attributes","NodeType","child","element","smallest","current","smallestArea","currentArea","distanceToCenter","distance","undefined","point1","point2","Math","samplePageDescription","describeUserPage","context","opt","screenshotBase64","width","height","imgSize","imageInfoOfBase64","idElementMap","flatElements","treeToList","console","pageDescription","visibleOnly","vlLocateMode","contentTree","descriptionOfTree","sizeDescription","idOrIndexId","generateElementByPosition"],"mappings":";;;;;AAWO,SAASA,aAAaC,IAAU;IACrC,OAAO,GAAGA,KAAK,KAAK,CAAC,GAAG,EAAEA,KAAK,MAAM,EAAE;AACzC;AAEO,SAASC,gBACdC,QAAoE;IAEpE,MAAMC,cAAc;IACpB,OAAOD,SACJ,GAAG,CAAC,CAACE,OACJ;YACEA,KAAK,EAAE;YACPA,KAAK,IAAI,CAAC,IAAI;YACdA,KAAK,IAAI,CAAC,GAAG;YACbA,KAAK,IAAI,CAAC,IAAI,GAAGA,KAAK,IAAI,CAAC,KAAK;YAChCA,KAAK,IAAI,CAAC,GAAG,GAAGA,KAAK,IAAI,CAAC,MAAM;YAChCA,KAAK,OAAO,CAAC,MAAM,GAAGD,cAClB,GAAGC,KAAK,OAAO,CAAC,KAAK,CAAC,GAAGD,aAAa,GAAG,CAAC,GAC1CC,KAAK,OAAO;SACjB,CAAC,IAAI,CAAC,OAER,IAAI,CAAC;AACV;AACO,MAAMC,oBAAoB;AAE1B,SAASC,iCACdC,QAAsC,EACtCC,QAGC,EACDC,OAGC;IAED,MAAMC,wBAAwBD,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,qBAAqB,AAAD,KAAK;IAChE,MAAME,yBAAyBF,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,sBAAsB,AAAD,KAAK;IAElEG,OAAO,AAAoB,WAAbJ,UAA0B;IAExC,MAAMK,mBAAkC,EAAE;IAE1C,SAASC,IAAIC,IAAkC;QAC7C,IAAIA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,IAAI,EAAE;YACd,MAAMX,OAAOW,KAAK,IAAI;YACtB,IACEX,KAAK,IAAI,CAAC,IAAI,IAAII,SAAS,CAAC,IAC5BA,SAAS,CAAC,IAAIJ,KAAK,IAAI,CAAC,IAAI,GAAGA,KAAK,IAAI,CAAC,KAAK,IAC9CA,KAAK,IAAI,CAAC,GAAG,IAAII,SAAS,CAAC,IAC3BA,SAAS,CAAC,IAAIJ,KAAK,IAAI,CAAC,GAAG,GAAGA,KAAK,IAAI,CAAC,MAAM,EAC9C;oBAIIY;gBAHJ,IACE,CACEL,CAAAA,0BACAK,AAAAA,SAAAA,CAAAA,mBAAAA,KAAK,UAAU,AAAD,IAAdA,KAAAA,IAAAA,iBAAiB,QAAQ,AAAD,MAAMC,SAAS,QAAO,KAEhDb,KAAK,SAAS,EAEdS,iBAAiB,IAAI,CAACT;YAE1B;QACF;QAEA,KAAK,MAAMc,SAASH,KAAK,QAAQ,CAC/BD,IAAII;IAER;IAEAJ,IAAIP;IAEJ,IAAIM,AAA4B,MAA5BA,iBAAiB,MAAM,EACzB;IAIF,MAAMM,UAAUN,iBAAiB,MAAM,CAAC,CAACO,UAAUC;QACjD,MAAMC,eAAeF,SAAS,IAAI,CAAC,KAAK,GAAGA,SAAS,IAAI,CAAC,MAAM;QAC/D,MAAMG,cAAcF,QAAQ,IAAI,CAAC,KAAK,GAAGA,QAAQ,IAAI,CAAC,MAAM;QAC5D,OAAOE,cAAcD,eAAeD,UAAUD;IAChD;IAEA,MAAMI,mBAAmBC,SACvB;QAAE,GAAGN,QAAQ,MAAM,CAAC,EAAE;QAAE,GAAGA,QAAQ,MAAM,CAAC,EAAE;IAAC,GAC7CX;IAGF,IAAIE,uBACF,OAAOc,oBAAoBnB,oBAAoBc,UAAUO;IAG3D,OAAOP;AACT;AAEO,SAASM,SACdE,MAAgC,EAChCC,MAAgC;IAEhC,OAAOC,KAAK,IAAI,CAAEF,AAAAA,CAAAA,OAAO,CAAC,GAAGC,OAAO,CAAC,AAAD,KAAM,IAAKD,AAAAA,CAAAA,OAAO,CAAC,GAAGC,OAAO,CAAC,AAAD,KAAM;AACzE;AAEO,MAAME,wBAAwB,CAAC;;;;;;;;;;;;;;AActC,CAAC;AAEM,eAAeC,iBAGpBC,OAAkD,EAClDC,GAKC;IAED,MAAM,EAAEC,gBAAgB,EAAE,GAAGF;IAC7B,IAAIG;IACJ,IAAIC;IAEJ,IAAIJ,QAAQ,IAAI,EACb,GAAEG,KAAK,EAAEC,MAAM,EAAE,GAAGJ,QAAQ,IAAG;SAC3B;QACL,MAAMK,UAAU,MAAMC,kBAAkBJ;QACvC,GAAEC,KAAK,EAAEC,MAAM,EAAE,GAAGC,OAAM;IAC7B;IAEA,MAAM9B,WAAWyB,QAAQ,IAAI;IAE7B,MAAMO,eAA4C,CAAC;IACnD,MAAMC,eAA8BC,WAAWlC;IAE/C,IAAI0B,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,MAAM,QAAQO,aAAa,MAAM,IAAI,MACtDE,QAAQ,IAAI,CACV;IAIJF,aAAa,OAAO,CAAC,CAACrB;QACpBoB,YAAY,CAACpB,QAAQ,EAAE,CAAC,GAAGA;QAC3B,IAAI,AAA2B,WAApBA,QAAQ,OAAO,EACxBoB,YAAY,CAAC,GAAGpB,QAAQ,OAAO,EAAE,CAAC,GAAGA;IAEzC;IAEA,IAAIwB,kBAAkB;IACtB,MAAMC,cAAcX,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAKA,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,MAAM;IAC7D,IAAIA,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAK,CAACY,gBAAgB;QAEvC,MAAMC,cAAc,MAAMC,kBACxBxC,UACA0B,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,kBAAkB,EACvBA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,oBAAoB,EACzBW;QAIF,MAAMI,kBAAkBjD,aAAa;YAAEoC;YAAOC;QAAO;QACrDO,kBAAkB,CAAC,sBAAsB,EAAEK,gBAAgB,6BAA6B,EAAEF,aAAa;IACzG;IAEA,OAAO;QACL,aAAaH;QACb,aAAYM,WAAmB;YAC7BrC,OAAO,AAAuB,WAAhBqC,aAA6B;YAC3C,MAAM7C,OAAOmC,YAAY,CAAC,GAAGU,aAAa,CAAC;YAC3C,OAAO7C;QACT;QACA,mBACEI,QAAkC,EAClCR,IAAuC;YAEvC,OAAOM,iCAAiCC,UAAUC;QACpD;QACA,yBAAwBA,QAAkC;YACxD,MAAMW,UAAU+B,0BAA0B1C;YAE1CD,SAAS,QAAQ,CAAC,IAAI,CAAC;gBACrB,MAAMY;gBACN,UAAU,EAAE;YACd;YACAqB,aAAa,IAAI,CAACrB;YAClBoB,YAAY,CAACpB,QAAQ,EAAE,CAAC,GAAGA;YAC3B,OAAOA;QACT;QACA,MAAM;YAAEgB;YAAOC;QAAO;IACxB;AACF"}
+{"version":3,"file":"ai-model/prompt/util.mjs","sources":["webpack://@midscene/core/./src/ai-model/prompt/util.ts"],"sourcesContent":["import { imageInfoOfBase64 } from '@/image/index';\nimport type { BaseElement, ElementTreeNode, Size, UIContext } from '@/types';\nimport { NodeType } from '@midscene/shared/constants';\nimport { type IModelPreferences, vlLocateMode } from '@midscene/shared/env';\nimport {\n  descriptionOfTree,\n  generateElementByPosition,\n  treeToList,\n} from '@midscene/shared/extractor';\nimport { assert } from '@midscene/shared/utils';\n\nexport function describeSize(size: Size) {\n  return `${size.width} x ${size.height}`;\n}\n\nexport function describeElement(\n  elements: (Pick<BaseElement, 'rect' | 'content'> & { id: string })[],\n) {\n  const sliceLength = 80;\n  return elements\n    .map((item) =>\n      [\n        item.id,\n        item.rect.left,\n        item.rect.top,\n        item.rect.left + item.rect.width,\n        item.rect.top + item.rect.height,\n        item.content.length > sliceLength\n          ? `${item.content.slice(0, sliceLength)}...`\n          : item.content,\n      ].join(', '),\n    )\n    .join('\\n');\n}\nexport const distanceThreshold = 16;\n\nexport function elementByPositionWithElementInfo(\n  treeRoot: ElementTreeNode<BaseElement>,\n  position: {\n    x: number;\n    y: number;\n  },\n  options?: {\n    requireStrictDistance?: boolean;\n    filterPositionElements?: boolean;\n  },\n) {\n  const requireStrictDistance = options?.requireStrictDistance ?? true;\n  const filterPositionElements = options?.filterPositionElements ?? false;\n\n  assert(typeof position !== 'undefined', 'position is required for query');\n\n  const matchingElements: BaseElement[] = [];\n\n  function dfs(node: ElementTreeNode<BaseElement>) {\n    if (node?.node) {\n      const item = node.node;\n      if (\n        item.rect.left <= position.x &&\n        position.x <= item.rect.left + item.rect.width &&\n        item.rect.top <= position.y &&\n        position.y <= item.rect.top + item.rect.height\n      ) {\n        if (\n          !(\n            filterPositionElements &&\n            item.attributes?.nodeType === NodeType.POSITION\n          ) &&\n          item.isVisible\n        ) {\n          matchingElements.push(item);\n        }\n      }\n    }\n\n    for (const child of node.children) {\n      dfs(child);\n    }\n  }\n\n  dfs(treeRoot);\n\n  if (matchingElements.length === 0) {\n    return undefined;\n  }\n\n  // Find the smallest element by area\n  const element = matchingElements.reduce((smallest, current) => {\n    const smallestArea = smallest.rect.width * smallest.rect.height;\n    const currentArea = current.rect.width * current.rect.height;\n    return currentArea < smallestArea ? current : smallest;\n  });\n\n  const distanceToCenter = distance(\n    { x: element.center[0], y: element.center[1] },\n    position,\n  );\n\n  if (requireStrictDistance) {\n    return distanceToCenter <= distanceThreshold ? element : undefined;\n  }\n\n  return element;\n}\n\nexport function distance(\n  point1: { x: number; y: number },\n  point2: { x: number; y: number },\n) {\n  return Math.sqrt((point1.x - point2.x) ** 2 + (point1.y - point2.y) ** 2);\n}\n\nexport const samplePageDescription = `\nAnd the page is described as follows:\n====================\nThe size of the page: 1280 x 720\nSome of the elements are marked with a rectangle in the screenshot corresponding to the markerId, some are not.\n\nDescription of all the elements in screenshot:\n<div id=\"969f1637\" markerId=\"1\" left=\"100\" top=\"100\" width=\"100\" height=\"100\"> // The markerId indicated by the rectangle label in the screenshot\n  <h4 id=\"b211ecb2\" markerId=\"5\" left=\"150\" top=\"150\" width=\"90\" height=\"60\">\n    The username is accepted\n  </h4>\n  ...many more\n</div>\n====================\n`;\n\nexport async function describeUserPage<\n  ElementType extends BaseElement = BaseElement,\n>(\n  context: Omit<UIContext<ElementType>, 'describer'>,\n  modelPreferences: IModelPreferences,\n  opt?: {\n    truncateTextLength?: number;\n    filterNonTextContent?: boolean;\n    domIncluded?: boolean | 'visible-only';\n    visibleOnly?: boolean;\n  },\n) {\n  const { screenshotBase64 } = context;\n  let width: number;\n  let height: number;\n\n  if (context.size) {\n    ({ width, height } = context.size);\n  } else {\n    const imgSize = await imageInfoOfBase64(screenshotBase64);\n    ({ width, height } = imgSize);\n  }\n\n  const treeRoot = context.tree;\n  // dfs tree, save the id and element info\n  const idElementMap: Record<string, ElementType> = {};\n  const flatElements: ElementType[] = treeToList(treeRoot);\n\n  if (opt?.domIncluded === true && flatElements.length >= 5000) {\n    console.warn(\n      'The number of elements is too large, it may cause the prompt to be too long, please use domIncluded: \"visible-only\" to reduce the number of elements',\n    );\n  }\n\n  flatElements.forEach((element) => {\n    idElementMap[element.id] = element;\n    if (typeof element.indexId !== 'undefined') {\n      idElementMap[`${element.indexId}`] = element;\n    }\n  });\n\n  let pageDescription = '';\n  const visibleOnly = opt?.visibleOnly ?? opt?.domIncluded === 'visible-only';\n  if (opt?.domIncluded || !vlLocateMode(modelPreferences)) {\n    // non-vl mode must provide the page description\n    const contentTree = await descriptionOfTree(\n      treeRoot,\n      opt?.truncateTextLength,\n      opt?.filterNonTextContent,\n      visibleOnly,\n    );\n\n    // if match by position, don't need to provide the page description\n    const sizeDescription = describeSize({ width, height });\n    pageDescription = `The size of the page: ${sizeDescription} \\n The page elements tree:\\n${contentTree}`;\n  }\n\n  return {\n    description: pageDescription,\n    elementById(idOrIndexId: string) {\n      assert(typeof idOrIndexId !== 'undefined', 'id is required for query');\n      const item = idElementMap[`${idOrIndexId}`];\n      return item;\n    },\n    elementByPosition(\n      position: { x: number; y: number },\n      size: { width: number; height: number },\n    ) {\n      return elementByPositionWithElementInfo(treeRoot, position);\n    },\n    insertElementByPosition(position: { x: number; y: number }) {\n      const element = generateElementByPosition(position) as ElementType;\n\n      treeRoot.children.push({\n        node: element,\n        children: [],\n      });\n      flatElements.push(element);\n      idElementMap[element.id] = element;\n      return element;\n    },\n    size: { width, height },\n  };\n}\n"],"names":["describeSize","size","describeElement","elements","sliceLength","item","distanceThreshold","elementByPositionWithElementInfo","treeRoot","position","options","requireStrictDistance","filterPositionElements","assert","matchingElements","dfs","node","_item_attributes","NodeType","child","element","smallest","current","smallestArea","currentArea","distanceToCenter","distance","undefined","point1","point2","Math","samplePageDescription","describeUserPage","context","modelPreferences","opt","screenshotBase64","width","height","imgSize","imageInfoOfBase64","idElementMap","flatElements","treeToList","console","pageDescription","visibleOnly","vlLocateMode","contentTree","descriptionOfTree","sizeDescription","idOrIndexId","generateElementByPosition"],"mappings":";;;;;AAWO,SAASA,aAAaC,IAAU;IACrC,OAAO,GAAGA,KAAK,KAAK,CAAC,GAAG,EAAEA,KAAK,MAAM,EAAE;AACzC;AAEO,SAASC,gBACdC,QAAoE;IAEpE,MAAMC,cAAc;IACpB,OAAOD,SACJ,GAAG,CAAC,CAACE,OACJ;YACEA,KAAK,EAAE;YACPA,KAAK,IAAI,CAAC,IAAI;YACdA,KAAK,IAAI,CAAC,GAAG;YACbA,KAAK,IAAI,CAAC,IAAI,GAAGA,KAAK,IAAI,CAAC,KAAK;YAChCA,KAAK,IAAI,CAAC,GAAG,GAAGA,KAAK,IAAI,CAAC,MAAM;YAChCA,KAAK,OAAO,CAAC,MAAM,GAAGD,cAClB,GAAGC,KAAK,OAAO,CAAC,KAAK,CAAC,GAAGD,aAAa,GAAG,CAAC,GAC1CC,KAAK,OAAO;SACjB,CAAC,IAAI,CAAC,OAER,IAAI,CAAC;AACV;AACO,MAAMC,oBAAoB;AAE1B,SAASC,iCACdC,QAAsC,EACtCC,QAGC,EACDC,OAGC;IAED,MAAMC,wBAAwBD,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,qBAAqB,AAAD,KAAK;IAChE,MAAME,yBAAyBF,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,sBAAsB,AAAD,KAAK;IAElEG,OAAO,AAAoB,WAAbJ,UAA0B;IAExC,MAAMK,mBAAkC,EAAE;IAE1C,SAASC,IAAIC,IAAkC;QAC7C,IAAIA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,IAAI,EAAE;YACd,MAAMX,OAAOW,KAAK,IAAI;YACtB,IACEX,KAAK,IAAI,CAAC,IAAI,IAAII,SAAS,CAAC,IAC5BA,SAAS,CAAC,IAAIJ,KAAK,IAAI,CAAC,IAAI,GAAGA,KAAK,IAAI,CAAC,KAAK,IAC9CA,KAAK,IAAI,CAAC,GAAG,IAAII,SAAS,CAAC,IAC3BA,SAAS,CAAC,IAAIJ,KAAK,IAAI,CAAC,GAAG,GAAGA,KAAK,IAAI,CAAC,MAAM,EAC9C;oBAIIY;gBAHJ,IACE,CACEL,CAAAA,0BACAK,AAAAA,SAAAA,CAAAA,mBAAAA,KAAK,UAAU,AAAD,IAAdA,KAAAA,IAAAA,iBAAiB,QAAQ,AAAD,MAAMC,SAAS,QAAO,KAEhDb,KAAK,SAAS,EAEdS,iBAAiB,IAAI,CAACT;YAE1B;QACF;QAEA,KAAK,MAAMc,SAASH,KAAK,QAAQ,CAC/BD,IAAII;IAER;IAEAJ,IAAIP;IAEJ,IAAIM,AAA4B,MAA5BA,iBAAiB,MAAM,EACzB;IAIF,MAAMM,UAAUN,iBAAiB,MAAM,CAAC,CAACO,UAAUC;QACjD,MAAMC,eAAeF,SAAS,IAAI,CAAC,KAAK,GAAGA,SAAS,IAAI,CAAC,MAAM;QAC/D,MAAMG,cAAcF,QAAQ,IAAI,CAAC,KAAK,GAAGA,QAAQ,IAAI,CAAC,MAAM;QAC5D,OAAOE,cAAcD,eAAeD,UAAUD;IAChD;IAEA,MAAMI,mBAAmBC,SACvB;QAAE,GAAGN,QAAQ,MAAM,CAAC,EAAE;QAAE,GAAGA,QAAQ,MAAM,CAAC,EAAE;IAAC,GAC7CX;IAGF,IAAIE,uBACF,OAAOc,oBAAoBnB,oBAAoBc,UAAUO;IAG3D,OAAOP;AACT;AAEO,SAASM,SACdE,MAAgC,EAChCC,MAAgC;IAEhC,OAAOC,KAAK,IAAI,CAAEF,AAAAA,CAAAA,OAAO,CAAC,GAAGC,OAAO,CAAC,AAAD,KAAM,IAAKD,AAAAA,CAAAA,OAAO,CAAC,GAAGC,OAAO,CAAC,AAAD,KAAM;AACzE;AAEO,MAAME,wBAAwB,CAAC;;;;;;;;;;;;;;AActC,CAAC;AAEM,eAAeC,iBAGpBC,OAAkD,EAClDC,gBAAmC,EACnCC,GAKC;IAED,MAAM,EAAEC,gBAAgB,EAAE,GAAGH;IAC7B,IAAII;IACJ,IAAIC;IAEJ,IAAIL,QAAQ,IAAI,EACb,GAAEI,KAAK,EAAEC,MAAM,EAAE,GAAGL,QAAQ,IAAG;SAC3B;QACL,MAAMM,UAAU,MAAMC,kBAAkBJ;QACvC,GAAEC,KAAK,EAAEC,MAAM,EAAE,GAAGC,OAAM;IAC7B;IAEA,MAAM/B,WAAWyB,QAAQ,IAAI;IAE7B,MAAMQ,eAA4C,CAAC;IACnD,MAAMC,eAA8BC,WAAWnC;IAE/C,IAAI2B,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,MAAM,QAAQO,aAAa,MAAM,IAAI,MACtDE,QAAQ,IAAI,CACV;IAIJF,aAAa,OAAO,CAAC,CAACtB;QACpBqB,YAAY,CAACrB,QAAQ,EAAE,CAAC,GAAGA;QAC3B,IAAI,AAA2B,WAApBA,QAAQ,OAAO,EACxBqB,YAAY,CAAC,GAAGrB,QAAQ,OAAO,EAAE,CAAC,GAAGA;IAEzC;IAEA,IAAIyB,kBAAkB;IACtB,MAAMC,cAAcX,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAKA,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,MAAM;IAC7D,IAAIA,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAK,CAACY,aAAab,mBAAmB;QAEvD,MAAMc,cAAc,MAAMC,kBACxBzC,UACA2B,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,kBAAkB,EACvBA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,oBAAoB,EACzBW;QAIF,MAAMI,kBAAkBlD,aAAa;YAAEqC;YAAOC;QAAO;QACrDO,kBAAkB,CAAC,sBAAsB,EAAEK,gBAAgB,6BAA6B,EAAEF,aAAa;IACzG;IAEA,OAAO;QACL,aAAaH;QACb,aAAYM,WAAmB;YAC7BtC,OAAO,AAAuB,WAAhBsC,aAA6B;YAC3C,MAAM9C,OAAOoC,YAAY,CAAC,GAAGU,aAAa,CAAC;YAC3C,OAAO9C;QACT;QACA,mBACEI,QAAkC,EAClCR,IAAuC;YAEvC,OAAOM,iCAAiCC,UAAUC;QACpD;QACA,yBAAwBA,QAAkC;YACxD,MAAMW,UAAUgC,0BAA0B3C;YAE1CD,SAAS,QAAQ,CAAC,IAAI,CAAC;gBACrB,MAAMY;gBACN,UAAU,EAAE;YACd;YACAsB,aAAa,IAAI,CAACtB;YAClBqB,YAAY,CAACrB,QAAQ,EAAE,CAAC,GAAGA;YAC3B,OAAOA;QACT;QACA,MAAM;YAAEiB;YAAOC;QAAO;IACxB;AACF"}

package/dist/es/ai-model/prompt/yaml-generator.mjs

@@ -141,7 +141,9 @@ Respond with YAML only, no explanations.`
                     }))
             });
         }
-        const response = await callAi(prompt, AIActionType.EXTRACT_DATA);
+        const response = await callAi(prompt, AIActionType.EXTRACT_DATA, {
+            intent: 'default'
+        });
         if ((null == response ? void 0 : response.content) && 'string' == typeof response.content) return response.content;
         throw new Error('Failed to generate YAML test configuration');
     } catch (error) {
@@ -197,12 +199,16 @@ Respond with YAML only, no explanations.`
                     }))
             });
         }
-        if (options.stream && options.onChunk) return await callAi(prompt, AIActionType.EXTRACT_DATA, void 0, {
+        if (options.stream && options.onChunk) return await callAi(prompt, AIActionType.EXTRACT_DATA, {
+            intent: 'default'
+        }, {
             stream: true,
             onChunk: options.onChunk
         });
         {
-            const response = await callAi(prompt, AIActionType.EXTRACT_DATA);
+            const response = await callAi(prompt, AIActionType.EXTRACT_DATA, {
+                intent: 'default'
+            });
             if ((null == response ? void 0 : response.content) && 'string' == typeof response.content) return {
                 content: response.content,
                 usage: response.usage,

package/dist/es/ai-model/prompt/yaml-generator.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/prompt/yaml-generator.mjs","sources":["webpack://@midscene/core/./src/ai-model/prompt/yaml-generator.ts"],"sourcesContent":["import type {\n  StreamingAIResponse,\n  StreamingCodeGenerationOptions,\n} from '@/types';\nimport { YAML_EXAMPLE_CODE } from '@midscene/shared/constants';\nimport {\n  AIActionType,\n  type ChatCompletionMessageParam,\n  callAi,\n} from '../index';\n\n// Common interfaces for test generation (shared between YAML and Playwright)\nexport interface EventCounts {\n  navigation: number;\n  click: number;\n  input: number;\n  scroll: number;\n  total: number;\n}\n\nexport interface InputDescription {\n  description: string;\n  value: string;\n}\n\nexport interface ProcessedEvent {\n  type: string;\n  timestamp: number;\n  url?: string;\n  title?: string;\n  elementDescription?: string;\n  value?: string;\n  pageInfo?: any;\n  elementRect?: any;\n}\n\nexport interface EventSummary {\n  testName: string;\n  startUrl: string;\n  eventCounts: EventCounts;\n  urls: string[];\n  clickDescriptions: string[];\n  inputDescriptions: InputDescription[];\n  events: ProcessedEvent[];\n}\n\n// Common ChromeRecordedEvent interface\nexport interface ChromeRecordedEvent {\n  type: string;\n  timestamp: number;\n  url?: string;\n  title?: string;\n  elementDescription?: string;\n  value?: string;\n  pageInfo?: any;\n  elementRect?: any;\n  screenshotBefore?: string;\n  screenshotAfter?: string;\n  screenshotWithBox?: string;\n}\n\nexport interface YamlGenerationOptions {\n  testName?: string;\n  includeTimestamps?: boolean;\n  maxScreenshots?: number;\n  description?: string;\n}\n\nexport interface FilteredEvents {\n  navigationEvents: ChromeRecordedEvent[];\n  clickEvents: ChromeRecordedEvent[];\n  inputEvents: ChromeRecordedEvent[];\n  scrollEvents: ChromeRecordedEvent[];\n}\n\n// Common utility functions (shared between YAML and Playwright generators)\n\n/**\n * Get screenshots from events for LLM context\n */\nexport const getScreenshotsForLLM = (\n  events: ChromeRecordedEvent[],\n  maxScreenshots = 1,\n): string[] => {\n  // Find events with screenshots, prioritizing navigation and click events\n  const eventsWithScreenshots = events.filter(\n    (event) =>\n      event.screenshotBefore ||\n      event.screenshotAfter ||\n      event.screenshotWithBox,\n  );\n\n  // Sort them by priority (navigation first, then clicks, then others)\n  const sortedEvents = [...eventsWithScreenshots].sort((a, b) => {\n    if (a.type === 'navigation' && b.type !== 'navigation') return -1;\n    if (a.type !== 'navigation' && b.type === 'navigation') return 1;\n    if (a.type === 'click' && b.type !== 'click') return -1;\n    if (a.type !== 'click' && b.type === 'click') return 1;\n    return 0;\n  });\n\n  // Extract up to maxScreenshots screenshots\n  const screenshots: string[] = [];\n  for (const event of sortedEvents) {\n    // Prefer the most informative screenshot\n    const screenshot =\n      event.screenshotWithBox ||\n      event.screenshotAfter ||\n      event.screenshotBefore;\n    if (screenshot && !screenshots.includes(screenshot)) {\n      screenshots.push(screenshot);\n      if (screenshots.length >= maxScreenshots) break;\n    }\n  }\n\n  return screenshots;\n};\n\n/**\n * Filter events by type for easier processing\n */\nexport const filterEventsByType = (\n  events: ChromeRecordedEvent[],\n): FilteredEvents => {\n  return {\n    navigationEvents: events.filter((event) => event.type === 'navigation'),\n    clickEvents: events.filter((event) => event.type === 'click'),\n    inputEvents: events.filter((event) => event.type === 'input'),\n    scrollEvents: events.filter((event) => event.type === 'scroll'),\n  };\n};\n\n/**\n * Create event counts summary\n */\nexport const createEventCounts = (\n  filteredEvents: FilteredEvents,\n  totalEvents: number,\n): EventCounts => {\n  return {\n    navigation: filteredEvents.navigationEvents.length,\n    click: filteredEvents.clickEvents.length,\n    input: filteredEvents.inputEvents.length,\n    scroll: filteredEvents.scrollEvents.length,\n    total: totalEvents,\n  };\n};\n\n/**\n * Extract input descriptions from input events\n */\nexport const extractInputDescriptions = (\n  inputEvents: ChromeRecordedEvent[],\n): InputDescription[] => {\n  return inputEvents\n    .map((event) => ({\n      description: event.elementDescription || '',\n      value: event.value || '',\n    }))\n    .filter((item) => item.description && item.value);\n};\n\n/**\n * Process events for LLM consumption\n */\nexport const processEventsForLLM = (\n  events: ChromeRecordedEvent[],\n): ProcessedEvent[] => {\n  return events.map((event) => ({\n    type: event.type,\n    timestamp: event.timestamp,\n    url: event.url,\n    title: event.title,\n    elementDescription: event.elementDescription,\n    value: event.value,\n    pageInfo: event.pageInfo,\n    elementRect: event.elementRect,\n  }));\n};\n\n/**\n * Prepare comprehensive event summary for LLM\n */\nexport const prepareEventSummary = (\n  events: ChromeRecordedEvent[],\n  options: { testName?: string; maxScreenshots?: number } = {},\n): EventSummary => {\n  const filteredEvents = filterEventsByType(events);\n  const eventCounts = createEventCounts(filteredEvents, events.length);\n\n  // Extract useful information from events\n  const startUrl =\n    filteredEvents.navigationEvents.length > 0\n      ? filteredEvents.navigationEvents[0].url || ''\n      : '';\n\n  const clickDescriptions = filteredEvents.clickEvents\n    .map((event) => event.elementDescription)\n    .filter((desc): desc is string => Boolean(desc))\n    .slice(0, 10);\n\n  const inputDescriptions = extractInputDescriptions(\n    filteredEvents.inputEvents,\n  ).slice(0, 10);\n\n  const urls = filteredEvents.navigationEvents\n    .map((e) => e.url)\n    .filter((url): url is string => Boolean(url))\n    .slice(0, 5);\n\n  const processedEvents = processEventsForLLM(events);\n\n  return {\n    testName: options.testName || 'Automated test from recorded events',\n    startUrl,\n    eventCounts,\n    urls,\n    clickDescriptions,\n    inputDescriptions,\n    events: processedEvents,\n  };\n};\n\n/**\n * Create message content for LLM with optional screenshots\n */\nexport const createMessageContent = (\n  promptText: string,\n  screenshots: string[] = [],\n  includeScreenshots = true,\n) => {\n  const messageContent: any[] = [\n    {\n      type: 'text',\n      text: promptText,\n    },\n  ];\n\n  // Add screenshots if available and requested\n  if (includeScreenshots && screenshots.length > 0) {\n    messageContent.unshift({\n      type: 'text',\n      text: 'Here are screenshots from the recording session to help you understand the context:',\n    });\n\n    screenshots.forEach((screenshot) => {\n      messageContent.push({\n        type: 'image_url',\n        image_url: {\n          url: screenshot,\n        },\n      });\n    });\n  }\n\n  return messageContent;\n};\n\n/**\n * Validate events before processing\n */\nexport const validateEvents = (events: ChromeRecordedEvent[]): void => {\n  if (!events.length) {\n    throw new Error('No events provided for test generation');\n  }\n};\n\n// YAML-specific generation functions\n\n/**\n * Generates YAML test configuration from recorded events using AI\n */\nexport const generateYamlTest = async (\n  events: ChromeRecordedEvent[],\n  options: YamlGenerationOptions = {},\n): Promise<string> => {\n  try {\n    // Validate input\n    validateEvents(events);\n\n    // Prepare event summary using shared utilities\n    const summary = prepareEventSummary(events, {\n      testName: options.testName,\n      maxScreenshots: options.maxScreenshots || 3,\n    });\n\n    // Add YAML-specific options to summary\n    const yamlSummary = {\n      ...summary,\n      includeTimestamps: options.includeTimestamps || false,\n    };\n\n    // Get screenshots for visual context\n    const screenshots = getScreenshotsForLLM(\n      events,\n      options.maxScreenshots || 3,\n    );\n\n    // Use LLM to generate the YAML test configuration\n    const prompt: ChatCompletionMessageParam[] = [\n      {\n        role: 'system',\n        content: `You are an expert in Midscene.js YAML test generation. Generate clean, accurate YAML following these rules: ${YAML_EXAMPLE_CODE}`,\n      },\n      {\n        role: 'user',\n        content: `Generate YAML test for Midscene.js automation from recorded browser events.\n\nEvent Summary:\n${JSON.stringify(yamlSummary, null, 2)}\n\nConvert events:\n- navigation → target.url\n- click → aiTap with element description\n- input → aiInput with value and locate\n- scroll → aiScroll with appropriate direction\n- Add aiAssert for important state changes\n\nRespond with YAML only, no explanations.`,\n      },\n    ];\n\n    // Add screenshots if available and requested\n    if (screenshots.length > 0) {\n      prompt.push({\n        role: 'user',\n        content:\n          'Here are screenshots from the recording session to help you understand the context:',\n      });\n\n      prompt.push({\n        role: 'user',\n        content: screenshots.map((screenshot) => ({\n          type: 'image_url',\n          image_url: {\n            url: screenshot,\n          },\n        })),\n      });\n    }\n\n    const response = await callAi(prompt, AIActionType.EXTRACT_DATA);\n\n    if (response?.content && typeof response.content === 'string') {\n      return response.content;\n    }\n\n    throw new Error('Failed to generate YAML test configuration');\n  } catch (error) {\n    throw new Error(`Failed to generate YAML test: ${error}`);\n  }\n};\n\n/**\n * Generates YAML test configuration from recorded events using AI with streaming support\n */\nexport const generateYamlTestStream = async (\n  events: ChromeRecordedEvent[],\n  options: YamlGenerationOptions & StreamingCodeGenerationOptions = {},\n): Promise<StreamingAIResponse> => {\n  try {\n    // Validate input\n    validateEvents(events);\n\n    // Prepare event summary using shared utilities\n    const summary = prepareEventSummary(events, {\n      testName: options.testName,\n      maxScreenshots: options.maxScreenshots || 3,\n    });\n\n    // Add YAML-specific options to summary\n    const yamlSummary = {\n      ...summary,\n      includeTimestamps: options.includeTimestamps || false,\n    };\n\n    // Get screenshots for visual context\n    const screenshots = getScreenshotsForLLM(\n      events,\n      options.maxScreenshots || 3,\n    );\n\n    // Use LLM to generate the YAML test configuration\n    const prompt: ChatCompletionMessageParam[] = [\n      {\n        role: 'system',\n        content: `You are an expert in Midscene.js YAML test generation. Generate clean, accurate YAML following these rules: ${YAML_EXAMPLE_CODE}`,\n      },\n      {\n        role: 'user',\n        content: `Generate YAML test for Midscene.js automation from recorded browser events.\n\nEvent Summary:\n${JSON.stringify(yamlSummary, null, 2)}\n\nConvert events:\n- navigation → target.url\n- click → aiTap with element description\n- input → aiInput with value and locate\n- scroll → aiScroll with appropriate direction\n- Add aiAssert for important state changes\n\nRespond with YAML only, no explanations.`,\n      },\n    ];\n\n    // Add screenshots if available and requested\n    if (screenshots.length > 0) {\n      prompt.push({\n        role: 'user',\n        content:\n          'Here are screenshots from the recording session to help you understand the context:',\n      });\n\n      prompt.push({\n        role: 'user',\n        content: screenshots.map((screenshot) => ({\n          type: 'image_url',\n          image_url: {\n            url: screenshot,\n          },\n        })),\n      });\n    }\n\n    if (options.stream && options.onChunk) {\n      // Use streaming\n      return await callAi(prompt, AIActionType.EXTRACT_DATA, undefined, {\n        stream: true,\n        onChunk: options.onChunk,\n      });\n    } else {\n      // Fallback to non-streaming\n      const response = await callAi(prompt, AIActionType.EXTRACT_DATA);\n\n      if (response?.content && typeof response.content === 'string') {\n        return {\n          content: response.content,\n          usage: response.usage,\n          isStreamed: false,\n        };\n      }\n\n      throw new Error('Failed to generate YAML test configuration');\n    }\n  } catch (error) {\n    throw new Error(`Failed to generate YAML test: ${error}`);\n  }\n};\n"],"names":["getScreenshotsForLLM","events","maxScreenshots","eventsWithScreenshots","event","sortedEvents","a","b","screenshots","screenshot","filterEventsByType","createEventCounts","filteredEvents","totalEvents","extractInputDescriptions","inputEvents","item","processEventsForLLM","prepareEventSummary","options","eventCounts","startUrl","clickDescriptions","desc","Boolean","inputDescriptions","urls","e","url","processedEvents","createMessageContent","promptText","includeScreenshots","messageContent","validateEvents","Error","generateYamlTest","summary","yamlSummary","prompt","YAML_EXAMPLE_CODE","JSON","response","callAi","AIActionType","error","generateYamlTestStream","undefined"],"mappings":";;AAgFO,MAAMA,uBAAuB,CAClCC,QACAC,iBAAiB,CAAC;IAGlB,MAAMC,wBAAwBF,OAAO,MAAM,CACzC,CAACG,QACCA,MAAM,gBAAgB,IACtBA,MAAM,eAAe,IACrBA,MAAM,iBAAiB;IAI3B,MAAMC,eAAe;WAAIF;KAAsB,CAAC,IAAI,CAAC,CAACG,GAAGC;QACvD,IAAID,AAAW,iBAAXA,EAAE,IAAI,IAAqBC,AAAW,iBAAXA,EAAE,IAAI,EAAmB,OAAO;QAC/D,IAAID,AAAW,iBAAXA,EAAE,IAAI,IAAqBC,AAAW,iBAAXA,EAAE,IAAI,EAAmB,OAAO;QAC/D,IAAID,AAAW,YAAXA,EAAE,IAAI,IAAgBC,AAAW,YAAXA,EAAE,IAAI,EAAc,OAAO;QACrD,IAAID,AAAW,YAAXA,EAAE,IAAI,IAAgBC,AAAW,YAAXA,EAAE,IAAI,EAAc,OAAO;QACrD,OAAO;IACT;IAGA,MAAMC,cAAwB,EAAE;IAChC,KAAK,MAAMJ,SAASC,aAAc;QAEhC,MAAMI,aACJL,MAAM,iBAAiB,IACvBA,MAAM,eAAe,IACrBA,MAAM,gBAAgB;QACxB,IAAIK,cAAc,CAACD,YAAY,QAAQ,CAACC,aAAa;YACnDD,YAAY,IAAI,CAACC;YACjB,IAAID,YAAY,MAAM,IAAIN,gBAAgB;QAC5C;IACF;IAEA,OAAOM;AACT;AAKO,MAAME,qBAAqB,CAChCT,SAEO;QACL,kBAAkBA,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,iBAAfA,MAAM,IAAI;QACrD,aAAaH,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,YAAfA,MAAM,IAAI;QAChD,aAAaH,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,YAAfA,MAAM,IAAI;QAChD,cAAcH,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,aAAfA,MAAM,IAAI;IACnD;AAMK,MAAMO,oBAAoB,CAC/BC,gBACAC,cAEO;QACL,YAAYD,eAAe,gBAAgB,CAAC,MAAM;QAClD,OAAOA,eAAe,WAAW,CAAC,MAAM;QACxC,OAAOA,eAAe,WAAW,CAAC,MAAM;QACxC,QAAQA,eAAe,YAAY,CAAC,MAAM;QAC1C,OAAOC;IACT;AAMK,MAAMC,2BAA2B,CACtCC,cAEOA,YACJ,GAAG,CAAC,CAACX,QAAW;YACf,aAAaA,MAAM,kBAAkB,IAAI;YACzC,OAAOA,MAAM,KAAK,IAAI;QACxB,IACC,MAAM,CAAC,CAACY,OAASA,KAAK,WAAW,IAAIA,KAAK,KAAK;AAM7C,MAAMC,sBAAsB,CACjChB,SAEOA,OAAO,GAAG,CAAC,CAACG,QAAW;YAC5B,MAAMA,MAAM,IAAI;YAChB,WAAWA,MAAM,SAAS;YAC1B,KAAKA,MAAM,GAAG;YACd,OAAOA,MAAM,KAAK;YAClB,oBAAoBA,MAAM,kBAAkB;YAC5C,OAAOA,MAAM,KAAK;YAClB,UAAUA,MAAM,QAAQ;YACxB,aAAaA,MAAM,WAAW;QAChC;AAMK,MAAMc,sBAAsB,CACjCjB,QACAkB,UAA0D,CAAC,CAAC;IAE5D,MAAMP,iBAAiBF,mBAAmBT;IAC1C,MAAMmB,cAAcT,kBAAkBC,gBAAgBX,OAAO,MAAM;IAGnE,MAAMoB,WACJT,eAAe,gBAAgB,CAAC,MAAM,GAAG,IACrCA,eAAe,gBAAgB,CAAC,EAAE,CAAC,GAAG,IAAI,KAC1C;IAEN,MAAMU,oBAAoBV,eAAe,WAAW,CACjD,GAAG,CAAC,CAACR,QAAUA,MAAM,kBAAkB,EACvC,MAAM,CAAC,CAACmB,OAAyBC,QAAQD,OACzC,KAAK,CAAC,GAAG;IAEZ,MAAME,oBAAoBX,yBACxBF,eAAe,WAAW,EAC1B,KAAK,CAAC,GAAG;IAEX,MAAMc,OAAOd,eAAe,gBAAgB,CACzC,GAAG,CAAC,CAACe,IAAMA,EAAE,GAAG,EAChB,MAAM,CAAC,CAACC,MAAuBJ,QAAQI,MACvC,KAAK,CAAC,GAAG;IAEZ,MAAMC,kBAAkBZ,oBAAoBhB;IAE5C,OAAO;QACL,UAAUkB,QAAQ,QAAQ,IAAI;QAC9BE;QACAD;QACAM;QACAJ;QACAG;QACA,QAAQI;IACV;AACF;AAKO,MAAMC,uBAAuB,CAClCC,YACAvB,cAAwB,EAAE,EAC1BwB,qBAAqB,IAAI;IAEzB,MAAMC,iBAAwB;QAC5B;YACE,MAAM;YACN,MAAMF;QACR;KACD;IAGD,IAAIC,sBAAsBxB,YAAY,MAAM,GAAG,GAAG;QAChDyB,eAAe,OAAO,CAAC;YACrB,MAAM;YACN,MAAM;QACR;QAEAzB,YAAY,OAAO,CAAC,CAACC;YACnBwB,eAAe,IAAI,CAAC;gBAClB,MAAM;gBACN,WAAW;oBACT,KAAKxB;gBACP;YACF;QACF;IACF;IAEA,OAAOwB;AACT;AAKO,MAAMC,iBAAiB,CAACjC;IAC7B,IAAI,CAACA,OAAO,MAAM,EAChB,MAAM,IAAIkC,MAAM;AAEpB;AAOO,MAAMC,mBAAmB,OAC9BnC,QACAkB,UAAiC,CAAC,CAAC;IAEnC,IAAI;QAEFe,eAAejC;QAGf,MAAMoC,UAAUnB,oBAAoBjB,QAAQ;YAC1C,UAAUkB,QAAQ,QAAQ;YAC1B,gBAAgBA,QAAQ,cAAc,IAAI;QAC5C;QAGA,MAAMmB,cAAc;YAClB,GAAGD,OAAO;YACV,mBAAmBlB,QAAQ,iBAAiB,IAAI;QAClD;QAGA,MAAMX,cAAcR,qBAClBC,QACAkB,QAAQ,cAAc,IAAI;QAI5B,MAAMoB,SAAuC;YAC3C;gBACE,MAAM;gBACN,SAAS,CAAC,4GAA4G,EAAEC,mBAAmB;YAC7I;YACA;gBACE,MAAM;gBACN,SAAS,CAAC;;;AAGlB,EAAEC,KAAK,SAAS,CAACH,aAAa,MAAM,GAAG;;;;;;;;;wCASC,CAAC;YACnC;SACD;QAGD,IAAI9B,YAAY,MAAM,GAAG,GAAG;YAC1B+B,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SACE;YACJ;YAEAA,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SAAS/B,YAAY,GAAG,CAAC,CAACC,aAAgB;wBACxC,MAAM;wBACN,WAAW;4BACT,KAAKA;wBACP;oBACF;YACF;QACF;QAEA,MAAMiC,WAAW,MAAMC,OAAOJ,QAAQK,aAAa,YAAY;QAE/D,IAAIF,AAAAA,CAAAA,QAAAA,WAAAA,KAAAA,IAAAA,SAAU,OAAO,AAAD,KAAK,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAOA,SAAS,OAAO;QAGzB,MAAM,IAAIP,MAAM;IAClB,EAAE,OAAOU,OAAO;QACd,MAAM,IAAIV,MAAM,CAAC,8BAA8B,EAAEU,OAAO;IAC1D;AACF;AAKO,MAAMC,yBAAyB,OACpC7C,QACAkB,UAAkE,CAAC,CAAC;IAEpE,IAAI;QAEFe,eAAejC;QAGf,MAAMoC,UAAUnB,oBAAoBjB,QAAQ;YAC1C,UAAUkB,QAAQ,QAAQ;YAC1B,gBAAgBA,QAAQ,cAAc,IAAI;QAC5C;QAGA,MAAMmB,cAAc;YAClB,GAAGD,OAAO;YACV,mBAAmBlB,QAAQ,iBAAiB,IAAI;QAClD;QAGA,MAAMX,cAAcR,qBAClBC,QACAkB,QAAQ,cAAc,IAAI;QAI5B,MAAMoB,SAAuC;YAC3C;gBACE,MAAM;gBACN,SAAS,CAAC,4GAA4G,EAAEC,mBAAmB;YAC7I;YACA;gBACE,MAAM;gBACN,SAAS,CAAC;;;AAGlB,EAAEC,KAAK,SAAS,CAACH,aAAa,MAAM,GAAG;;;;;;;;;wCASC,CAAC;YACnC;SACD;QAGD,IAAI9B,YAAY,MAAM,GAAG,GAAG;YAC1B+B,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SACE;YACJ;YAEAA,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SAAS/B,YAAY,GAAG,CAAC,CAACC,aAAgB;wBACxC,MAAM;wBACN,WAAW;4BACT,KAAKA;wBACP;oBACF;YACF;QACF;QAEA,IAAIU,QAAQ,MAAM,IAAIA,QAAQ,OAAO,EAEnC,OAAO,MAAMwB,OAAOJ,QAAQK,aAAa,YAAY,EAAEG,QAAW;YAChE,QAAQ;YACR,SAAS5B,QAAQ,OAAO;QAC1B;QACK;YAEL,MAAMuB,WAAW,MAAMC,OAAOJ,QAAQK,aAAa,YAAY;YAE/D,IAAIF,AAAAA,CAAAA,QAAAA,WAAAA,KAAAA,IAAAA,SAAU,OAAO,AAAD,KAAK,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAO;gBACL,SAASA,SAAS,OAAO;gBACzB,OAAOA,SAAS,KAAK;gBACrB,YAAY;YACd;YAGF,MAAM,IAAIP,MAAM;QAClB;IACF,EAAE,OAAOU,OAAO;QACd,MAAM,IAAIV,MAAM,CAAC,8BAA8B,EAAEU,OAAO;IAC1D;AACF"}
+{"version":3,"file":"ai-model/prompt/yaml-generator.mjs","sources":["webpack://@midscene/core/./src/ai-model/prompt/yaml-generator.ts"],"sourcesContent":["import type {\n  StreamingAIResponse,\n  StreamingCodeGenerationOptions,\n} from '@/types';\nimport { YAML_EXAMPLE_CODE } from '@midscene/shared/constants';\nimport {\n  AIActionType,\n  type ChatCompletionMessageParam,\n  callAi,\n} from '../index';\n\n// Common interfaces for test generation (shared between YAML and Playwright)\nexport interface EventCounts {\n  navigation: number;\n  click: number;\n  input: number;\n  scroll: number;\n  total: number;\n}\n\nexport interface InputDescription {\n  description: string;\n  value: string;\n}\n\nexport interface ProcessedEvent {\n  type: string;\n  timestamp: number;\n  url?: string;\n  title?: string;\n  elementDescription?: string;\n  value?: string;\n  pageInfo?: any;\n  elementRect?: any;\n}\n\nexport interface EventSummary {\n  testName: string;\n  startUrl: string;\n  eventCounts: EventCounts;\n  urls: string[];\n  clickDescriptions: string[];\n  inputDescriptions: InputDescription[];\n  events: ProcessedEvent[];\n}\n\n// Common ChromeRecordedEvent interface\nexport interface ChromeRecordedEvent {\n  type: string;\n  timestamp: number;\n  url?: string;\n  title?: string;\n  elementDescription?: string;\n  value?: string;\n  pageInfo?: any;\n  elementRect?: any;\n  screenshotBefore?: string;\n  screenshotAfter?: string;\n  screenshotWithBox?: string;\n}\n\nexport interface YamlGenerationOptions {\n  testName?: string;\n  includeTimestamps?: boolean;\n  maxScreenshots?: number;\n  description?: string;\n}\n\nexport interface FilteredEvents {\n  navigationEvents: ChromeRecordedEvent[];\n  clickEvents: ChromeRecordedEvent[];\n  inputEvents: ChromeRecordedEvent[];\n  scrollEvents: ChromeRecordedEvent[];\n}\n\n// Common utility functions (shared between YAML and Playwright generators)\n\n/**\n * Get screenshots from events for LLM context\n */\nexport const getScreenshotsForLLM = (\n  events: ChromeRecordedEvent[],\n  maxScreenshots = 1,\n): string[] => {\n  // Find events with screenshots, prioritizing navigation and click events\n  const eventsWithScreenshots = events.filter(\n    (event) =>\n      event.screenshotBefore ||\n      event.screenshotAfter ||\n      event.screenshotWithBox,\n  );\n\n  // Sort them by priority (navigation first, then clicks, then others)\n  const sortedEvents = [...eventsWithScreenshots].sort((a, b) => {\n    if (a.type === 'navigation' && b.type !== 'navigation') return -1;\n    if (a.type !== 'navigation' && b.type === 'navigation') return 1;\n    if (a.type === 'click' && b.type !== 'click') return -1;\n    if (a.type !== 'click' && b.type === 'click') return 1;\n    return 0;\n  });\n\n  // Extract up to maxScreenshots screenshots\n  const screenshots: string[] = [];\n  for (const event of sortedEvents) {\n    // Prefer the most informative screenshot\n    const screenshot =\n      event.screenshotWithBox ||\n      event.screenshotAfter ||\n      event.screenshotBefore;\n    if (screenshot && !screenshots.includes(screenshot)) {\n      screenshots.push(screenshot);\n      if (screenshots.length >= maxScreenshots) break;\n    }\n  }\n\n  return screenshots;\n};\n\n/**\n * Filter events by type for easier processing\n */\nexport const filterEventsByType = (\n  events: ChromeRecordedEvent[],\n): FilteredEvents => {\n  return {\n    navigationEvents: events.filter((event) => event.type === 'navigation'),\n    clickEvents: events.filter((event) => event.type === 'click'),\n    inputEvents: events.filter((event) => event.type === 'input'),\n    scrollEvents: events.filter((event) => event.type === 'scroll'),\n  };\n};\n\n/**\n * Create event counts summary\n */\nexport const createEventCounts = (\n  filteredEvents: FilteredEvents,\n  totalEvents: number,\n): EventCounts => {\n  return {\n    navigation: filteredEvents.navigationEvents.length,\n    click: filteredEvents.clickEvents.length,\n    input: filteredEvents.inputEvents.length,\n    scroll: filteredEvents.scrollEvents.length,\n    total: totalEvents,\n  };\n};\n\n/**\n * Extract input descriptions from input events\n */\nexport const extractInputDescriptions = (\n  inputEvents: ChromeRecordedEvent[],\n): InputDescription[] => {\n  return inputEvents\n    .map((event) => ({\n      description: event.elementDescription || '',\n      value: event.value || '',\n    }))\n    .filter((item) => item.description && item.value);\n};\n\n/**\n * Process events for LLM consumption\n */\nexport const processEventsForLLM = (\n  events: ChromeRecordedEvent[],\n): ProcessedEvent[] => {\n  return events.map((event) => ({\n    type: event.type,\n    timestamp: event.timestamp,\n    url: event.url,\n    title: event.title,\n    elementDescription: event.elementDescription,\n    value: event.value,\n    pageInfo: event.pageInfo,\n    elementRect: event.elementRect,\n  }));\n};\n\n/**\n * Prepare comprehensive event summary for LLM\n */\nexport const prepareEventSummary = (\n  events: ChromeRecordedEvent[],\n  options: { testName?: string; maxScreenshots?: number } = {},\n): EventSummary => {\n  const filteredEvents = filterEventsByType(events);\n  const eventCounts = createEventCounts(filteredEvents, events.length);\n\n  // Extract useful information from events\n  const startUrl =\n    filteredEvents.navigationEvents.length > 0\n      ? filteredEvents.navigationEvents[0].url || ''\n      : '';\n\n  const clickDescriptions = filteredEvents.clickEvents\n    .map((event) => event.elementDescription)\n    .filter((desc): desc is string => Boolean(desc))\n    .slice(0, 10);\n\n  const inputDescriptions = extractInputDescriptions(\n    filteredEvents.inputEvents,\n  ).slice(0, 10);\n\n  const urls = filteredEvents.navigationEvents\n    .map((e) => e.url)\n    .filter((url): url is string => Boolean(url))\n    .slice(0, 5);\n\n  const processedEvents = processEventsForLLM(events);\n\n  return {\n    testName: options.testName || 'Automated test from recorded events',\n    startUrl,\n    eventCounts,\n    urls,\n    clickDescriptions,\n    inputDescriptions,\n    events: processedEvents,\n  };\n};\n\n/**\n * Create message content for LLM with optional screenshots\n */\nexport const createMessageContent = (\n  promptText: string,\n  screenshots: string[] = [],\n  includeScreenshots = true,\n) => {\n  const messageContent: any[] = [\n    {\n      type: 'text',\n      text: promptText,\n    },\n  ];\n\n  // Add screenshots if available and requested\n  if (includeScreenshots && screenshots.length > 0) {\n    messageContent.unshift({\n      type: 'text',\n      text: 'Here are screenshots from the recording session to help you understand the context:',\n    });\n\n    screenshots.forEach((screenshot) => {\n      messageContent.push({\n        type: 'image_url',\n        image_url: {\n          url: screenshot,\n        },\n      });\n    });\n  }\n\n  return messageContent;\n};\n\n/**\n * Validate events before processing\n */\nexport const validateEvents = (events: ChromeRecordedEvent[]): void => {\n  if (!events.length) {\n    throw new Error('No events provided for test generation');\n  }\n};\n\n// YAML-specific generation functions\n\n/**\n * Generates YAML test configuration from recorded events using AI\n */\nexport const generateYamlTest = async (\n  events: ChromeRecordedEvent[],\n  options: YamlGenerationOptions = {},\n): Promise<string> => {\n  try {\n    // Validate input\n    validateEvents(events);\n\n    // Prepare event summary using shared utilities\n    const summary = prepareEventSummary(events, {\n      testName: options.testName,\n      maxScreenshots: options.maxScreenshots || 3,\n    });\n\n    // Add YAML-specific options to summary\n    const yamlSummary = {\n      ...summary,\n      includeTimestamps: options.includeTimestamps || false,\n    };\n\n    // Get screenshots for visual context\n    const screenshots = getScreenshotsForLLM(\n      events,\n      options.maxScreenshots || 3,\n    );\n\n    // Use LLM to generate the YAML test configuration\n    const prompt: ChatCompletionMessageParam[] = [\n      {\n        role: 'system',\n        content: `You are an expert in Midscene.js YAML test generation. Generate clean, accurate YAML following these rules: ${YAML_EXAMPLE_CODE}`,\n      },\n      {\n        role: 'user',\n        content: `Generate YAML test for Midscene.js automation from recorded browser events.\n\nEvent Summary:\n${JSON.stringify(yamlSummary, null, 2)}\n\nConvert events:\n- navigation → target.url\n- click → aiTap with element description\n- input → aiInput with value and locate\n- scroll → aiScroll with appropriate direction\n- Add aiAssert for important state changes\n\nRespond with YAML only, no explanations.`,\n      },\n    ];\n\n    // Add screenshots if available and requested\n    if (screenshots.length > 0) {\n      prompt.push({\n        role: 'user',\n        content:\n          'Here are screenshots from the recording session to help you understand the context:',\n      });\n\n      prompt.push({\n        role: 'user',\n        content: screenshots.map((screenshot) => ({\n          type: 'image_url',\n          image_url: {\n            url: screenshot,\n          },\n        })),\n      });\n    }\n\n    const response = await callAi(prompt, AIActionType.EXTRACT_DATA, {\n      intent: 'default',\n    });\n\n    if (response?.content && typeof response.content === 'string') {\n      return response.content;\n    }\n\n    throw new Error('Failed to generate YAML test configuration');\n  } catch (error) {\n    throw new Error(`Failed to generate YAML test: ${error}`);\n  }\n};\n\n/**\n * Generates YAML test configuration from recorded events using AI with streaming support\n */\nexport const generateYamlTestStream = async (\n  events: ChromeRecordedEvent[],\n  options: YamlGenerationOptions & StreamingCodeGenerationOptions = {},\n): Promise<StreamingAIResponse> => {\n  try {\n    // Validate input\n    validateEvents(events);\n\n    // Prepare event summary using shared utilities\n    const summary = prepareEventSummary(events, {\n      testName: options.testName,\n      maxScreenshots: options.maxScreenshots || 3,\n    });\n\n    // Add YAML-specific options to summary\n    const yamlSummary = {\n      ...summary,\n      includeTimestamps: options.includeTimestamps || false,\n    };\n\n    // Get screenshots for visual context\n    const screenshots = getScreenshotsForLLM(\n      events,\n      options.maxScreenshots || 3,\n    );\n\n    // Use LLM to generate the YAML test configuration\n    const prompt: ChatCompletionMessageParam[] = [\n      {\n        role: 'system',\n        content: `You are an expert in Midscene.js YAML test generation. Generate clean, accurate YAML following these rules: ${YAML_EXAMPLE_CODE}`,\n      },\n      {\n        role: 'user',\n        content: `Generate YAML test for Midscene.js automation from recorded browser events.\n\nEvent Summary:\n${JSON.stringify(yamlSummary, null, 2)}\n\nConvert events:\n- navigation → target.url\n- click → aiTap with element description\n- input → aiInput with value and locate\n- scroll → aiScroll with appropriate direction\n- Add aiAssert for important state changes\n\nRespond with YAML only, no explanations.`,\n      },\n    ];\n\n    // Add screenshots if available and requested\n    if (screenshots.length > 0) {\n      prompt.push({\n        role: 'user',\n        content:\n          'Here are screenshots from the recording session to help you understand the context:',\n      });\n\n      prompt.push({\n        role: 'user',\n        content: screenshots.map((screenshot) => ({\n          type: 'image_url',\n          image_url: {\n            url: screenshot,\n          },\n        })),\n      });\n    }\n\n    if (options.stream && options.onChunk) {\n      // Use streaming\n      return await callAi(\n        prompt,\n        AIActionType.EXTRACT_DATA,\n        {\n          intent: 'default',\n        },\n        {\n          stream: true,\n          onChunk: options.onChunk,\n        },\n      );\n    } else {\n      // Fallback to non-streaming\n      const response = await callAi(prompt, AIActionType.EXTRACT_DATA, {\n        intent: 'default',\n      });\n\n      if (response?.content && typeof response.content === 'string') {\n        return {\n          content: response.content,\n          usage: response.usage,\n          isStreamed: false,\n        };\n      }\n\n      throw new Error('Failed to generate YAML test configuration');\n    }\n  } catch (error) {\n    throw new Error(`Failed to generate YAML test: ${error}`);\n  }\n};\n"],"names":["getScreenshotsForLLM","events","maxScreenshots","eventsWithScreenshots","event","sortedEvents","a","b","screenshots","screenshot","filterEventsByType","createEventCounts","filteredEvents","totalEvents","extractInputDescriptions","inputEvents","item","processEventsForLLM","prepareEventSummary","options","eventCounts","startUrl","clickDescriptions","desc","Boolean","inputDescriptions","urls","e","url","processedEvents","createMessageContent","promptText","includeScreenshots","messageContent","validateEvents","Error","generateYamlTest","summary","yamlSummary","prompt","YAML_EXAMPLE_CODE","JSON","response","callAi","AIActionType","error","generateYamlTestStream"],"mappings":";;AAgFO,MAAMA,uBAAuB,CAClCC,QACAC,iBAAiB,CAAC;IAGlB,MAAMC,wBAAwBF,OAAO,MAAM,CACzC,CAACG,QACCA,MAAM,gBAAgB,IACtBA,MAAM,eAAe,IACrBA,MAAM,iBAAiB;IAI3B,MAAMC,eAAe;WAAIF;KAAsB,CAAC,IAAI,CAAC,CAACG,GAAGC;QACvD,IAAID,AAAW,iBAAXA,EAAE,IAAI,IAAqBC,AAAW,iBAAXA,EAAE,IAAI,EAAmB,OAAO;QAC/D,IAAID,AAAW,iBAAXA,EAAE,IAAI,IAAqBC,AAAW,iBAAXA,EAAE,IAAI,EAAmB,OAAO;QAC/D,IAAID,AAAW,YAAXA,EAAE,IAAI,IAAgBC,AAAW,YAAXA,EAAE,IAAI,EAAc,OAAO;QACrD,IAAID,AAAW,YAAXA,EAAE,IAAI,IAAgBC,AAAW,YAAXA,EAAE,IAAI,EAAc,OAAO;QACrD,OAAO;IACT;IAGA,MAAMC,cAAwB,EAAE;IAChC,KAAK,MAAMJ,SAASC,aAAc;QAEhC,MAAMI,aACJL,MAAM,iBAAiB,IACvBA,MAAM,eAAe,IACrBA,MAAM,gBAAgB;QACxB,IAAIK,cAAc,CAACD,YAAY,QAAQ,CAACC,aAAa;YACnDD,YAAY,IAAI,CAACC;YACjB,IAAID,YAAY,MAAM,IAAIN,gBAAgB;QAC5C;IACF;IAEA,OAAOM;AACT;AAKO,MAAME,qBAAqB,CAChCT,SAEO;QACL,kBAAkBA,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,iBAAfA,MAAM,IAAI;QACrD,aAAaH,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,YAAfA,MAAM,IAAI;QAChD,aAAaH,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,YAAfA,MAAM,IAAI;QAChD,cAAcH,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,aAAfA,MAAM,IAAI;IACnD;AAMK,MAAMO,oBAAoB,CAC/BC,gBACAC,cAEO;QACL,YAAYD,eAAe,gBAAgB,CAAC,MAAM;QAClD,OAAOA,eAAe,WAAW,CAAC,MAAM;QACxC,OAAOA,eAAe,WAAW,CAAC,MAAM;QACxC,QAAQA,eAAe,YAAY,CAAC,MAAM;QAC1C,OAAOC;IACT;AAMK,MAAMC,2BAA2B,CACtCC,cAEOA,YACJ,GAAG,CAAC,CAACX,QAAW;YACf,aAAaA,MAAM,kBAAkB,IAAI;YACzC,OAAOA,MAAM,KAAK,IAAI;QACxB,IACC,MAAM,CAAC,CAACY,OAASA,KAAK,WAAW,IAAIA,KAAK,KAAK;AAM7C,MAAMC,sBAAsB,CACjChB,SAEOA,OAAO,GAAG,CAAC,CAACG,QAAW;YAC5B,MAAMA,MAAM,IAAI;YAChB,WAAWA,MAAM,SAAS;YAC1B,KAAKA,MAAM,GAAG;YACd,OAAOA,MAAM,KAAK;YAClB,oBAAoBA,MAAM,kBAAkB;YAC5C,OAAOA,MAAM,KAAK;YAClB,UAAUA,MAAM,QAAQ;YACxB,aAAaA,MAAM,WAAW;QAChC;AAMK,MAAMc,sBAAsB,CACjCjB,QACAkB,UAA0D,CAAC,CAAC;IAE5D,MAAMP,iBAAiBF,mBAAmBT;IAC1C,MAAMmB,cAAcT,kBAAkBC,gBAAgBX,OAAO,MAAM;IAGnE,MAAMoB,WACJT,eAAe,gBAAgB,CAAC,MAAM,GAAG,IACrCA,eAAe,gBAAgB,CAAC,EAAE,CAAC,GAAG,IAAI,KAC1C;IAEN,MAAMU,oBAAoBV,eAAe,WAAW,CACjD,GAAG,CAAC,CAACR,QAAUA,MAAM,kBAAkB,EACvC,MAAM,CAAC,CAACmB,OAAyBC,QAAQD,OACzC,KAAK,CAAC,GAAG;IAEZ,MAAME,oBAAoBX,yBACxBF,eAAe,WAAW,EAC1B,KAAK,CAAC,GAAG;IAEX,MAAMc,OAAOd,eAAe,gBAAgB,CACzC,GAAG,CAAC,CAACe,IAAMA,EAAE,GAAG,EAChB,MAAM,CAAC,CAACC,MAAuBJ,QAAQI,MACvC,KAAK,CAAC,GAAG;IAEZ,MAAMC,kBAAkBZ,oBAAoBhB;IAE5C,OAAO;QACL,UAAUkB,QAAQ,QAAQ,IAAI;QAC9BE;QACAD;QACAM;QACAJ;QACAG;QACA,QAAQI;IACV;AACF;AAKO,MAAMC,uBAAuB,CAClCC,YACAvB,cAAwB,EAAE,EAC1BwB,qBAAqB,IAAI;IAEzB,MAAMC,iBAAwB;QAC5B;YACE,MAAM;YACN,MAAMF;QACR;KACD;IAGD,IAAIC,sBAAsBxB,YAAY,MAAM,GAAG,GAAG;QAChDyB,eAAe,OAAO,CAAC;YACrB,MAAM;YACN,MAAM;QACR;QAEAzB,YAAY,OAAO,CAAC,CAACC;YACnBwB,eAAe,IAAI,CAAC;gBAClB,MAAM;gBACN,WAAW;oBACT,KAAKxB;gBACP;YACF;QACF;IACF;IAEA,OAAOwB;AACT;AAKO,MAAMC,iBAAiB,CAACjC;IAC7B,IAAI,CAACA,OAAO,MAAM,EAChB,MAAM,IAAIkC,MAAM;AAEpB;AAOO,MAAMC,mBAAmB,OAC9BnC,QACAkB,UAAiC,CAAC,CAAC;IAEnC,IAAI;QAEFe,eAAejC;QAGf,MAAMoC,UAAUnB,oBAAoBjB,QAAQ;YAC1C,UAAUkB,QAAQ,QAAQ;YAC1B,gBAAgBA,QAAQ,cAAc,IAAI;QAC5C;QAGA,MAAMmB,cAAc;YAClB,GAAGD,OAAO;YACV,mBAAmBlB,QAAQ,iBAAiB,IAAI;QAClD;QAGA,MAAMX,cAAcR,qBAClBC,QACAkB,QAAQ,cAAc,IAAI;QAI5B,MAAMoB,SAAuC;YAC3C;gBACE,MAAM;gBACN,SAAS,CAAC,4GAA4G,EAAEC,mBAAmB;YAC7I;YACA;gBACE,MAAM;gBACN,SAAS,CAAC;;;AAGlB,EAAEC,KAAK,SAAS,CAACH,aAAa,MAAM,GAAG;;;;;;;;;wCASC,CAAC;YACnC;SACD;QAGD,IAAI9B,YAAY,MAAM,GAAG,GAAG;YAC1B+B,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SACE;YACJ;YAEAA,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SAAS/B,YAAY,GAAG,CAAC,CAACC,aAAgB;wBACxC,MAAM;wBACN,WAAW;4BACT,KAAKA;wBACP;oBACF;YACF;QACF;QAEA,MAAMiC,WAAW,MAAMC,OAAOJ,QAAQK,aAAa,YAAY,EAAE;YAC/D,QAAQ;QACV;QAEA,IAAIF,AAAAA,CAAAA,QAAAA,WAAAA,KAAAA,IAAAA,SAAU,OAAO,AAAD,KAAK,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAOA,SAAS,OAAO;QAGzB,MAAM,IAAIP,MAAM;IAClB,EAAE,OAAOU,OAAO;QACd,MAAM,IAAIV,MAAM,CAAC,8BAA8B,EAAEU,OAAO;IAC1D;AACF;AAKO,MAAMC,yBAAyB,OACpC7C,QACAkB,UAAkE,CAAC,CAAC;IAEpE,IAAI;QAEFe,eAAejC;QAGf,MAAMoC,UAAUnB,oBAAoBjB,QAAQ;YAC1C,UAAUkB,QAAQ,QAAQ;YAC1B,gBAAgBA,QAAQ,cAAc,IAAI;QAC5C;QAGA,MAAMmB,cAAc;YAClB,GAAGD,OAAO;YACV,mBAAmBlB,QAAQ,iBAAiB,IAAI;QAClD;QAGA,MAAMX,cAAcR,qBAClBC,QACAkB,QAAQ,cAAc,IAAI;QAI5B,MAAMoB,SAAuC;YAC3C;gBACE,MAAM;gBACN,SAAS,CAAC,4GAA4G,EAAEC,mBAAmB;YAC7I;YACA;gBACE,MAAM;gBACN,SAAS,CAAC;;;AAGlB,EAAEC,KAAK,SAAS,CAACH,aAAa,MAAM,GAAG;;;;;;;;;wCASC,CAAC;YACnC;SACD;QAGD,IAAI9B,YAAY,MAAM,GAAG,GAAG;YAC1B+B,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SACE;YACJ;YAEAA,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SAAS/B,YAAY,GAAG,CAAC,CAACC,aAAgB;wBACxC,MAAM;wBACN,WAAW;4BACT,KAAKA;wBACP;oBACF;YACF;QACF;QAEA,IAAIU,QAAQ,MAAM,IAAIA,QAAQ,OAAO,EAEnC,OAAO,MAAMwB,OACXJ,QACAK,aAAa,YAAY,EACzB;YACE,QAAQ;QACV,GACA;YACE,QAAQ;YACR,SAASzB,QAAQ,OAAO;QAC1B;QAEG;YAEL,MAAMuB,WAAW,MAAMC,OAAOJ,QAAQK,aAAa,YAAY,EAAE;gBAC/D,QAAQ;YACV;YAEA,IAAIF,AAAAA,CAAAA,QAAAA,WAAAA,KAAAA,IAAAA,SAAU,OAAO,AAAD,KAAK,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAO;gBACL,SAASA,SAAS,OAAO;gBACzB,OAAOA,SAAS,KAAK;gBACrB,YAAY;YACd;YAGF,MAAM,IAAIP,MAAM;QAClB;IACF,EAAE,OAAOU,OAAO;QACd,MAAM,IAAIV,MAAM,CAAC,8BAA8B,EAAEU,OAAO;IAC1D;AACF"}

package/dist/es/ai-model/service-caller/index.mjs

@@ -1,9 +1,9 @@
 import { AIResponseFormat } from "../../types.mjs";
 import { Anthropic } from "@anthropic-ai/sdk";
 import { DefaultAzureCredential, getBearerTokenProvider } from "@azure/identity";
-import { ANTHROPIC_API_KEY, AZURE_OPENAI_API_VERSION, AZURE_OPENAI_DEPLOYMENT, AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_KEY, MIDSCENE_API_TYPE, MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON, MIDSCENE_AZURE_OPENAI_SCOPE, MIDSCENE_DEBUG_AI_PROFILE, MIDSCENE_DEBUG_AI_RESPONSE, MIDSCENE_LANGSMITH_DEBUG, MIDSCENE_MODEL_NAME, MIDSCENE_OPENAI_HTTP_PROXY, MIDSCENE_OPENAI_INIT_CONFIG_JSON, MIDSCENE_OPENAI_SOCKS_PROXY, MIDSCENE_USE_ANTHROPIC_SDK, MIDSCENE_USE_AZURE_OPENAI, OPENAI_API_KEY, OPENAI_BASE_URL, OPENAI_MAX_TOKENS, OPENAI_USE_AZURE, getAIConfig, getAIConfigInBoolean, getAIConfigInJson, uiTarsModelVersion, vlLocateMode } from "@midscene/shared/env";
+import { MIDSCENE_API_TYPE, MIDSCENE_LANGSMITH_DEBUG, OPENAI_MAX_TOKENS, decideModelConfig, getAIConfig, getAIConfigInBoolean, uiTarsModelVersion, vlLocateMode } from "@midscene/shared/env";
 import { parseBase64 } from "@midscene/shared/img";
-import { enableDebug, getDebug } from "@midscene/shared/logger";
+import { getDebug } from "@midscene/shared/logger";
 import { assert, ifInBrowser } from "@midscene/shared/utils";
 import { HttpsProxyAgent } from "https-proxy-agent";
 import { jsonrepair } from "jsonrepair";
@@ -13,46 +13,9 @@ import { AIActionType } from "../common.mjs";
 import { assertSchema } from "../prompt/assertion.mjs";
 import { locatorSchema } from "../prompt/llm-locator.mjs";
 import { planSchema } from "../prompt/llm-planning.mjs";
-function checkAIConfig() {
-    const openaiKey = getAIConfig(OPENAI_API_KEY);
-    const azureConfig = getAIConfig(MIDSCENE_USE_AZURE_OPENAI);
-    const anthropicKey = getAIConfig(ANTHROPIC_API_KEY);
-    const initConfigJson = getAIConfig(MIDSCENE_OPENAI_INIT_CONFIG_JSON);
-    if (openaiKey) return true;
-    if (azureConfig) return true;
-    if (anthropicKey) return true;
-    return Boolean(initConfigJson);
-}
-let debugConfigInitialized = false;
-function initDebugConfig() {
-    if (debugConfigInitialized) return;
-    const shouldPrintTiming = getAIConfigInBoolean(MIDSCENE_DEBUG_AI_PROFILE);
-    let debugConfig = '';
-    if (shouldPrintTiming) {
-        console.warn('MIDSCENE_DEBUG_AI_PROFILE is deprecated, use DEBUG=midscene:ai:profile instead');
-        debugConfig = 'ai:profile';
-    }
-    const shouldPrintAIResponse = getAIConfigInBoolean(MIDSCENE_DEBUG_AI_RESPONSE);
-    if (shouldPrintAIResponse) {
-        console.warn('MIDSCENE_DEBUG_AI_RESPONSE is deprecated, use DEBUG=midscene:ai:response instead');
-        debugConfig = debugConfig ? 'ai:*' : 'ai:call';
-    }
-    if (debugConfig) enableDebug(debugConfig);
-    debugConfigInitialized = true;
-}
-const defaultModel = 'gpt-4o';
-function getModelName() {
-    let modelName = defaultModel;
-    const nameInConfig = getAIConfig(MIDSCENE_MODEL_NAME);
-    if (nameInConfig) modelName = nameInConfig;
-    return modelName;
-}
-async function createChatClient({ AIActionTypeValue }) {
-    initDebugConfig();
+async function createChatClient({ AIActionTypeValue, modelPreferences }) {
+    const { socksProxy, httpProxy, modelName, openaiBaseURL, openaiApiKey, openaiExtraConfig, openaiUseAzureDeprecated, useAzureOpenai, azureOpenaiScope, azureOpenaiKey, azureOpenaiEndpoint, azureOpenaiApiVersion, azureOpenaiDeployment, azureExtraConfig, useAnthropicSdk, anthropicApiKey } = decideModelConfig(modelPreferences, true);
     let openai;
-    const extraConfig = getAIConfigInJson(MIDSCENE_OPENAI_INIT_CONFIG_JSON);
-    const socksProxy = getAIConfig(MIDSCENE_OPENAI_SOCKS_PROXY);
-    const httpProxy = getAIConfig(MIDSCENE_OPENAI_HTTP_PROXY);
     let proxyAgent;
     const debugProxy = getDebug('ai:call:proxy');
     if (httpProxy) {
@@ -62,56 +25,47 @@ async function createChatClient({ AIActionTypeValue }) {
         debugProxy('using socks proxy', socksProxy);
         proxyAgent = new SocksProxyAgent(socksProxy);
     }
-    if (getAIConfig(OPENAI_USE_AZURE)) openai = new AzureOpenAI({
-        baseURL: getAIConfig(OPENAI_BASE_URL),
-        apiKey: getAIConfig(OPENAI_API_KEY),
+    if (openaiUseAzureDeprecated) openai = new AzureOpenAI({
+        baseURL: openaiBaseURL,
+        apiKey: openaiApiKey,
         httpAgent: proxyAgent,
-        ...extraConfig,
+        ...openaiExtraConfig,
         dangerouslyAllowBrowser: true
     });
-    else if (getAIConfig(MIDSCENE_USE_AZURE_OPENAI)) {
-        const extraAzureConfig = getAIConfigInJson(MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON);
-        const scope = getAIConfig(MIDSCENE_AZURE_OPENAI_SCOPE);
+    else if (useAzureOpenai) {
         let tokenProvider;
-        if (scope) {
+        if (azureOpenaiScope) {
             assert(!ifInBrowser, 'Azure OpenAI is not supported in browser with Midscene.');
             const credential = new DefaultAzureCredential();
-            assert(scope, 'MIDSCENE_AZURE_OPENAI_SCOPE is required');
-            tokenProvider = getBearerTokenProvider(credential, scope);
+            tokenProvider = getBearerTokenProvider(credential, azureOpenaiScope);
             openai = new AzureOpenAI({
                 azureADTokenProvider: tokenProvider,
-                endpoint: getAIConfig(AZURE_OPENAI_ENDPOINT),
-                apiVersion: getAIConfig(AZURE_OPENAI_API_VERSION),
-                deployment: getAIConfig(AZURE_OPENAI_DEPLOYMENT),
-                ...extraConfig,
-                ...extraAzureConfig
+                endpoint: azureOpenaiEndpoint,
+                apiVersion: azureOpenaiApiVersion,
+                deployment: azureOpenaiDeployment,
+                ...openaiExtraConfig,
+                ...azureExtraConfig
             });
         } else openai = new AzureOpenAI({
-            apiKey: getAIConfig(AZURE_OPENAI_KEY),
-            endpoint: getAIConfig(AZURE_OPENAI_ENDPOINT),
-            apiVersion: getAIConfig(AZURE_OPENAI_API_VERSION),
-            deployment: getAIConfig(AZURE_OPENAI_DEPLOYMENT),
+            apiKey: azureOpenaiKey,
+            endpoint: azureOpenaiEndpoint,
+            apiVersion: azureOpenaiApiVersion,
+            deployment: azureOpenaiDeployment,
             dangerouslyAllowBrowser: true,
-            ...extraConfig,
-            ...extraAzureConfig
-        });
-    } else if (!getAIConfig(MIDSCENE_USE_ANTHROPIC_SDK)) {
-        const baseURL = getAIConfig(OPENAI_BASE_URL);
-        if ('string' == typeof baseURL) {
-            if (!/^https?:\/\//.test(baseURL)) throw new Error(`OPENAI_BASE_URL must be a valid URL starting with http:// or https://, but got: ${baseURL}\nPlease check your config.`);
-        }
-        openai = new openai_0({
-            baseURL: getAIConfig(OPENAI_BASE_URL),
-            apiKey: getAIConfig(OPENAI_API_KEY),
-            httpAgent: proxyAgent,
-            ...extraConfig,
-            defaultHeaders: {
-                ...(null == extraConfig ? void 0 : extraConfig.defaultHeaders) || {},
-                [MIDSCENE_API_TYPE]: AIActionTypeValue.toString()
-            },
-            dangerouslyAllowBrowser: true
+            ...openaiExtraConfig,
+            ...azureExtraConfig
         });
-    }
+    } else if (!useAnthropicSdk) openai = new openai_0({
+        baseURL: openaiBaseURL,
+        apiKey: openaiApiKey,
+        httpAgent: proxyAgent,
+        ...openaiExtraConfig,
+        defaultHeaders: {
+            ...(null == openaiExtraConfig ? void 0 : openaiExtraConfig.defaultHeaders) || {},
+            [MIDSCENE_API_TYPE]: AIActionTypeValue.toString()
+        },
+        dangerouslyAllowBrowser: true
+    });
     if (openai && getAIConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)) {
         if (ifInBrowser) throw new Error('langsmith is not supported in browser');
         console.log('DEBUGGING MODE: langsmith wrapper enabled');
@@ -120,53 +74,51 @@ async function createChatClient({ AIActionTypeValue }) {
     }
     if (void 0 !== openai) return {
         completion: openai.chat.completions,
-        style: 'openai'
+        style: 'openai',
+        modelName
     };
-    if (getAIConfig(MIDSCENE_USE_ANTHROPIC_SDK)) {
-        const apiKey = getAIConfig(ANTHROPIC_API_KEY);
-        assert(apiKey, 'ANTHROPIC_API_KEY is required');
-        openai = new Anthropic({
-            apiKey,
-            httpAgent: proxyAgent,
-            dangerouslyAllowBrowser: true
-        });
-    }
+    if (useAnthropicSdk) openai = new Anthropic({
+        apiKey: anthropicApiKey,
+        httpAgent: proxyAgent,
+        dangerouslyAllowBrowser: true
+    });
     if (void 0 !== openai && openai.messages) return {
         completion: openai.messages,
-        style: 'anthropic'
+        style: 'anthropic',
+        modelName
     };
     throw new Error('Openai SDK or Anthropic SDK is not initialized');
 }
-async function call(messages, AIActionTypeValue, responseFormat, options) {
-    assert(checkAIConfig(), 'Cannot find config for AI model service. If you are using a self-hosted model without validating the API key, please set `OPENAI_API_KEY` to any non-null value. https://midscenejs.com/model-provider.html');
-    const { completion, style } = await createChatClient({
-        AIActionTypeValue
+async function call(messages, AIActionTypeValue, modelPreferences, options) {
+    const { completion, style, modelName } = await createChatClient({
+        AIActionTypeValue,
+        modelPreferences
     });
+    const responseFormat = getResponseFormat(modelName, AIActionTypeValue);
     const maxTokens = getAIConfig(OPENAI_MAX_TOKENS);
     const debugCall = getDebug('ai:call');
     const debugProfileStats = getDebug('ai:profile:stats');
     const debugProfileDetail = getDebug('ai:profile:detail');
     const startTime = Date.now();
-    const model = getModelName();
     const isStreaming = (null == options ? void 0 : options.stream) && (null == options ? void 0 : options.onChunk);
     let content;
     let accumulated = '';
     let usage;
     let timeCost;
     const commonConfig = {
-        temperature: 'vlm-ui-tars' === vlLocateMode() ? 0.0 : 0.1,
+        temperature: 'vlm-ui-tars' === vlLocateMode(modelPreferences) ? 0.0 : 0.1,
         stream: !!isStreaming,
         max_tokens: 'number' == typeof maxTokens ? maxTokens : Number.parseInt(maxTokens || '2048', 10),
-        ...'qwen-vl' === vlLocateMode() ? {
+        ...'qwen-vl' === vlLocateMode(modelPreferences) ? {
             vl_high_resolution_images: true
         } : {}
     };
     try {
         if ('openai' === style) {
-            debugCall(`sending ${isStreaming ? 'streaming ' : ''}request to ${model}`);
+            debugCall(`sending ${isStreaming ? 'streaming ' : ''}request to ${modelName}`);
             if (isStreaming) {
                 const stream = await completion.create({
-                    model,
+                    model: modelName,
                     messages,
                     response_format: responseFormat,
                     ...commonConfig
@@ -208,7 +160,8 @@ async function call(messages, AIActionTypeValue, responseFormat, options) {
                                 prompt_tokens: usage.prompt_tokens ?? 0,
                                 completion_tokens: usage.completion_tokens ?? 0,
                                 total_tokens: usage.total_tokens ?? 0,
-                                time_cost: timeCost ?? 0
+                                time_cost: timeCost ?? 0,
+                                model_name: modelName
                             }
                         };
                         options.onChunk(finalChunk);
@@ -216,17 +169,17 @@ async function call(messages, AIActionTypeValue, responseFormat, options) {
                     }
                 }
                 content = accumulated;
-                debugProfileStats(`streaming model, ${model}, mode, ${vlLocateMode() || 'default'}, cost-ms, ${timeCost}`);
+                debugProfileStats(`streaming model, ${modelName}, mode, ${vlLocateMode(modelPreferences) || 'default'}, cost-ms, ${timeCost}`);
             } else {
                 var _result_usage, _result_usage1, _result_usage2;
                 const result = await completion.create({
-                    model,
+                    model: modelName,
                     messages,
                     response_format: responseFormat,
                     ...commonConfig
                 });
                 timeCost = Date.now() - startTime;
-                debugProfileStats(`model, ${model}, mode, ${vlLocateMode() || 'default'}, ui-tars-version, ${uiTarsModelVersion()}, prompt-tokens, ${(null == (_result_usage = result.usage) ? void 0 : _result_usage.prompt_tokens) || ''}, completion-tokens, ${(null == (_result_usage1 = result.usage) ? void 0 : _result_usage1.completion_tokens) || ''}, total-tokens, ${(null == (_result_usage2 = result.usage) ? void 0 : _result_usage2.total_tokens) || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}`);
+                debugProfileStats(`model, ${modelName}, mode, ${vlLocateMode(modelPreferences) || 'default'}, ui-tars-version, ${uiTarsModelVersion(modelPreferences)}, prompt-tokens, ${(null == (_result_usage = result.usage) ? void 0 : _result_usage.prompt_tokens) || ''}, completion-tokens, ${(null == (_result_usage1 = result.usage) ? void 0 : _result_usage1.completion_tokens) || ''}, total-tokens, ${(null == (_result_usage2 = result.usage) ? void 0 : _result_usage2.total_tokens) || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}`);
                 debugProfileDetail(`model usage detail: ${JSON.stringify(result.usage)}`);
                 assert(result.choices, `invalid response from LLM service: ${JSON.stringify(result)}`);
                 content = result.choices[0].message.content;
@@ -253,7 +206,7 @@ async function call(messages, AIActionTypeValue, responseFormat, options) {
             };
             if (isStreaming) {
                 const stream = await completion.create({
-                    model,
+                    model: modelName,
                     system: 'You are a versatile professional in software UI automation',
                     messages: messages.map((m)=>({
                             role: 'user',
@@ -288,7 +241,8 @@ async function call(messages, AIActionTypeValue, responseFormat, options) {
                                 prompt_tokens: anthropicUsage.input_tokens ?? 0,
                                 completion_tokens: anthropicUsage.output_tokens ?? 0,
                                 total_tokens: (anthropicUsage.input_tokens ?? 0) + (anthropicUsage.output_tokens ?? 0),
-                                time_cost: timeCost ?? 0
+                                time_cost: timeCost ?? 0,
+                                model_name: modelName
                             } : void 0
                         };
                         options.onChunk(finalChunk);
@@ -298,7 +252,7 @@ async function call(messages, AIActionTypeValue, responseFormat, options) {
                 content = accumulated;
             } else {
                 const result = await completion.create({
-                    model,
+                    model: modelName,
                     system: 'You are a versatile professional in software UI automation',
                     messages: messages.map((m)=>({
                             role: 'user',
@@ -327,7 +281,8 @@ async function call(messages, AIActionTypeValue, responseFormat, options) {
                 prompt_tokens: usage.prompt_tokens ?? 0,
                 completion_tokens: usage.completion_tokens ?? 0,
                 total_tokens: usage.total_tokens ?? 0,
-                time_cost: timeCost ?? 0
+                time_cost: timeCost ?? 0,
+                model_name: modelName
             } : void 0,
             isStreamed: !!isStreaming
         };
@@ -339,10 +294,9 @@ async function call(messages, AIActionTypeValue, responseFormat, options) {
         throw newError;
     }
 }
-async function callToGetJSONObject(messages, AIActionTypeValue) {
+const getResponseFormat = (modelName, AIActionTypeValue)=>{
     let responseFormat;
-    const model = getModelName();
-    if (model.includes('gpt-4')) switch(AIActionTypeValue){
+    if (modelName.includes('gpt-4')) switch(AIActionTypeValue){
         case AIActionType.ASSERT:
             responseFormat = assertSchema;
             break;
@@ -359,19 +313,22 @@ async function callToGetJSONObject(messages, AIActionTypeValue) {
             };
             break;
     }
-    if ('gpt-4o-2024-05-13' === model) responseFormat = {
+    if ('gpt-4o-2024-05-13' === modelName) responseFormat = {
         type: AIResponseFormat.JSON
     };
-    const response = await call(messages, AIActionTypeValue, responseFormat);
+    return responseFormat;
+};
+async function callToGetJSONObject(messages, AIActionTypeValue, modelPreferences) {
+    const response = await call(messages, AIActionTypeValue, modelPreferences);
     assert(response, 'empty response');
-    const jsonContent = safeParseJson(response.content);
+    const jsonContent = safeParseJson(response.content, modelPreferences);
     return {
         content: jsonContent,
         usage: response.usage
     };
 }
-async function callAiFnWithStringResponse(msgs, AIActionTypeValue) {
-    const { content, usage } = await call(msgs, AIActionTypeValue);
+async function callAiFnWithStringResponse(msgs, AIActionTypeValue, modelPreferences) {
+    const { content, usage } = await call(msgs, AIActionTypeValue, modelPreferences);
     return {
         content,
         usage
@@ -392,7 +349,7 @@ function preprocessDoubaoBboxJson(input) {
     if (input.includes('bbox')) while(/\d+\s+\d+/.test(input))input = input.replace(/(\d+)\s+(\d+)/g, '$1,$2');
     return input;
 }
-function safeParseJson(input) {
+function safeParseJson(input, modelPreferences) {
     const cleanJsonString = extractJSONFromCodeBlock(input);
     if (null == cleanJsonString ? void 0 : cleanJsonString.match(/\((\d+),(\d+)\)/)) {
         var _cleanJsonString_match;
@@ -404,12 +361,12 @@ function safeParseJson(input) {
     try {
         return JSON.parse(jsonrepair(cleanJsonString));
     } catch (e) {}
-    if ('doubao-vision' === vlLocateMode() || 'vlm-ui-tars' === vlLocateMode()) {
+    if ('doubao-vision' === vlLocateMode(modelPreferences) || 'vlm-ui-tars' === vlLocateMode(modelPreferences)) {
         const jsonString = preprocessDoubaoBboxJson(cleanJsonString);
         return JSON.parse(jsonrepair(jsonString));
     }
     throw Error(`failed to parse json response: ${input}`);
 }
-export { call, callAiFnWithStringResponse, callToGetJSONObject, checkAIConfig, extractJSONFromCodeBlock, getModelName, preprocessDoubaoBboxJson, safeParseJson };
+export { call, callAiFnWithStringResponse, callToGetJSONObject, extractJSONFromCodeBlock, getResponseFormat, preprocessDoubaoBboxJson, safeParseJson };
 
 //# sourceMappingURL=index.mjs.map