@midscene/core 0.26.7-beta-20250818081955.0 → 0.26.7-beta-20250820150415.0

package/dist/lib/ai-model/inspect.js

@@ -87,20 +87,23 @@ const promptsToChatParam = async (multimodalPrompt)=>{
 async function AiLocateElement(options) {
     const { context, targetElementDescription, callAI } = options;
     const { screenshotBase64 } = context;
-    const { description, elementById, insertElementByPosition } = await (0, util_js_namespaceObject.describeUserPage)(context);
+    const modelPreferences = {
+        intent: 'grounding'
+    };
+    const { description, elementById, insertElementByPosition } = await (0, util_js_namespaceObject.describeUserPage)(context, modelPreferences);
     (0, utils_namespaceObject.assert)(targetElementDescription, "cannot find the target element description");
     const userInstructionPrompt = await llm_locator_js_namespaceObject.findElementPrompt.format({
         pageDescription: description,
         targetElementDescription: extraTextFromUserPrompt(targetElementDescription)
     });
-    const systemPrompt = (0, llm_locator_js_namespaceObject.systemPromptToLocateElement)((0, env_namespaceObject.vlLocateMode)());
+    const systemPrompt = (0, llm_locator_js_namespaceObject.systemPromptToLocateElement)((0, env_namespaceObject.vlLocateMode)(modelPreferences));
     let imagePayload = screenshotBase64;
     if (options.searchConfig) {
         (0, utils_namespaceObject.assert)(options.searchConfig.rect, 'searchArea is provided but its rect cannot be found. Failed to locate element');
         (0, utils_namespaceObject.assert)(options.searchConfig.imageBase64, 'searchArea is provided but its imageBase64 cannot be found. Failed to locate element');
         imagePayload = options.searchConfig.imageBase64;
-    } else if ('qwen-vl' === (0, env_namespaceObject.vlLocateMode)()) imagePayload = await (0, img_namespaceObject.paddingToMatchBlockByBase64)(imagePayload);
-    else if (!(0, env_namespaceObject.vlLocateMode)()) imagePayload = await (0, external_common_js_namespaceObject.markupImageForLLM)(screenshotBase64, context.tree, context.size);
+    } else if ('qwen-vl' === (0, env_namespaceObject.vlLocateMode)(modelPreferences)) imagePayload = await (0, img_namespaceObject.paddingToMatchBlockByBase64)(imagePayload);
+    else if (!(0, env_namespaceObject.vlLocateMode)(modelPreferences)) imagePayload = await (0, external_common_js_namespaceObject.markupImageForLLM)(screenshotBase64, context.tree, context.size);
     const msgs = [
         {
             role: 'system',
@@ -131,7 +134,9 @@ async function AiLocateElement(options) {
         msgs.push(...addOns);
     }
     const callAIFn = callAI || index_js_namespaceObject.callToGetJSONObject;
-    const res = await callAIFn(msgs, external_common_js_namespaceObject.AIActionType.INSPECT_ELEMENT);
+    const res = await callAIFn(msgs, external_common_js_namespaceObject.AIActionType.INSPECT_ELEMENT, {
+        intent: 'grounding'
+    });
     const rawResponse = JSON.stringify(res.content);
     let resRect;
     let matchedElements = 'elements' in res.content ? res.content.elements : [];
@@ -139,7 +144,7 @@ async function AiLocateElement(options) {
     try {
         if ('bbox' in res.content && Array.isArray(res.content.bbox)) {
             var _options_searchConfig_rect, _options_searchConfig, _options_searchConfig_rect1, _options_searchConfig1, _options_searchConfig_rect2, _options_searchConfig2, _options_searchConfig_rect3, _options_searchConfig3;
-            resRect = (0, external_common_js_namespaceObject.adaptBboxToRect)(res.content.bbox, (null == (_options_searchConfig = options.searchConfig) ? void 0 : null == (_options_searchConfig_rect = _options_searchConfig.rect) ? void 0 : _options_searchConfig_rect.width) || context.size.width, (null == (_options_searchConfig1 = options.searchConfig) ? void 0 : null == (_options_searchConfig_rect1 = _options_searchConfig1.rect) ? void 0 : _options_searchConfig_rect1.height) || context.size.height, null == (_options_searchConfig2 = options.searchConfig) ? void 0 : null == (_options_searchConfig_rect2 = _options_searchConfig2.rect) ? void 0 : _options_searchConfig_rect2.left, null == (_options_searchConfig3 = options.searchConfig) ? void 0 : null == (_options_searchConfig_rect3 = _options_searchConfig3.rect) ? void 0 : _options_searchConfig_rect3.top);
+            resRect = (0, external_common_js_namespaceObject.adaptBboxToRect)(res.content.bbox, (null == (_options_searchConfig = options.searchConfig) ? void 0 : null == (_options_searchConfig_rect = _options_searchConfig.rect) ? void 0 : _options_searchConfig_rect.width) || context.size.width, (null == (_options_searchConfig1 = options.searchConfig) ? void 0 : null == (_options_searchConfig_rect1 = _options_searchConfig1.rect) ? void 0 : _options_searchConfig_rect1.height) || context.size.height, modelPreferences, null == (_options_searchConfig2 = options.searchConfig) ? void 0 : null == (_options_searchConfig_rect2 = _options_searchConfig2.rect) ? void 0 : _options_searchConfig_rect2.left, null == (_options_searchConfig3 = options.searchConfig) ? void 0 : null == (_options_searchConfig_rect3 = _options_searchConfig3.rect) ? void 0 : _options_searchConfig_rect3.top);
             debugInspect('resRect', resRect);
             const rectCenter = {
                 x: resRect.left + resRect.width / 2,
@@ -180,7 +185,10 @@ async function AiLocateElement(options) {
 async function AiLocateSection(options) {
     const { context, sectionDescription } = options;
     const { screenshotBase64 } = context;
-    const systemPrompt = (0, llm_section_locator_js_namespaceObject.systemPromptToLocateSection)((0, env_namespaceObject.vlLocateMode)());
+    const modelPreferences = {
+        intent: 'grounding'
+    };
+    const systemPrompt = (0, llm_section_locator_js_namespaceObject.systemPromptToLocateSection)((0, env_namespaceObject.vlLocateMode)(modelPreferences));
     const sectionLocatorInstructionText = await llm_section_locator_js_namespaceObject.sectionLocatorInstruction.format({
         sectionDescription: extraTextFromUserPrompt(sectionDescription)
     });
@@ -213,26 +221,30 @@ async function AiLocateSection(options) {
         });
         msgs.push(...addOns);
     }
-    const result = await (0, external_common_js_namespaceObject.callAiFn)(msgs, external_common_js_namespaceObject.AIActionType.EXTRACT_DATA);
+    const result = await (0, external_common_js_namespaceObject.callAiFn)(msgs, external_common_js_namespaceObject.AIActionType.EXTRACT_DATA, {
+        intent: 'grounding'
+    });
     let sectionRect;
     const sectionBbox = result.content.bbox;
     if (sectionBbox) {
-        const targetRect = (0, external_common_js_namespaceObject.adaptBboxToRect)(sectionBbox, context.size.width, context.size.height);
+        const targetRect = (0, external_common_js_namespaceObject.adaptBboxToRect)(sectionBbox, context.size.width, context.size.height, modelPreferences);
         debugSection('original targetRect %j', targetRect);
         const referenceBboxList = result.content.references_bbox || [];
         debugSection('referenceBboxList %j', referenceBboxList);
-        const referenceRects = referenceBboxList.filter((bbox)=>Array.isArray(bbox)).map((bbox)=>(0, external_common_js_namespaceObject.adaptBboxToRect)(bbox, context.size.width, context.size.height));
+        const referenceRects = referenceBboxList.filter((bbox)=>Array.isArray(bbox)).map((bbox)=>(0, external_common_js_namespaceObject.adaptBboxToRect)(bbox, context.size.width, context.size.height, modelPreferences));
         debugSection('referenceRects %j', referenceRects);
         const mergedRect = (0, external_common_js_namespaceObject.mergeRects)([
             targetRect,
             ...referenceRects
         ]);
         debugSection('mergedRect %j', mergedRect);
-        sectionRect = (0, external_common_js_namespaceObject.expandSearchArea)(mergedRect, context.size);
+        sectionRect = (0, external_common_js_namespaceObject.expandSearchArea)(mergedRect, context.size, modelPreferences);
         debugSection('expanded sectionRect %j', sectionRect);
     }
     let imageBase64 = screenshotBase64;
-    if (sectionRect) imageBase64 = await (0, img_namespaceObject.cropByRect)(screenshotBase64, sectionRect, (0, env_namespaceObject.getAIConfigInBoolean)(env_namespaceObject.MIDSCENE_USE_QWEN_VL));
+    if (sectionRect) imageBase64 = await (0, img_namespaceObject.cropByRect)(screenshotBase64, sectionRect, (0, env_namespaceObject.getIsUseQwenVl)({
+        intent: 'grounding'
+    }));
     return {
         rect: sectionRect,
         imageBase64,
@@ -243,10 +255,10 @@ async function AiLocateSection(options) {
 }
 async function AiExtractElementInfo(options) {
     var _options_extractOption;
-    const { dataQuery, context, extractOption, multimodalPrompt } = options;
+    const { dataQuery, context, extractOption, multimodalPrompt, modelPreferences } = options;
     const systemPrompt = (0, extraction_js_namespaceObject.systemPromptToExtract)();
     const { screenshotBase64 } = context;
-    const { description, elementById } = await (0, util_js_namespaceObject.describeUserPage)(context, {
+    const { description, elementById } = await (0, util_js_namespaceObject.describeUserPage)(context, modelPreferences, {
         truncateTextLength: 200,
         filterNonTextContent: false,
         visibleOnly: false,
@@ -286,7 +298,7 @@ async function AiExtractElementInfo(options) {
         });
         msgs.push(...addOns);
     }
-    const result = await (0, external_common_js_namespaceObject.callAiFn)(msgs, external_common_js_namespaceObject.AIActionType.EXTRACT_DATA);
+    const result = await (0, external_common_js_namespaceObject.callAiFn)(msgs, external_common_js_namespaceObject.AIActionType.EXTRACT_DATA, modelPreferences);
     return {
         parseResult: result.content,
         elementById,

package/dist/lib/ai-model/inspect.js.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/inspect.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/inspect.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n  AIDataExtractionResponse,\n  AIElementLocatorResponse,\n  AIElementResponse,\n  AISectionLocatorResponse,\n  AIUsageInfo,\n  BaseElement,\n  ElementById,\n  InsightExtractOption,\n  Rect,\n  ReferenceImage,\n  TMultimodalPrompt,\n  TUserPrompt,\n  UIContext,\n} from '@/types';\nimport {\n  MIDSCENE_USE_QWEN_VL,\n  getAIConfigInBoolean,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport {\n  cropByRect,\n  paddingToMatchBlockByBase64,\n  preProcessImageUrl,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type {\n  ChatCompletionSystemMessageParam,\n  ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport {\n  AIActionType,\n  adaptBboxToRect,\n  callAiFn,\n  expandSearchArea,\n  markupImageForLLM,\n  mergeRects,\n} from './common';\nimport {\n  extractDataQueryPrompt,\n  systemPromptToExtract,\n} from './prompt/extraction';\nimport {\n  findElementPrompt,\n  systemPromptToLocateElement,\n} from './prompt/llm-locator';\nimport {\n  sectionLocatorInstruction,\n  systemPromptToLocateSection,\n} from './prompt/llm-section-locator';\nimport {\n  describeUserPage,\n  distance,\n  distanceThreshold,\n  elementByPositionWithElementInfo,\n} from './prompt/util';\nimport { callToGetJSONObject } from './service-caller/index';\n\nexport type AIArgs = [\n  ChatCompletionSystemMessageParam,\n  ...ChatCompletionUserMessageParam[],\n];\n\nconst debugInspect = getDebug('ai:inspect');\nconst debugSection = getDebug('ai:section');\n\nconst extraTextFromUserPrompt = (prompt: TUserPrompt): string => {\n  if (typeof prompt === 'string') {\n    return prompt;\n  } else {\n    return prompt.prompt;\n  }\n};\n\nconst promptsToChatParam = async (\n  multimodalPrompt: TMultimodalPrompt,\n): Promise<ChatCompletionUserMessageParam[]> => {\n  const msgs: ChatCompletionUserMessageParam[] = [];\n  if (multimodalPrompt?.images?.length) {\n    msgs.push({\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: 'Next, I will provide all the reference images.',\n        },\n      ],\n    });\n\n    for (const item of multimodalPrompt.images) {\n      const base64 = await preProcessImageUrl(\n        item.url,\n        !!multimodalPrompt.convertHttpImage2Base64,\n      );\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'text',\n            text: `reference image ${item.name}:`,\n          },\n        ],\n      });\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: base64,\n              detail: 'high',\n            },\n          },\n        ],\n      });\n    }\n  }\n  return msgs;\n};\n\nexport async function AiLocateElement<\n  ElementType extends BaseElement = BaseElement,\n>(options: {\n  context: UIContext<ElementType>;\n  targetElementDescription: TUserPrompt;\n  referenceImage?: ReferenceImage;\n  callAI?: typeof callAiFn<AIElementResponse | [number, number]>;\n  searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n}): Promise<{\n  parseResult: AIElementLocatorResponse;\n  rect?: Rect;\n  rawResponse: string;\n  elementById: ElementById;\n  usage?: AIUsageInfo;\n  isOrderSensitive?: boolean;\n}> {\n  const { context, targetElementDescription, callAI } = options;\n  const { screenshotBase64 } = context;\n  const { description, elementById, insertElementByPosition } =\n    await describeUserPage(context);\n\n  assert(\n    targetElementDescription,\n    'cannot find the target element description',\n  );\n\n  const userInstructionPrompt = await findElementPrompt.format({\n    pageDescription: description,\n    targetElementDescription: extraTextFromUserPrompt(targetElementDescription),\n  });\n  const systemPrompt = systemPromptToLocateElement(vlLocateMode());\n\n  let imagePayload = screenshotBase64;\n\n  if (options.searchConfig) {\n    assert(\n      options.searchConfig.rect,\n      'searchArea is provided but its rect cannot be found. Failed to locate element',\n    );\n    assert(\n      options.searchConfig.imageBase64,\n      'searchArea is provided but its imageBase64 cannot be found. Failed to locate element',\n    );\n\n    imagePayload = options.searchConfig.imageBase64;\n  } else if (vlLocateMode() === 'qwen-vl') {\n    imagePayload = await paddingToMatchBlockByBase64(imagePayload);\n  } else if (!vlLocateMode()) {\n    imagePayload = await markupImageForLLM(\n      screenshotBase64,\n      context.tree,\n      context.size,\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          type: 'text',\n          text: userInstructionPrompt,\n        },\n      ],\n    },\n  ];\n\n  if (typeof targetElementDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: targetElementDescription.images,\n      convertHttpImage2Base64: targetElementDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const callAIFn =\n    callAI || callToGetJSONObject<AIElementResponse | [number, number]>;\n\n  const res = await callAIFn(msgs, AIActionType.INSPECT_ELEMENT);\n\n  const rawResponse = JSON.stringify(res.content);\n\n  let resRect: Rect | undefined;\n  let matchedElements: AIElementLocatorResponse['elements'] =\n    'elements' in res.content ? res.content.elements : [];\n  let errors: AIElementLocatorResponse['errors'] | undefined =\n    'errors' in res.content ? res.content.errors : [];\n  try {\n    if ('bbox' in res.content && Array.isArray(res.content.bbox)) {\n      resRect = adaptBboxToRect(\n        res.content.bbox,\n        options.searchConfig?.rect?.width || context.size.width,\n        options.searchConfig?.rect?.height || context.size.height,\n        options.searchConfig?.rect?.left,\n        options.searchConfig?.rect?.top,\n      );\n      debugInspect('resRect', resRect);\n\n      const rectCenter = {\n        x: resRect.left + resRect.width / 2,\n        y: resRect.top + resRect.height / 2,\n      };\n      let element = elementByPositionWithElementInfo(context.tree, rectCenter);\n\n      const distanceToCenter = element\n        ? distance({ x: element.center[0], y: element.center[1] }, rectCenter)\n        : 0;\n\n      if (!element || distanceToCenter > distanceThreshold) {\n        element = insertElementByPosition(rectCenter);\n      }\n\n      if (element) {\n        matchedElements = [element];\n        errors = [];\n      }\n    }\n  } catch (e) {\n    const msg =\n      e instanceof Error\n        ? `Failed to parse bbox: ${e.message}`\n        : 'unknown error in locate';\n    if (!errors || errors?.length === 0) {\n      errors = [msg];\n    } else {\n      errors.push(`(${msg})`);\n    }\n  }\n\n  return {\n    rect: resRect,\n    parseResult: {\n      elements: matchedElements,\n      errors,\n    },\n    rawResponse,\n    elementById,\n    usage: res.usage,\n    isOrderSensitive:\n      typeof res.content === 'object' &&\n      res.content !== null &&\n      'isOrderSensitive' in res.content\n        ? (res.content as any).isOrderSensitive\n        : undefined,\n  };\n}\n\nexport async function AiLocateSection(options: {\n  context: UIContext<BaseElement>;\n  sectionDescription: TUserPrompt;\n  callAI?: typeof callAiFn<AISectionLocatorResponse>;\n}): Promise<{\n  rect?: Rect;\n  imageBase64?: string;\n  error?: string;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n}> {\n  const { context, sectionDescription } = options;\n  const { screenshotBase64 } = context;\n\n  const systemPrompt = systemPromptToLocateSection(vlLocateMode());\n  const sectionLocatorInstructionText = await sectionLocatorInstruction.format({\n    sectionDescription: extraTextFromUserPrompt(sectionDescription),\n  });\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: screenshotBase64,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: sectionLocatorInstructionText,\n        },\n      ],\n    },\n  ];\n\n  if (typeof sectionDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: sectionDescription.images,\n      convertHttpImage2Base64: sectionDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const result = await callAiFn<AISectionLocatorResponse>(\n    msgs,\n    AIActionType.EXTRACT_DATA,\n  );\n\n  let sectionRect: Rect | undefined;\n  const sectionBbox = result.content.bbox;\n  if (sectionBbox) {\n    const targetRect = adaptBboxToRect(\n      sectionBbox,\n      context.size.width,\n      context.size.height,\n    );\n    debugSection('original targetRect %j', targetRect);\n\n    const referenceBboxList = result.content.references_bbox || [];\n    debugSection('referenceBboxList %j', referenceBboxList);\n\n    const referenceRects = referenceBboxList\n      .filter((bbox) => Array.isArray(bbox))\n      .map((bbox) => {\n        return adaptBboxToRect(bbox, context.size.width, context.size.height);\n      });\n    debugSection('referenceRects %j', referenceRects);\n\n    // merge the sectionRect and referenceRects\n    const mergedRect = mergeRects([targetRect, ...referenceRects]);\n    debugSection('mergedRect %j', mergedRect);\n\n    // expand search area to at least 200 x 200\n    sectionRect = expandSearchArea(mergedRect, context.size);\n    debugSection('expanded sectionRect %j', sectionRect);\n  }\n\n  let imageBase64 = screenshotBase64;\n  if (sectionRect) {\n    imageBase64 = await cropByRect(\n      screenshotBase64,\n      sectionRect,\n      getAIConfigInBoolean(MIDSCENE_USE_QWEN_VL),\n    );\n  }\n\n  return {\n    rect: sectionRect,\n    imageBase64,\n    error: result.content.error,\n    rawResponse: JSON.stringify(result.content),\n    usage: result.usage,\n  };\n}\n\nexport async function AiExtractElementInfo<\n  T,\n  ElementType extends BaseElement = BaseElement,\n>(options: {\n  dataQuery: string | Record<string, string>;\n  multimodalPrompt?: TMultimodalPrompt;\n  context: UIContext<ElementType>;\n  extractOption?: InsightExtractOption;\n}) {\n  const { dataQuery, context, extractOption, multimodalPrompt } = options;\n  const systemPrompt = systemPromptToExtract();\n\n  const { screenshotBase64 } = context;\n  const { description, elementById } = await describeUserPage(context, {\n    truncateTextLength: 200,\n    filterNonTextContent: false,\n    visibleOnly: false,\n    domIncluded: extractOption?.domIncluded,\n  });\n\n  const extractDataPromptText = await extractDataQueryPrompt(\n    description,\n    dataQuery,\n  );\n\n  const userContent: ChatCompletionUserMessageParam['content'] = [];\n\n  if (extractOption?.screenshotIncluded !== false) {\n    userContent.push({\n      type: 'image_url',\n      image_url: {\n        url: screenshotBase64,\n        detail: 'high',\n      },\n    });\n  }\n\n  userContent.push({\n    type: 'text',\n    text: extractDataPromptText,\n  });\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userContent,\n    },\n  ];\n\n  if (options.extractOption?.returnThought) {\n    msgs.push({\n      role: 'user',\n      content: 'Please provide reasons.',\n    });\n  }\n\n  if (multimodalPrompt) {\n    const addOns = await promptsToChatParam({\n      images: multimodalPrompt.images,\n      convertHttpImage2Base64: multimodalPrompt.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const result = await callAiFn<AIDataExtractionResponse<T>>(\n    msgs,\n    AIActionType.EXTRACT_DATA,\n  );\n  return {\n    parseResult: result.content,\n    elementById,\n    usage: result.usage,\n  };\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","_multimodalPrompt_images","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","callAI","screenshotBase64","description","elementById","insertElementByPosition","describeUserPage","assert","userInstructionPrompt","findElementPrompt","systemPrompt","systemPromptToLocateElement","vlLocateMode","imagePayload","paddingToMatchBlockByBase64","markupImageForLLM","addOns","callAIFn","callToGetJSONObject","res","AIActionType","rawResponse","JSON","resRect","matchedElements","errors","Array","_options_searchConfig_rect","_options_searchConfig_rect1","_options_searchConfig_rect2","_options_searchConfig_rect3","adaptBboxToRect","rectCenter","element","elementByPositionWithElementInfo","distanceToCenter","distance","distanceThreshold","e","msg","Error","undefined","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","callAiFn","sectionRect","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandSearchArea","imageBase64","cropByRect","getAIConfigInBoolean","MIDSCENE_USE_QWEN_VL","AiExtractElementInfo","_options_extractOption","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;AC0DA,MAAMI,eAAeC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAC9B,MAAMC,eAAeD,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAE9B,MAAME,0BAA0B,CAACC;IAC/B,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAEP,OAAOA,OAAO,MAAM;AAExB;AAEA,MAAMC,qBAAqB,OACzBC;QAGIC;IADJ,MAAMC,OAAyC,EAAE;IACjD,IAAID,QAAAA,mBAAAA,KAAAA,IAAAA,QAAAA,CAAAA,2BAAAA,iBAAkB,MAAM,AAAD,IAAvBA,KAAAA,IAAAA,yBAA0B,MAAM,EAAE;QACpCC,KAAK,IAAI,CAAC;YACR,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM;gBACR;aACD;QACH;QAEA,KAAK,MAAMC,QAAQH,iBAAiB,MAAM,CAAE;YAC1C,MAAMI,SAAS,MAAMC,AAAAA,IAAAA,oBAAAA,kBAAAA,AAAAA,EACnBF,KAAK,GAAG,EACR,CAAC,CAACH,iBAAiB,uBAAuB;YAG5CE,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,MAAM,CAAC,gBAAgB,EAAEC,KAAK,IAAI,CAAC,CAAC,CAAC;oBACvC;iBACD;YACH;YAEAD,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKE;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;QACF;IACF;IACA,OAAOF;AACT;AAEO,eAAeI,gBAEpBC,OAMD;IAQC,MAAM,EAAEC,OAAO,EAAEC,wBAAwB,EAAEC,MAAM,EAAE,GAAGH;IACtD,MAAM,EAAEI,gBAAgB,EAAE,GAAGH;IAC7B,MAAM,EAAEI,WAAW,EAAEC,WAAW,EAAEC,uBAAuB,EAAE,GACzD,MAAMC,AAAAA,IAAAA,wBAAAA,gBAAAA,AAAAA,EAAiBP;IAEzBQ,IAAAA,sBAAAA,MAAAA,AAAAA,EACEP,0BACA;IAGF,MAAMQ,wBAAwB,MAAMC,+BAAAA,iBAAAA,CAAAA,MAAwB,CAAC;QAC3D,iBAAiBN;QACjB,0BAA0Bf,wBAAwBY;IACpD;IACA,MAAMU,eAAeC,AAAAA,IAAAA,+BAAAA,2BAAAA,AAAAA,EAA4BC,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA;IAEjD,IAAIC,eAAeX;IAEnB,IAAIJ,QAAQ,YAAY,EAAE;QACxBS,IAAAA,sBAAAA,MAAAA,AAAAA,EACET,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFS,IAAAA,sBAAAA,MAAAA,AAAAA,EACET,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGFe,eAAef,QAAQ,YAAY,CAAC,WAAW;IACjD,OAAO,IAAIc,AAAmB,cAAnBA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,KACTC,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,2BAAAA,AAAAA,EAA4BD;SAC5C,IAAI,CAACD,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,KACVC,eAAe,MAAME,AAAAA,IAAAA,mCAAAA,iBAAAA,AAAAA,EACnBb,kBACAH,QAAQ,IAAI,EACZA,QAAQ,IAAI;IAIhB,MAAMN,OAAe;QACnB;YAAE,MAAM;YAAU,SAASiB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKG;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAML;gBACR;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAOR,0BAAuC;QAChD,MAAMgB,SAAS,MAAM1B,mBAAmB;YACtC,QAAQU,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAP,KAAK,IAAI,IAAIuB;IACf;IAEA,MAAMC,WACJhB,UAAUiB,yBAAAA,mBAAmBA;IAE/B,MAAMC,MAAM,MAAMF,SAASxB,MAAM2B,mCAAAA,YAAAA,CAAAA,eAA4B;IAE7D,MAAMC,cAAcC,KAAK,SAAS,CAACH,IAAI,OAAO;IAE9C,IAAII;IACJ,IAAIC,kBACF,cAAcL,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,QAAQ,GAAG,EAAE;IACvD,IAAIM,SACF,YAAYN,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IAAI,UAAUA,IAAI,OAAO,IAAIO,MAAM,OAAO,CAACP,IAAI,OAAO,CAAC,IAAI,GAAG;gBAG1DQ,4BAAAA,uBACAC,6BAAAA,wBACAC,6BAAAA,wBACAC,6BAAAA;YALFP,UAAUQ,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EACRZ,IAAI,OAAO,CAAC,IAAI,EAChBQ,AAAAA,SAAAA,CAAAA,wBAAAA,QAAQ,YAAY,AAAD,IAAnBA,KAAAA,IAAAA,QAAAA,CAAAA,6BAAAA,sBAAsB,IAAI,AAAD,IAAzBA,KAAAA,IAAAA,2BAA4B,KAAK,AAAD,KAAK5B,QAAQ,IAAI,CAAC,KAAK,EACvD6B,AAAAA,SAAAA,CAAAA,yBAAAA,QAAQ,YAAY,AAAD,IAAnBA,KAAAA,IAAAA,QAAAA,CAAAA,8BAAAA,uBAAsB,IAAI,AAAD,IAAzBA,KAAAA,IAAAA,4BAA4B,MAAM,AAAD,KAAK7B,QAAQ,IAAI,CAAC,MAAM,UACzD8B,CAAAA,yBAAAA,QAAQ,YAAY,AAAD,IAAnBA,KAAAA,IAAAA,QAAAA,CAAAA,8BAAAA,uBAAsB,IAAI,AAAD,IAAzBA,KAAAA,IAAAA,4BAA4B,IAAI,UAChCC,CAAAA,yBAAAA,QAAQ,YAAY,AAAD,IAAnBA,KAAAA,IAAAA,QAAAA,CAAAA,8BAAAA,uBAAsB,IAAI,AAAD,IAAzBA,KAAAA,IAAAA,4BAA4B,GAAG;YAEjC7C,aAAa,WAAWsC;YAExB,MAAMS,aAAa;gBACjB,GAAGT,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YACA,IAAIU,UAAUC,AAAAA,IAAAA,wBAAAA,gCAAAA,AAAAA,EAAiCnC,QAAQ,IAAI,EAAEiC;YAE7D,MAAMG,mBAAmBF,UACrBG,AAAAA,IAAAA,wBAAAA,QAAAA,AAAAA,EAAS;gBAAE,GAAGH,QAAQ,MAAM,CAAC,EAAE;gBAAE,GAAGA,QAAQ,MAAM,CAAC,EAAE;YAAC,GAAGD,cACzD;YAEJ,IAAI,CAACC,WAAWE,mBAAmBE,wBAAAA,iBAAiBA,EAClDJ,UAAU5B,wBAAwB2B;YAGpC,IAAIC,SAAS;gBACXT,kBAAkB;oBAACS;iBAAQ;gBAC3BR,SAAS,EAAE;YACb;QACF;IACF,EAAE,OAAOa,GAAG;QACV,MAAMC,MACJD,aAAaE,QACT,CAAC,sBAAsB,EAAEF,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACb,UAAUA,AAAAA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM,AAAD,MAAM,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAEc,IAAI,CAAC,CAAC;aAFtBd,SAAS;YAACc;SAAI;IAIlB;IAEA,OAAO;QACL,MAAMhB;QACN,aAAa;YACX,UAAUC;YACVC;QACF;QACAJ;QACAjB;QACA,OAAOe,IAAI,KAAK;QAChB,kBACE,AAAuB,YAAvB,OAAOA,IAAI,OAAO,IAClBA,AAAgB,SAAhBA,IAAI,OAAO,IACX,sBAAsBA,IAAI,OAAO,GAC5BA,IAAI,OAAO,CAAS,gBAAgB,GACrCsB;IACR;AACF;AAEO,eAAeC,gBAAgB5C,OAIrC;IAOC,MAAM,EAAEC,OAAO,EAAE4C,kBAAkB,EAAE,GAAG7C;IACxC,MAAM,EAAEI,gBAAgB,EAAE,GAAGH;IAE7B,MAAMW,eAAekC,AAAAA,IAAAA,uCAAAA,2BAAAA,AAAAA,EAA4BhC,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA;IACjD,MAAMiC,gCAAgC,MAAMC,uCAAAA,yBAAAA,CAAAA,MAAgC,CAAC;QAC3E,oBAAoB1D,wBAAwBuD;IAC9C;IACA,MAAMlD,OAAe;QACnB;YAAE,MAAM;YAAU,SAASiB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKR;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAM2C;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAM3B,SAAS,MAAM1B,mBAAmB;YACtC,QAAQqD,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACAlD,KAAK,IAAI,IAAIuB;IACf;IAEA,MAAM+B,SAAS,MAAMC,AAAAA,IAAAA,mCAAAA,QAAAA,AAAAA,EACnBvD,MACA2B,mCAAAA,YAAAA,CAAAA,YAAyB;IAG3B,IAAI6B;IACJ,MAAMC,cAAcH,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIG,aAAa;QACf,MAAMC,aAAapB,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EACjBmB,aACAnD,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM;QAErBZ,aAAa,0BAA0BgE;QAEvC,MAAMC,oBAAoBL,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9D5D,aAAa,wBAAwBiE;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAAS5B,MAAM,OAAO,CAAC4B,OAC/B,GAAG,CAAC,CAACA,OACGvB,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EAAgBuB,MAAMvD,QAAQ,IAAI,CAAC,KAAK,EAAEA,QAAQ,IAAI,CAAC,MAAM;QAExEZ,aAAa,qBAAqBkE;QAGlC,MAAME,aAAaC,AAAAA,IAAAA,mCAAAA,UAAAA,AAAAA,EAAW;YAACL;eAAeE;SAAe;QAC7DlE,aAAa,iBAAiBoE;QAG9BN,cAAcQ,AAAAA,IAAAA,mCAAAA,gBAAAA,AAAAA,EAAiBF,YAAYxD,QAAQ,IAAI;QACvDZ,aAAa,2BAA2B8D;IAC1C;IAEA,IAAIS,cAAcxD;IAClB,IAAI+C,aACFS,cAAc,MAAMC,AAAAA,IAAAA,oBAAAA,UAAAA,AAAAA,EAClBzD,kBACA+C,aACAW,AAAAA,IAAAA,oBAAAA,oBAAAA,AAAAA,EAAqBC,oBAAAA,oBAAoBA;IAI7C,OAAO;QACL,MAAMZ;QACNS;QACA,OAAOX,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAazB,KAAK,SAAS,CAACyB,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAee,qBAGpBhE,OAKD;QA0CKiE;IAzCJ,MAAM,EAAEC,SAAS,EAAEjE,OAAO,EAAEkE,aAAa,EAAE1E,gBAAgB,EAAE,GAAGO;IAChE,MAAMY,eAAewD,AAAAA,IAAAA,8BAAAA,qBAAAA,AAAAA;IAErB,MAAM,EAAEhE,gBAAgB,EAAE,GAAGH;IAC7B,MAAM,EAAEI,WAAW,EAAEC,WAAW,EAAE,GAAG,MAAME,AAAAA,IAAAA,wBAAAA,gBAAAA,AAAAA,EAAiBP,SAAS;QACnE,oBAAoB;QACpB,sBAAsB;QACtB,aAAa;QACb,aAAakE,QAAAA,gBAAAA,KAAAA,IAAAA,cAAe,WAAW;IACzC;IAEA,MAAME,wBAAwB,MAAMC,AAAAA,IAAAA,8BAAAA,sBAAAA,AAAAA,EAClCjE,aACA6D;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,AAAAA,CAAAA,QAAAA,gBAAAA,KAAAA,IAAAA,cAAe,kBAAkB,AAAD,MAAM,OACxCI,YAAY,IAAI,CAAC;QACf,MAAM;QACN,WAAW;YACT,KAAKnE;YACL,QAAQ;QACV;IACF;IAGFmE,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAM1E,OAAe;QACnB;YAAE,MAAM;YAAU,SAASiB;QAAa;QACxC;YACE,MAAM;YACN,SAAS2D;QACX;KACD;IAED,IAAI,QAAAN,CAAAA,yBAAAA,QAAQ,aAAa,AAAD,IAApBA,KAAAA,IAAAA,uBAAuB,aAAa,EACtCtE,KAAK,IAAI,CAAC;QACR,MAAM;QACN,SAAS;IACX;IAGF,IAAIF,kBAAkB;QACpB,MAAMyB,SAAS,MAAM1B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAE,KAAK,IAAI,IAAIuB;IACf;IAEA,MAAM+B,SAAS,MAAMC,AAAAA,IAAAA,mCAAAA,QAAAA,AAAAA,EACnBvD,MACA2B,mCAAAA,YAAAA,CAAAA,YAAyB;IAE3B,OAAO;QACL,aAAa2B,OAAO,OAAO;QAC3B3C;QACA,OAAO2C,OAAO,KAAK;IACrB;AACF"}
+{"version":3,"file":"ai-model/inspect.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/inspect.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n  AIDataExtractionResponse,\n  AIElementLocatorResponse,\n  AIElementResponse,\n  AISectionLocatorResponse,\n  AIUsageInfo,\n  BaseElement,\n  ElementById,\n  InsightExtractOption,\n  Rect,\n  ReferenceImage,\n  UIContext,\n} from '@/types';\nimport {\n  type IModelPreferences,\n  getIsUseQwenVl,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport {\n  cropByRect,\n  paddingToMatchBlockByBase64,\n  preProcessImageUrl,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type {\n  ChatCompletionSystemMessageParam,\n  ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport type { TMultimodalPrompt, TUserPrompt } from './common';\nimport {\n  AIActionType,\n  adaptBboxToRect,\n  callAiFn,\n  expandSearchArea,\n  markupImageForLLM,\n  mergeRects,\n} from './common';\nimport {\n  extractDataQueryPrompt,\n  systemPromptToExtract,\n} from './prompt/extraction';\nimport {\n  findElementPrompt,\n  systemPromptToLocateElement,\n} from './prompt/llm-locator';\nimport {\n  sectionLocatorInstruction,\n  systemPromptToLocateSection,\n} from './prompt/llm-section-locator';\nimport {\n  describeUserPage,\n  distance,\n  distanceThreshold,\n  elementByPositionWithElementInfo,\n} from './prompt/util';\nimport { callToGetJSONObject } from './service-caller/index';\n\nexport type AIArgs = [\n  ChatCompletionSystemMessageParam,\n  ...ChatCompletionUserMessageParam[],\n];\n\nconst debugInspect = getDebug('ai:inspect');\nconst debugSection = getDebug('ai:section');\n\nconst extraTextFromUserPrompt = (prompt: TUserPrompt): string => {\n  if (typeof prompt === 'string') {\n    return prompt;\n  } else {\n    return prompt.prompt;\n  }\n};\n\nconst promptsToChatParam = async (\n  multimodalPrompt: TMultimodalPrompt,\n): Promise<ChatCompletionUserMessageParam[]> => {\n  const msgs: ChatCompletionUserMessageParam[] = [];\n  if (multimodalPrompt?.images?.length) {\n    msgs.push({\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: 'Next, I will provide all the reference images.',\n        },\n      ],\n    });\n\n    for (const item of multimodalPrompt.images) {\n      const base64 = await preProcessImageUrl(\n        item.url,\n        !!multimodalPrompt.convertHttpImage2Base64,\n      );\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'text',\n            text: `reference image ${item.name}:`,\n          },\n        ],\n      });\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: base64,\n              detail: 'high',\n            },\n          },\n        ],\n      });\n    }\n  }\n  return msgs;\n};\n\nexport async function AiLocateElement<\n  ElementType extends BaseElement = BaseElement,\n>(options: {\n  context: UIContext<ElementType>;\n  targetElementDescription: TUserPrompt;\n  referenceImage?: ReferenceImage;\n  callAI?: typeof callAiFn<AIElementResponse | [number, number]>;\n  searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n}): Promise<{\n  parseResult: AIElementLocatorResponse;\n  rect?: Rect;\n  rawResponse: string;\n  elementById: ElementById;\n  usage?: AIUsageInfo;\n  isOrderSensitive?: boolean;\n}> {\n  const { context, targetElementDescription, callAI } = options;\n  const { screenshotBase64 } = context;\n\n  const modelPreferences: IModelPreferences = {\n    intent: 'grounding',\n  };\n\n  const { description, elementById, insertElementByPosition } =\n    await describeUserPage(context, modelPreferences);\n\n  assert(\n    targetElementDescription,\n    'cannot find the target element description',\n  );\n\n  const userInstructionPrompt = await findElementPrompt.format({\n    pageDescription: description,\n    targetElementDescription: extraTextFromUserPrompt(targetElementDescription),\n  });\n  const systemPrompt = systemPromptToLocateElement(\n    vlLocateMode(modelPreferences),\n  );\n\n  let imagePayload = screenshotBase64;\n\n  if (options.searchConfig) {\n    assert(\n      options.searchConfig.rect,\n      'searchArea is provided but its rect cannot be found. Failed to locate element',\n    );\n    assert(\n      options.searchConfig.imageBase64,\n      'searchArea is provided but its imageBase64 cannot be found. Failed to locate element',\n    );\n\n    imagePayload = options.searchConfig.imageBase64;\n  } else if (vlLocateMode(modelPreferences) === 'qwen-vl') {\n    imagePayload = await paddingToMatchBlockByBase64(imagePayload);\n  } else if (!vlLocateMode(modelPreferences)) {\n    imagePayload = await markupImageForLLM(\n      screenshotBase64,\n      context.tree,\n      context.size,\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          type: 'text',\n          text: userInstructionPrompt,\n        },\n      ],\n    },\n  ];\n\n  if (typeof targetElementDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: targetElementDescription.images,\n      convertHttpImage2Base64: targetElementDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const callAIFn =\n    callAI || callToGetJSONObject<AIElementResponse | [number, number]>;\n\n  const res = await callAIFn(msgs, AIActionType.INSPECT_ELEMENT, {\n    intent: 'grounding',\n  });\n\n  const rawResponse = JSON.stringify(res.content);\n\n  let resRect: Rect | undefined;\n  let matchedElements: AIElementLocatorResponse['elements'] =\n    'elements' in res.content ? res.content.elements : [];\n  let errors: AIElementLocatorResponse['errors'] | undefined =\n    'errors' in res.content ? res.content.errors : [];\n  try {\n    if ('bbox' in res.content && Array.isArray(res.content.bbox)) {\n      resRect = adaptBboxToRect(\n        res.content.bbox,\n        options.searchConfig?.rect?.width || context.size.width,\n        options.searchConfig?.rect?.height || context.size.height,\n        modelPreferences,\n        options.searchConfig?.rect?.left,\n        options.searchConfig?.rect?.top,\n      );\n      debugInspect('resRect', resRect);\n\n      const rectCenter = {\n        x: resRect.left + resRect.width / 2,\n        y: resRect.top + resRect.height / 2,\n      };\n      let element = elementByPositionWithElementInfo(context.tree, rectCenter);\n\n      const distanceToCenter = element\n        ? distance({ x: element.center[0], y: element.center[1] }, rectCenter)\n        : 0;\n\n      if (!element || distanceToCenter > distanceThreshold) {\n        element = insertElementByPosition(rectCenter);\n      }\n\n      if (element) {\n        matchedElements = [element];\n        errors = [];\n      }\n    }\n  } catch (e) {\n    const msg =\n      e instanceof Error\n        ? `Failed to parse bbox: ${e.message}`\n        : 'unknown error in locate';\n    if (!errors || errors?.length === 0) {\n      errors = [msg];\n    } else {\n      errors.push(`(${msg})`);\n    }\n  }\n\n  return {\n    rect: resRect,\n    parseResult: {\n      elements: matchedElements,\n      errors,\n    },\n    rawResponse,\n    elementById,\n    usage: res.usage,\n    isOrderSensitive:\n      typeof res.content === 'object' &&\n      res.content !== null &&\n      'isOrderSensitive' in res.content\n        ? (res.content as any).isOrderSensitive\n        : undefined,\n  };\n}\n\nexport async function AiLocateSection(options: {\n  context: UIContext<BaseElement>;\n  sectionDescription: TUserPrompt;\n  callAI?: typeof callAiFn<AISectionLocatorResponse>;\n}): Promise<{\n  rect?: Rect;\n  imageBase64?: string;\n  error?: string;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n}> {\n  const { context, sectionDescription } = options;\n  const { screenshotBase64 } = context;\n\n  const modelPreferences: IModelPreferences = {\n    intent: 'grounding',\n  };\n\n  const systemPrompt = systemPromptToLocateSection(\n    vlLocateMode(modelPreferences),\n  );\n  const sectionLocatorInstructionText = await sectionLocatorInstruction.format({\n    sectionDescription: extraTextFromUserPrompt(sectionDescription),\n  });\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: screenshotBase64,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: sectionLocatorInstructionText,\n        },\n      ],\n    },\n  ];\n\n  if (typeof sectionDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: sectionDescription.images,\n      convertHttpImage2Base64: sectionDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const result = await callAiFn<AISectionLocatorResponse>(\n    msgs,\n    AIActionType.EXTRACT_DATA,\n    {\n      intent: 'grounding',\n    },\n  );\n\n  let sectionRect: Rect | undefined;\n  const sectionBbox = result.content.bbox;\n  if (sectionBbox) {\n    const targetRect = adaptBboxToRect(\n      sectionBbox,\n      context.size.width,\n      context.size.height,\n      modelPreferences,\n    );\n    debugSection('original targetRect %j', targetRect);\n\n    const referenceBboxList = result.content.references_bbox || [];\n    debugSection('referenceBboxList %j', referenceBboxList);\n\n    const referenceRects = referenceBboxList\n      .filter((bbox) => Array.isArray(bbox))\n      .map((bbox) => {\n        return adaptBboxToRect(\n          bbox,\n          context.size.width,\n          context.size.height,\n          modelPreferences,\n        );\n      });\n    debugSection('referenceRects %j', referenceRects);\n\n    // merge the sectionRect and referenceRects\n    const mergedRect = mergeRects([targetRect, ...referenceRects]);\n    debugSection('mergedRect %j', mergedRect);\n\n    // expand search area to at least 200 x 200\n    sectionRect = expandSearchArea(mergedRect, context.size, modelPreferences);\n    debugSection('expanded sectionRect %j', sectionRect);\n  }\n\n  let imageBase64 = screenshotBase64;\n  if (sectionRect) {\n    imageBase64 = await cropByRect(\n      screenshotBase64,\n      sectionRect,\n      getIsUseQwenVl({\n        intent: 'grounding',\n      }),\n    );\n  }\n\n  return {\n    rect: sectionRect,\n    imageBase64,\n    error: result.content.error,\n    rawResponse: JSON.stringify(result.content),\n    usage: result.usage,\n  };\n}\n\nexport async function AiExtractElementInfo<\n  T,\n  ElementType extends BaseElement = BaseElement,\n>(options: {\n  dataQuery: string | Record<string, string>;\n  multimodalPrompt?: TMultimodalPrompt;\n  context: UIContext<ElementType>;\n  extractOption?: InsightExtractOption;\n  modelPreferences: IModelPreferences;\n}) {\n  const {\n    dataQuery,\n    context,\n    extractOption,\n    multimodalPrompt,\n    modelPreferences,\n  } = options;\n  const systemPrompt = systemPromptToExtract();\n\n  const { screenshotBase64 } = context;\n  const { description, elementById } = await describeUserPage(\n    context,\n    modelPreferences,\n    {\n      truncateTextLength: 200,\n      filterNonTextContent: false,\n      visibleOnly: false,\n      domIncluded: extractOption?.domIncluded,\n    },\n  );\n\n  const extractDataPromptText = await extractDataQueryPrompt(\n    description,\n    dataQuery,\n  );\n\n  const userContent: ChatCompletionUserMessageParam['content'] = [];\n\n  if (extractOption?.screenshotIncluded !== false) {\n    userContent.push({\n      type: 'image_url',\n      image_url: {\n        url: screenshotBase64,\n        detail: 'high',\n      },\n    });\n  }\n\n  userContent.push({\n    type: 'text',\n    text: extractDataPromptText,\n  });\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userContent,\n    },\n  ];\n\n  if (options.extractOption?.returnThought) {\n    msgs.push({\n      role: 'user',\n      content: 'Please provide reasons.',\n    });\n  }\n\n  if (multimodalPrompt) {\n    const addOns = await promptsToChatParam({\n      images: multimodalPrompt.images,\n      convertHttpImage2Base64: multimodalPrompt.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const result = await callAiFn<AIDataExtractionResponse<T>>(\n    msgs,\n    AIActionType.EXTRACT_DATA,\n    modelPreferences,\n  );\n  return {\n    parseResult: result.content,\n    elementById,\n    usage: result.usage,\n  };\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","_multimodalPrompt_images","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","callAI","screenshotBase64","modelPreferences","description","elementById","insertElementByPosition","describeUserPage","assert","userInstructionPrompt","findElementPrompt","systemPrompt","systemPromptToLocateElement","vlLocateMode","imagePayload","paddingToMatchBlockByBase64","markupImageForLLM","addOns","callAIFn","callToGetJSONObject","res","AIActionType","rawResponse","JSON","resRect","matchedElements","errors","Array","_options_searchConfig_rect","_options_searchConfig_rect1","_options_searchConfig_rect2","_options_searchConfig_rect3","adaptBboxToRect","rectCenter","element","elementByPositionWithElementInfo","distanceToCenter","distance","distanceThreshold","e","msg","Error","undefined","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","callAiFn","sectionRect","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandSearchArea","imageBase64","cropByRect","getIsUseQwenVl","AiExtractElementInfo","_options_extractOption","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;ACyDA,MAAMI,eAAeC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAC9B,MAAMC,eAAeD,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAE9B,MAAME,0BAA0B,CAACC;IAC/B,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAEP,OAAOA,OAAO,MAAM;AAExB;AAEA,MAAMC,qBAAqB,OACzBC;QAGIC;IADJ,MAAMC,OAAyC,EAAE;IACjD,IAAID,QAAAA,mBAAAA,KAAAA,IAAAA,QAAAA,CAAAA,2BAAAA,iBAAkB,MAAM,AAAD,IAAvBA,KAAAA,IAAAA,yBAA0B,MAAM,EAAE;QACpCC,KAAK,IAAI,CAAC;YACR,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM;gBACR;aACD;QACH;QAEA,KAAK,MAAMC,QAAQH,iBAAiB,MAAM,CAAE;YAC1C,MAAMI,SAAS,MAAMC,AAAAA,IAAAA,oBAAAA,kBAAAA,AAAAA,EACnBF,KAAK,GAAG,EACR,CAAC,CAACH,iBAAiB,uBAAuB;YAG5CE,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,MAAM,CAAC,gBAAgB,EAAEC,KAAK,IAAI,CAAC,CAAC,CAAC;oBACvC;iBACD;YACH;YAEAD,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKE;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;QACF;IACF;IACA,OAAOF;AACT;AAEO,eAAeI,gBAEpBC,OAMD;IAQC,MAAM,EAAEC,OAAO,EAAEC,wBAAwB,EAAEC,MAAM,EAAE,GAAGH;IACtD,MAAM,EAAEI,gBAAgB,EAAE,GAAGH;IAE7B,MAAMI,mBAAsC;QAC1C,QAAQ;IACV;IAEA,MAAM,EAAEC,WAAW,EAAEC,WAAW,EAAEC,uBAAuB,EAAE,GACzD,MAAMC,AAAAA,IAAAA,wBAAAA,gBAAAA,AAAAA,EAAiBR,SAASI;IAElCK,IAAAA,sBAAAA,MAAAA,AAAAA,EACER,0BACA;IAGF,MAAMS,wBAAwB,MAAMC,+BAAAA,iBAAAA,CAAAA,MAAwB,CAAC;QAC3D,iBAAiBN;QACjB,0BAA0BhB,wBAAwBY;IACpD;IACA,MAAMW,eAAeC,AAAAA,IAAAA,+BAAAA,2BAAAA,AAAAA,EACnBC,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAaV;IAGf,IAAIW,eAAeZ;IAEnB,IAAIJ,QAAQ,YAAY,EAAE;QACxBU,IAAAA,sBAAAA,MAAAA,AAAAA,EACEV,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFU,IAAAA,sBAAAA,MAAAA,AAAAA,EACEV,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGFgB,eAAehB,QAAQ,YAAY,CAAC,WAAW;IACjD,OAAO,IAAIe,AAAmC,cAAnCA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAaV,mBACtBW,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,2BAAAA,AAAAA,EAA4BD;SAC5C,IAAI,CAACD,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAaV,mBACvBW,eAAe,MAAME,AAAAA,IAAAA,mCAAAA,iBAAAA,AAAAA,EACnBd,kBACAH,QAAQ,IAAI,EACZA,QAAQ,IAAI;IAIhB,MAAMN,OAAe;QACnB;YAAE,MAAM;YAAU,SAASkB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKG;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAML;gBACR;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAOT,0BAAuC;QAChD,MAAMiB,SAAS,MAAM3B,mBAAmB;YACtC,QAAQU,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAP,KAAK,IAAI,IAAIwB;IACf;IAEA,MAAMC,WACJjB,UAAUkB,yBAAAA,mBAAmBA;IAE/B,MAAMC,MAAM,MAAMF,SAASzB,MAAM4B,mCAAAA,YAAAA,CAAAA,eAA4B,EAAE;QAC7D,QAAQ;IACV;IAEA,MAAMC,cAAcC,KAAK,SAAS,CAACH,IAAI,OAAO;IAE9C,IAAII;IACJ,IAAIC,kBACF,cAAcL,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,QAAQ,GAAG,EAAE;IACvD,IAAIM,SACF,YAAYN,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IAAI,UAAUA,IAAI,OAAO,IAAIO,MAAM,OAAO,CAACP,IAAI,OAAO,CAAC,IAAI,GAAG;gBAG1DQ,4BAAAA,uBACAC,6BAAAA,wBAEAC,6BAAAA,wBACAC,6BAAAA;YANFP,UAAUQ,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EACRZ,IAAI,OAAO,CAAC,IAAI,EAChBQ,AAAAA,SAAAA,CAAAA,wBAAAA,QAAQ,YAAY,AAAD,IAAnBA,KAAAA,IAAAA,QAAAA,CAAAA,6BAAAA,sBAAsB,IAAI,AAAD,IAAzBA,KAAAA,IAAAA,2BAA4B,KAAK,AAAD,KAAK7B,QAAQ,IAAI,CAAC,KAAK,EACvD8B,AAAAA,SAAAA,CAAAA,yBAAAA,QAAQ,YAAY,AAAD,IAAnBA,KAAAA,IAAAA,QAAAA,CAAAA,8BAAAA,uBAAsB,IAAI,AAAD,IAAzBA,KAAAA,IAAAA,4BAA4B,MAAM,AAAD,KAAK9B,QAAQ,IAAI,CAAC,MAAM,EACzDI,kBAAAA,QACA2B,CAAAA,yBAAAA,QAAQ,YAAY,AAAD,IAAnBA,KAAAA,IAAAA,QAAAA,CAAAA,8BAAAA,uBAAsB,IAAI,AAAD,IAAzBA,KAAAA,IAAAA,4BAA4B,IAAI,UAChCC,CAAAA,yBAAAA,QAAQ,YAAY,AAAD,IAAnBA,KAAAA,IAAAA,QAAAA,CAAAA,8BAAAA,uBAAsB,IAAI,AAAD,IAAzBA,KAAAA,IAAAA,4BAA4B,GAAG;YAEjC9C,aAAa,WAAWuC;YAExB,MAAMS,aAAa;gBACjB,GAAGT,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YACA,IAAIU,UAAUC,AAAAA,IAAAA,wBAAAA,gCAAAA,AAAAA,EAAiCpC,QAAQ,IAAI,EAAEkC;YAE7D,MAAMG,mBAAmBF,UACrBG,AAAAA,IAAAA,wBAAAA,QAAAA,AAAAA,EAAS;gBAAE,GAAGH,QAAQ,MAAM,CAAC,EAAE;gBAAE,GAAGA,QAAQ,MAAM,CAAC,EAAE;YAAC,GAAGD,cACzD;YAEJ,IAAI,CAACC,WAAWE,mBAAmBE,wBAAAA,iBAAiBA,EAClDJ,UAAU5B,wBAAwB2B;YAGpC,IAAIC,SAAS;gBACXT,kBAAkB;oBAACS;iBAAQ;gBAC3BR,SAAS,EAAE;YACb;QACF;IACF,EAAE,OAAOa,GAAG;QACV,MAAMC,MACJD,aAAaE,QACT,CAAC,sBAAsB,EAAEF,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACb,UAAUA,AAAAA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,MAAM,AAAD,MAAM,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAEc,IAAI,CAAC,CAAC;aAFtBd,SAAS;YAACc;SAAI;IAIlB;IAEA,OAAO;QACL,MAAMhB;QACN,aAAa;YACX,UAAUC;YACVC;QACF;QACAJ;QACAjB;QACA,OAAOe,IAAI,KAAK;QAChB,kBACE,AAAuB,YAAvB,OAAOA,IAAI,OAAO,IAClBA,AAAgB,SAAhBA,IAAI,OAAO,IACX,sBAAsBA,IAAI,OAAO,GAC5BA,IAAI,OAAO,CAAS,gBAAgB,GACrCsB;IACR;AACF;AAEO,eAAeC,gBAAgB7C,OAIrC;IAOC,MAAM,EAAEC,OAAO,EAAE6C,kBAAkB,EAAE,GAAG9C;IACxC,MAAM,EAAEI,gBAAgB,EAAE,GAAGH;IAE7B,MAAMI,mBAAsC;QAC1C,QAAQ;IACV;IAEA,MAAMQ,eAAekC,AAAAA,IAAAA,uCAAAA,2BAAAA,AAAAA,EACnBhC,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAaV;IAEf,MAAM2C,gCAAgC,MAAMC,uCAAAA,yBAAAA,CAAAA,MAAgC,CAAC;QAC3E,oBAAoB3D,wBAAwBwD;IAC9C;IACA,MAAMnD,OAAe;QACnB;YAAE,MAAM;YAAU,SAASkB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKT;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAM4C;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAM3B,SAAS,MAAM3B,mBAAmB;YACtC,QAAQsD,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACAnD,KAAK,IAAI,IAAIwB;IACf;IAEA,MAAM+B,SAAS,MAAMC,AAAAA,IAAAA,mCAAAA,QAAAA,AAAAA,EACnBxD,MACA4B,mCAAAA,YAAAA,CAAAA,YAAyB,EACzB;QACE,QAAQ;IACV;IAGF,IAAI6B;IACJ,MAAMC,cAAcH,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIG,aAAa;QACf,MAAMC,aAAapB,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EACjBmB,aACApD,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAEFhB,aAAa,0BAA0BiE;QAEvC,MAAMC,oBAAoBL,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9D7D,aAAa,wBAAwBkE;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAAS5B,MAAM,OAAO,CAAC4B,OAC/B,GAAG,CAAC,CAACA,OACGvB,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EACLuB,MACAxD,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAGNhB,aAAa,qBAAqBmE;QAGlC,MAAME,aAAaC,AAAAA,IAAAA,mCAAAA,UAAAA,AAAAA,EAAW;YAACL;eAAeE;SAAe;QAC7DnE,aAAa,iBAAiBqE;QAG9BN,cAAcQ,AAAAA,IAAAA,mCAAAA,gBAAAA,AAAAA,EAAiBF,YAAYzD,QAAQ,IAAI,EAAEI;QACzDhB,aAAa,2BAA2B+D;IAC1C;IAEA,IAAIS,cAAczD;IAClB,IAAIgD,aACFS,cAAc,MAAMC,AAAAA,IAAAA,oBAAAA,UAAAA,AAAAA,EAClB1D,kBACAgD,aACAW,AAAAA,IAAAA,oBAAAA,cAAAA,AAAAA,EAAe;QACb,QAAQ;IACV;IAIJ,OAAO;QACL,MAAMX;QACNS;QACA,OAAOX,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAazB,KAAK,SAAS,CAACyB,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAec,qBAGpBhE,OAMD;QAoDKiE;IAnDJ,MAAM,EACJC,SAAS,EACTjE,OAAO,EACPkE,aAAa,EACb1E,gBAAgB,EAChBY,gBAAgB,EACjB,GAAGL;IACJ,MAAMa,eAAeuD,AAAAA,IAAAA,8BAAAA,qBAAAA,AAAAA;IAErB,MAAM,EAAEhE,gBAAgB,EAAE,GAAGH;IAC7B,MAAM,EAAEK,WAAW,EAAEC,WAAW,EAAE,GAAG,MAAME,AAAAA,IAAAA,wBAAAA,gBAAAA,AAAAA,EACzCR,SACAI,kBACA;QACE,oBAAoB;QACpB,sBAAsB;QACtB,aAAa;QACb,aAAa8D,QAAAA,gBAAAA,KAAAA,IAAAA,cAAe,WAAW;IACzC;IAGF,MAAME,wBAAwB,MAAMC,AAAAA,IAAAA,8BAAAA,sBAAAA,AAAAA,EAClChE,aACA4D;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,AAAAA,CAAAA,QAAAA,gBAAAA,KAAAA,IAAAA,cAAe,kBAAkB,AAAD,MAAM,OACxCI,YAAY,IAAI,CAAC;QACf,MAAM;QACN,WAAW;YACT,KAAKnE;YACL,QAAQ;QACV;IACF;IAGFmE,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAM1E,OAAe;QACnB;YAAE,MAAM;YAAU,SAASkB;QAAa;QACxC;YACE,MAAM;YACN,SAAS0D;QACX;KACD;IAED,IAAI,QAAAN,CAAAA,yBAAAA,QAAQ,aAAa,AAAD,IAApBA,KAAAA,IAAAA,uBAAuB,aAAa,EACtCtE,KAAK,IAAI,CAAC;QACR,MAAM;QACN,SAAS;IACX;IAGF,IAAIF,kBAAkB;QACpB,MAAM0B,SAAS,MAAM3B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAE,KAAK,IAAI,IAAIwB;IACf;IAEA,MAAM+B,SAAS,MAAMC,AAAAA,IAAAA,mCAAAA,QAAAA,AAAAA,EACnBxD,MACA4B,mCAAAA,YAAAA,CAAAA,YAAyB,EACzBlB;IAEF,OAAO;QACL,aAAa6C,OAAO,OAAO;QAC3B3C;QACA,OAAO2C,OAAO,KAAK;IACrB;AACF"}

package/dist/lib/ai-model/llm-planning.js

@@ -28,28 +28,33 @@ __webpack_require__.d(__webpack_exports__, {
 });
 const env_namespaceObject = require("@midscene/shared/env");
 const img_namespaceObject = require("@midscene/shared/img");
+const logger_namespaceObject = require("@midscene/shared/logger");
 const utils_namespaceObject = require("@midscene/shared/utils");
 const external_common_js_namespaceObject = require("./common.js");
 const llm_planning_js_namespaceObject = require("./prompt/llm-planning.js");
 const util_js_namespaceObject = require("./prompt/util.js");
+const debug = (0, logger_namespaceObject.getDebug)('planning');
 async function plan(userInstruction, opts) {
     var _planFromAI_action;
     const { callAI, context } = opts || {};
     const { screenshotBase64, size } = context;
-    const { description: pageDescription, elementById } = await (0, util_js_namespaceObject.describeUserPage)(context);
+    const modelPreferences = {
+        intent: 'planning'
+    };
+    const { description: pageDescription, elementById } = await (0, util_js_namespaceObject.describeUserPage)(context, modelPreferences);
     const systemPrompt = await (0, llm_planning_js_namespaceObject.systemPromptToTaskPlanning)({
         actionSpace: opts.actionSpace,
-        vlMode: (0, env_namespaceObject.vlLocateMode)()
+        vlMode: (0, env_namespaceObject.vlLocateMode)(modelPreferences)
     });
     const taskBackgroundContextText = (0, llm_planning_js_namespaceObject.generateTaskBackgroundContext)(userInstruction, opts.log, opts.actionContext);
-    const userInstructionPrompt = await (0, llm_planning_js_namespaceObject.automationUserPrompt)((0, env_namespaceObject.vlLocateMode)()).format({
+    const userInstructionPrompt = await (0, llm_planning_js_namespaceObject.automationUserPrompt)((0, env_namespaceObject.vlLocateMode)(modelPreferences)).format({
         pageDescription,
         taskBackgroundContext: taskBackgroundContextText
     });
     let imagePayload = screenshotBase64;
-    if ('qwen-vl' === (0, env_namespaceObject.vlLocateMode)()) imagePayload = await (0, img_namespaceObject.paddingToMatchBlockByBase64)(imagePayload);
-    else if (!(0, env_namespaceObject.vlLocateMode)()) imagePayload = await (0, external_common_js_namespaceObject.markupImageForLLM)(screenshotBase64, context.tree, context.size);
-    (0, external_common_js_namespaceObject.warnGPT4oSizeLimit)(size);
+    if ('qwen-vl' === (0, env_namespaceObject.vlLocateMode)(modelPreferences)) imagePayload = await (0, img_namespaceObject.paddingToMatchBlockByBase64)(imagePayload);
+    else if (!(0, env_namespaceObject.vlLocateMode)(modelPreferences)) imagePayload = await (0, external_common_js_namespaceObject.markupImageForLLM)(screenshotBase64, context.tree, context.size);
+    (0, external_common_js_namespaceObject.warnGPT4oSizeLimit)(size, modelPreferences);
     const msgs = [
         {
             role: 'system',
@@ -73,7 +78,7 @@ async function plan(userInstruction, opts) {
         }
     ];
     const call = callAI || external_common_js_namespaceObject.callAiFn;
-    const { content, usage } = await call(msgs, external_common_js_namespaceObject.AIActionType.PLAN);
+    const { content, usage } = await call(msgs, external_common_js_namespaceObject.AIActionType.PLAN, modelPreferences);
     const rawResponse = JSON.stringify(content, void 0, 2);
     const planFromAI = content;
     const actions = ((null == (_planFromAI_action = planFromAI.action) ? void 0 : _planFromAI_action.type) ? [
@@ -87,24 +92,21 @@ async function plan(userInstruction, opts) {
         yamlFlow: (0, external_common_js_namespaceObject.buildYamlFlowFromPlans)(actions, opts.actionSpace, planFromAI.sleep)
     };
     (0, utils_namespaceObject.assert)(planFromAI, "can't get plans from AI");
-    if ((0, env_namespaceObject.vlLocateMode)()) {
-        actions.forEach((action)=>{
-            if (action.locate) try {
-                action.locate = (0, external_common_js_namespaceObject.fillBboxParam)(action.locate, size.width, size.height);
-            } catch (e) {
-                throw new Error(`Failed to fill locate param: ${planFromAI.error} (${e instanceof Error ? e.message : 'unknown error'})`, {
-                    cause: e
-                });
+    actions.forEach((action)=>{
+        const type = action.type;
+        const actionInActionSpace = opts.actionSpace.find((action)=>action.name === type);
+        const locateFields = actionInActionSpace ? (0, external_common_js_namespaceObject.findAllMidsceneLocatorField)(actionInActionSpace.paramSchema) : [];
+        debug('locateFields', locateFields);
+        locateFields.forEach((field)=>{
+            const locateResult = action.param[field];
+            if (locateResult) if ((0, env_namespaceObject.vlLocateMode)(modelPreferences)) action.param[field] = (0, external_common_js_namespaceObject.fillBboxParam)(locateResult, size.width, size.height, modelPreferences);
+            else {
+                const element = elementById(locateResult);
+                if (element) action.param[field].id = element.id;
             }
         });
-        (0, utils_namespaceObject.assert)(!planFromAI.error, `Failed to plan actions: ${planFromAI.error}`);
-    } else actions.forEach((action)=>{
-        var _action_locate;
-        if (null == (_action_locate = action.locate) ? void 0 : _action_locate.id) {
-            const element = elementById(action.locate.id);
-            if (element) action.locate.id = element.id;
-        }
     });
+    (0, utils_namespaceObject.assert)(!planFromAI.error, `Failed to plan actions: ${planFromAI.error}`);
     if (0 === actions.length && returnValue.more_actions_needed_by_instruction && !returnValue.sleep) console.warn('No actions planned for the prompt, but model said more actions are needed:', userInstruction);
     return returnValue;
 }

package/dist/lib/ai-model/llm-planning.js.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/llm-planning.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/llm-planning.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n  DeviceAction,\n  PageType,\n  PlanningAIResponse,\n  UIContext,\n} from '@/types';\nimport { vlLocateMode } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { assert } from '@midscene/shared/utils';\nimport {\n  AIActionType,\n  type AIArgs,\n  buildYamlFlowFromPlans,\n  callAiFn,\n  fillBboxParam,\n  markupImageForLLM,\n  warnGPT4oSizeLimit,\n} from './common';\nimport {\n  automationUserPrompt,\n  generateTaskBackgroundContext,\n  systemPromptToTaskPlanning,\n} from './prompt/llm-planning';\nimport { describeUserPage } from './prompt/util';\n\nexport async function plan(\n  userInstruction: string,\n  opts: {\n    context: UIContext;\n    pageType: PageType;\n    actionSpace: DeviceAction[];\n    callAI?: typeof callAiFn<PlanningAIResponse>;\n    log?: string;\n    actionContext?: string;\n  },\n): Promise<PlanningAIResponse> {\n  const { callAI, context } = opts || {};\n  const { screenshotBase64, size } = context;\n  const { description: pageDescription, elementById } =\n    await describeUserPage(context);\n\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    vlMode: vlLocateMode(),\n  });\n  const taskBackgroundContextText = generateTaskBackgroundContext(\n    userInstruction,\n    opts.log,\n    opts.actionContext,\n  );\n  const userInstructionPrompt = await automationUserPrompt(\n    vlLocateMode(),\n  ).format({\n    pageDescription,\n    taskBackgroundContext: taskBackgroundContextText,\n  });\n\n  let imagePayload = screenshotBase64;\n  if (vlLocateMode() === 'qwen-vl') {\n    imagePayload = await paddingToMatchBlockByBase64(imagePayload);\n  } else if (!vlLocateMode()) {\n    imagePayload = await markupImageForLLM(\n      screenshotBase64,\n      context.tree,\n      context.size,\n    );\n  }\n\n  warnGPT4oSizeLimit(size);\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: userInstructionPrompt,\n        },\n      ],\n    },\n  ];\n\n  const call = callAI || callAiFn;\n  const { content, usage } = await call(msgs, AIActionType.PLAN);\n  const rawResponse = JSON.stringify(content, undefined, 2);\n  const planFromAI = content;\n\n  const actions =\n    (planFromAI.action?.type ? [planFromAI.action] : planFromAI.actions) || [];\n  const returnValue: PlanningAIResponse = {\n    ...planFromAI,\n    actions,\n    rawResponse,\n    usage,\n    yamlFlow: buildYamlFlowFromPlans(\n      actions,\n      opts.actionSpace,\n      planFromAI.sleep,\n    ),\n  };\n\n  assert(planFromAI, \"can't get plans from AI\");\n\n  if (vlLocateMode()) {\n    actions.forEach((action) => {\n      if (action.locate) {\n        try {\n          action.locate = fillBboxParam(action.locate, size.width, size.height);\n        } catch (e) {\n          throw new Error(\n            `Failed to fill locate param: ${planFromAI.error} (${\n              e instanceof Error ? e.message : 'unknown error'\n            })`,\n            {\n              cause: e,\n            },\n          );\n        }\n      }\n    });\n    // in Qwen-VL, error means error. In GPT-4o, error may mean more actions are needed.\n    assert(!planFromAI.error, `Failed to plan actions: ${planFromAI.error}`);\n  } else {\n    actions.forEach((action) => {\n      if (action.locate?.id) {\n        // The model may return indexId, need to perform a query correction to avoid exceptions\n        const element = elementById(action.locate.id);\n        if (element) {\n          action.locate.id = element.id;\n        }\n      }\n    });\n  }\n\n  if (\n    actions.length === 0 &&\n    returnValue.more_actions_needed_by_instruction &&\n    !returnValue.sleep\n  ) {\n    console.warn(\n      'No actions planned for the prompt, but model said more actions are needed:',\n      userInstruction,\n    );\n  }\n\n  return returnValue;\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","plan","userInstruction","opts","_planFromAI_action","callAI","context","screenshotBase64","size","pageDescription","elementById","describeUserPage","systemPrompt","systemPromptToTaskPlanning","vlLocateMode","taskBackgroundContextText","generateTaskBackgroundContext","userInstructionPrompt","automationUserPrompt","imagePayload","paddingToMatchBlockByBase64","markupImageForLLM","warnGPT4oSizeLimit","msgs","call","callAiFn","content","usage","AIActionType","rawResponse","JSON","undefined","planFromAI","actions","returnValue","buildYamlFlowFromPlans","assert","action","fillBboxParam","e","Error","_action_locate","element","console"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;ACmBO,eAAeI,KACpBC,eAAuB,EACvBC,IAOC;QA8DEC;IA5DH,MAAM,EAAEC,MAAM,EAAEC,OAAO,EAAE,GAAGH,QAAQ,CAAC;IACrC,MAAM,EAAEI,gBAAgB,EAAEC,IAAI,EAAE,GAAGF;IACnC,MAAM,EAAE,aAAaG,eAAe,EAAEC,WAAW,EAAE,GACjD,MAAMC,AAAAA,IAAAA,wBAAAA,gBAAAA,AAAAA,EAAiBL;IAEzB,MAAMM,eAAe,MAAMC,AAAAA,IAAAA,gCAAAA,0BAAAA,AAAAA,EAA2B;QACpD,aAAaV,KAAK,WAAW;QAC7B,QAAQW,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA;IACV;IACA,MAAMC,4BAA4BC,AAAAA,IAAAA,gCAAAA,6BAAAA,AAAAA,EAChCd,iBACAC,KAAK,GAAG,EACRA,KAAK,aAAa;IAEpB,MAAMc,wBAAwB,MAAMC,AAAAA,IAAAA,gCAAAA,oBAAAA,AAAAA,EAClCJ,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,KACA,MAAM,CAAC;QACPL;QACA,uBAAuBM;IACzB;IAEA,IAAII,eAAeZ;IACnB,IAAIO,AAAmB,cAAnBA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,KACFK,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,2BAAAA,AAAAA,EAA4BD;SAC5C,IAAI,CAACL,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,KACVK,eAAe,MAAME,AAAAA,IAAAA,mCAAAA,iBAAAA,AAAAA,EACnBd,kBACAD,QAAQ,IAAI,EACZA,QAAQ,IAAI;IAIhBgB,IAAAA,mCAAAA,kBAAAA,AAAAA,EAAmBd;IAEnB,MAAMe,OAAe;QACnB;YAAE,MAAM;YAAU,SAASX;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKO;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMF;gBACR;aACD;QACH;KACD;IAED,MAAMO,OAAOnB,UAAUoB,mCAAAA,QAAQA;IAC/B,MAAM,EAAEC,OAAO,EAAEC,KAAK,EAAE,GAAG,MAAMH,KAAKD,MAAMK,mCAAAA,YAAAA,CAAAA,IAAiB;IAC7D,MAAMC,cAAcC,KAAK,SAAS,CAACJ,SAASK,QAAW;IACvD,MAAMC,aAAaN;IAEnB,MAAMO,UACH7B,AAAAA,CAAAA,SAAAA,CAAAA,qBAAAA,WAAW,MAAM,AAAD,IAAhBA,KAAAA,IAAAA,mBAAmB,IAAI,AAAD,IAAI;QAAC4B,WAAW,MAAM;KAAC,GAAGA,WAAW,OAAM,KAAM,EAAE;IAC5E,MAAME,cAAkC;QACtC,GAAGF,UAAU;QACbC;QACAJ;QACAF;QACA,UAAUQ,AAAAA,IAAAA,mCAAAA,sBAAAA,AAAAA,EACRF,SACA9B,KAAK,WAAW,EAChB6B,WAAW,KAAK;IAEpB;IAEAI,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOJ,YAAY;IAEnB,IAAIlB,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,KAAgB;QAClBmB,QAAQ,OAAO,CAAC,CAACI;YACf,IAAIA,OAAO,MAAM,EACf,IAAI;gBACFA,OAAO,MAAM,GAAGC,AAAAA,IAAAA,mCAAAA,aAAAA,AAAAA,EAAcD,OAAO,MAAM,EAAE7B,KAAK,KAAK,EAAEA,KAAK,MAAM;YACtE,EAAE,OAAO+B,GAAG;gBACV,MAAM,IAAIC,MACR,CAAC,6BAA6B,EAAER,WAAW,KAAK,CAAC,EAAE,EACjDO,aAAaC,QAAQD,EAAE,OAAO,GAAG,gBAClC,CAAC,CAAC,EACH;oBACE,OAAOA;gBACT;YAEJ;QAEJ;QAEAH,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,CAACJ,WAAW,KAAK,EAAE,CAAC,wBAAwB,EAAEA,WAAW,KAAK,EAAE;IACzE,OACEC,QAAQ,OAAO,CAAC,CAACI;YACXI;QAAJ,IAAI,QAAAA,CAAAA,iBAAAA,OAAO,MAAM,AAAD,IAAZA,KAAAA,IAAAA,eAAe,EAAE,EAAE;YAErB,MAAMC,UAAUhC,YAAY2B,OAAO,MAAM,CAAC,EAAE;YAC5C,IAAIK,SACFL,OAAO,MAAM,CAAC,EAAE,GAAGK,QAAQ,EAAE;QAEjC;IACF;IAGF,IACET,AAAmB,MAAnBA,QAAQ,MAAM,IACdC,YAAY,kCAAkC,IAC9C,CAACA,YAAY,KAAK,EAElBS,QAAQ,IAAI,CACV,8EACAzC;IAIJ,OAAOgC;AACT"}
+{"version":3,"file":"ai-model/llm-planning.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/llm-planning.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n  DeviceAction,\n  PageType,\n  PlanningAIResponse,\n  UIContext,\n} from '@/types';\nimport { type IModelPreferences, vlLocateMode } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport {\n  AIActionType,\n  type AIArgs,\n  buildYamlFlowFromPlans,\n  callAiFn,\n  fillBboxParam,\n  findAllMidsceneLocatorField,\n  markupImageForLLM,\n  warnGPT4oSizeLimit,\n} from './common';\nimport {\n  automationUserPrompt,\n  generateTaskBackgroundContext,\n  systemPromptToTaskPlanning,\n} from './prompt/llm-planning';\nimport { describeUserPage } from './prompt/util';\n\nconst debug = getDebug('planning');\n\nexport async function plan(\n  userInstruction: string,\n  opts: {\n    context: UIContext;\n    pageType: PageType;\n    actionSpace: DeviceAction<any>[];\n    callAI?: typeof callAiFn<PlanningAIResponse>;\n    log?: string;\n    actionContext?: string;\n  },\n): Promise<PlanningAIResponse> {\n  const { callAI, context } = opts || {};\n  const { screenshotBase64, size } = context;\n\n  const modelPreferences: IModelPreferences = {\n    intent: 'planning',\n  };\n  const { description: pageDescription, elementById } = await describeUserPage(\n    context,\n    modelPreferences,\n  );\n\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    vlMode: vlLocateMode(modelPreferences),\n  });\n  const taskBackgroundContextText = generateTaskBackgroundContext(\n    userInstruction,\n    opts.log,\n    opts.actionContext,\n  );\n  const userInstructionPrompt = await automationUserPrompt(\n    vlLocateMode(modelPreferences),\n  ).format({\n    pageDescription,\n    taskBackgroundContext: taskBackgroundContextText,\n  });\n\n  let imagePayload = screenshotBase64;\n  if (vlLocateMode(modelPreferences) === 'qwen-vl') {\n    imagePayload = await paddingToMatchBlockByBase64(imagePayload);\n  } else if (!vlLocateMode(modelPreferences)) {\n    imagePayload = await markupImageForLLM(\n      screenshotBase64,\n      context.tree,\n      context.size,\n    );\n  }\n\n  warnGPT4oSizeLimit(size, modelPreferences);\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: userInstructionPrompt,\n        },\n      ],\n    },\n  ];\n\n  const call = callAI || callAiFn;\n  const { content, usage } = await call(\n    msgs,\n    AIActionType.PLAN,\n    modelPreferences,\n  );\n  const rawResponse = JSON.stringify(content, undefined, 2);\n  const planFromAI = content;\n\n  const actions =\n    (planFromAI.action?.type ? [planFromAI.action] : planFromAI.actions) || [];\n  const returnValue: PlanningAIResponse = {\n    ...planFromAI,\n    actions,\n    rawResponse,\n    usage,\n    yamlFlow: buildYamlFlowFromPlans(\n      actions,\n      opts.actionSpace,\n      planFromAI.sleep,\n    ),\n  };\n\n  assert(planFromAI, \"can't get plans from AI\");\n\n  // TODO: use zod.parse to parse the action.param, and then fill the bbox param.\n  actions.forEach((action) => {\n    const type = action.type;\n    const actionInActionSpace = opts.actionSpace.find(\n      (action) => action.name === type,\n    );\n    const locateFields = actionInActionSpace\n      ? findAllMidsceneLocatorField(actionInActionSpace.paramSchema)\n      : [];\n\n    debug('locateFields', locateFields);\n\n    locateFields.forEach((field) => {\n      const locateResult = action.param[field];\n      if (locateResult) {\n        if (vlLocateMode(modelPreferences)) {\n          action.param[field] = fillBboxParam(\n            locateResult,\n            size.width,\n            size.height,\n            modelPreferences,\n          );\n        } else {\n          const element = elementById(locateResult);\n          if (element) {\n            action.param[field].id = element.id;\n          }\n        }\n      }\n    });\n  });\n  // in Qwen-VL, error means error. In GPT-4o, error may mean more actions are needed.\n  assert(!planFromAI.error, `Failed to plan actions: ${planFromAI.error}`);\n\n  if (\n    actions.length === 0 &&\n    returnValue.more_actions_needed_by_instruction &&\n    !returnValue.sleep\n  ) {\n    console.warn(\n      'No actions planned for the prompt, but model said more actions are needed:',\n      userInstruction,\n    );\n  }\n\n  return returnValue;\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","plan","userInstruction","opts","_planFromAI_action","callAI","context","screenshotBase64","size","modelPreferences","pageDescription","elementById","describeUserPage","systemPrompt","systemPromptToTaskPlanning","vlLocateMode","taskBackgroundContextText","generateTaskBackgroundContext","userInstructionPrompt","automationUserPrompt","imagePayload","paddingToMatchBlockByBase64","markupImageForLLM","warnGPT4oSizeLimit","msgs","call","callAiFn","content","usage","AIActionType","rawResponse","JSON","undefined","planFromAI","actions","returnValue","buildYamlFlowFromPlans","assert","action","type","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam","element","console"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;ACqBA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAEhB,eAAeC,KACpBC,eAAuB,EACvBC,IAOC;QAwEEC;IAtEH,MAAM,EAAEC,MAAM,EAAEC,OAAO,EAAE,GAAGH,QAAQ,CAAC;IACrC,MAAM,EAAEI,gBAAgB,EAAEC,IAAI,EAAE,GAAGF;IAEnC,MAAMG,mBAAsC;QAC1C,QAAQ;IACV;IACA,MAAM,EAAE,aAAaC,eAAe,EAAEC,WAAW,EAAE,GAAG,MAAMC,AAAAA,IAAAA,wBAAAA,gBAAAA,AAAAA,EAC1DN,SACAG;IAGF,MAAMI,eAAe,MAAMC,AAAAA,IAAAA,gCAAAA,0BAAAA,AAAAA,EAA2B;QACpD,aAAaX,KAAK,WAAW;QAC7B,QAAQY,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAaN;IACvB;IACA,MAAMO,4BAA4BC,AAAAA,IAAAA,gCAAAA,6BAAAA,AAAAA,EAChCf,iBACAC,KAAK,GAAG,EACRA,KAAK,aAAa;IAEpB,MAAMe,wBAAwB,MAAMC,AAAAA,IAAAA,gCAAAA,oBAAAA,AAAAA,EAClCJ,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAaN,mBACb,MAAM,CAAC;QACPC;QACA,uBAAuBM;IACzB;IAEA,IAAII,eAAeb;IACnB,IAAIQ,AAAmC,cAAnCA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAaN,mBACfW,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,2BAAAA,AAAAA,EAA4BD;SAC5C,IAAI,CAACL,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAaN,mBACvBW,eAAe,MAAME,AAAAA,IAAAA,mCAAAA,iBAAAA,AAAAA,EACnBf,kBACAD,QAAQ,IAAI,EACZA,QAAQ,IAAI;IAIhBiB,IAAAA,mCAAAA,kBAAAA,AAAAA,EAAmBf,MAAMC;IAEzB,MAAMe,OAAe;QACnB;YAAE,MAAM;YAAU,SAASX;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKO;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMF;gBACR;aACD;QACH;KACD;IAED,MAAMO,OAAOpB,UAAUqB,mCAAAA,QAAQA;IAC/B,MAAM,EAAEC,OAAO,EAAEC,KAAK,EAAE,GAAG,MAAMH,KAC/BD,MACAK,mCAAAA,YAAAA,CAAAA,IAAiB,EACjBpB;IAEF,MAAMqB,cAAcC,KAAK,SAAS,CAACJ,SAASK,QAAW;IACvD,MAAMC,aAAaN;IAEnB,MAAMO,UACH9B,AAAAA,CAAAA,SAAAA,CAAAA,qBAAAA,WAAW,MAAM,AAAD,IAAhBA,KAAAA,IAAAA,mBAAmB,IAAI,AAAD,IAAI;QAAC6B,WAAW,MAAM;KAAC,GAAGA,WAAW,OAAM,KAAM,EAAE;IAC5E,MAAME,cAAkC;QACtC,GAAGF,UAAU;QACbC;QACAJ;QACAF;QACA,UAAUQ,AAAAA,IAAAA,mCAAAA,sBAAAA,AAAAA,EACRF,SACA/B,KAAK,WAAW,EAChB8B,WAAW,KAAK;IAEpB;IAEAI,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOJ,YAAY;IAGnBC,QAAQ,OAAO,CAAC,CAACI;QACf,MAAMC,OAAOD,OAAO,IAAI;QACxB,MAAME,sBAAsBrC,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACmC,SAAWA,OAAO,IAAI,KAAKC;QAE9B,MAAME,eAAeD,sBACjBE,AAAAA,IAAAA,mCAAAA,2BAAAA,AAAAA,EAA4BF,oBAAoB,WAAW,IAC3D,EAAE;QAENzC,MAAM,gBAAgB0C;QAEtBA,aAAa,OAAO,CAAC,CAACE;YACpB,MAAMC,eAAeN,OAAO,KAAK,CAACK,MAAM;YACxC,IAAIC,cACF,IAAI7B,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAaN,mBACf6B,OAAO,KAAK,CAACK,MAAM,GAAGE,AAAAA,IAAAA,mCAAAA,aAAAA,AAAAA,EACpBD,cACApC,KAAK,KAAK,EACVA,KAAK,MAAM,EACXC;iBAEG;gBACL,MAAMqC,UAAUnC,YAAYiC;gBAC5B,IAAIE,SACFR,OAAO,KAAK,CAACK,MAAM,CAAC,EAAE,GAAGG,QAAQ,EAAE;YAEvC;QAEJ;IACF;IAEAT,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,CAACJ,WAAW,KAAK,EAAE,CAAC,wBAAwB,EAAEA,WAAW,KAAK,EAAE;IAEvE,IACEC,AAAmB,MAAnBA,QAAQ,MAAM,IACdC,YAAY,kCAAkC,IAC9C,CAACA,YAAY,KAAK,EAElBY,QAAQ,IAAI,CACV,8EACA7C;IAIJ,OAAOiC;AACT"}

package/dist/lib/ai-model/prompt/llm-planning.js

@@ -42,41 +42,87 @@ __webpack_require__.d(__webpack_exports__, {
 const external_node_assert_namespaceObject = require("node:assert");
 var external_node_assert_default = /*#__PURE__*/ __webpack_require__.n(external_node_assert_namespaceObject);
 const prompts_namespaceObject = require("@langchain/core/prompts");
-const external_common_js_namespaceObject = require("./common.js");
+const external_zod_namespaceObject = require("zod");
+const external_common_js_namespaceObject = require("../common.js");
+const external_common_js_namespaceObject_1 = require("./common.js");
 const vlCoTLog = '"what_the_user_wants_to_do_next_by_instruction": string, // What the user wants to do according to the instruction and previous logs. ';
 const vlCurrentLog = '"log": string, // Log what the next one action (ONLY ONE!) you can do according to the screenshot and the instruction. The typical log looks like "Now i want to use action \'{ action-type }\' to do .. first". If no action should be done, log the reason. ". Use the same language as the user\'s instruction.';
 const llmCurrentLog = '"log": string, // Log what the next actions you can do according to the screenshot and the instruction. The typical log looks like "Now i want to use action \'{ action-type }\' to do ..". If no action should be done, log the reason. ". Use the same language as the user\'s instruction.';
 const commonOutputFields = `"error"?: string, // Error messages about unexpected situations, if any. Only think it is an error when the situation is not foreseeable according to the instruction. Use the same language as the user's instruction.
   "more_actions_needed_by_instruction": boolean, // Consider if there is still more action(s) to do after the action in "Log" is done, according to the instruction. If so, set this field to true. Otherwise, set it to false.`;
-const vlLocateParam = (required)=>`locate${required ? '' : '?'}: {bbox: [number, number, number, number], prompt: string }`;
-const llmLocateParam = (required)=>`locate${required ? '' : '?'}: {"id": string, "prompt": string}`;
-const descriptionForAction = (action, locatorScheme)=>{
+const vlLocateParam = ()=>'{bbox: [number, number, number, number], prompt: string }';
+const llmLocateParam = ()=>'{"id": string, "prompt": string}';
+const descriptionForAction = (action, locatorSchemaTypeDescription)=>{
     const tab = '  ';
-    let locateParam = '';
-    if ('required' === action.location) locateParam = locatorScheme;
-    else if ('optional' === action.location) locateParam = `${locatorScheme} | null`;
-    else if (false === action.location) locateParam = '';
-    const locatorParam = locateParam ? `- ${locateParam}` : '';
-    if (action.whatToLocate) if (locateParam) locateParam += ` // ${action.whatToLocate}`;
-    else console.warn(`whatToLocate is provided for action ${action.name}, but location is not required or optional. The whatToLocate will be ignored.`);
-    let paramSchema = '';
-    if (action.paramSchema) paramSchema = `- param: ${action.paramSchema}`;
-    if (action.paramDescription) {
-        external_node_assert_default()(paramSchema, `paramSchema is required when paramDescription is provided for action ${action.name}, but got ${action.paramSchema}`);
-        paramSchema += ` // ${action.paramDescription}`;
+    const fields = [];
+    fields.push(`- type: "${action.name}"`);
+    if (action.paramSchema) {
+        external_node_assert_default()(action.paramSchema instanceof external_zod_namespaceObject.z.ZodObject, 'paramSchema must be a zod object');
+        const shape = action.paramSchema.shape;
+        const paramLines = [];
+        const getTypeName = (field)=>{
+            var _actualField__def;
+            const unwrapField = (f)=>{
+                if (!f._def) return f;
+                const typeName = f._def.typeName;
+                if ('ZodOptional' === typeName || 'ZodNullable' === typeName || 'ZodDefault' === typeName) return unwrapField(f._def.innerType);
+                return f;
+            };
+            const actualField = unwrapField(field);
+            const fieldTypeName = null == (_actualField__def = actualField._def) ? void 0 : _actualField__def.typeName;
+            if ('ZodString' === fieldTypeName) return 'string';
+            if ('ZodNumber' === fieldTypeName) return 'number';
+            if ('ZodBoolean' === fieldTypeName) return 'boolean';
+            if ('ZodArray' === fieldTypeName) return 'array';
+            if ('ZodObject' === fieldTypeName) {
+                if ((0, external_common_js_namespaceObject.ifMidsceneLocatorField)(actualField)) return locatorSchemaTypeDescription;
+                return 'object';
+            }
+            if ('ZodEnum' === fieldTypeName) {
+                var _actualField__def_values, _actualField__def1;
+                const values = (null == (_actualField__def1 = actualField._def) ? void 0 : null == (_actualField__def_values = _actualField__def1.values) ? void 0 : _actualField__def_values.map((option)=>String(`'${option}'`)).join(', ')) ?? 'enum';
+                return `enum(${values})`;
+            }
+            console.warn('failed to parse Zod type. This may lead to wrong params from the LLM.\n', actualField._def);
+            return actualField.toString();
+        };
+        const getDescription = (field)=>{
+            var _actualField__def;
+            const unwrapField = (f)=>{
+                if (!f._def) return f;
+                const typeName = f._def.typeName;
+                if ('ZodOptional' === typeName || 'ZodNullable' === typeName || 'ZodDefault' === typeName) return unwrapField(f._def.innerType);
+                return f;
+            };
+            if ("description" in field) return field.description || null;
+            const actualField = unwrapField(field);
+            if ("description" in actualField) return actualField.description || null;
+            if ((null == (_actualField__def = actualField._def) ? void 0 : _actualField__def.typeName) === 'ZodObject') {
+                if ('midscene_location_field_flag' in actualField._def.shape()) return 'Location information for the target element';
+            }
+            return null;
+        };
+        for (const [key, field] of Object.entries(shape))if (field && 'object' == typeof field) {
+            const isOptional = 'function' == typeof field.isOptional && field.isOptional();
+            const keyWithOptional = isOptional ? `${key}?` : key;
+            const typeName = getTypeName(field);
+            const description = getDescription(field);
+            let paramLine = `${keyWithOptional}: ${typeName}`;
+            if (description) paramLine += ` // ${description}`;
+            paramLines.push(paramLine);
+        }
+        if (paramLines.length > 0) {
+            fields.push('- param:');
+            for (const paramLine of paramLines)fields.push(`  - ${paramLine}`);
+        }
     }
-    const fields = [
-        paramSchema,
-        locatorParam
-    ].filter(Boolean);
-    return `- ${action.name}, ${action.description}
-${tab}- type: "${action.name}"
+    return `- ${action.name}, ${action.description || "No description provided"}
 ${tab}${fields.join(`\n${tab}`)}
 `.trim();
 };
 const systemTemplateOfVLPlanning = ({ actionSpace, vlMode })=>{
     const actionNameList = actionSpace.map((action)=>action.name).join(', ');
-    const actionDescriptionList = actionSpace.map((action)=>descriptionForAction(action, vlLocateParam('required' === action.location)));
+    const actionDescriptionList = actionSpace.map((action)=>descriptionForAction(action, vlLocateParam()));
     const actionList = actionDescriptionList.join('\n');
     return `
 Target: User will give you a screenshot, an instruction and some previous logs indicating what have been done. Please tell what the next one action is (or null if no action should be done) to do the tasks the instruction requires. 
@@ -85,7 +131,7 @@ Restriction:
 - Don't give extra actions or plans beyond the instruction. ONLY plan for what the instruction requires. For example, don't try to submit the form if the instruction is only to fill something.
 - Always give ONLY ONE action in \`log\` field (or null if no action should be done), instead of multiple actions. Supported actions are ${actionNameList}.
 - Don't repeat actions in the previous logs.
-- Bbox is the bounding box of the element to be located. It's an array of 4 numbers, representing ${(0, external_common_js_namespaceObject.bboxDescription)(vlMode)}.
+- Bbox is the bounding box of the element to be located. It's an array of 4 numbers, representing ${(0, external_common_js_namespaceObject_1.bboxDescription)(vlMode)}.
 
 Supporting actions:
 ${actionList}
@@ -126,7 +172,7 @@ this and output the JSON:
 };
 const systemTemplateOfLLM = ({ actionSpace })=>{
     const actionNameList = actionSpace.map((action)=>action.name).join(' / ');
-    const actionDescriptionList = actionSpace.map((action)=>descriptionForAction(action, llmLocateParam('required' === action.location)));
+    const actionDescriptionList = actionSpace.map((action)=>descriptionForAction(action, llmLocateParam()));
     const actionList = actionDescriptionList.join('\n');
     return `
 ## Role