@midscene/core 0.26.7-beta-20250815153024.0 → 0.26.7-beta-20250818024834.0

package/dist/es/ai-model/common.mjs

@@ -13,7 +13,6 @@ var common_AIActionType = /*#__PURE__*/ function(AIActionType) {
     AIActionType[AIActionType["DESCRIBE_ELEMENT"] = 4] = "DESCRIBE_ELEMENT";
     return AIActionType;
 }({});
-const actionSpaceTypePrefix = 'action_space_';
 async function callAiFn(msgs, AIActionTypeValue) {
     const jsonObject = await callToGetJSONObject(msgs, AIActionTypeValue);
     return {
@@ -170,29 +169,51 @@ async function markupImageForLLM(screenshotBase64, tree, size) {
     });
     return imagePayload;
 }
-function buildYamlFlowFromPlans(plans, actionSpace, sleep) {
+function buildYamlFlowFromPlans(plans, sleep) {
     const flow = [];
     for (const plan of plans){
         var _plan_locate;
-        const verb = plan.type;
-        const action = actionSpace.find((action)=>action.name === verb);
-        if (!action) {
-            console.warn(`Cannot convert action ${verb} to yaml flow. Will ignore it.`);
-            continue;
-        }
+        const type = plan.type;
         const locate = null == (_plan_locate = plan.locate) ? void 0 : _plan_locate.prompt;
-        const flowKey = action.interfaceAlias || `${actionSpaceTypePrefix}${verb}`;
-        const flowItem = {
-            [flowKey]: locate || '',
-            ...plan.param || {}
-        };
-        flow.push(flowItem);
+        if ('Tap' === type) flow.push({
+            aiTap: locate
+        });
+        else if ('Hover' === type) flow.push({
+            aiHover: locate
+        });
+        else if ('Input' === type) {
+            const param = plan.param;
+            flow.push({
+                aiInput: param.value,
+                locate
+            });
+        } else if ('KeyboardPress' === type) {
+            const param = plan.param;
+            flow.push({
+                aiKeyboardPress: param.value,
+                locate
+            });
+        } else if ('Scroll' === type) {
+            const param = plan.param;
+            flow.push({
+                aiScroll: null,
+                locate,
+                direction: param.direction,
+                scrollType: param.scrollType,
+                distance: param.distance
+            });
+        } else if ('Sleep' === type) {
+            const param = plan.param;
+            flow.push({
+                sleep: param.timeMs
+            });
+        } else 'AndroidBackButton' === type || 'AndroidHomeButton' === type || 'AndroidRecentAppsButton' === type || 'AndroidLongPress' === type || 'AndroidPull' === type || 'Error' === type || 'Assert' === type || 'AssertWithoutThrow' === type || 'Finished' === type || console.warn(`Cannot convert action ${type} to yaml flow. This should be a bug of Midscene.`);
     }
     if (sleep) flow.push({
-        sleep
+        sleep: sleep
     });
     return flow;
 }
-export { common_AIActionType as AIActionType, actionSpaceTypePrefix, adaptBbox, adaptBboxToRect, adaptDoubaoBbox, adaptGeminiBbox, adaptQwenBbox, buildYamlFlowFromPlans, callAiFn, expandSearchArea, fillBboxParam, markupImageForLLM, mergeRects, warnGPT4oSizeLimit };
+export { common_AIActionType as AIActionType, adaptBbox, adaptBboxToRect, adaptDoubaoBbox, adaptGeminiBbox, adaptQwenBbox, buildYamlFlowFromPlans, callAiFn, expandSearchArea, fillBboxParam, markupImageForLLM, mergeRects, warnGPT4oSizeLimit };
 
 //# sourceMappingURL=common.mjs.map

package/dist/es/ai-model/common.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/common.mjs","sources":["webpack://@midscene/core/./src/ai-model/common.ts"],"sourcesContent":["import type {\n  AIUsageInfo,\n  BaseElement,\n  DeviceAction,\n  ElementTreeNode,\n  MidsceneYamlFlowItem,\n  PlanningAction,\n  Rect,\n  Size,\n} from '@/types';\nimport { assert } from '@midscene/shared/utils';\n\nimport type {\n  ChatCompletionSystemMessageParam,\n  ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport {\n  call,\n  callToGetJSONObject,\n  getModelName,\n} from './service-caller/index';\n\nimport type { PlanningLocateParam } from '@/types';\nimport { NodeType } from '@midscene/shared/constants';\nimport { vlLocateMode } from '@midscene/shared/env';\nimport { treeToList } from '@midscene/shared/extractor';\nimport { compositeElementInfoImg } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\n\nexport type AIArgs = [\n  ChatCompletionSystemMessageParam,\n  ...ChatCompletionUserMessageParam[],\n];\n\nexport enum AIActionType {\n  ASSERT = 0,\n  INSPECT_ELEMENT = 1,\n  EXTRACT_DATA = 2,\n  PLAN = 3,\n  DESCRIBE_ELEMENT = 4,\n}\n\nexport const actionSpaceTypePrefix = 'action_space_';\n\nexport async function callAiFn<T>(\n  msgs: AIArgs,\n  AIActionTypeValue: AIActionType,\n): Promise<{ content: T; usage?: AIUsageInfo }> {\n  const jsonObject = await callToGetJSONObject<T>(msgs, AIActionTypeValue);\n\n  return {\n    content: jsonObject.content,\n    usage: jsonObject.usage,\n  };\n}\n\nconst defaultBboxSize = 20; // must be even number\nconst debugInspectUtils = getDebug('ai:common');\n\n// transform the param of locate from qwen mode\nexport function fillBboxParam(\n  locate: PlanningLocateParam,\n  width: number,\n  height: number,\n) {\n  // The Qwen model might have hallucinations of naming bbox as bbox_2d.\n  if ((locate as any).bbox_2d && !locate?.bbox) {\n    locate.bbox = (locate as any).bbox_2d;\n    // biome-ignore lint/performance/noDelete: <explanation>\n    delete (locate as any).bbox_2d;\n  }\n\n  if (locate?.bbox) {\n    locate.bbox = adaptBbox(locate.bbox, width, height);\n  }\n\n  return locate;\n}\n\nexport function adaptQwenBbox(\n  bbox: number[],\n): [number, number, number, number] {\n  if (bbox.length < 2) {\n    const msg = `invalid bbox data for qwen-vl mode: ${JSON.stringify(bbox)} `;\n    throw new Error(msg);\n  }\n\n  const result: [number, number, number, number] = [\n    Math.round(bbox[0]),\n    Math.round(bbox[1]),\n    typeof bbox[2] === 'number'\n      ? Math.round(bbox[2])\n      : Math.round(bbox[0] + defaultBboxSize),\n    typeof bbox[3] === 'number'\n      ? Math.round(bbox[3])\n      : Math.round(bbox[1] + defaultBboxSize),\n  ];\n  return result;\n}\n\nexport function adaptDoubaoBbox(\n  bbox: string[] | number[] | string,\n  width: number,\n  height: number,\n): [number, number, number, number] {\n  assert(\n    width > 0 && height > 0,\n    'width and height must be greater than 0 in doubao mode',\n  );\n\n  if (typeof bbox === 'string') {\n    assert(\n      /^(\\d+)\\s(\\d+)\\s(\\d+)\\s(\\d+)$/.test(bbox.trim()),\n      `invalid bbox data string for doubao-vision mode: ${bbox}`,\n    );\n    const splitted = bbox.split(' ');\n    if (splitted.length === 4) {\n      return [\n        Math.round((Number(splitted[0]) * width) / 1000),\n        Math.round((Number(splitted[1]) * height) / 1000),\n        Math.round((Number(splitted[2]) * width) / 1000),\n        Math.round((Number(splitted[3]) * height) / 1000),\n      ];\n    }\n    throw new Error(`invalid bbox data string for doubao-vision mode: ${bbox}`);\n  }\n\n  if (Array.isArray(bbox) && Array.isArray(bbox[0])) {\n    bbox = bbox[0];\n  }\n\n  let bboxList: number[] = [];\n  if (Array.isArray(bbox) && typeof bbox[0] === 'string') {\n    bbox.forEach((item) => {\n      if (typeof item === 'string' && item.includes(',')) {\n        const [x, y] = item.split(',');\n        bboxList.push(Number(x.trim()), Number(y.trim()));\n      } else if (typeof item === 'string' && item.includes(' ')) {\n        const [x, y] = item.split(' ');\n        bboxList.push(Number(x.trim()), Number(y.trim()));\n      } else {\n        bboxList.push(Number(item));\n      }\n    });\n  } else {\n    bboxList = bbox as any;\n  }\n\n  if (bboxList.length === 4 || bboxList.length === 5) {\n    return [\n      Math.round((bboxList[0] * width) / 1000),\n      Math.round((bboxList[1] * height) / 1000),\n      Math.round((bboxList[2] * width) / 1000),\n      Math.round((bboxList[3] * height) / 1000),\n    ];\n  }\n\n  // treat the bbox as a center point\n  if (\n    bboxList.length === 6 ||\n    bboxList.length === 2 ||\n    bboxList.length === 3 ||\n    bboxList.length === 7\n  ) {\n    return [\n      Math.max(\n        0,\n        Math.round((bboxList[0] * width) / 1000) - defaultBboxSize / 2,\n      ),\n      Math.max(\n        0,\n        Math.round((bboxList[1] * height) / 1000) - defaultBboxSize / 2,\n      ),\n      Math.min(\n        width,\n        Math.round((bboxList[0] * width) / 1000) + defaultBboxSize / 2,\n      ),\n      Math.min(\n        height,\n        Math.round((bboxList[1] * height) / 1000) + defaultBboxSize / 2,\n      ),\n    ];\n  }\n\n  if (bbox.length === 8) {\n    return [\n      Math.round((bboxList[0] * width) / 1000),\n      Math.round((bboxList[1] * height) / 1000),\n      Math.round((bboxList[4] * width) / 1000),\n      Math.round((bboxList[5] * height) / 1000),\n    ];\n  }\n\n  const msg = `invalid bbox data for doubao-vision mode: ${JSON.stringify(bbox)} `;\n  throw new Error(msg);\n}\n\nexport function adaptBbox(\n  bbox: number[],\n  width: number,\n  height: number,\n): [number, number, number, number] {\n  if (vlLocateMode() === 'doubao-vision' || vlLocateMode() === 'vlm-ui-tars') {\n    return adaptDoubaoBbox(bbox, width, height);\n  }\n\n  if (vlLocateMode() === 'gemini') {\n    return adaptGeminiBbox(bbox, width, height);\n  }\n\n  return adaptQwenBbox(bbox);\n}\n\nexport function adaptGeminiBbox(\n  bbox: number[],\n  width: number,\n  height: number,\n): [number, number, number, number] {\n  const left = Math.round((bbox[1] * width) / 1000);\n  const top = Math.round((bbox[0] * height) / 1000);\n  const right = Math.round((bbox[3] * width) / 1000);\n  const bottom = Math.round((bbox[2] * height) / 1000);\n  return [left, top, right, bottom];\n}\n\nexport function adaptBboxToRect(\n  bbox: number[],\n  width: number,\n  height: number,\n  offsetX = 0,\n  offsetY = 0,\n): Rect {\n  debugInspectUtils('adaptBboxToRect', bbox, width, height, offsetX, offsetY);\n  const [left, top, right, bottom] = adaptBbox(bbox, width, height);\n  const rect = {\n    left: left + offsetX,\n    top: top + offsetY,\n    width: right - left,\n    height: bottom - top,\n  };\n  debugInspectUtils('adaptBboxToRect, result=', rect);\n  return rect;\n}\n\nlet warned = false;\nexport function warnGPT4oSizeLimit(size: Size) {\n  if (warned) return;\n  if (getModelName()?.toLowerCase().includes('gpt-4o')) {\n    const warningMsg = `GPT-4o has a maximum image input size of 2000x768 or 768x2000, but got ${size.width}x${size.height}. Please set your page to a smaller resolution. Otherwise, the result may be inaccurate.`;\n\n    if (\n      Math.max(size.width, size.height) > 2000 ||\n      Math.min(size.width, size.height) > 768\n    ) {\n      console.warn(warningMsg);\n      warned = true;\n    }\n  } else if (size.width > 1800 || size.height > 1800) {\n    console.warn(\n      `The image size seems too large (${size.width}x${size.height}). It may lead to more token usage, slower response, and inaccurate result.`,\n    );\n    warned = true;\n  }\n}\n\nexport function mergeRects(rects: Rect[]) {\n  const minLeft = Math.min(...rects.map((r) => r.left));\n  const minTop = Math.min(...rects.map((r) => r.top));\n  const maxRight = Math.max(...rects.map((r) => r.left + r.width));\n  const maxBottom = Math.max(...rects.map((r) => r.top + r.height));\n  return {\n    left: minLeft,\n    top: minTop,\n    width: maxRight - minLeft,\n    height: maxBottom - minTop,\n  };\n}\n\n// expand the search area to at least 300 x 300, or add a default padding\nexport function expandSearchArea(rect: Rect, screenSize: Size) {\n  const minEdgeSize = vlLocateMode() === 'doubao-vision' ? 500 : 300;\n  const defaultPadding = 160;\n\n  const paddingSizeHorizontal =\n    rect.width < minEdgeSize\n      ? Math.ceil((minEdgeSize - rect.width) / 2)\n      : defaultPadding;\n  const paddingSizeVertical =\n    rect.height < minEdgeSize\n      ? Math.ceil((minEdgeSize - rect.height) / 2)\n      : defaultPadding;\n  rect.left = Math.max(0, rect.left - paddingSizeHorizontal);\n  rect.width = Math.min(\n    rect.width + paddingSizeHorizontal * 2,\n    screenSize.width - rect.left,\n  );\n  rect.top = Math.max(0, rect.top - paddingSizeVertical);\n  rect.height = Math.min(\n    rect.height + paddingSizeVertical * 2,\n    screenSize.height - rect.top,\n  );\n  return rect;\n}\n\nexport async function markupImageForLLM(\n  screenshotBase64: string,\n  tree: ElementTreeNode<BaseElement>,\n  size: Size,\n) {\n  const elementsInfo = treeToList(tree);\n  const elementsPositionInfoWithoutText = elementsInfo!.filter(\n    (elementInfo) => {\n      if (elementInfo.attributes.nodeType === NodeType.TEXT) {\n        return false;\n      }\n      return true;\n    },\n  );\n\n  const imagePayload = await compositeElementInfoImg({\n    inputImgBase64: screenshotBase64,\n    elementsPositionInfo: elementsPositionInfoWithoutText,\n    size,\n  });\n  return imagePayload;\n}\n\nexport function buildYamlFlowFromPlans(\n  plans: PlanningAction[],\n  actionSpace: DeviceAction[],\n  sleep?: number,\n): MidsceneYamlFlowItem[] {\n  const flow: MidsceneYamlFlowItem[] = [];\n\n  for (const plan of plans) {\n    const verb = plan.type;\n\n    const action = actionSpace.find((action) => action.name === verb);\n    if (!action) {\n      console.warn(\n        `Cannot convert action ${verb} to yaml flow. Will ignore it.`,\n      );\n      continue;\n    }\n\n    const locate = plan.locate?.prompt;\n    const flowKey = action.interfaceAlias || `${actionSpaceTypePrefix}${verb}`;\n\n    const flowItem: MidsceneYamlFlowItem = {\n      [flowKey]: locate || '',\n      ...(plan.param || {}),\n    };\n\n    flow.push(flowItem);\n  }\n\n  if (sleep) {\n    flow.push({\n      sleep,\n    });\n  }\n\n  return flow;\n}\n"],"names":["AIActionType","actionSpaceTypePrefix","callAiFn","msgs","AIActionTypeValue","jsonObject","callToGetJSONObject","defaultBboxSize","debugInspectUtils","getDebug","fillBboxParam","locate","width","height","adaptBbox","adaptQwenBbox","bbox","msg","JSON","Error","result","Math","adaptDoubaoBbox","assert","splitted","Number","Array","bboxList","item","x","y","vlLocateMode","adaptGeminiBbox","left","top","right","bottom","adaptBboxToRect","offsetX","offsetY","rect","warned","warnGPT4oSizeLimit","size","_getModelName","warningMsg","console","mergeRects","rects","minLeft","r","minTop","maxRight","maxBottom","expandSearchArea","screenSize","minEdgeSize","defaultPadding","paddingSizeHorizontal","paddingSizeVertical","markupImageForLLM","screenshotBase64","tree","elementsInfo","treeToList","elementsPositionInfoWithoutText","elementInfo","NodeType","imagePayload","compositeElementInfoImg","buildYamlFlowFromPlans","plans","actionSpace","sleep","flow","plan","_plan_locate","verb","action","flowKey","flowItem"],"mappings":";;;;;;;AAkCO,IAAKA,sBAAYA,WAAAA,GAAAA,SAAZA,YAAY;;;;;;WAAZA;;AAQL,MAAMC,wBAAwB;AAE9B,eAAeC,SACpBC,IAAY,EACZC,iBAA+B;IAE/B,MAAMC,aAAa,MAAMC,oBAAuBH,MAAMC;IAEtD,OAAO;QACL,SAASC,WAAW,OAAO;QAC3B,OAAOA,WAAW,KAAK;IACzB;AACF;AAEA,MAAME,kBAAkB;AACxB,MAAMC,oBAAoBC,SAAS;AAG5B,SAASC,cACdC,MAA2B,EAC3BC,KAAa,EACbC,MAAc;IAGd,IAAKF,OAAe,OAAO,IAAI,CAACA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,IAAI,AAAD,GAAG;QAC5CA,OAAO,IAAI,GAAIA,OAAe,OAAO;QAErC,OAAQA,OAAe,OAAO;IAChC;IAEA,IAAIA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,IAAI,EACdA,OAAO,IAAI,GAAGG,UAAUH,OAAO,IAAI,EAAEC,OAAOC;IAG9C,OAAOF;AACT;AAEO,SAASI,cACdC,IAAc;IAEd,IAAIA,KAAK,MAAM,GAAG,GAAG;QACnB,MAAMC,MAAM,CAAC,oCAAoC,EAAEC,KAAK,SAAS,CAACF,MAAM,CAAC,CAAC;QAC1E,MAAM,IAAIG,MAAMF;IAClB;IAEA,MAAMG,SAA2C;QAC/CC,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE;QAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE;QACC,YAAnB,OAAOA,IAAI,CAAC,EAAE,GACVK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,IAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,GAAGT;QACN,YAAnB,OAAOS,IAAI,CAAC,EAAE,GACVK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,IAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,GAAGT;KAC1B;IACD,OAAOa;AACT;AAEO,SAASE,gBACdN,IAAkC,EAClCJ,KAAa,EACbC,MAAc;IAEdU,OACEX,QAAQ,KAAKC,SAAS,GACtB;IAGF,IAAI,AAAgB,YAAhB,OAAOG,MAAmB;QAC5BO,OACE,+BAA+B,IAAI,CAACP,KAAK,IAAI,KAC7C,CAAC,iDAAiD,EAAEA,MAAM;QAE5D,MAAMQ,WAAWR,KAAK,KAAK,CAAC;QAC5B,IAAIQ,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACLH,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIZ,QAAS;YAC3CS,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIX,SAAU;YAC5CQ,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIZ,QAAS;YAC3CS,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIX,SAAU;SAC7C;QAEH,MAAM,IAAIM,MAAM,CAAC,iDAAiD,EAAEH,MAAM;IAC5E;IAEA,IAAIU,MAAM,OAAO,CAACV,SAASU,MAAM,OAAO,CAACV,IAAI,CAAC,EAAE,GAC9CA,OAAOA,IAAI,CAAC,EAAE;IAGhB,IAAIW,WAAqB,EAAE;IAC3B,IAAID,MAAM,OAAO,CAACV,SAAS,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EACvCA,KAAK,OAAO,CAAC,CAACY;QACZ,IAAI,AAAgB,YAAhB,OAAOA,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YAClD,MAAM,CAACC,GAAGC,EAAE,GAAGF,KAAK,KAAK,CAAC;YAC1BD,SAAS,IAAI,CAACF,OAAOI,EAAE,IAAI,KAAKJ,OAAOK,EAAE,IAAI;QAC/C,OAAO,IAAI,AAAgB,YAAhB,OAAOF,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YACzD,MAAM,CAACC,GAAGC,EAAE,GAAGF,KAAK,KAAK,CAAC;YAC1BD,SAAS,IAAI,CAACF,OAAOI,EAAE,IAAI,KAAKJ,OAAOK,EAAE,IAAI;QAC/C,OACEH,SAAS,IAAI,CAACF,OAAOG;IAEzB;SAEAD,WAAWX;IAGb,IAAIW,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAO;QACLN,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGf,QAAS;QACnCS,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGd,SAAU;QACpCQ,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGf,QAAS;QACnCS,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGd,SAAU;KACrC;IAIH,IACEc,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAO;QACLN,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGf,QAAS,QAAQL,kBAAkB;QAE/Dc,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGd,SAAU,QAAQN,kBAAkB;QAEhEc,KAAK,GAAG,CACNT,OACAS,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGf,QAAS,QAAQL,kBAAkB;QAE/Dc,KAAK,GAAG,CACNR,QACAQ,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGd,SAAU,QAAQN,kBAAkB;KAEjE;IAGH,IAAIS,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAO;QACLK,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGf,QAAS;QACnCS,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGd,SAAU;QACpCQ,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGf,QAAS;QACnCS,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGd,SAAU;KACrC;IAGH,MAAMI,MAAM,CAAC,0CAA0C,EAAEC,KAAK,SAAS,CAACF,MAAM,CAAC,CAAC;IAChF,MAAM,IAAIG,MAAMF;AAClB;AAEO,SAASH,UACdE,IAAc,EACdJ,KAAa,EACbC,MAAc;IAEd,IAAIkB,AAAmB,oBAAnBA,kBAAsCA,AAAmB,kBAAnBA,gBACxC,OAAOT,gBAAgBN,MAAMJ,OAAOC;IAGtC,IAAIkB,AAAmB,aAAnBA,gBACF,OAAOC,gBAAgBhB,MAAMJ,OAAOC;IAGtC,OAAOE,cAAcC;AACvB;AAEO,SAASgB,gBACdhB,IAAc,EACdJ,KAAa,EACbC,MAAc;IAEd,MAAMoB,OAAOZ,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGJ,QAAS;IAC5C,MAAMsB,MAAMb,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGH,SAAU;IAC5C,MAAMsB,QAAQd,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGJ,QAAS;IAC7C,MAAMwB,SAASf,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGH,SAAU;IAC/C,OAAO;QAACoB;QAAMC;QAAKC;QAAOC;KAAO;AACnC;AAEO,SAASC,gBACdrB,IAAc,EACdJ,KAAa,EACbC,MAAc,EACdyB,UAAU,CAAC,EACXC,UAAU,CAAC;IAEX/B,kBAAkB,mBAAmBQ,MAAMJ,OAAOC,QAAQyB,SAASC;IACnE,MAAM,CAACN,MAAMC,KAAKC,OAAOC,OAAO,GAAGtB,UAAUE,MAAMJ,OAAOC;IAC1D,MAAM2B,OAAO;QACX,MAAMP,OAAOK;QACb,KAAKJ,MAAMK;QACX,OAAOJ,QAAQF;QACf,QAAQG,SAASF;IACnB;IACA1B,kBAAkB,4BAA4BgC;IAC9C,OAAOA;AACT;AAEA,IAAIC,SAAS;AACN,SAASC,mBAAmBC,IAAU;QAEvCC;IADJ,IAAIH,QAAQ;IACZ,IAAI,QAAAG,CAAAA,gBAAAA,cAAa,IAAbA,KAAAA,IAAAA,cAAgB,WAAW,GAAG,QAAQ,CAAC,WAAW;QACpD,MAAMC,aAAa,CAAC,uEAAuE,EAAEF,KAAK,KAAK,CAAC,CAAC,EAAEA,KAAK,MAAM,CAAC,wFAAwF,CAAC;QAEhN,IACEtB,KAAK,GAAG,CAACsB,KAAK,KAAK,EAAEA,KAAK,MAAM,IAAI,QACpCtB,KAAK,GAAG,CAACsB,KAAK,KAAK,EAAEA,KAAK,MAAM,IAAI,KACpC;YACAG,QAAQ,IAAI,CAACD;YACbJ,SAAS;QACX;IACF,OAAO,IAAIE,KAAK,KAAK,GAAG,QAAQA,KAAK,MAAM,GAAG,MAAM;QAClDG,QAAQ,IAAI,CACV,CAAC,gCAAgC,EAAEH,KAAK,KAAK,CAAC,CAAC,EAAEA,KAAK,MAAM,CAAC,2EAA2E,CAAC;QAE3IF,SAAS;IACX;AACF;AAEO,SAASM,WAAWC,KAAa;IACtC,MAAMC,UAAU5B,KAAK,GAAG,IAAI2B,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,IAAI;IACnD,MAAMC,SAAS9B,KAAK,GAAG,IAAI2B,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,GAAG;IACjD,MAAME,WAAW/B,KAAK,GAAG,IAAI2B,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,IAAI,GAAGA,EAAE,KAAK;IAC9D,MAAMG,YAAYhC,KAAK,GAAG,IAAI2B,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,GAAG,GAAGA,EAAE,MAAM;IAC/D,OAAO;QACL,MAAMD;QACN,KAAKE;QACL,OAAOC,WAAWH;QAClB,QAAQI,YAAYF;IACtB;AACF;AAGO,SAASG,iBAAiBd,IAAU,EAAEe,UAAgB;IAC3D,MAAMC,cAAczB,AAAmB,oBAAnBA,iBAAqC,MAAM;IAC/D,MAAM0B,iBAAiB;IAEvB,MAAMC,wBACJlB,KAAK,KAAK,GAAGgB,cACTnC,KAAK,IAAI,CAAEmC,AAAAA,CAAAA,cAAchB,KAAK,KAAI,IAAK,KACvCiB;IACN,MAAME,sBACJnB,KAAK,MAAM,GAAGgB,cACVnC,KAAK,IAAI,CAAEmC,AAAAA,CAAAA,cAAchB,KAAK,MAAK,IAAK,KACxCiB;IACNjB,KAAK,IAAI,GAAGnB,KAAK,GAAG,CAAC,GAAGmB,KAAK,IAAI,GAAGkB;IACpClB,KAAK,KAAK,GAAGnB,KAAK,GAAG,CACnBmB,KAAK,KAAK,GAAGkB,AAAwB,IAAxBA,uBACbH,WAAW,KAAK,GAAGf,KAAK,IAAI;IAE9BA,KAAK,GAAG,GAAGnB,KAAK,GAAG,CAAC,GAAGmB,KAAK,GAAG,GAAGmB;IAClCnB,KAAK,MAAM,GAAGnB,KAAK,GAAG,CACpBmB,KAAK,MAAM,GAAGmB,AAAsB,IAAtBA,qBACdJ,WAAW,MAAM,GAAGf,KAAK,GAAG;IAE9B,OAAOA;AACT;AAEO,eAAeoB,kBACpBC,gBAAwB,EACxBC,IAAkC,EAClCnB,IAAU;IAEV,MAAMoB,eAAeC,WAAWF;IAChC,MAAMG,kCAAkCF,aAAc,MAAM,CAC1D,CAACG;QACC,IAAIA,YAAY,UAAU,CAAC,QAAQ,KAAKC,SAAS,IAAI,EACnD,OAAO;QAET,OAAO;IACT;IAGF,MAAMC,eAAe,MAAMC,wBAAwB;QACjD,gBAAgBR;QAChB,sBAAsBI;QACtBtB;IACF;IACA,OAAOyB;AACT;AAEO,SAASE,uBACdC,KAAuB,EACvBC,WAA2B,EAC3BC,KAAc;IAEd,MAAMC,OAA+B,EAAE;IAEvC,KAAK,MAAMC,QAAQJ,MAAO;YAWTK;QAVf,MAAMC,OAAOF,KAAK,IAAI;QAEtB,MAAMG,SAASN,YAAY,IAAI,CAAC,CAACM,SAAWA,OAAO,IAAI,KAAKD;QAC5D,IAAI,CAACC,QAAQ;YACXhC,QAAQ,IAAI,CACV,CAAC,sBAAsB,EAAE+B,KAAK,8BAA8B,CAAC;YAE/D;QACF;QAEA,MAAMlE,SAAS,QAAAiE,CAAAA,eAAAA,KAAK,MAAM,AAAD,IAAVA,KAAAA,IAAAA,aAAa,MAAM;QAClC,MAAMG,UAAUD,OAAO,cAAc,IAAI,GAAG7E,wBAAwB4E,MAAM;QAE1E,MAAMG,WAAiC;YACrC,CAACD,QAAQ,EAAEpE,UAAU;YACrB,GAAIgE,KAAK,KAAK,IAAI,CAAC,CAAC;QACtB;QAEAD,KAAK,IAAI,CAACM;IACZ;IAEA,IAAIP,OACFC,KAAK,IAAI,CAAC;QACRD;IACF;IAGF,OAAOC;AACT"}
+{"version":3,"file":"ai-model/common.mjs","sources":["webpack://@midscene/core/./src/ai-model/common.ts"],"sourcesContent":["import type {\n  AIUsageInfo,\n  BaseElement,\n  ElementTreeNode,\n  MidsceneYamlFlowItem,\n  PlanningAction,\n  PlanningActionParamInputOrKeyPress,\n  PlanningActionParamSleep,\n  Rect,\n  ScrollParam,\n  Size,\n} from '@/types';\nimport { assert } from '@midscene/shared/utils';\n\nimport type {\n  ChatCompletionSystemMessageParam,\n  ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport {\n  call,\n  callToGetJSONObject,\n  getModelName,\n} from './service-caller/index';\n\nimport type { PlanningLocateParam } from '@/types';\nimport { NodeType } from '@midscene/shared/constants';\nimport { vlLocateMode } from '@midscene/shared/env';\nimport { treeToList } from '@midscene/shared/extractor';\nimport { compositeElementInfoImg } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\n\nexport type AIArgs = [\n  ChatCompletionSystemMessageParam,\n  ...ChatCompletionUserMessageParam[],\n];\n\nexport enum AIActionType {\n  ASSERT = 0,\n  INSPECT_ELEMENT = 1,\n  EXTRACT_DATA = 2,\n  PLAN = 3,\n  DESCRIBE_ELEMENT = 4,\n}\n\nexport async function callAiFn<T>(\n  msgs: AIArgs,\n  AIActionTypeValue: AIActionType,\n): Promise<{ content: T; usage?: AIUsageInfo }> {\n  const jsonObject = await callToGetJSONObject<T>(msgs, AIActionTypeValue);\n\n  return {\n    content: jsonObject.content,\n    usage: jsonObject.usage,\n  };\n}\n\nconst defaultBboxSize = 20; // must be even number\nconst debugInspectUtils = getDebug('ai:common');\n\n// transform the param of locate from qwen mode\nexport function fillBboxParam(\n  locate: PlanningLocateParam,\n  width: number,\n  height: number,\n) {\n  // The Qwen model might have hallucinations of naming bbox as bbox_2d.\n  if ((locate as any).bbox_2d && !locate?.bbox) {\n    locate.bbox = (locate as any).bbox_2d;\n    // biome-ignore lint/performance/noDelete: <explanation>\n    delete (locate as any).bbox_2d;\n  }\n\n  if (locate?.bbox) {\n    locate.bbox = adaptBbox(locate.bbox, width, height);\n  }\n\n  return locate;\n}\n\nexport function adaptQwenBbox(\n  bbox: number[],\n): [number, number, number, number] {\n  if (bbox.length < 2) {\n    const msg = `invalid bbox data for qwen-vl mode: ${JSON.stringify(bbox)} `;\n    throw new Error(msg);\n  }\n\n  const result: [number, number, number, number] = [\n    Math.round(bbox[0]),\n    Math.round(bbox[1]),\n    typeof bbox[2] === 'number'\n      ? Math.round(bbox[2])\n      : Math.round(bbox[0] + defaultBboxSize),\n    typeof bbox[3] === 'number'\n      ? Math.round(bbox[3])\n      : Math.round(bbox[1] + defaultBboxSize),\n  ];\n  return result;\n}\n\nexport function adaptDoubaoBbox(\n  bbox: string[] | number[] | string,\n  width: number,\n  height: number,\n): [number, number, number, number] {\n  assert(\n    width > 0 && height > 0,\n    'width and height must be greater than 0 in doubao mode',\n  );\n\n  if (typeof bbox === 'string') {\n    assert(\n      /^(\\d+)\\s(\\d+)\\s(\\d+)\\s(\\d+)$/.test(bbox.trim()),\n      `invalid bbox data string for doubao-vision mode: ${bbox}`,\n    );\n    const splitted = bbox.split(' ');\n    if (splitted.length === 4) {\n      return [\n        Math.round((Number(splitted[0]) * width) / 1000),\n        Math.round((Number(splitted[1]) * height) / 1000),\n        Math.round((Number(splitted[2]) * width) / 1000),\n        Math.round((Number(splitted[3]) * height) / 1000),\n      ];\n    }\n    throw new Error(`invalid bbox data string for doubao-vision mode: ${bbox}`);\n  }\n\n  if (Array.isArray(bbox) && Array.isArray(bbox[0])) {\n    bbox = bbox[0];\n  }\n\n  let bboxList: number[] = [];\n  if (Array.isArray(bbox) && typeof bbox[0] === 'string') {\n    bbox.forEach((item) => {\n      if (typeof item === 'string' && item.includes(',')) {\n        const [x, y] = item.split(',');\n        bboxList.push(Number(x.trim()), Number(y.trim()));\n      } else if (typeof item === 'string' && item.includes(' ')) {\n        const [x, y] = item.split(' ');\n        bboxList.push(Number(x.trim()), Number(y.trim()));\n      } else {\n        bboxList.push(Number(item));\n      }\n    });\n  } else {\n    bboxList = bbox as any;\n  }\n\n  if (bboxList.length === 4 || bboxList.length === 5) {\n    return [\n      Math.round((bboxList[0] * width) / 1000),\n      Math.round((bboxList[1] * height) / 1000),\n      Math.round((bboxList[2] * width) / 1000),\n      Math.round((bboxList[3] * height) / 1000),\n    ];\n  }\n\n  // treat the bbox as a center point\n  if (\n    bboxList.length === 6 ||\n    bboxList.length === 2 ||\n    bboxList.length === 3 ||\n    bboxList.length === 7\n  ) {\n    return [\n      Math.max(\n        0,\n        Math.round((bboxList[0] * width) / 1000) - defaultBboxSize / 2,\n      ),\n      Math.max(\n        0,\n        Math.round((bboxList[1] * height) / 1000) - defaultBboxSize / 2,\n      ),\n      Math.min(\n        width,\n        Math.round((bboxList[0] * width) / 1000) + defaultBboxSize / 2,\n      ),\n      Math.min(\n        height,\n        Math.round((bboxList[1] * height) / 1000) + defaultBboxSize / 2,\n      ),\n    ];\n  }\n\n  if (bbox.length === 8) {\n    return [\n      Math.round((bboxList[0] * width) / 1000),\n      Math.round((bboxList[1] * height) / 1000),\n      Math.round((bboxList[4] * width) / 1000),\n      Math.round((bboxList[5] * height) / 1000),\n    ];\n  }\n\n  const msg = `invalid bbox data for doubao-vision mode: ${JSON.stringify(bbox)} `;\n  throw new Error(msg);\n}\n\nexport function adaptBbox(\n  bbox: number[],\n  width: number,\n  height: number,\n): [number, number, number, number] {\n  if (vlLocateMode() === 'doubao-vision' || vlLocateMode() === 'vlm-ui-tars') {\n    return adaptDoubaoBbox(bbox, width, height);\n  }\n\n  if (vlLocateMode() === 'gemini') {\n    return adaptGeminiBbox(bbox, width, height);\n  }\n\n  return adaptQwenBbox(bbox);\n}\n\nexport function adaptGeminiBbox(\n  bbox: number[],\n  width: number,\n  height: number,\n): [number, number, number, number] {\n  const left = Math.round((bbox[1] * width) / 1000);\n  const top = Math.round((bbox[0] * height) / 1000);\n  const right = Math.round((bbox[3] * width) / 1000);\n  const bottom = Math.round((bbox[2] * height) / 1000);\n  return [left, top, right, bottom];\n}\n\nexport function adaptBboxToRect(\n  bbox: number[],\n  width: number,\n  height: number,\n  offsetX = 0,\n  offsetY = 0,\n): Rect {\n  debugInspectUtils('adaptBboxToRect', bbox, width, height, offsetX, offsetY);\n  const [left, top, right, bottom] = adaptBbox(bbox, width, height);\n  const rect = {\n    left: left + offsetX,\n    top: top + offsetY,\n    width: right - left,\n    height: bottom - top,\n  };\n  debugInspectUtils('adaptBboxToRect, result=', rect);\n  return rect;\n}\n\nlet warned = false;\nexport function warnGPT4oSizeLimit(size: Size) {\n  if (warned) return;\n  if (getModelName()?.toLowerCase().includes('gpt-4o')) {\n    const warningMsg = `GPT-4o has a maximum image input size of 2000x768 or 768x2000, but got ${size.width}x${size.height}. Please set your page to a smaller resolution. Otherwise, the result may be inaccurate.`;\n\n    if (\n      Math.max(size.width, size.height) > 2000 ||\n      Math.min(size.width, size.height) > 768\n    ) {\n      console.warn(warningMsg);\n      warned = true;\n    }\n  } else if (size.width > 1800 || size.height > 1800) {\n    console.warn(\n      `The image size seems too large (${size.width}x${size.height}). It may lead to more token usage, slower response, and inaccurate result.`,\n    );\n    warned = true;\n  }\n}\n\nexport function mergeRects(rects: Rect[]) {\n  const minLeft = Math.min(...rects.map((r) => r.left));\n  const minTop = Math.min(...rects.map((r) => r.top));\n  const maxRight = Math.max(...rects.map((r) => r.left + r.width));\n  const maxBottom = Math.max(...rects.map((r) => r.top + r.height));\n  return {\n    left: minLeft,\n    top: minTop,\n    width: maxRight - minLeft,\n    height: maxBottom - minTop,\n  };\n}\n\n// expand the search area to at least 300 x 300, or add a default padding\nexport function expandSearchArea(rect: Rect, screenSize: Size) {\n  const minEdgeSize = vlLocateMode() === 'doubao-vision' ? 500 : 300;\n  const defaultPadding = 160;\n\n  const paddingSizeHorizontal =\n    rect.width < minEdgeSize\n      ? Math.ceil((minEdgeSize - rect.width) / 2)\n      : defaultPadding;\n  const paddingSizeVertical =\n    rect.height < minEdgeSize\n      ? Math.ceil((minEdgeSize - rect.height) / 2)\n      : defaultPadding;\n  rect.left = Math.max(0, rect.left - paddingSizeHorizontal);\n  rect.width = Math.min(\n    rect.width + paddingSizeHorizontal * 2,\n    screenSize.width - rect.left,\n  );\n  rect.top = Math.max(0, rect.top - paddingSizeVertical);\n  rect.height = Math.min(\n    rect.height + paddingSizeVertical * 2,\n    screenSize.height - rect.top,\n  );\n  return rect;\n}\n\nexport async function markupImageForLLM(\n  screenshotBase64: string,\n  tree: ElementTreeNode<BaseElement>,\n  size: Size,\n) {\n  const elementsInfo = treeToList(tree);\n  const elementsPositionInfoWithoutText = elementsInfo!.filter(\n    (elementInfo) => {\n      if (elementInfo.attributes.nodeType === NodeType.TEXT) {\n        return false;\n      }\n      return true;\n    },\n  );\n\n  const imagePayload = await compositeElementInfoImg({\n    inputImgBase64: screenshotBase64,\n    elementsPositionInfo: elementsPositionInfoWithoutText,\n    size,\n  });\n  return imagePayload;\n}\n\nexport function buildYamlFlowFromPlans(\n  plans: PlanningAction[],\n  sleep?: number,\n): MidsceneYamlFlowItem[] {\n  const flow: MidsceneYamlFlowItem[] = [];\n\n  for (const plan of plans) {\n    const type = plan.type;\n    const locate = plan.locate?.prompt!; // TODO: check if locate is null\n\n    if (type === 'Tap') {\n      flow.push({\n        aiTap: locate!,\n      });\n    } else if (type === 'Hover') {\n      flow.push({\n        aiHover: locate!,\n      });\n    } else if (type === 'Input') {\n      const param = plan.param as PlanningActionParamInputOrKeyPress;\n      flow.push({\n        aiInput: param.value,\n        locate,\n      });\n    } else if (type === 'KeyboardPress') {\n      const param = plan.param as PlanningActionParamInputOrKeyPress;\n      flow.push({\n        aiKeyboardPress: param.value,\n        locate,\n      });\n    } else if (type === 'Scroll') {\n      const param = plan.param as ScrollParam;\n      flow.push({\n        aiScroll: null,\n        locate,\n        direction: param.direction,\n        scrollType: param.scrollType,\n        distance: param.distance,\n      });\n    } else if (type === 'Sleep') {\n      const param = plan.param as PlanningActionParamSleep;\n      flow.push({\n        sleep: param.timeMs,\n      });\n    } else if (\n      type === 'AndroidBackButton' ||\n      type === 'AndroidHomeButton' ||\n      type === 'AndroidRecentAppsButton' ||\n      type === 'AndroidLongPress' ||\n      type === 'AndroidPull'\n    ) {\n      // not implemented in yaml yet\n    } else if (\n      type === 'Error' ||\n      type === 'Assert' ||\n      type === 'AssertWithoutThrow' ||\n      type === 'Finished'\n    ) {\n      // do nothing\n    } else {\n      console.warn(\n        `Cannot convert action ${type} to yaml flow. This should be a bug of Midscene.`,\n      );\n    }\n  }\n\n  if (sleep) {\n    flow.push({\n      sleep: sleep,\n    });\n  }\n\n  return flow;\n}\n"],"names":["AIActionType","callAiFn","msgs","AIActionTypeValue","jsonObject","callToGetJSONObject","defaultBboxSize","debugInspectUtils","getDebug","fillBboxParam","locate","width","height","adaptBbox","adaptQwenBbox","bbox","msg","JSON","Error","result","Math","adaptDoubaoBbox","assert","splitted","Number","Array","bboxList","item","x","y","vlLocateMode","adaptGeminiBbox","left","top","right","bottom","adaptBboxToRect","offsetX","offsetY","rect","warned","warnGPT4oSizeLimit","size","_getModelName","warningMsg","console","mergeRects","rects","minLeft","r","minTop","maxRight","maxBottom","expandSearchArea","screenSize","minEdgeSize","defaultPadding","paddingSizeHorizontal","paddingSizeVertical","markupImageForLLM","screenshotBase64","tree","elementsInfo","treeToList","elementsPositionInfoWithoutText","elementInfo","NodeType","imagePayload","compositeElementInfoImg","buildYamlFlowFromPlans","plans","sleep","flow","plan","_plan_locate","type","param"],"mappings":";;;;;;;AAoCO,IAAKA,sBAAYA,WAAAA,GAAAA,SAAZA,YAAY;;;;;;WAAZA;;AAQL,eAAeC,SACpBC,IAAY,EACZC,iBAA+B;IAE/B,MAAMC,aAAa,MAAMC,oBAAuBH,MAAMC;IAEtD,OAAO;QACL,SAASC,WAAW,OAAO;QAC3B,OAAOA,WAAW,KAAK;IACzB;AACF;AAEA,MAAME,kBAAkB;AACxB,MAAMC,oBAAoBC,SAAS;AAG5B,SAASC,cACdC,MAA2B,EAC3BC,KAAa,EACbC,MAAc;IAGd,IAAKF,OAAe,OAAO,IAAI,CAACA,CAAAA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,IAAI,AAAD,GAAG;QAC5CA,OAAO,IAAI,GAAIA,OAAe,OAAO;QAErC,OAAQA,OAAe,OAAO;IAChC;IAEA,IAAIA,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,IAAI,EACdA,OAAO,IAAI,GAAGG,UAAUH,OAAO,IAAI,EAAEC,OAAOC;IAG9C,OAAOF;AACT;AAEO,SAASI,cACdC,IAAc;IAEd,IAAIA,KAAK,MAAM,GAAG,GAAG;QACnB,MAAMC,MAAM,CAAC,oCAAoC,EAAEC,KAAK,SAAS,CAACF,MAAM,CAAC,CAAC;QAC1E,MAAM,IAAIG,MAAMF;IAClB;IAEA,MAAMG,SAA2C;QAC/CC,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE;QAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE;QACC,YAAnB,OAAOA,IAAI,CAAC,EAAE,GACVK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,IAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,GAAGT;QACN,YAAnB,OAAOS,IAAI,CAAC,EAAE,GACVK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,IAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,GAAGT;KAC1B;IACD,OAAOa;AACT;AAEO,SAASE,gBACdN,IAAkC,EAClCJ,KAAa,EACbC,MAAc;IAEdU,OACEX,QAAQ,KAAKC,SAAS,GACtB;IAGF,IAAI,AAAgB,YAAhB,OAAOG,MAAmB;QAC5BO,OACE,+BAA+B,IAAI,CAACP,KAAK,IAAI,KAC7C,CAAC,iDAAiD,EAAEA,MAAM;QAE5D,MAAMQ,WAAWR,KAAK,KAAK,CAAC;QAC5B,IAAIQ,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACLH,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIZ,QAAS;YAC3CS,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIX,SAAU;YAC5CQ,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIZ,QAAS;YAC3CS,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIX,SAAU;SAC7C;QAEH,MAAM,IAAIM,MAAM,CAAC,iDAAiD,EAAEH,MAAM;IAC5E;IAEA,IAAIU,MAAM,OAAO,CAACV,SAASU,MAAM,OAAO,CAACV,IAAI,CAAC,EAAE,GAC9CA,OAAOA,IAAI,CAAC,EAAE;IAGhB,IAAIW,WAAqB,EAAE;IAC3B,IAAID,MAAM,OAAO,CAACV,SAAS,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EACvCA,KAAK,OAAO,CAAC,CAACY;QACZ,IAAI,AAAgB,YAAhB,OAAOA,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YAClD,MAAM,CAACC,GAAGC,EAAE,GAAGF,KAAK,KAAK,CAAC;YAC1BD,SAAS,IAAI,CAACF,OAAOI,EAAE,IAAI,KAAKJ,OAAOK,EAAE,IAAI;QAC/C,OAAO,IAAI,AAAgB,YAAhB,OAAOF,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YACzD,MAAM,CAACC,GAAGC,EAAE,GAAGF,KAAK,KAAK,CAAC;YAC1BD,SAAS,IAAI,CAACF,OAAOI,EAAE,IAAI,KAAKJ,OAAOK,EAAE,IAAI;QAC/C,OACEH,SAAS,IAAI,CAACF,OAAOG;IAEzB;SAEAD,WAAWX;IAGb,IAAIW,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAO;QACLN,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGf,QAAS;QACnCS,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGd,SAAU;QACpCQ,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGf,QAAS;QACnCS,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGd,SAAU;KACrC;IAIH,IACEc,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAO;QACLN,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGf,QAAS,QAAQL,kBAAkB;QAE/Dc,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGd,SAAU,QAAQN,kBAAkB;QAEhEc,KAAK,GAAG,CACNT,OACAS,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGf,QAAS,QAAQL,kBAAkB;QAE/Dc,KAAK,GAAG,CACNR,QACAQ,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGd,SAAU,QAAQN,kBAAkB;KAEjE;IAGH,IAAIS,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAO;QACLK,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGf,QAAS;QACnCS,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGd,SAAU;QACpCQ,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGf,QAAS;QACnCS,KAAK,KAAK,CAAEM,QAAQ,CAAC,EAAE,GAAGd,SAAU;KACrC;IAGH,MAAMI,MAAM,CAAC,0CAA0C,EAAEC,KAAK,SAAS,CAACF,MAAM,CAAC,CAAC;IAChF,MAAM,IAAIG,MAAMF;AAClB;AAEO,SAASH,UACdE,IAAc,EACdJ,KAAa,EACbC,MAAc;IAEd,IAAIkB,AAAmB,oBAAnBA,kBAAsCA,AAAmB,kBAAnBA,gBACxC,OAAOT,gBAAgBN,MAAMJ,OAAOC;IAGtC,IAAIkB,AAAmB,aAAnBA,gBACF,OAAOC,gBAAgBhB,MAAMJ,OAAOC;IAGtC,OAAOE,cAAcC;AACvB;AAEO,SAASgB,gBACdhB,IAAc,EACdJ,KAAa,EACbC,MAAc;IAEd,MAAMoB,OAAOZ,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGJ,QAAS;IAC5C,MAAMsB,MAAMb,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGH,SAAU;IAC5C,MAAMsB,QAAQd,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGJ,QAAS;IAC7C,MAAMwB,SAASf,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGH,SAAU;IAC/C,OAAO;QAACoB;QAAMC;QAAKC;QAAOC;KAAO;AACnC;AAEO,SAASC,gBACdrB,IAAc,EACdJ,KAAa,EACbC,MAAc,EACdyB,UAAU,CAAC,EACXC,UAAU,CAAC;IAEX/B,kBAAkB,mBAAmBQ,MAAMJ,OAAOC,QAAQyB,SAASC;IACnE,MAAM,CAACN,MAAMC,KAAKC,OAAOC,OAAO,GAAGtB,UAAUE,MAAMJ,OAAOC;IAC1D,MAAM2B,OAAO;QACX,MAAMP,OAAOK;QACb,KAAKJ,MAAMK;QACX,OAAOJ,QAAQF;QACf,QAAQG,SAASF;IACnB;IACA1B,kBAAkB,4BAA4BgC;IAC9C,OAAOA;AACT;AAEA,IAAIC,SAAS;AACN,SAASC,mBAAmBC,IAAU;QAEvCC;IADJ,IAAIH,QAAQ;IACZ,IAAI,QAAAG,CAAAA,gBAAAA,cAAa,IAAbA,KAAAA,IAAAA,cAAgB,WAAW,GAAG,QAAQ,CAAC,WAAW;QACpD,MAAMC,aAAa,CAAC,uEAAuE,EAAEF,KAAK,KAAK,CAAC,CAAC,EAAEA,KAAK,MAAM,CAAC,wFAAwF,CAAC;QAEhN,IACEtB,KAAK,GAAG,CAACsB,KAAK,KAAK,EAAEA,KAAK,MAAM,IAAI,QACpCtB,KAAK,GAAG,CAACsB,KAAK,KAAK,EAAEA,KAAK,MAAM,IAAI,KACpC;YACAG,QAAQ,IAAI,CAACD;YACbJ,SAAS;QACX;IACF,OAAO,IAAIE,KAAK,KAAK,GAAG,QAAQA,KAAK,MAAM,GAAG,MAAM;QAClDG,QAAQ,IAAI,CACV,CAAC,gCAAgC,EAAEH,KAAK,KAAK,CAAC,CAAC,EAAEA,KAAK,MAAM,CAAC,2EAA2E,CAAC;QAE3IF,SAAS;IACX;AACF;AAEO,SAASM,WAAWC,KAAa;IACtC,MAAMC,UAAU5B,KAAK,GAAG,IAAI2B,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,IAAI;IACnD,MAAMC,SAAS9B,KAAK,GAAG,IAAI2B,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,GAAG;IACjD,MAAME,WAAW/B,KAAK,GAAG,IAAI2B,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,IAAI,GAAGA,EAAE,KAAK;IAC9D,MAAMG,YAAYhC,KAAK,GAAG,IAAI2B,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,GAAG,GAAGA,EAAE,MAAM;IAC/D,OAAO;QACL,MAAMD;QACN,KAAKE;QACL,OAAOC,WAAWH;QAClB,QAAQI,YAAYF;IACtB;AACF;AAGO,SAASG,iBAAiBd,IAAU,EAAEe,UAAgB;IAC3D,MAAMC,cAAczB,AAAmB,oBAAnBA,iBAAqC,MAAM;IAC/D,MAAM0B,iBAAiB;IAEvB,MAAMC,wBACJlB,KAAK,KAAK,GAAGgB,cACTnC,KAAK,IAAI,CAAEmC,AAAAA,CAAAA,cAAchB,KAAK,KAAI,IAAK,KACvCiB;IACN,MAAME,sBACJnB,KAAK,MAAM,GAAGgB,cACVnC,KAAK,IAAI,CAAEmC,AAAAA,CAAAA,cAAchB,KAAK,MAAK,IAAK,KACxCiB;IACNjB,KAAK,IAAI,GAAGnB,KAAK,GAAG,CAAC,GAAGmB,KAAK,IAAI,GAAGkB;IACpClB,KAAK,KAAK,GAAGnB,KAAK,GAAG,CACnBmB,KAAK,KAAK,GAAGkB,AAAwB,IAAxBA,uBACbH,WAAW,KAAK,GAAGf,KAAK,IAAI;IAE9BA,KAAK,GAAG,GAAGnB,KAAK,GAAG,CAAC,GAAGmB,KAAK,GAAG,GAAGmB;IAClCnB,KAAK,MAAM,GAAGnB,KAAK,GAAG,CACpBmB,KAAK,MAAM,GAAGmB,AAAsB,IAAtBA,qBACdJ,WAAW,MAAM,GAAGf,KAAK,GAAG;IAE9B,OAAOA;AACT;AAEO,eAAeoB,kBACpBC,gBAAwB,EACxBC,IAAkC,EAClCnB,IAAU;IAEV,MAAMoB,eAAeC,WAAWF;IAChC,MAAMG,kCAAkCF,aAAc,MAAM,CAC1D,CAACG;QACC,IAAIA,YAAY,UAAU,CAAC,QAAQ,KAAKC,SAAS,IAAI,EACnD,OAAO;QAET,OAAO;IACT;IAGF,MAAMC,eAAe,MAAMC,wBAAwB;QACjD,gBAAgBR;QAChB,sBAAsBI;QACtBtB;IACF;IACA,OAAOyB;AACT;AAEO,SAASE,uBACdC,KAAuB,EACvBC,KAAc;IAEd,MAAMC,OAA+B,EAAE;IAEvC,KAAK,MAAMC,QAAQH,MAAO;YAETI;QADf,MAAMC,OAAOF,KAAK,IAAI;QACtB,MAAM/D,SAAS,QAAAgE,CAAAA,eAAAA,KAAK,MAAM,AAAD,IAAVA,KAAAA,IAAAA,aAAa,MAAM;QAElC,IAAIC,AAAS,UAATA,MACFH,KAAK,IAAI,CAAC;YACR,OAAO9D;QACT;aACK,IAAIiE,AAAS,YAATA,MACTH,KAAK,IAAI,CAAC;YACR,SAAS9D;QACX;aACK,IAAIiE,AAAS,YAATA,MAAkB;YAC3B,MAAMC,QAAQH,KAAK,KAAK;YACxBD,KAAK,IAAI,CAAC;gBACR,SAASI,MAAM,KAAK;gBACpBlE;YACF;QACF,OAAO,IAAIiE,AAAS,oBAATA,MAA0B;YACnC,MAAMC,QAAQH,KAAK,KAAK;YACxBD,KAAK,IAAI,CAAC;gBACR,iBAAiBI,MAAM,KAAK;gBAC5BlE;YACF;QACF,OAAO,IAAIiE,AAAS,aAATA,MAAmB;YAC5B,MAAMC,QAAQH,KAAK,KAAK;YACxBD,KAAK,IAAI,CAAC;gBACR,UAAU;gBACV9D;gBACA,WAAWkE,MAAM,SAAS;gBAC1B,YAAYA,MAAM,UAAU;gBAC5B,UAAUA,MAAM,QAAQ;YAC1B;QACF,OAAO,IAAID,AAAS,YAATA,MAAkB;YAC3B,MAAMC,QAAQH,KAAK,KAAK;YACxBD,KAAK,IAAI,CAAC;gBACR,OAAOI,MAAM,MAAM;YACrB;QACF,OACW,wBAATD,QACAA,AAAS,wBAATA,QACAA,AAAS,8BAATA,QACAA,AAAS,uBAATA,QACAA,AAAS,kBAATA,QAIAA,AAAS,YAATA,QACAA,AAAS,aAATA,QACAA,AAAS,yBAATA,QACAA,AAAS,eAATA,QAIA9B,QAAQ,IAAI,CACV,CAAC,sBAAsB,EAAE8B,KAAK,gDAAgD,CAAC;IAGrF;IAEA,IAAIJ,OACFC,KAAK,IAAI,CAAC;QACR,OAAOD;IACT;IAGF,OAAOC;AACT"}

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

@@ -5,6 +5,6 @@ import { generatePlaywrightTest, generatePlaywrightTestStream } from "./prompt/p
 import { generateYamlTest, generateYamlTestStream } from "./prompt/yaml-generator.mjs";
 import { AiAssert, AiExtractElementInfo, AiLocateElement, AiLocateSection } from "./inspect.mjs";
 import { plan } from "./llm-planning.mjs";
-import { AIActionType, actionSpaceTypePrefix, adaptBboxToRect, callAiFn } from "./common.mjs";
+import { AIActionType, adaptBboxToRect, callAiFn } from "./common.mjs";
 import { resizeImageForUiTars, vlmPlanning } from "./ui-tars-planning.mjs";
-export { AIActionType, AiAssert, AiExtractElementInfo, AiLocateElement, AiLocateSection, actionSpaceTypePrefix, adaptBboxToRect, call as callAi, callAiFn, callAiFnWithStringResponse, callToGetJSONObject, describeUserPage, elementByPositionWithElementInfo, generatePlaywrightTest, generatePlaywrightTestStream, generateYamlTest, generateYamlTestStream, plan, resizeImageForUiTars, systemPromptToLocateElement, vlmPlanning };
+export { AIActionType, AiAssert, AiExtractElementInfo, AiLocateElement, AiLocateSection, adaptBboxToRect, call as callAi, callAiFn, callAiFnWithStringResponse, callToGetJSONObject, describeUserPage, elementByPositionWithElementInfo, generatePlaywrightTest, generatePlaywrightTestStream, generateYamlTest, generateYamlTestStream, plan, resizeImageForUiTars, systemPromptToLocateElement, vlmPlanning };

package/dist/es/ai-model/llm-planning.mjs

@@ -56,7 +56,7 @@ async function plan(userInstruction, opts) {
         actions,
         rawResponse,
         usage,
-        yamlFlow: buildYamlFlowFromPlans(actions, opts.actionSpace, planFromAI.sleep)
+        yamlFlow: buildYamlFlowFromPlans(actions, planFromAI.sleep)
     };
     assert(planFromAI, "can't get plans from AI");
     if (vlLocateMode()) {

package/dist/es/ai-model/llm-planning.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/llm-planning.mjs","sources":["webpack://@midscene/core/./src/ai-model/llm-planning.ts"],"sourcesContent":["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":["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":";;;;;;AAyBO,eAAeA,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,iBAAiBL;IAEzB,MAAMM,eAAe,MAAMC,2BAA2B;QACpD,aAAaV,KAAK,WAAW;QAC7B,QAAQW;IACV;IACA,MAAMC,4BAA4BC,8BAChCd,iBACAC,KAAK,GAAG,EACRA,KAAK,aAAa;IAEpB,MAAMc,wBAAwB,MAAMC,qBAClCJ,gBACA,MAAM,CAAC;QACPL;QACA,uBAAuBM;IACzB;IAEA,IAAII,eAAeZ;IACnB,IAAIO,AAAmB,cAAnBA,gBACFK,eAAe,MAAMC,4BAA4BD;SAC5C,IAAI,CAACL,gBACVK,eAAe,MAAME,kBACnBd,kBACAD,QAAQ,IAAI,EACZA,QAAQ,IAAI;IAIhBgB,mBAAmBd;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;IACvB,MAAM,EAAEC,OAAO,EAAEC,KAAK,EAAE,GAAG,MAAMH,KAAKD,MAAMK,aAAa,IAAI;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,uBACRF,SACA9B,KAAK,WAAW,EAChB6B,WAAW,KAAK;IAEpB;IAEAI,OAAOJ,YAAY;IAEnB,IAAIlB,gBAAgB;QAClBmB,QAAQ,OAAO,CAAC,CAACI;YACf,IAAIA,OAAO,MAAM,EACf,IAAI;gBACFA,OAAO,MAAM,GAAGC,cAAcD,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,OAAO,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.mjs","sources":["webpack://@midscene/core/./src/ai-model/llm-planning.ts"],"sourcesContent":["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(actions, planFromAI.sleep),\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":["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":";;;;;;AAyBO,eAAeA,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,iBAAiBL;IAEzB,MAAMM,eAAe,MAAMC,2BAA2B;QACpD,aAAaV,KAAK,WAAW;QAC7B,QAAQW;IACV;IACA,MAAMC,4BAA4BC,8BAChCd,iBACAC,KAAK,GAAG,EACRA,KAAK,aAAa;IAEpB,MAAMc,wBAAwB,MAAMC,qBAClCJ,gBACA,MAAM,CAAC;QACPL;QACA,uBAAuBM;IACzB;IAEA,IAAII,eAAeZ;IACnB,IAAIO,AAAmB,cAAnBA,gBACFK,eAAe,MAAMC,4BAA4BD;SAC5C,IAAI,CAACL,gBACVK,eAAe,MAAME,kBACnBd,kBACAD,QAAQ,IAAI,EACZA,QAAQ,IAAI;IAIhBgB,mBAAmBd;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;IACvB,MAAM,EAAEC,OAAO,EAAEC,KAAK,EAAE,GAAG,MAAMH,KAAKD,MAAMK,aAAa,IAAI;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,uBAAuBF,SAASD,WAAW,KAAK;IAC5D;IAEAI,OAAOJ,YAAY;IAEnB,IAAIlB,gBAAgB;QAClBmB,QAAQ,OAAO,CAAC,CAACI;YACf,IAAIA,OAAO,MAAM,EACf,IAAI;gBACFA,OAAO,MAAM,GAAGC,cAAcD,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,OAAO,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"}

package/dist/es/types.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"types.mjs","sources":["webpack://@midscene/core/./src/types.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport type { NodeType } from '@midscene/shared/constants';\nimport type {\n  BaseElement,\n  ElementTreeNode,\n  Rect,\n  Size,\n} from '@midscene/shared/types';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { DetailedLocateParam, MidsceneYamlFlowItem } from './yaml';\n\nexport type {\n  ElementTreeNode,\n  BaseElement,\n  Rect,\n  Size,\n  Point,\n} from '@midscene/shared/types';\nexport * from './yaml';\n\nexport type AIUsageInfo = Record<string, any> & {\n  prompt_tokens: number | undefined;\n  completion_tokens: number | undefined;\n  total_tokens: number | undefined;\n  time_cost: number | undefined;\n};\n\n/**\n * openai\n *\n */\nexport enum AIResponseFormat {\n  JSON = 'json_object',\n  TEXT = 'text',\n}\n\nexport type AISingleElementResponseById = {\n  id: string;\n  reason?: string;\n  text?: string;\n  xpaths?: string[];\n};\n\nexport type AISingleElementResponseByPosition = {\n  position?: {\n    x: number;\n    y: number;\n  };\n  bbox?: [number, number, number, number];\n  reason: string;\n  text: string;\n};\n\nexport type AISingleElementResponse = AISingleElementResponseById;\nexport interface AIElementLocatorResponse {\n  elements: {\n    id: string;\n    reason?: string;\n    text?: string;\n    xpaths?: string[];\n  }[];\n  bbox?: [number, number, number, number];\n  isOrderSensitive?: boolean;\n  errors?: string[];\n}\n\nexport interface AIElementCoordinatesResponse {\n  bbox: [number, number, number, number];\n  isOrderSensitive?: boolean;\n  errors?: string[];\n}\n\nexport type AIElementResponse =\n  | AIElementLocatorResponse\n  | AIElementCoordinatesResponse;\n\nexport interface AIDataExtractionResponse<DataDemand> {\n  data: DataDemand;\n  errors?: string[];\n  thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n  bbox: [number, number, number, number];\n  references_bbox?: [number, number, number, number][];\n  error?: string;\n}\n\nexport interface AIAssertionResponse {\n  pass: boolean;\n  thought: string;\n}\n\nexport interface AIDescribeElementResponse {\n  description: string;\n  error?: string;\n}\n\nexport interface LocatorValidatorOption {\n  centerDistanceThreshold?: number;\n}\n\nexport interface LocateValidatorResult {\n  pass: boolean;\n  rect: Rect;\n  center: [number, number];\n  centerDistance?: number;\n}\n\nexport interface AgentDescribeElementAtPointResult {\n  prompt: string;\n  deepThink: boolean;\n  verifyResult?: LocateValidatorResult;\n}\n\n/**\n * context\n */\n\nexport abstract class UIContext<ElementType extends BaseElement = BaseElement> {\n  abstract screenshotBase64: string;\n\n  abstract tree: ElementTreeNode<ElementType>;\n\n  abstract size: Size;\n}\n\n/**\n * insight\n */\n\nexport type CallAIFn = <T>(\n  messages: ChatCompletionMessageParam[],\n) => Promise<T>;\n\nexport interface InsightOptions {\n  taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n  aiVendorFn?: CallAIFn;\n}\n\nexport type EnsureObject<T> = { [K in keyof T]: any };\n\nexport type InsightAction = 'locate' | 'extract' | 'assert' | 'describe';\n\nexport type InsightExtractParam = string | Record<string, string>;\n\nexport type LocateResultElement = {\n  id: string;\n  indexId?: number;\n  center: [number, number];\n  rect: Rect;\n  xpaths: string[];\n  attributes: {\n    nodeType: NodeType;\n    [key: string]: string;\n  };\n  isOrderSensitive?: boolean;\n};\n\nexport interface LocateResult {\n  element: LocateResultElement | null;\n  rect?: Rect;\n}\n\nexport interface InsightTaskInfo {\n  durationMs: number;\n  formatResponse?: string;\n  rawResponse?: string;\n  usage?: AIUsageInfo;\n  searchArea?: Rect;\n  searchAreaRawResponse?: string;\n  searchAreaUsage?: AIUsageInfo;\n}\n\nexport interface DumpMeta {\n  sdkVersion: string;\n  logTime: number;\n  model_name: string;\n  model_description?: string;\n}\n\nexport interface ReportDumpWithAttributes {\n  dumpString: string;\n  attributes?: Record<string, any>;\n}\n\nexport interface InsightDump extends DumpMeta {\n  type: 'locate' | 'extract' | 'assert';\n  logId: string;\n  userQuery: {\n    element?: TUserPrompt;\n    dataDemand?: InsightExtractParam;\n    assertion?: TUserPrompt;\n  };\n  matchedElement: BaseElement[];\n  matchedRect?: Rect;\n  deepThink?: boolean;\n  data: any;\n  assertionPass?: boolean;\n  assertionThought?: string;\n  taskInfo: InsightTaskInfo;\n  error?: string;\n  output?: any;\n}\n\nexport type PartialInsightDumpFromSDK = Omit<\n  InsightDump,\n  'sdkVersion' | 'logTime' | 'logId' | 'model_name'\n>;\n\nexport type DumpSubscriber = (dump: InsightDump) => Promise<void> | void;\n\n// intermediate variables to optimize the return value by AI\nexport interface LiteUISection {\n  name: string;\n  description: string;\n  sectionCharacteristics: string;\n  textIds: string[];\n}\n\nexport type ElementById = (id: string) => BaseElement | null;\n\nexport type InsightAssertionResponse = AIAssertionResponse & {\n  usage?: AIUsageInfo;\n};\n\n/**\n * agent\n */\n\nexport type OnTaskStartTip = (tip: string) => Promise<void> | void;\n\nexport interface AgentWaitForOpt {\n  checkIntervalMs?: number;\n  timeoutMs?: number;\n}\n\nexport interface AgentAssertOpt {\n  keepRawResponse?: boolean;\n}\n\n/**\n * planning\n *\n */\n\nexport interface PlanningLocateParam extends DetailedLocateParam {\n  id?: string;\n  bbox?: [number, number, number, number];\n}\n\nexport interface PlanningAction<ParamType = any> {\n  thought?: string;\n  type:\n    | 'Locate'\n    | 'Tap'\n    | 'RightClick'\n    | 'Hover'\n    | 'Drag'\n    | 'Input'\n    | 'KeyboardPress'\n    | 'Scroll'\n    | 'Error'\n    | 'Assert'\n    | 'AssertWithoutThrow'\n    | 'Sleep'\n    | 'Finished'\n    | 'AndroidBackButton'\n    | 'AndroidHomeButton'\n    | 'AndroidRecentAppsButton'\n    | 'AndroidLongPress'\n    | 'AndroidPull';\n  param: ParamType;\n  locate?: PlanningLocateParam | null;\n}\n\nexport interface PlanningAIResponse {\n  action?: PlanningAction; // this is the qwen mode\n  actions?: PlanningAction[];\n  more_actions_needed_by_instruction: boolean;\n  log: string;\n  sleep?: number;\n  error?: string;\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  yamlString?: string;\n}\n\nexport type PlanningActionParamTap = null;\nexport type PlanningActionParamHover = null;\nexport type PlanningActionParamRightClick = null;\n\nexport interface PlanningActionParamInputOrKeyPress {\n  value: string;\n  autoDismissKeyboard?: boolean;\n}\n\nexport interface PlanningActionParamAssert {\n  assertion: TUserPrompt;\n}\n\nexport interface PlanningActionParamSleep {\n  timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n  thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {\n  assertion: string;\n};\n\nexport interface AndroidLongPressParam {\n  duration?: number;\n}\n\nexport interface AndroidPullParam {\n  direction: 'up' | 'down';\n  distance?: number;\n  duration?: number;\n}\n/**\n * misc\n */\n\nexport interface Color {\n  name: string;\n  hex: string;\n}\n\nexport interface BaseAgentParserOpt {\n  selector?: string;\n  ignoreMarker?: boolean;\n}\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PuppeteerParserOpt extends BaseAgentParserOpt {}\n\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PlaywrightParserOpt extends BaseAgentParserOpt {}\n\n/*\naction\n*/\nexport interface ExecutionTaskProgressOptions {\n  onTaskStart?: (task: ExecutionTask) => Promise<void> | void;\n}\n\nexport interface ExecutionRecorderItem {\n  type: 'screenshot';\n  ts: number;\n  screenshot?: string;\n  timing?: string;\n}\n\nexport type ExecutionTaskType =\n  | 'Planning'\n  | 'Insight'\n  | 'Action'\n  | 'Assertion'\n  | 'Log';\n\nexport interface ExecutorContext {\n  task: ExecutionTask;\n  element?: LocateResultElement | null;\n}\n\nexport interface ExecutionTaskApply<\n  Type extends ExecutionTaskType = any,\n  TaskParam = any,\n  TaskOutput = any,\n  TaskLog = any,\n> {\n  type: Type;\n  subType?: string;\n  param?: TaskParam;\n  thought?: string;\n  locate?: PlanningLocateParam | null;\n  pageContext?: UIContext;\n  executor: (\n    param: TaskParam,\n    context: ExecutorContext,\n  ) => // biome-ignore lint/suspicious/noConfusingVoidType: <explanation>\n    | Promise<ExecutionTaskReturn<TaskOutput, TaskLog> | undefined | void>\n    | undefined\n    | void;\n}\n\nexport interface ExecutionTaskHitBy {\n  from: string;\n  context: Record<string, any>;\n}\n\nexport interface ExecutionTaskReturn<TaskOutput = unknown, TaskLog = unknown> {\n  output?: TaskOutput;\n  log?: TaskLog;\n  recorder?: ExecutionRecorderItem[];\n  hitBy?: ExecutionTaskHitBy;\n}\n\nexport type ExecutionTask<\n  E extends ExecutionTaskApply<any, any, any> = ExecutionTaskApply<\n    any,\n    any,\n    any\n  >,\n> = E &\n  ExecutionTaskReturn<\n    E extends ExecutionTaskApply<any, any, infer TaskOutput, any>\n      ? TaskOutput\n      : unknown,\n    E extends ExecutionTaskApply<any, any, any, infer TaskLog>\n      ? TaskLog\n      : unknown\n  > & {\n    status: 'pending' | 'running' | 'finished' | 'failed' | 'cancelled';\n    error?: Error;\n    errorMessage?: string;\n    errorStack?: string;\n    timing?: {\n      start: number;\n      end?: number;\n      cost?: number;\n    };\n    usage?: AIUsageInfo;\n  };\n\nexport interface ExecutionDump extends DumpMeta {\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n}\n\n/*\ntask - insight-locate\n*/\nexport type ExecutionTaskInsightLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskInsightLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport interface ExecutionTaskInsightDumpLog {\n  dump?: InsightDump;\n}\n\nexport type ExecutionTaskInsightLocateApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightLocateParam,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightLocate =\n  ExecutionTask<ExecutionTaskInsightLocateApply>;\n\n/*\ntask - insight-query\n*/\nexport interface ExecutionTaskInsightQueryParam {\n  dataDemand: InsightExtractParam;\n}\n\nexport interface ExecutionTaskInsightQueryOutput {\n  data: any;\n}\n\nexport type ExecutionTaskInsightQueryApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightQueryParam,\n  any,\n  ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightQuery =\n  ExecutionTask<ExecutionTaskInsightQueryApply>;\n\n/*\ntask - assertion\n*/\nexport interface ExecutionTaskInsightAssertionParam {\n  assertion: string;\n}\n\nexport type ExecutionTaskInsightAssertionApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightAssertionParam,\n  InsightAssertionResponse,\n  ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightAssertion =\n  ExecutionTask<ExecutionTaskInsightAssertionApply>;\n\n/*\ntask - action (i.e. interact) \n*/\nexport type ExecutionTaskActionApply<ActionParam = any> = ExecutionTaskApply<\n  'Action',\n  ActionParam,\n  void,\n  void\n>;\n\nexport type ExecutionTaskAction = ExecutionTask<ExecutionTaskActionApply>;\n\n/*\ntask - Log\n*/\n\nexport type ExecutionTaskLogApply<\n  LogParam = {\n    content: string;\n  },\n> = ExecutionTaskApply<'Log', LogParam, void, void>;\n\nexport type ExecutionTaskLog = ExecutionTask<ExecutionTaskLogApply>;\n\n/*\ntask - planning\n*/\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n  'Planning',\n  {\n    userInstruction: string;\n    log?: string;\n  },\n  PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\nGrouped dump\n*/\nexport interface GroupedActionDump {\n  groupName: string;\n  groupDescription?: string;\n  executions: ExecutionDump[];\n}\n\nexport type PageType =\n  | 'puppeteer'\n  | 'playwright'\n  | 'static'\n  | 'chrome-extension-proxy'\n  | 'android';\n\nexport interface StreamingCodeGenerationOptions {\n  /** Whether to enable streaming output */\n  stream?: boolean;\n  /** Callback function to handle streaming chunks */\n  onChunk?: StreamingCallback;\n  /** Callback function to handle streaming completion */\n  onComplete?: (finalCode: string) => void;\n  /** Callback function to handle streaming errors */\n  onError?: (error: Error) => void;\n}\n\nexport type StreamingCallback = (chunk: CodeGenerationChunk) => void;\n\nexport interface CodeGenerationChunk {\n  /** The incremental content chunk */\n  content: string;\n  /** The reasoning content */\n  reasoning_content: string;\n  /** The accumulated content so far */\n  accumulated: string;\n  /** Whether this is the final chunk */\n  isComplete: boolean;\n  /** Token usage information if available */\n  usage?: AIUsageInfo;\n}\n\nexport interface StreamingAIResponse {\n  /** The final accumulated content */\n  content: string;\n  /** Token usage information */\n  usage?: AIUsageInfo;\n  /** Whether the response was streamed */\n  isStreamed: boolean;\n}\n\nexport type TMultimodalPrompt = {\n  /**\n   * Support use image to inspect elements.\n   * The \"images\" field is an object that uses image name as key and image url as value.\n   * The image url can be a local path, a http link , or a base64 string.\n   */\n  images?: {\n    name: string;\n    url: string;\n  }[];\n  /**\n   * By default, the image url in the \"images\" filed starts with `https://` or `http://` will be directly sent to the LLM.\n   * In case the images are not accessible to the LLM (One common case is that image url is internal network only.), you can enable this option.\n   * Then image will be download and convert to base64 format.\n   */\n  convertHttpImage2Base64?: boolean;\n};\n\nexport type TUserPrompt =\n  | string\n  | ({\n      prompt: string;\n    } & Partial<TMultimodalPrompt>);\n\nexport interface DeviceAction<ParamType = any> {\n  name: string;\n  interfaceAlias?: string;\n  description?: string;\n  paramSchema?: string;\n  paramDescription?: string;\n  location?: 'required' | 'optional' | false;\n  whatToLocate?: string; // what to locate if location is required or optional\n  call: (context: ExecutorContext, param: ParamType) => Promise<void> | void;\n}\n"],"names":["AIResponseFormat","UIContext"],"mappings":";AAgCO,IAAKA,yBAAgBA,WAAAA,GAAAA,SAAhBA,gBAAgB;;;WAAhBA;;AAwFL,MAAeC;AAMtB"}
+{"version":3,"file":"types.mjs","sources":["webpack://@midscene/core/./src/types.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport type { NodeType } from '@midscene/shared/constants';\nimport type {\n  BaseElement,\n  ElementTreeNode,\n  Rect,\n  Size,\n} from '@midscene/shared/types';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { DetailedLocateParam, MidsceneYamlFlowItem } from './yaml';\n\nexport type {\n  ElementTreeNode,\n  BaseElement,\n  Rect,\n  Size,\n  Point,\n} from '@midscene/shared/types';\nexport * from './yaml';\n\nexport type AIUsageInfo = Record<string, any> & {\n  prompt_tokens: number | undefined;\n  completion_tokens: number | undefined;\n  total_tokens: number | undefined;\n  time_cost: number | undefined;\n};\n\n/**\n * openai\n *\n */\nexport enum AIResponseFormat {\n  JSON = 'json_object',\n  TEXT = 'text',\n}\n\nexport type AISingleElementResponseById = {\n  id: string;\n  reason?: string;\n  text?: string;\n  xpaths?: string[];\n};\n\nexport type AISingleElementResponseByPosition = {\n  position?: {\n    x: number;\n    y: number;\n  };\n  bbox?: [number, number, number, number];\n  reason: string;\n  text: string;\n};\n\nexport type AISingleElementResponse = AISingleElementResponseById;\nexport interface AIElementLocatorResponse {\n  elements: {\n    id: string;\n    reason?: string;\n    text?: string;\n    xpaths?: string[];\n  }[];\n  bbox?: [number, number, number, number];\n  isOrderSensitive?: boolean;\n  errors?: string[];\n}\n\nexport interface AIElementCoordinatesResponse {\n  bbox: [number, number, number, number];\n  isOrderSensitive?: boolean;\n  errors?: string[];\n}\n\nexport type AIElementResponse =\n  | AIElementLocatorResponse\n  | AIElementCoordinatesResponse;\n\nexport interface AIDataExtractionResponse<DataDemand> {\n  data: DataDemand;\n  errors?: string[];\n  thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n  bbox: [number, number, number, number];\n  references_bbox?: [number, number, number, number][];\n  error?: string;\n}\n\nexport interface AIAssertionResponse {\n  pass: boolean;\n  thought: string;\n}\n\nexport interface AIDescribeElementResponse {\n  description: string;\n  error?: string;\n}\n\nexport interface LocatorValidatorOption {\n  centerDistanceThreshold?: number;\n}\n\nexport interface LocateValidatorResult {\n  pass: boolean;\n  rect: Rect;\n  center: [number, number];\n  centerDistance?: number;\n}\n\nexport interface AgentDescribeElementAtPointResult {\n  prompt: string;\n  deepThink: boolean;\n  verifyResult?: LocateValidatorResult;\n}\n\n/**\n * context\n */\n\nexport abstract class UIContext<ElementType extends BaseElement = BaseElement> {\n  abstract screenshotBase64: string;\n\n  abstract tree: ElementTreeNode<ElementType>;\n\n  abstract size: Size;\n}\n\n/**\n * insight\n */\n\nexport type CallAIFn = <T>(\n  messages: ChatCompletionMessageParam[],\n) => Promise<T>;\n\nexport interface InsightOptions {\n  taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n  aiVendorFn?: CallAIFn;\n}\n\nexport type EnsureObject<T> = { [K in keyof T]: any };\n\nexport type InsightAction = 'locate' | 'extract' | 'assert' | 'describe';\n\nexport type InsightExtractParam = string | Record<string, string>;\n\nexport type LocateResultElement = {\n  id: string;\n  indexId?: number;\n  center: [number, number];\n  rect: Rect;\n  xpaths: string[];\n  attributes: {\n    nodeType: NodeType;\n    [key: string]: string;\n  };\n  isOrderSensitive?: boolean;\n};\n\nexport interface LocateResult {\n  element: LocateResultElement | null;\n  rect?: Rect;\n}\n\nexport interface InsightTaskInfo {\n  durationMs: number;\n  formatResponse?: string;\n  rawResponse?: string;\n  usage?: AIUsageInfo;\n  searchArea?: Rect;\n  searchAreaRawResponse?: string;\n  searchAreaUsage?: AIUsageInfo;\n}\n\nexport interface DumpMeta {\n  sdkVersion: string;\n  logTime: number;\n  model_name: string;\n  model_description?: string;\n}\n\nexport interface ReportDumpWithAttributes {\n  dumpString: string;\n  attributes?: Record<string, any>;\n}\n\nexport interface InsightDump extends DumpMeta {\n  type: 'locate' | 'extract' | 'assert';\n  logId: string;\n  userQuery: {\n    element?: TUserPrompt;\n    dataDemand?: InsightExtractParam;\n    assertion?: TUserPrompt;\n  };\n  matchedElement: BaseElement[];\n  matchedRect?: Rect;\n  deepThink?: boolean;\n  data: any;\n  assertionPass?: boolean;\n  assertionThought?: string;\n  taskInfo: InsightTaskInfo;\n  error?: string;\n  output?: any;\n}\n\nexport type PartialInsightDumpFromSDK = Omit<\n  InsightDump,\n  'sdkVersion' | 'logTime' | 'logId' | 'model_name'\n>;\n\nexport type DumpSubscriber = (dump: InsightDump) => Promise<void> | void;\n\n// intermediate variables to optimize the return value by AI\nexport interface LiteUISection {\n  name: string;\n  description: string;\n  sectionCharacteristics: string;\n  textIds: string[];\n}\n\nexport type ElementById = (id: string) => BaseElement | null;\n\nexport type InsightAssertionResponse = AIAssertionResponse & {\n  usage?: AIUsageInfo;\n};\n\n/**\n * agent\n */\n\nexport type OnTaskStartTip = (tip: string) => Promise<void> | void;\n\nexport interface AgentWaitForOpt {\n  checkIntervalMs?: number;\n  timeoutMs?: number;\n}\n\nexport interface AgentAssertOpt {\n  keepRawResponse?: boolean;\n}\n\n/**\n * planning\n *\n */\n\nexport interface PlanningLocateParam extends DetailedLocateParam {\n  id?: string;\n  bbox?: [number, number, number, number];\n}\n\nexport interface PlanningAction<ParamType = any> {\n  thought?: string;\n  type:\n    | 'Locate'\n    | 'Tap'\n    | 'RightClick'\n    | 'Hover'\n    | 'Drag'\n    | 'Input'\n    | 'KeyboardPress'\n    | 'Scroll'\n    | 'Error'\n    | 'Assert'\n    | 'AssertWithoutThrow'\n    | 'Sleep'\n    | 'Finished'\n    | 'AndroidBackButton'\n    | 'AndroidHomeButton'\n    | 'AndroidRecentAppsButton'\n    | 'AndroidLongPress'\n    | 'AndroidPull';\n  param: ParamType;\n  locate?: PlanningLocateParam | null;\n}\n\nexport interface PlanningAIResponse {\n  action?: PlanningAction; // this is the qwen mode\n  actions?: PlanningAction[];\n  more_actions_needed_by_instruction: boolean;\n  log: string;\n  sleep?: number;\n  error?: string;\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  yamlString?: string;\n}\n\nexport type PlanningActionParamTap = null;\nexport type PlanningActionParamHover = null;\nexport type PlanningActionParamRightClick = null;\n\nexport interface PlanningActionParamInputOrKeyPress {\n  value: string;\n  autoDismissKeyboard?: boolean;\n}\n\nexport interface PlanningActionParamAssert {\n  assertion: TUserPrompt;\n}\n\nexport interface PlanningActionParamSleep {\n  timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n  thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {\n  assertion: string;\n};\n\nexport interface AndroidLongPressParam {\n  duration?: number;\n}\n\nexport interface AndroidPullParam {\n  direction: 'up' | 'down';\n  distance?: number;\n  duration?: number;\n}\n/**\n * misc\n */\n\nexport interface Color {\n  name: string;\n  hex: string;\n}\n\nexport interface BaseAgentParserOpt {\n  selector?: string;\n  ignoreMarker?: boolean;\n}\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PuppeteerParserOpt extends BaseAgentParserOpt {}\n\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PlaywrightParserOpt extends BaseAgentParserOpt {}\n\n/*\naction\n*/\nexport interface ExecutionTaskProgressOptions {\n  onTaskStart?: (task: ExecutionTask) => Promise<void> | void;\n}\n\nexport interface ExecutionRecorderItem {\n  type: 'screenshot';\n  ts: number;\n  screenshot?: string;\n  timing?: string;\n}\n\nexport type ExecutionTaskType =\n  | 'Planning'\n  | 'Insight'\n  | 'Action'\n  | 'Assertion'\n  | 'Log';\n\nexport interface ExecutorContext {\n  task: ExecutionTask;\n  element?: LocateResultElement | null;\n}\n\nexport interface ExecutionTaskApply<\n  Type extends ExecutionTaskType = any,\n  TaskParam = any,\n  TaskOutput = any,\n  TaskLog = any,\n> {\n  type: Type;\n  subType?: string;\n  param?: TaskParam;\n  thought?: string;\n  locate?: PlanningLocateParam | null;\n  pageContext?: UIContext;\n  executor: (\n    param: TaskParam,\n    context: ExecutorContext,\n  ) => // biome-ignore lint/suspicious/noConfusingVoidType: <explanation>\n    | Promise<ExecutionTaskReturn<TaskOutput, TaskLog> | undefined | void>\n    | undefined\n    | void;\n}\n\nexport interface ExecutionTaskHitBy {\n  from: string;\n  context: Record<string, any>;\n}\n\nexport interface ExecutionTaskReturn<TaskOutput = unknown, TaskLog = unknown> {\n  output?: TaskOutput;\n  log?: TaskLog;\n  recorder?: ExecutionRecorderItem[];\n  hitBy?: ExecutionTaskHitBy;\n}\n\nexport type ExecutionTask<\n  E extends ExecutionTaskApply<any, any, any> = ExecutionTaskApply<\n    any,\n    any,\n    any\n  >,\n> = E &\n  ExecutionTaskReturn<\n    E extends ExecutionTaskApply<any, any, infer TaskOutput, any>\n      ? TaskOutput\n      : unknown,\n    E extends ExecutionTaskApply<any, any, any, infer TaskLog>\n      ? TaskLog\n      : unknown\n  > & {\n    status: 'pending' | 'running' | 'finished' | 'failed' | 'cancelled';\n    error?: Error;\n    errorMessage?: string;\n    errorStack?: string;\n    timing?: {\n      start: number;\n      end?: number;\n      cost?: number;\n    };\n    usage?: AIUsageInfo;\n  };\n\nexport interface ExecutionDump extends DumpMeta {\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n}\n\n/*\ntask - insight-locate\n*/\nexport type ExecutionTaskInsightLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskInsightLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport interface ExecutionTaskInsightDumpLog {\n  dump?: InsightDump;\n}\n\nexport type ExecutionTaskInsightLocateApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightLocateParam,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightLocate =\n  ExecutionTask<ExecutionTaskInsightLocateApply>;\n\n/*\ntask - insight-query\n*/\nexport interface ExecutionTaskInsightQueryParam {\n  dataDemand: InsightExtractParam;\n}\n\nexport interface ExecutionTaskInsightQueryOutput {\n  data: any;\n}\n\nexport type ExecutionTaskInsightQueryApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightQueryParam,\n  any,\n  ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightQuery =\n  ExecutionTask<ExecutionTaskInsightQueryApply>;\n\n/*\ntask - assertion\n*/\nexport interface ExecutionTaskInsightAssertionParam {\n  assertion: string;\n}\n\nexport type ExecutionTaskInsightAssertionApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightAssertionParam,\n  InsightAssertionResponse,\n  ExecutionTaskInsightDumpLog\n>;\n\nexport type ExecutionTaskInsightAssertion =\n  ExecutionTask<ExecutionTaskInsightAssertionApply>;\n\n/*\ntask - action (i.e. interact) \n*/\nexport type ExecutionTaskActionApply<ActionParam = any> = ExecutionTaskApply<\n  'Action',\n  ActionParam,\n  void,\n  void\n>;\n\nexport type ExecutionTaskAction = ExecutionTask<ExecutionTaskActionApply>;\n\n/*\ntask - Log\n*/\n\nexport type ExecutionTaskLogApply<\n  LogParam = {\n    content: string;\n  },\n> = ExecutionTaskApply<'Log', LogParam, void, void>;\n\nexport type ExecutionTaskLog = ExecutionTask<ExecutionTaskLogApply>;\n\n/*\ntask - planning\n*/\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n  'Planning',\n  {\n    userInstruction: string;\n    log?: string;\n  },\n  PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\nGrouped dump\n*/\nexport interface GroupedActionDump {\n  groupName: string;\n  groupDescription?: string;\n  executions: ExecutionDump[];\n}\n\nexport type PageType =\n  | 'puppeteer'\n  | 'playwright'\n  | 'static'\n  | 'chrome-extension-proxy'\n  | 'android';\n\nexport interface StreamingCodeGenerationOptions {\n  /** Whether to enable streaming output */\n  stream?: boolean;\n  /** Callback function to handle streaming chunks */\n  onChunk?: StreamingCallback;\n  /** Callback function to handle streaming completion */\n  onComplete?: (finalCode: string) => void;\n  /** Callback function to handle streaming errors */\n  onError?: (error: Error) => void;\n}\n\nexport type StreamingCallback = (chunk: CodeGenerationChunk) => void;\n\nexport interface CodeGenerationChunk {\n  /** The incremental content chunk */\n  content: string;\n  /** The reasoning content */\n  reasoning_content: string;\n  /** The accumulated content so far */\n  accumulated: string;\n  /** Whether this is the final chunk */\n  isComplete: boolean;\n  /** Token usage information if available */\n  usage?: AIUsageInfo;\n}\n\nexport interface StreamingAIResponse {\n  /** The final accumulated content */\n  content: string;\n  /** Token usage information */\n  usage?: AIUsageInfo;\n  /** Whether the response was streamed */\n  isStreamed: boolean;\n}\n\nexport type TMultimodalPrompt = {\n  /**\n   * Support use image to inspect elements.\n   * The \"images\" field is an object that uses image name as key and image url as value.\n   * The image url can be a local path, a http link , or a base64 string.\n   */\n  images?: {\n    name: string;\n    url: string;\n  }[];\n  /**\n   * By default, the image url in the \"images\" filed starts with `https://` or `http://` will be directly sent to the LLM.\n   * In case the images are not accessible to the LLM (One common case is that image url is internal network only.), you can enable this option.\n   * Then image will be download and convert to base64 format.\n   */\n  convertHttpImage2Base64?: boolean;\n};\n\nexport type TUserPrompt =\n  | string\n  | ({\n      prompt: string;\n    } & Partial<TMultimodalPrompt>);\n\nexport interface DeviceAction<ParamType = any> {\n  name: string;\n  description?: string;\n  paramSchema?: string;\n  paramDescription?: string;\n  location?: 'required' | 'optional' | false;\n  whatToLocate?: string; // what to locate if location is required or optional\n  call: (context: ExecutorContext, param: ParamType) => Promise<void> | void;\n}\n"],"names":["AIResponseFormat","UIContext"],"mappings":";AAgCO,IAAKA,yBAAgBA,WAAAA,GAAAA,SAAhBA,gBAAgB;;;WAAhBA;;AAwFL,MAAeC;AAMtB"}