@midscene/core 1.0.1-beta-20251202112442.0 → 1.0.1-beta-20251203073716.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/es/agent/agent.mjs +33 -40
- package/dist/es/agent/agent.mjs.map +1 -1
- package/dist/es/agent/execution-session.mjs.map +1 -1
- package/dist/es/agent/task-builder.mjs +25 -25
- package/dist/es/agent/task-builder.mjs.map +1 -1
- package/dist/es/agent/task-cache.mjs +3 -3
- package/dist/es/agent/task-cache.mjs.map +1 -1
- package/dist/es/agent/tasks.mjs +13 -14
- package/dist/es/agent/tasks.mjs.map +1 -1
- package/dist/es/agent/ui-utils.mjs +9 -21
- package/dist/es/agent/ui-utils.mjs.map +1 -1
- package/dist/es/agent/utils.mjs +5 -8
- package/dist/es/agent/utils.mjs.map +1 -1
- package/dist/es/ai-model/conversation-history.mjs +1 -2
- package/dist/es/ai-model/conversation-history.mjs.map +1 -1
- package/dist/es/ai-model/inspect.mjs +6 -9
- package/dist/es/ai-model/inspect.mjs.map +1 -1
- package/dist/es/ai-model/llm-planning.mjs.map +1 -1
- package/dist/es/ai-model/prompt/assertion.mjs.map +1 -1
- package/dist/es/ai-model/prompt/common.mjs.map +1 -1
- package/dist/es/ai-model/prompt/describe.mjs.map +1 -1
- package/dist/es/ai-model/prompt/extraction.mjs.map +1 -1
- package/dist/es/ai-model/prompt/llm-locator.mjs.map +1 -1
- package/dist/es/ai-model/prompt/llm-planning.mjs +6 -12
- package/dist/es/ai-model/prompt/llm-planning.mjs.map +1 -1
- package/dist/es/ai-model/prompt/llm-section-locator.mjs.map +1 -1
- package/dist/es/ai-model/prompt/order-sensitive-judge.mjs.map +1 -1
- package/dist/es/ai-model/prompt/playwright-generator.mjs +2 -2
- package/dist/es/ai-model/prompt/playwright-generator.mjs.map +1 -1
- package/dist/es/ai-model/prompt/ui-tars-locator.mjs.map +1 -1
- package/dist/es/ai-model/prompt/ui-tars-planning.mjs.map +1 -1
- package/dist/es/ai-model/prompt/util.mjs.map +1 -1
- package/dist/es/ai-model/prompt/yaml-generator.mjs +10 -10
- package/dist/es/ai-model/prompt/yaml-generator.mjs.map +1 -1
- package/dist/es/ai-model/service-caller/index.mjs +11 -14
- package/dist/es/ai-model/service-caller/index.mjs.map +1 -1
- package/dist/es/ai-model/ui-tars-planning.mjs +68 -7
- package/dist/es/ai-model/ui-tars-planning.mjs.map +1 -1
- package/dist/es/common.mjs +6 -11
- package/dist/es/common.mjs.map +1 -1
- package/dist/es/device/index.mjs.map +1 -1
- package/dist/es/index.mjs.map +1 -1
- package/dist/es/report.mjs.map +1 -1
- package/dist/es/service/index.mjs +6 -8
- package/dist/es/service/index.mjs.map +1 -1
- package/dist/es/service/utils.mjs.map +1 -1
- package/dist/es/task-runner.mjs +17 -21
- package/dist/es/task-runner.mjs.map +1 -1
- package/dist/es/tree.mjs.map +1 -1
- package/dist/es/types.mjs.map +1 -1
- package/dist/es/utils.mjs +6 -7
- package/dist/es/utils.mjs.map +1 -1
- package/dist/es/yaml/builder.mjs.map +1 -1
- package/dist/es/yaml/player.mjs +11 -16
- package/dist/es/yaml/player.mjs.map +1 -1
- package/dist/es/yaml/utils.mjs +1 -1
- package/dist/es/yaml/utils.mjs.map +1 -1
- package/dist/lib/agent/agent.js +33 -40
- package/dist/lib/agent/agent.js.map +1 -1
- package/dist/lib/agent/execution-session.js.map +1 -1
- package/dist/lib/agent/index.js +10 -10
- package/dist/lib/agent/index.js.map +1 -1
- package/dist/lib/agent/task-builder.js +25 -25
- package/dist/lib/agent/task-builder.js.map +1 -1
- package/dist/lib/agent/task-cache.js +5 -5
- package/dist/lib/agent/task-cache.js.map +1 -1
- package/dist/lib/agent/tasks.js +14 -15
- package/dist/lib/agent/tasks.js.map +1 -1
- package/dist/lib/agent/ui-utils.js +9 -21
- package/dist/lib/agent/ui-utils.js.map +1 -1
- package/dist/lib/agent/utils.js +12 -15
- package/dist/lib/agent/utils.js.map +1 -1
- package/dist/lib/ai-model/conversation-history.js +1 -2
- package/dist/lib/ai-model/conversation-history.js.map +1 -1
- package/dist/lib/ai-model/index.js +22 -22
- package/dist/lib/ai-model/index.js.map +1 -1
- package/dist/lib/ai-model/inspect.js +9 -12
- package/dist/lib/ai-model/inspect.js.map +1 -1
- package/dist/lib/ai-model/llm-planning.js.map +1 -1
- package/dist/lib/ai-model/prompt/assertion.js.map +1 -1
- package/dist/lib/ai-model/prompt/common.js.map +1 -1
- package/dist/lib/ai-model/prompt/describe.js.map +1 -1
- package/dist/lib/ai-model/prompt/extraction.js.map +1 -1
- package/dist/lib/ai-model/prompt/llm-locator.js +2 -2
- package/dist/lib/ai-model/prompt/llm-locator.js.map +1 -1
- package/dist/lib/ai-model/prompt/llm-planning.js +6 -12
- package/dist/lib/ai-model/prompt/llm-planning.js.map +1 -1
- package/dist/lib/ai-model/prompt/llm-section-locator.js.map +1 -1
- package/dist/lib/ai-model/prompt/order-sensitive-judge.js.map +1 -1
- package/dist/lib/ai-model/prompt/playwright-generator.js +9 -9
- package/dist/lib/ai-model/prompt/playwright-generator.js.map +1 -1
- package/dist/lib/ai-model/prompt/ui-tars-locator.js.map +1 -1
- package/dist/lib/ai-model/prompt/ui-tars-planning.js.map +1 -1
- package/dist/lib/ai-model/prompt/util.js.map +1 -1
- package/dist/lib/ai-model/prompt/yaml-generator.js +16 -16
- package/dist/lib/ai-model/prompt/yaml-generator.js.map +1 -1
- package/dist/lib/ai-model/service-caller/index.js +14 -17
- package/dist/lib/ai-model/service-caller/index.js.map +1 -1
- package/dist/lib/ai-model/ui-tars-planning.js +68 -7
- package/dist/lib/ai-model/ui-tars-planning.js.map +1 -1
- package/dist/lib/common.js +24 -29
- package/dist/lib/common.js.map +1 -1
- package/dist/lib/device/device-options.js.map +1 -1
- package/dist/lib/device/index.js +19 -19
- package/dist/lib/device/index.js.map +1 -1
- package/dist/lib/image/index.js +3 -3
- package/dist/lib/image/index.js.map +1 -1
- package/dist/lib/index.js +13 -13
- package/dist/lib/index.js.map +1 -1
- package/dist/lib/report.js.map +1 -1
- package/dist/lib/service/index.js +6 -8
- package/dist/lib/service/index.js.map +1 -1
- package/dist/lib/service/utils.js.map +1 -1
- package/dist/lib/task-runner.js +17 -21
- package/dist/lib/task-runner.js.map +1 -1
- package/dist/lib/tree.js +2 -2
- package/dist/lib/tree.js.map +1 -1
- package/dist/lib/types.js +7 -9
- package/dist/lib/types.js.map +1 -1
- package/dist/lib/utils.js +14 -15
- package/dist/lib/utils.js.map +1 -1
- package/dist/lib/yaml/builder.js.map +1 -1
- package/dist/lib/yaml/index.js +12 -18
- package/dist/lib/yaml/index.js.map +1 -1
- package/dist/lib/yaml/player.js +11 -16
- package/dist/lib/yaml/player.js.map +1 -1
- package/dist/lib/yaml/utils.js +4 -4
- package/dist/lib/yaml/utils.js.map +1 -1
- package/dist/lib/yaml.js.map +1 -1
- package/dist/types/ai-model/ui-tars-planning.d.ts +15 -2
- package/dist/types/yaml.d.ts +44 -1
- package/package.json +3 -3
package/dist/es/common.mjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"common.mjs","sources":["webpack://@midscene/core/./src/common.ts"],"sourcesContent":["import type {\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 { ChatCompletionMessageParam } from 'openai/resources/index';\n\nimport type { PlanningLocateParam } from '@/types';\nimport { NodeType } from '@midscene/shared/constants';\nimport type { TVlModeTypes } from '@midscene/shared/env';\nimport { treeToList } from '@midscene/shared/extractor';\nimport { compositeElementInfoImg } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { z } from 'zod';\n\nexport type AIArgs = ChatCompletionMessageParam[];\n\nexport enum AIActionType {\n ASSERT = 0,\n INSPECT_ELEMENT = 1,\n EXTRACT_DATA = 2,\n PLAN = 3,\n DESCRIBE_ELEMENT = 4,\n TEXT = 5,\n}\n\nconst defaultBboxSize = 20; // must be even number\nconst debugInspectUtils = getDebug('ai:common');\ntype AdaptBboxInput = number[] | string[] | string | (number[] | string[])[];\n\n// transform the param of locate from qwen mode\nexport function fillBboxParam(\n locate: PlanningLocateParam,\n width: number,\n height: number,\n rightLimit: number,\n bottomLimit: number,\n vlMode: TVlModeTypes | undefined,\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(\n locate.bbox,\n width,\n height,\n rightLimit,\n bottomLimit,\n vlMode,\n );\n }\n\n return locate;\n}\n\nexport function adaptQwen2_5Bbox(\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 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\nfunction normalizeBboxInput(\n bbox: AdaptBboxInput,\n): number[] | string[] | string {\n if (Array.isArray(bbox)) {\n if (Array.isArray(bbox[0])) {\n return bbox[0] as number[] | string[];\n }\n return bbox as number[] | string[];\n }\n return bbox as string;\n}\n\nexport function adaptBbox(\n bbox: AdaptBboxInput,\n width: number,\n height: number,\n rightLimit: number,\n bottomLimit: number,\n vlMode: TVlModeTypes | undefined,\n): [number, number, number, number] {\n const normalizedBbox = normalizeBboxInput(bbox);\n\n let result: [number, number, number, number] = [0, 0, 0, 0];\n if (vlMode === 'doubao-vision' || vlMode === 'vlm-ui-tars') {\n result = adaptDoubaoBbox(normalizedBbox, width, height);\n } else if (vlMode === 'gemini') {\n result = adaptGeminiBbox(normalizedBbox as number[], width, height);\n } else if (vlMode === 'qwen3-vl') {\n result = normalized01000(normalizedBbox as number[], width, height);\n } else {\n result = adaptQwen2_5Bbox(normalizedBbox as number[]);\n }\n\n result[2] = Math.min(result[2], rightLimit);\n result[3] = Math.min(result[3], bottomLimit);\n\n return result;\n}\n\n// x1, y1, x2, y2 -> 0-1000\nexport function normalized01000(\n bbox: number[],\n width: number,\n height: number,\n): [number, number, number, number] {\n return [\n Math.round((bbox[0] * width) / 1000),\n Math.round((bbox[1] * height) / 1000),\n Math.round((bbox[2] * width) / 1000),\n Math.round((bbox[3] * height) / 1000),\n ];\n}\n\n// y1, x1, y2, x2 -> 0-1000\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 rightLimit = width,\n bottomLimit = height,\n vlMode?: TVlModeTypes | undefined,\n): Rect {\n debugInspectUtils(\n 'adaptBboxToRect',\n bbox,\n width,\n height,\n 'offset',\n offsetX,\n offsetY,\n 'limit',\n rightLimit,\n bottomLimit,\n 'vlMode',\n vlMode,\n );\n const [left, top, right, bottom] = adaptBbox(\n bbox,\n width,\n height,\n rightLimit,\n bottomLimit,\n vlMode,\n );\n\n // Calculate initial rect dimensions\n const rectLeft = left;\n const rectTop = top;\n let rectWidth = right - left;\n let rectHeight = bottom - top;\n\n // Ensure the rect doesn't exceed image boundaries\n // If right edge exceeds width, adjust the width\n if (rectLeft + rectWidth > width) {\n rectWidth = width - rectLeft;\n }\n\n // If bottom edge exceeds height, adjust the height\n if (rectTop + rectHeight > height) {\n rectHeight = height - rectTop;\n }\n\n // Ensure minimum dimensions (width and height should be at least 1)\n rectWidth = Math.max(1, rectWidth);\n rectHeight = Math.max(1, rectHeight);\n\n const rect = {\n left: rectLeft + offsetX,\n top: rectTop + offsetY,\n width: rectWidth,\n height: rectHeight,\n };\n debugInspectUtils('adaptBboxToRect, result=', rect);\n\n return rect;\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(\n rect: Rect,\n screenSize: Size,\n vlMode: TVlModeTypes | undefined,\n) {\n const minEdgeSize = vlMode === 'doubao-vision' ? 500 : 300;\n const defaultPadding = 160;\n\n // Calculate padding needed to reach minimum edge size\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\n // Calculate new dimensions (ensure minimum edge size)\n let newWidth = Math.max(minEdgeSize, rect.width + paddingSizeHorizontal * 2);\n let newHeight = Math.max(minEdgeSize, rect.height + paddingSizeVertical * 2);\n\n // Calculate initial position with padding\n let newLeft = rect.left - paddingSizeHorizontal;\n let newTop = rect.top - paddingSizeVertical;\n\n // Ensure the rect doesn't exceed screen boundaries by adjusting position\n // If the rect goes beyond the right edge, shift it left\n if (newLeft + newWidth > screenSize.width) {\n newLeft = screenSize.width - newWidth;\n }\n\n // If the rect goes beyond the bottom edge, shift it up\n if (newTop + newHeight > screenSize.height) {\n newTop = screenSize.height - newHeight;\n }\n\n // Ensure the rect doesn't go beyond the left/top edges\n newLeft = Math.max(0, newLeft);\n newTop = Math.max(0, newTop);\n\n // If after position adjustment, the rect still exceeds screen boundaries,\n // clamp the dimensions to fit within screen\n if (newLeft + newWidth > screenSize.width) {\n newWidth = screenSize.width - newLeft;\n }\n if (newTop + newHeight > screenSize.height) {\n newHeight = screenSize.height - newTop;\n }\n\n rect.left = newLeft;\n rect.top = newTop;\n rect.width = newWidth;\n rect.height = newHeight;\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<any>[],\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 flowKey = action.interfaceAlias || verb;\n const flowParam = action.paramSchema\n ? dumpActionParam(plan.param || {}, action.paramSchema)\n : {};\n\n const flowItem: MidsceneYamlFlowItem = {\n [flowKey]: '',\n ...flowParam,\n };\n\n flow.push(flowItem);\n }\n\n if (sleep) {\n flow.push({\n sleep,\n });\n }\n\n return flow;\n}\n\n// Zod schemas for shared types\nexport const PointSchema = z.object({\n left: z.number(),\n top: z.number(),\n});\n\nexport const SizeSchema = z.object({\n width: z.number(),\n height: z.number(),\n dpr: z.number().optional(),\n});\n\nexport const RectSchema = PointSchema.and(SizeSchema).and(\n z.object({\n zoom: z.number().optional(),\n }),\n);\n\n// Zod schema for TMultimodalPrompt\nexport const TMultimodalPromptSchema = z.object({\n images: z\n .array(\n z.object({\n name: z.string(),\n url: z.string(),\n }),\n )\n .optional(),\n convertHttpImage2Base64: z.boolean().optional(),\n});\n\n// Zod schema for TUserPrompt\nexport const TUserPromptSchema = z.union([\n z.string(),\n z\n .object({\n prompt: z.string(),\n })\n .and(TMultimodalPromptSchema.partial()),\n]);\n\n// Generate TypeScript types from Zod schemas\nexport type TMultimodalPrompt = z.infer<typeof TMultimodalPromptSchema>;\nexport type TUserPrompt = z.infer<typeof TUserPromptSchema>;\n\nconst locateFieldFlagName = 'midscene_location_field_flag';\n\n// Schema for locator field input (when users provide locate parameters)\nconst MidsceneLocationInput = z\n .object({\n prompt: TUserPromptSchema,\n deepThink: z.boolean().optional(),\n cacheable: z.boolean().optional(),\n xpath: z.union([z.string(), z.boolean()]).optional(),\n })\n .passthrough();\n\n// Schema for locator field result (when AI returns locate results)\nconst MidsceneLocationResult = z\n .object({\n [locateFieldFlagName]: z.literal(true),\n prompt: TUserPromptSchema,\n\n // optional fields\n deepThink: z.boolean().optional(), // only available in vl model\n cacheable: z.boolean().optional(),\n xpath: z.boolean().optional(), // preset result for xpath\n\n // these two fields will only appear in the result\n center: z.tuple([z.number(), z.number()]),\n rect: RectSchema,\n })\n .passthrough();\n\n// Export the result type - this is used for runtime results that include center and rect\nexport type MidsceneLocationResultType = z.infer<typeof MidsceneLocationResult>;\n\n// Export the input type - this is the inferred type from getMidsceneLocationSchema()\nexport type MidsceneLocationInputType = z.infer<typeof MidsceneLocationInput>;\n\n/**\n * Returns the schema for locator fields.\n * This now returns the input schema which is more permissive and suitable for validation.\n */\nexport const getMidsceneLocationSchema = () => {\n return MidsceneLocationInput;\n};\n\nexport const ifMidsceneLocatorField = (field: any): boolean => {\n // Handle optional fields by getting the inner type\n let actualField = field;\n if (actualField._def?.typeName === 'ZodOptional') {\n actualField = actualField._def.innerType;\n }\n\n // Check if this is a ZodObject\n if (actualField._def?.typeName === 'ZodObject') {\n const shape = actualField._def.shape();\n\n // Method 1: Check for the location field flag (for result schema)\n if (locateFieldFlagName in shape) {\n return true;\n }\n\n // Method 2: Check if it's the input schema by checking for 'prompt' field\n // Input schema has 'prompt' as a required field\n if ('prompt' in shape && shape.prompt) {\n return true;\n }\n }\n\n return false;\n};\n\nexport const dumpMidsceneLocatorField = (field: any): string => {\n assert(\n ifMidsceneLocatorField(field),\n 'field is not a midscene locator field',\n );\n\n // If field is a string, return it directly\n if (typeof field === 'string') {\n return field;\n }\n\n // If field is an object with prompt property\n if (field && typeof field === 'object' && field.prompt) {\n // If prompt is a string, return it directly\n if (typeof field.prompt === 'string') {\n return field.prompt;\n }\n // If prompt is a TUserPrompt object, extract the prompt string\n if (typeof field.prompt === 'object' && field.prompt.prompt) {\n return field.prompt.prompt; // TODO: dump images if necessary\n }\n }\n\n // Fallback: try to convert to string\n return String(field);\n};\n\nexport const findAllMidsceneLocatorField = (\n zodType?: z.ZodType<any>,\n requiredOnly?: boolean,\n): string[] => {\n if (!zodType) {\n return [];\n }\n\n // Check if this is a ZodObject by checking if it has a shape property\n const zodObject = zodType as any;\n if (zodObject._def?.typeName === 'ZodObject' && zodObject.shape) {\n const keys = Object.keys(zodObject.shape);\n return keys.filter((key) => {\n const field = zodObject.shape[key];\n if (!ifMidsceneLocatorField(field)) {\n return false;\n }\n\n // If requiredOnly is true, filter out optional fields\n if (requiredOnly) {\n return field._def?.typeName !== 'ZodOptional';\n }\n\n return true;\n });\n }\n\n // For other ZodType instances, we can't extract field names\n return [];\n};\n\nexport const dumpActionParam = (\n jsonObject: Record<string, any>,\n zodSchema: z.ZodType<any>,\n): Record<string, any> => {\n const locatorFields = findAllMidsceneLocatorField(zodSchema);\n const result = { ...jsonObject };\n\n for (const fieldName of locatorFields) {\n const fieldValue = result[fieldName];\n if (fieldValue) {\n // If it's already a string, keep it as is\n if (typeof fieldValue === 'string') {\n result[fieldName] = fieldValue;\n } else if (typeof fieldValue === 'object') {\n // Check if this field is actually a MidsceneLocationType object\n if (fieldValue.prompt) {\n // If prompt is a string, use it directly\n if (typeof fieldValue.prompt === 'string') {\n result[fieldName] = fieldValue.prompt;\n } else if (\n typeof fieldValue.prompt === 'object' &&\n fieldValue.prompt.prompt\n ) {\n // If prompt is a TUserPrompt object, extract the prompt string\n result[fieldName] = fieldValue.prompt.prompt;\n }\n }\n }\n }\n }\n\n return result;\n};\n\nexport const loadActionParam = (\n jsonObject: Record<string, any>,\n zodSchema: z.ZodType<any>,\n): Record<string, any> => {\n const locatorFields = findAllMidsceneLocatorField(zodSchema);\n const result = { ...jsonObject };\n\n for (const fieldName of locatorFields) {\n const fieldValue = result[fieldName];\n if (fieldValue && typeof fieldValue === 'string') {\n result[fieldName] = {\n [locateFieldFlagName]: true,\n prompt: fieldValue,\n };\n }\n }\n\n return result;\n};\n\n/**\n * Parse and validate action parameters using Zod schema.\n * All fields are validated through Zod, EXCEPT locator fields which are skipped.\n * Default values defined in the schema are automatically applied.\n *\n * Locator fields are special business logic fields with complex validation requirements,\n * so they are intentionally excluded from Zod parsing and use existing validation logic.\n */\nexport const parseActionParam = (\n rawParam: Record<string, any> | undefined,\n zodSchema?: z.ZodType<any>,\n): Record<string, any> | undefined => {\n // If no schema is provided, return undefined (action takes no parameters)\n if (!zodSchema) {\n return undefined;\n }\n\n // Handle undefined or null rawParam by providing an empty object\n const param = rawParam ?? {};\n\n // Find all locate fields in the schema\n const locateFields = findAllMidsceneLocatorField(zodSchema);\n\n // If there are no locate fields, just do normal validation\n if (locateFields.length === 0) {\n return zodSchema.parse(param);\n }\n\n // Extract locate field values to restore later\n const locateFieldValues: Record<string, any> = {};\n for (const fieldName of locateFields) {\n if (fieldName in param) {\n locateFieldValues[fieldName] = param[fieldName];\n }\n }\n\n // Build params for validation - skip locate fields and use dummy values\n const paramsForValidation: Record<string, any> = {};\n for (const key in param) {\n if (locateFields.includes(key)) {\n // Use dummy value to satisfy schema validation\n paramsForValidation[key] = { prompt: '_dummy_' };\n } else {\n paramsForValidation[key] = param[key];\n }\n }\n\n // Validate with dummy locate values\n const validated = zodSchema.parse(paramsForValidation);\n\n // Restore the actual locate field values (unvalidated, as per business requirement)\n for (const fieldName in locateFieldValues) {\n validated[fieldName] = locateFieldValues[fieldName];\n }\n\n return validated;\n};\n"],"names":["AIActionType","defaultBboxSize","debugInspectUtils","getDebug","fillBboxParam","locate","width","height","rightLimit","bottomLimit","vlMode","adaptBbox","adaptQwen2_5Bbox","bbox","msg","JSON","Error","result","Math","adaptDoubaoBbox","assert","splitted","Number","bboxList","Array","item","x","y","normalizeBboxInput","normalizedBbox","adaptGeminiBbox","normalized01000","left","top","right","bottom","adaptBboxToRect","offsetX","offsetY","rectLeft","rectTop","rectWidth","rectHeight","rect","mergeRects","rects","minLeft","r","minTop","maxRight","maxBottom","expandSearchArea","screenSize","minEdgeSize","defaultPadding","paddingSizeHorizontal","paddingSizeVertical","newWidth","newHeight","newLeft","newTop","markupImageForLLM","screenshotBase64","tree","size","elementsInfo","treeToList","elementsPositionInfoWithoutText","elementInfo","NodeType","imagePayload","compositeElementInfoImg","buildYamlFlowFromPlans","plans","actionSpace","sleep","flow","plan","verb","action","console","flowKey","flowParam","dumpActionParam","flowItem","PointSchema","z","SizeSchema","RectSchema","TMultimodalPromptSchema","TUserPromptSchema","locateFieldFlagName","MidsceneLocationInput","getMidsceneLocationSchema","ifMidsceneLocatorField","field","_actualField__def","_actualField__def1","actualField","shape","dumpMidsceneLocatorField","String","findAllMidsceneLocatorField","zodType","requiredOnly","_zodObject__def","zodObject","keys","Object","key","_field__def","jsonObject","zodSchema","locatorFields","fieldName","fieldValue","loadActionParam","parseActionParam","rawParam","param","locateFields","locateFieldValues","paramsForValidation","validated"],"mappings":";;;;;;AAuBO,IAAKA,sBAAYA,WAAAA,GAAAA,SAAZA,YAAY;;;;;;;WAAZA;;AASZ,MAAMC,kBAAkB;AACxB,MAAMC,oBAAoBC,SAAS;AAI5B,SAASC,cACdC,MAA2B,EAC3BC,KAAa,EACbC,MAAc,EACdC,UAAkB,EAClBC,WAAmB,EACnBC,MAAgC;IAGhC,IAAKL,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,GAAGM,UACZN,OAAO,IAAI,EACXC,OACAC,QACAC,YACAC,aACAC;IAIJ,OAAOL;AACT;AAEO,SAASO,iBACdC,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,GAAGZ;QACN,YAAnB,OAAOY,IAAI,CAAC,EAAE,GACVK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,IAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,GAAGZ;KAC1B;IACD,OAAOgB;AACT;AAEO,SAASE,gBACdN,IAAkC,EAClCP,KAAa,EACbC,MAAc;IAEda,OACEd,QAAQ,KAAKC,SAAS,GACtB;IAGF,IAAI,AAAgB,YAAhB,OAAOM,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,IAAIf,QAAS;YAC3CY,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAId,SAAU;YAC5CW,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIf,QAAS;YAC3CY,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAId,SAAU;SAC7C;QAEH,MAAM,IAAIS,MAAM,CAAC,iDAAiD,EAAEH,MAAM;IAC5E;IAEA,IAAIU,WAAqB,EAAE;IAC3B,IAAIC,MAAM,OAAO,CAACX,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;YAC1BF,SAAS,IAAI,CAACD,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;YAC1BF,SAAS,IAAI,CAACD,OAAOI,EAAE,IAAI,KAAKJ,OAAOK,EAAE,IAAI;QAC/C,OACEJ,SAAS,IAAI,CAACD,OAAOG;IAEzB;SAEAF,WAAWV;IAGb,IAAIU,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAO;QACLL,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGjB,QAAS;QACnCY,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGhB,SAAU;QACpCW,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGjB,QAAS;QACnCY,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGhB,SAAU;KACrC;IAIH,IACEgB,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAO;QACLL,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGjB,QAAS,QAAQL,kBAAkB;QAE/DiB,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGhB,SAAU,QAAQN,kBAAkB;QAEhEiB,KAAK,GAAG,CACNZ,OACAY,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGjB,QAAS,QAAQL,kBAAkB;QAE/DiB,KAAK,GAAG,CACNX,QACAW,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGhB,SAAU,QAAQN,kBAAkB;KAEjE;IAGH,IAAIY,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAO;QACLK,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGjB,QAAS;QACnCY,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGhB,SAAU;QACpCW,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGjB,QAAS;QACnCY,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGhB,SAAU;KACrC;IAGH,MAAMO,MAAM,CAAC,0CAA0C,EAAEC,KAAK,SAAS,CAACF,MAAM,CAAC,CAAC;IAChF,MAAM,IAAIG,MAAMF;AAClB;AAEA,SAASc,mBACPf,IAAoB;IAEpB,IAAIW,MAAM,OAAO,CAACX,OAChB;QAAA,IAAIW,MAAM,OAAO,CAACX,IAAI,CAAC,EAAE,GACvB,OAAOA,IAAI,CAAC,EAAE;IAChB;IAGF,OAAOA;AACT;AAEO,SAASF,UACdE,IAAoB,EACpBP,KAAa,EACbC,MAAc,EACdC,UAAkB,EAClBC,WAAmB,EACnBC,MAAgC;IAEhC,MAAMmB,iBAAiBD,mBAAmBf;IAE1C,IAAII,SAA2C;QAAC;QAAG;QAAG;QAAG;KAAE;IAEzDA,SADEP,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,SACvBS,gBAAgBU,gBAAgBvB,OAAOC,UACvCG,AAAW,aAAXA,SACAoB,gBAAgBD,gBAA4BvB,OAAOC,UACnDG,AAAW,eAAXA,SACAqB,gBAAgBF,gBAA4BvB,OAAOC,UAEnDK,iBAAiBiB;IAG5BZ,MAAM,CAAC,EAAE,GAAGC,KAAK,GAAG,CAACD,MAAM,CAAC,EAAE,EAAET;IAChCS,MAAM,CAAC,EAAE,GAAGC,KAAK,GAAG,CAACD,MAAM,CAAC,EAAE,EAAER;IAEhC,OAAOQ;AACT;AAGO,SAASc,gBACdlB,IAAc,EACdP,KAAa,EACbC,MAAc;IAEd,OAAO;QACLW,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;QAC/BY,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;QAChCW,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;QAC/BY,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;KACjC;AACH;AAGO,SAASuB,gBACdjB,IAAc,EACdP,KAAa,EACbC,MAAc;IAEd,MAAMyB,OAAOd,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;IAC5C,MAAM2B,MAAMf,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;IAC5C,MAAM2B,QAAQhB,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;IAC7C,MAAM6B,SAASjB,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;IAC/C,OAAO;QAACyB;QAAMC;QAAKC;QAAOC;KAAO;AACnC;AAEO,SAASC,gBACdvB,IAAc,EACdP,KAAa,EACbC,MAAc,EACd8B,UAAU,CAAC,EACXC,UAAU,CAAC,EACX9B,aAAaF,KAAK,EAClBG,cAAcF,MAAM,EACpBG,MAAiC;IAEjCR,kBACE,mBACAW,MACAP,OACAC,QACA,UACA8B,SACAC,SACA,SACA9B,YACAC,aACA,UACAC;IAEF,MAAM,CAACsB,MAAMC,KAAKC,OAAOC,OAAO,GAAGxB,UACjCE,MACAP,OACAC,QACAC,YACAC,aACAC;IAIF,MAAM6B,WAAWP;IACjB,MAAMQ,UAAUP;IAChB,IAAIQ,YAAYP,QAAQF;IACxB,IAAIU,aAAaP,SAASF;IAI1B,IAAIM,WAAWE,YAAYnC,OACzBmC,YAAYnC,QAAQiC;IAItB,IAAIC,UAAUE,aAAanC,QACzBmC,aAAanC,SAASiC;IAIxBC,YAAYvB,KAAK,GAAG,CAAC,GAAGuB;IACxBC,aAAaxB,KAAK,GAAG,CAAC,GAAGwB;IAEzB,MAAMC,OAAO;QACX,MAAMJ,WAAWF;QACjB,KAAKG,UAAUF;QACf,OAAOG;QACP,QAAQC;IACV;IACAxC,kBAAkB,4BAA4ByC;IAE9C,OAAOA;AACT;AAEO,SAASC,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,iBACdR,IAAU,EACVS,UAAgB,EAChB1C,MAAgC;IAEhC,MAAM2C,cAAc3C,AAAW,oBAAXA,SAA6B,MAAM;IACvD,MAAM4C,iBAAiB;IAGvB,MAAMC,wBACJZ,KAAK,KAAK,GAAGU,cACTnC,KAAK,IAAI,CAAEmC,AAAAA,CAAAA,cAAcV,KAAK,KAAI,IAAK,KACvCW;IACN,MAAME,sBACJb,KAAK,MAAM,GAAGU,cACVnC,KAAK,IAAI,CAAEmC,AAAAA,CAAAA,cAAcV,KAAK,MAAK,IAAK,KACxCW;IAGN,IAAIG,WAAWvC,KAAK,GAAG,CAACmC,aAAaV,KAAK,KAAK,GAAGY,AAAwB,IAAxBA;IAClD,IAAIG,YAAYxC,KAAK,GAAG,CAACmC,aAAaV,KAAK,MAAM,GAAGa,AAAsB,IAAtBA;IAGpD,IAAIG,UAAUhB,KAAK,IAAI,GAAGY;IAC1B,IAAIK,SAASjB,KAAK,GAAG,GAAGa;IAIxB,IAAIG,UAAUF,WAAWL,WAAW,KAAK,EACvCO,UAAUP,WAAW,KAAK,GAAGK;IAI/B,IAAIG,SAASF,YAAYN,WAAW,MAAM,EACxCQ,SAASR,WAAW,MAAM,GAAGM;IAI/BC,UAAUzC,KAAK,GAAG,CAAC,GAAGyC;IACtBC,SAAS1C,KAAK,GAAG,CAAC,GAAG0C;IAIrB,IAAID,UAAUF,WAAWL,WAAW,KAAK,EACvCK,WAAWL,WAAW,KAAK,GAAGO;IAEhC,IAAIC,SAASF,YAAYN,WAAW,MAAM,EACxCM,YAAYN,WAAW,MAAM,GAAGQ;IAGlCjB,KAAK,IAAI,GAAGgB;IACZhB,KAAK,GAAG,GAAGiB;IACXjB,KAAK,KAAK,GAAGc;IACbd,KAAK,MAAM,GAAGe;IAEd,OAAOf;AACT;AAEO,eAAekB,kBACpBC,gBAAwB,EACxBC,IAAkC,EAClCC,IAAU;IAEV,MAAMC,eAAeC,WAAWH;IAChC,MAAMI,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,gBAAgBT;QAChB,sBAAsBK;QACtBH;IACF;IACA,OAAOM;AACT;AAEO,SAASE,uBACdC,KAAuB,EACvBC,WAAgC,EAChCC,KAAc;IAEd,MAAMC,OAA+B,EAAE;IAEvC,KAAK,MAAMC,QAAQJ,MAAO;QACxB,MAAMK,OAAOD,KAAK,IAAI;QAEtB,MAAME,SAASL,YAAY,IAAI,CAAC,CAACK,SAAWA,OAAO,IAAI,KAAKD;QAC5D,IAAI,CAACC,QAAQ;YACXC,QAAQ,IAAI,CACV,CAAC,sBAAsB,EAAEF,KAAK,8BAA8B,CAAC;YAE/D;QACF;QAEA,MAAMG,UAAUF,OAAO,cAAc,IAAID;QACzC,MAAMI,YAAYH,OAAO,WAAW,GAChCI,gBAAgBN,KAAK,KAAK,IAAI,CAAC,GAAGE,OAAO,WAAW,IACpD,CAAC;QAEL,MAAMK,WAAiC;YACrC,CAACH,QAAQ,EAAE;YACX,GAAGC,SAAS;QACd;QAEAN,KAAK,IAAI,CAACQ;IACZ;IAEA,IAAIT,OACFC,KAAK,IAAI,CAAC;QACRD;IACF;IAGF,OAAOC;AACT;AAGO,MAAMS,cAAcC,EAAE,MAAM,CAAC;IAClC,MAAMA,EAAE,MAAM;IACd,KAAKA,EAAE,MAAM;AACf;AAEO,MAAMC,aAAaD,EAAE,MAAM,CAAC;IACjC,OAAOA,EAAE,MAAM;IACf,QAAQA,EAAE,MAAM;IAChB,KAAKA,EAAE,MAAM,GAAG,QAAQ;AAC1B;AAEO,MAAME,aAAaH,YAAY,GAAG,CAACE,YAAY,GAAG,CACvDD,EAAE,MAAM,CAAC;IACP,MAAMA,EAAE,MAAM,GAAG,QAAQ;AAC3B;AAIK,MAAMG,0BAA0BH,EAAE,MAAM,CAAC;IAC9C,QAAQA,EAAAA,KACA,CACJA,EAAE,MAAM,CAAC;QACP,MAAMA,EAAE,MAAM;QACd,KAAKA,EAAE,MAAM;IACf,IAED,QAAQ;IACX,yBAAyBA,EAAE,OAAO,GAAG,QAAQ;AAC/C;AAGO,MAAMI,oBAAoBJ,EAAE,KAAK,CAAC;IACvCA,EAAE,MAAM;IACRA,EAAAA,MACS,CAAC;QACN,QAAQA,EAAE,MAAM;IAClB,GACC,GAAG,CAACG,wBAAwB,OAAO;CACvC;AAMD,MAAME,sBAAsB;AAG5B,MAAMC,wBAAwBN,EAAAA,MACrB,CAAC;IACN,QAAQI;IACR,WAAWJ,EAAE,OAAO,GAAG,QAAQ;IAC/B,WAAWA,EAAE,OAAO,GAAG,QAAQ;IAC/B,OAAOA,EAAE,KAAK,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,OAAO;KAAG,EAAE,QAAQ;AACpD,GACC,WAAW;AAGiBA,EAAAA,MACtB,CAAC;IACN,CAACK,oBAAoB,EAAEL,EAAE,OAAO,CAAC;IACjC,QAAQI;IAGR,WAAWJ,EAAE,OAAO,GAAG,QAAQ;IAC/B,WAAWA,EAAE,OAAO,GAAG,QAAQ;IAC/B,OAAOA,EAAE,OAAO,GAAG,QAAQ;IAG3B,QAAQA,EAAE,KAAK,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,MAAM;KAAG;IACxC,MAAME;AACR,GACC,WAAW;AAYP,MAAMK,4BAA4B,IAChCD;AAGF,MAAME,yBAAyB,CAACC;QAGjCC,mBAKAC;IANJ,IAAIC,cAAcH;IAClB,IAAIC,AAAAA,SAAAA,CAAAA,oBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,kBAAkB,QAAQ,AAAD,MAAM,eACjCE,cAAcA,YAAY,IAAI,CAAC,SAAS;IAI1C,IAAID,AAAAA,SAAAA,CAAAA,qBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,mBAAkB,QAAQ,AAAD,MAAM,aAAa;QAC9C,MAAME,QAAQD,YAAY,IAAI,CAAC,KAAK;QAGpC,IAAIP,uBAAuBQ,OACzB,OAAO;QAKT,IAAI,YAAYA,SAASA,MAAM,MAAM,EACnC,OAAO;IAEX;IAEA,OAAO;AACT;AAEO,MAAMC,2BAA2B,CAACL;IACvC3E,OACE0E,uBAAuBC,QACvB;IAIF,IAAI,AAAiB,YAAjB,OAAOA,OACT,OAAOA;IAIT,IAAIA,SAAS,AAAiB,YAAjB,OAAOA,SAAsBA,MAAM,MAAM,EAAE;QAEtD,IAAI,AAAwB,YAAxB,OAAOA,MAAM,MAAM,EACrB,OAAOA,MAAM,MAAM;QAGrB,IAAI,AAAwB,YAAxB,OAAOA,MAAM,MAAM,IAAiBA,MAAM,MAAM,CAAC,MAAM,EACzD,OAAOA,MAAM,MAAM,CAAC,MAAM;IAE9B;IAGA,OAAOM,OAAON;AAChB;AAEO,MAAMO,8BAA8B,CACzCC,SACAC;QAQIC;IANJ,IAAI,CAACF,SACH,OAAO,EAAE;IAIX,MAAMG,YAAYH;IAClB,IAAIE,AAAAA,SAAAA,CAAAA,kBAAAA,UAAU,IAAI,AAAD,IAAbA,KAAAA,IAAAA,gBAAgB,QAAQ,AAAD,MAAM,eAAeC,UAAU,KAAK,EAAE;QAC/D,MAAMC,OAAOC,OAAO,IAAI,CAACF,UAAU,KAAK;QACxC,OAAOC,KAAK,MAAM,CAAC,CAACE;YAClB,MAAMd,QAAQW,UAAU,KAAK,CAACG,IAAI;YAClC,IAAI,CAACf,uBAAuBC,QAC1B,OAAO;YAIT,IAAIS,cAAc;oBACTM;gBAAP,OAAOA,AAAAA,SAAAA,CAAAA,cAAAA,MAAM,IAAI,AAAD,IAATA,KAAAA,IAAAA,YAAY,QAAQ,AAAD,MAAM;YAClC;YAEA,OAAO;QACT;IACF;IAGA,OAAO,EAAE;AACX;AAEO,MAAM3B,kBAAkB,CAC7B4B,YACAC;IAEA,MAAMC,gBAAgBX,4BAA4BU;IAClD,MAAM/F,SAAS;QAAE,GAAG8F,UAAU;IAAC;IAE/B,KAAK,MAAMG,aAAaD,cAAe;QACrC,MAAME,aAAalG,MAAM,CAACiG,UAAU;QACpC,IAAIC,YAEF;YAAA,IAAI,AAAsB,YAAtB,OAAOA,YACTlG,MAAM,CAACiG,UAAU,GAAGC;iBACf,IAAI,AAAsB,YAAtB,OAAOA,YAEhB;gBAAA,IAAIA,WAAW,MAAM,EAEnB;oBAAA,IAAI,AAA6B,YAA7B,OAAOA,WAAW,MAAM,EAC1BlG,MAAM,CAACiG,UAAU,GAAGC,WAAW,MAAM;yBAChC,IACL,AAA6B,YAA7B,OAAOA,WAAW,MAAM,IACxBA,WAAW,MAAM,CAAC,MAAM,EAGxBlG,MAAM,CAACiG,UAAU,GAAGC,WAAW,MAAM,CAAC,MAAM;gBAC9C;YACF;QACF;IAEJ;IAEA,OAAOlG;AACT;AAEO,MAAMmG,kBAAkB,CAC7BL,YACAC;IAEA,MAAMC,gBAAgBX,4BAA4BU;IAClD,MAAM/F,SAAS;QAAE,GAAG8F,UAAU;IAAC;IAE/B,KAAK,MAAMG,aAAaD,cAAe;QACrC,MAAME,aAAalG,MAAM,CAACiG,UAAU;QACpC,IAAIC,cAAc,AAAsB,YAAtB,OAAOA,YACvBlG,MAAM,CAACiG,UAAU,GAAG;YAClB,CAACvB,oBAAoB,EAAE;YACvB,QAAQwB;QACV;IAEJ;IAEA,OAAOlG;AACT;AAUO,MAAMoG,mBAAmB,CAC9BC,UACAN;IAGA,IAAI,CAACA,WACH;IAIF,MAAMO,QAAQD,YAAY,CAAC;IAG3B,MAAME,eAAelB,4BAA4BU;IAGjD,IAAIQ,AAAwB,MAAxBA,aAAa,MAAM,EACrB,OAAOR,UAAU,KAAK,CAACO;IAIzB,MAAME,oBAAyC,CAAC;IAChD,KAAK,MAAMP,aAAaM,aACtB,IAAIN,aAAaK,OACfE,iBAAiB,CAACP,UAAU,GAAGK,KAAK,CAACL,UAAU;IAKnD,MAAMQ,sBAA2C,CAAC;IAClD,IAAK,MAAMb,OAAOU,MAChB,IAAIC,aAAa,QAAQ,CAACX,MAExBa,mBAAmB,CAACb,IAAI,GAAG;QAAE,QAAQ;IAAU;SAE/Ca,mBAAmB,CAACb,IAAI,GAAGU,KAAK,CAACV,IAAI;IAKzC,MAAMc,YAAYX,UAAU,KAAK,CAACU;IAGlC,IAAK,MAAMR,aAAaO,kBACtBE,SAAS,CAACT,UAAU,GAAGO,iBAAiB,CAACP,UAAU;IAGrD,OAAOS;AACT"}
|
|
1
|
+
{"version":3,"file":"common.mjs","sources":["../../src/common.ts"],"sourcesContent":["import type {\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 { ChatCompletionMessageParam } from 'openai/resources/index';\n\nimport type { PlanningLocateParam } from '@/types';\nimport { NodeType } from '@midscene/shared/constants';\nimport type { TVlModeTypes } from '@midscene/shared/env';\nimport { treeToList } from '@midscene/shared/extractor';\nimport { compositeElementInfoImg } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { z } from 'zod';\n\nexport type AIArgs = ChatCompletionMessageParam[];\n\nexport enum AIActionType {\n ASSERT = 0,\n INSPECT_ELEMENT = 1,\n EXTRACT_DATA = 2,\n PLAN = 3,\n DESCRIBE_ELEMENT = 4,\n TEXT = 5,\n}\n\nconst defaultBboxSize = 20; // must be even number\nconst debugInspectUtils = getDebug('ai:common');\ntype AdaptBboxInput = number[] | string[] | string | (number[] | string[])[];\n\n// transform the param of locate from qwen mode\nexport function fillBboxParam(\n locate: PlanningLocateParam,\n width: number,\n height: number,\n rightLimit: number,\n bottomLimit: number,\n vlMode: TVlModeTypes | undefined,\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(\n locate.bbox,\n width,\n height,\n rightLimit,\n bottomLimit,\n vlMode,\n );\n }\n\n return locate;\n}\n\nexport function adaptQwen2_5Bbox(\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 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\nfunction normalizeBboxInput(\n bbox: AdaptBboxInput,\n): number[] | string[] | string {\n if (Array.isArray(bbox)) {\n if (Array.isArray(bbox[0])) {\n return bbox[0] as number[] | string[];\n }\n return bbox as number[] | string[];\n }\n return bbox as string;\n}\n\nexport function adaptBbox(\n bbox: AdaptBboxInput,\n width: number,\n height: number,\n rightLimit: number,\n bottomLimit: number,\n vlMode: TVlModeTypes | undefined,\n): [number, number, number, number] {\n const normalizedBbox = normalizeBboxInput(bbox);\n\n let result: [number, number, number, number] = [0, 0, 0, 0];\n if (vlMode === 'doubao-vision' || vlMode === 'vlm-ui-tars') {\n result = adaptDoubaoBbox(normalizedBbox, width, height);\n } else if (vlMode === 'gemini') {\n result = adaptGeminiBbox(normalizedBbox as number[], width, height);\n } else if (vlMode === 'qwen3-vl') {\n result = normalized01000(normalizedBbox as number[], width, height);\n } else {\n result = adaptQwen2_5Bbox(normalizedBbox as number[]);\n }\n\n result[2] = Math.min(result[2], rightLimit);\n result[3] = Math.min(result[3], bottomLimit);\n\n return result;\n}\n\n// x1, y1, x2, y2 -> 0-1000\nexport function normalized01000(\n bbox: number[],\n width: number,\n height: number,\n): [number, number, number, number] {\n return [\n Math.round((bbox[0] * width) / 1000),\n Math.round((bbox[1] * height) / 1000),\n Math.round((bbox[2] * width) / 1000),\n Math.round((bbox[3] * height) / 1000),\n ];\n}\n\n// y1, x1, y2, x2 -> 0-1000\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 rightLimit = width,\n bottomLimit = height,\n vlMode?: TVlModeTypes | undefined,\n): Rect {\n debugInspectUtils(\n 'adaptBboxToRect',\n bbox,\n width,\n height,\n 'offset',\n offsetX,\n offsetY,\n 'limit',\n rightLimit,\n bottomLimit,\n 'vlMode',\n vlMode,\n );\n const [left, top, right, bottom] = adaptBbox(\n bbox,\n width,\n height,\n rightLimit,\n bottomLimit,\n vlMode,\n );\n\n // Calculate initial rect dimensions\n const rectLeft = left;\n const rectTop = top;\n let rectWidth = right - left;\n let rectHeight = bottom - top;\n\n // Ensure the rect doesn't exceed image boundaries\n // If right edge exceeds width, adjust the width\n if (rectLeft + rectWidth > width) {\n rectWidth = width - rectLeft;\n }\n\n // If bottom edge exceeds height, adjust the height\n if (rectTop + rectHeight > height) {\n rectHeight = height - rectTop;\n }\n\n // Ensure minimum dimensions (width and height should be at least 1)\n rectWidth = Math.max(1, rectWidth);\n rectHeight = Math.max(1, rectHeight);\n\n const rect = {\n left: rectLeft + offsetX,\n top: rectTop + offsetY,\n width: rectWidth,\n height: rectHeight,\n };\n debugInspectUtils('adaptBboxToRect, result=', rect);\n\n return rect;\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(\n rect: Rect,\n screenSize: Size,\n vlMode: TVlModeTypes | undefined,\n) {\n const minEdgeSize = vlMode === 'doubao-vision' ? 500 : 300;\n const defaultPadding = 160;\n\n // Calculate padding needed to reach minimum edge size\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\n // Calculate new dimensions (ensure minimum edge size)\n let newWidth = Math.max(minEdgeSize, rect.width + paddingSizeHorizontal * 2);\n let newHeight = Math.max(minEdgeSize, rect.height + paddingSizeVertical * 2);\n\n // Calculate initial position with padding\n let newLeft = rect.left - paddingSizeHorizontal;\n let newTop = rect.top - paddingSizeVertical;\n\n // Ensure the rect doesn't exceed screen boundaries by adjusting position\n // If the rect goes beyond the right edge, shift it left\n if (newLeft + newWidth > screenSize.width) {\n newLeft = screenSize.width - newWidth;\n }\n\n // If the rect goes beyond the bottom edge, shift it up\n if (newTop + newHeight > screenSize.height) {\n newTop = screenSize.height - newHeight;\n }\n\n // Ensure the rect doesn't go beyond the left/top edges\n newLeft = Math.max(0, newLeft);\n newTop = Math.max(0, newTop);\n\n // If after position adjustment, the rect still exceeds screen boundaries,\n // clamp the dimensions to fit within screen\n if (newLeft + newWidth > screenSize.width) {\n newWidth = screenSize.width - newLeft;\n }\n if (newTop + newHeight > screenSize.height) {\n newHeight = screenSize.height - newTop;\n }\n\n rect.left = newLeft;\n rect.top = newTop;\n rect.width = newWidth;\n rect.height = newHeight;\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<any>[],\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 flowKey = action.interfaceAlias || verb;\n const flowParam = action.paramSchema\n ? dumpActionParam(plan.param || {}, action.paramSchema)\n : {};\n\n const flowItem: MidsceneYamlFlowItem = {\n [flowKey]: '',\n ...flowParam,\n };\n\n flow.push(flowItem);\n }\n\n if (sleep) {\n flow.push({\n sleep,\n });\n }\n\n return flow;\n}\n\n// Zod schemas for shared types\nexport const PointSchema = z.object({\n left: z.number(),\n top: z.number(),\n});\n\nexport const SizeSchema = z.object({\n width: z.number(),\n height: z.number(),\n dpr: z.number().optional(),\n});\n\nexport const RectSchema = PointSchema.and(SizeSchema).and(\n z.object({\n zoom: z.number().optional(),\n }),\n);\n\n// Zod schema for TMultimodalPrompt\nexport const TMultimodalPromptSchema = z.object({\n images: z\n .array(\n z.object({\n name: z.string(),\n url: z.string(),\n }),\n )\n .optional(),\n convertHttpImage2Base64: z.boolean().optional(),\n});\n\n// Zod schema for TUserPrompt\nexport const TUserPromptSchema = z.union([\n z.string(),\n z\n .object({\n prompt: z.string(),\n })\n .and(TMultimodalPromptSchema.partial()),\n]);\n\n// Generate TypeScript types from Zod schemas\nexport type TMultimodalPrompt = z.infer<typeof TMultimodalPromptSchema>;\nexport type TUserPrompt = z.infer<typeof TUserPromptSchema>;\n\nconst locateFieldFlagName = 'midscene_location_field_flag';\n\n// Schema for locator field input (when users provide locate parameters)\nconst MidsceneLocationInput = z\n .object({\n prompt: TUserPromptSchema,\n deepThink: z.boolean().optional(),\n cacheable: z.boolean().optional(),\n xpath: z.union([z.string(), z.boolean()]).optional(),\n })\n .passthrough();\n\n// Schema for locator field result (when AI returns locate results)\nconst MidsceneLocationResult = z\n .object({\n [locateFieldFlagName]: z.literal(true),\n prompt: TUserPromptSchema,\n\n // optional fields\n deepThink: z.boolean().optional(), // only available in vl model\n cacheable: z.boolean().optional(),\n xpath: z.boolean().optional(), // preset result for xpath\n\n // these two fields will only appear in the result\n center: z.tuple([z.number(), z.number()]),\n rect: RectSchema,\n })\n .passthrough();\n\n// Export the result type - this is used for runtime results that include center and rect\nexport type MidsceneLocationResultType = z.infer<typeof MidsceneLocationResult>;\n\n// Export the input type - this is the inferred type from getMidsceneLocationSchema()\nexport type MidsceneLocationInputType = z.infer<typeof MidsceneLocationInput>;\n\n/**\n * Returns the schema for locator fields.\n * This now returns the input schema which is more permissive and suitable for validation.\n */\nexport const getMidsceneLocationSchema = () => {\n return MidsceneLocationInput;\n};\n\nexport const ifMidsceneLocatorField = (field: any): boolean => {\n // Handle optional fields by getting the inner type\n let actualField = field;\n if (actualField._def?.typeName === 'ZodOptional') {\n actualField = actualField._def.innerType;\n }\n\n // Check if this is a ZodObject\n if (actualField._def?.typeName === 'ZodObject') {\n const shape = actualField._def.shape();\n\n // Method 1: Check for the location field flag (for result schema)\n if (locateFieldFlagName in shape) {\n return true;\n }\n\n // Method 2: Check if it's the input schema by checking for 'prompt' field\n // Input schema has 'prompt' as a required field\n if ('prompt' in shape && shape.prompt) {\n return true;\n }\n }\n\n return false;\n};\n\nexport const dumpMidsceneLocatorField = (field: any): string => {\n assert(\n ifMidsceneLocatorField(field),\n 'field is not a midscene locator field',\n );\n\n // If field is a string, return it directly\n if (typeof field === 'string') {\n return field;\n }\n\n // If field is an object with prompt property\n if (field && typeof field === 'object' && field.prompt) {\n // If prompt is a string, return it directly\n if (typeof field.prompt === 'string') {\n return field.prompt;\n }\n // If prompt is a TUserPrompt object, extract the prompt string\n if (typeof field.prompt === 'object' && field.prompt.prompt) {\n return field.prompt.prompt; // TODO: dump images if necessary\n }\n }\n\n // Fallback: try to convert to string\n return String(field);\n};\n\nexport const findAllMidsceneLocatorField = (\n zodType?: z.ZodType<any>,\n requiredOnly?: boolean,\n): string[] => {\n if (!zodType) {\n return [];\n }\n\n // Check if this is a ZodObject by checking if it has a shape property\n const zodObject = zodType as any;\n if (zodObject._def?.typeName === 'ZodObject' && zodObject.shape) {\n const keys = Object.keys(zodObject.shape);\n return keys.filter((key) => {\n const field = zodObject.shape[key];\n if (!ifMidsceneLocatorField(field)) {\n return false;\n }\n\n // If requiredOnly is true, filter out optional fields\n if (requiredOnly) {\n return field._def?.typeName !== 'ZodOptional';\n }\n\n return true;\n });\n }\n\n // For other ZodType instances, we can't extract field names\n return [];\n};\n\nexport const dumpActionParam = (\n jsonObject: Record<string, any>,\n zodSchema: z.ZodType<any>,\n): Record<string, any> => {\n const locatorFields = findAllMidsceneLocatorField(zodSchema);\n const result = { ...jsonObject };\n\n for (const fieldName of locatorFields) {\n const fieldValue = result[fieldName];\n if (fieldValue) {\n // If it's already a string, keep it as is\n if (typeof fieldValue === 'string') {\n result[fieldName] = fieldValue;\n } else if (typeof fieldValue === 'object') {\n // Check if this field is actually a MidsceneLocationType object\n if (fieldValue.prompt) {\n // If prompt is a string, use it directly\n if (typeof fieldValue.prompt === 'string') {\n result[fieldName] = fieldValue.prompt;\n } else if (\n typeof fieldValue.prompt === 'object' &&\n fieldValue.prompt.prompt\n ) {\n // If prompt is a TUserPrompt object, extract the prompt string\n result[fieldName] = fieldValue.prompt.prompt;\n }\n }\n }\n }\n }\n\n return result;\n};\n\nexport const loadActionParam = (\n jsonObject: Record<string, any>,\n zodSchema: z.ZodType<any>,\n): Record<string, any> => {\n const locatorFields = findAllMidsceneLocatorField(zodSchema);\n const result = { ...jsonObject };\n\n for (const fieldName of locatorFields) {\n const fieldValue = result[fieldName];\n if (fieldValue && typeof fieldValue === 'string') {\n result[fieldName] = {\n [locateFieldFlagName]: true,\n prompt: fieldValue,\n };\n }\n }\n\n return result;\n};\n\n/**\n * Parse and validate action parameters using Zod schema.\n * All fields are validated through Zod, EXCEPT locator fields which are skipped.\n * Default values defined in the schema are automatically applied.\n *\n * Locator fields are special business logic fields with complex validation requirements,\n * so they are intentionally excluded from Zod parsing and use existing validation logic.\n */\nexport const parseActionParam = (\n rawParam: Record<string, any> | undefined,\n zodSchema?: z.ZodType<any>,\n): Record<string, any> | undefined => {\n // If no schema is provided, return undefined (action takes no parameters)\n if (!zodSchema) {\n return undefined;\n }\n\n // Handle undefined or null rawParam by providing an empty object\n const param = rawParam ?? {};\n\n // Find all locate fields in the schema\n const locateFields = findAllMidsceneLocatorField(zodSchema);\n\n // If there are no locate fields, just do normal validation\n if (locateFields.length === 0) {\n return zodSchema.parse(param);\n }\n\n // Extract locate field values to restore later\n const locateFieldValues: Record<string, any> = {};\n for (const fieldName of locateFields) {\n if (fieldName in param) {\n locateFieldValues[fieldName] = param[fieldName];\n }\n }\n\n // Build params for validation - skip locate fields and use dummy values\n const paramsForValidation: Record<string, any> = {};\n for (const key in param) {\n if (locateFields.includes(key)) {\n // Use dummy value to satisfy schema validation\n paramsForValidation[key] = { prompt: '_dummy_' };\n } else {\n paramsForValidation[key] = param[key];\n }\n }\n\n // Validate with dummy locate values\n const validated = zodSchema.parse(paramsForValidation);\n\n // Restore the actual locate field values (unvalidated, as per business requirement)\n for (const fieldName in locateFieldValues) {\n validated[fieldName] = locateFieldValues[fieldName];\n }\n\n return validated;\n};\n"],"names":["AIActionType","defaultBboxSize","debugInspectUtils","getDebug","fillBboxParam","locate","width","height","rightLimit","bottomLimit","vlMode","adaptBbox","adaptQwen2_5Bbox","bbox","msg","JSON","Error","result","Math","adaptDoubaoBbox","assert","splitted","Number","bboxList","Array","item","x","y","normalizeBboxInput","normalizedBbox","adaptGeminiBbox","normalized01000","left","top","right","bottom","adaptBboxToRect","offsetX","offsetY","rectLeft","rectTop","rectWidth","rectHeight","rect","mergeRects","rects","minLeft","r","minTop","maxRight","maxBottom","expandSearchArea","screenSize","minEdgeSize","defaultPadding","paddingSizeHorizontal","paddingSizeVertical","newWidth","newHeight","newLeft","newTop","markupImageForLLM","screenshotBase64","tree","size","elementsInfo","treeToList","elementsPositionInfoWithoutText","elementInfo","NodeType","imagePayload","compositeElementInfoImg","buildYamlFlowFromPlans","plans","actionSpace","sleep","flow","plan","verb","action","console","flowKey","flowParam","dumpActionParam","flowItem","PointSchema","z","SizeSchema","RectSchema","TMultimodalPromptSchema","TUserPromptSchema","locateFieldFlagName","MidsceneLocationInput","getMidsceneLocationSchema","ifMidsceneLocatorField","field","actualField","shape","dumpMidsceneLocatorField","String","findAllMidsceneLocatorField","zodType","requiredOnly","zodObject","keys","Object","key","jsonObject","zodSchema","locatorFields","fieldName","fieldValue","loadActionParam","parseActionParam","rawParam","param","locateFields","locateFieldValues","paramsForValidation","validated"],"mappings":";;;;;;AAuBO,IAAKA,sBAAYA,WAAAA,GAAAA,SAAZA,YAAY;;;;;;;WAAZA;;AASZ,MAAMC,kBAAkB;AACxB,MAAMC,oBAAoBC,SAAS;AAI5B,SAASC,cACdC,MAA2B,EAC3BC,KAAa,EACbC,MAAc,EACdC,UAAkB,EAClBC,WAAmB,EACnBC,MAAgC;IAGhC,IAAKL,OAAe,OAAO,IAAI,CAACA,QAAQ,MAAM;QAC5CA,OAAO,IAAI,GAAIA,OAAe,OAAO;QAErC,OAAQA,OAAe,OAAO;IAChC;IAEA,IAAIA,QAAQ,MACVA,OAAO,IAAI,GAAGM,UACZN,OAAO,IAAI,EACXC,OACAC,QACAC,YACAC,aACAC;IAIJ,OAAOL;AACT;AAEO,SAASO,iBACdC,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,GAAGZ;QACN,YAAnB,OAAOY,IAAI,CAAC,EAAE,GACVK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,IAClBK,KAAK,KAAK,CAACL,IAAI,CAAC,EAAE,GAAGZ;KAC1B;IACD,OAAOgB;AACT;AAEO,SAASE,gBACdN,IAAkC,EAClCP,KAAa,EACbC,MAAc;IAEda,OACEd,QAAQ,KAAKC,SAAS,GACtB;IAGF,IAAI,AAAgB,YAAhB,OAAOM,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,IAAIf,QAAS;YAC3CY,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAId,SAAU;YAC5CW,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAIf,QAAS;YAC3CY,KAAK,KAAK,CAAEI,OAAOD,QAAQ,CAAC,EAAE,IAAId,SAAU;SAC7C;QAEH,MAAM,IAAIS,MAAM,CAAC,iDAAiD,EAAEH,MAAM;IAC5E;IAEA,IAAIU,WAAqB,EAAE;IAC3B,IAAIC,MAAM,OAAO,CAACX,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;YAC1BF,SAAS,IAAI,CAACD,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;YAC1BF,SAAS,IAAI,CAACD,OAAOI,EAAE,IAAI,KAAKJ,OAAOK,EAAE,IAAI;QAC/C,OACEJ,SAAS,IAAI,CAACD,OAAOG;IAEzB;SAEAF,WAAWV;IAGb,IAAIU,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAO;QACLL,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGjB,QAAS;QACnCY,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGhB,SAAU;QACpCW,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGjB,QAAS;QACnCY,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGhB,SAAU;KACrC;IAIH,IACEgB,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAO;QACLL,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGjB,QAAS,QAAQL,kBAAkB;QAE/DiB,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGhB,SAAU,QAAQN,kBAAkB;QAEhEiB,KAAK,GAAG,CACNZ,OACAY,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGjB,QAAS,QAAQL,kBAAkB;QAE/DiB,KAAK,GAAG,CACNX,QACAW,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGhB,SAAU,QAAQN,kBAAkB;KAEjE;IAGH,IAAIY,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAO;QACLK,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGjB,QAAS;QACnCY,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGhB,SAAU;QACpCW,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGjB,QAAS;QACnCY,KAAK,KAAK,CAAEK,QAAQ,CAAC,EAAE,GAAGhB,SAAU;KACrC;IAGH,MAAMO,MAAM,CAAC,0CAA0C,EAAEC,KAAK,SAAS,CAACF,MAAM,CAAC,CAAC;IAChF,MAAM,IAAIG,MAAMF;AAClB;AAEA,SAASc,mBACPf,IAAoB;IAEpB,IAAIW,MAAM,OAAO,CAACX,OAChB;QAAA,IAAIW,MAAM,OAAO,CAACX,IAAI,CAAC,EAAE,GACvB,OAAOA,IAAI,CAAC,EAAE;IAChB;IAGF,OAAOA;AACT;AAEO,SAASF,UACdE,IAAoB,EACpBP,KAAa,EACbC,MAAc,EACdC,UAAkB,EAClBC,WAAmB,EACnBC,MAAgC;IAEhC,MAAMmB,iBAAiBD,mBAAmBf;IAE1C,IAAII,SAA2C;QAAC;QAAG;QAAG;QAAG;KAAE;IAEzDA,SADEP,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,SACvBS,gBAAgBU,gBAAgBvB,OAAOC,UACvCG,AAAW,aAAXA,SACAoB,gBAAgBD,gBAA4BvB,OAAOC,UACnDG,AAAW,eAAXA,SACAqB,gBAAgBF,gBAA4BvB,OAAOC,UAEnDK,iBAAiBiB;IAG5BZ,MAAM,CAAC,EAAE,GAAGC,KAAK,GAAG,CAACD,MAAM,CAAC,EAAE,EAAET;IAChCS,MAAM,CAAC,EAAE,GAAGC,KAAK,GAAG,CAACD,MAAM,CAAC,EAAE,EAAER;IAEhC,OAAOQ;AACT;AAGO,SAASc,gBACdlB,IAAc,EACdP,KAAa,EACbC,MAAc;IAEd,OAAO;QACLW,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;QAC/BY,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;QAChCW,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;QAC/BY,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;KACjC;AACH;AAGO,SAASuB,gBACdjB,IAAc,EACdP,KAAa,EACbC,MAAc;IAEd,MAAMyB,OAAOd,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;IAC5C,MAAM2B,MAAMf,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;IAC5C,MAAM2B,QAAQhB,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGP,QAAS;IAC7C,MAAM6B,SAASjB,KAAK,KAAK,CAAEL,IAAI,CAAC,EAAE,GAAGN,SAAU;IAC/C,OAAO;QAACyB;QAAMC;QAAKC;QAAOC;KAAO;AACnC;AAEO,SAASC,gBACdvB,IAAc,EACdP,KAAa,EACbC,MAAc,EACd8B,UAAU,CAAC,EACXC,UAAU,CAAC,EACX9B,aAAaF,KAAK,EAClBG,cAAcF,MAAM,EACpBG,MAAiC;IAEjCR,kBACE,mBACAW,MACAP,OACAC,QACA,UACA8B,SACAC,SACA,SACA9B,YACAC,aACA,UACAC;IAEF,MAAM,CAACsB,MAAMC,KAAKC,OAAOC,OAAO,GAAGxB,UACjCE,MACAP,OACAC,QACAC,YACAC,aACAC;IAIF,MAAM6B,WAAWP;IACjB,MAAMQ,UAAUP;IAChB,IAAIQ,YAAYP,QAAQF;IACxB,IAAIU,aAAaP,SAASF;IAI1B,IAAIM,WAAWE,YAAYnC,OACzBmC,YAAYnC,QAAQiC;IAItB,IAAIC,UAAUE,aAAanC,QACzBmC,aAAanC,SAASiC;IAIxBC,YAAYvB,KAAK,GAAG,CAAC,GAAGuB;IACxBC,aAAaxB,KAAK,GAAG,CAAC,GAAGwB;IAEzB,MAAMC,OAAO;QACX,MAAMJ,WAAWF;QACjB,KAAKG,UAAUF;QACf,OAAOG;QACP,QAAQC;IACV;IACAxC,kBAAkB,4BAA4ByC;IAE9C,OAAOA;AACT;AAEO,SAASC,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,iBACdR,IAAU,EACVS,UAAgB,EAChB1C,MAAgC;IAEhC,MAAM2C,cAAc3C,AAAW,oBAAXA,SAA6B,MAAM;IACvD,MAAM4C,iBAAiB;IAGvB,MAAMC,wBACJZ,KAAK,KAAK,GAAGU,cACTnC,KAAK,IAAI,CAAEmC,AAAAA,CAAAA,cAAcV,KAAK,KAAI,IAAK,KACvCW;IACN,MAAME,sBACJb,KAAK,MAAM,GAAGU,cACVnC,KAAK,IAAI,CAAEmC,AAAAA,CAAAA,cAAcV,KAAK,MAAK,IAAK,KACxCW;IAGN,IAAIG,WAAWvC,KAAK,GAAG,CAACmC,aAAaV,KAAK,KAAK,GAAGY,AAAwB,IAAxBA;IAClD,IAAIG,YAAYxC,KAAK,GAAG,CAACmC,aAAaV,KAAK,MAAM,GAAGa,AAAsB,IAAtBA;IAGpD,IAAIG,UAAUhB,KAAK,IAAI,GAAGY;IAC1B,IAAIK,SAASjB,KAAK,GAAG,GAAGa;IAIxB,IAAIG,UAAUF,WAAWL,WAAW,KAAK,EACvCO,UAAUP,WAAW,KAAK,GAAGK;IAI/B,IAAIG,SAASF,YAAYN,WAAW,MAAM,EACxCQ,SAASR,WAAW,MAAM,GAAGM;IAI/BC,UAAUzC,KAAK,GAAG,CAAC,GAAGyC;IACtBC,SAAS1C,KAAK,GAAG,CAAC,GAAG0C;IAIrB,IAAID,UAAUF,WAAWL,WAAW,KAAK,EACvCK,WAAWL,WAAW,KAAK,GAAGO;IAEhC,IAAIC,SAASF,YAAYN,WAAW,MAAM,EACxCM,YAAYN,WAAW,MAAM,GAAGQ;IAGlCjB,KAAK,IAAI,GAAGgB;IACZhB,KAAK,GAAG,GAAGiB;IACXjB,KAAK,KAAK,GAAGc;IACbd,KAAK,MAAM,GAAGe;IAEd,OAAOf;AACT;AAEO,eAAekB,kBACpBC,gBAAwB,EACxBC,IAAkC,EAClCC,IAAU;IAEV,MAAMC,eAAeC,WAAWH;IAChC,MAAMI,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,gBAAgBT;QAChB,sBAAsBK;QACtBH;IACF;IACA,OAAOM;AACT;AAEO,SAASE,uBACdC,KAAuB,EACvBC,WAAgC,EAChCC,KAAc;IAEd,MAAMC,OAA+B,EAAE;IAEvC,KAAK,MAAMC,QAAQJ,MAAO;QACxB,MAAMK,OAAOD,KAAK,IAAI;QAEtB,MAAME,SAASL,YAAY,IAAI,CAAC,CAACK,SAAWA,OAAO,IAAI,KAAKD;QAC5D,IAAI,CAACC,QAAQ;YACXC,QAAQ,IAAI,CACV,CAAC,sBAAsB,EAAEF,KAAK,8BAA8B,CAAC;YAE/D;QACF;QAEA,MAAMG,UAAUF,OAAO,cAAc,IAAID;QACzC,MAAMI,YAAYH,OAAO,WAAW,GAChCI,gBAAgBN,KAAK,KAAK,IAAI,CAAC,GAAGE,OAAO,WAAW,IACpD,CAAC;QAEL,MAAMK,WAAiC;YACrC,CAACH,QAAQ,EAAE;YACX,GAAGC,SAAS;QACd;QAEAN,KAAK,IAAI,CAACQ;IACZ;IAEA,IAAIT,OACFC,KAAK,IAAI,CAAC;QACRD;IACF;IAGF,OAAOC;AACT;AAGO,MAAMS,cAAcC,EAAE,MAAM,CAAC;IAClC,MAAMA,EAAE,MAAM;IACd,KAAKA,EAAE,MAAM;AACf;AAEO,MAAMC,aAAaD,EAAE,MAAM,CAAC;IACjC,OAAOA,EAAE,MAAM;IACf,QAAQA,EAAE,MAAM;IAChB,KAAKA,EAAE,MAAM,GAAG,QAAQ;AAC1B;AAEO,MAAME,aAAaH,YAAY,GAAG,CAACE,YAAY,GAAG,CACvDD,EAAE,MAAM,CAAC;IACP,MAAMA,EAAE,MAAM,GAAG,QAAQ;AAC3B;AAIK,MAAMG,0BAA0BH,EAAE,MAAM,CAAC;IAC9C,QAAQA,EAAAA,KACA,CACJA,EAAE,MAAM,CAAC;QACP,MAAMA,EAAE,MAAM;QACd,KAAKA,EAAE,MAAM;IACf,IAED,QAAQ;IACX,yBAAyBA,EAAE,OAAO,GAAG,QAAQ;AAC/C;AAGO,MAAMI,oBAAoBJ,EAAE,KAAK,CAAC;IACvCA,EAAE,MAAM;IACRA,EAAAA,MACS,CAAC;QACN,QAAQA,EAAE,MAAM;IAClB,GACC,GAAG,CAACG,wBAAwB,OAAO;CACvC;AAMD,MAAME,sBAAsB;AAG5B,MAAMC,wBAAwBN,EAAAA,MACrB,CAAC;IACN,QAAQI;IACR,WAAWJ,EAAE,OAAO,GAAG,QAAQ;IAC/B,WAAWA,EAAE,OAAO,GAAG,QAAQ;IAC/B,OAAOA,EAAE,KAAK,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,OAAO;KAAG,EAAE,QAAQ;AACpD,GACC,WAAW;AAGiBA,EAAAA,MACtB,CAAC;IACN,CAACK,oBAAoB,EAAEL,EAAE,OAAO,CAAC;IACjC,QAAQI;IAGR,WAAWJ,EAAE,OAAO,GAAG,QAAQ;IAC/B,WAAWA,EAAE,OAAO,GAAG,QAAQ;IAC/B,OAAOA,EAAE,OAAO,GAAG,QAAQ;IAG3B,QAAQA,EAAE,KAAK,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,MAAM;KAAG;IACxC,MAAME;AACR,GACC,WAAW;AAYP,MAAMK,4BAA4B,IAChCD;AAGF,MAAME,yBAAyB,CAACC;IAErC,IAAIC,cAAcD;IAClB,IAAIC,YAAY,IAAI,EAAE,aAAa,eACjCA,cAAcA,YAAY,IAAI,CAAC,SAAS;IAI1C,IAAIA,YAAY,IAAI,EAAE,aAAa,aAAa;QAC9C,MAAMC,QAAQD,YAAY,IAAI,CAAC,KAAK;QAGpC,IAAIL,uBAAuBM,OACzB,OAAO;QAKT,IAAI,YAAYA,SAASA,MAAM,MAAM,EACnC,OAAO;IAEX;IAEA,OAAO;AACT;AAEO,MAAMC,2BAA2B,CAACH;IACvC3E,OACE0E,uBAAuBC,QACvB;IAIF,IAAI,AAAiB,YAAjB,OAAOA,OACT,OAAOA;IAIT,IAAIA,SAAS,AAAiB,YAAjB,OAAOA,SAAsBA,MAAM,MAAM,EAAE;QAEtD,IAAI,AAAwB,YAAxB,OAAOA,MAAM,MAAM,EACrB,OAAOA,MAAM,MAAM;QAGrB,IAAI,AAAwB,YAAxB,OAAOA,MAAM,MAAM,IAAiBA,MAAM,MAAM,CAAC,MAAM,EACzD,OAAOA,MAAM,MAAM,CAAC,MAAM;IAE9B;IAGA,OAAOI,OAAOJ;AAChB;AAEO,MAAMK,8BAA8B,CACzCC,SACAC;IAEA,IAAI,CAACD,SACH,OAAO,EAAE;IAIX,MAAME,YAAYF;IAClB,IAAIE,UAAU,IAAI,EAAE,aAAa,eAAeA,UAAU,KAAK,EAAE;QAC/D,MAAMC,OAAOC,OAAO,IAAI,CAACF,UAAU,KAAK;QACxC,OAAOC,KAAK,MAAM,CAAC,CAACE;YAClB,MAAMX,QAAQQ,UAAU,KAAK,CAACG,IAAI;YAClC,IAAI,CAACZ,uBAAuBC,QAC1B,OAAO;YAIT,IAAIO,cACF,OAAOP,MAAM,IAAI,EAAE,aAAa;YAGlC,OAAO;QACT;IACF;IAGA,OAAO,EAAE;AACX;AAEO,MAAMZ,kBAAkB,CAC7BwB,YACAC;IAEA,MAAMC,gBAAgBT,4BAA4BQ;IAClD,MAAM3F,SAAS;QAAE,GAAG0F,UAAU;IAAC;IAE/B,KAAK,MAAMG,aAAaD,cAAe;QACrC,MAAME,aAAa9F,MAAM,CAAC6F,UAAU;QACpC,IAAIC,YAEF;YAAA,IAAI,AAAsB,YAAtB,OAAOA,YACT9F,MAAM,CAAC6F,UAAU,GAAGC;iBACf,IAAI,AAAsB,YAAtB,OAAOA,YAEhB;gBAAA,IAAIA,WAAW,MAAM,EAEnB;oBAAA,IAAI,AAA6B,YAA7B,OAAOA,WAAW,MAAM,EAC1B9F,MAAM,CAAC6F,UAAU,GAAGC,WAAW,MAAM;yBAChC,IACL,AAA6B,YAA7B,OAAOA,WAAW,MAAM,IACxBA,WAAW,MAAM,CAAC,MAAM,EAGxB9F,MAAM,CAAC6F,UAAU,GAAGC,WAAW,MAAM,CAAC,MAAM;gBAC9C;YACF;QACF;IAEJ;IAEA,OAAO9F;AACT;AAEO,MAAM+F,kBAAkB,CAC7BL,YACAC;IAEA,MAAMC,gBAAgBT,4BAA4BQ;IAClD,MAAM3F,SAAS;QAAE,GAAG0F,UAAU;IAAC;IAE/B,KAAK,MAAMG,aAAaD,cAAe;QACrC,MAAME,aAAa9F,MAAM,CAAC6F,UAAU;QACpC,IAAIC,cAAc,AAAsB,YAAtB,OAAOA,YACvB9F,MAAM,CAAC6F,UAAU,GAAG;YAClB,CAACnB,oBAAoB,EAAE;YACvB,QAAQoB;QACV;IAEJ;IAEA,OAAO9F;AACT;AAUO,MAAMgG,mBAAmB,CAC9BC,UACAN;IAGA,IAAI,CAACA,WACH;IAIF,MAAMO,QAAQD,YAAY,CAAC;IAG3B,MAAME,eAAehB,4BAA4BQ;IAGjD,IAAIQ,AAAwB,MAAxBA,aAAa,MAAM,EACrB,OAAOR,UAAU,KAAK,CAACO;IAIzB,MAAME,oBAAyC,CAAC;IAChD,KAAK,MAAMP,aAAaM,aACtB,IAAIN,aAAaK,OACfE,iBAAiB,CAACP,UAAU,GAAGK,KAAK,CAACL,UAAU;IAKnD,MAAMQ,sBAA2C,CAAC;IAClD,IAAK,MAAMZ,OAAOS,MAChB,IAAIC,aAAa,QAAQ,CAACV,MAExBY,mBAAmB,CAACZ,IAAI,GAAG;QAAE,QAAQ;IAAU;SAE/CY,mBAAmB,CAACZ,IAAI,GAAGS,KAAK,CAACT,IAAI;IAKzC,MAAMa,YAAYX,UAAU,KAAK,CAACU;IAGlC,IAAK,MAAMR,aAAaO,kBACtBE,SAAS,CAACT,UAAU,GAAGO,iBAAiB,CAACP,UAAU;IAGrD,OAAOS;AACT"}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"device/index.mjs","sources":["webpack://@midscene/core/./src/device/index.ts"],"sourcesContent":["import { getMidsceneLocationSchema } from '@/common';\nimport type {\n ActionScrollParam,\n DeviceAction,\n LocateResultElement,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport type { ElementNode } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { z } from 'zod';\nimport type { ElementCacheFeature, Rect, Size, UIContext } from '../types';\n\nexport abstract class AbstractInterface {\n abstract interfaceType: string;\n\n abstract screenshotBase64(): Promise<string>;\n abstract size(): Promise<Size>;\n abstract actionSpace(): DeviceAction[] | Promise<DeviceAction[]>;\n\n abstract cacheFeatureForRect?(\n rect: Rect,\n options?: {\n targetDescription?: string;\n modelConfig?: IModelConfig;\n },\n ): Promise<ElementCacheFeature>;\n abstract rectMatchesCacheFeature?(\n feature: ElementCacheFeature,\n ): Promise<Rect>;\n\n abstract destroy?(): Promise<void>;\n\n abstract describe?(): string;\n abstract beforeInvokeAction?(actionName: string, param: any): Promise<void>;\n abstract afterInvokeAction?(actionName: string, param: any): Promise<void>;\n\n // @deprecated do NOT extend this method\n abstract getElementsNodeTree?: () => Promise<ElementNode>;\n\n // @deprecated do NOT extend this method\n abstract url?: () => string | Promise<string>;\n\n // @deprecated do NOT extend this method\n abstract evaluateJavaScript?<T = any>(script: string): Promise<T>;\n\n // @deprecated do NOT extend this method\n abstract getContext?(): Promise<UIContext>;\n}\n\n// Generic function to define actions with proper type inference\n// TRuntime allows specifying a different type for the runtime parameter (after location resolution)\n// TReturn allows specifying the return type of the action\nexport const defineAction = <\n TSchema extends z.ZodType | undefined = undefined,\n TRuntime = TSchema extends z.ZodType ? z.infer<TSchema> : undefined,\n TReturn = any,\n>(\n config: {\n name: string;\n description: string;\n interfaceAlias?: string;\n paramSchema?: TSchema;\n call: (param: TRuntime) => Promise<TReturn> | TReturn;\n } & Partial<\n Omit<\n DeviceAction<TRuntime, TReturn>,\n 'name' | 'description' | 'interfaceAlias' | 'paramSchema' | 'call'\n >\n >,\n): DeviceAction<TRuntime, TReturn> => {\n return config as any; // Type assertion needed because schema validation type differs from runtime type\n};\n\n// Tap\nexport const actionTapParamSchema = z.object({\n locate: getMidsceneLocationSchema().describe('The element to be tapped'),\n});\n// Override the inferred type to use LocateResultElement for the runtime locate field\nexport type ActionTapParam = {\n locate: LocateResultElement;\n};\n\nexport const defineActionTap = (\n call: (param: ActionTapParam) => Promise<void>,\n): DeviceAction<ActionTapParam> => {\n return defineAction<typeof actionTapParamSchema, ActionTapParam>({\n name: 'Tap',\n description: 'Tap the element',\n interfaceAlias: 'aiTap',\n paramSchema: actionTapParamSchema,\n call,\n });\n};\n\n// RightClick\nexport const actionRightClickParamSchema = z.object({\n locate: getMidsceneLocationSchema().describe(\n 'The element to be right clicked',\n ),\n});\nexport type ActionRightClickParam = {\n locate: LocateResultElement;\n};\n\nexport const defineActionRightClick = (\n call: (param: ActionRightClickParam) => Promise<void>,\n): DeviceAction<ActionRightClickParam> => {\n return defineAction<\n typeof actionRightClickParamSchema,\n ActionRightClickParam\n >({\n name: 'RightClick',\n description: 'Right click the element',\n interfaceAlias: 'aiRightClick',\n paramSchema: actionRightClickParamSchema,\n call,\n });\n};\n\n// DoubleClick\nexport const actionDoubleClickParamSchema = z.object({\n locate: getMidsceneLocationSchema().describe(\n 'The element to be double clicked',\n ),\n});\nexport type ActionDoubleClickParam = {\n locate: LocateResultElement;\n};\n\nexport const defineActionDoubleClick = (\n call: (param: ActionDoubleClickParam) => Promise<void>,\n): DeviceAction<ActionDoubleClickParam> => {\n return defineAction<\n typeof actionDoubleClickParamSchema,\n ActionDoubleClickParam\n >({\n name: 'DoubleClick',\n description: 'Double click the element',\n interfaceAlias: 'aiDoubleClick',\n paramSchema: actionDoubleClickParamSchema,\n call,\n });\n};\n\n// Hover\nexport const actionHoverParamSchema = z.object({\n locate: getMidsceneLocationSchema().describe('The element to be hovered'),\n});\nexport type ActionHoverParam = {\n locate: LocateResultElement;\n};\n\nexport const defineActionHover = (\n call: (param: ActionHoverParam) => Promise<void>,\n): DeviceAction<ActionHoverParam> => {\n return defineAction<typeof actionHoverParamSchema, ActionHoverParam>({\n name: 'Hover',\n description: 'Move the mouse to the element',\n interfaceAlias: 'aiHover',\n paramSchema: actionHoverParamSchema,\n call,\n });\n};\n\n// Input\nconst inputLocateDescription =\n 'the position of the placeholder or text content in the target input field. If there is no content, locate the center of the input field.';\nexport const actionInputParamSchema = z.object({\n value: z\n .union([z.string(), z.number()])\n .transform((val) => String(val))\n .describe(\n 'The text to input. Provide the final content for replace/append modes, or an empty string when using clear mode to remove existing text.',\n ),\n locate: getMidsceneLocationSchema()\n .describe(inputLocateDescription)\n .optional(),\n mode: z\n .enum(['replace', 'clear', 'append'])\n .default('replace')\n .optional()\n .describe(\n 'Input mode: \"replace\" (default) - clear the field and input the value; \"append\" - append the value to existing content; \"clear\" - clear the field without inputting new text.',\n ),\n});\nexport type ActionInputParam = {\n value: string;\n locate?: LocateResultElement;\n mode?: 'replace' | 'clear' | 'append';\n};\n\nexport const defineActionInput = (\n call: (param: ActionInputParam) => Promise<void>,\n): DeviceAction<ActionInputParam> => {\n return defineAction<typeof actionInputParamSchema, ActionInputParam>({\n name: 'Input',\n description: 'Input the value into the element',\n interfaceAlias: 'aiInput',\n paramSchema: actionInputParamSchema,\n call,\n });\n};\n\n// KeyboardPress\nexport const actionKeyboardPressParamSchema = z.object({\n locate: getMidsceneLocationSchema()\n .describe('The element to be clicked before pressing the key')\n .optional(),\n keyName: z\n .string()\n .describe(\n \"The key to be pressed. Use '+' for key combinations, e.g., 'Control+A', 'Shift+Enter'\",\n ),\n});\nexport type ActionKeyboardPressParam = {\n locate?: LocateResultElement;\n keyName: string;\n};\n\nexport const defineActionKeyboardPress = (\n call: (param: ActionKeyboardPressParam) => Promise<void>,\n): DeviceAction<ActionKeyboardPressParam> => {\n return defineAction<\n typeof actionKeyboardPressParamSchema,\n ActionKeyboardPressParam\n >({\n name: 'KeyboardPress',\n description:\n 'Press a key or key combination, like \"Enter\", \"Tab\", \"Escape\", or \"Control+A\", \"Shift+Enter\". Do not use this to type text.',\n interfaceAlias: 'aiKeyboardPress',\n paramSchema: actionKeyboardPressParamSchema,\n call,\n });\n};\n\n// Scroll\nexport const actionScrollParamSchema = z.object({\n direction: z\n .enum(['down', 'up', 'right', 'left'])\n .default('down')\n .describe('The direction to scroll'),\n scrollType: z\n .enum([\n 'singleAction',\n 'scrollToBottom',\n 'scrollToTop',\n 'scrollToRight',\n 'scrollToLeft',\n ])\n .default('singleAction')\n .describe(\n 'The scroll behavior: \"singleAction\" for a single scroll action, \"scrollToBottom\" for scrolling to the bottom, \"scrollToTop\" for scrolling to the top, \"scrollToRight\" for scrolling to the right, \"scrollToLeft\" for scrolling to the left',\n ),\n distance: z\n .number()\n .nullable()\n .optional()\n .describe('The distance in pixels to scroll'),\n locate: getMidsceneLocationSchema()\n .optional()\n .describe('The target element to be scrolled'),\n});\n\nexport const defineActionScroll = (\n call: (param: ActionScrollParam) => Promise<void>,\n): DeviceAction<ActionScrollParam> => {\n return defineAction<typeof actionScrollParamSchema, ActionScrollParam>({\n name: 'Scroll',\n description:\n 'Scroll the page or an element. The direction to scroll, the scroll type, and the distance to scroll. The distance is the number of pixels to scroll. If not specified, use `down` direction, `once` scroll type, and `null` distance.',\n interfaceAlias: 'aiScroll',\n paramSchema: actionScrollParamSchema,\n call,\n });\n};\n\n// DragAndDrop\nexport const actionDragAndDropParamSchema = z.object({\n from: getMidsceneLocationSchema().describe('The position to be dragged'),\n to: getMidsceneLocationSchema().describe('The position to be dropped'),\n});\nexport type ActionDragAndDropParam = {\n from: LocateResultElement;\n to: LocateResultElement;\n};\n\nexport const defineActionDragAndDrop = (\n call: (param: ActionDragAndDropParam) => Promise<void>,\n): DeviceAction<ActionDragAndDropParam> => {\n return defineAction<\n typeof actionDragAndDropParamSchema,\n ActionDragAndDropParam\n >({\n name: 'DragAndDrop',\n description: 'Drag and drop the element',\n interfaceAlias: 'aiDragAndDrop',\n paramSchema: actionDragAndDropParamSchema,\n call,\n });\n};\n\nexport const ActionLongPressParamSchema = z.object({\n locate: getMidsceneLocationSchema().describe(\n 'The element to be long pressed',\n ),\n duration: z\n .number()\n .default(500)\n .optional()\n .describe('Long press duration in milliseconds'),\n});\n\nexport type ActionLongPressParam = {\n locate: LocateResultElement;\n duration?: number;\n};\nexport const defineActionLongPress = (\n call: (param: ActionLongPressParam) => Promise<void>,\n): DeviceAction<ActionLongPressParam> => {\n return defineAction<typeof ActionLongPressParamSchema, ActionLongPressParam>({\n name: 'LongPress',\n description: 'Long press the element',\n paramSchema: ActionLongPressParamSchema,\n call,\n });\n};\n\nexport const ActionSwipeParamSchema = z.object({\n start: getMidsceneLocationSchema()\n .optional()\n .describe(\n 'Starting point of the swipe gesture, if not specified, the center of the page will be used',\n ),\n direction: z\n .enum(['up', 'down', 'left', 'right'])\n .optional()\n .describe(\n 'The direction to swipe (required when using distance). The direction means the direction of the finger swipe.',\n ),\n distance: z\n .number()\n .optional()\n .describe('The distance in pixels to swipe (mutually exclusive with end)'),\n end: getMidsceneLocationSchema()\n .optional()\n .describe(\n 'Ending point of the swipe gesture (mutually exclusive with distance)',\n ),\n duration: z\n .number()\n .default(300)\n .describe('Duration of the swipe gesture in milliseconds'),\n repeat: z\n .number()\n .optional()\n .describe(\n 'The number of times to repeat the swipe gesture. 1 for default, 0 for infinite (e.g. endless swipe until the end of the page)',\n ),\n});\n\nexport type ActionSwipeParam = {\n start?: LocateResultElement;\n direction?: 'up' | 'down' | 'left' | 'right';\n distance?: number;\n end?: LocateResultElement;\n duration?: number;\n repeat?: number;\n};\n\nexport const defineActionSwipe = (\n call: (param: ActionSwipeParam) => Promise<void>,\n): DeviceAction<ActionSwipeParam> => {\n return defineAction<typeof ActionSwipeParamSchema, ActionSwipeParam>({\n name: 'Swipe',\n description:\n 'Perform a swipe gesture. You must specify either \"end\" (target location) or \"distance\" + \"direction\" - they are mutually exclusive. Use \"end\" for precise location-based swipes, or \"distance\" + \"direction\" for relative movement.',\n paramSchema: ActionSwipeParamSchema,\n call,\n });\n};\n\n// ClearInput\nexport const actionClearInputParamSchema = z.object({\n locate: getMidsceneLocationSchema().describe('The input field to be cleared'),\n});\nexport type ActionClearInputParam = {\n locate: LocateResultElement;\n};\n\nexport const defineActionClearInput = (\n call: (param: ActionClearInputParam) => Promise<void>,\n): DeviceAction<ActionClearInputParam> => {\n return defineAction<\n typeof actionClearInputParamSchema,\n ActionClearInputParam\n >({\n name: 'ClearInput',\n description: inputLocateDescription,\n interfaceAlias: 'aiClearInput',\n paramSchema: actionClearInputParamSchema,\n call,\n });\n};\n\n// Assert\nexport const actionAssertParamSchema = z.object({\n thought: z\n .string()\n .describe('The thinking process of analyzing the assertion'),\n result: z\n .boolean()\n .describe('Whether the assertion is truthy, return true or false'),\n});\nexport type ActionAssertParam = {\n thought: string;\n result: boolean;\n};\n\nexport const defineActionAssert = (\n call: (param: ActionAssertParam) => Promise<void>,\n): DeviceAction<ActionAssertParam> => {\n return defineAction<typeof actionAssertParamSchema, ActionAssertParam>({\n name: 'Assert',\n description:\n 'If the user explicitly requires making an assertion (like \"there should be a button with text \"Yes\" in the popup\"), think about it, provide the reasoning and result here.',\n interfaceAlias: 'aiAssert',\n paramSchema: actionAssertParamSchema,\n call,\n });\n};\n\n/**\n * Creates an Assert action with default implementation.\n * This action can be used across all interfaces without modification.\n * If pass=true, the assertion succeeds silently.\n * If pass=false, the assertion fails and throws an error with the thought as the error message.\n */\nexport const createAssertAction = (): DeviceAction<ActionAssertParam> => {\n const debug = getDebug('core:planning-assert');\n return defineActionAssert(async (param: ActionAssertParam) => {\n debug('assertion', param.thought, param.result);\n if (!param.result) {\n throw new Error(`Assertion failed: ${param.thought}`);\n }\n // If pass is true, do nothing (assertion succeeds)\n });\n};\n\nexport type { DeviceAction } from '../types';\nexport type {\n AndroidDeviceOpt,\n AndroidDeviceInputOpt,\n IOSDeviceOpt,\n IOSDeviceInputOpt,\n} from './device-options';\n"],"names":["AbstractInterface","defineAction","config","actionTapParamSchema","z","getMidsceneLocationSchema","defineActionTap","call","actionRightClickParamSchema","defineActionRightClick","actionDoubleClickParamSchema","defineActionDoubleClick","actionHoverParamSchema","defineActionHover","inputLocateDescription","actionInputParamSchema","val","String","defineActionInput","actionKeyboardPressParamSchema","defineActionKeyboardPress","actionScrollParamSchema","defineActionScroll","actionDragAndDropParamSchema","defineActionDragAndDrop","ActionLongPressParamSchema","defineActionLongPress","ActionSwipeParamSchema","defineActionSwipe","actionClearInputParamSchema","defineActionClearInput","actionAssertParamSchema","defineActionAssert","createAssertAction","debug","getDebug","param","Error"],"mappings":";;;AAaO,MAAeA;AAmCtB;AAKO,MAAMC,eAAe,CAK1BC,SAaOA;AAIF,MAAMC,uBAAuBC,EAAE,MAAM,CAAC;IAC3C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAMO,MAAMC,kBAAkB,CAC7BC,OAEON,aAA0D;QAC/D,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaE;QACbI;IACF;AAIK,MAAMC,8BAA8BJ,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMI,yBAAyB,CACpCF,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaO;QACbD;IACF;AAIK,MAAMG,+BAA+BN,EAAE,MAAM,CAAC;IACnD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMM,0BAA0B,CACrCJ,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaS;QACbH;IACF;AAIK,MAAMK,yBAAyBR,EAAE,MAAM,CAAC;IAC7C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMQ,oBAAoB,CAC/BN,OAEON,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaW;QACbL;IACF;AAIF,MAAMO,yBACJ;AACK,MAAMC,yBAAyBX,EAAE,MAAM,CAAC;IAC7C,OAAOA,EAAAA,KACC,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,MAAM;KAAG,EAC9B,SAAS,CAAC,CAACY,MAAQC,OAAOD,MAC1B,QAAQ,CACP;IAEJ,QAAQX,4BACL,QAAQ,CAACS,wBACT,QAAQ;IACX,MAAMV,CAAC,CAADA,OACC,CAAC;QAAC;QAAW;QAAS;KAAS,EACnC,OAAO,CAAC,WACR,QAAQ,GACR,QAAQ,CACP;AAEN;AAOO,MAAMc,oBAAoB,CAC/BX,OAEON,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAac;QACbR;IACF;AAIK,MAAMY,iCAAiCf,EAAE,MAAM,CAAC;IACrD,QAAQC,4BACL,QAAQ,CAAC,qDACT,QAAQ;IACX,SAASD,EAAAA,MACA,GACN,QAAQ,CACP;AAEN;AAMO,MAAMgB,4BAA4B,CACvCb,OAEON,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAakB;QACbZ;IACF;AAIK,MAAMc,0BAA0BjB,EAAE,MAAM,CAAC;IAC9C,WAAWA,CAAC,CAADA,OACJ,CAAC;QAAC;QAAQ;QAAM;QAAS;KAAO,EACpC,OAAO,CAAC,QACR,QAAQ,CAAC;IACZ,YAAYA,CAAC,CAADA,OACL,CAAC;QACJ;QACA;QACA;QACA;QACA;KACD,EACA,OAAO,CAAC,gBACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,CAAC;IACZ,QAAQC,4BACL,QAAQ,GACR,QAAQ,CAAC;AACd;AAEO,MAAMiB,qBAAqB,CAChCf,OAEON,aAAgE;QACrE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAaoB;QACbd;IACF;AAIK,MAAMgB,+BAA+BnB,EAAE,MAAM,CAAC;IACnD,MAAMC,4BAA4B,QAAQ,CAAC;IAC3C,IAAIA,4BAA4B,QAAQ,CAAC;AAC3C;AAMO,MAAMmB,0BAA0B,CACrCjB,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAasB;QACbhB;IACF;AAGK,MAAMkB,6BAA6BrB,EAAE,MAAM,CAAC;IACjD,QAAQC,4BAA4B,QAAQ,CAC1C;IAEF,UAAUD,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,GACR,QAAQ,CAAC;AACd;AAMO,MAAMsB,wBAAwB,CACnCnB,OAEON,aAAsE;QAC3E,MAAM;QACN,aAAa;QACb,aAAawB;QACblB;IACF;AAGK,MAAMoB,yBAAyBvB,EAAE,MAAM,CAAC;IAC7C,OAAOC,4BACJ,QAAQ,GACR,QAAQ,CACP;IAEJ,WAAWD,CAAC,CAADA,OACJ,CAAC;QAAC;QAAM;QAAQ;QAAQ;KAAQ,EACpC,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,CAAC;IACZ,KAAKC,4BACF,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUD,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,CAAC;IACZ,QAAQA,EAAAA,MACC,GACN,QAAQ,GACR,QAAQ,CACP;AAEN;AAWO,MAAMwB,oBAAoB,CAC/BrB,OAEON,aAA8D;QACnE,MAAM;QACN,aACE;QACF,aAAa0B;QACbpB;IACF;AAIK,MAAMsB,8BAA8BzB,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMyB,yBAAyB,CACpCvB,OAEON,aAGL;QACA,MAAM;QACN,aAAaa;QACb,gBAAgB;QAChB,aAAae;QACbtB;IACF;AAIK,MAAMwB,0BAA0B3B,EAAE,MAAM,CAAC;IAC9C,SAASA,EAAAA,MACA,GACN,QAAQ,CAAC;IACZ,QAAQA,EAAAA,OACE,GACP,QAAQ,CAAC;AACd;AAMO,MAAM4B,qBAAqB,CAChCzB,OAEON,aAAgE;QACrE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAa8B;QACbxB;IACF;AASK,MAAM0B,qBAAqB;IAChC,MAAMC,QAAQC,SAAS;IACvB,OAAOH,mBAAmB,OAAOI;QAC/BF,MAAM,aAAaE,MAAM,OAAO,EAAEA,MAAM,MAAM;QAC9C,IAAI,CAACA,MAAM,MAAM,EACf,MAAM,IAAIC,MAAM,CAAC,kBAAkB,EAAED,MAAM,OAAO,EAAE;IAGxD;AACF"}
|
|
1
|
+
{"version":3,"file":"device/index.mjs","sources":["../../../src/device/index.ts"],"sourcesContent":["import { getMidsceneLocationSchema } from '@/common';\nimport type {\n ActionScrollParam,\n DeviceAction,\n LocateResultElement,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport type { ElementNode } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { z } from 'zod';\nimport type { ElementCacheFeature, Rect, Size, UIContext } from '../types';\n\nexport abstract class AbstractInterface {\n abstract interfaceType: string;\n\n abstract screenshotBase64(): Promise<string>;\n abstract size(): Promise<Size>;\n abstract actionSpace(): DeviceAction[] | Promise<DeviceAction[]>;\n\n abstract cacheFeatureForRect?(\n rect: Rect,\n options?: {\n targetDescription?: string;\n modelConfig?: IModelConfig;\n },\n ): Promise<ElementCacheFeature>;\n abstract rectMatchesCacheFeature?(\n feature: ElementCacheFeature,\n ): Promise<Rect>;\n\n abstract destroy?(): Promise<void>;\n\n abstract describe?(): string;\n abstract beforeInvokeAction?(actionName: string, param: any): Promise<void>;\n abstract afterInvokeAction?(actionName: string, param: any): Promise<void>;\n\n // @deprecated do NOT extend this method\n abstract getElementsNodeTree?: () => Promise<ElementNode>;\n\n // @deprecated do NOT extend this method\n abstract url?: () => string | Promise<string>;\n\n // @deprecated do NOT extend this method\n abstract evaluateJavaScript?<T = any>(script: string): Promise<T>;\n\n // @deprecated do NOT extend this method\n abstract getContext?(): Promise<UIContext>;\n}\n\n// Generic function to define actions with proper type inference\n// TRuntime allows specifying a different type for the runtime parameter (after location resolution)\n// TReturn allows specifying the return type of the action\nexport const defineAction = <\n TSchema extends z.ZodType | undefined = undefined,\n TRuntime = TSchema extends z.ZodType ? z.infer<TSchema> : undefined,\n TReturn = any,\n>(\n config: {\n name: string;\n description: string;\n interfaceAlias?: string;\n paramSchema?: TSchema;\n call: (param: TRuntime) => Promise<TReturn> | TReturn;\n } & Partial<\n Omit<\n DeviceAction<TRuntime, TReturn>,\n 'name' | 'description' | 'interfaceAlias' | 'paramSchema' | 'call'\n >\n >,\n): DeviceAction<TRuntime, TReturn> => {\n return config as any; // Type assertion needed because schema validation type differs from runtime type\n};\n\n// Tap\nexport const actionTapParamSchema = z.object({\n locate: getMidsceneLocationSchema().describe('The element to be tapped'),\n});\n// Override the inferred type to use LocateResultElement for the runtime locate field\nexport type ActionTapParam = {\n locate: LocateResultElement;\n};\n\nexport const defineActionTap = (\n call: (param: ActionTapParam) => Promise<void>,\n): DeviceAction<ActionTapParam> => {\n return defineAction<typeof actionTapParamSchema, ActionTapParam>({\n name: 'Tap',\n description: 'Tap the element',\n interfaceAlias: 'aiTap',\n paramSchema: actionTapParamSchema,\n call,\n });\n};\n\n// RightClick\nexport const actionRightClickParamSchema = z.object({\n locate: getMidsceneLocationSchema().describe(\n 'The element to be right clicked',\n ),\n});\nexport type ActionRightClickParam = {\n locate: LocateResultElement;\n};\n\nexport const defineActionRightClick = (\n call: (param: ActionRightClickParam) => Promise<void>,\n): DeviceAction<ActionRightClickParam> => {\n return defineAction<\n typeof actionRightClickParamSchema,\n ActionRightClickParam\n >({\n name: 'RightClick',\n description: 'Right click the element',\n interfaceAlias: 'aiRightClick',\n paramSchema: actionRightClickParamSchema,\n call,\n });\n};\n\n// DoubleClick\nexport const actionDoubleClickParamSchema = z.object({\n locate: getMidsceneLocationSchema().describe(\n 'The element to be double clicked',\n ),\n});\nexport type ActionDoubleClickParam = {\n locate: LocateResultElement;\n};\n\nexport const defineActionDoubleClick = (\n call: (param: ActionDoubleClickParam) => Promise<void>,\n): DeviceAction<ActionDoubleClickParam> => {\n return defineAction<\n typeof actionDoubleClickParamSchema,\n ActionDoubleClickParam\n >({\n name: 'DoubleClick',\n description: 'Double click the element',\n interfaceAlias: 'aiDoubleClick',\n paramSchema: actionDoubleClickParamSchema,\n call,\n });\n};\n\n// Hover\nexport const actionHoverParamSchema = z.object({\n locate: getMidsceneLocationSchema().describe('The element to be hovered'),\n});\nexport type ActionHoverParam = {\n locate: LocateResultElement;\n};\n\nexport const defineActionHover = (\n call: (param: ActionHoverParam) => Promise<void>,\n): DeviceAction<ActionHoverParam> => {\n return defineAction<typeof actionHoverParamSchema, ActionHoverParam>({\n name: 'Hover',\n description: 'Move the mouse to the element',\n interfaceAlias: 'aiHover',\n paramSchema: actionHoverParamSchema,\n call,\n });\n};\n\n// Input\nconst inputLocateDescription =\n 'the position of the placeholder or text content in the target input field. If there is no content, locate the center of the input field.';\nexport const actionInputParamSchema = z.object({\n value: z\n .union([z.string(), z.number()])\n .transform((val) => String(val))\n .describe(\n 'The text to input. Provide the final content for replace/append modes, or an empty string when using clear mode to remove existing text.',\n ),\n locate: getMidsceneLocationSchema()\n .describe(inputLocateDescription)\n .optional(),\n mode: z\n .enum(['replace', 'clear', 'append'])\n .default('replace')\n .optional()\n .describe(\n 'Input mode: \"replace\" (default) - clear the field and input the value; \"append\" - append the value to existing content; \"clear\" - clear the field without inputting new text.',\n ),\n});\nexport type ActionInputParam = {\n value: string;\n locate?: LocateResultElement;\n mode?: 'replace' | 'clear' | 'append';\n};\n\nexport const defineActionInput = (\n call: (param: ActionInputParam) => Promise<void>,\n): DeviceAction<ActionInputParam> => {\n return defineAction<typeof actionInputParamSchema, ActionInputParam>({\n name: 'Input',\n description: 'Input the value into the element',\n interfaceAlias: 'aiInput',\n paramSchema: actionInputParamSchema,\n call,\n });\n};\n\n// KeyboardPress\nexport const actionKeyboardPressParamSchema = z.object({\n locate: getMidsceneLocationSchema()\n .describe('The element to be clicked before pressing the key')\n .optional(),\n keyName: z\n .string()\n .describe(\n \"The key to be pressed. Use '+' for key combinations, e.g., 'Control+A', 'Shift+Enter'\",\n ),\n});\nexport type ActionKeyboardPressParam = {\n locate?: LocateResultElement;\n keyName: string;\n};\n\nexport const defineActionKeyboardPress = (\n call: (param: ActionKeyboardPressParam) => Promise<void>,\n): DeviceAction<ActionKeyboardPressParam> => {\n return defineAction<\n typeof actionKeyboardPressParamSchema,\n ActionKeyboardPressParam\n >({\n name: 'KeyboardPress',\n description:\n 'Press a key or key combination, like \"Enter\", \"Tab\", \"Escape\", or \"Control+A\", \"Shift+Enter\". Do not use this to type text.',\n interfaceAlias: 'aiKeyboardPress',\n paramSchema: actionKeyboardPressParamSchema,\n call,\n });\n};\n\n// Scroll\nexport const actionScrollParamSchema = z.object({\n direction: z\n .enum(['down', 'up', 'right', 'left'])\n .default('down')\n .describe('The direction to scroll'),\n scrollType: z\n .enum([\n 'singleAction',\n 'scrollToBottom',\n 'scrollToTop',\n 'scrollToRight',\n 'scrollToLeft',\n ])\n .default('singleAction')\n .describe(\n 'The scroll behavior: \"singleAction\" for a single scroll action, \"scrollToBottom\" for scrolling to the bottom, \"scrollToTop\" for scrolling to the top, \"scrollToRight\" for scrolling to the right, \"scrollToLeft\" for scrolling to the left',\n ),\n distance: z\n .number()\n .nullable()\n .optional()\n .describe('The distance in pixels to scroll'),\n locate: getMidsceneLocationSchema()\n .optional()\n .describe('The target element to be scrolled'),\n});\n\nexport const defineActionScroll = (\n call: (param: ActionScrollParam) => Promise<void>,\n): DeviceAction<ActionScrollParam> => {\n return defineAction<typeof actionScrollParamSchema, ActionScrollParam>({\n name: 'Scroll',\n description:\n 'Scroll the page or an element. The direction to scroll, the scroll type, and the distance to scroll. The distance is the number of pixels to scroll. If not specified, use `down` direction, `once` scroll type, and `null` distance.',\n interfaceAlias: 'aiScroll',\n paramSchema: actionScrollParamSchema,\n call,\n });\n};\n\n// DragAndDrop\nexport const actionDragAndDropParamSchema = z.object({\n from: getMidsceneLocationSchema().describe('The position to be dragged'),\n to: getMidsceneLocationSchema().describe('The position to be dropped'),\n});\nexport type ActionDragAndDropParam = {\n from: LocateResultElement;\n to: LocateResultElement;\n};\n\nexport const defineActionDragAndDrop = (\n call: (param: ActionDragAndDropParam) => Promise<void>,\n): DeviceAction<ActionDragAndDropParam> => {\n return defineAction<\n typeof actionDragAndDropParamSchema,\n ActionDragAndDropParam\n >({\n name: 'DragAndDrop',\n description: 'Drag and drop the element',\n interfaceAlias: 'aiDragAndDrop',\n paramSchema: actionDragAndDropParamSchema,\n call,\n });\n};\n\nexport const ActionLongPressParamSchema = z.object({\n locate: getMidsceneLocationSchema().describe(\n 'The element to be long pressed',\n ),\n duration: z\n .number()\n .default(500)\n .optional()\n .describe('Long press duration in milliseconds'),\n});\n\nexport type ActionLongPressParam = {\n locate: LocateResultElement;\n duration?: number;\n};\nexport const defineActionLongPress = (\n call: (param: ActionLongPressParam) => Promise<void>,\n): DeviceAction<ActionLongPressParam> => {\n return defineAction<typeof ActionLongPressParamSchema, ActionLongPressParam>({\n name: 'LongPress',\n description: 'Long press the element',\n paramSchema: ActionLongPressParamSchema,\n call,\n });\n};\n\nexport const ActionSwipeParamSchema = z.object({\n start: getMidsceneLocationSchema()\n .optional()\n .describe(\n 'Starting point of the swipe gesture, if not specified, the center of the page will be used',\n ),\n direction: z\n .enum(['up', 'down', 'left', 'right'])\n .optional()\n .describe(\n 'The direction to swipe (required when using distance). The direction means the direction of the finger swipe.',\n ),\n distance: z\n .number()\n .optional()\n .describe('The distance in pixels to swipe (mutually exclusive with end)'),\n end: getMidsceneLocationSchema()\n .optional()\n .describe(\n 'Ending point of the swipe gesture (mutually exclusive with distance)',\n ),\n duration: z\n .number()\n .default(300)\n .describe('Duration of the swipe gesture in milliseconds'),\n repeat: z\n .number()\n .optional()\n .describe(\n 'The number of times to repeat the swipe gesture. 1 for default, 0 for infinite (e.g. endless swipe until the end of the page)',\n ),\n});\n\nexport type ActionSwipeParam = {\n start?: LocateResultElement;\n direction?: 'up' | 'down' | 'left' | 'right';\n distance?: number;\n end?: LocateResultElement;\n duration?: number;\n repeat?: number;\n};\n\nexport const defineActionSwipe = (\n call: (param: ActionSwipeParam) => Promise<void>,\n): DeviceAction<ActionSwipeParam> => {\n return defineAction<typeof ActionSwipeParamSchema, ActionSwipeParam>({\n name: 'Swipe',\n description:\n 'Perform a swipe gesture. You must specify either \"end\" (target location) or \"distance\" + \"direction\" - they are mutually exclusive. Use \"end\" for precise location-based swipes, or \"distance\" + \"direction\" for relative movement.',\n paramSchema: ActionSwipeParamSchema,\n call,\n });\n};\n\n// ClearInput\nexport const actionClearInputParamSchema = z.object({\n locate: getMidsceneLocationSchema().describe('The input field to be cleared'),\n});\nexport type ActionClearInputParam = {\n locate: LocateResultElement;\n};\n\nexport const defineActionClearInput = (\n call: (param: ActionClearInputParam) => Promise<void>,\n): DeviceAction<ActionClearInputParam> => {\n return defineAction<\n typeof actionClearInputParamSchema,\n ActionClearInputParam\n >({\n name: 'ClearInput',\n description: inputLocateDescription,\n interfaceAlias: 'aiClearInput',\n paramSchema: actionClearInputParamSchema,\n call,\n });\n};\n\n// Assert\nexport const actionAssertParamSchema = z.object({\n thought: z\n .string()\n .describe('The thinking process of analyzing the assertion'),\n result: z\n .boolean()\n .describe('Whether the assertion is truthy, return true or false'),\n});\nexport type ActionAssertParam = {\n thought: string;\n result: boolean;\n};\n\nexport const defineActionAssert = (\n call: (param: ActionAssertParam) => Promise<void>,\n): DeviceAction<ActionAssertParam> => {\n return defineAction<typeof actionAssertParamSchema, ActionAssertParam>({\n name: 'Assert',\n description:\n 'If the user explicitly requires making an assertion (like \"there should be a button with text \"Yes\" in the popup\"), think about it, provide the reasoning and result here.',\n interfaceAlias: 'aiAssert',\n paramSchema: actionAssertParamSchema,\n call,\n });\n};\n\n/**\n * Creates an Assert action with default implementation.\n * This action can be used across all interfaces without modification.\n * If pass=true, the assertion succeeds silently.\n * If pass=false, the assertion fails and throws an error with the thought as the error message.\n */\nexport const createAssertAction = (): DeviceAction<ActionAssertParam> => {\n const debug = getDebug('core:planning-assert');\n return defineActionAssert(async (param: ActionAssertParam) => {\n debug('assertion', param.thought, param.result);\n if (!param.result) {\n throw new Error(`Assertion failed: ${param.thought}`);\n }\n // If pass is true, do nothing (assertion succeeds)\n });\n};\n\nexport type { DeviceAction } from '../types';\nexport type {\n AndroidDeviceOpt,\n AndroidDeviceInputOpt,\n IOSDeviceOpt,\n IOSDeviceInputOpt,\n} from './device-options';\n"],"names":["AbstractInterface","defineAction","config","actionTapParamSchema","z","getMidsceneLocationSchema","defineActionTap","call","actionRightClickParamSchema","defineActionRightClick","actionDoubleClickParamSchema","defineActionDoubleClick","actionHoverParamSchema","defineActionHover","inputLocateDescription","actionInputParamSchema","val","String","defineActionInput","actionKeyboardPressParamSchema","defineActionKeyboardPress","actionScrollParamSchema","defineActionScroll","actionDragAndDropParamSchema","defineActionDragAndDrop","ActionLongPressParamSchema","defineActionLongPress","ActionSwipeParamSchema","defineActionSwipe","actionClearInputParamSchema","defineActionClearInput","actionAssertParamSchema","defineActionAssert","createAssertAction","debug","getDebug","param","Error"],"mappings":";;;AAaO,MAAeA;AAmCtB;AAKO,MAAMC,eAAe,CAK1BC,SAaOA;AAIF,MAAMC,uBAAuBC,EAAE,MAAM,CAAC;IAC3C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAMO,MAAMC,kBAAkB,CAC7BC,OAEON,aAA0D;QAC/D,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaE;QACbI;IACF;AAIK,MAAMC,8BAA8BJ,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMI,yBAAyB,CACpCF,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaO;QACbD;IACF;AAIK,MAAMG,+BAA+BN,EAAE,MAAM,CAAC;IACnD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMM,0BAA0B,CACrCJ,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaS;QACbH;IACF;AAIK,MAAMK,yBAAyBR,EAAE,MAAM,CAAC;IAC7C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMQ,oBAAoB,CAC/BN,OAEON,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaW;QACbL;IACF;AAIF,MAAMO,yBACJ;AACK,MAAMC,yBAAyBX,EAAE,MAAM,CAAC;IAC7C,OAAOA,EAAAA,KACC,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,MAAM;KAAG,EAC9B,SAAS,CAAC,CAACY,MAAQC,OAAOD,MAC1B,QAAQ,CACP;IAEJ,QAAQX,4BACL,QAAQ,CAACS,wBACT,QAAQ;IACX,MAAMV,CAAC,CAADA,OACC,CAAC;QAAC;QAAW;QAAS;KAAS,EACnC,OAAO,CAAC,WACR,QAAQ,GACR,QAAQ,CACP;AAEN;AAOO,MAAMc,oBAAoB,CAC/BX,OAEON,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAac;QACbR;IACF;AAIK,MAAMY,iCAAiCf,EAAE,MAAM,CAAC;IACrD,QAAQC,4BACL,QAAQ,CAAC,qDACT,QAAQ;IACX,SAASD,EAAAA,MACA,GACN,QAAQ,CACP;AAEN;AAMO,MAAMgB,4BAA4B,CACvCb,OAEON,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAakB;QACbZ;IACF;AAIK,MAAMc,0BAA0BjB,EAAE,MAAM,CAAC;IAC9C,WAAWA,CAAC,CAADA,OACJ,CAAC;QAAC;QAAQ;QAAM;QAAS;KAAO,EACpC,OAAO,CAAC,QACR,QAAQ,CAAC;IACZ,YAAYA,CAAC,CAADA,OACL,CAAC;QACJ;QACA;QACA;QACA;QACA;KACD,EACA,OAAO,CAAC,gBACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,CAAC;IACZ,QAAQC,4BACL,QAAQ,GACR,QAAQ,CAAC;AACd;AAEO,MAAMiB,qBAAqB,CAChCf,OAEON,aAAgE;QACrE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAaoB;QACbd;IACF;AAIK,MAAMgB,+BAA+BnB,EAAE,MAAM,CAAC;IACnD,MAAMC,4BAA4B,QAAQ,CAAC;IAC3C,IAAIA,4BAA4B,QAAQ,CAAC;AAC3C;AAMO,MAAMmB,0BAA0B,CACrCjB,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAasB;QACbhB;IACF;AAGK,MAAMkB,6BAA6BrB,EAAE,MAAM,CAAC;IACjD,QAAQC,4BAA4B,QAAQ,CAC1C;IAEF,UAAUD,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,GACR,QAAQ,CAAC;AACd;AAMO,MAAMsB,wBAAwB,CACnCnB,OAEON,aAAsE;QAC3E,MAAM;QACN,aAAa;QACb,aAAawB;QACblB;IACF;AAGK,MAAMoB,yBAAyBvB,EAAE,MAAM,CAAC;IAC7C,OAAOC,4BACJ,QAAQ,GACR,QAAQ,CACP;IAEJ,WAAWD,CAAC,CAADA,OACJ,CAAC;QAAC;QAAM;QAAQ;QAAQ;KAAQ,EACpC,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,CAAC;IACZ,KAAKC,4BACF,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUD,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,CAAC;IACZ,QAAQA,EAAAA,MACC,GACN,QAAQ,GACR,QAAQ,CACP;AAEN;AAWO,MAAMwB,oBAAoB,CAC/BrB,OAEON,aAA8D;QACnE,MAAM;QACN,aACE;QACF,aAAa0B;QACbpB;IACF;AAIK,MAAMsB,8BAA8BzB,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMyB,yBAAyB,CACpCvB,OAEON,aAGL;QACA,MAAM;QACN,aAAaa;QACb,gBAAgB;QAChB,aAAae;QACbtB;IACF;AAIK,MAAMwB,0BAA0B3B,EAAE,MAAM,CAAC;IAC9C,SAASA,EAAAA,MACA,GACN,QAAQ,CAAC;IACZ,QAAQA,EAAAA,OACE,GACP,QAAQ,CAAC;AACd;AAMO,MAAM4B,qBAAqB,CAChCzB,OAEON,aAAgE;QACrE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAa8B;QACbxB;IACF;AASK,MAAM0B,qBAAqB;IAChC,MAAMC,QAAQC,SAAS;IACvB,OAAOH,mBAAmB,OAAOI;QAC/BF,MAAM,aAAaE,MAAM,OAAO,EAAEA,MAAM,MAAM;QAC9C,IAAI,CAACA,MAAM,MAAM,EACf,MAAM,IAAIC,MAAM,CAAC,kBAAkB,EAAED,MAAM,OAAO,EAAE;IAGxD;AACF"}
|
package/dist/es/index.mjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.mjs","sources":["
|
|
1
|
+
{"version":3,"file":"index.mjs","sources":["../../src/index.ts"],"sourcesContent":["import { z } from 'zod';\nimport Service from './service/index';\nimport { TaskRunner } from './task-runner';\nimport { getVersion } from './utils';\n\nexport {\n plan,\n describeUserPage,\n AiLocateElement,\n getMidsceneLocationSchema,\n type MidsceneLocationResultType,\n PointSchema,\n SizeSchema,\n RectSchema,\n TMultimodalPromptSchema,\n TUserPromptSchema,\n type TMultimodalPrompt,\n type TUserPrompt,\n} from './ai-model/index';\n\nexport {\n MIDSCENE_MODEL_NAME,\n type CreateOpenAIClientFn,\n} from '@midscene/shared/env';\n\nexport type * from './types';\nexport { ServiceError } from './types';\n\nexport { z };\n\nexport default Service;\nexport { TaskRunner, Service, getVersion };\n\nexport type {\n MidsceneYamlScript,\n MidsceneYamlTask,\n MidsceneYamlFlowItem,\n MidsceneYamlConfigResult,\n MidsceneYamlConfig,\n MidsceneYamlScriptWebEnv,\n MidsceneYamlScriptAndroidEnv,\n MidsceneYamlScriptIOSEnv,\n MidsceneYamlScriptEnv,\n LocateOption,\n DetailedLocateParam,\n} from './yaml';\n\nexport { Agent, type AgentOpt, createAgent } from './agent';\n"],"names":["Service"],"mappings":";;;;;;;;AA8BA,YAAeA"}
|
package/dist/es/report.mjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"report.mjs","sources":["
|
|
1
|
+
{"version":3,"file":"report.mjs","sources":["../../src/report.ts"],"sourcesContent":["import * as fs from 'node:fs';\nimport * as path from 'node:path';\nimport { getMidsceneRunSubDir } from '@midscene/shared/common';\nimport { getReportFileName } from './agent';\nimport type { ReportFileWithAttributes } from './types';\nimport { getReportTpl, reportHTMLContent } from './utils';\n\nexport class ReportMergingTool {\n private reportInfos: ReportFileWithAttributes[] = [];\n public append(reportInfo: ReportFileWithAttributes) {\n this.reportInfos.push(reportInfo);\n }\n public clear() {\n this.reportInfos = [];\n }\n private extractScriptContent(filePath: string): string {\n // Regular expression to match content between script tags\n // Requires newline before <script and </script>\n const scriptRegex =\n /\\n<script type=\"midscene_web_dump\" type=\"application\\/json\"[^>]*>([\\s\\S]*?)\\n<\\/script>/;\n\n const fileContent = fs.readFileSync(filePath, 'utf-8');\n const match = scriptRegex.exec(fileContent);\n\n return match ? match[1].trim() : '';\n }\n\n public mergeReports(\n reportFileName: 'AUTO' | string = 'AUTO', // user custom report filename, save into midscene report dir if undefined\n opts?: {\n rmOriginalReports?: boolean; // whether to remove origin report files\n overwrite?: boolean; // if output filepath specified, throw an error when overwrite = true, otherwise overwrite the file\n },\n ): string | null {\n if (this.reportInfos.length <= 1) {\n console.log('Not enough report to merge');\n return null;\n }\n opts = Object.assign(\n {\n rmOriginalReports: false,\n overwrite: false,\n },\n opts || {},\n );\n const { rmOriginalReports, overwrite } = opts;\n let outputFilePath;\n const targetDir = `${getMidsceneRunSubDir('report')}`;\n if (reportFileName === 'AUTO') {\n outputFilePath = path.resolve(\n targetDir,\n `${getReportFileName('merged-report')}.html`,\n );\n } else {\n // user specified a output filepath\n outputFilePath = path.resolve(targetDir, `${reportFileName}.html`);\n if (fs.existsSync(outputFilePath) && !overwrite) {\n throw Error(\n `report file already existed: ${outputFilePath}\\nset override to true to overwrite this file.`,\n );\n } else if (fs.existsSync(outputFilePath) && overwrite) {\n fs.unlinkSync(outputFilePath);\n }\n }\n\n console.log(\n `Start merging ${this.reportInfos.length} reports...\\nCreating template file...`,\n );\n\n try {\n // Write template\n fs.appendFileSync(outputFilePath, getReportTpl());\n\n // Process all reports one by one\n for (let i = 0; i < this.reportInfos.length; i++) {\n const reportInfo = this.reportInfos[i];\n console.log(`Processing report ${i + 1}/${this.reportInfos.length}`);\n\n const dumpString = this.extractScriptContent(reportInfo.reportFilePath);\n const reportAttributes = reportInfo.reportAttributes;\n\n const reportHtmlStr = `${reportHTMLContent(\n {\n dumpString,\n attributes: {\n playwright_test_duration: reportAttributes.testDuration,\n playwright_test_status: reportAttributes.testStatus,\n playwright_test_title: reportAttributes.testTitle,\n playwright_test_id: reportAttributes.testId,\n playwright_test_description: reportAttributes.testDescription,\n },\n },\n undefined,\n undefined,\n false,\n )}\\n`; // use existed function to achieve report script content\n\n fs.appendFileSync(outputFilePath, reportHtmlStr);\n }\n\n console.log(`Successfully merged new report: ${outputFilePath}`);\n\n // Remove original reports if needed\n if (rmOriginalReports) {\n for (const info of this.reportInfos) {\n try {\n fs.unlinkSync(info.reportFilePath);\n } catch (error) {\n console.error(\n `Error deleting report ${info.reportFilePath}:`,\n error,\n );\n }\n }\n console.log(`Removed ${this.reportInfos.length} original reports`);\n }\n return outputFilePath;\n } catch (error) {\n console.error('Error in mergeReports:', error);\n throw error;\n }\n }\n}\n"],"names":["ReportMergingTool","reportInfo","filePath","scriptRegex","fileContent","fs","match","reportFileName","opts","console","Object","rmOriginalReports","overwrite","outputFilePath","targetDir","getMidsceneRunSubDir","path","getReportFileName","Error","getReportTpl","i","dumpString","reportAttributes","reportHtmlStr","reportHTMLContent","undefined","info","error"],"mappings":";;;;;;;;;;;;;;;AAOO,MAAMA;IAEJ,OAAOC,UAAoC,EAAE;QAClD,IAAI,CAAC,WAAW,CAAC,IAAI,CAACA;IACxB;IACO,QAAQ;QACb,IAAI,CAAC,WAAW,GAAG,EAAE;IACvB;IACQ,qBAAqBC,QAAgB,EAAU;QAGrD,MAAMC,cACJ;QAEF,MAAMC,cAAcC,aAAgBH,UAAU;QAC9C,MAAMI,QAAQH,YAAY,IAAI,CAACC;QAE/B,OAAOE,QAAQA,KAAK,CAAC,EAAE,CAAC,IAAI,KAAK;IACnC;IAEO,aACLC,iBAAkC,MAAM,EACxCC,IAGC,EACc;QACf,IAAI,IAAI,CAAC,WAAW,CAAC,MAAM,IAAI,GAAG;YAChCC,QAAQ,GAAG,CAAC;YACZ,OAAO;QACT;QACAD,OAAOE,OAAO,MAAM,CAClB;YACE,mBAAmB;YACnB,WAAW;QACb,GACAF,QAAQ,CAAC;QAEX,MAAM,EAAEG,iBAAiB,EAAEC,SAAS,EAAE,GAAGJ;QACzC,IAAIK;QACJ,MAAMC,YAAY,GAAGC,qBAAqB,WAAW;QACrD,IAAIR,AAAmB,WAAnBA,gBACFM,iBAAiBG,QACfF,WACA,GAAGG,kBAAkB,iBAAiB,KAAK,CAAC;aAEzC;YAELJ,iBAAiBG,QAAaF,WAAW,GAAGP,eAAe,KAAK,CAAC;YACjE,IAAIF,WAAcQ,mBAAmB,CAACD,WACpC,MAAMM,MACJ,CAAC,6BAA6B,EAAEL,eAAe,8CAA8C,CAAC;YAE3F,IAAIR,WAAcQ,mBAAmBD,WAC1CP,WAAcQ;QAElB;QAEAJ,QAAQ,GAAG,CACT,CAAC,cAAc,EAAE,IAAI,CAAC,WAAW,CAAC,MAAM,CAAC,sCAAsC,CAAC;QAGlF,IAAI;YAEFJ,eAAkBQ,gBAAgBM;YAGlC,IAAK,IAAIC,IAAI,GAAGA,IAAI,IAAI,CAAC,WAAW,CAAC,MAAM,EAAEA,IAAK;gBAChD,MAAMnB,aAAa,IAAI,CAAC,WAAW,CAACmB,EAAE;gBACtCX,QAAQ,GAAG,CAAC,CAAC,kBAAkB,EAAEW,IAAI,EAAE,CAAC,EAAE,IAAI,CAAC,WAAW,CAAC,MAAM,EAAE;gBAEnE,MAAMC,aAAa,IAAI,CAAC,oBAAoB,CAACpB,WAAW,cAAc;gBACtE,MAAMqB,mBAAmBrB,WAAW,gBAAgB;gBAEpD,MAAMsB,gBAAgB,GAAGC,kBACvB;oBACEH;oBACA,YAAY;wBACV,0BAA0BC,iBAAiB,YAAY;wBACvD,wBAAwBA,iBAAiB,UAAU;wBACnD,uBAAuBA,iBAAiB,SAAS;wBACjD,oBAAoBA,iBAAiB,MAAM;wBAC3C,6BAA6BA,iBAAiB,eAAe;oBAC/D;gBACF,GACAG,QACAA,QACA,OACA,EAAE,CAAC;gBAELpB,eAAkBQ,gBAAgBU;YACpC;YAEAd,QAAQ,GAAG,CAAC,CAAC,gCAAgC,EAAEI,gBAAgB;YAG/D,IAAIF,mBAAmB;gBACrB,KAAK,MAAMe,QAAQ,IAAI,CAAC,WAAW,CACjC,IAAI;oBACFrB,WAAcqB,KAAK,cAAc;gBACnC,EAAE,OAAOC,OAAO;oBACdlB,QAAQ,KAAK,CACX,CAAC,sBAAsB,EAAEiB,KAAK,cAAc,CAAC,CAAC,CAAC,EAC/CC;gBAEJ;gBAEFlB,QAAQ,GAAG,CAAC,CAAC,QAAQ,EAAE,IAAI,CAAC,WAAW,CAAC,MAAM,CAAC,iBAAiB,CAAC;YACnE;YACA,OAAOI;QACT,EAAE,OAAOc,OAAO;YACdlB,QAAQ,KAAK,CAAC,0BAA0BkB;YACxC,MAAMA;QACR;IACF;;QAjHA,uBAAQ,eAA0C,EAAE;;AAkHtD"}
|
|
@@ -21,7 +21,6 @@ function _define_property(obj, key, value) {
|
|
|
21
21
|
const debug = getDebug('ai:service');
|
|
22
22
|
class Service {
|
|
23
23
|
async locate(query, opt, modelConfig) {
|
|
24
|
-
var _parseResult_errors;
|
|
25
24
|
const queryPrompt = 'string' == typeof query ? query : query.prompt;
|
|
26
25
|
assert(queryPrompt, 'query is required for locate');
|
|
27
26
|
assert('object' == typeof query, 'query should be an object for locate');
|
|
@@ -34,7 +33,7 @@ class Service {
|
|
|
34
33
|
console.warn('The "deepThink" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model');
|
|
35
34
|
searchAreaPrompt = void 0;
|
|
36
35
|
}
|
|
37
|
-
const context =
|
|
36
|
+
const context = opt?.context || await this.contextRetrieverFn();
|
|
38
37
|
let searchArea;
|
|
39
38
|
let searchAreaRawResponse;
|
|
40
39
|
let searchAreaUsage;
|
|
@@ -70,7 +69,7 @@ class Service {
|
|
|
70
69
|
searchAreaUsage
|
|
71
70
|
};
|
|
72
71
|
let errorLog;
|
|
73
|
-
if (
|
|
72
|
+
if (parseResult.errors?.length) errorLog = `AI model failed to locate: \n${parseResult.errors.join('\n')}`;
|
|
74
73
|
const dumpData = {
|
|
75
74
|
type: 'locate',
|
|
76
75
|
userQuery: {
|
|
@@ -106,7 +105,6 @@ class Service {
|
|
|
106
105
|
};
|
|
107
106
|
}
|
|
108
107
|
async extract(dataDemand, modelConfig, opt, pageDescription, multimodalPrompt) {
|
|
109
|
-
var _parseResult_errors;
|
|
110
108
|
assert('object' == typeof dataDemand || 'string' == typeof dataDemand, `dataDemand should be object or string, but get ${typeof dataDemand}`);
|
|
111
109
|
const context = await this.contextRetrieverFn();
|
|
112
110
|
const startTime = Date.now();
|
|
@@ -125,7 +123,7 @@ class Service {
|
|
|
125
123
|
rawResponse: JSON.stringify(parseResult)
|
|
126
124
|
};
|
|
127
125
|
let errorLog;
|
|
128
|
-
if (
|
|
126
|
+
if (parseResult.errors?.length) errorLog = `AI response error: \n${parseResult.errors.join('\n')}`;
|
|
129
127
|
const dumpData = {
|
|
130
128
|
type: 'extract',
|
|
131
129
|
userQuery: {
|
|
@@ -173,7 +171,7 @@ class Service {
|
|
|
173
171
|
],
|
|
174
172
|
borderThickness: 3
|
|
175
173
|
});
|
|
176
|
-
if (
|
|
174
|
+
if (opt?.deepThink) {
|
|
177
175
|
const searchArea = expandSearchArea(targetRect, context.size, vlMode);
|
|
178
176
|
debug('describe: set searchArea', searchArea);
|
|
179
177
|
const croppedResult = await cropByRect(imagePayload, searchArea, 'qwen2.5-vl' === vlMode);
|
|
@@ -211,8 +209,8 @@ class Service {
|
|
|
211
209
|
assert(context, 'context is required for Service');
|
|
212
210
|
if ('function' == typeof context) this.contextRetrieverFn = context;
|
|
213
211
|
else this.contextRetrieverFn = ()=>Promise.resolve(context);
|
|
214
|
-
if (void 0 !==
|
|
215
|
-
if (void 0 !==
|
|
212
|
+
if (void 0 !== opt?.aiVendorFn) this.aiVendorFn = opt.aiVendorFn;
|
|
213
|
+
if (void 0 !== opt?.taskInfo) this.taskInfo = opt.taskInfo;
|
|
216
214
|
}
|
|
217
215
|
}
|
|
218
216
|
export { Service as default };
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"service/index.mjs","sources":["webpack://@midscene/core/./src/service/index.ts"],"sourcesContent":["import {\n AiExtractElementInfo,\n AiLocateElement,\n callAIWithObjectResponse,\n} from '@/ai-model/index';\nimport { AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport { AIActionType, type AIArgs, expandSearchArea } from '@/common';\nimport type {\n AIDescribeElementResponse,\n AIUsageInfo,\n DetailedLocateParam,\n LocateResultWithDump,\n PartialServiceDumpFromSDK,\n Rect,\n ServiceExtractOption,\n ServiceExtractParam,\n ServiceExtractResult,\n ServiceTaskInfo,\n UIContext,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport {\n type IModelConfig,\n MIDSCENE_FORCE_DEEP_THINK,\n globalConfigManager,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TMultimodalPrompt } from '../common';\nimport { createServiceDump } from './utils';\n\nexport interface LocateOpts {\n context?: UIContext;\n}\n\nexport type AnyValue<T> = {\n [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\ninterface ServiceOptions {\n taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n aiVendorFn?: typeof callAIWithObjectResponse;\n}\n\nconst debug = getDebug('ai:service');\nexport default class Service {\n contextRetrieverFn: () => Promise<UIContext> | UIContext;\n\n aiVendorFn: Exclude<ServiceOptions['aiVendorFn'], undefined> =\n callAIWithObjectResponse;\n\n taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n\n constructor(\n context: UIContext | (() => Promise<UIContext> | UIContext),\n opt?: ServiceOptions,\n ) {\n assert(context, 'context is required for Service');\n if (typeof context === 'function') {\n this.contextRetrieverFn = context;\n } else {\n this.contextRetrieverFn = () => Promise.resolve(context);\n }\n\n // just for unit test, aiVendorFn is callAIWithObjectResponse by default\n if (typeof opt?.aiVendorFn !== 'undefined') {\n this.aiVendorFn = opt.aiVendorFn;\n }\n if (typeof opt?.taskInfo !== 'undefined') {\n this.taskInfo = opt.taskInfo;\n }\n }\n\n async locate(\n query: DetailedLocateParam,\n opt: LocateOpts,\n modelConfig: IModelConfig,\n ): Promise<LocateResultWithDump> {\n const queryPrompt = typeof query === 'string' ? query : query.prompt;\n assert(queryPrompt, 'query is required for locate');\n\n assert(typeof query === 'object', 'query should be an object for locate');\n\n const globalDeepThinkSwitch = globalConfigManager.getEnvConfigInBoolean(\n MIDSCENE_FORCE_DEEP_THINK,\n );\n if (globalDeepThinkSwitch) {\n debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n }\n let searchAreaPrompt;\n if (query.deepThink || globalDeepThinkSwitch) {\n searchAreaPrompt = query.prompt;\n }\n\n const { vlMode } = modelConfig;\n\n if (searchAreaPrompt && !vlMode) {\n console.warn(\n 'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n );\n searchAreaPrompt = undefined;\n }\n\n const context = opt?.context || (await this.contextRetrieverFn());\n\n let searchArea: Rect | undefined = undefined;\n let searchAreaRawResponse: string | undefined = undefined;\n let searchAreaUsage: AIUsageInfo | undefined = undefined;\n let searchAreaResponse:\n | Awaited<ReturnType<typeof AiLocateSection>>\n | undefined = undefined;\n if (searchAreaPrompt) {\n searchAreaResponse = await AiLocateSection({\n context,\n sectionDescription: searchAreaPrompt,\n modelConfig,\n });\n assert(\n searchAreaResponse.rect,\n `cannot find search area for \"${searchAreaPrompt}\"${\n searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n }`,\n );\n searchAreaRawResponse = searchAreaResponse.rawResponse;\n searchAreaUsage = searchAreaResponse.usage;\n searchArea = searchAreaResponse.rect;\n }\n\n const startTime = Date.now();\n const { parseResult, rect, rawResponse, usage } = await AiLocateElement({\n callAIFn: this.aiVendorFn,\n context,\n targetElementDescription: queryPrompt,\n searchConfig: searchAreaResponse,\n modelConfig,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: ServiceTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(rawResponse),\n formatResponse: JSON.stringify(parseResult),\n usage,\n searchArea,\n searchAreaRawResponse,\n searchAreaUsage,\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialServiceDumpFromSDK = {\n type: 'locate',\n userQuery: {\n element: queryPrompt,\n },\n matchedElement: [],\n matchedRect: rect,\n data: null,\n taskInfo,\n deepThink: !!searchArea,\n error: errorLog,\n };\n\n const elements = parseResult.elements || [];\n\n const dump = createServiceDump({\n ...dumpData,\n matchedElement: elements,\n });\n\n if (errorLog) {\n throw new ServiceError(errorLog, dump);\n }\n\n if (elements.length > 1) {\n throw new ServiceError(\n `locate: multiple elements found, length = ${elements.length}`,\n dump,\n );\n }\n\n if (elements.length === 1) {\n return {\n element: {\n center: elements[0]!.center,\n rect: elements[0]!.rect,\n description: elements[0]!.description,\n },\n rect,\n dump,\n };\n }\n\n return {\n element: null,\n rect,\n dump,\n };\n }\n\n async extract<T>(\n dataDemand: ServiceExtractParam,\n modelConfig: IModelConfig,\n opt?: ServiceExtractOption,\n pageDescription?: string,\n multimodalPrompt?: TMultimodalPrompt,\n ): Promise<ServiceExtractResult<T>> {\n assert(\n typeof dataDemand === 'object' || typeof dataDemand === 'string',\n `dataDemand should be object or string, but get ${typeof dataDemand}`,\n );\n const context = await this.contextRetrieverFn();\n\n const startTime = Date.now();\n\n const { parseResult, usage } = await AiExtractElementInfo<T>({\n context,\n dataQuery: dataDemand,\n multimodalPrompt,\n extractOption: opt,\n modelConfig,\n pageDescription,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: ServiceTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(parseResult),\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialServiceDumpFromSDK = {\n type: 'extract',\n userQuery: {\n dataDemand,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n error: errorLog,\n };\n\n const { data, thought } = parseResult || {};\n\n // 4\n const dump = createServiceDump({\n ...dumpData,\n data,\n });\n\n if (errorLog && !data) {\n throw new ServiceError(errorLog, dump);\n }\n\n return {\n data,\n thought,\n usage,\n dump,\n };\n }\n\n async describe(\n target: Rect | [number, number],\n modelConfig: IModelConfig,\n opt?: {\n deepThink?: boolean;\n },\n ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n assert(target, 'target is required for service.describe');\n const context = await this.contextRetrieverFn();\n const { screenshotBase64, size } = context;\n assert(screenshotBase64, 'screenshot is required for service.describe');\n // The result of the \"describe\" function will be used for positioning, so essentially it is a form of grounding.\n const { vlMode } = modelConfig;\n const systemPrompt = elementDescriberInstruction();\n\n // Convert [x,y] center point to Rect if needed\n const defaultRectSize = 30;\n const targetRect: Rect = Array.isArray(target)\n ? {\n left: Math.floor(target[0] - defaultRectSize / 2),\n top: Math.floor(target[1] - defaultRectSize / 2),\n width: defaultRectSize,\n height: defaultRectSize,\n }\n : target;\n\n let imagePayload = await compositeElementInfoImg({\n inputImgBase64: screenshotBase64,\n size,\n elementsPositionInfo: [\n {\n rect: targetRect,\n },\n ],\n borderThickness: 3,\n });\n\n if (opt?.deepThink) {\n const searchArea = expandSearchArea(targetRect, context.size, vlMode);\n debug('describe: set searchArea', searchArea);\n const croppedResult = await cropByRect(\n imagePayload,\n searchArea,\n vlMode === 'qwen2.5-vl',\n );\n imagePayload = croppedResult.imageBase64;\n }\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n detail: 'high',\n },\n },\n ],\n },\n ];\n\n const callAIFn = this\n .aiVendorFn as typeof callAIWithObjectResponse<AIDescribeElementResponse>;\n\n const res = await callAIFn(\n msgs,\n AIActionType.DESCRIBE_ELEMENT,\n modelConfig,\n );\n\n const { content } = res;\n assert(!content.error, `describe failed: ${content.error}`);\n assert(content.description, 'failed to describe the element');\n return content;\n }\n}\n"],"names":["debug","getDebug","Service","query","opt","modelConfig","_parseResult_errors","queryPrompt","assert","globalDeepThinkSwitch","globalConfigManager","MIDSCENE_FORCE_DEEP_THINK","searchAreaPrompt","vlMode","console","undefined","context","searchArea","searchAreaRawResponse","searchAreaUsage","searchAreaResponse","AiLocateSection","startTime","Date","parseResult","rect","rawResponse","usage","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","elements","dump","createServiceDump","ServiceError","dataDemand","pageDescription","multimodalPrompt","AiExtractElementInfo","data","thought","target","screenshotBase64","size","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","croppedResult","cropByRect","msgs","callAIFn","res","AIActionType","content","callAIWithObjectResponse","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;;AA8CA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IA4BnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,WAAyB,EACM;YAyE3BC;QAxEJ,MAAMC,cAAc,AAAiB,YAAjB,OAAOJ,QAAqBA,QAAQA,MAAM,MAAM;QACpEK,OAAOD,aAAa;QAEpBC,OAAO,AAAiB,YAAjB,OAAOL,OAAoB;QAElC,MAAMM,wBAAwBC,oBAAoB,qBAAqB,CACrEC;QAEF,IAAIF,uBACFT,MAAM,yBAAyBS;QAEjC,IAAIG;QACJ,IAAIT,MAAM,SAAS,IAAIM,uBACrBG,mBAAmBT,MAAM,MAAM;QAGjC,MAAM,EAAEU,MAAM,EAAE,GAAGR;QAEnB,IAAIO,oBAAoB,CAACC,QAAQ;YAC/BC,QAAQ,IAAI,CACV;YAEFF,mBAAmBG;QACrB;QAEA,MAAMC,UAAUZ,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAM,MAAM,IAAI,CAAC,kBAAkB;QAE9D,IAAIa;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAGJ,IAAIR,kBAAkB;YACpBQ,qBAAqB,MAAMC,gBAAgB;gBACzCL;gBACA,oBAAoBJ;gBACpBP;YACF;YACAG,OACEY,mBAAmB,IAAI,EACvB,CAAC,6BAA6B,EAAER,iBAAiB,CAAC,EAChDQ,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAEJF,wBAAwBE,mBAAmB,WAAW;YACtDD,kBAAkBC,mBAAmB,KAAK;YAC1CH,aAAaG,mBAAmB,IAAI;QACtC;QAEA,MAAME,YAAYC,KAAK,GAAG;QAC1B,MAAM,EAAEC,WAAW,EAAEC,IAAI,EAAEC,WAAW,EAAEC,KAAK,EAAE,GAAG,MAAMC,gBAAgB;YACtE,UAAU,IAAI,CAAC,UAAU;YACzBZ;YACA,0BAA0BT;YAC1B,cAAca;YACdf;QACF;QAEA,MAAMwB,WAAWN,KAAK,GAAG,KAAKD;QAC9B,MAAMQ,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACL;YAC5B,gBAAgBK,KAAK,SAAS,CAACP;YAC/BG;YACAV;YACAC;YACAC;QACF;QAEA,IAAIa;QACJ,IAAI,QAAA1B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B0B,WAAW,CAAC,6BAA6B,EAAER,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG5E,MAAMS,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAAS1B;YACX;YACA,gBAAgB,EAAE;YAClB,aAAakB;YACb,MAAM;YACNK;YACA,WAAW,CAAC,CAACb;YACb,OAAOe;QACT;QAEA,MAAME,WAAWV,YAAY,QAAQ,IAAI,EAAE;QAE3C,MAAMW,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACX,gBAAgBC;QAClB;QAEA,IAAIF,UACF,MAAM,IAAIK,aAAaL,UAAUG;QAGnC,IAAID,SAAS,MAAM,GAAG,GACpB,MAAM,IAAIG,aACR,CAAC,0CAA0C,EAAEH,SAAS,MAAM,EAAE,EAC9DC;QAIJ,IAAID,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,SAAS;gBACP,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM;gBAC3B,MAAMA,QAAQ,CAAC,EAAE,CAAE,IAAI;gBACvB,aAAaA,QAAQ,CAAC,EAAE,CAAE,WAAW;YACvC;YACAT;YACAU;QACF;QAGF,OAAO;YACL,SAAS;YACTV;YACAU;QACF;IACF;IAEA,MAAM,QACJG,UAA+B,EAC/BjC,WAAyB,EACzBD,GAA0B,EAC1BmC,eAAwB,EACxBC,gBAAoC,EACF;YA0B9BlC;QAzBJE,OACE,AAAsB,YAAtB,OAAO8B,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAEvE,MAAMtB,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAE7C,MAAMM,YAAYC,KAAK,GAAG;QAE1B,MAAM,EAAEC,WAAW,EAAEG,KAAK,EAAE,GAAG,MAAMc,qBAAwB;YAC3DzB;YACA,WAAWsB;YACXE;YACA,eAAepC;YACfC;YACAkC;QACF;QAEA,MAAMV,WAAWN,KAAK,GAAG,KAAKD;QAC9B,MAAMQ,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACP;QAC9B;QAEA,IAAIQ;QACJ,IAAI,QAAA1B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B0B,WAAW,CAAC,qBAAqB,EAAER,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMS,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTK;YACF;YACA,gBAAgB,EAAE;YAClB,MAAM;YACNR;YACA,OAAOE;QACT;QAEA,MAAM,EAAEU,IAAI,EAAEC,OAAO,EAAE,GAAGnB,eAAe,CAAC;QAG1C,MAAMW,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACXS;QACF;QAEA,IAAIV,YAAY,CAACU,MACf,MAAM,IAAIL,aAAaL,UAAUG;QAGnC,OAAO;YACLO;YACAC;YACAhB;YACAQ;QACF;IACF;IAEA,MAAM,SACJS,MAA+B,EAC/BvC,WAAyB,EACzBD,GAEC,EACwD;QACzDI,OAAOoC,QAAQ;QACf,MAAM5B,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAC7C,MAAM,EAAE6B,gBAAgB,EAAEC,IAAI,EAAE,GAAG9B;QACnCR,OAAOqC,kBAAkB;QAEzB,MAAM,EAAEhC,MAAM,EAAE,GAAGR;QACnB,MAAM0C,eAAeC;QAGrB,MAAMC,kBAAkB;QACxB,MAAMC,aAAmBC,MAAM,OAAO,CAACP,UACnC;YACE,MAAMQ,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC/C,KAAKG,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC9C,OAAOA;YACP,QAAQA;QACV,IACAL;QAEJ,IAAIS,eAAe,MAAMC,wBAAwB;YAC/C,gBAAgBT;YAChBC;YACA,sBAAsB;gBACpB;oBACE,MAAMI;gBACR;aACD;YACD,iBAAiB;QACnB;QAEA,IAAI9C,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,EAAE;YAClB,MAAMa,aAAasC,iBAAiBL,YAAYlC,QAAQ,IAAI,EAAEH;YAC9Db,MAAM,4BAA4BiB;YAClC,MAAMuC,gBAAgB,MAAMC,WAC1BJ,cACApC,YACAJ,AAAW,iBAAXA;YAEFwC,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,WAAW,IAAI,CAClB,UAAU;QAEb,MAAMC,MAAM,MAAMD,SAChBD,MACAG,aAAa,gBAAgB,EAC7BxD;QAGF,MAAM,EAAEyD,OAAO,EAAE,GAAGF;QACpBpD,OAAO,CAACsD,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1DtD,OAAOsD,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IAvSA,YACE9C,OAA2D,EAC3DZ,GAAoB,CACpB;QAVF;QAEA,qCACE2D;QAEF;QAMEvD,OAAOQ,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMgD,QAAQ,OAAO,CAAChD;QAIlD,IAAI,AAA2B,WAApBZ,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,UAAU,AAAD,GACvB,IAAI,CAAC,UAAU,GAAGA,IAAI,UAAU;QAElC,IAAI,AAAyB,WAAlBA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,QAAQ,AAAD,GACrB,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAsRF"}
|
|
1
|
+
{"version":3,"file":"service/index.mjs","sources":["../../../src/service/index.ts"],"sourcesContent":["import {\n AiExtractElementInfo,\n AiLocateElement,\n callAIWithObjectResponse,\n} from '@/ai-model/index';\nimport { AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport { AIActionType, type AIArgs, expandSearchArea } from '@/common';\nimport type {\n AIDescribeElementResponse,\n AIUsageInfo,\n DetailedLocateParam,\n LocateResultWithDump,\n PartialServiceDumpFromSDK,\n Rect,\n ServiceExtractOption,\n ServiceExtractParam,\n ServiceExtractResult,\n ServiceTaskInfo,\n UIContext,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport {\n type IModelConfig,\n MIDSCENE_FORCE_DEEP_THINK,\n globalConfigManager,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TMultimodalPrompt } from '../common';\nimport { createServiceDump } from './utils';\n\nexport interface LocateOpts {\n context?: UIContext;\n}\n\nexport type AnyValue<T> = {\n [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\ninterface ServiceOptions {\n taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n aiVendorFn?: typeof callAIWithObjectResponse;\n}\n\nconst debug = getDebug('ai:service');\nexport default class Service {\n contextRetrieverFn: () => Promise<UIContext> | UIContext;\n\n aiVendorFn: Exclude<ServiceOptions['aiVendorFn'], undefined> =\n callAIWithObjectResponse;\n\n taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n\n constructor(\n context: UIContext | (() => Promise<UIContext> | UIContext),\n opt?: ServiceOptions,\n ) {\n assert(context, 'context is required for Service');\n if (typeof context === 'function') {\n this.contextRetrieverFn = context;\n } else {\n this.contextRetrieverFn = () => Promise.resolve(context);\n }\n\n // just for unit test, aiVendorFn is callAIWithObjectResponse by default\n if (typeof opt?.aiVendorFn !== 'undefined') {\n this.aiVendorFn = opt.aiVendorFn;\n }\n if (typeof opt?.taskInfo !== 'undefined') {\n this.taskInfo = opt.taskInfo;\n }\n }\n\n async locate(\n query: DetailedLocateParam,\n opt: LocateOpts,\n modelConfig: IModelConfig,\n ): Promise<LocateResultWithDump> {\n const queryPrompt = typeof query === 'string' ? query : query.prompt;\n assert(queryPrompt, 'query is required for locate');\n\n assert(typeof query === 'object', 'query should be an object for locate');\n\n const globalDeepThinkSwitch = globalConfigManager.getEnvConfigInBoolean(\n MIDSCENE_FORCE_DEEP_THINK,\n );\n if (globalDeepThinkSwitch) {\n debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n }\n let searchAreaPrompt;\n if (query.deepThink || globalDeepThinkSwitch) {\n searchAreaPrompt = query.prompt;\n }\n\n const { vlMode } = modelConfig;\n\n if (searchAreaPrompt && !vlMode) {\n console.warn(\n 'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n );\n searchAreaPrompt = undefined;\n }\n\n const context = opt?.context || (await this.contextRetrieverFn());\n\n let searchArea: Rect | undefined = undefined;\n let searchAreaRawResponse: string | undefined = undefined;\n let searchAreaUsage: AIUsageInfo | undefined = undefined;\n let searchAreaResponse:\n | Awaited<ReturnType<typeof AiLocateSection>>\n | undefined = undefined;\n if (searchAreaPrompt) {\n searchAreaResponse = await AiLocateSection({\n context,\n sectionDescription: searchAreaPrompt,\n modelConfig,\n });\n assert(\n searchAreaResponse.rect,\n `cannot find search area for \"${searchAreaPrompt}\"${\n searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n }`,\n );\n searchAreaRawResponse = searchAreaResponse.rawResponse;\n searchAreaUsage = searchAreaResponse.usage;\n searchArea = searchAreaResponse.rect;\n }\n\n const startTime = Date.now();\n const { parseResult, rect, rawResponse, usage } = await AiLocateElement({\n callAIFn: this.aiVendorFn,\n context,\n targetElementDescription: queryPrompt,\n searchConfig: searchAreaResponse,\n modelConfig,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: ServiceTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(rawResponse),\n formatResponse: JSON.stringify(parseResult),\n usage,\n searchArea,\n searchAreaRawResponse,\n searchAreaUsage,\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialServiceDumpFromSDK = {\n type: 'locate',\n userQuery: {\n element: queryPrompt,\n },\n matchedElement: [],\n matchedRect: rect,\n data: null,\n taskInfo,\n deepThink: !!searchArea,\n error: errorLog,\n };\n\n const elements = parseResult.elements || [];\n\n const dump = createServiceDump({\n ...dumpData,\n matchedElement: elements,\n });\n\n if (errorLog) {\n throw new ServiceError(errorLog, dump);\n }\n\n if (elements.length > 1) {\n throw new ServiceError(\n `locate: multiple elements found, length = ${elements.length}`,\n dump,\n );\n }\n\n if (elements.length === 1) {\n return {\n element: {\n center: elements[0]!.center,\n rect: elements[0]!.rect,\n description: elements[0]!.description,\n },\n rect,\n dump,\n };\n }\n\n return {\n element: null,\n rect,\n dump,\n };\n }\n\n async extract<T>(\n dataDemand: ServiceExtractParam,\n modelConfig: IModelConfig,\n opt?: ServiceExtractOption,\n pageDescription?: string,\n multimodalPrompt?: TMultimodalPrompt,\n ): Promise<ServiceExtractResult<T>> {\n assert(\n typeof dataDemand === 'object' || typeof dataDemand === 'string',\n `dataDemand should be object or string, but get ${typeof dataDemand}`,\n );\n const context = await this.contextRetrieverFn();\n\n const startTime = Date.now();\n\n const { parseResult, usage } = await AiExtractElementInfo<T>({\n context,\n dataQuery: dataDemand,\n multimodalPrompt,\n extractOption: opt,\n modelConfig,\n pageDescription,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: ServiceTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(parseResult),\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialServiceDumpFromSDK = {\n type: 'extract',\n userQuery: {\n dataDemand,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n error: errorLog,\n };\n\n const { data, thought } = parseResult || {};\n\n // 4\n const dump = createServiceDump({\n ...dumpData,\n data,\n });\n\n if (errorLog && !data) {\n throw new ServiceError(errorLog, dump);\n }\n\n return {\n data,\n thought,\n usage,\n dump,\n };\n }\n\n async describe(\n target: Rect | [number, number],\n modelConfig: IModelConfig,\n opt?: {\n deepThink?: boolean;\n },\n ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n assert(target, 'target is required for service.describe');\n const context = await this.contextRetrieverFn();\n const { screenshotBase64, size } = context;\n assert(screenshotBase64, 'screenshot is required for service.describe');\n // The result of the \"describe\" function will be used for positioning, so essentially it is a form of grounding.\n const { vlMode } = modelConfig;\n const systemPrompt = elementDescriberInstruction();\n\n // Convert [x,y] center point to Rect if needed\n const defaultRectSize = 30;\n const targetRect: Rect = Array.isArray(target)\n ? {\n left: Math.floor(target[0] - defaultRectSize / 2),\n top: Math.floor(target[1] - defaultRectSize / 2),\n width: defaultRectSize,\n height: defaultRectSize,\n }\n : target;\n\n let imagePayload = await compositeElementInfoImg({\n inputImgBase64: screenshotBase64,\n size,\n elementsPositionInfo: [\n {\n rect: targetRect,\n },\n ],\n borderThickness: 3,\n });\n\n if (opt?.deepThink) {\n const searchArea = expandSearchArea(targetRect, context.size, vlMode);\n debug('describe: set searchArea', searchArea);\n const croppedResult = await cropByRect(\n imagePayload,\n searchArea,\n vlMode === 'qwen2.5-vl',\n );\n imagePayload = croppedResult.imageBase64;\n }\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n detail: 'high',\n },\n },\n ],\n },\n ];\n\n const callAIFn = this\n .aiVendorFn as typeof callAIWithObjectResponse<AIDescribeElementResponse>;\n\n const res = await callAIFn(\n msgs,\n AIActionType.DESCRIBE_ELEMENT,\n modelConfig,\n );\n\n const { content } = res;\n assert(!content.error, `describe failed: ${content.error}`);\n assert(content.description, 'failed to describe the element');\n return content;\n }\n}\n"],"names":["debug","getDebug","Service","query","opt","modelConfig","queryPrompt","assert","globalDeepThinkSwitch","globalConfigManager","MIDSCENE_FORCE_DEEP_THINK","searchAreaPrompt","vlMode","console","undefined","context","searchArea","searchAreaRawResponse","searchAreaUsage","searchAreaResponse","AiLocateSection","startTime","Date","parseResult","rect","rawResponse","usage","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","elements","dump","createServiceDump","ServiceError","dataDemand","pageDescription","multimodalPrompt","AiExtractElementInfo","data","thought","target","screenshotBase64","size","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","croppedResult","cropByRect","msgs","callAIFn","res","AIActionType","content","callAIWithObjectResponse","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;;AA8CA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IA4BnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,WAAyB,EACM;QAC/B,MAAMC,cAAc,AAAiB,YAAjB,OAAOH,QAAqBA,QAAQA,MAAM,MAAM;QACpEI,OAAOD,aAAa;QAEpBC,OAAO,AAAiB,YAAjB,OAAOJ,OAAoB;QAElC,MAAMK,wBAAwBC,oBAAoB,qBAAqB,CACrEC;QAEF,IAAIF,uBACFR,MAAM,yBAAyBQ;QAEjC,IAAIG;QACJ,IAAIR,MAAM,SAAS,IAAIK,uBACrBG,mBAAmBR,MAAM,MAAM;QAGjC,MAAM,EAAES,MAAM,EAAE,GAAGP;QAEnB,IAAIM,oBAAoB,CAACC,QAAQ;YAC/BC,QAAQ,IAAI,CACV;YAEFF,mBAAmBG;QACrB;QAEA,MAAMC,UAAUX,KAAK,WAAY,MAAM,IAAI,CAAC,kBAAkB;QAE9D,IAAIY;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAGJ,IAAIR,kBAAkB;YACpBQ,qBAAqB,MAAMC,gBAAgB;gBACzCL;gBACA,oBAAoBJ;gBACpBN;YACF;YACAE,OACEY,mBAAmB,IAAI,EACvB,CAAC,6BAA6B,EAAER,iBAAiB,CAAC,EAChDQ,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAEJF,wBAAwBE,mBAAmB,WAAW;YACtDD,kBAAkBC,mBAAmB,KAAK;YAC1CH,aAAaG,mBAAmB,IAAI;QACtC;QAEA,MAAME,YAAYC,KAAK,GAAG;QAC1B,MAAM,EAAEC,WAAW,EAAEC,IAAI,EAAEC,WAAW,EAAEC,KAAK,EAAE,GAAG,MAAMC,gBAAgB;YACtE,UAAU,IAAI,CAAC,UAAU;YACzBZ;YACA,0BAA0BT;YAC1B,cAAca;YACdd;QACF;QAEA,MAAMuB,WAAWN,KAAK,GAAG,KAAKD;QAC9B,MAAMQ,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACL;YAC5B,gBAAgBK,KAAK,SAAS,CAACP;YAC/BG;YACAV;YACAC;YACAC;QACF;QAEA,IAAIa;QACJ,IAAIR,YAAY,MAAM,EAAE,QACtBQ,WAAW,CAAC,6BAA6B,EAAER,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG5E,MAAMS,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAAS1B;YACX;YACA,gBAAgB,EAAE;YAClB,aAAakB;YACb,MAAM;YACNK;YACA,WAAW,CAAC,CAACb;YACb,OAAOe;QACT;QAEA,MAAME,WAAWV,YAAY,QAAQ,IAAI,EAAE;QAE3C,MAAMW,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACX,gBAAgBC;QAClB;QAEA,IAAIF,UACF,MAAM,IAAIK,aAAaL,UAAUG;QAGnC,IAAID,SAAS,MAAM,GAAG,GACpB,MAAM,IAAIG,aACR,CAAC,0CAA0C,EAAEH,SAAS,MAAM,EAAE,EAC9DC;QAIJ,IAAID,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,SAAS;gBACP,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM;gBAC3B,MAAMA,QAAQ,CAAC,EAAE,CAAE,IAAI;gBACvB,aAAaA,QAAQ,CAAC,EAAE,CAAE,WAAW;YACvC;YACAT;YACAU;QACF;QAGF,OAAO;YACL,SAAS;YACTV;YACAU;QACF;IACF;IAEA,MAAM,QACJG,UAA+B,EAC/BhC,WAAyB,EACzBD,GAA0B,EAC1BkC,eAAwB,EACxBC,gBAAoC,EACF;QAClChC,OACE,AAAsB,YAAtB,OAAO8B,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAEvE,MAAMtB,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAE7C,MAAMM,YAAYC,KAAK,GAAG;QAE1B,MAAM,EAAEC,WAAW,EAAEG,KAAK,EAAE,GAAG,MAAMc,qBAAwB;YAC3DzB;YACA,WAAWsB;YACXE;YACA,eAAenC;YACfC;YACAiC;QACF;QAEA,MAAMV,WAAWN,KAAK,GAAG,KAAKD;QAC9B,MAAMQ,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACP;QAC9B;QAEA,IAAIQ;QACJ,IAAIR,YAAY,MAAM,EAAE,QACtBQ,WAAW,CAAC,qBAAqB,EAAER,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMS,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTK;YACF;YACA,gBAAgB,EAAE;YAClB,MAAM;YACNR;YACA,OAAOE;QACT;QAEA,MAAM,EAAEU,IAAI,EAAEC,OAAO,EAAE,GAAGnB,eAAe,CAAC;QAG1C,MAAMW,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACXS;QACF;QAEA,IAAIV,YAAY,CAACU,MACf,MAAM,IAAIL,aAAaL,UAAUG;QAGnC,OAAO;YACLO;YACAC;YACAhB;YACAQ;QACF;IACF;IAEA,MAAM,SACJS,MAA+B,EAC/BtC,WAAyB,EACzBD,GAEC,EACwD;QACzDG,OAAOoC,QAAQ;QACf,MAAM5B,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAC7C,MAAM,EAAE6B,gBAAgB,EAAEC,IAAI,EAAE,GAAG9B;QACnCR,OAAOqC,kBAAkB;QAEzB,MAAM,EAAEhC,MAAM,EAAE,GAAGP;QACnB,MAAMyC,eAAeC;QAGrB,MAAMC,kBAAkB;QACxB,MAAMC,aAAmBC,MAAM,OAAO,CAACP,UACnC;YACE,MAAMQ,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC/C,KAAKG,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC9C,OAAOA;YACP,QAAQA;QACV,IACAL;QAEJ,IAAIS,eAAe,MAAMC,wBAAwB;YAC/C,gBAAgBT;YAChBC;YACA,sBAAsB;gBACpB;oBACE,MAAMI;gBACR;aACD;YACD,iBAAiB;QACnB;QAEA,IAAI7C,KAAK,WAAW;YAClB,MAAMY,aAAasC,iBAAiBL,YAAYlC,QAAQ,IAAI,EAAEH;YAC9DZ,MAAM,4BAA4BgB;YAClC,MAAMuC,gBAAgB,MAAMC,WAC1BJ,cACApC,YACAJ,AAAW,iBAAXA;YAEFwC,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,WAAW,IAAI,CAClB,UAAU;QAEb,MAAMC,MAAM,MAAMD,SAChBD,MACAG,aAAa,gBAAgB,EAC7BvD;QAGF,MAAM,EAAEwD,OAAO,EAAE,GAAGF;QACpBpD,OAAO,CAACsD,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1DtD,OAAOsD,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IAvSA,YACE9C,OAA2D,EAC3DX,GAAoB,CACpB;QAVF;QAEA,qCACE0D;QAEF;QAMEvD,OAAOQ,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMgD,QAAQ,OAAO,CAAChD;QAIlD,IAAI,AAA2B,WAApBX,KAAK,YACd,IAAI,CAAC,UAAU,GAAGA,IAAI,UAAU;QAElC,IAAI,AAAyB,WAAlBA,KAAK,UACd,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAsRF"}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"service/utils.mjs","sources":["
|
|
1
|
+
{"version":3,"file":"service/utils.mjs","sources":["../../../src/service/utils.ts"],"sourcesContent":["import type { DumpMeta, PartialServiceDumpFromSDK, ServiceDump } from '@/types';\nimport { uuid } from '@midscene/shared/utils';\n\nexport function createServiceDump(\n data: PartialServiceDumpFromSDK,\n): ServiceDump {\n const baseData: DumpMeta = {\n logTime: Date.now(),\n };\n const finalData: ServiceDump = {\n logId: uuid(),\n ...baseData,\n ...data,\n };\n\n return finalData;\n}\n"],"names":["createServiceDump","data","baseData","Date","finalData","uuid"],"mappings":";AAGO,SAASA,kBACdC,IAA+B;IAE/B,MAAMC,WAAqB;QACzB,SAASC,KAAK,GAAG;IACnB;IACA,MAAMC,YAAyB;QAC7B,OAAOC;QACP,GAAGH,QAAQ;QACX,GAAGD,IAAI;IACT;IAEA,OAAOG;AACT"}
|
package/dist/es/task-runner.mjs
CHANGED
|
@@ -18,13 +18,11 @@ class TaskRunner {
|
|
|
18
18
|
await this.onTaskUpdate(this, error);
|
|
19
19
|
}
|
|
20
20
|
async getUiContext(options) {
|
|
21
|
-
var _this_lastUiContext;
|
|
22
21
|
const now = Date.now();
|
|
23
|
-
const shouldReuse = !
|
|
24
|
-
if (shouldReuse &&
|
|
25
|
-
var _this_lastUiContext1;
|
|
22
|
+
const shouldReuse = !options?.forceRefresh && this.lastUiContext && now - this.lastUiContext.capturedAt <= UI_CONTEXT_CACHE_TTL_MS;
|
|
23
|
+
if (shouldReuse && this.lastUiContext?.context) {
|
|
26
24
|
debug(`reuse cached uiContext captured ${now - this.lastUiContext.capturedAt}ms ago`);
|
|
27
|
-
return
|
|
25
|
+
return this.lastUiContext?.context;
|
|
28
26
|
}
|
|
29
27
|
try {
|
|
30
28
|
const uiContext = await this.uiContextBuilder();
|
|
@@ -44,14 +42,14 @@ class TaskRunner {
|
|
|
44
42
|
const uiContext = await this.getUiContext({
|
|
45
43
|
forceRefresh: true
|
|
46
44
|
});
|
|
47
|
-
return
|
|
45
|
+
return uiContext?.screenshotBase64;
|
|
48
46
|
} catch (error) {
|
|
49
47
|
console.error('error while capturing screenshot', error);
|
|
50
48
|
}
|
|
51
49
|
}
|
|
52
50
|
attachRecorderItem(task, contextOrScreenshot, phase) {
|
|
53
51
|
const timing = phase;
|
|
54
|
-
const screenshot = 'string' == typeof contextOrScreenshot ? contextOrScreenshot :
|
|
52
|
+
const screenshot = 'string' == typeof contextOrScreenshot ? contextOrScreenshot : contextOrScreenshot?.screenshotBase64;
|
|
55
53
|
if (!timing || !screenshot) return;
|
|
56
54
|
const recorderItem = {
|
|
57
55
|
type: 'screenshot',
|
|
@@ -74,9 +72,8 @@ class TaskRunner {
|
|
|
74
72
|
};
|
|
75
73
|
}
|
|
76
74
|
normalizeStatusFromError(options, errorMessage) {
|
|
77
|
-
var _this_latestErrorTask, _this_latestErrorTask1;
|
|
78
75
|
if ('error' !== this.status) return;
|
|
79
|
-
assert(
|
|
76
|
+
assert(options?.allowWhenError, errorMessage || `task runner is in error state, cannot proceed\nerror=${this.latestErrorTask()?.error}\n${this.latestErrorTask()?.errorStack}`);
|
|
80
77
|
this.status = this.tasks.length > 0 ? 'pending' : 'init';
|
|
81
78
|
}
|
|
82
79
|
findPreviousNonSubTaskUIContext(currentIndex) {
|
|
@@ -88,8 +85,7 @@ class TaskRunner {
|
|
|
88
85
|
}
|
|
89
86
|
}
|
|
90
87
|
async append(task, options) {
|
|
91
|
-
|
|
92
|
-
this.normalizeStatusFromError(options, `task runner is in error state, cannot append task\nerror=${null == (_this_latestErrorTask = this.latestErrorTask()) ? void 0 : _this_latestErrorTask.error}\n${null == (_this_latestErrorTask1 = this.latestErrorTask()) ? void 0 : _this_latestErrorTask1.errorStack}`);
|
|
88
|
+
this.normalizeStatusFromError(options, `task runner is in error state, cannot append task\nerror=${this.latestErrorTask()?.error}\n${this.latestErrorTask()?.errorStack}`);
|
|
93
89
|
if (Array.isArray(task)) this.tasks.push(...task.map((item)=>this.markTaskAsPending(item)));
|
|
94
90
|
else this.tasks.push(this.markTaskAsPending(task));
|
|
95
91
|
if ('running' !== this.status) this.status = 'pending';
|
|
@@ -141,7 +137,7 @@ class TaskRunner {
|
|
|
141
137
|
task.uiContext = uiContext;
|
|
142
138
|
const executorContext = {
|
|
143
139
|
task,
|
|
144
|
-
element:
|
|
140
|
+
element: previousFindOutput?.element,
|
|
145
141
|
uiContext
|
|
146
142
|
};
|
|
147
143
|
if ('Insight' === task.type) {
|
|
@@ -149,7 +145,7 @@ class TaskRunner {
|
|
|
149
145
|
returnValue = await task.executor(param, executorContext);
|
|
150
146
|
} else if ('Planning' === task.type) {
|
|
151
147
|
returnValue = await task.executor(param, executorContext);
|
|
152
|
-
if ('Locate' === task.subType) previousFindOutput =
|
|
148
|
+
if ('Locate' === task.subType) previousFindOutput = returnValue?.output;
|
|
153
149
|
} else if ('Action Space' === task.type) returnValue = await task.executor(param, executorContext);
|
|
154
150
|
else {
|
|
155
151
|
console.warn(`unsupported task type: ${task.type}, will try to execute it directly`);
|
|
@@ -169,7 +165,7 @@ class TaskRunner {
|
|
|
169
165
|
} catch (e) {
|
|
170
166
|
successfullyCompleted = false;
|
|
171
167
|
task.error = e;
|
|
172
|
-
task.errorMessage =
|
|
168
|
+
task.errorMessage = e?.message || ('string' == typeof e ? e : 'error-without-message');
|
|
173
169
|
task.errorStack = e.stack;
|
|
174
170
|
task.status = 'failed';
|
|
175
171
|
task.timing.end = Date.now();
|
|
@@ -187,11 +183,11 @@ class TaskRunner {
|
|
|
187
183
|
} else {
|
|
188
184
|
this.status = 'error';
|
|
189
185
|
const errorTask = this.latestErrorTask();
|
|
190
|
-
const messageBase =
|
|
191
|
-
const stack =
|
|
186
|
+
const messageBase = errorTask?.errorMessage || (errorTask?.error ? String(errorTask.error) : 'Task execution failed');
|
|
187
|
+
const stack = errorTask?.errorStack;
|
|
192
188
|
const message = stack ? `${messageBase}\n${stack}` : messageBase;
|
|
193
189
|
finalizeError = new TaskExecutionError(message, this, errorTask, {
|
|
194
|
-
cause:
|
|
190
|
+
cause: errorTask?.error
|
|
195
191
|
});
|
|
196
192
|
await this.emitOnTaskUpdate(finalizeError);
|
|
197
193
|
}
|
|
@@ -248,12 +244,12 @@ class TaskRunner {
|
|
|
248
244
|
_define_property(this, "uiContextBuilder", void 0);
|
|
249
245
|
_define_property(this, "onTaskUpdate", void 0);
|
|
250
246
|
_define_property(this, "lastUiContext", void 0);
|
|
251
|
-
this.status =
|
|
247
|
+
this.status = options?.tasks && options.tasks.length > 0 ? 'pending' : 'init';
|
|
252
248
|
this.name = name;
|
|
253
|
-
this.tasks = (
|
|
254
|
-
this.onTaskStart =
|
|
249
|
+
this.tasks = (options?.tasks || []).map((item)=>this.markTaskAsPending(item));
|
|
250
|
+
this.onTaskStart = options?.onTaskStart;
|
|
255
251
|
this.uiContextBuilder = uiContextBuilder;
|
|
256
|
-
this.onTaskUpdate =
|
|
252
|
+
this.onTaskUpdate = options?.onTaskUpdate;
|
|
257
253
|
}
|
|
258
254
|
}
|
|
259
255
|
class TaskExecutionError extends Error {
|