@midscene/core 1.2.2-beta-20260116114131.0 → 1.2.2-beta-20260119111553.0
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 +4 -4
- package/dist/es/agent/agent.mjs.map +1 -1
- package/dist/es/agent/tasks.mjs +3 -4
- package/dist/es/agent/tasks.mjs.map +1 -1
- package/dist/es/agent/utils.mjs +1 -1
- package/dist/es/ai-model/auto-glm/index.mjs +2 -2
- package/dist/es/ai-model/auto-glm/planning.mjs +1 -1
- package/dist/es/ai-model/auto-glm/planning.mjs.map +1 -1
- package/dist/es/ai-model/auto-glm/prompt.mjs +8 -8
- package/dist/es/ai-model/auto-glm/prompt.mjs.map +1 -1
- package/dist/es/ai-model/auto-glm/util.mjs +6 -3
- package/dist/es/ai-model/auto-glm/util.mjs.map +1 -1
- package/dist/es/ai-model/index.mjs +2 -2
- package/dist/es/ai-model/inspect.mjs +12 -12
- package/dist/es/ai-model/inspect.mjs.map +1 -1
- package/dist/es/ai-model/llm-planning.mjs +4 -4
- package/dist/es/ai-model/llm-planning.mjs.map +1 -1
- package/dist/es/ai-model/prompt/common.mjs +2 -2
- package/dist/es/ai-model/prompt/common.mjs.map +1 -1
- package/dist/es/ai-model/prompt/extraction.mjs +3 -1
- package/dist/es/ai-model/prompt/extraction.mjs.map +1 -1
- package/dist/es/ai-model/prompt/llm-locator.mjs +5 -3
- package/dist/es/ai-model/prompt/llm-locator.mjs.map +1 -1
- package/dist/es/ai-model/prompt/llm-planning.mjs +15 -13
- package/dist/es/ai-model/prompt/llm-planning.mjs.map +1 -1
- package/dist/es/ai-model/prompt/llm-section-locator.mjs +5 -3
- package/dist/es/ai-model/prompt/llm-section-locator.mjs.map +1 -1
- package/dist/es/ai-model/service-caller/index.mjs +12 -13
- package/dist/es/ai-model/service-caller/index.mjs.map +1 -1
- package/dist/es/ai-model/ui-tars-planning.mjs +2 -24
- package/dist/es/ai-model/ui-tars-planning.mjs.map +1 -1
- package/dist/es/common.mjs +10 -9
- package/dist/es/common.mjs.map +1 -1
- package/dist/es/device/index.mjs +3 -3
- package/dist/es/device/index.mjs.map +1 -1
- package/dist/es/service/index.mjs +6 -6
- package/dist/es/service/index.mjs.map +1 -1
- package/dist/es/utils.mjs +2 -2
- package/dist/lib/agent/agent.js +3 -3
- package/dist/lib/agent/agent.js.map +1 -1
- package/dist/lib/agent/tasks.js +2 -3
- package/dist/lib/agent/tasks.js.map +1 -1
- package/dist/lib/agent/utils.js +1 -1
- package/dist/lib/ai-model/auto-glm/index.js +3 -0
- package/dist/lib/ai-model/auto-glm/planning.js +1 -1
- package/dist/lib/ai-model/auto-glm/planning.js.map +1 -1
- package/dist/lib/ai-model/auto-glm/prompt.js +8 -8
- package/dist/lib/ai-model/auto-glm/prompt.js.map +1 -1
- package/dist/lib/ai-model/auto-glm/util.js +10 -4
- package/dist/lib/ai-model/auto-glm/util.js.map +1 -1
- package/dist/lib/ai-model/index.js +0 -3
- package/dist/lib/ai-model/inspect.js +12 -12
- package/dist/lib/ai-model/inspect.js.map +1 -1
- package/dist/lib/ai-model/llm-planning.js +4 -4
- package/dist/lib/ai-model/llm-planning.js.map +1 -1
- package/dist/lib/ai-model/prompt/common.js +2 -2
- package/dist/lib/ai-model/prompt/common.js.map +1 -1
- package/dist/lib/ai-model/prompt/extraction.js +5 -3
- package/dist/lib/ai-model/prompt/extraction.js.map +1 -1
- package/dist/lib/ai-model/prompt/llm-locator.js +5 -3
- package/dist/lib/ai-model/prompt/llm-locator.js.map +1 -1
- package/dist/lib/ai-model/prompt/llm-planning.js +15 -13
- package/dist/lib/ai-model/prompt/llm-planning.js.map +1 -1
- package/dist/lib/ai-model/prompt/llm-section-locator.js +5 -3
- package/dist/lib/ai-model/prompt/llm-section-locator.js.map +1 -1
- package/dist/lib/ai-model/service-caller/index.js +11 -12
- package/dist/lib/ai-model/service-caller/index.js.map +1 -1
- package/dist/lib/ai-model/ui-tars-planning.js +1 -26
- package/dist/lib/ai-model/ui-tars-planning.js.map +1 -1
- package/dist/lib/common.js +10 -9
- package/dist/lib/common.js.map +1 -1
- package/dist/lib/device/index.js +3 -3
- package/dist/lib/device/index.js.map +1 -1
- package/dist/lib/service/index.js +6 -6
- package/dist/lib/service/index.js.map +1 -1
- package/dist/lib/utils.js +2 -2
- package/dist/types/agent/agent.d.ts +2 -2
- package/dist/types/ai-model/auto-glm/index.d.ts +1 -1
- package/dist/types/ai-model/auto-glm/prompt.d.ts +3 -3
- package/dist/types/ai-model/auto-glm/util.d.ts +11 -5
- package/dist/types/ai-model/index.d.ts +1 -1
- package/dist/types/ai-model/prompt/common.d.ts +2 -2
- package/dist/types/ai-model/prompt/llm-locator.d.ts +2 -2
- package/dist/types/ai-model/prompt/llm-planning.d.ts +3 -3
- package/dist/types/ai-model/prompt/llm-section-locator.d.ts +2 -2
- package/dist/types/ai-model/service-caller/index.d.ts +2 -2
- package/dist/types/ai-model/ui-tars-planning.d.ts +2 -3
- package/dist/types/common.d.ts +5 -5
- package/dist/types/device/index.d.ts +4 -4
- package/package.json +2 -2
- package/dist/es/ai-model/prompt/assertion.mjs +0 -31
- package/dist/es/ai-model/prompt/assertion.mjs.map +0 -1
- package/dist/lib/ai-model/prompt/assertion.js +0 -65
- package/dist/lib/ai-model/prompt/assertion.js.map +0 -1
- package/dist/types/ai-model/prompt/assertion.d.ts +0 -2
|
@@ -24,11 +24,8 @@ var __webpack_require__ = {};
|
|
|
24
24
|
var __webpack_exports__ = {};
|
|
25
25
|
__webpack_require__.r(__webpack_exports__);
|
|
26
26
|
__webpack_require__.d(__webpack_exports__, {
|
|
27
|
-
resizeImageForUiTars: ()=>resizeImageForUiTars,
|
|
28
27
|
uiTarsPlanning: ()=>uiTarsPlanning
|
|
29
28
|
});
|
|
30
|
-
const env_namespaceObject = require("@midscene/shared/env");
|
|
31
|
-
const img_namespaceObject = require("@midscene/shared/img");
|
|
32
29
|
const logger_namespaceObject = require("@midscene/shared/logger");
|
|
33
30
|
const us_keyboard_layout_namespaceObject = require("@midscene/shared/us-keyboard-layout");
|
|
34
31
|
const utils_namespaceObject = require("@midscene/shared/utils");
|
|
@@ -50,14 +47,13 @@ async function uiTarsPlanning(userInstruction, options) {
|
|
|
50
47
|
if (actionContext) instruction = `<high_priority_knowledge>${actionContext}</high_priority_knowledge>\n<user_instruction>${userInstruction}</user_instruction>`;
|
|
51
48
|
const systemPrompt = (0, ui_tars_planning_js_namespaceObject.getUiTarsPlanningPrompt)() + instruction;
|
|
52
49
|
const screenshotBase64 = context.screenshot.base64;
|
|
53
|
-
const imagePayload = await resizeImageForUiTars(screenshotBase64, context.size, uiTarsModelVersion);
|
|
54
50
|
conversationHistory.append({
|
|
55
51
|
role: 'user',
|
|
56
52
|
content: [
|
|
57
53
|
{
|
|
58
54
|
type: 'image_url',
|
|
59
55
|
image_url: {
|
|
60
|
-
url:
|
|
56
|
+
url: screenshotBase64
|
|
61
57
|
}
|
|
62
58
|
}
|
|
63
59
|
]
|
|
@@ -270,29 +266,8 @@ function getPoint(startBox, size) {
|
|
|
270
266
|
y * size.height
|
|
271
267
|
];
|
|
272
268
|
}
|
|
273
|
-
async function resizeImageForUiTars(imageBase64, size, uiTarsVersion) {
|
|
274
|
-
if (uiTarsVersion === env_namespaceObject.UITarsModelVersion.V1_5) {
|
|
275
|
-
debug('ui-tars-v1.5, will check image size', size);
|
|
276
|
-
const currentPixels = size.width * size.height;
|
|
277
|
-
const maxPixels = 12845056;
|
|
278
|
-
if (currentPixels > maxPixels) {
|
|
279
|
-
const resizeFactor = Math.sqrt(maxPixels / currentPixels);
|
|
280
|
-
const newWidth = Math.floor(size.width * resizeFactor);
|
|
281
|
-
const newHeight = Math.floor(size.height * resizeFactor);
|
|
282
|
-
debug('resize image for ui-tars, new width: %s, new height: %s', newWidth, newHeight);
|
|
283
|
-
const resizedImage = await (0, img_namespaceObject.resizeImgBase64)(imageBase64, {
|
|
284
|
-
width: newWidth,
|
|
285
|
-
height: newHeight
|
|
286
|
-
});
|
|
287
|
-
return resizedImage;
|
|
288
|
-
}
|
|
289
|
-
}
|
|
290
|
-
return imageBase64;
|
|
291
|
-
}
|
|
292
|
-
exports.resizeImageForUiTars = __webpack_exports__.resizeImageForUiTars;
|
|
293
269
|
exports.uiTarsPlanning = __webpack_exports__.uiTarsPlanning;
|
|
294
270
|
for(var __rspack_i in __webpack_exports__)if (-1 === [
|
|
295
|
-
"resizeImageForUiTars",
|
|
296
271
|
"uiTarsPlanning"
|
|
297
272
|
].indexOf(__rspack_i)) exports[__rspack_i] = __webpack_exports__[__rspack_i];
|
|
298
273
|
Object.defineProperty(exports, '__esModule', {
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"ai-model/ui-tars-planning.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../src/ai-model/ui-tars-planning.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n PlanningAIResponse,\n PlanningAction,\n Size,\n UIContext,\n} from '@/types';\nimport { type IModelConfig, UITarsModelVersion } from '@midscene/shared/env';\nimport { resizeImgBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { transformHotkeyInput } from '@midscene/shared/us-keyboard-layout';\nimport { assert } from '@midscene/shared/utils';\nimport { actionParser } from '@ui-tars/action-parser';\nimport type { ConversationHistory } from './conversation-history';\nimport { getSummary, getUiTarsPlanningPrompt } from './prompt/ui-tars-planning';\nimport { callAIWithStringResponse } from './service-caller/index';\n\ntype ActionType =\n | 'click'\n | 'left_double'\n | 'right_single'\n | 'drag'\n | 'type'\n | 'hotkey'\n | 'finished'\n | 'scroll'\n | 'wait';\n\nconst debug = getDebug('ui-tars-planning');\nconst bboxSize = 10;\nconst pointToBbox = (\n point: { x: number; y: number },\n width: number,\n height: number,\n): [number, number, number, number] => {\n return [\n Math.round(Math.max(point.x - bboxSize / 2, 0)),\n Math.round(Math.max(point.y - bboxSize / 2, 0)),\n Math.round(Math.min(point.x + bboxSize / 2, width)),\n Math.round(Math.min(point.y + bboxSize / 2, height)),\n ];\n};\n\nexport async function uiTarsPlanning(\n userInstruction: string,\n options: {\n conversationHistory: ConversationHistory;\n context: UIContext;\n modelConfig: IModelConfig;\n actionContext?: string;\n },\n): Promise<PlanningAIResponse> {\n const { conversationHistory, context, modelConfig, actionContext } = options;\n const { uiTarsModelVersion } = modelConfig;\n\n let instruction = userInstruction;\n if (actionContext) {\n instruction = `<high_priority_knowledge>${actionContext}</high_priority_knowledge>\\n<user_instruction>${userInstruction}</user_instruction>`;\n }\n\n const systemPrompt = getUiTarsPlanningPrompt() + instruction;\n\n const screenshotBase64 = context.screenshot.base64;\n const imagePayload = await resizeImageForUiTars(\n screenshotBase64,\n context.size,\n uiTarsModelVersion,\n );\n\n conversationHistory.append({\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n },\n },\n ],\n });\n\n const res = await callAIWithStringResponse(\n [\n {\n role: 'user',\n content: systemPrompt,\n },\n ...conversationHistory.snapshot(),\n ],\n modelConfig,\n );\n const convertedText = convertBboxToCoordinates(res.content);\n\n const { size } = context;\n const { parsed } = actionParser({\n prediction: convertedText,\n factor: [1000, 1000],\n screenContext: {\n width: size.width,\n height: size.height,\n },\n modelVer: uiTarsModelVersion,\n });\n\n debug(\n 'ui-tars modelVer',\n uiTarsModelVersion,\n ', parsed',\n JSON.stringify(parsed),\n );\n\n const transformActions: PlanningAction[] = [];\n const unhandledActions: Array<{ type: string; thought: string }> = [];\n let shouldContinue = true;\n parsed.forEach((action) => {\n const actionType = (action.action_type || '').toLowerCase();\n if (actionType === 'click') {\n assert(action.action_inputs.start_box, 'start_box is required');\n const point = getPoint(action.action_inputs.start_box, size);\n\n const locate = {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: point[0], y: point[1] },\n size.width,\n size.height,\n ),\n };\n\n transformActions.push({\n type: 'Tap',\n param: {\n locate: locate,\n },\n });\n } else if (actionType === 'left_double') {\n assert(action.action_inputs.start_box, 'start_box is required');\n const point = getPoint(action.action_inputs.start_box, size);\n\n const locate = {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: point[0], y: point[1] },\n size.width,\n size.height,\n ),\n };\n\n transformActions.push({\n type: 'DoubleClick',\n param: {\n locate: locate,\n },\n thought: action.thought || '',\n });\n } else if (actionType === 'right_single') {\n assert(action.action_inputs.start_box, 'start_box is required');\n const point = getPoint(action.action_inputs.start_box, size);\n\n const locate = {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: point[0], y: point[1] },\n size.width,\n size.height,\n ),\n };\n\n transformActions.push({\n type: 'RightClick',\n param: {\n locate: locate,\n },\n thought: action.thought || '',\n });\n } else if (actionType === 'drag') {\n assert(action.action_inputs.start_box, 'start_box is required');\n assert(action.action_inputs.end_box, 'end_box is required');\n const startPoint = getPoint(action.action_inputs.start_box, size);\n const endPoint = getPoint(action.action_inputs.end_box, size);\n transformActions.push({\n type: 'DragAndDrop',\n param: {\n from: {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: startPoint[0], y: startPoint[1] },\n size.width,\n size.height,\n ),\n },\n to: {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: endPoint[0], y: endPoint[1] },\n size.width,\n size.height,\n ),\n },\n },\n thought: action.thought || '',\n });\n } else if (actionType === 'type') {\n transformActions.push({\n type: 'Input',\n param: {\n value: action.action_inputs.content,\n },\n thought: action.thought || '',\n });\n } else if (actionType === 'scroll') {\n transformActions.push({\n type: 'Scroll',\n param: {\n direction: action.action_inputs.direction,\n },\n thought: action.thought || '',\n });\n } else if (actionType === 'finished') {\n shouldContinue = false;\n transformActions.push({\n type: 'Finished',\n param: {},\n thought: action.thought || '',\n });\n } else if (actionType === 'hotkey') {\n if (!action.action_inputs.key) {\n console.warn(\n 'No key found in action: hotkey. Will not perform action.',\n );\n } else {\n const keys = transformHotkeyInput(action.action_inputs.key);\n\n transformActions.push({\n type: 'KeyboardPress',\n param: {\n keyName: keys.join('+'),\n },\n thought: action.thought || '',\n });\n }\n } else if (actionType === 'wait') {\n transformActions.push({\n type: 'Sleep',\n param: {\n timeMs: 1000,\n },\n thought: action.thought || '',\n });\n } else if (actionType) {\n // Track unhandled action types\n unhandledActions.push({\n type: actionType,\n thought: action.thought || '',\n });\n debug('Unhandled action type:', actionType, 'thought:', action.thought);\n }\n });\n\n if (transformActions.length === 0) {\n const errorDetails: string[] = [];\n\n // Check if parsing failed\n if (parsed.length === 0) {\n errorDetails.push('Action parser returned no actions');\n\n // Check if response has Thought but no Action\n if (\n res.content.includes('Thought:') &&\n !res.content.includes('Action:')\n ) {\n errorDetails.push(\n 'Response contains \"Thought:\" but missing \"Action:\" line',\n );\n } else {\n errorDetails.push('Response may be malformed or empty');\n }\n }\n\n // Check if we have unhandled action types\n if (unhandledActions.length > 0) {\n const types = unhandledActions.map((a) => a.type).join(', ');\n errorDetails.push(`Unhandled action types: ${types}`);\n }\n\n const errorMessage = [\n 'No actions found in UI-TARS response.',\n ...errorDetails,\n `\\nRaw response: ${res.content}`,\n ].join('\\n');\n\n throw new Error(errorMessage, {\n cause: {\n prediction: res.content,\n parsed,\n unhandledActions,\n convertedText,\n },\n });\n }\n\n debug('transformActions', JSON.stringify(transformActions, null, 2));\n const log = getSummary(res.content);\n\n conversationHistory.append({\n role: 'assistant',\n content: log,\n });\n\n return {\n actions: transformActions,\n log,\n usage: res.usage,\n rawResponse: JSON.stringify(res.content, undefined, 2),\n shouldContinuePlanning: shouldContinue,\n };\n}\n\n/**\n * Converts bounding box notation to coordinate points\n * @param text - The text containing bbox tags to be converted\n * @returns The text with bbox tags replaced by coordinate points\n */\nfunction convertBboxToCoordinates(text: string): string {\n // Match the four numbers after <bbox>\n const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n function replaceMatch(\n match: string,\n x1: string,\n y1: string,\n x2: string,\n y2: string,\n ): string {\n // Convert strings to numbers and calculate center point\n const x1Num = Number.parseInt(x1, 10);\n const y1Num = Number.parseInt(y1, 10);\n const x2Num = Number.parseInt(x2, 10);\n const y2Num = Number.parseInt(y2, 10);\n\n // Use Math.floor to truncate and calculate center point\n const x = Math.floor((x1Num + x2Num) / 2);\n const y = Math.floor((y1Num + y2Num) / 2);\n\n // Return formatted coordinate string\n return `(${x},${y})`;\n }\n\n // Remove [EOS] and replace <bbox> coordinates\n const cleanedText = text.replace(/\\[EOS\\]/g, '');\n return cleanedText.replace(pattern, replaceMatch).trim();\n}\n\nfunction getPoint(startBox: string, size: { width: number; height: number }) {\n const [x, y] = JSON.parse(startBox);\n return [x * size.width, y * size.height];\n}\n\ninterface BaseAction {\n action_type: ActionType;\n action_inputs: Record<string, any>;\n reflection: string | null;\n thought: string | null;\n}\n\ninterface ClickAction extends BaseAction {\n action_type: 'click';\n action_inputs: {\n start_box: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface DragAction extends BaseAction {\n action_type: 'drag';\n action_inputs: {\n start_box: string; // JSON string of [x, y] coordinates\n end_box: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface WaitAction extends BaseAction {\n action_type: 'wait';\n action_inputs: {\n time: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface LeftDoubleAction extends BaseAction {\n action_type: 'left_double';\n action_inputs: {\n start_box: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface RightSingleAction extends BaseAction {\n action_type: 'right_single';\n action_inputs: {\n start_box: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface TypeAction extends BaseAction {\n action_type: 'type';\n action_inputs: {\n content: string;\n };\n}\n\ninterface HotkeyAction extends BaseAction {\n action_type: 'hotkey';\n action_inputs: {\n key: string;\n };\n}\n\ninterface ScrollAction extends BaseAction {\n action_type: 'scroll';\n action_inputs: {\n direction: 'up' | 'down';\n };\n}\n\ninterface FinishedAction extends BaseAction {\n action_type: 'finished';\n action_inputs: Record<string, never>;\n}\n\nexport type Action =\n | ClickAction\n | LeftDoubleAction\n | RightSingleAction\n | DragAction\n | TypeAction\n | HotkeyAction\n | ScrollAction\n | FinishedAction\n | WaitAction;\n\nexport async function resizeImageForUiTars(\n imageBase64: string,\n size: Size,\n uiTarsVersion: UITarsModelVersion | undefined,\n) {\n if (uiTarsVersion === UITarsModelVersion.V1_5) {\n debug('ui-tars-v1.5, will check image size', size);\n const currentPixels = size.width * size.height;\n const maxPixels = 16384 * 28 * 28; //\n if (currentPixels > maxPixels) {\n const resizeFactor = Math.sqrt(maxPixels / currentPixels);\n const newWidth = Math.floor(size.width * resizeFactor);\n const newHeight = Math.floor(size.height * resizeFactor);\n debug(\n 'resize image for ui-tars, new width: %s, new height: %s',\n newWidth,\n newHeight,\n );\n const resizedImage = await resizeImgBase64(imageBase64, {\n width: newWidth,\n height: newHeight,\n });\n return resizedImage;\n }\n }\n return imageBase64;\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","bboxSize","pointToBbox","point","width","height","Math","uiTarsPlanning","userInstruction","options","conversationHistory","context","modelConfig","actionContext","uiTarsModelVersion","instruction","systemPrompt","getUiTarsPlanningPrompt","screenshotBase64","imagePayload","resizeImageForUiTars","res","callAIWithStringResponse","convertedText","convertBboxToCoordinates","size","parsed","actionParser","JSON","transformActions","unhandledActions","shouldContinue","action","actionType","assert","getPoint","locate","startPoint","endPoint","keys","transformHotkeyInput","console","errorDetails","types","a","errorMessage","Error","log","getSummary","undefined","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","y","cleanedText","startBox","imageBase64","uiTarsVersion","UITarsModelVersion","currentPixels","maxPixels","resizeFactor","newWidth","newHeight","resizedImage","resizeImgBase64"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;ACqBA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AACvB,MAAMC,WAAW;AACjB,MAAMC,cAAc,CAClBC,OACAC,OACAC,SAEO;QACLC,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGG;QAC5CE,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGI;KAC7C;AAGI,eAAeE,eACpBC,eAAuB,EACvBC,OAKC;IAED,MAAM,EAAEC,mBAAmB,EAAEC,OAAO,EAAEC,WAAW,EAAEC,aAAa,EAAE,GAAGJ;IACrE,MAAM,EAAEK,kBAAkB,EAAE,GAAGF;IAE/B,IAAIG,cAAcP;IAClB,IAAIK,eACFE,cAAc,CAAC,yBAAyB,EAAEF,cAAc,8CAA8C,EAAEL,gBAAgB,mBAAmB,CAAC;IAG9I,MAAMQ,eAAeC,AAAAA,IAAAA,oCAAAA,uBAAAA,AAAAA,MAA4BF;IAEjD,MAAMG,mBAAmBP,QAAQ,UAAU,CAAC,MAAM;IAClD,MAAMQ,eAAe,MAAMC,qBACzBF,kBACAP,QAAQ,IAAI,EACZG;IAGFJ,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKS;gBACP;YACF;SACD;IACH;IAEA,MAAME,MAAM,MAAMC,AAAAA,IAAAA,yBAAAA,wBAAAA,AAAAA,EAChB;QACE;YACE,MAAM;YACN,SAASN;QACX;WACGN,oBAAoB,QAAQ;KAChC,EACDE;IAEF,MAAMW,gBAAgBC,yBAAyBH,IAAI,OAAO;IAE1D,MAAM,EAAEI,IAAI,EAAE,GAAGd;IACjB,MAAM,EAAEe,MAAM,EAAE,GAAGC,AAAAA,IAAAA,8BAAAA,YAAAA,AAAAA,EAAa;QAC9B,YAAYJ;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOE,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUX;IACZ;IAEAf,MACE,oBACAe,oBACA,YACAc,KAAK,SAAS,CAACF;IAGjB,MAAMG,mBAAqC,EAAE;IAC7C,MAAMC,mBAA6D,EAAE;IACrE,IAAIC,iBAAiB;IACrBL,OAAO,OAAO,CAAC,CAACM;QACd,MAAMC,aAAcD,AAAAA,CAAAA,OAAO,WAAW,IAAI,EAAC,EAAG,WAAW;QACzD,IAAIC,AAAe,YAAfA,YAAwB;YAC1BC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM7B,QAAQgC,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAEvD,MAAMW,SAAS;gBACb,QAAQJ,OAAO,OAAO,IAAI;gBAC1B,MAAM9B,YACJ;oBAAE,GAAGC,KAAK,CAAC,EAAE;oBAAE,GAAGA,KAAK,CAAC,EAAE;gBAAC,GAC3BsB,KAAK,KAAK,EACVA,KAAK,MAAM;YAEf;YAEAI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQO;gBACV;YACF;QACF,OAAO,IAAIH,AAAe,kBAAfA,YAA8B;YACvCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM7B,QAAQgC,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAEvD,MAAMW,SAAS;gBACb,QAAQJ,OAAO,OAAO,IAAI;gBAC1B,MAAM9B,YACJ;oBAAE,GAAGC,KAAK,CAAC,EAAE;oBAAE,GAAGA,KAAK,CAAC,EAAE;gBAAC,GAC3BsB,KAAK,KAAK,EACVA,KAAK,MAAM;YAEf;YAEAI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQO;gBACV;gBACA,SAASJ,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,mBAAfA,YAA+B;YACxCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM7B,QAAQgC,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAEvD,MAAMW,SAAS;gBACb,QAAQJ,OAAO,OAAO,IAAI;gBAC1B,MAAM9B,YACJ;oBAAE,GAAGC,KAAK,CAAC,EAAE;oBAAE,GAAGA,KAAK,CAAC,EAAE;gBAAC,GAC3BsB,KAAK,KAAK,EACVA,KAAK,MAAM;YAEf;YAEAI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQO;gBACV;gBACA,SAASJ,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YAAuB;YAChCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCE,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMK,aAAaF,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAC5D,MAAMa,WAAWH,SAASH,OAAO,aAAa,CAAC,OAAO,EAAEP;YACxDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,MAAM;wBACJ,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM9B,YACJ;4BAAE,GAAGmC,UAAU,CAAC,EAAE;4BAAE,GAAGA,UAAU,CAAC,EAAE;wBAAC,GACrCZ,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;oBACA,IAAI;wBACF,QAAQO,OAAO,OAAO,IAAI;wBAC1B,MAAM9B,YACJ;4BAAE,GAAGoC,QAAQ,CAAC,EAAE;4BAAE,GAAGA,QAAQ,CAAC,EAAE;wBAAC,GACjCb,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASO,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,OAAOG,OAAO,aAAa,CAAC,OAAO;YACrC;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,aAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWG,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,eAAfA,YAA2B;YACpCF,iBAAiB;YACjBF,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO,CAAC;gBACR,SAASG,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,aAAfA,YACT,IAAKD,OAAO,aAAa,CAAC,GAAG,EAItB;YACL,MAAMO,OAAOC,AAAAA,IAAAA,mCAAAA,oBAAAA,AAAAA,EAAqBR,OAAO,aAAa,CAAC,GAAG;YAE1DH,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,SAASU,KAAK,IAAI,CAAC;gBACrB;gBACA,SAASP,OAAO,OAAO,IAAI;YAC7B;QACF,OAbES,QAAQ,IAAI,CACV;aAaC,IAAIR,AAAe,WAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,SAASG,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,YAAY;YAErBH,iBAAiB,IAAI,CAAC;gBACpB,MAAMG;gBACN,SAASD,OAAO,OAAO,IAAI;YAC7B;YACAjC,MAAM,0BAA0BkC,YAAY,YAAYD,OAAO,OAAO;QACxE;IACF;IAEA,IAAIH,AAA4B,MAA5BA,iBAAiB,MAAM,EAAQ;QACjC,MAAMa,eAAyB,EAAE;QAGjC,IAAIhB,AAAkB,MAAlBA,OAAO,MAAM,EAAQ;YACvBgB,aAAa,IAAI,CAAC;YAGlB,IACErB,IAAI,OAAO,CAAC,QAAQ,CAAC,eACrB,CAACA,IAAI,OAAO,CAAC,QAAQ,CAAC,YAEtBqB,aAAa,IAAI,CACf;iBAGFA,aAAa,IAAI,CAAC;QAEtB;QAGA,IAAIZ,iBAAiB,MAAM,GAAG,GAAG;YAC/B,MAAMa,QAAQb,iBAAiB,GAAG,CAAC,CAACc,IAAMA,EAAE,IAAI,EAAE,IAAI,CAAC;YACvDF,aAAa,IAAI,CAAC,CAAC,wBAAwB,EAAEC,OAAO;QACtD;QAEA,MAAME,eAAe;YACnB;eACGH;YACH,CAAC,gBAAgB,EAAErB,IAAI,OAAO,EAAE;SACjC,CAAC,IAAI,CAAC;QAEP,MAAM,IAAIyB,MAAMD,cAAc;YAC5B,OAAO;gBACL,YAAYxB,IAAI,OAAO;gBACvBK;gBACAI;gBACAP;YACF;QACF;IACF;IAEAxB,MAAM,oBAAoB6B,KAAK,SAAS,CAACC,kBAAkB,MAAM;IACjE,MAAMkB,MAAMC,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAW3B,IAAI,OAAO;IAElCX,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAASqC;IACX;IAEA,OAAO;QACL,SAASlB;QACTkB;QACA,OAAO1B,IAAI,KAAK;QAChB,aAAaO,KAAK,SAAS,CAACP,IAAI,OAAO,EAAE4B,QAAW;QACpD,wBAAwBlB;IAC1B;AACF;AAOA,SAASP,yBAAyB0B,IAAY;IAE5C,MAAMC,UAAU;IAEhB,SAASC,aACPC,KAAa,EACbC,EAAU,EACVC,EAAU,EACVC,EAAU,EACVC,EAAU;QAGV,MAAMC,QAAQC,OAAO,QAAQ,CAACL,IAAI;QAClC,MAAMM,QAAQD,OAAO,QAAQ,CAACJ,IAAI;QAClC,MAAMM,QAAQF,OAAO,QAAQ,CAACH,IAAI;QAClC,MAAMM,QAAQH,OAAO,QAAQ,CAACF,IAAI;QAGlC,MAAMM,IAAIzD,KAAK,KAAK,CAAEoD,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMG,IAAI1D,KAAK,KAAK,CAAEsD,AAAAA,CAAAA,QAAQE,KAAI,IAAK;QAGvC,OAAO,CAAC,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC,CAAC;IACtB;IAGA,MAAMC,cAAcf,KAAK,OAAO,CAAC,YAAY;IAC7C,OAAOe,YAAY,OAAO,CAACd,SAASC,cAAc,IAAI;AACxD;AAEA,SAASjB,SAAS+B,QAAgB,EAAEzC,IAAuC;IACzE,MAAM,CAACsC,GAAGC,EAAE,GAAGpC,KAAK,KAAK,CAACsC;IAC1B,OAAO;QAACH,IAAItC,KAAK,KAAK;QAAEuC,IAAIvC,KAAK,MAAM;KAAC;AAC1C;AAkFO,eAAeL,qBACpB+C,WAAmB,EACnB1C,IAAU,EACV2C,aAA6C;IAE7C,IAAIA,kBAAkBC,oBAAAA,kBAAAA,CAAAA,IAAuB,EAAE;QAC7CtE,MAAM,uCAAuC0B;QAC7C,MAAM6C,gBAAgB7C,KAAK,KAAK,GAAGA,KAAK,MAAM;QAC9C,MAAM8C,YAAY;QAClB,IAAID,gBAAgBC,WAAW;YAC7B,MAAMC,eAAelE,KAAK,IAAI,CAACiE,YAAYD;YAC3C,MAAMG,WAAWnE,KAAK,KAAK,CAACmB,KAAK,KAAK,GAAG+C;YACzC,MAAME,YAAYpE,KAAK,KAAK,CAACmB,KAAK,MAAM,GAAG+C;YAC3CzE,MACE,2DACA0E,UACAC;YAEF,MAAMC,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,eAAAA,AAAAA,EAAgBT,aAAa;gBACtD,OAAOM;gBACP,QAAQC;YACV;YACA,OAAOC;QACT;IACF;IACA,OAAOR;AACT"}
|
|
1
|
+
{"version":3,"file":"ai-model/ui-tars-planning.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../src/ai-model/ui-tars-planning.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n PlanningAIResponse,\n PlanningAction,\n Size,\n UIContext,\n} from '@/types';\nimport { type IModelConfig, UITarsModelVersion } from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { transformHotkeyInput } from '@midscene/shared/us-keyboard-layout';\nimport { assert } from '@midscene/shared/utils';\nimport { actionParser } from '@ui-tars/action-parser';\nimport type { ConversationHistory } from './conversation-history';\nimport { getSummary, getUiTarsPlanningPrompt } from './prompt/ui-tars-planning';\nimport { callAIWithStringResponse } from './service-caller/index';\n\ntype ActionType =\n | 'click'\n | 'left_double'\n | 'right_single'\n | 'drag'\n | 'type'\n | 'hotkey'\n | 'finished'\n | 'scroll'\n | 'wait';\n\nconst debug = getDebug('ui-tars-planning');\nconst bboxSize = 10;\nconst pointToBbox = (\n point: { x: number; y: number },\n width: number,\n height: number,\n): [number, number, number, number] => {\n return [\n Math.round(Math.max(point.x - bboxSize / 2, 0)),\n Math.round(Math.max(point.y - bboxSize / 2, 0)),\n Math.round(Math.min(point.x + bboxSize / 2, width)),\n Math.round(Math.min(point.y + bboxSize / 2, height)),\n ];\n};\n\nexport async function uiTarsPlanning(\n userInstruction: string,\n options: {\n conversationHistory: ConversationHistory;\n context: UIContext;\n modelConfig: IModelConfig;\n actionContext?: string;\n },\n): Promise<PlanningAIResponse> {\n const { conversationHistory, context, modelConfig, actionContext } = options;\n const { uiTarsModelVersion } = modelConfig;\n\n let instruction = userInstruction;\n if (actionContext) {\n instruction = `<high_priority_knowledge>${actionContext}</high_priority_knowledge>\\n<user_instruction>${userInstruction}</user_instruction>`;\n }\n\n const systemPrompt = getUiTarsPlanningPrompt() + instruction;\n\n const screenshotBase64 = context.screenshot.base64;\n\n conversationHistory.append({\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: screenshotBase64,\n },\n },\n ],\n });\n\n const res = await callAIWithStringResponse(\n [\n {\n role: 'user',\n content: systemPrompt,\n },\n ...conversationHistory.snapshot(),\n ],\n modelConfig,\n );\n const convertedText = convertBboxToCoordinates(res.content);\n\n const { size } = context;\n const { parsed } = actionParser({\n prediction: convertedText,\n factor: [1000, 1000],\n screenContext: {\n width: size.width,\n height: size.height,\n },\n modelVer: uiTarsModelVersion,\n });\n\n debug(\n 'ui-tars modelVer',\n uiTarsModelVersion,\n ', parsed',\n JSON.stringify(parsed),\n );\n\n const transformActions: PlanningAction[] = [];\n const unhandledActions: Array<{ type: string; thought: string }> = [];\n let shouldContinue = true;\n parsed.forEach((action) => {\n const actionType = (action.action_type || '').toLowerCase();\n if (actionType === 'click') {\n assert(action.action_inputs.start_box, 'start_box is required');\n const point = getPoint(action.action_inputs.start_box, size);\n\n const locate = {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: point[0], y: point[1] },\n size.width,\n size.height,\n ),\n };\n\n transformActions.push({\n type: 'Tap',\n param: {\n locate: locate,\n },\n });\n } else if (actionType === 'left_double') {\n assert(action.action_inputs.start_box, 'start_box is required');\n const point = getPoint(action.action_inputs.start_box, size);\n\n const locate = {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: point[0], y: point[1] },\n size.width,\n size.height,\n ),\n };\n\n transformActions.push({\n type: 'DoubleClick',\n param: {\n locate: locate,\n },\n thought: action.thought || '',\n });\n } else if (actionType === 'right_single') {\n assert(action.action_inputs.start_box, 'start_box is required');\n const point = getPoint(action.action_inputs.start_box, size);\n\n const locate = {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: point[0], y: point[1] },\n size.width,\n size.height,\n ),\n };\n\n transformActions.push({\n type: 'RightClick',\n param: {\n locate: locate,\n },\n thought: action.thought || '',\n });\n } else if (actionType === 'drag') {\n assert(action.action_inputs.start_box, 'start_box is required');\n assert(action.action_inputs.end_box, 'end_box is required');\n const startPoint = getPoint(action.action_inputs.start_box, size);\n const endPoint = getPoint(action.action_inputs.end_box, size);\n transformActions.push({\n type: 'DragAndDrop',\n param: {\n from: {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: startPoint[0], y: startPoint[1] },\n size.width,\n size.height,\n ),\n },\n to: {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: endPoint[0], y: endPoint[1] },\n size.width,\n size.height,\n ),\n },\n },\n thought: action.thought || '',\n });\n } else if (actionType === 'type') {\n transformActions.push({\n type: 'Input',\n param: {\n value: action.action_inputs.content,\n },\n thought: action.thought || '',\n });\n } else if (actionType === 'scroll') {\n transformActions.push({\n type: 'Scroll',\n param: {\n direction: action.action_inputs.direction,\n },\n thought: action.thought || '',\n });\n } else if (actionType === 'finished') {\n shouldContinue = false;\n transformActions.push({\n type: 'Finished',\n param: {},\n thought: action.thought || '',\n });\n } else if (actionType === 'hotkey') {\n if (!action.action_inputs.key) {\n console.warn(\n 'No key found in action: hotkey. Will not perform action.',\n );\n } else {\n const keys = transformHotkeyInput(action.action_inputs.key);\n\n transformActions.push({\n type: 'KeyboardPress',\n param: {\n keyName: keys.join('+'),\n },\n thought: action.thought || '',\n });\n }\n } else if (actionType === 'wait') {\n transformActions.push({\n type: 'Sleep',\n param: {\n timeMs: 1000,\n },\n thought: action.thought || '',\n });\n } else if (actionType) {\n // Track unhandled action types\n unhandledActions.push({\n type: actionType,\n thought: action.thought || '',\n });\n debug('Unhandled action type:', actionType, 'thought:', action.thought);\n }\n });\n\n if (transformActions.length === 0) {\n const errorDetails: string[] = [];\n\n // Check if parsing failed\n if (parsed.length === 0) {\n errorDetails.push('Action parser returned no actions');\n\n // Check if response has Thought but no Action\n if (\n res.content.includes('Thought:') &&\n !res.content.includes('Action:')\n ) {\n errorDetails.push(\n 'Response contains \"Thought:\" but missing \"Action:\" line',\n );\n } else {\n errorDetails.push('Response may be malformed or empty');\n }\n }\n\n // Check if we have unhandled action types\n if (unhandledActions.length > 0) {\n const types = unhandledActions.map((a) => a.type).join(', ');\n errorDetails.push(`Unhandled action types: ${types}`);\n }\n\n const errorMessage = [\n 'No actions found in UI-TARS response.',\n ...errorDetails,\n `\\nRaw response: ${res.content}`,\n ].join('\\n');\n\n throw new Error(errorMessage, {\n cause: {\n prediction: res.content,\n parsed,\n unhandledActions,\n convertedText,\n },\n });\n }\n\n debug('transformActions', JSON.stringify(transformActions, null, 2));\n const log = getSummary(res.content);\n\n conversationHistory.append({\n role: 'assistant',\n content: log,\n });\n\n return {\n actions: transformActions,\n log,\n usage: res.usage,\n rawResponse: JSON.stringify(res.content, undefined, 2),\n shouldContinuePlanning: shouldContinue,\n };\n}\n\n/**\n * Converts bounding box notation to coordinate points\n * @param text - The text containing bbox tags to be converted\n * @returns The text with bbox tags replaced by coordinate points\n */\nfunction convertBboxToCoordinates(text: string): string {\n // Match the four numbers after <bbox>\n const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n function replaceMatch(\n match: string,\n x1: string,\n y1: string,\n x2: string,\n y2: string,\n ): string {\n // Convert strings to numbers and calculate center point\n const x1Num = Number.parseInt(x1, 10);\n const y1Num = Number.parseInt(y1, 10);\n const x2Num = Number.parseInt(x2, 10);\n const y2Num = Number.parseInt(y2, 10);\n\n // Use Math.floor to truncate and calculate center point\n const x = Math.floor((x1Num + x2Num) / 2);\n const y = Math.floor((y1Num + y2Num) / 2);\n\n // Return formatted coordinate string\n return `(${x},${y})`;\n }\n\n // Remove [EOS] and replace <bbox> coordinates\n const cleanedText = text.replace(/\\[EOS\\]/g, '');\n return cleanedText.replace(pattern, replaceMatch).trim();\n}\n\nfunction getPoint(startBox: string, size: { width: number; height: number }) {\n const [x, y] = JSON.parse(startBox);\n return [x * size.width, y * size.height];\n}\n\ninterface BaseAction {\n action_type: ActionType;\n action_inputs: Record<string, any>;\n reflection: string | null;\n thought: string | null;\n}\n\ninterface ClickAction extends BaseAction {\n action_type: 'click';\n action_inputs: {\n start_box: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface DragAction extends BaseAction {\n action_type: 'drag';\n action_inputs: {\n start_box: string; // JSON string of [x, y] coordinates\n end_box: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface WaitAction extends BaseAction {\n action_type: 'wait';\n action_inputs: {\n time: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface LeftDoubleAction extends BaseAction {\n action_type: 'left_double';\n action_inputs: {\n start_box: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface RightSingleAction extends BaseAction {\n action_type: 'right_single';\n action_inputs: {\n start_box: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface TypeAction extends BaseAction {\n action_type: 'type';\n action_inputs: {\n content: string;\n };\n}\n\ninterface HotkeyAction extends BaseAction {\n action_type: 'hotkey';\n action_inputs: {\n key: string;\n };\n}\n\ninterface ScrollAction extends BaseAction {\n action_type: 'scroll';\n action_inputs: {\n direction: 'up' | 'down';\n };\n}\n\ninterface FinishedAction extends BaseAction {\n action_type: 'finished';\n action_inputs: Record<string, never>;\n}\n\nexport type Action =\n | ClickAction\n | LeftDoubleAction\n | RightSingleAction\n | DragAction\n | TypeAction\n | HotkeyAction\n | ScrollAction\n | FinishedAction\n | WaitAction;\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","bboxSize","pointToBbox","point","width","height","Math","uiTarsPlanning","userInstruction","options","conversationHistory","context","modelConfig","actionContext","uiTarsModelVersion","instruction","systemPrompt","getUiTarsPlanningPrompt","screenshotBase64","res","callAIWithStringResponse","convertedText","convertBboxToCoordinates","size","parsed","actionParser","JSON","transformActions","unhandledActions","shouldContinue","action","actionType","assert","getPoint","locate","startPoint","endPoint","keys","transformHotkeyInput","console","errorDetails","types","a","errorMessage","Error","log","getSummary","undefined","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","y","cleanedText","startBox"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;ACoBA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AACvB,MAAMC,WAAW;AACjB,MAAMC,cAAc,CAClBC,OACAC,OACAC,SAEO;QACLC,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGG;QAC5CE,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGI;KAC7C;AAGI,eAAeE,eACpBC,eAAuB,EACvBC,OAKC;IAED,MAAM,EAAEC,mBAAmB,EAAEC,OAAO,EAAEC,WAAW,EAAEC,aAAa,EAAE,GAAGJ;IACrE,MAAM,EAAEK,kBAAkB,EAAE,GAAGF;IAE/B,IAAIG,cAAcP;IAClB,IAAIK,eACFE,cAAc,CAAC,yBAAyB,EAAEF,cAAc,8CAA8C,EAAEL,gBAAgB,mBAAmB,CAAC;IAG9I,MAAMQ,eAAeC,AAAAA,IAAAA,oCAAAA,uBAAAA,AAAAA,MAA4BF;IAEjD,MAAMG,mBAAmBP,QAAQ,UAAU,CAAC,MAAM;IAElDD,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKQ;gBACP;YACF;SACD;IACH;IAEA,MAAMC,MAAM,MAAMC,AAAAA,IAAAA,yBAAAA,wBAAAA,AAAAA,EAChB;QACE;YACE,MAAM;YACN,SAASJ;QACX;WACGN,oBAAoB,QAAQ;KAChC,EACDE;IAEF,MAAMS,gBAAgBC,yBAAyBH,IAAI,OAAO;IAE1D,MAAM,EAAEI,IAAI,EAAE,GAAGZ;IACjB,MAAM,EAAEa,MAAM,EAAE,GAAGC,AAAAA,IAAAA,8BAAAA,YAAAA,AAAAA,EAAa;QAC9B,YAAYJ;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOE,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUT;IACZ;IAEAf,MACE,oBACAe,oBACA,YACAY,KAAK,SAAS,CAACF;IAGjB,MAAMG,mBAAqC,EAAE;IAC7C,MAAMC,mBAA6D,EAAE;IACrE,IAAIC,iBAAiB;IACrBL,OAAO,OAAO,CAAC,CAACM;QACd,MAAMC,aAAcD,AAAAA,CAAAA,OAAO,WAAW,IAAI,EAAC,EAAG,WAAW;QACzD,IAAIC,AAAe,YAAfA,YAAwB;YAC1BC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM3B,QAAQ8B,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAEvD,MAAMW,SAAS;gBACb,QAAQJ,OAAO,OAAO,IAAI;gBAC1B,MAAM5B,YACJ;oBAAE,GAAGC,KAAK,CAAC,EAAE;oBAAE,GAAGA,KAAK,CAAC,EAAE;gBAAC,GAC3BoB,KAAK,KAAK,EACVA,KAAK,MAAM;YAEf;YAEAI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQO;gBACV;YACF;QACF,OAAO,IAAIH,AAAe,kBAAfA,YAA8B;YACvCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM3B,QAAQ8B,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAEvD,MAAMW,SAAS;gBACb,QAAQJ,OAAO,OAAO,IAAI;gBAC1B,MAAM5B,YACJ;oBAAE,GAAGC,KAAK,CAAC,EAAE;oBAAE,GAAGA,KAAK,CAAC,EAAE;gBAAC,GAC3BoB,KAAK,KAAK,EACVA,KAAK,MAAM;YAEf;YAEAI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQO;gBACV;gBACA,SAASJ,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,mBAAfA,YAA+B;YACxCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM3B,QAAQ8B,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAEvD,MAAMW,SAAS;gBACb,QAAQJ,OAAO,OAAO,IAAI;gBAC1B,MAAM5B,YACJ;oBAAE,GAAGC,KAAK,CAAC,EAAE;oBAAE,GAAGA,KAAK,CAAC,EAAE;gBAAC,GAC3BoB,KAAK,KAAK,EACVA,KAAK,MAAM;YAEf;YAEAI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQO;gBACV;gBACA,SAASJ,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YAAuB;YAChCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCE,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMK,aAAaF,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAC5D,MAAMa,WAAWH,SAASH,OAAO,aAAa,CAAC,OAAO,EAAEP;YACxDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,MAAM;wBACJ,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM5B,YACJ;4BAAE,GAAGiC,UAAU,CAAC,EAAE;4BAAE,GAAGA,UAAU,CAAC,EAAE;wBAAC,GACrCZ,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;oBACA,IAAI;wBACF,QAAQO,OAAO,OAAO,IAAI;wBAC1B,MAAM5B,YACJ;4BAAE,GAAGkC,QAAQ,CAAC,EAAE;4BAAE,GAAGA,QAAQ,CAAC,EAAE;wBAAC,GACjCb,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASO,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,OAAOG,OAAO,aAAa,CAAC,OAAO;YACrC;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,aAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWG,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,eAAfA,YAA2B;YACpCF,iBAAiB;YACjBF,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO,CAAC;gBACR,SAASG,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,aAAfA,YACT,IAAKD,OAAO,aAAa,CAAC,GAAG,EAItB;YACL,MAAMO,OAAOC,AAAAA,IAAAA,mCAAAA,oBAAAA,AAAAA,EAAqBR,OAAO,aAAa,CAAC,GAAG;YAE1DH,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,SAASU,KAAK,IAAI,CAAC;gBACrB;gBACA,SAASP,OAAO,OAAO,IAAI;YAC7B;QACF,OAbES,QAAQ,IAAI,CACV;aAaC,IAAIR,AAAe,WAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,SAASG,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,YAAY;YAErBH,iBAAiB,IAAI,CAAC;gBACpB,MAAMG;gBACN,SAASD,OAAO,OAAO,IAAI;YAC7B;YACA/B,MAAM,0BAA0BgC,YAAY,YAAYD,OAAO,OAAO;QACxE;IACF;IAEA,IAAIH,AAA4B,MAA5BA,iBAAiB,MAAM,EAAQ;QACjC,MAAMa,eAAyB,EAAE;QAGjC,IAAIhB,AAAkB,MAAlBA,OAAO,MAAM,EAAQ;YACvBgB,aAAa,IAAI,CAAC;YAGlB,IACErB,IAAI,OAAO,CAAC,QAAQ,CAAC,eACrB,CAACA,IAAI,OAAO,CAAC,QAAQ,CAAC,YAEtBqB,aAAa,IAAI,CACf;iBAGFA,aAAa,IAAI,CAAC;QAEtB;QAGA,IAAIZ,iBAAiB,MAAM,GAAG,GAAG;YAC/B,MAAMa,QAAQb,iBAAiB,GAAG,CAAC,CAACc,IAAMA,EAAE,IAAI,EAAE,IAAI,CAAC;YACvDF,aAAa,IAAI,CAAC,CAAC,wBAAwB,EAAEC,OAAO;QACtD;QAEA,MAAME,eAAe;YACnB;eACGH;YACH,CAAC,gBAAgB,EAAErB,IAAI,OAAO,EAAE;SACjC,CAAC,IAAI,CAAC;QAEP,MAAM,IAAIyB,MAAMD,cAAc;YAC5B,OAAO;gBACL,YAAYxB,IAAI,OAAO;gBACvBK;gBACAI;gBACAP;YACF;QACF;IACF;IAEAtB,MAAM,oBAAoB2B,KAAK,SAAS,CAACC,kBAAkB,MAAM;IACjE,MAAMkB,MAAMC,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAW3B,IAAI,OAAO;IAElCT,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAASmC;IACX;IAEA,OAAO;QACL,SAASlB;QACTkB;QACA,OAAO1B,IAAI,KAAK;QAChB,aAAaO,KAAK,SAAS,CAACP,IAAI,OAAO,EAAE4B,QAAW;QACpD,wBAAwBlB;IAC1B;AACF;AAOA,SAASP,yBAAyB0B,IAAY;IAE5C,MAAMC,UAAU;IAEhB,SAASC,aACPC,KAAa,EACbC,EAAU,EACVC,EAAU,EACVC,EAAU,EACVC,EAAU;QAGV,MAAMC,QAAQC,OAAO,QAAQ,CAACL,IAAI;QAClC,MAAMM,QAAQD,OAAO,QAAQ,CAACJ,IAAI;QAClC,MAAMM,QAAQF,OAAO,QAAQ,CAACH,IAAI;QAClC,MAAMM,QAAQH,OAAO,QAAQ,CAACF,IAAI;QAGlC,MAAMM,IAAIvD,KAAK,KAAK,CAAEkD,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMG,IAAIxD,KAAK,KAAK,CAAEoD,AAAAA,CAAAA,QAAQE,KAAI,IAAK;QAGvC,OAAO,CAAC,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC,CAAC;IACtB;IAGA,MAAMC,cAAcf,KAAK,OAAO,CAAC,YAAY;IAC7C,OAAOe,YAAY,OAAO,CAACd,SAASC,cAAc,IAAI;AACxD;AAEA,SAASjB,SAAS+B,QAAgB,EAAEzC,IAAuC;IACzE,MAAM,CAACsC,GAAGC,EAAE,GAAGpC,KAAK,KAAK,CAACsC;IAC1B,OAAO;QAACH,IAAItC,KAAK,KAAK;QAAEuC,IAAIvC,KAAK,MAAM;KAAC;AAC1C"}
|
package/dist/lib/common.js
CHANGED
|
@@ -51,6 +51,7 @@ __webpack_require__.d(__webpack_exports__, {
|
|
|
51
51
|
adaptBbox: ()=>adaptBbox
|
|
52
52
|
});
|
|
53
53
|
const utils_namespaceObject = require("@midscene/shared/utils");
|
|
54
|
+
const util_js_namespaceObject = require("./ai-model/auto-glm/util.js");
|
|
54
55
|
const constants_namespaceObject = require("@midscene/shared/constants");
|
|
55
56
|
const extractor_namespaceObject = require("@midscene/shared/extractor");
|
|
56
57
|
const img_namespaceObject = require("@midscene/shared/img");
|
|
@@ -71,12 +72,12 @@ function pointToBbox(x, y, bboxSize = defaultBboxSize) {
|
|
|
71
72
|
y2
|
|
72
73
|
];
|
|
73
74
|
}
|
|
74
|
-
function fillBboxParam(locate, width, height, rightLimit, bottomLimit,
|
|
75
|
+
function fillBboxParam(locate, width, height, rightLimit, bottomLimit, modelFamily) {
|
|
75
76
|
if (locate.bbox_2d && !locate?.bbox) {
|
|
76
77
|
locate.bbox = locate.bbox_2d;
|
|
77
78
|
delete locate.bbox_2d;
|
|
78
79
|
}
|
|
79
|
-
if (locate?.bbox) locate.bbox = adaptBbox(locate.bbox, width, height, rightLimit, bottomLimit,
|
|
80
|
+
if (locate?.bbox) locate.bbox = adaptBbox(locate.bbox, width, height, rightLimit, bottomLimit, modelFamily);
|
|
80
81
|
return locate;
|
|
81
82
|
}
|
|
82
83
|
function adaptQwen2_5Bbox(bbox) {
|
|
@@ -143,7 +144,7 @@ function normalizeBboxInput(bbox) {
|
|
|
143
144
|
}
|
|
144
145
|
return bbox;
|
|
145
146
|
}
|
|
146
|
-
function adaptBbox(bbox, width, height, rightLimit, bottomLimit,
|
|
147
|
+
function adaptBbox(bbox, width, height, rightLimit, bottomLimit, modelFamily) {
|
|
147
148
|
const normalizedBbox = normalizeBboxInput(bbox);
|
|
148
149
|
let result = [
|
|
149
150
|
0,
|
|
@@ -151,7 +152,7 @@ function adaptBbox(bbox, width, height, rightLimit, bottomLimit, vlMode) {
|
|
|
151
152
|
0,
|
|
152
153
|
0
|
|
153
154
|
];
|
|
154
|
-
result = 'doubao-vision' ===
|
|
155
|
+
result = 'doubao-vision' === modelFamily || (0, util_js_namespaceObject.isUITars)(modelFamily) ? adaptDoubaoBbox(normalizedBbox, width, height) : 'gemini' === modelFamily ? adaptGeminiBbox(normalizedBbox, width, height) : 'qwen2.5-vl' === modelFamily ? adaptQwen2_5Bbox(normalizedBbox) : normalized01000(normalizedBbox, width, height);
|
|
155
156
|
result[2] = Math.min(result[2], rightLimit);
|
|
156
157
|
result[3] = Math.min(result[3], bottomLimit);
|
|
157
158
|
return result;
|
|
@@ -176,9 +177,9 @@ function adaptGeminiBbox(bbox, width, height) {
|
|
|
176
177
|
bottom
|
|
177
178
|
];
|
|
178
179
|
}
|
|
179
|
-
function adaptBboxToRect(bbox, width, height, offsetX = 0, offsetY = 0, rightLimit = width, bottomLimit = height,
|
|
180
|
-
debugInspectUtils('adaptBboxToRect', bbox, width, height, 'offset', offsetX, offsetY, 'limit', rightLimit, bottomLimit, '
|
|
181
|
-
const [left, top, right, bottom] = adaptBbox(bbox, width, height, rightLimit, bottomLimit,
|
|
180
|
+
function adaptBboxToRect(bbox, width, height, offsetX = 0, offsetY = 0, rightLimit = width, bottomLimit = height, modelFamily) {
|
|
181
|
+
debugInspectUtils('adaptBboxToRect', bbox, width, height, 'offset', offsetX, offsetY, 'limit', rightLimit, bottomLimit, 'modelFamily', modelFamily);
|
|
182
|
+
const [left, top, right, bottom] = adaptBbox(bbox, width, height, rightLimit, bottomLimit, modelFamily);
|
|
182
183
|
const rectLeft = left;
|
|
183
184
|
const rectTop = top;
|
|
184
185
|
let rectWidth = right - left;
|
|
@@ -208,9 +209,9 @@ function mergeRects(rects) {
|
|
|
208
209
|
height: maxBottom - minTop
|
|
209
210
|
};
|
|
210
211
|
}
|
|
211
|
-
function expandSearchArea(rect, screenSize,
|
|
212
|
+
function expandSearchArea(rect, screenSize, modelFamily) {
|
|
212
213
|
let minEdgeSize = 500;
|
|
213
|
-
if ('qwen3-vl' ===
|
|
214
|
+
if ('qwen3-vl' === modelFamily) minEdgeSize = 1200;
|
|
214
215
|
const defaultPadding = 160;
|
|
215
216
|
const paddingSizeHorizontal = rect.width < minEdgeSize ? Math.ceil((minEdgeSize - rect.width) / 2) : defaultPadding;
|
|
216
217
|
const paddingSizeVertical = rect.height < minEdgeSize ? Math.ceil((minEdgeSize - rect.height) / 2) : defaultPadding;
|
package/dist/lib/common.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"common.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../src/common.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n BaseElement,\n DeviceAction,\n ElementTreeNode,\n MidsceneYamlFlowItem,\n PlanningAction,\n Rect,\n Size,\n} from '@/types';\nimport { assert, isPlainObject } 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\nconst defaultBboxSize = 20; // must be even number\nconst debugInspectUtils = getDebug('ai:common');\ntype AdaptBboxInput = number[] | string[] | string | (number[] | string[])[];\n\n/**\n * Convert a point coordinate [0, 1000] to a small bbox [0, 1000]\n * Creates a small bbox around the center point in the same coordinate space\n *\n * @param x - X coordinate in [0, 1000] range\n * @param y - Y coordinate in [0, 1000] range\n * @param bboxSize - Size of the bbox to create (default: 20)\n * @returns [x1, y1, x2, y2] bbox in [0, 1000] coordinate space\n */\nexport function pointToBbox(\n x: number,\n y: number,\n bboxSize = defaultBboxSize,\n): [number, number, number, number] {\n const halfSize = bboxSize / 2;\n const x1 = Math.max(x - halfSize, 0);\n const y1 = Math.max(y - halfSize, 0);\n const x2 = Math.min(x + halfSize, 1000);\n const y2 = Math.min(y + halfSize, 1000);\n\n return [x1, y1, x2, y2];\n}\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 === 'qwen2.5-vl') {\n result = adaptQwen2_5Bbox(normalizedBbox as number[]);\n } else {\n // Default: normalized 0-1000 coordinate system\n // Includes: qwen3-vl, glm-v, auto-glm, auto-glm-multilingual, and future models\n result = normalized01000(normalizedBbox as number[], width, height);\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 let minEdgeSize = 500;\n if (vlMode === 'qwen3-vl') {\n minEdgeSize = 1200;\n }\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): 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 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 // Prevent spreading strings into {0: 'c', 1: 'o', ...}\n if (!isPlainObject(jsonObject)) {\n return {};\n }\n\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\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\nexport const finalizeActionName = 'Finalize';\n\n/**\n * Get a readable time string for the current time\n * @param format - Optional format string. Supports: YYYY, MM, DD, HH, mm, ss. Default: 'YYYY-MM-DD HH:mm:ss'\n * @returns A formatted time string with format label\n */\nexport const getReadableTimeString = (\n format = 'YYYY-MM-DD HH:mm:ss',\n): string => {\n const now = new Date();\n const year = now.getFullYear();\n const month = String(now.getMonth() + 1).padStart(2, '0');\n const day = String(now.getDate()).padStart(2, '0');\n const hours = String(now.getHours()).padStart(2, '0');\n const minutes = String(now.getMinutes()).padStart(2, '0');\n const seconds = String(now.getSeconds()).padStart(2, '0');\n\n const timeString = format\n .replace('YYYY', String(year))\n .replace('MM', month)\n .replace('DD', day)\n .replace('HH', hours)\n .replace('mm', minutes)\n .replace('ss', seconds);\n\n return `${timeString} (${format})`;\n};\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","defaultBboxSize","debugInspectUtils","getDebug","pointToBbox","x","y","bboxSize","halfSize","x1","Math","y1","x2","y2","fillBboxParam","locate","width","height","rightLimit","bottomLimit","vlMode","adaptBbox","adaptQwen2_5Bbox","bbox","msg","JSON","Error","result","adaptDoubaoBbox","assert","splitted","Number","bboxList","Array","item","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","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","jsonObject","zodSchema","isPlainObject","locatorFields","fieldName","fieldValue","parseActionParam","rawParam","param","locateFields","locateFieldValues","paramsForValidation","validated","finalizeActionName","getReadableTimeString","format","now","Date","year","month","day","hours","minutes","seconds","timeString"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;ACiBA,MAAMI,kBAAkB;AACxB,MAAMC,oBAAoBC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAY5B,SAASC,YACdC,CAAS,EACTC,CAAS,EACTC,WAAWN,eAAe;IAE1B,MAAMO,WAAWD,WAAW;IAC5B,MAAME,KAAKC,KAAK,GAAG,CAACL,IAAIG,UAAU;IAClC,MAAMG,KAAKD,KAAK,GAAG,CAACJ,IAAIE,UAAU;IAClC,MAAMI,KAAKF,KAAK,GAAG,CAACL,IAAIG,UAAU;IAClC,MAAMK,KAAKH,KAAK,GAAG,CAACJ,IAAIE,UAAU;IAElC,OAAO;QAACC;QAAIE;QAAIC;QAAIC;KAAG;AACzB;AAGO,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/CjB,KAAK,KAAK,CAACa,IAAI,CAAC,EAAE;QAClBb,KAAK,KAAK,CAACa,IAAI,CAAC,EAAE;QACC,YAAnB,OAAOA,IAAI,CAAC,EAAE,GACVb,KAAK,KAAK,CAACa,IAAI,CAAC,EAAE,IAClBb,KAAK,KAAK,CAACa,IAAI,CAAC,EAAE,GAAGtB;QACN,YAAnB,OAAOsB,IAAI,CAAC,EAAE,GACVb,KAAK,KAAK,CAACa,IAAI,CAAC,EAAE,IAClBb,KAAK,KAAK,CAACa,IAAI,CAAC,EAAE,GAAGtB;KAC1B;IACD,OAAO0B;AACT;AAEO,SAASC,gBACdL,IAAkC,EAClCP,KAAa,EACbC,MAAc;IAEdY,IAAAA,sBAAAA,MAAAA,AAAAA,EACEb,QAAQ,KAAKC,SAAS,GACtB;IAGF,IAAI,AAAgB,YAAhB,OAAOM,MAAmB;QAC5BM,IAAAA,sBAAAA,MAAAA,AAAAA,EACE,+BAA+B,IAAI,CAACN,KAAK,IAAI,KAC7C,CAAC,iDAAiD,EAAEA,MAAM;QAE5D,MAAMO,WAAWP,KAAK,KAAK,CAAC;QAC5B,IAAIO,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACLpB,KAAK,KAAK,CAAEqB,OAAOD,QAAQ,CAAC,EAAE,IAAId,QAAS;YAC3CN,KAAK,KAAK,CAAEqB,OAAOD,QAAQ,CAAC,EAAE,IAAIb,SAAU;YAC5CP,KAAK,KAAK,CAAEqB,OAAOD,QAAQ,CAAC,EAAE,IAAId,QAAS;YAC3CN,KAAK,KAAK,CAAEqB,OAAOD,QAAQ,CAAC,EAAE,IAAIb,SAAU;SAC7C;QAEH,MAAM,IAAIS,MAAM,CAAC,iDAAiD,EAAEH,MAAM;IAC5E;IAEA,IAAIS,WAAqB,EAAE;IAC3B,IAAIC,MAAM,OAAO,CAACV,SAAS,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EACvCA,KAAK,OAAO,CAAC,CAACW;QACZ,IAAI,AAAgB,YAAhB,OAAOA,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YAClD,MAAM,CAAC7B,GAAGC,EAAE,GAAG4B,KAAK,KAAK,CAAC;YAC1BF,SAAS,IAAI,CAACD,OAAO1B,EAAE,IAAI,KAAK0B,OAAOzB,EAAE,IAAI;QAC/C,OAAO,IAAI,AAAgB,YAAhB,OAAO4B,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YACzD,MAAM,CAAC7B,GAAGC,EAAE,GAAG4B,KAAK,KAAK,CAAC;YAC1BF,SAAS,IAAI,CAACD,OAAO1B,EAAE,IAAI,KAAK0B,OAAOzB,EAAE,IAAI;QAC/C,OACE0B,SAAS,IAAI,CAACD,OAAOG;IAEzB;SAEAF,WAAWT;IAGb,IAAIS,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAO;QACLtB,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGhB,QAAS;QACnCN,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGf,SAAU;QACpCP,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGhB,QAAS;QACnCN,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGf,SAAU;KACrC;IAIH,IACEe,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAO;QACLtB,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGhB,QAAS,QAAQf,kBAAkB;QAE/DS,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGf,SAAU,QAAQhB,kBAAkB;QAEhES,KAAK,GAAG,CACNM,OACAN,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGhB,QAAS,QAAQf,kBAAkB;QAE/DS,KAAK,GAAG,CACNO,QACAP,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGf,SAAU,QAAQhB,kBAAkB;KAEjE;IAGH,IAAIsB,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAO;QACLb,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGhB,QAAS;QACnCN,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGf,SAAU;QACpCP,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGhB,QAAS;QACnCN,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGf,SAAU;KACrC;IAGH,MAAMO,MAAM,CAAC,0CAA0C,EAAEC,KAAK,SAAS,CAACF,MAAM,CAAC,CAAC;IAChF,MAAM,IAAIG,MAAMF;AAClB;AAEA,SAASW,mBACPZ,IAAoB;IAEpB,IAAIU,MAAM,OAAO,CAACV,OAChB;QAAA,IAAIU,MAAM,OAAO,CAACV,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,MAAMgB,iBAAiBD,mBAAmBZ;IAE1C,IAAII,SAA2C;QAAC;QAAG;QAAG;QAAG;KAAE;IAEzDA,SADEP,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,SACvBQ,gBAAgBQ,gBAAgBpB,OAAOC,UACvCG,AAAW,aAAXA,SACAiB,gBAAgBD,gBAA4BpB,OAAOC,UACnDG,AAAW,iBAAXA,SACAE,iBAAiBc,kBAIjBE,gBAAgBF,gBAA4BpB,OAAOC;IAG9DU,MAAM,CAAC,EAAE,GAAGjB,KAAK,GAAG,CAACiB,MAAM,CAAC,EAAE,EAAET;IAChCS,MAAM,CAAC,EAAE,GAAGjB,KAAK,GAAG,CAACiB,MAAM,CAAC,EAAE,EAAER;IAEhC,OAAOQ;AACT;AAGO,SAASW,gBACdf,IAAc,EACdP,KAAa,EACbC,MAAc;IAEd,OAAO;QACLP,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGP,QAAS;QAC/BN,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGN,SAAU;QAChCP,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGP,QAAS;QAC/BN,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGN,SAAU;KACjC;AACH;AAGO,SAASoB,gBACdd,IAAc,EACdP,KAAa,EACbC,MAAc;IAEd,MAAMsB,OAAO7B,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGP,QAAS;IAC5C,MAAMwB,MAAM9B,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGN,SAAU;IAC5C,MAAMwB,QAAQ/B,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGP,QAAS;IAC7C,MAAM0B,SAAShC,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGN,SAAU;IAC/C,OAAO;QAACsB;QAAMC;QAAKC;QAAOC;KAAO;AACnC;AAEO,SAASC,gBACdpB,IAAc,EACdP,KAAa,EACbC,MAAc,EACd2B,UAAU,CAAC,EACXC,UAAU,CAAC,EACX3B,aAAaF,KAAK,EAClBG,cAAcF,MAAM,EACpBG,MAAiC;IAEjClB,kBACE,mBACAqB,MACAP,OACAC,QACA,UACA2B,SACAC,SACA,SACA3B,YACAC,aACA,UACAC;IAEF,MAAM,CAACmB,MAAMC,KAAKC,OAAOC,OAAO,GAAGrB,UACjCE,MACAP,OACAC,QACAC,YACAC,aACAC;IAIF,MAAM0B,WAAWP;IACjB,MAAMQ,UAAUP;IAChB,IAAIQ,YAAYP,QAAQF;IACxB,IAAIU,aAAaP,SAASF;IAI1B,IAAIM,WAAWE,YAAYhC,OACzBgC,YAAYhC,QAAQ8B;IAItB,IAAIC,UAAUE,aAAahC,QACzBgC,aAAahC,SAAS8B;IAIxBC,YAAYtC,KAAK,GAAG,CAAC,GAAGsC;IACxBC,aAAavC,KAAK,GAAG,CAAC,GAAGuC;IAEzB,MAAMC,OAAO;QACX,MAAMJ,WAAWF;QACjB,KAAKG,UAAUF;QACf,OAAOG;QACP,QAAQC;IACV;IACA/C,kBAAkB,4BAA4BgD;IAE9C,OAAOA;AACT;AAEO,SAASC,WAAWC,KAAa;IACtC,MAAMC,UAAU3C,KAAK,GAAG,IAAI0C,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,IAAI;IACnD,MAAMC,SAAS7C,KAAK,GAAG,IAAI0C,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,GAAG;IACjD,MAAME,WAAW9C,KAAK,GAAG,IAAI0C,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,IAAI,GAAGA,EAAE,KAAK;IAC9D,MAAMG,YAAY/C,KAAK,GAAG,IAAI0C,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,EAChBvC,MAAgC;IAEhC,IAAIwC,cAAc;IAClB,IAAIxC,AAAW,eAAXA,QACFwC,cAAc;IAEhB,MAAMC,iBAAiB;IAGvB,MAAMC,wBACJZ,KAAK,KAAK,GAAGU,cACTlD,KAAK,IAAI,CAAEkD,AAAAA,CAAAA,cAAcV,KAAK,KAAI,IAAK,KACvCW;IACN,MAAME,sBACJb,KAAK,MAAM,GAAGU,cACVlD,KAAK,IAAI,CAAEkD,AAAAA,CAAAA,cAAcV,KAAK,MAAK,IAAK,KACxCW;IAGN,IAAIG,WAAWtD,KAAK,GAAG,CAACkD,aAAaV,KAAK,KAAK,GAAGY,AAAwB,IAAxBA;IAClD,IAAIG,YAAYvD,KAAK,GAAG,CAACkD,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,UAAUxD,KAAK,GAAG,CAAC,GAAGwD;IACtBC,SAASzD,KAAK,GAAG,CAAC,GAAGyD;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,AAAAA,IAAAA,0BAAAA,UAAAA,AAAAA,EAAWH;IAChC,MAAMI,kCAAkCF,aAAc,MAAM,CAC1D,CAACG;QACC,IAAIA,YAAY,UAAU,CAAC,QAAQ,KAAKC,0BAAAA,QAAAA,CAAAA,IAAa,EACnD,OAAO;QAET,OAAO;IACT;IAGF,MAAMC,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,uBAAAA,AAAAA,EAAwB;QACjD,gBAAgBT;QAChB,sBAAsBK;QACtBH;IACF;IACA,OAAOM;AACT;AAEO,SAASE,uBACdC,KAAuB,EACvBC,WAAgC;IAEhC,MAAMC,OAA+B,EAAE;IAEvC,KAAK,MAAMC,QAAQH,MAAO;QACxB,MAAMI,OAAOD,KAAK,IAAI;QAEtB,MAAME,SAASJ,YAAY,IAAI,CAAC,CAACI,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,OAAOR;AACT;AAGO,MAAMS,cAAcC,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAClC,MAAMA,6BAAAA,CAAAA,CAAAA,MAAQ;IACd,KAAKA,6BAAAA,CAAAA,CAAAA,MAAQ;AACf;AAEO,MAAMC,aAAaD,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IACjC,OAAOA,6BAAAA,CAAAA,CAAAA,MAAQ;IACf,QAAQA,6BAAAA,CAAAA,CAAAA,MAAQ;IAChB,KAAKA,6BAAAA,CAAAA,CAAAA,MAAQ,GAAG,QAAQ;AAC1B;AAEO,MAAME,aAAaH,YAAY,GAAG,CAACE,YAAY,GAAG,CACvDD,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IACP,MAAMA,6BAAAA,CAAAA,CAAAA,MAAQ,GAAG,QAAQ;AAC3B;AAIK,MAAMG,0BAA0BH,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC9C,QAAQA,6BAAAA,CAAAA,CAAAA,KACA,CACJA,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;QACP,MAAMA,6BAAAA,CAAAA,CAAAA,MAAQ;QACd,KAAKA,6BAAAA,CAAAA,CAAAA,MAAQ;IACf,IAED,QAAQ;IACX,yBAAyBA,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ;AAC/C;AAGO,MAAMI,oBAAoBJ,6BAAAA,CAAAA,CAAAA,KAAO,CAAC;IACvCA,6BAAAA,CAAAA,CAAAA,MAAQ;IACRA,6BAAAA,CAAAA,CAAAA,MACS,CAAC;QACN,QAAQA,6BAAAA,CAAAA,CAAAA,MAAQ;IAClB,GACC,GAAG,CAACG,wBAAwB,OAAO;CACvC;AAMD,MAAME,sBAAsB;AAG5B,MAAMC,wBAAwBN,6BAAAA,CAAAA,CAAAA,MACrB,CAAC;IACN,QAAQI;IACR,WAAWJ,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ;IAC/B,WAAWA,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ;IAC/B,OAAOA,6BAAAA,CAAAA,CAAAA,KAAO,CAAC;QAACA,6BAAAA,CAAAA,CAAAA,MAAQ;QAAIA,6BAAAA,CAAAA,CAAAA,OAAS;KAAG,EAAE,QAAQ;AACpD,GACC,WAAW;AAGiBA,6BAAAA,CAAAA,CAAAA,MACtB,CAAC;IACN,CAACK,oBAAoB,EAAEL,6BAAAA,CAAAA,CAAAA,OAAS,CAAC;IACjC,QAAQI;IAGR,WAAWJ,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ;IAC/B,WAAWA,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ;IAC/B,OAAOA,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ;IAG3B,QAAQA,6BAAAA,CAAAA,CAAAA,KAAO,CAAC;QAACA,6BAAAA,CAAAA,CAAAA,MAAQ;QAAIA,6BAAAA,CAAAA,CAAAA,MAAQ;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;IACvCxE,IAAAA,sBAAAA,MAAAA,AAAAA,EACEuE,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,OAAOjH,OAAO,IAAI,CAACgH,UAAU,KAAK;QACxC,OAAOC,KAAK,MAAM,CAAC,CAAClH;YAClB,MAAMyG,QAAQQ,UAAU,KAAK,CAACjH,IAAI;YAClC,IAAI,CAACwG,uBAAuBC,QAC1B,OAAO;YAIT,IAAIO,cACF,OAAOP,MAAM,IAAI,EAAE,aAAa;YAGlC,OAAO;QACT;IACF;IAGA,OAAO,EAAE;AACX;AAEO,MAAMZ,kBAAkB,CAC7BsB,YACAC;IAGA,IAAI,CAACC,AAAAA,IAAAA,sBAAAA,aAAAA,AAAAA,EAAcF,aACjB,OAAO,CAAC;IAGV,MAAMG,gBAAgBR,4BAA4BM;IAClD,MAAMrF,SAAS;QAAE,GAAGoF,UAAU;IAAC;IAE/B,KAAK,MAAMI,aAAaD,cAAe;QACrC,MAAME,aAAazF,MAAM,CAACwF,UAAU;QACpC,IAAIC,YAEF;YAAA,IAAI,AAAsB,YAAtB,OAAOA,YACTzF,MAAM,CAACwF,UAAU,GAAGC;iBACf,IAAI,AAAsB,YAAtB,OAAOA,YAEhB;gBAAA,IAAIA,WAAW,MAAM,EAEnB;oBAAA,IAAI,AAA6B,YAA7B,OAAOA,WAAW,MAAM,EAC1BzF,MAAM,CAACwF,UAAU,GAAGC,WAAW,MAAM;yBAChC,IACL,AAA6B,YAA7B,OAAOA,WAAW,MAAM,IACxBA,WAAW,MAAM,CAAC,MAAM,EAGxBzF,MAAM,CAACwF,UAAU,GAAGC,WAAW,MAAM,CAAC,MAAM;gBAC9C;YACF;QACF;IAEJ;IAEA,OAAOzF;AACT;AAUO,MAAM0F,mBAAmB,CAC9BC,UACAN;IAGA,IAAI,CAACA,WACH;IAIF,MAAMO,QAAQD,YAAY,CAAC;IAG3B,MAAME,eAAed,4BAA4BM;IAGjD,IAAIQ,AAAwB,MAAxBA,aAAa,MAAM,EACrB,OAAOR,UAAU,KAAK,CAACO;IAIzB,MAAME,oBAAyC,CAAC;IAChD,KAAK,MAAMN,aAAaK,aACtB,IAAIL,aAAaI,OACfE,iBAAiB,CAACN,UAAU,GAAGI,KAAK,CAACJ,UAAU;IAKnD,MAAMO,sBAA2C,CAAC;IAClD,IAAK,MAAM9H,OAAO2H,MAChB,IAAIC,aAAa,QAAQ,CAAC5H,MAExB8H,mBAAmB,CAAC9H,IAAI,GAAG;QAAE,QAAQ;IAAU;SAE/C8H,mBAAmB,CAAC9H,IAAI,GAAG2H,KAAK,CAAC3H,IAAI;IAKzC,MAAM+H,YAAYX,UAAU,KAAK,CAACU;IAGlC,IAAK,MAAMP,aAAaM,kBACtBE,SAAS,CAACR,UAAU,GAAGM,iBAAiB,CAACN,UAAU;IAGrD,OAAOQ;AACT;AAEO,MAAMC,qBAAqB;AAO3B,MAAMC,wBAAwB,CACnCC,SAAS,qBAAqB;IAE9B,MAAMC,MAAM,IAAIC;IAChB,MAAMC,OAAOF,IAAI,WAAW;IAC5B,MAAMG,QAAQzB,OAAOsB,IAAI,QAAQ,KAAK,GAAG,QAAQ,CAAC,GAAG;IACrD,MAAMI,MAAM1B,OAAOsB,IAAI,OAAO,IAAI,QAAQ,CAAC,GAAG;IAC9C,MAAMK,QAAQ3B,OAAOsB,IAAI,QAAQ,IAAI,QAAQ,CAAC,GAAG;IACjD,MAAMM,UAAU5B,OAAOsB,IAAI,UAAU,IAAI,QAAQ,CAAC,GAAG;IACrD,MAAMO,UAAU7B,OAAOsB,IAAI,UAAU,IAAI,QAAQ,CAAC,GAAG;IAErD,MAAMQ,aAAaT,OAChB,OAAO,CAAC,QAAQrB,OAAOwB,OACvB,OAAO,CAAC,MAAMC,OACd,OAAO,CAAC,MAAMC,KACd,OAAO,CAAC,MAAMC,OACd,OAAO,CAAC,MAAMC,SACd,OAAO,CAAC,MAAMC;IAEjB,OAAO,GAAGC,WAAW,EAAE,EAAET,OAAO,CAAC,CAAC;AACpC"}
|
|
1
|
+
{"version":3,"file":"common.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../src/common.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n BaseElement,\n DeviceAction,\n ElementTreeNode,\n MidsceneYamlFlowItem,\n PlanningAction,\n Rect,\n Size,\n} from '@/types';\nimport { assert, isPlainObject } from '@midscene/shared/utils';\n\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\n\nimport { isUITars } from '@/ai-model/auto-glm/util';\nimport type { PlanningLocateParam } from '@/types';\nimport { NodeType } from '@midscene/shared/constants';\nimport type { TModelFamily } 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\nconst defaultBboxSize = 20; // must be even number\nconst debugInspectUtils = getDebug('ai:common');\ntype AdaptBboxInput = number[] | string[] | string | (number[] | string[])[];\n\n/**\n * Convert a point coordinate [0, 1000] to a small bbox [0, 1000]\n * Creates a small bbox around the center point in the same coordinate space\n *\n * @param x - X coordinate in [0, 1000] range\n * @param y - Y coordinate in [0, 1000] range\n * @param bboxSize - Size of the bbox to create (default: 20)\n * @returns [x1, y1, x2, y2] bbox in [0, 1000] coordinate space\n */\nexport function pointToBbox(\n x: number,\n y: number,\n bboxSize = defaultBboxSize,\n): [number, number, number, number] {\n const halfSize = bboxSize / 2;\n const x1 = Math.max(x - halfSize, 0);\n const y1 = Math.max(y - halfSize, 0);\n const x2 = Math.min(x + halfSize, 1000);\n const y2 = Math.min(y + halfSize, 1000);\n\n return [x1, y1, x2, y2];\n}\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 modelFamily: TModelFamily | 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 modelFamily,\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 modelFamily: TModelFamily | 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 (modelFamily === 'doubao-vision' || isUITars(modelFamily)) {\n result = adaptDoubaoBbox(normalizedBbox, width, height);\n } else if (modelFamily === 'gemini') {\n result = adaptGeminiBbox(normalizedBbox as number[], width, height);\n } else if (modelFamily === 'qwen2.5-vl') {\n result = adaptQwen2_5Bbox(normalizedBbox as number[]);\n } else {\n // Default: normalized 0-1000 coordinate system\n // Includes: qwen3-vl, glm-v, auto-glm, auto-glm-multilingual, and future models\n result = normalized01000(normalizedBbox as number[], width, height);\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 modelFamily?: TModelFamily | 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 'modelFamily',\n modelFamily,\n );\n const [left, top, right, bottom] = adaptBbox(\n bbox,\n width,\n height,\n rightLimit,\n bottomLimit,\n modelFamily,\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 modelFamily: TModelFamily | undefined,\n) {\n let minEdgeSize = 500;\n if (modelFamily === 'qwen3-vl') {\n minEdgeSize = 1200;\n }\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): 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 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 // Prevent spreading strings into {0: 'c', 1: 'o', ...}\n if (!isPlainObject(jsonObject)) {\n return {};\n }\n\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\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\nexport const finalizeActionName = 'Finalize';\n\n/**\n * Get a readable time string for the current time\n * @param format - Optional format string. Supports: YYYY, MM, DD, HH, mm, ss. Default: 'YYYY-MM-DD HH:mm:ss'\n * @returns A formatted time string with format label\n */\nexport const getReadableTimeString = (\n format = 'YYYY-MM-DD HH:mm:ss',\n): string => {\n const now = new Date();\n const year = now.getFullYear();\n const month = String(now.getMonth() + 1).padStart(2, '0');\n const day = String(now.getDate()).padStart(2, '0');\n const hours = String(now.getHours()).padStart(2, '0');\n const minutes = String(now.getMinutes()).padStart(2, '0');\n const seconds = String(now.getSeconds()).padStart(2, '0');\n\n const timeString = format\n .replace('YYYY', String(year))\n .replace('MM', month)\n .replace('DD', day)\n .replace('HH', hours)\n .replace('mm', minutes)\n .replace('ss', seconds);\n\n return `${timeString} (${format})`;\n};\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","defaultBboxSize","debugInspectUtils","getDebug","pointToBbox","x","y","bboxSize","halfSize","x1","Math","y1","x2","y2","fillBboxParam","locate","width","height","rightLimit","bottomLimit","modelFamily","adaptBbox","adaptQwen2_5Bbox","bbox","msg","JSON","Error","result","adaptDoubaoBbox","assert","splitted","Number","bboxList","Array","item","normalizeBboxInput","normalizedBbox","isUITars","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","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","jsonObject","zodSchema","isPlainObject","locatorFields","fieldName","fieldValue","parseActionParam","rawParam","param","locateFields","locateFieldValues","paramsForValidation","validated","finalizeActionName","getReadableTimeString","format","now","Date","year","month","day","hours","minutes","seconds","timeString"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;ACkBA,MAAMI,kBAAkB;AACxB,MAAMC,oBAAoBC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAY5B,SAASC,YACdC,CAAS,EACTC,CAAS,EACTC,WAAWN,eAAe;IAE1B,MAAMO,WAAWD,WAAW;IAC5B,MAAME,KAAKC,KAAK,GAAG,CAACL,IAAIG,UAAU;IAClC,MAAMG,KAAKD,KAAK,GAAG,CAACJ,IAAIE,UAAU;IAClC,MAAMI,KAAKF,KAAK,GAAG,CAACL,IAAIG,UAAU;IAClC,MAAMK,KAAKH,KAAK,GAAG,CAACJ,IAAIE,UAAU;IAElC,OAAO;QAACC;QAAIE;QAAIC;QAAIC;KAAG;AACzB;AAGO,SAASC,cACdC,MAA2B,EAC3BC,KAAa,EACbC,MAAc,EACdC,UAAkB,EAClBC,WAAmB,EACnBC,WAAqC;IAGrC,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/CjB,KAAK,KAAK,CAACa,IAAI,CAAC,EAAE;QAClBb,KAAK,KAAK,CAACa,IAAI,CAAC,EAAE;QACC,YAAnB,OAAOA,IAAI,CAAC,EAAE,GACVb,KAAK,KAAK,CAACa,IAAI,CAAC,EAAE,IAClBb,KAAK,KAAK,CAACa,IAAI,CAAC,EAAE,GAAGtB;QACN,YAAnB,OAAOsB,IAAI,CAAC,EAAE,GACVb,KAAK,KAAK,CAACa,IAAI,CAAC,EAAE,IAClBb,KAAK,KAAK,CAACa,IAAI,CAAC,EAAE,GAAGtB;KAC1B;IACD,OAAO0B;AACT;AAEO,SAASC,gBACdL,IAAkC,EAClCP,KAAa,EACbC,MAAc;IAEdY,IAAAA,sBAAAA,MAAAA,AAAAA,EACEb,QAAQ,KAAKC,SAAS,GACtB;IAGF,IAAI,AAAgB,YAAhB,OAAOM,MAAmB;QAC5BM,IAAAA,sBAAAA,MAAAA,AAAAA,EACE,+BAA+B,IAAI,CAACN,KAAK,IAAI,KAC7C,CAAC,iDAAiD,EAAEA,MAAM;QAE5D,MAAMO,WAAWP,KAAK,KAAK,CAAC;QAC5B,IAAIO,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACLpB,KAAK,KAAK,CAAEqB,OAAOD,QAAQ,CAAC,EAAE,IAAId,QAAS;YAC3CN,KAAK,KAAK,CAAEqB,OAAOD,QAAQ,CAAC,EAAE,IAAIb,SAAU;YAC5CP,KAAK,KAAK,CAAEqB,OAAOD,QAAQ,CAAC,EAAE,IAAId,QAAS;YAC3CN,KAAK,KAAK,CAAEqB,OAAOD,QAAQ,CAAC,EAAE,IAAIb,SAAU;SAC7C;QAEH,MAAM,IAAIS,MAAM,CAAC,iDAAiD,EAAEH,MAAM;IAC5E;IAEA,IAAIS,WAAqB,EAAE;IAC3B,IAAIC,MAAM,OAAO,CAACV,SAAS,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EACvCA,KAAK,OAAO,CAAC,CAACW;QACZ,IAAI,AAAgB,YAAhB,OAAOA,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YAClD,MAAM,CAAC7B,GAAGC,EAAE,GAAG4B,KAAK,KAAK,CAAC;YAC1BF,SAAS,IAAI,CAACD,OAAO1B,EAAE,IAAI,KAAK0B,OAAOzB,EAAE,IAAI;QAC/C,OAAO,IAAI,AAAgB,YAAhB,OAAO4B,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YACzD,MAAM,CAAC7B,GAAGC,EAAE,GAAG4B,KAAK,KAAK,CAAC;YAC1BF,SAAS,IAAI,CAACD,OAAO1B,EAAE,IAAI,KAAK0B,OAAOzB,EAAE,IAAI;QAC/C,OACE0B,SAAS,IAAI,CAACD,OAAOG;IAEzB;SAEAF,WAAWT;IAGb,IAAIS,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAO;QACLtB,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGhB,QAAS;QACnCN,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGf,SAAU;QACpCP,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGhB,QAAS;QACnCN,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGf,SAAU;KACrC;IAIH,IACEe,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAO;QACLtB,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGhB,QAAS,QAAQf,kBAAkB;QAE/DS,KAAK,GAAG,CACN,GACAA,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGf,SAAU,QAAQhB,kBAAkB;QAEhES,KAAK,GAAG,CACNM,OACAN,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGhB,QAAS,QAAQf,kBAAkB;QAE/DS,KAAK,GAAG,CACNO,QACAP,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGf,SAAU,QAAQhB,kBAAkB;KAEjE;IAGH,IAAIsB,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAO;QACLb,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGhB,QAAS;QACnCN,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGf,SAAU;QACpCP,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGhB,QAAS;QACnCN,KAAK,KAAK,CAAEsB,QAAQ,CAAC,EAAE,GAAGf,SAAU;KACrC;IAGH,MAAMO,MAAM,CAAC,0CAA0C,EAAEC,KAAK,SAAS,CAACF,MAAM,CAAC,CAAC;IAChF,MAAM,IAAIG,MAAMF;AAClB;AAEA,SAASW,mBACPZ,IAAoB;IAEpB,IAAIU,MAAM,OAAO,CAACV,OAChB;QAAA,IAAIU,MAAM,OAAO,CAACV,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,WAAqC;IAErC,MAAMgB,iBAAiBD,mBAAmBZ;IAE1C,IAAII,SAA2C;QAAC;QAAG;QAAG;QAAG;KAAE;IAEzDA,SADEP,AAAgB,oBAAhBA,eAAmCiB,AAAAA,IAAAA,wBAAAA,QAAAA,AAAAA,EAASjB,eACrCQ,gBAAgBQ,gBAAgBpB,OAAOC,UACvCG,AAAgB,aAAhBA,cACAkB,gBAAgBF,gBAA4BpB,OAAOC,UACnDG,AAAgB,iBAAhBA,cACAE,iBAAiBc,kBAIjBG,gBAAgBH,gBAA4BpB,OAAOC;IAG9DU,MAAM,CAAC,EAAE,GAAGjB,KAAK,GAAG,CAACiB,MAAM,CAAC,EAAE,EAAET;IAChCS,MAAM,CAAC,EAAE,GAAGjB,KAAK,GAAG,CAACiB,MAAM,CAAC,EAAE,EAAER;IAEhC,OAAOQ;AACT;AAGO,SAASY,gBACdhB,IAAc,EACdP,KAAa,EACbC,MAAc;IAEd,OAAO;QACLP,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGP,QAAS;QAC/BN,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGN,SAAU;QAChCP,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGP,QAAS;QAC/BN,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGN,SAAU;KACjC;AACH;AAGO,SAASqB,gBACdf,IAAc,EACdP,KAAa,EACbC,MAAc;IAEd,MAAMuB,OAAO9B,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGP,QAAS;IAC5C,MAAMyB,MAAM/B,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGN,SAAU;IAC5C,MAAMyB,QAAQhC,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGP,QAAS;IAC7C,MAAM2B,SAASjC,KAAK,KAAK,CAAEa,IAAI,CAAC,EAAE,GAAGN,SAAU;IAC/C,OAAO;QAACuB;QAAMC;QAAKC;QAAOC;KAAO;AACnC;AAEO,SAASC,gBACdrB,IAAc,EACdP,KAAa,EACbC,MAAc,EACd4B,UAAU,CAAC,EACXC,UAAU,CAAC,EACX5B,aAAaF,KAAK,EAClBG,cAAcF,MAAM,EACpBG,WAAsC;IAEtClB,kBACE,mBACAqB,MACAP,OACAC,QACA,UACA4B,SACAC,SACA,SACA5B,YACAC,aACA,eACAC;IAEF,MAAM,CAACoB,MAAMC,KAAKC,OAAOC,OAAO,GAAGtB,UACjCE,MACAP,OACAC,QACAC,YACAC,aACAC;IAIF,MAAM2B,WAAWP;IACjB,MAAMQ,UAAUP;IAChB,IAAIQ,YAAYP,QAAQF;IACxB,IAAIU,aAAaP,SAASF;IAI1B,IAAIM,WAAWE,YAAYjC,OACzBiC,YAAYjC,QAAQ+B;IAItB,IAAIC,UAAUE,aAAajC,QACzBiC,aAAajC,SAAS+B;IAIxBC,YAAYvC,KAAK,GAAG,CAAC,GAAGuC;IACxBC,aAAaxC,KAAK,GAAG,CAAC,GAAGwC;IAEzB,MAAMC,OAAO;QACX,MAAMJ,WAAWF;QACjB,KAAKG,UAAUF;QACf,OAAOG;QACP,QAAQC;IACV;IACAhD,kBAAkB,4BAA4BiD;IAE9C,OAAOA;AACT;AAEO,SAASC,WAAWC,KAAa;IACtC,MAAMC,UAAU5C,KAAK,GAAG,IAAI2C,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,IAAI;IACnD,MAAMC,SAAS9C,KAAK,GAAG,IAAI2C,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,GAAG;IACjD,MAAME,WAAW/C,KAAK,GAAG,IAAI2C,MAAM,GAAG,CAAC,CAACE,IAAMA,EAAE,IAAI,GAAGA,EAAE,KAAK;IAC9D,MAAMG,YAAYhD,KAAK,GAAG,IAAI2C,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,EAChBxC,WAAqC;IAErC,IAAIyC,cAAc;IAClB,IAAIzC,AAAgB,eAAhBA,aACFyC,cAAc;IAEhB,MAAMC,iBAAiB;IAGvB,MAAMC,wBACJZ,KAAK,KAAK,GAAGU,cACTnD,KAAK,IAAI,CAAEmD,AAAAA,CAAAA,cAAcV,KAAK,KAAI,IAAK,KACvCW;IACN,MAAME,sBACJb,KAAK,MAAM,GAAGU,cACVnD,KAAK,IAAI,CAAEmD,AAAAA,CAAAA,cAAcV,KAAK,MAAK,IAAK,KACxCW;IAGN,IAAIG,WAAWvD,KAAK,GAAG,CAACmD,aAAaV,KAAK,KAAK,GAAGY,AAAwB,IAAxBA;IAClD,IAAIG,YAAYxD,KAAK,GAAG,CAACmD,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,UAAUzD,KAAK,GAAG,CAAC,GAAGyD;IACtBC,SAAS1D,KAAK,GAAG,CAAC,GAAG0D;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,AAAAA,IAAAA,0BAAAA,UAAAA,AAAAA,EAAWH;IAChC,MAAMI,kCAAkCF,aAAc,MAAM,CAC1D,CAACG;QACC,IAAIA,YAAY,UAAU,CAAC,QAAQ,KAAKC,0BAAAA,QAAAA,CAAAA,IAAa,EACnD,OAAO;QAET,OAAO;IACT;IAGF,MAAMC,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,uBAAAA,AAAAA,EAAwB;QACjD,gBAAgBT;QAChB,sBAAsBK;QACtBH;IACF;IACA,OAAOM;AACT;AAEO,SAASE,uBACdC,KAAuB,EACvBC,WAAgC;IAEhC,MAAMC,OAA+B,EAAE;IAEvC,KAAK,MAAMC,QAAQH,MAAO;QACxB,MAAMI,OAAOD,KAAK,IAAI;QAEtB,MAAME,SAASJ,YAAY,IAAI,CAAC,CAACI,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,OAAOR;AACT;AAGO,MAAMS,cAAcC,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAClC,MAAMA,6BAAAA,CAAAA,CAAAA,MAAQ;IACd,KAAKA,6BAAAA,CAAAA,CAAAA,MAAQ;AACf;AAEO,MAAMC,aAAaD,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IACjC,OAAOA,6BAAAA,CAAAA,CAAAA,MAAQ;IACf,QAAQA,6BAAAA,CAAAA,CAAAA,MAAQ;IAChB,KAAKA,6BAAAA,CAAAA,CAAAA,MAAQ,GAAG,QAAQ;AAC1B;AAEO,MAAME,aAAaH,YAAY,GAAG,CAACE,YAAY,GAAG,CACvDD,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IACP,MAAMA,6BAAAA,CAAAA,CAAAA,MAAQ,GAAG,QAAQ;AAC3B;AAIK,MAAMG,0BAA0BH,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC9C,QAAQA,6BAAAA,CAAAA,CAAAA,KACA,CACJA,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;QACP,MAAMA,6BAAAA,CAAAA,CAAAA,MAAQ;QACd,KAAKA,6BAAAA,CAAAA,CAAAA,MAAQ;IACf,IAED,QAAQ;IACX,yBAAyBA,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ;AAC/C;AAGO,MAAMI,oBAAoBJ,6BAAAA,CAAAA,CAAAA,KAAO,CAAC;IACvCA,6BAAAA,CAAAA,CAAAA,MAAQ;IACRA,6BAAAA,CAAAA,CAAAA,MACS,CAAC;QACN,QAAQA,6BAAAA,CAAAA,CAAAA,MAAQ;IAClB,GACC,GAAG,CAACG,wBAAwB,OAAO;CACvC;AAMD,MAAME,sBAAsB;AAG5B,MAAMC,wBAAwBN,6BAAAA,CAAAA,CAAAA,MACrB,CAAC;IACN,QAAQI;IACR,WAAWJ,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ;IAC/B,WAAWA,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ;IAC/B,OAAOA,6BAAAA,CAAAA,CAAAA,KAAO,CAAC;QAACA,6BAAAA,CAAAA,CAAAA,MAAQ;QAAIA,6BAAAA,CAAAA,CAAAA,OAAS;KAAG,EAAE,QAAQ;AACpD,GACC,WAAW;AAGiBA,6BAAAA,CAAAA,CAAAA,MACtB,CAAC;IACN,CAACK,oBAAoB,EAAEL,6BAAAA,CAAAA,CAAAA,OAAS,CAAC;IACjC,QAAQI;IAGR,WAAWJ,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ;IAC/B,WAAWA,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ;IAC/B,OAAOA,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ;IAG3B,QAAQA,6BAAAA,CAAAA,CAAAA,KAAO,CAAC;QAACA,6BAAAA,CAAAA,CAAAA,MAAQ;QAAIA,6BAAAA,CAAAA,CAAAA,MAAQ;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;IACvCzE,IAAAA,sBAAAA,MAAAA,AAAAA,EACEwE,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,OAAOlH,OAAO,IAAI,CAACiH,UAAU,KAAK;QACxC,OAAOC,KAAK,MAAM,CAAC,CAACnH;YAClB,MAAM0G,QAAQQ,UAAU,KAAK,CAAClH,IAAI;YAClC,IAAI,CAACyG,uBAAuBC,QAC1B,OAAO;YAIT,IAAIO,cACF,OAAOP,MAAM,IAAI,EAAE,aAAa;YAGlC,OAAO;QACT;IACF;IAGA,OAAO,EAAE;AACX;AAEO,MAAMZ,kBAAkB,CAC7BsB,YACAC;IAGA,IAAI,CAACC,AAAAA,IAAAA,sBAAAA,aAAAA,AAAAA,EAAcF,aACjB,OAAO,CAAC;IAGV,MAAMG,gBAAgBR,4BAA4BM;IAClD,MAAMtF,SAAS;QAAE,GAAGqF,UAAU;IAAC;IAE/B,KAAK,MAAMI,aAAaD,cAAe;QACrC,MAAME,aAAa1F,MAAM,CAACyF,UAAU;QACpC,IAAIC,YAEF;YAAA,IAAI,AAAsB,YAAtB,OAAOA,YACT1F,MAAM,CAACyF,UAAU,GAAGC;iBACf,IAAI,AAAsB,YAAtB,OAAOA,YAEhB;gBAAA,IAAIA,WAAW,MAAM,EAEnB;oBAAA,IAAI,AAA6B,YAA7B,OAAOA,WAAW,MAAM,EAC1B1F,MAAM,CAACyF,UAAU,GAAGC,WAAW,MAAM;yBAChC,IACL,AAA6B,YAA7B,OAAOA,WAAW,MAAM,IACxBA,WAAW,MAAM,CAAC,MAAM,EAGxB1F,MAAM,CAACyF,UAAU,GAAGC,WAAW,MAAM,CAAC,MAAM;gBAC9C;YACF;QACF;IAEJ;IAEA,OAAO1F;AACT;AAUO,MAAM2F,mBAAmB,CAC9BC,UACAN;IAGA,IAAI,CAACA,WACH;IAIF,MAAMO,QAAQD,YAAY,CAAC;IAG3B,MAAME,eAAed,4BAA4BM;IAGjD,IAAIQ,AAAwB,MAAxBA,aAAa,MAAM,EACrB,OAAOR,UAAU,KAAK,CAACO;IAIzB,MAAME,oBAAyC,CAAC;IAChD,KAAK,MAAMN,aAAaK,aACtB,IAAIL,aAAaI,OACfE,iBAAiB,CAACN,UAAU,GAAGI,KAAK,CAACJ,UAAU;IAKnD,MAAMO,sBAA2C,CAAC;IAClD,IAAK,MAAM/H,OAAO4H,MAChB,IAAIC,aAAa,QAAQ,CAAC7H,MAExB+H,mBAAmB,CAAC/H,IAAI,GAAG;QAAE,QAAQ;IAAU;SAE/C+H,mBAAmB,CAAC/H,IAAI,GAAG4H,KAAK,CAAC5H,IAAI;IAKzC,MAAMgI,YAAYX,UAAU,KAAK,CAACU;IAGlC,IAAK,MAAMP,aAAaM,kBACtBE,SAAS,CAACR,UAAU,GAAGM,iBAAiB,CAACN,UAAU;IAGrD,OAAOQ;AACT;AAEO,MAAMC,qBAAqB;AAO3B,MAAMC,wBAAwB,CACnCC,SAAS,qBAAqB;IAE9B,MAAMC,MAAM,IAAIC;IAChB,MAAMC,OAAOF,IAAI,WAAW;IAC5B,MAAMG,QAAQzB,OAAOsB,IAAI,QAAQ,KAAK,GAAG,QAAQ,CAAC,GAAG;IACrD,MAAMI,MAAM1B,OAAOsB,IAAI,OAAO,IAAI,QAAQ,CAAC,GAAG;IAC9C,MAAMK,QAAQ3B,OAAOsB,IAAI,QAAQ,IAAI,QAAQ,CAAC,GAAG;IACjD,MAAMM,UAAU5B,OAAOsB,IAAI,UAAU,IAAI,QAAQ,CAAC,GAAG;IACrD,MAAMO,UAAU7B,OAAOsB,IAAI,UAAU,IAAI,QAAQ,CAAC,GAAG;IAErD,MAAMQ,aAAaT,OAChB,OAAO,CAAC,QAAQrB,OAAOwB,OACvB,OAAO,CAAC,MAAMC,OACd,OAAO,CAAC,MAAMC,KACd,OAAO,CAAC,MAAMC,OACd,OAAO,CAAC,MAAMC,SACd,OAAO,CAAC,MAAMC;IAEjB,OAAO,GAAGC,WAAW,EAAE,EAAET,OAAO,CAAC,CAAC;AACpC"}
|
package/dist/lib/device/index.js
CHANGED
|
@@ -108,11 +108,11 @@ const actionInputParamSchema = external_zod_namespaceObject.z.object({
|
|
|
108
108
|
external_zod_namespaceObject.z.number()
|
|
109
109
|
]).transform((val)=>String(val)).describe('The text to input. Provide the final content for replace/append modes, or an empty string when using clear mode to remove existing text.'),
|
|
110
110
|
locate: (0, external_common_js_namespaceObject.getMidsceneLocationSchema)().describe(inputLocateDescription).optional(),
|
|
111
|
-
mode: external_zod_namespaceObject.z["enum"]([
|
|
111
|
+
mode: external_zod_namespaceObject.z.preprocess((val)=>'append' === val ? 'typeOnly' : val, external_zod_namespaceObject.z["enum"]([
|
|
112
112
|
'replace',
|
|
113
113
|
'clear',
|
|
114
|
-
'
|
|
115
|
-
]).default('replace').optional().describe('Input mode: "replace" (default) - clear the field and input the value; "
|
|
114
|
+
'typeOnly'
|
|
115
|
+
]).default('replace').optional().describe('Input mode: "replace" (default) - clear the field and input the value; "typeOnly" - type the value directly without clearing the field first; "clear" - clear the field without inputting new text.'))
|
|
116
116
|
});
|
|
117
117
|
const defineActionInput = (call)=>defineAction({
|
|
118
118
|
name: 'Input',
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"device/index.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../src/device/index.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import { finalizeActionName, 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 interface FileChooserHandler {\n accept(files: string[]): Promise<void>;\n}\n\nexport abstract class AbstractInterface {\n abstract interfaceType: string;\n\n abstract screenshotBase64(): Promise<string>;\n abstract size(): Promise<Size>;\n abstract actionSpace(): 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 // for web only\n registerFileChooserListener?(\n handler: (chooser: FileChooserHandler) => Promise<void>,\n ): Promise<{ dispose: () => void; getError: () => Error | undefined }>;\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});\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 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 direction: z\n .enum(['down', 'up', 'right', 'left'])\n .default('down')\n .describe(\n 'The direction to scroll. Only effective when scrollType is \"singleAction\".',\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(\n 'Describe the target element to be scrolled on, like \"the table\" or \"the list\" or \"the content area\" or \"the scrollable area\". Do NOT provide a general intent like \"scroll to find some element\"',\n ),\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:\n 'Drag and drop (hold the mouse or finger down and move the mouse) ',\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()\n .describe('The input field to be cleared')\n .optional(),\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 condition: z.string().describe('The condition of the assertion'),\n thought: z\n .string()\n .describe(\n 'The thought of the assertion, like \"I can see there are A, B, C elements on the page, which means ... , so the assertion is true\"',\n ),\n result: z.boolean().describe('The result of the assertion, true or false'),\n});\nexport type ActionAssertParam = {\n condition: string;\n thought: string;\n result: boolean;\n};\n\nexport const defineActionAssert = (): DeviceAction<ActionAssertParam> => {\n return defineAction<typeof actionAssertParamSchema, ActionAssertParam>({\n name: 'Print_Assert_Result',\n description:\n 'Print the result of the assertion. Use this only when the user asks for an assertion',\n paramSchema: actionAssertParamSchema,\n call: async (param) => {\n if (typeof param?.result !== 'boolean') {\n throw new Error(\n `The result of the assertion must be a boolean, but got: ${typeof param?.result}. ${param.thought || '(no thought)'}`,\n );\n }\n\n getDebug('device:common-action')(\n `Assert: ${param.condition}, Thought: ${param.thought}, Result: ${param.result}`,\n );\n\n if (!param.result) {\n throw new Error(\n `Assertion failed: ${param.thought || '(no thought)'} (Assertion = ${param.condition})`,\n );\n }\n },\n });\n};\n\n// Sleep\nexport const ActionSleepParamSchema = z.object({\n millisecond: z\n .number()\n .default(1000)\n .optional()\n .describe('Sleep duration in milliseconds, defaults to 1000ms (1 second)'),\n});\n\nexport type ActionSleepParam = {\n millisecond?: number;\n};\n\nexport const defineActionSleep = (): DeviceAction<ActionSleepParam> => {\n return defineAction<typeof ActionSleepParamSchema, ActionSleepParam>({\n name: 'Sleep',\n description:\n 'Wait for a specified duration before continuing. Defaults to 1 second (1000ms) if not specified.',\n paramSchema: ActionSleepParamSchema,\n call: async (param) => {\n const duration = param?.millisecond ?? 1000;\n getDebug('device:common-action')(`Sleeping for ${duration}ms`);\n await new Promise((resolve) => setTimeout(resolve, duration));\n },\n });\n};\n\n// Finalize\nexport const actionFinalizeParamSchema = z.object({\n message: z\n .string()\n .optional()\n .describe(\n 'The conclusion, data, or return value that the user needs. This message will be provided to the user when the task is finalized.',\n ),\n});\nexport type ActionFinalizeParam = {\n message?: string;\n};\n\nexport const defineActionFinalize = (): DeviceAction<ActionFinalizeParam> => {\n return defineAction<typeof actionFinalizeParamSchema, ActionFinalizeParam>({\n name: finalizeActionName,\n description:\n 'Finalize the task. You can provide the conclusion, data, or return value that the user needs in the message.',\n paramSchema: actionFinalizeParamSchema,\n call: async (param) => {\n getDebug('device:common-action')(\n `Task finalized${param?.message ? `: ${param.message}` : ''}`,\n );\n return param.message;\n },\n });\n};\n\nexport type { DeviceAction } from '../types';\nexport type {\n AndroidDeviceOpt,\n AndroidDeviceInputOpt,\n IOSDeviceOpt,\n IOSDeviceInputOpt,\n} from './device-options';\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","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","param","Error","getDebug","ActionSleepParamSchema","defineActionSleep","duration","Promise","resolve","setTimeout","actionFinalizeParamSchema","defineActionFinalize","finalizeActionName"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;ACWO,MAAeI;AAwCtB;AAKO,MAAMC,eAAe,CAK1BC,SAaOA;AAIF,MAAMC,uBAAuBC,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC3C,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMC,kBAAkB,CAC7BC,OAEON,aAA0D;QAC/D,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaE;QACbI;IACF;AAIK,MAAMC,8BAA8BJ,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAClD,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMI,yBAAyB,CACpCF,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaO;QACbD;IACF;AAIK,MAAMG,+BAA+BN,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IACnD,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMM,0BAA0B,CACrCJ,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaS;QACbH;IACF;AAIK,MAAMK,yBAAyBR,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC7C,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,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,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC7C,OAAOA,6BAAAA,CAAAA,CAAAA,KACC,CAAC;QAACA,6BAAAA,CAAAA,CAAAA,MAAQ;QAAIA,6BAAAA,CAAAA,CAAAA,MAAQ;KAAG,EAC9B,SAAS,CAAC,CAACY,MAAQC,OAAOD,MAC1B,QAAQ,CACP;IAEJ,QAAQX,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IACL,QAAQ,CAACS,wBACT,QAAQ;IACX,MAAMV,6BAAAA,CAAAA,CAAAA,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,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IACrD,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IACL,QAAQ,CAAC,qDACT,QAAQ;IACX,SAASD,6BAAAA,CAAAA,CAAAA,MACA,GACN,QAAQ,CACP;AAEN;AAMO,MAAMgB,4BAA4B,CACvCb,OAEON,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAakB;QACbZ;IACF;AAIK,MAAMc,0BAA0BjB,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC9C,YAAYA,6BAAAA,CAAAA,CAAAA,OACL,CAAC;QACJ;QACA;QACA;QACA;QACA;KACD,EACA,OAAO,CAAC,gBACR,QAAQ,CACP;IAEJ,WAAWA,6BAAAA,CAAAA,CAAAA,OACJ,CAAC;QAAC;QAAQ;QAAM;QAAS;KAAO,EACpC,OAAO,CAAC,QACR,QAAQ,CACP;IAEJ,UAAUA,6BAAAA,CAAAA,CAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,CAAC;IACZ,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IACL,QAAQ,GACR,QAAQ,CACP;AAEN;AAEO,MAAMiB,qBAAqB,CAChCf,OAEON,aAAgE;QACrE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAaoB;QACbd;IACF;AAIK,MAAMgB,+BAA+BnB,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IACnD,MAAMC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,QAAQ,CAAC;IAC3C,IAAIA,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,QAAQ,CAAC;AAC3C;AAMO,MAAMmB,0BAA0B,CACrCjB,OAEON,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAasB;QACbhB;IACF;AAGK,MAAMkB,6BAA6BrB,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IACjD,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,QAAQ,CAC1C;IAEF,UAAUD,6BAAAA,CAAAA,CAAAA,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,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC7C,OAAOC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IACJ,QAAQ,GACR,QAAQ,CACP;IAEJ,WAAWD,6BAAAA,CAAAA,CAAAA,OACJ,CAAC;QAAC;QAAM;QAAQ;QAAQ;KAAQ,EACpC,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUA,6BAAAA,CAAAA,CAAAA,MACD,GACN,QAAQ,GACR,QAAQ,CAAC;IACZ,KAAKC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IACF,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUD,6BAAAA,CAAAA,CAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,CAAC;IACZ,QAAQA,6BAAAA,CAAAA,CAAAA,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,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAClD,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IACL,QAAQ,CAAC,iCACT,QAAQ;AACb;AAKO,MAAMyB,yBAAyB,CACpCvB,OAEON,aAGL;QACA,MAAM;QACN,aAAaa;QACb,gBAAgB;QAChB,aAAae;QACbtB;IACF;AAIK,MAAMwB,0BAA0B3B,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC9C,WAAWA,6BAAAA,CAAAA,CAAAA,MAAQ,GAAG,QAAQ,CAAC;IAC/B,SAASA,6BAAAA,CAAAA,CAAAA,MACA,GACN,QAAQ,CACP;IAEJ,QAAQA,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ,CAAC;AAC/B;AAOO,MAAM4B,qBAAqB,IACzB/B,aAAgE;QACrE,MAAM;QACN,aACE;QACF,aAAa8B;QACb,MAAM,OAAOE;YACX,IAAI,AAAyB,aAAzB,OAAOA,OAAO,QAChB,MAAM,IAAIC,MACR,CAAC,wDAAwD,EAAE,OAAOD,OAAO,OAAO,EAAE,EAAEA,MAAM,OAAO,IAAI,gBAAgB;YAIzHE,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS,wBACP,CAAC,QAAQ,EAAEF,MAAM,SAAS,CAAC,WAAW,EAAEA,MAAM,OAAO,CAAC,UAAU,EAAEA,MAAM,MAAM,EAAE;YAGlF,IAAI,CAACA,MAAM,MAAM,EACf,MAAM,IAAIC,MACR,CAAC,kBAAkB,EAAED,MAAM,OAAO,IAAI,eAAe,cAAc,EAAEA,MAAM,SAAS,CAAC,CAAC,CAAC;QAG7F;IACF;AAIK,MAAMG,yBAAyBhC,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC7C,aAAaA,6BAAAA,CAAAA,CAAAA,MACJ,GACN,OAAO,CAAC,MACR,QAAQ,GACR,QAAQ,CAAC;AACd;AAMO,MAAMiC,oBAAoB,IACxBpC,aAA8D;QACnE,MAAM;QACN,aACE;QACF,aAAamC;QACb,MAAM,OAAOH;YACX,MAAMK,WAAWL,OAAO,eAAe;YACvCE,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS,wBAAwB,CAAC,aAAa,EAAEG,SAAS,EAAE,CAAC;YAC7D,MAAM,IAAIC,QAAQ,CAACC,UAAYC,WAAWD,SAASF;QACrD;IACF;AAIK,MAAMI,4BAA4BtC,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAChD,SAASA,6BAAAA,CAAAA,CAAAA,MACA,GACN,QAAQ,GACR,QAAQ,CACP;AAEN;AAKO,MAAMuC,uBAAuB,IAC3B1C,aAAoE;QACzE,MAAM2C,mCAAAA,kBAAkBA;QACxB,aACE;QACF,aAAaF;QACb,MAAM,OAAOT;YACXE,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS,wBACP,CAAC,cAAc,EAAEF,OAAO,UAAU,CAAC,EAAE,EAAEA,MAAM,OAAO,EAAE,GAAG,IAAI;YAE/D,OAAOA,MAAM,OAAO;QACtB;IACF"}
|
|
1
|
+
{"version":3,"file":"device/index.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../src/device/index.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import { finalizeActionName, 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 interface FileChooserHandler {\n accept(files: string[]): Promise<void>;\n}\n\nexport abstract class AbstractInterface {\n abstract interfaceType: string;\n\n abstract screenshotBase64(): Promise<string>;\n abstract size(): Promise<Size>;\n abstract actionSpace(): 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 // for web only\n registerFileChooserListener?(\n handler: (chooser: FileChooserHandler) => Promise<void>,\n ): Promise<{ dispose: () => void; getError: () => Error | undefined }>;\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});\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.preprocess(\n (val) => (val === 'append' ? 'typeOnly' : val),\n z\n .enum(['replace', 'clear', 'typeOnly'])\n .default('replace')\n .optional()\n .describe(\n 'Input mode: \"replace\" (default) - clear the field and input the value; \"typeOnly\" - type the value directly without clearing the field first; \"clear\" - clear the field without inputting new text.',\n ),\n ),\n});\nexport type ActionInputParam = {\n value: string;\n locate?: LocateResultElement;\n mode?: 'replace' | 'clear' | 'typeOnly' | '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 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 direction: z\n .enum(['down', 'up', 'right', 'left'])\n .default('down')\n .describe(\n 'The direction to scroll. Only effective when scrollType is \"singleAction\".',\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(\n 'Describe the target element to be scrolled on, like \"the table\" or \"the list\" or \"the content area\" or \"the scrollable area\". Do NOT provide a general intent like \"scroll to find some element\"',\n ),\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:\n 'Drag and drop (hold the mouse or finger down and move the mouse) ',\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()\n .describe('The input field to be cleared')\n .optional(),\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 condition: z.string().describe('The condition of the assertion'),\n thought: z\n .string()\n .describe(\n 'The thought of the assertion, like \"I can see there are A, B, C elements on the page, which means ... , so the assertion is true\"',\n ),\n result: z.boolean().describe('The result of the assertion, true or false'),\n});\nexport type ActionAssertParam = {\n condition: string;\n thought: string;\n result: boolean;\n};\n\nexport const defineActionAssert = (): DeviceAction<ActionAssertParam> => {\n return defineAction<typeof actionAssertParamSchema, ActionAssertParam>({\n name: 'Print_Assert_Result',\n description:\n 'Print the result of the assertion. Use this only when the user asks for an assertion',\n paramSchema: actionAssertParamSchema,\n call: async (param) => {\n if (typeof param?.result !== 'boolean') {\n throw new Error(\n `The result of the assertion must be a boolean, but got: ${typeof param?.result}. ${param.thought || '(no thought)'}`,\n );\n }\n\n getDebug('device:common-action')(\n `Assert: ${param.condition}, Thought: ${param.thought}, Result: ${param.result}`,\n );\n\n if (!param.result) {\n throw new Error(\n `Assertion failed: ${param.thought || '(no thought)'} (Assertion = ${param.condition})`,\n );\n }\n },\n });\n};\n\n// Sleep\nexport const ActionSleepParamSchema = z.object({\n millisecond: z\n .number()\n .default(1000)\n .optional()\n .describe('Sleep duration in milliseconds, defaults to 1000ms (1 second)'),\n});\n\nexport type ActionSleepParam = {\n millisecond?: number;\n};\n\nexport const defineActionSleep = (): DeviceAction<ActionSleepParam> => {\n return defineAction<typeof ActionSleepParamSchema, ActionSleepParam>({\n name: 'Sleep',\n description:\n 'Wait for a specified duration before continuing. Defaults to 1 second (1000ms) if not specified.',\n paramSchema: ActionSleepParamSchema,\n call: async (param) => {\n const duration = param?.millisecond ?? 1000;\n getDebug('device:common-action')(`Sleeping for ${duration}ms`);\n await new Promise((resolve) => setTimeout(resolve, duration));\n },\n });\n};\n\n// Finalize\nexport const actionFinalizeParamSchema = z.object({\n message: z\n .string()\n .optional()\n .describe(\n 'The conclusion, data, or return value that the user needs. This message will be provided to the user when the task is finalized.',\n ),\n});\nexport type ActionFinalizeParam = {\n message?: string;\n};\n\nexport const defineActionFinalize = (): DeviceAction<ActionFinalizeParam> => {\n return defineAction<typeof actionFinalizeParamSchema, ActionFinalizeParam>({\n name: finalizeActionName,\n description:\n 'Finalize the task. You can provide the conclusion, data, or return value that the user needs in the message.',\n paramSchema: actionFinalizeParamSchema,\n call: async (param) => {\n getDebug('device:common-action')(\n `Task finalized${param?.message ? `: ${param.message}` : ''}`,\n );\n return param.message;\n },\n });\n};\n\nexport type { DeviceAction } from '../types';\nexport type {\n AndroidDeviceOpt,\n AndroidDeviceInputOpt,\n IOSDeviceOpt,\n IOSDeviceInputOpt,\n} from './device-options';\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","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","param","Error","getDebug","ActionSleepParamSchema","defineActionSleep","duration","Promise","resolve","setTimeout","actionFinalizeParamSchema","defineActionFinalize","finalizeActionName"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;ACWO,MAAeI;AAwCtB;AAKO,MAAMC,eAAe,CAK1BC,SAaOA;AAIF,MAAMC,uBAAuBC,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC3C,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMC,kBAAkB,CAC7BC,OAEON,aAA0D;QAC/D,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaE;QACbI;IACF;AAIK,MAAMC,8BAA8BJ,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAClD,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMI,yBAAyB,CACpCF,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaO;QACbD;IACF;AAIK,MAAMG,+BAA+BN,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IACnD,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMM,0BAA0B,CACrCJ,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaS;QACbH;IACF;AAIK,MAAMK,yBAAyBR,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC7C,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,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,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC7C,OAAOA,6BAAAA,CAAAA,CAAAA,KACC,CAAC;QAACA,6BAAAA,CAAAA,CAAAA,MAAQ;QAAIA,6BAAAA,CAAAA,CAAAA,MAAQ;KAAG,EAC9B,SAAS,CAAC,CAACY,MAAQC,OAAOD,MAC1B,QAAQ,CACP;IAEJ,QAAQX,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IACL,QAAQ,CAACS,wBACT,QAAQ;IACX,MAAMV,6BAAAA,CAAAA,CAAAA,UAAY,CAChB,CAACY,MAASA,AAAQ,aAARA,MAAmB,aAAaA,KAC1CZ,6BAAAA,CAAAA,CAAAA,OACO,CAAC;QAAC;QAAW;QAAS;KAAW,EACrC,OAAO,CAAC,WACR,QAAQ,GACR,QAAQ,CACP;AAGR;AAOO,MAAMc,oBAAoB,CAC/BX,OAEON,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAac;QACbR;IACF;AAIK,MAAMY,iCAAiCf,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IACrD,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IACL,QAAQ,CAAC,qDACT,QAAQ;IACX,SAASD,6BAAAA,CAAAA,CAAAA,MACA,GACN,QAAQ,CACP;AAEN;AAMO,MAAMgB,4BAA4B,CACvCb,OAEON,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAakB;QACbZ;IACF;AAIK,MAAMc,0BAA0BjB,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC9C,YAAYA,6BAAAA,CAAAA,CAAAA,OACL,CAAC;QACJ;QACA;QACA;QACA;QACA;KACD,EACA,OAAO,CAAC,gBACR,QAAQ,CACP;IAEJ,WAAWA,6BAAAA,CAAAA,CAAAA,OACJ,CAAC;QAAC;QAAQ;QAAM;QAAS;KAAO,EACpC,OAAO,CAAC,QACR,QAAQ,CACP;IAEJ,UAAUA,6BAAAA,CAAAA,CAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,CAAC;IACZ,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IACL,QAAQ,GACR,QAAQ,CACP;AAEN;AAEO,MAAMiB,qBAAqB,CAChCf,OAEON,aAAgE;QACrE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAaoB;QACbd;IACF;AAIK,MAAMgB,+BAA+BnB,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IACnD,MAAMC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,QAAQ,CAAC;IAC3C,IAAIA,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,QAAQ,CAAC;AAC3C;AAMO,MAAMmB,0BAA0B,CACrCjB,OAEON,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAasB;QACbhB;IACF;AAGK,MAAMkB,6BAA6BrB,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IACjD,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IAA4B,QAAQ,CAC1C;IAEF,UAAUD,6BAAAA,CAAAA,CAAAA,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,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC7C,OAAOC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IACJ,QAAQ,GACR,QAAQ,CACP;IAEJ,WAAWD,6BAAAA,CAAAA,CAAAA,OACJ,CAAC;QAAC;QAAM;QAAQ;QAAQ;KAAQ,EACpC,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUA,6BAAAA,CAAAA,CAAAA,MACD,GACN,QAAQ,GACR,QAAQ,CAAC;IACZ,KAAKC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IACF,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUD,6BAAAA,CAAAA,CAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,CAAC;IACZ,QAAQA,6BAAAA,CAAAA,CAAAA,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,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAClD,QAAQC,AAAAA,IAAAA,mCAAAA,yBAAAA,AAAAA,IACL,QAAQ,CAAC,iCACT,QAAQ;AACb;AAKO,MAAMyB,yBAAyB,CACpCvB,OAEON,aAGL;QACA,MAAM;QACN,aAAaa;QACb,gBAAgB;QAChB,aAAae;QACbtB;IACF;AAIK,MAAMwB,0BAA0B3B,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC9C,WAAWA,6BAAAA,CAAAA,CAAAA,MAAQ,GAAG,QAAQ,CAAC;IAC/B,SAASA,6BAAAA,CAAAA,CAAAA,MACA,GACN,QAAQ,CACP;IAEJ,QAAQA,6BAAAA,CAAAA,CAAAA,OAAS,GAAG,QAAQ,CAAC;AAC/B;AAOO,MAAM4B,qBAAqB,IACzB/B,aAAgE;QACrE,MAAM;QACN,aACE;QACF,aAAa8B;QACb,MAAM,OAAOE;YACX,IAAI,AAAyB,aAAzB,OAAOA,OAAO,QAChB,MAAM,IAAIC,MACR,CAAC,wDAAwD,EAAE,OAAOD,OAAO,OAAO,EAAE,EAAEA,MAAM,OAAO,IAAI,gBAAgB;YAIzHE,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS,wBACP,CAAC,QAAQ,EAAEF,MAAM,SAAS,CAAC,WAAW,EAAEA,MAAM,OAAO,CAAC,UAAU,EAAEA,MAAM,MAAM,EAAE;YAGlF,IAAI,CAACA,MAAM,MAAM,EACf,MAAM,IAAIC,MACR,CAAC,kBAAkB,EAAED,MAAM,OAAO,IAAI,eAAe,cAAc,EAAEA,MAAM,SAAS,CAAC,CAAC,CAAC;QAG7F;IACF;AAIK,MAAMG,yBAAyBhC,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAC7C,aAAaA,6BAAAA,CAAAA,CAAAA,MACJ,GACN,OAAO,CAAC,MACR,QAAQ,GACR,QAAQ,CAAC;AACd;AAMO,MAAMiC,oBAAoB,IACxBpC,aAA8D;QACnE,MAAM;QACN,aACE;QACF,aAAamC;QACb,MAAM,OAAOH;YACX,MAAMK,WAAWL,OAAO,eAAe;YACvCE,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS,wBAAwB,CAAC,aAAa,EAAEG,SAAS,EAAE,CAAC;YAC7D,MAAM,IAAIC,QAAQ,CAACC,UAAYC,WAAWD,SAASF;QACrD;IACF;AAIK,MAAMI,4BAA4BtC,6BAAAA,CAAAA,CAAAA,MAAQ,CAAC;IAChD,SAASA,6BAAAA,CAAAA,CAAAA,MACA,GACN,QAAQ,GACR,QAAQ,CACP;AAEN;AAKO,MAAMuC,uBAAuB,IAC3B1C,aAAoE;QACzE,MAAM2C,mCAAAA,kBAAkBA;QACxB,aACE;QACF,aAAaF;QACb,MAAM,OAAOT;YACXE,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS,wBACP,CAAC,cAAc,EAAEF,OAAO,UAAU,CAAC,EAAE,EAAEA,MAAM,OAAO,EAAE,GAAG,IAAI;YAE/D,OAAOA,MAAM,OAAO;QACtB;IACF"}
|
|
@@ -57,12 +57,12 @@ class Service {
|
|
|
57
57
|
if (globalDeepThinkSwitch) debug('globalDeepThinkSwitch', globalDeepThinkSwitch);
|
|
58
58
|
let searchAreaPrompt;
|
|
59
59
|
if (query.deepThink || globalDeepThinkSwitch) searchAreaPrompt = query.prompt;
|
|
60
|
-
const {
|
|
61
|
-
if (searchAreaPrompt && !
|
|
60
|
+
const { modelFamily } = modelConfig;
|
|
61
|
+
if (searchAreaPrompt && !modelFamily) {
|
|
62
62
|
console.warn('The "deepThink" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/model-config');
|
|
63
63
|
searchAreaPrompt = void 0;
|
|
64
64
|
}
|
|
65
|
-
if (searchAreaPrompt && (0, util_js_namespaceObject.isAutoGLM)(
|
|
65
|
+
if (searchAreaPrompt && (0, util_js_namespaceObject.isAutoGLM)(modelFamily)) {
|
|
66
66
|
console.warn('The "deepThink" feature is not supported with AutoGLM.');
|
|
67
67
|
searchAreaPrompt = void 0;
|
|
68
68
|
}
|
|
@@ -189,7 +189,7 @@ class Service {
|
|
|
189
189
|
const { size } = context;
|
|
190
190
|
const screenshotBase64 = context.screenshot.base64;
|
|
191
191
|
(0, utils_namespaceObject.assert)(screenshotBase64, 'screenshot is required for service.describe');
|
|
192
|
-
const {
|
|
192
|
+
const { modelFamily } = modelConfig;
|
|
193
193
|
const systemPrompt = (0, describe_js_namespaceObject.elementDescriberInstruction)();
|
|
194
194
|
const defaultRectSize = 30;
|
|
195
195
|
const targetRect = Array.isArray(target) ? {
|
|
@@ -209,9 +209,9 @@ class Service {
|
|
|
209
209
|
borderThickness: 3
|
|
210
210
|
});
|
|
211
211
|
if (opt?.deepThink) {
|
|
212
|
-
const searchArea = (0, external_common_js_namespaceObject.expandSearchArea)(targetRect, context.size,
|
|
212
|
+
const searchArea = (0, external_common_js_namespaceObject.expandSearchArea)(targetRect, context.size, modelFamily);
|
|
213
213
|
debug('describe: set searchArea', searchArea);
|
|
214
|
-
const croppedResult = await (0, img_namespaceObject.cropByRect)(imagePayload, searchArea, 'qwen2.5-vl' ===
|
|
214
|
+
const croppedResult = await (0, img_namespaceObject.cropByRect)(imagePayload, searchArea, 'qwen2.5-vl' === modelFamily);
|
|
215
215
|
imagePayload = croppedResult.imageBase64;
|
|
216
216
|
}
|
|
217
217
|
const msgs = [
|