@midscene/core 1.0.1-beta-20251124014653.0 → 1.0.1-beta-20251124032544.0

package/dist/es/ai-model/conversation-history.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/conversation-history.mjs","sources":["webpack://@midscene/core/./src/ai-model/conversation-history.ts"],"sourcesContent":["import type { ChatCompletionMessageParam } from 'openai/resources/index';\n\nexport interface ConversationHistoryOptions {\n  maxUserImageMessages?: number;\n  initialMessages?: ChatCompletionMessageParam[];\n}\n\nexport class ConversationHistory {\n  private readonly maxUserImageMessages: number;\n  private readonly messages: ChatCompletionMessageParam[] = [];\n\n  constructor(options?: ConversationHistoryOptions) {\n    this.maxUserImageMessages = options?.maxUserImageMessages ?? 4;\n    if (options?.initialMessages?.length) {\n      this.seed(options.initialMessages);\n    }\n  }\n\n  append(message: ChatCompletionMessageParam) {\n    if (message.role === 'user') {\n      this.pruneOldestUserMessageIfNecessary();\n    }\n\n    this.messages.push(message);\n  }\n\n  seed(messages: ChatCompletionMessageParam[]) {\n    this.reset();\n    messages.forEach((message) => {\n      this.append(message);\n    });\n  }\n\n  reset() {\n    this.messages.length = 0;\n  }\n\n  snapshot(): ChatCompletionMessageParam[] {\n    return [...this.messages];\n  }\n\n  get length(): number {\n    return this.messages.length;\n  }\n\n  [Symbol.iterator](): IterableIterator<ChatCompletionMessageParam> {\n    return this.messages[Symbol.iterator]();\n  }\n\n  toJSON(): ChatCompletionMessageParam[] {\n    return this.snapshot();\n  }\n\n  private pruneOldestUserMessageIfNecessary() {\n    const userMessages = this.messages.filter((item) => item.role === 'user');\n    if (userMessages.length < this.maxUserImageMessages) {\n      return;\n    }\n\n    const firstUserMessageIndex = this.messages.findIndex(\n      (item) => item.role === 'user',\n    );\n\n    if (firstUserMessageIndex >= 0) {\n      this.messages.splice(firstUserMessageIndex, 1);\n    }\n  }\n}\n"],"names":["Symbol","ConversationHistory","message","messages","userMessages","item","firstUserMessageIndex","options","_options_initialMessages"],"mappings":";;;;;;;;;;;eA6CGA,OAAO,QAAQ;;AAtCX,MAAMC;IAWX,OAAOC,OAAmC,EAAE;QAC1C,IAAIA,AAAiB,WAAjBA,QAAQ,IAAI,EACd,IAAI,CAAC,iCAAiC;QAGxC,IAAI,CAAC,QAAQ,CAAC,IAAI,CAACA;IACrB;IAEA,KAAKC,QAAsC,EAAE;QAC3C,IAAI,CAAC,KAAK;QACVA,SAAS,OAAO,CAAC,CAACD;YAChB,IAAI,CAAC,MAAM,CAACA;QACd;IACF;IAEA,QAAQ;QACN,IAAI,CAAC,QAAQ,CAAC,MAAM,GAAG;IACzB;IAEA,WAAyC;QACvC,OAAO;eAAI,IAAI,CAAC,QAAQ;SAAC;IAC3B;IAEA,IAAI,SAAiB;QACnB,OAAO,IAAI,CAAC,QAAQ,CAAC,MAAM;IAC7B;IAEA,CAAC,cAAD,GAAkE;QAChE,OAAO,IAAI,CAAC,QAAQ,CAACF,OAAO,QAAQ,CAAC;IACvC;IAEA,SAAuC;QACrC,OAAO,IAAI,CAAC,QAAQ;IACtB;IAEQ,oCAAoC;QAC1C,MAAMI,eAAe,IAAI,CAAC,QAAQ,CAAC,MAAM,CAAC,CAACC,OAASA,AAAc,WAAdA,KAAK,IAAI;QAC7D,IAAID,aAAa,MAAM,GAAG,IAAI,CAAC,oBAAoB,EACjD;QAGF,MAAME,wBAAwB,IAAI,CAAC,QAAQ,CAAC,SAAS,CACnD,CAACD,OAASA,AAAc,WAAdA,KAAK,IAAI;QAGrB,IAAIC,yBAAyB,GAC3B,IAAI,CAAC,QAAQ,CAAC,MAAM,CAACA,uBAAuB;IAEhD;IAvDA,YAAYC,OAAoC,CAAE;YAE5CC;QALN,uBAAiB,wBAAjB;QACA,uBAAiB,YAAyC,EAAE;QAG1D,IAAI,CAAC,oBAAoB,GAAGD,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,oBAAoB,AAAD,KAAK;QAC7D,IAAIC,QAAAA,UAAAA,KAAAA,IAAAA,QAAAA,CAAAA,2BAAAA,QAAS,eAAe,AAAD,IAAvBA,KAAAA,IAAAA,yBAA0B,MAAM,EAClC,IAAI,CAAC,IAAI,CAACD,QAAQ,eAAe;IAErC;AAmDF"}
+{"version":3,"file":"ai-model/conversation-history.mjs","sources":["webpack://@midscene/core/./src/ai-model/conversation-history.ts"],"sourcesContent":["import type { ChatCompletionMessageParam } from 'openai/resources/index';\n\nexport interface ConversationHistoryOptions {\n  maxUserImageMessages?: number;\n  initialMessages?: ChatCompletionMessageParam[];\n}\n\nconst defaultMaxUserImagesCount = 6;\n\nexport class ConversationHistory {\n  private readonly maxUserImageMessages: number;\n  private readonly messages: ChatCompletionMessageParam[] = [];\n\n  constructor(options?: ConversationHistoryOptions) {\n    this.maxUserImageMessages =\n      options?.maxUserImageMessages ?? defaultMaxUserImagesCount;\n    if (options?.initialMessages?.length) {\n      this.seed(options.initialMessages);\n    }\n  }\n\n  append(message: ChatCompletionMessageParam) {\n    this.messages.push(message);\n  }\n\n  seed(messages: ChatCompletionMessageParam[]) {\n    this.reset();\n    messages.forEach((message) => {\n      this.append(message);\n    });\n  }\n\n  reset() {\n    this.messages.length = 0;\n  }\n\n  snapshot(options?: {\n    maxImageMessages?: number;\n  }): ChatCompletionMessageParam[] {\n    const maxImageMessages =\n      options?.maxImageMessages ?? this.maxUserImageMessages;\n\n    // Count image_url messages from back to front\n    let imageCount = 0;\n    const processedMessages = [...this.messages]\n      .reverse()\n      .map((message): ChatCompletionMessageParam => {\n        if (\n          typeof message.content !== 'string' &&\n          Array.isArray(message.content)\n        ) {\n          // Also process content items from back to front\n          const processedContent = [...message.content]\n            .reverse()\n            .map((item) => {\n              if (item.type === 'image_url') {\n                imageCount++;\n                if (imageCount > maxImageMessages) {\n                  // Replace with text type\n                  return {\n                    type: 'text' as const,\n                    text: '(omitted due to size limit)',\n                  };\n                }\n              }\n              return item;\n            })\n            .reverse();\n          return {\n            ...message,\n            content: processedContent,\n          } as ChatCompletionMessageParam;\n        }\n        return message;\n      });\n\n    return processedMessages.reverse();\n  }\n\n  get length(): number {\n    return this.messages.length;\n  }\n\n  [Symbol.iterator](): IterableIterator<ChatCompletionMessageParam> {\n    return this.messages[Symbol.iterator]();\n  }\n\n  toJSON(): ChatCompletionMessageParam[] {\n    return this.snapshot();\n  }\n}\n"],"names":["defaultMaxUserImagesCount","Symbol","ConversationHistory","message","messages","options","maxImageMessages","imageCount","processedMessages","Array","processedContent","item","_options_initialMessages"],"mappings":";;;;;;;;;;;AAOA,MAAMA,4BAA4B;eA4E/BC,OAAO,QAAQ;;AA1EX,MAAMC;IAYX,OAAOC,OAAmC,EAAE;QAC1C,IAAI,CAAC,QAAQ,CAAC,IAAI,CAACA;IACrB;IAEA,KAAKC,QAAsC,EAAE;QAC3C,IAAI,CAAC,KAAK;QACVA,SAAS,OAAO,CAAC,CAACD;YAChB,IAAI,CAAC,MAAM,CAACA;QACd;IACF;IAEA,QAAQ;QACN,IAAI,CAAC,QAAQ,CAAC,MAAM,GAAG;IACzB;IAEA,SAASE,OAER,EAAgC;QAC/B,MAAMC,mBACJD,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,gBAAgB,AAAD,KAAK,IAAI,CAAC,oBAAoB;QAGxD,IAAIE,aAAa;QACjB,MAAMC,oBAAoB;eAAI,IAAI,CAAC,QAAQ;SAAC,CACzC,OAAO,GACP,GAAG,CAAC,CAACL;YACJ,IACE,AAA2B,YAA3B,OAAOA,QAAQ,OAAO,IACtBM,MAAM,OAAO,CAACN,QAAQ,OAAO,GAC7B;gBAEA,MAAMO,mBAAmB;uBAAIP,QAAQ,OAAO;iBAAC,CAC1C,OAAO,GACP,GAAG,CAAC,CAACQ;oBACJ,IAAIA,AAAc,gBAAdA,KAAK,IAAI,EAAkB;wBAC7BJ;wBACA,IAAIA,aAAaD,kBAEf,OAAO;4BACL,MAAM;4BACN,MAAM;wBACR;oBAEJ;oBACA,OAAOK;gBACT,GACC,OAAO;gBACV,OAAO;oBACL,GAAGR,OAAO;oBACV,SAASO;gBACX;YACF;YACA,OAAOP;QACT;QAEF,OAAOK,kBAAkB,OAAO;IAClC;IAEA,IAAI,SAAiB;QACnB,OAAO,IAAI,CAAC,QAAQ,CAAC,MAAM;IAC7B;IAEA,CAAC,cAAD,GAAkE;QAChE,OAAO,IAAI,CAAC,QAAQ,CAACP,OAAO,QAAQ,CAAC;IACvC;IAEA,SAAuC;QACrC,OAAO,IAAI,CAAC,QAAQ;IACtB;IA5EA,YAAYI,OAAoC,CAAE;YAG5CO;QANN,uBAAiB,wBAAjB;QACA,uBAAiB,YAAyC,EAAE;QAG1D,IAAI,CAAC,oBAAoB,GACvBP,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,oBAAoB,AAAD,KAAKL;QACnC,IAAIY,QAAAA,UAAAA,KAAAA,IAAAA,QAAAA,CAAAA,2BAAAA,QAAS,eAAe,AAAD,IAAvBA,KAAAA,IAAAA,yBAA0B,MAAM,EAClC,IAAI,CAAC,IAAI,CAACP,QAAQ,eAAe;IAErC;AAuEF"}

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

@@ -13,7 +13,8 @@ async function plan(userInstruction, opts) {
     const systemPrompt = await systemPromptToTaskPlanning({
         actionSpace: opts.actionSpace,
         vlMode,
-        includeBbox: opts.includeBbox
+        includeBbox: opts.includeBbox,
+        thinkingStrategy: opts.thinkingStrategy
     });
     let imagePayload = screenshotBase64;
     let imageWidth = size.width;
@@ -49,6 +50,22 @@ async function plan(userInstruction, opts) {
             ]
         }
     ];
+    const latestImageMessage = {
+        role: 'user',
+        content: [
+            {
+                type: 'text',
+                text: 'I have finished the action previously planned, and the last screenshot is as follows:'
+            },
+            {
+                type: 'image_url',
+                image_url: {
+                    url: imagePayload,
+                    detail: 'high'
+                }
+            }
+        ]
+    };
     const msgs = [
         {
             role: 'system',
@@ -57,18 +74,7 @@ async function plan(userInstruction, opts) {
         ...knowledgeContext,
         ...instruction,
         ...historyLog,
-        {
-            role: 'user',
-            content: [
-                {
-                    type: 'image_url',
-                    image_url: {
-                        url: imagePayload,
-                        detail: 'high'
-                    }
-                }
-            ]
-        }
+        latestImageMessage
     ];
     const { content: planFromAI, usage } = await callAIWithObjectResponse(msgs, AIActionType.PLAN, modelConfig);
     const rawResponse = JSON.stringify(planFromAI, void 0, 2);
@@ -96,6 +102,7 @@ async function plan(userInstruction, opts) {
     });
     assert(!planFromAI.error, `Failed to plan actions: ${planFromAI.error}`);
     if (0 === actions.length && returnValue.more_actions_needed_by_instruction && !returnValue.sleep) console.warn('No actions planned for the prompt, but model said more actions are needed:', userInstruction);
+    null == conversationHistory || conversationHistory.append(latestImageMessage);
     null == conversationHistory || conversationHistory.append({
         role: 'assistant',
         content: [
@@ -105,15 +112,6 @@ async function plan(userInstruction, opts) {
             }
         ]
     });
-    null == conversationHistory || conversationHistory.append({
-        role: 'user',
-        content: [
-            {
-                type: 'text',
-                text: 'I have finished the action previously planned'
-            }
-        ]
-    });
     return returnValue;
 }
 export { plan };

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

@@ -1 +1 @@
-{"version":3,"file":"ai-model/llm-planning.mjs","sources":["webpack://@midscene/core/./src/ai-model/llm-planning.ts"],"sourcesContent":["import type {\n  DeviceAction,\n  InterfaceType,\n  PlanningAIResponse,\n  RawResponsePlanningAIResponse,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport {\n  AIActionType,\n  buildYamlFlowFromPlans,\n  fillBboxParam,\n  findAllMidsceneLocatorField,\n} from '../common';\nimport type { ConversationHistory } from './conversation-history';\nimport { systemPromptToTaskPlanning } from './prompt/llm-planning';\nimport { callAIWithObjectResponse } from './service-caller/index';\n\nconst debug = getDebug('planning');\n\nexport async function plan(\n  userInstruction: string,\n  opts: {\n    context: UIContext;\n    interfaceType: InterfaceType;\n    actionSpace: DeviceAction<any>[];\n    actionContext?: string;\n    modelConfig: IModelConfig;\n    conversationHistory?: ConversationHistory;\n    includeBbox: boolean;\n  },\n): Promise<PlanningAIResponse> {\n  const { context, modelConfig, conversationHistory } = opts;\n  const { screenshotBase64, size } = context;\n\n  const { vlMode } = modelConfig;\n\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    vlMode,\n    includeBbox: opts.includeBbox,\n  });\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = size.width;\n  let imageHeight = size.height;\n  const rightLimit = imageWidth;\n  const bottomLimit = imageHeight;\n\n  // Process image based on VL mode requirements\n  if (vlMode === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const historyLog = opts.conversationHistory?.snapshot() || [];\n  // .filter((item) => item.role === 'assistant') || [];\n\n  const knowledgeContext: ChatCompletionMessageParam[] = opts.actionContext\n    ? [\n        {\n          role: 'user',\n          content: [\n            {\n              type: 'text',\n              text: `<high_priority_knowledge>${opts.actionContext}</high_priority_knowledge>`,\n            },\n          ],\n        },\n      ]\n    : [];\n\n  const instruction: ChatCompletionMessageParam[] = [\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `<user_instruction>${userInstruction}</user_instruction>`,\n        },\n      ],\n    },\n  ];\n\n  const msgs: ChatCompletionMessageParam[] = [\n    { role: 'system', content: systemPrompt },\n    ...knowledgeContext,\n    ...instruction,\n    ...historyLog,\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n        // Planning uses pure vision mode, no DOM description needed\n      ],\n    },\n  ];\n\n  const { content: planFromAI, usage } =\n    await callAIWithObjectResponse<RawResponsePlanningAIResponse>(\n      msgs,\n      AIActionType.PLAN,\n      modelConfig,\n    );\n  const rawResponse = JSON.stringify(planFromAI, undefined, 2);\n\n  const actions = planFromAI.action ? [planFromAI.action] : [];\n  const returnValue: PlanningAIResponse = {\n    ...planFromAI,\n    actions,\n    rawResponse,\n    usage,\n    yamlFlow: buildYamlFlowFromPlans(\n      actions,\n      opts.actionSpace,\n      planFromAI.sleep,\n    ),\n  };\n\n  assert(planFromAI, \"can't get plans from AI\");\n\n  actions.forEach((action) => {\n    const type = action.type;\n    const actionInActionSpace = opts.actionSpace.find(\n      (action) => action.name === type,\n    );\n\n    debug('actionInActionSpace matched', actionInActionSpace);\n    const locateFields = actionInActionSpace\n      ? findAllMidsceneLocatorField(actionInActionSpace.paramSchema)\n      : [];\n\n    debug('locateFields', locateFields);\n\n    locateFields.forEach((field) => {\n      const locateResult = action.param[field];\n      if (locateResult && vlMode !== undefined) {\n        // Always use VL mode to fill bbox parameters\n        action.param[field] = fillBboxParam(\n          locateResult,\n          imageWidth,\n          imageHeight,\n          rightLimit,\n          bottomLimit,\n          vlMode,\n        );\n      }\n    });\n  });\n  // in Qwen-VL, error means error. In GPT-4o, error may mean more actions are needed.\n  assert(!planFromAI.error, `Failed to plan actions: ${planFromAI.error}`);\n\n  if (\n    actions.length === 0 &&\n    returnValue.more_actions_needed_by_instruction &&\n    !returnValue.sleep\n  ) {\n    console.warn(\n      'No actions planned for the prompt, but model said more actions are needed:',\n      userInstruction,\n    );\n  }\n\n  conversationHistory?.append({\n    role: 'assistant',\n    content: [\n      {\n        type: 'text',\n        text: rawResponse,\n      },\n    ],\n  });\n  conversationHistory?.append({\n    role: 'user',\n    content: [\n      {\n        type: 'text',\n        text: 'I have finished the action previously planned',\n      },\n    ],\n  });\n\n  return returnValue;\n}\n"],"names":["debug","getDebug","plan","userInstruction","opts","_opts_conversationHistory","context","modelConfig","conversationHistory","screenshotBase64","size","vlMode","systemPrompt","systemPromptToTaskPlanning","imagePayload","imageWidth","imageHeight","rightLimit","bottomLimit","paddedResult","paddingToMatchBlockByBase64","historyLog","knowledgeContext","instruction","msgs","planFromAI","usage","callAIWithObjectResponse","AIActionType","rawResponse","JSON","undefined","actions","returnValue","buildYamlFlowFromPlans","assert","action","type","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam","console"],"mappings":";;;;;;AAsBA,MAAMA,QAAQC,SAAS;AAEhB,eAAeC,KACpBC,eAAuB,EACvBC,IAQC;QA2BkBC;IAzBnB,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAEC,mBAAmB,EAAE,GAAGJ;IACtD,MAAM,EAAEK,gBAAgB,EAAEC,IAAI,EAAE,GAAGJ;IAEnC,MAAM,EAAEK,MAAM,EAAE,GAAGJ;IAEnB,MAAMK,eAAe,MAAMC,2BAA2B;QACpD,aAAaT,KAAK,WAAW;QAC7BO;QACA,aAAaP,KAAK,WAAW;IAC/B;IAEA,IAAIU,eAAeL;IACnB,IAAIM,aAAaL,KAAK,KAAK;IAC3B,IAAIM,cAAcN,KAAK,MAAM;IAC7B,MAAMO,aAAaF;IACnB,MAAMG,cAAcF;IAGpB,IAAIL,AAAW,iBAAXA,QAAyB;QAC3B,MAAMQ,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAME,aAAahB,AAAAA,SAAAA,CAAAA,4BAAAA,KAAK,mBAAmB,AAAD,IAAvBA,KAAAA,IAAAA,0BAA0B,QAAQ,EAAC,KAAK,EAAE;IAG7D,MAAMiB,mBAAiDlB,KAAK,aAAa,GACrE;QACE;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,CAAC,yBAAyB,EAAEA,KAAK,aAAa,CAAC,0BAA0B,CAAC;gBAClF;aACD;QACH;KACD,GACD,EAAE;IAEN,MAAMmB,cAA4C;QAChD;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,CAAC,kBAAkB,EAAEpB,gBAAgB,mBAAmB,CAAC;gBACjE;aACD;QACH;KACD;IAED,MAAMqB,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASZ;QAAa;WACrCU;WACAC;WACAF;QACH;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKP;wBACL,QAAQ;oBACV;gBACF;aAED;QACH;KACD;IAED,MAAM,EAAE,SAASW,UAAU,EAAEC,KAAK,EAAE,GAClC,MAAMC,yBACJH,MACAI,aAAa,IAAI,EACjBrB;IAEJ,MAAMsB,cAAcC,KAAK,SAAS,CAACL,YAAYM,QAAW;IAE1D,MAAMC,UAAUP,WAAW,MAAM,GAAG;QAACA,WAAW,MAAM;KAAC,GAAG,EAAE;IAC5D,MAAMQ,cAAkC;QACtC,GAAGR,UAAU;QACbO;QACAH;QACAH;QACA,UAAUQ,uBACRF,SACA5B,KAAK,WAAW,EAChBqB,WAAW,KAAK;IAEpB;IAEAU,OAAOV,YAAY;IAEnBO,QAAQ,OAAO,CAAC,CAACI;QACf,MAAMC,OAAOD,OAAO,IAAI;QACxB,MAAME,sBAAsBlC,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACgC,SAAWA,OAAO,IAAI,KAAKC;QAG9BrC,MAAM,+BAA+BsC;QACrC,MAAMC,eAAeD,sBACjBE,4BAA4BF,oBAAoB,WAAW,IAC3D,EAAE;QAENtC,MAAM,gBAAgBuC;QAEtBA,aAAa,OAAO,CAAC,CAACE;YACpB,MAAMC,eAAeN,OAAO,KAAK,CAACK,MAAM;YACxC,IAAIC,gBAAgB/B,AAAWoB,WAAXpB,QAElByB,OAAO,KAAK,CAACK,MAAM,GAAGE,cACpBD,cACA3B,YACAC,aACAC,YACAC,aACAP;QAGN;IACF;IAEAwB,OAAO,CAACV,WAAW,KAAK,EAAE,CAAC,wBAAwB,EAAEA,WAAW,KAAK,EAAE;IAEvE,IACEO,AAAmB,MAAnBA,QAAQ,MAAM,IACdC,YAAY,kCAAkC,IAC9C,CAACA,YAAY,KAAK,EAElBW,QAAQ,IAAI,CACV,8EACAzC;IAIJK,QAAAA,uBAAAA,oBAAqB,MAAM,CAAC;QAC1B,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAMqB;YACR;SACD;IACH;IACArB,QAAAA,uBAAAA,oBAAqB,MAAM,CAAC;QAC1B,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAM;YACR;SACD;IACH;IAEA,OAAOyB;AACT"}
+{"version":3,"file":"ai-model/llm-planning.mjs","sources":["webpack://@midscene/core/./src/ai-model/llm-planning.ts"],"sourcesContent":["import type {\n  DeviceAction,\n  InterfaceType,\n  PlanningAIResponse,\n  RawResponsePlanningAIResponse,\n  ThinkingStrategy,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport {\n  AIActionType,\n  buildYamlFlowFromPlans,\n  fillBboxParam,\n  findAllMidsceneLocatorField,\n} from '../common';\nimport type { ConversationHistory } from './conversation-history';\nimport { systemPromptToTaskPlanning } from './prompt/llm-planning';\nimport { callAIWithObjectResponse } from './service-caller/index';\n\nconst debug = getDebug('planning');\n\nexport async function plan(\n  userInstruction: string,\n  opts: {\n    context: UIContext;\n    interfaceType: InterfaceType;\n    actionSpace: DeviceAction<any>[];\n    actionContext?: string;\n    modelConfig: IModelConfig;\n    conversationHistory?: ConversationHistory;\n    includeBbox: boolean;\n    thinkingStrategy: ThinkingStrategy;\n  },\n): Promise<PlanningAIResponse> {\n  const { context, modelConfig, conversationHistory } = opts;\n  const { screenshotBase64, size } = context;\n\n  const { vlMode } = modelConfig;\n\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    vlMode,\n    includeBbox: opts.includeBbox,\n    thinkingStrategy: opts.thinkingStrategy,\n  });\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = size.width;\n  let imageHeight = size.height;\n  const rightLimit = imageWidth;\n  const bottomLimit = imageHeight;\n\n  // Process image based on VL mode requirements\n  if (vlMode === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const historyLog = opts.conversationHistory?.snapshot() || [];\n  // .filter((item) => item.role === 'assistant') || [];\n\n  const knowledgeContext: ChatCompletionMessageParam[] = opts.actionContext\n    ? [\n        {\n          role: 'user',\n          content: [\n            {\n              type: 'text',\n              text: `<high_priority_knowledge>${opts.actionContext}</high_priority_knowledge>`,\n            },\n          ],\n        },\n      ]\n    : [];\n\n  const instruction: ChatCompletionMessageParam[] = [\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `<user_instruction>${userInstruction}</user_instruction>`,\n        },\n      ],\n    },\n  ];\n\n  const latestImageMessage: ChatCompletionMessageParam = {\n    role: 'user',\n    content: [\n      {\n        type: 'text',\n        text: 'I have finished the action previously planned, and the last screenshot is as follows:',\n      },\n      {\n        type: 'image_url',\n        image_url: {\n          url: imagePayload,\n          detail: 'high',\n        },\n      },\n      // Planning uses pure vision mode, no DOM description needed\n    ],\n  };\n\n  const msgs: ChatCompletionMessageParam[] = [\n    { role: 'system', content: systemPrompt },\n    ...knowledgeContext,\n    ...instruction,\n    ...historyLog,\n    latestImageMessage,\n  ];\n\n  const { content: planFromAI, usage } =\n    await callAIWithObjectResponse<RawResponsePlanningAIResponse>(\n      msgs,\n      AIActionType.PLAN,\n      modelConfig,\n    );\n  const rawResponse = JSON.stringify(planFromAI, undefined, 2);\n\n  const actions = planFromAI.action ? [planFromAI.action] : [];\n  const returnValue: PlanningAIResponse = {\n    ...planFromAI,\n    actions,\n    rawResponse,\n    usage,\n    yamlFlow: buildYamlFlowFromPlans(\n      actions,\n      opts.actionSpace,\n      planFromAI.sleep,\n    ),\n  };\n\n  assert(planFromAI, \"can't get plans from AI\");\n\n  actions.forEach((action) => {\n    const type = action.type;\n    const actionInActionSpace = opts.actionSpace.find(\n      (action) => action.name === type,\n    );\n\n    debug('actionInActionSpace matched', actionInActionSpace);\n    const locateFields = actionInActionSpace\n      ? findAllMidsceneLocatorField(actionInActionSpace.paramSchema)\n      : [];\n\n    debug('locateFields', locateFields);\n\n    locateFields.forEach((field) => {\n      const locateResult = action.param[field];\n      if (locateResult && vlMode !== undefined) {\n        // Always use VL mode to fill bbox parameters\n        action.param[field] = fillBboxParam(\n          locateResult,\n          imageWidth,\n          imageHeight,\n          rightLimit,\n          bottomLimit,\n          vlMode,\n        );\n      }\n    });\n  });\n  // in Qwen-VL, error means error. In GPT-4o, error may mean more actions are needed.\n  assert(!planFromAI.error, `Failed to plan actions: ${planFromAI.error}`);\n\n  if (\n    actions.length === 0 &&\n    returnValue.more_actions_needed_by_instruction &&\n    !returnValue.sleep\n  ) {\n    console.warn(\n      'No actions planned for the prompt, but model said more actions are needed:',\n      userInstruction,\n    );\n  }\n\n  conversationHistory?.append(latestImageMessage);\n\n  conversationHistory?.append({\n    role: 'assistant',\n    content: [\n      {\n        type: 'text',\n        text: rawResponse,\n      },\n    ],\n  });\n\n  return returnValue;\n}\n"],"names":["debug","getDebug","plan","userInstruction","opts","_opts_conversationHistory","context","modelConfig","conversationHistory","screenshotBase64","size","vlMode","systemPrompt","systemPromptToTaskPlanning","imagePayload","imageWidth","imageHeight","rightLimit","bottomLimit","paddedResult","paddingToMatchBlockByBase64","historyLog","knowledgeContext","instruction","latestImageMessage","msgs","planFromAI","usage","callAIWithObjectResponse","AIActionType","rawResponse","JSON","undefined","actions","returnValue","buildYamlFlowFromPlans","assert","action","type","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam","console"],"mappings":";;;;;;AAuBA,MAAMA,QAAQC,SAAS;AAEhB,eAAeC,KACpBC,eAAuB,EACvBC,IASC;QA4BkBC;IA1BnB,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAEC,mBAAmB,EAAE,GAAGJ;IACtD,MAAM,EAAEK,gBAAgB,EAAEC,IAAI,EAAE,GAAGJ;IAEnC,MAAM,EAAEK,MAAM,EAAE,GAAGJ;IAEnB,MAAMK,eAAe,MAAMC,2BAA2B;QACpD,aAAaT,KAAK,WAAW;QAC7BO;QACA,aAAaP,KAAK,WAAW;QAC7B,kBAAkBA,KAAK,gBAAgB;IACzC;IAEA,IAAIU,eAAeL;IACnB,IAAIM,aAAaL,KAAK,KAAK;IAC3B,IAAIM,cAAcN,KAAK,MAAM;IAC7B,MAAMO,aAAaF;IACnB,MAAMG,cAAcF;IAGpB,IAAIL,AAAW,iBAAXA,QAAyB;QAC3B,MAAMQ,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAME,aAAahB,AAAAA,SAAAA,CAAAA,4BAAAA,KAAK,mBAAmB,AAAD,IAAvBA,KAAAA,IAAAA,0BAA0B,QAAQ,EAAC,KAAK,EAAE;IAG7D,MAAMiB,mBAAiDlB,KAAK,aAAa,GACrE;QACE;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,CAAC,yBAAyB,EAAEA,KAAK,aAAa,CAAC,0BAA0B,CAAC;gBAClF;aACD;QACH;KACD,GACD,EAAE;IAEN,MAAMmB,cAA4C;QAChD;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,CAAC,kBAAkB,EAAEpB,gBAAgB,mBAAmB,CAAC;gBACjE;aACD;QACH;KACD;IAED,MAAMqB,qBAAiD;QACrD,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAM;YACR;YACA;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKV;oBACL,QAAQ;gBACV;YACF;SAED;IACH;IAEA,MAAMW,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASb;QAAa;WACrCU;WACAC;WACAF;QACHG;KACD;IAED,MAAM,EAAE,SAASE,UAAU,EAAEC,KAAK,EAAE,GAClC,MAAMC,yBACJH,MACAI,aAAa,IAAI,EACjBtB;IAEJ,MAAMuB,cAAcC,KAAK,SAAS,CAACL,YAAYM,QAAW;IAE1D,MAAMC,UAAUP,WAAW,MAAM,GAAG;QAACA,WAAW,MAAM;KAAC,GAAG,EAAE;IAC5D,MAAMQ,cAAkC;QACtC,GAAGR,UAAU;QACbO;QACAH;QACAH;QACA,UAAUQ,uBACRF,SACA7B,KAAK,WAAW,EAChBsB,WAAW,KAAK;IAEpB;IAEAU,OAAOV,YAAY;IAEnBO,QAAQ,OAAO,CAAC,CAACI;QACf,MAAMC,OAAOD,OAAO,IAAI;QACxB,MAAME,sBAAsBnC,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACiC,SAAWA,OAAO,IAAI,KAAKC;QAG9BtC,MAAM,+BAA+BuC;QACrC,MAAMC,eAAeD,sBACjBE,4BAA4BF,oBAAoB,WAAW,IAC3D,EAAE;QAENvC,MAAM,gBAAgBwC;QAEtBA,aAAa,OAAO,CAAC,CAACE;YACpB,MAAMC,eAAeN,OAAO,KAAK,CAACK,MAAM;YACxC,IAAIC,gBAAgBhC,AAAWqB,WAAXrB,QAElB0B,OAAO,KAAK,CAACK,MAAM,GAAGE,cACpBD,cACA5B,YACAC,aACAC,YACAC,aACAP;QAGN;IACF;IAEAyB,OAAO,CAACV,WAAW,KAAK,EAAE,CAAC,wBAAwB,EAAEA,WAAW,KAAK,EAAE;IAEvE,IACEO,AAAmB,MAAnBA,QAAQ,MAAM,IACdC,YAAY,kCAAkC,IAC9C,CAACA,YAAY,KAAK,EAElBW,QAAQ,IAAI,CACV,8EACA1C;IAIJK,QAAAA,uBAAAA,oBAAqB,MAAM,CAACgB;IAE5BhB,QAAAA,uBAAAA,oBAAqB,MAAM,CAAC;QAC1B,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAMsB;YACR;SACD;IACH;IAEA,OAAOI;AACT"}

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

@@ -107,10 +107,12 @@ const descriptionForAction = (action, locatorSchemaTypeDescription)=>{
 ${tab}${fields.join(`\n${tab}`)}
 `.trim();
 };
-async function systemPromptToTaskPlanning({ actionSpace, vlMode, includeBbox }) {
+async function systemPromptToTaskPlanning({ actionSpace, vlMode, includeBbox, thinkingStrategy }) {
     if (includeBbox && !vlMode) throw new Error('vlMode cannot be undefined when includeBbox is true. A valid vlMode is required for bbox-based location.');
     const actionDescriptionList = actionSpace.map((action)=>descriptionForAction(action, vlLocateParam(includeBbox ? vlMode : void 0)));
     const actionList = actionDescriptionList.join('\n');
+    const logFieldDefinition = 'off' === thinkingStrategy ? '' : '"log": string, // Log your thoughts and what the next one action (ONLY ONE!) you can do according to the screenshot and the instruction. The log should contain the following information: "The user wants to do ... . According to the instruction and the previous logs, next step is to .... Now i am going to compose an action \'{ action-type }\' to do this". If no action should be done, log the reason. Use the same language as the user\'s instruction.\n  ';
+    const exampleLogField = 'off' === thinkingStrategy ? '' : "\"log\": \"The user wants to do click 'Confirm' button, and click 'Yes' in popup. According to the instruction and the previous logs, next step is to tap the 'Yes' button in the popup. Now i am going to compose an action 'Tap' to click 'Yes' in popup.\",\n  ";
     return `
 Target: User will give you an instruction, some screenshots and previous logs indicating what have been done. Your task is to plan the next one action to accomplish the instruction.
 
@@ -119,14 +121,14 @@ Please tell what the next one action is (or null if no action should be done) to
 Restriction:
 - Don't give extra actions or plans beyond the instruction. ONLY plan for what the instruction requires. For example, don't try to submit the form if the instruction is only to fill something.
 - Give just the next ONE action you should do
+- If the user mentions something to assert and the condition is not met, you should think this is an error and set the "error" field to the error message.
 
 Supporting actions:
 ${actionList}
 
 Return in JSON format:
 {
-  "log": string, // Log your thoughts and what the next one action (ONLY ONE!) you can do according to the screenshot and the instruction. The log should contain the following information: "The user wants to do ... . According to the instruction and the previous logs, next step is to .... Now i am going to compose an action '{ action-type }' to do this". If no action should be done, log the reason. Use the same language as the user's instruction.
-  ${commonOutputFields}
+  ${logFieldDefinition}${commonOutputFields}
   "action": 
     {
       // one of the supporting actions
@@ -140,8 +142,7 @@ For example, when the instruction is "click 'Confirm' button, and click 'Yes' in
 this and output the JSON:
 
 {
-  "log": "The user wants to do click 'Confirm' button, and click 'Yes' in popup. According to the instruction and the previous logs, next step is to tap the 'Yes' button in the popup. Now i am going to compose an action 'Tap' to click 'Yes' in popup.",
-  "action": {
+  ${exampleLogField}"action": {
     "type": "Tap",
     "param": {
       "locate": {

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

@@ -1 +1 @@
-{"version":3,"file":"ai-model/prompt/llm-planning.mjs","sources":["webpack://@midscene/core/./src/ai-model/prompt/llm-planning.ts"],"sourcesContent":["import type { DeviceAction } from '@/types';\nimport type { TVlModeTypes } from '@midscene/shared/env';\nimport type { ResponseFormatJSONSchema } from 'openai/resources/index';\nimport type { z } from 'zod';\nimport { ifMidsceneLocatorField } from '../../common';\nimport { bboxDescription } from './common';\n\n// Note: put the log field first to trigger the CoT\n\nconst commonOutputFields = `\"error\"?: string, // Error messages about unexpected situations, if any. Only think it is an error when the situation is not foreseeable according to the instruction. Use the same language as the user's instruction.\n  \"more_actions_needed_by_instruction\": boolean, // Consider if there is still more action(s) to do after the action in \"Log\" is done, according to the instruction. If so, set this field to true. Otherwise, set it to false.`;\n\nconst vlLocateParam = (vlMode: TVlModeTypes | undefined) => {\n  if (vlMode) {\n    return `{bbox: [number, number, number, number], prompt: string } // ${bboxDescription(vlMode)}`;\n  }\n  return '{ prompt: string /* description of the target element */ }';\n};\n\nexport const descriptionForAction = (\n  action: DeviceAction<any>,\n  locatorSchemaTypeDescription: string,\n) => {\n  const tab = '  ';\n  const fields: string[] = [];\n\n  // Add the action type field\n  fields.push(`- type: \"${action.name}\"`);\n\n  // Handle paramSchema if it exists\n  if (action.paramSchema) {\n    const paramLines: string[] = [];\n\n    // Check if paramSchema is a ZodObject with shape\n    const schema = action.paramSchema as any;\n    const isZodObject = schema._def?.typeName === 'ZodObject';\n\n    if (isZodObject && schema.shape) {\n      // Original logic for ZodObject schemas\n      const shape = schema.shape;\n\n      // Helper function to get type name from zod schema\n      const getTypeName = (field: any): string => {\n        // Recursively unwrap optional, nullable, and other wrapper types to get the actual inner type\n        const unwrapField = (f: any): any => {\n          if (!f._def) return f;\n\n          const typeName = f._def.typeName;\n\n          // Handle wrapper types that have innerType\n          if (\n            typeName === 'ZodOptional' ||\n            typeName === 'ZodNullable' ||\n            typeName === 'ZodDefault'\n          ) {\n            return unwrapField(f._def.innerType);\n          }\n\n          // Handle ZodEffects (transformations, refinements, preprocessors)\n          if (typeName === 'ZodEffects') {\n            // For ZodEffects, unwrap the schema field which contains the underlying type\n            if (f._def.schema) {\n              return unwrapField(f._def.schema);\n            }\n          }\n\n          return f;\n        };\n\n        const actualField = unwrapField(field);\n        const fieldTypeName = actualField._def?.typeName;\n\n        if (fieldTypeName === 'ZodString') return 'string';\n        if (fieldTypeName === 'ZodNumber') return 'number';\n        if (fieldTypeName === 'ZodBoolean') return 'boolean';\n        if (fieldTypeName === 'ZodArray') return 'array';\n        if (fieldTypeName === 'ZodObject') {\n          // Check if this is a passthrough object (like MidsceneLocation)\n          if (ifMidsceneLocatorField(actualField)) {\n            return locatorSchemaTypeDescription;\n          }\n          return 'object';\n        }\n        if (fieldTypeName === 'ZodEnum') {\n          const values =\n            (actualField._def?.values as unknown[] | undefined)\n              ?.map((option: unknown) => String(`'${option}'`))\n              .join(', ') ?? 'enum';\n\n          return `enum(${values})`;\n        }\n        // Handle ZodUnion by taking the first option (for display purposes)\n        if (fieldTypeName === 'ZodUnion') {\n          const options = actualField._def?.options as any[] | undefined;\n          if (options && options.length > 0) {\n            // For unions, list all types\n            const types = options.map((opt: any) => getTypeName(opt));\n            return types.join(' | ');\n          }\n          return 'union';\n        }\n\n        console.warn(\n          'failed to parse Zod type. This may lead to wrong params from the LLM.\\n',\n          actualField._def,\n        );\n        return actualField.toString();\n      };\n\n      // Helper function to get description from zod schema\n      const getDescription = (field: z.ZodTypeAny): string | null => {\n        // Recursively unwrap optional, nullable, and other wrapper types to get the actual inner type\n        const unwrapField = (f: any): any => {\n          if (!f._def) return f;\n\n          const typeName = f._def.typeName;\n\n          // Handle wrapper types that have innerType\n          if (\n            typeName === 'ZodOptional' ||\n            typeName === 'ZodNullable' ||\n            typeName === 'ZodDefault'\n          ) {\n            return unwrapField(f._def.innerType);\n          }\n\n          // Handle ZodEffects (transformations, refinements, preprocessors)\n          if (typeName === 'ZodEffects') {\n            // For ZodEffects, unwrap the schema field which contains the underlying type\n            if (f._def.schema) {\n              return unwrapField(f._def.schema);\n            }\n          }\n\n          return f;\n        };\n\n        // Check for direct description on the original field (wrapper may have description)\n        if ('description' in field) {\n          return field.description || null;\n        }\n\n        const actualField = unwrapField(field);\n\n        // Check for description on the unwrapped field\n        if ('description' in actualField) {\n          return actualField.description || null;\n        }\n\n        // Check for MidsceneLocation fields and add description\n        if (actualField._def?.typeName === 'ZodObject') {\n          if ('midscene_location_field_flag' in actualField._def.shape()) {\n            return 'Location information for the target element';\n          }\n        }\n\n        return null;\n      };\n\n      for (const [key, field] of Object.entries(shape)) {\n        if (field && typeof field === 'object') {\n          // Check if field is optional\n          const isOptional =\n            typeof (field as any).isOptional === 'function' &&\n            (field as any).isOptional();\n          const keyWithOptional = isOptional ? `${key}?` : key;\n\n          // Get the type name\n          const typeName = getTypeName(field);\n\n          // Get description\n          const description = getDescription(field as z.ZodTypeAny);\n\n          // Build param line for this field\n          let paramLine = `${keyWithOptional}: ${typeName}`;\n          if (description) {\n            paramLine += ` // ${description}`;\n          }\n\n          paramLines.push(paramLine);\n        }\n      }\n\n      // Add the param section to fields if there are paramLines\n      if (paramLines.length > 0) {\n        fields.push('- param:');\n        paramLines.forEach((line) => {\n          fields.push(`  - ${line}`);\n        });\n      }\n    } else {\n      // Handle non-object schemas (string, number, etc.)\n      // For simple primitive types, the param should be passed directly as the value\n      const schemaTypeName = schema._def?.typeName;\n      let typeName = 'unknown';\n\n      if (schemaTypeName === 'ZodString') typeName = 'string';\n      else if (schemaTypeName === 'ZodNumber') typeName = 'number';\n      else if (schemaTypeName === 'ZodBoolean') typeName = 'boolean';\n\n      // Get description if available\n      const description = 'description' in schema ? schema.description : null;\n\n      // For simple types, indicate that param should be the direct value, not an object\n      let paramDescription = `- param: ${typeName}`;\n      if (description) {\n        paramDescription += ` // ${description}`;\n      }\n      paramDescription += ' (pass the value directly, not as an object)';\n\n      fields.push(paramDescription);\n    }\n  }\n\n  return `- ${action.name}, ${action.description || 'No description provided'}\n${tab}${fields.join(`\\n${tab}`)}\n`.trim();\n};\n\nexport async function systemPromptToTaskPlanning({\n  actionSpace,\n  vlMode,\n  includeBbox,\n}: {\n  actionSpace: DeviceAction<any>[];\n  vlMode: TVlModeTypes | undefined;\n  includeBbox: boolean;\n}) {\n  // Validate parameters: if includeBbox is true, vlMode must be defined\n  if (includeBbox && !vlMode) {\n    throw new Error(\n      'vlMode cannot be undefined when includeBbox is true. A valid vlMode is required for bbox-based location.',\n    );\n  }\n\n  const actionDescriptionList = actionSpace.map((action) => {\n    return descriptionForAction(\n      action,\n      vlLocateParam(includeBbox ? vlMode : undefined),\n    );\n  });\n  const actionList = actionDescriptionList.join('\\n');\n\n  return `\nTarget: User will give you an instruction, some screenshots and previous logs indicating what have been done. Your task is to plan the next one action to accomplish the instruction.\n\nPlease tell what the next one action is (or null if no action should be done) to do the tasks the instruction requires. \n\nRestriction:\n- Don't give extra actions or plans beyond the instruction. ONLY plan for what the instruction requires. For example, don't try to submit the form if the instruction is only to fill something.\n- Give just the next ONE action you should do\n\nSupporting actions:\n${actionList}\n\nReturn in JSON format:\n{\n  \"log\": string, // Log your thoughts and what the next one action (ONLY ONE!) you can do according to the screenshot and the instruction. The log should contain the following information: \"The user wants to do ... . According to the instruction and the previous logs, next step is to .... Now i am going to compose an action '{ action-type }' to do this\". If no action should be done, log the reason. Use the same language as the user's instruction.\n  ${commonOutputFields}\n  \"action\": \n    {\n      // one of the supporting actions\n    } | null,\n  ,\n  \"sleep\"?: number, // The sleep time after the action, in milliseconds.\n}\n\nFor example, when the instruction is \"click 'Confirm' button, and click 'Yes' in popup\" and the previous log shows \"The 'Confirm' button has been clicked\", by viewing the screenshot and previous logs, you should consider: We have already clicked the 'Confirm' button, so next we should find and click 'Yes' in popup.\n\nthis and output the JSON:\n\n{\n  \"log\": \"The user wants to do click 'Confirm' button, and click 'Yes' in popup. According to the instruction and the previous logs, next step is to tap the 'Yes' button in the popup. Now i am going to compose an action 'Tap' to click 'Yes' in popup.\",\n  \"action\": {\n    \"type\": \"Tap\",\n    \"param\": {\n      \"locate\": {\n        ${vlMode ? `\"bbox\": [100, 100, 200, 200],` : ''}\n        \"prompt\": \"The 'Yes' button in popup\"\n      }\n    }\n  },\n  \"more_actions_needed_by_instruction\": false,\n}\n`;\n}\n\nexport const planSchema: ResponseFormatJSONSchema = {\n  type: 'json_schema',\n  json_schema: {\n    name: 'action_items',\n    strict: false,\n    schema: {\n      type: 'object',\n      strict: false,\n      properties: {\n        actions: {\n          type: 'array',\n          items: {\n            type: 'object',\n            strict: false,\n            properties: {\n              thought: {\n                type: 'string',\n                description:\n                  'Reasons for generating this task, and why this task is feasible on this page',\n              },\n              type: {\n                type: 'string',\n                description: 'Type of action',\n              },\n              param: {\n                anyOf: [\n                  { type: 'null' },\n                  {\n                    type: 'object',\n                    additionalProperties: true,\n                  },\n                ],\n                description: 'Parameter of the action',\n              },\n              locate: {\n                type: ['object', 'null'],\n                properties: {\n                  id: { type: 'string' },\n                  prompt: { type: 'string' },\n                },\n                required: ['id', 'prompt'],\n                additionalProperties: false,\n                description: 'Location information for the target element',\n              },\n            },\n            required: ['thought', 'type', 'param', 'locate'],\n            additionalProperties: false,\n          },\n          description: 'List of actions to be performed',\n        },\n        more_actions_needed_by_instruction: {\n          type: 'boolean',\n          description:\n            'If all the actions described in the instruction have been covered by this action and logs, set this field to false.',\n        },\n        log: {\n          type: 'string',\n          description:\n            'Log what these planned actions do. Do not include further actions that have not been planned.',\n        },\n        error: {\n          type: ['string', 'null'],\n          description: 'Error messages about unexpected situations',\n        },\n      },\n      required: [\n        'actions',\n        'more_actions_needed_by_instruction',\n        'log',\n        'error',\n      ],\n      additionalProperties: false,\n    },\n  },\n};\n"],"names":["commonOutputFields","vlLocateParam","vlMode","bboxDescription","descriptionForAction","action","locatorSchemaTypeDescription","tab","fields","_schema__def","paramLines","schema","isZodObject","shape","getTypeName","field","_actualField__def","unwrapField","f","typeName","actualField","fieldTypeName","ifMidsceneLocatorField","_actualField__def_values","values","option","String","_actualField__def2","options","types","opt","console","getDescription","key","Object","isOptional","keyWithOptional","description","paramLine","line","_schema__def1","schemaTypeName","paramDescription","systemPromptToTaskPlanning","actionSpace","includeBbox","Error","actionDescriptionList","undefined","actionList","planSchema"],"mappings":";;AASA,MAAMA,qBAAqB,CAAC;+NACmM,CAAC;AAEhO,MAAMC,gBAAgB,CAACC;IACrB,IAAIA,QACF,OAAO,CAAC,6DAA6D,EAAEC,gBAAgBD,SAAS;IAElG,OAAO;AACT;AAEO,MAAME,uBAAuB,CAClCC,QACAC;IAEA,MAAMC,MAAM;IACZ,MAAMC,SAAmB,EAAE;IAG3BA,OAAO,IAAI,CAAC,CAAC,SAAS,EAAEH,OAAO,IAAI,CAAC,CAAC,CAAC;IAGtC,IAAIA,OAAO,WAAW,EAAE;YAKFI;QAJpB,MAAMC,aAAuB,EAAE;QAG/B,MAAMC,SAASN,OAAO,WAAW;QACjC,MAAMO,cAAcH,AAAAA,SAAAA,CAAAA,eAAAA,OAAO,IAAI,AAAD,IAAVA,KAAAA,IAAAA,aAAa,QAAQ,AAAD,MAAM;QAE9C,IAAIG,eAAeD,OAAO,KAAK,EAAE;YAE/B,MAAME,QAAQF,OAAO,KAAK;YAG1B,MAAMG,cAAc,CAACC;oBA4BGC;gBA1BtB,MAAMC,cAAc,CAACC;oBACnB,IAAI,CAACA,EAAE,IAAI,EAAE,OAAOA;oBAEpB,MAAMC,WAAWD,EAAE,IAAI,CAAC,QAAQ;oBAGhC,IACEC,AAAa,kBAAbA,YACAA,AAAa,kBAAbA,YACAA,AAAa,iBAAbA,UAEA,OAAOF,YAAYC,EAAE,IAAI,CAAC,SAAS;oBAIrC,IAAIC,AAAa,iBAAbA,UAEF;wBAAA,IAAID,EAAE,IAAI,CAAC,MAAM,EACf,OAAOD,YAAYC,EAAE,IAAI,CAAC,MAAM;oBAClC;oBAGF,OAAOA;gBACT;gBAEA,MAAME,cAAcH,YAAYF;gBAChC,MAAMM,gBAAgB,QAAAL,CAAAA,oBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,kBAAkB,QAAQ;gBAEhD,IAAIK,AAAkB,gBAAlBA,eAA+B,OAAO;gBAC1C,IAAIA,AAAkB,gBAAlBA,eAA+B,OAAO;gBAC1C,IAAIA,AAAkB,iBAAlBA,eAAgC,OAAO;gBAC3C,IAAIA,AAAkB,eAAlBA,eAA8B,OAAO;gBACzC,IAAIA,AAAkB,gBAAlBA,eAA+B;oBAEjC,IAAIC,uBAAuBF,cACzB,OAAOd;oBAET,OAAO;gBACT;gBACA,IAAIe,AAAkB,cAAlBA,eAA6B;wBAE5BE,0BAAAA;oBADH,MAAMC,SACJ,AAAC,SAAAD,CAAAA,qBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,QAAAA,CAAAA,2BAAAA,mBAAkB,MAAM,AAAD,IAAvBA,KAAAA,IAAAA,yBACG,GAAG,CAAC,CAACE,SAAoBC,OAAO,CAAC,CAAC,EAAED,OAAO,CAAC,CAAC,GAC9C,IAAI,CAAC,KAAI,KAAK;oBAEnB,OAAO,CAAC,KAAK,EAAED,OAAO,CAAC,CAAC;gBAC1B;gBAEA,IAAIH,AAAkB,eAAlBA,eAA8B;wBAChBM;oBAAhB,MAAMC,UAAU,QAAAD,CAAAA,qBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,mBAAkB,OAAO;oBACzC,IAAIC,WAAWA,QAAQ,MAAM,GAAG,GAAG;wBAEjC,MAAMC,QAAQD,QAAQ,GAAG,CAAC,CAACE,MAAahB,YAAYgB;wBACpD,OAAOD,MAAM,IAAI,CAAC;oBACpB;oBACA,OAAO;gBACT;gBAEAE,QAAQ,IAAI,CACV,2EACAX,YAAY,IAAI;gBAElB,OAAOA,YAAY,QAAQ;YAC7B;YAGA,MAAMY,iBAAiB,CAACjB;oBAwClBC;gBAtCJ,MAAMC,cAAc,CAACC;oBACnB,IAAI,CAACA,EAAE,IAAI,EAAE,OAAOA;oBAEpB,MAAMC,WAAWD,EAAE,IAAI,CAAC,QAAQ;oBAGhC,IACEC,AAAa,kBAAbA,YACAA,AAAa,kBAAbA,YACAA,AAAa,iBAAbA,UAEA,OAAOF,YAAYC,EAAE,IAAI,CAAC,SAAS;oBAIrC,IAAIC,AAAa,iBAAbA,UAEF;wBAAA,IAAID,EAAE,IAAI,CAAC,MAAM,EACf,OAAOD,YAAYC,EAAE,IAAI,CAAC,MAAM;oBAClC;oBAGF,OAAOA;gBACT;gBAGA,IAAI,iBAAiBH,OACnB,OAAOA,MAAM,WAAW,IAAI;gBAG9B,MAAMK,cAAcH,YAAYF;gBAGhC,IAAI,iBAAiBK,aACnB,OAAOA,YAAY,WAAW,IAAI;gBAIpC,IAAIJ,AAAAA,SAAAA,CAAAA,oBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,kBAAkB,QAAQ,AAAD,MAAM,aACjC;oBAAA,IAAI,kCAAkCI,YAAY,IAAI,CAAC,KAAK,IAC1D,OAAO;gBACT;gBAGF,OAAO;YACT;YAEA,KAAK,MAAM,CAACa,KAAKlB,MAAM,IAAImB,OAAO,OAAO,CAACrB,OACxC,IAAIE,SAAS,AAAiB,YAAjB,OAAOA,OAAoB;gBAEtC,MAAMoB,aACJ,AAAqC,cAArC,OAAQpB,MAAc,UAAU,IAC/BA,MAAc,UAAU;gBAC3B,MAAMqB,kBAAkBD,aAAa,GAAGF,IAAI,CAAC,CAAC,GAAGA;gBAGjD,MAAMd,WAAWL,YAAYC;gBAG7B,MAAMsB,cAAcL,eAAejB;gBAGnC,IAAIuB,YAAY,GAAGF,gBAAgB,EAAE,EAAEjB,UAAU;gBACjD,IAAIkB,aACFC,aAAa,CAAC,IAAI,EAAED,aAAa;gBAGnC3B,WAAW,IAAI,CAAC4B;YAClB;YAIF,IAAI5B,WAAW,MAAM,GAAG,GAAG;gBACzBF,OAAO,IAAI,CAAC;gBACZE,WAAW,OAAO,CAAC,CAAC6B;oBAClB/B,OAAO,IAAI,CAAC,CAAC,IAAI,EAAE+B,MAAM;gBAC3B;YACF;QACF,OAAO;gBAGkBC;YAAvB,MAAMC,iBAAiB,QAAAD,CAAAA,gBAAAA,OAAO,IAAI,AAAD,IAAVA,KAAAA,IAAAA,cAAa,QAAQ;YAC5C,IAAIrB,WAAW;YAEf,IAAIsB,AAAmB,gBAAnBA,gBAAgCtB,WAAW;iBAC1C,IAAIsB,AAAmB,gBAAnBA,gBAAgCtB,WAAW;iBAC/C,IAAIsB,AAAmB,iBAAnBA,gBAAiCtB,WAAW;YAGrD,MAAMkB,cAAc,iBAAiB1B,SAASA,OAAO,WAAW,GAAG;YAGnE,IAAI+B,mBAAmB,CAAC,SAAS,EAAEvB,UAAU;YAC7C,IAAIkB,aACFK,oBAAoB,CAAC,IAAI,EAAEL,aAAa;YAE1CK,oBAAoB;YAEpBlC,OAAO,IAAI,CAACkC;QACd;IACF;IAEA,OAAO,CAAC,EAAE,EAAErC,OAAO,IAAI,CAAC,EAAE,EAAEA,OAAO,WAAW,IAAI,0BAA0B;AAC9E,EAAEE,MAAMC,OAAO,IAAI,CAAC,CAAC,EAAE,EAAED,KAAK,EAAE;AAChC,CAAC,CAAC,IAAI;AACN;AAEO,eAAeoC,2BAA2B,EAC/CC,WAAW,EACX1C,MAAM,EACN2C,WAAW,EAKZ;IAEC,IAAIA,eAAe,CAAC3C,QAClB,MAAM,IAAI4C,MACR;IAIJ,MAAMC,wBAAwBH,YAAY,GAAG,CAAC,CAACvC,SACtCD,qBACLC,QACAJ,cAAc4C,cAAc3C,SAAS8C;IAGzC,MAAMC,aAAaF,sBAAsB,IAAI,CAAC;IAE9C,OAAO,CAAC;;;;;;;;;;AAUV,EAAEE,WAAW;;;;;EAKX,EAAEjD,mBAAmB;;;;;;;;;;;;;;;;;;;QAmBf,EAAEE,SAAS,kCAAkC,GAAG;;;;;;;AAOxD,CAAC;AACD;AAEO,MAAMgD,aAAuC;IAClD,MAAM;IACN,aAAa;QACX,MAAM;QACN,QAAQ;QACR,QAAQ;YACN,MAAM;YACN,QAAQ;YACR,YAAY;gBACV,SAAS;oBACP,MAAM;oBACN,OAAO;wBACL,MAAM;wBACN,QAAQ;wBACR,YAAY;4BACV,SAAS;gCACP,MAAM;gCACN,aACE;4BACJ;4BACA,MAAM;gCACJ,MAAM;gCACN,aAAa;4BACf;4BACA,OAAO;gCACL,OAAO;oCACL;wCAAE,MAAM;oCAAO;oCACf;wCACE,MAAM;wCACN,sBAAsB;oCACxB;iCACD;gCACD,aAAa;4BACf;4BACA,QAAQ;gCACN,MAAM;oCAAC;oCAAU;iCAAO;gCACxB,YAAY;oCACV,IAAI;wCAAE,MAAM;oCAAS;oCACrB,QAAQ;wCAAE,MAAM;oCAAS;gCAC3B;gCACA,UAAU;oCAAC;oCAAM;iCAAS;gCAC1B,sBAAsB;gCACtB,aAAa;4BACf;wBACF;wBACA,UAAU;4BAAC;4BAAW;4BAAQ;4BAAS;yBAAS;wBAChD,sBAAsB;oBACxB;oBACA,aAAa;gBACf;gBACA,oCAAoC;oBAClC,MAAM;oBACN,aACE;gBACJ;gBACA,KAAK;oBACH,MAAM;oBACN,aACE;gBACJ;gBACA,OAAO;oBACL,MAAM;wBAAC;wBAAU;qBAAO;oBACxB,aAAa;gBACf;YACF;YACA,UAAU;gBACR;gBACA;gBACA;gBACA;aACD;YACD,sBAAsB;QACxB;IACF;AACF"}
+{"version":3,"file":"ai-model/prompt/llm-planning.mjs","sources":["webpack://@midscene/core/./src/ai-model/prompt/llm-planning.ts"],"sourcesContent":["import type { DeviceAction, ThinkingStrategy } from '@/types';\nimport type { TVlModeTypes } from '@midscene/shared/env';\nimport type { ResponseFormatJSONSchema } from 'openai/resources/index';\nimport type { z } from 'zod';\nimport { ifMidsceneLocatorField } from '../../common';\nimport { bboxDescription } from './common';\n\n// Note: put the log field first to trigger the CoT\n\nconst commonOutputFields = `\"error\"?: string, // Error messages about unexpected situations, if any. Only think it is an error when the situation is not foreseeable according to the instruction. Use the same language as the user's instruction.\n  \"more_actions_needed_by_instruction\": boolean, // Consider if there is still more action(s) to do after the action in \"Log\" is done, according to the instruction. If so, set this field to true. Otherwise, set it to false.`;\n\nconst vlLocateParam = (vlMode: TVlModeTypes | undefined) => {\n  if (vlMode) {\n    return `{bbox: [number, number, number, number], prompt: string } // ${bboxDescription(vlMode)}`;\n  }\n  return '{ prompt: string /* description of the target element */ }';\n};\n\nexport const descriptionForAction = (\n  action: DeviceAction<any>,\n  locatorSchemaTypeDescription: string,\n) => {\n  const tab = '  ';\n  const fields: string[] = [];\n\n  // Add the action type field\n  fields.push(`- type: \"${action.name}\"`);\n\n  // Handle paramSchema if it exists\n  if (action.paramSchema) {\n    const paramLines: string[] = [];\n\n    // Check if paramSchema is a ZodObject with shape\n    const schema = action.paramSchema as any;\n    const isZodObject = schema._def?.typeName === 'ZodObject';\n\n    if (isZodObject && schema.shape) {\n      // Original logic for ZodObject schemas\n      const shape = schema.shape;\n\n      // Helper function to get type name from zod schema\n      const getTypeName = (field: any): string => {\n        // Recursively unwrap optional, nullable, and other wrapper types to get the actual inner type\n        const unwrapField = (f: any): any => {\n          if (!f._def) return f;\n\n          const typeName = f._def.typeName;\n\n          // Handle wrapper types that have innerType\n          if (\n            typeName === 'ZodOptional' ||\n            typeName === 'ZodNullable' ||\n            typeName === 'ZodDefault'\n          ) {\n            return unwrapField(f._def.innerType);\n          }\n\n          // Handle ZodEffects (transformations, refinements, preprocessors)\n          if (typeName === 'ZodEffects') {\n            // For ZodEffects, unwrap the schema field which contains the underlying type\n            if (f._def.schema) {\n              return unwrapField(f._def.schema);\n            }\n          }\n\n          return f;\n        };\n\n        const actualField = unwrapField(field);\n        const fieldTypeName = actualField._def?.typeName;\n\n        if (fieldTypeName === 'ZodString') return 'string';\n        if (fieldTypeName === 'ZodNumber') return 'number';\n        if (fieldTypeName === 'ZodBoolean') return 'boolean';\n        if (fieldTypeName === 'ZodArray') return 'array';\n        if (fieldTypeName === 'ZodObject') {\n          // Check if this is a passthrough object (like MidsceneLocation)\n          if (ifMidsceneLocatorField(actualField)) {\n            return locatorSchemaTypeDescription;\n          }\n          return 'object';\n        }\n        if (fieldTypeName === 'ZodEnum') {\n          const values =\n            (actualField._def?.values as unknown[] | undefined)\n              ?.map((option: unknown) => String(`'${option}'`))\n              .join(', ') ?? 'enum';\n\n          return `enum(${values})`;\n        }\n        // Handle ZodUnion by taking the first option (for display purposes)\n        if (fieldTypeName === 'ZodUnion') {\n          const options = actualField._def?.options as any[] | undefined;\n          if (options && options.length > 0) {\n            // For unions, list all types\n            const types = options.map((opt: any) => getTypeName(opt));\n            return types.join(' | ');\n          }\n          return 'union';\n        }\n\n        console.warn(\n          'failed to parse Zod type. This may lead to wrong params from the LLM.\\n',\n          actualField._def,\n        );\n        return actualField.toString();\n      };\n\n      // Helper function to get description from zod schema\n      const getDescription = (field: z.ZodTypeAny): string | null => {\n        // Recursively unwrap optional, nullable, and other wrapper types to get the actual inner type\n        const unwrapField = (f: any): any => {\n          if (!f._def) return f;\n\n          const typeName = f._def.typeName;\n\n          // Handle wrapper types that have innerType\n          if (\n            typeName === 'ZodOptional' ||\n            typeName === 'ZodNullable' ||\n            typeName === 'ZodDefault'\n          ) {\n            return unwrapField(f._def.innerType);\n          }\n\n          // Handle ZodEffects (transformations, refinements, preprocessors)\n          if (typeName === 'ZodEffects') {\n            // For ZodEffects, unwrap the schema field which contains the underlying type\n            if (f._def.schema) {\n              return unwrapField(f._def.schema);\n            }\n          }\n\n          return f;\n        };\n\n        // Check for direct description on the original field (wrapper may have description)\n        if ('description' in field) {\n          return field.description || null;\n        }\n\n        const actualField = unwrapField(field);\n\n        // Check for description on the unwrapped field\n        if ('description' in actualField) {\n          return actualField.description || null;\n        }\n\n        // Check for MidsceneLocation fields and add description\n        if (actualField._def?.typeName === 'ZodObject') {\n          if ('midscene_location_field_flag' in actualField._def.shape()) {\n            return 'Location information for the target element';\n          }\n        }\n\n        return null;\n      };\n\n      for (const [key, field] of Object.entries(shape)) {\n        if (field && typeof field === 'object') {\n          // Check if field is optional\n          const isOptional =\n            typeof (field as any).isOptional === 'function' &&\n            (field as any).isOptional();\n          const keyWithOptional = isOptional ? `${key}?` : key;\n\n          // Get the type name\n          const typeName = getTypeName(field);\n\n          // Get description\n          const description = getDescription(field as z.ZodTypeAny);\n\n          // Build param line for this field\n          let paramLine = `${keyWithOptional}: ${typeName}`;\n          if (description) {\n            paramLine += ` // ${description}`;\n          }\n\n          paramLines.push(paramLine);\n        }\n      }\n\n      // Add the param section to fields if there are paramLines\n      if (paramLines.length > 0) {\n        fields.push('- param:');\n        paramLines.forEach((line) => {\n          fields.push(`  - ${line}`);\n        });\n      }\n    } else {\n      // Handle non-object schemas (string, number, etc.)\n      // For simple primitive types, the param should be passed directly as the value\n      const schemaTypeName = schema._def?.typeName;\n      let typeName = 'unknown';\n\n      if (schemaTypeName === 'ZodString') typeName = 'string';\n      else if (schemaTypeName === 'ZodNumber') typeName = 'number';\n      else if (schemaTypeName === 'ZodBoolean') typeName = 'boolean';\n\n      // Get description if available\n      const description = 'description' in schema ? schema.description : null;\n\n      // For simple types, indicate that param should be the direct value, not an object\n      let paramDescription = `- param: ${typeName}`;\n      if (description) {\n        paramDescription += ` // ${description}`;\n      }\n      paramDescription += ' (pass the value directly, not as an object)';\n\n      fields.push(paramDescription);\n    }\n  }\n\n  return `- ${action.name}, ${action.description || 'No description provided'}\n${tab}${fields.join(`\\n${tab}`)}\n`.trim();\n};\n\nexport async function systemPromptToTaskPlanning({\n  actionSpace,\n  vlMode,\n  includeBbox,\n  thinkingStrategy,\n}: {\n  actionSpace: DeviceAction<any>[];\n  vlMode: TVlModeTypes | undefined;\n  includeBbox: boolean;\n  thinkingStrategy: ThinkingStrategy;\n}) {\n  // Validate parameters: if includeBbox is true, vlMode must be defined\n  if (includeBbox && !vlMode) {\n    throw new Error(\n      'vlMode cannot be undefined when includeBbox is true. A valid vlMode is required for bbox-based location.',\n    );\n  }\n\n  const actionDescriptionList = actionSpace.map((action) => {\n    return descriptionForAction(\n      action,\n      vlLocateParam(includeBbox ? vlMode : undefined),\n    );\n  });\n  const actionList = actionDescriptionList.join('\\n');\n\n  // Conditionally include log field based on thinkingStrategy\n  const logFieldDefinition =\n    thinkingStrategy === 'off'\n      ? ''\n      : '\"log\": string, // Log your thoughts and what the next one action (ONLY ONE!) you can do according to the screenshot and the instruction. The log should contain the following information: \"The user wants to do ... . According to the instruction and the previous logs, next step is to .... Now i am going to compose an action \\'{ action-type }\\' to do this\". If no action should be done, log the reason. Use the same language as the user\\'s instruction.\\n  ';\n\n  const exampleLogField =\n    thinkingStrategy === 'off'\n      ? ''\n      : \"\\\"log\\\": \\\"The user wants to do click 'Confirm' button, and click 'Yes' in popup. According to the instruction and the previous logs, next step is to tap the 'Yes' button in the popup. Now i am going to compose an action 'Tap' to click 'Yes' in popup.\\\",\\n  \";\n\n  return `\nTarget: User will give you an instruction, some screenshots and previous logs indicating what have been done. Your task is to plan the next one action to accomplish the instruction.\n\nPlease tell what the next one action is (or null if no action should be done) to do the tasks the instruction requires. \n\nRestriction:\n- Don't give extra actions or plans beyond the instruction. ONLY plan for what the instruction requires. For example, don't try to submit the form if the instruction is only to fill something.\n- Give just the next ONE action you should do\n- If the user mentions something to assert and the condition is not met, you should think this is an error and set the \"error\" field to the error message.\n\nSupporting actions:\n${actionList}\n\nReturn in JSON format:\n{\n  ${logFieldDefinition}${commonOutputFields}\n  \"action\": \n    {\n      // one of the supporting actions\n    } | null,\n  ,\n  \"sleep\"?: number, // The sleep time after the action, in milliseconds.\n}\n\nFor example, when the instruction is \"click 'Confirm' button, and click 'Yes' in popup\" and the previous log shows \"The 'Confirm' button has been clicked\", by viewing the screenshot and previous logs, you should consider: We have already clicked the 'Confirm' button, so next we should find and click 'Yes' in popup.\n\nthis and output the JSON:\n\n{\n  ${exampleLogField}\"action\": {\n    \"type\": \"Tap\",\n    \"param\": {\n      \"locate\": {\n        ${vlMode ? `\"bbox\": [100, 100, 200, 200],` : ''}\n        \"prompt\": \"The 'Yes' button in popup\"\n      }\n    }\n  },\n  \"more_actions_needed_by_instruction\": false,\n}\n`;\n}\n\nexport const planSchema: ResponseFormatJSONSchema = {\n  type: 'json_schema',\n  json_schema: {\n    name: 'action_items',\n    strict: false,\n    schema: {\n      type: 'object',\n      strict: false,\n      properties: {\n        actions: {\n          type: 'array',\n          items: {\n            type: 'object',\n            strict: false,\n            properties: {\n              thought: {\n                type: 'string',\n                description:\n                  'Reasons for generating this task, and why this task is feasible on this page',\n              },\n              type: {\n                type: 'string',\n                description: 'Type of action',\n              },\n              param: {\n                anyOf: [\n                  { type: 'null' },\n                  {\n                    type: 'object',\n                    additionalProperties: true,\n                  },\n                ],\n                description: 'Parameter of the action',\n              },\n              locate: {\n                type: ['object', 'null'],\n                properties: {\n                  id: { type: 'string' },\n                  prompt: { type: 'string' },\n                },\n                required: ['id', 'prompt'],\n                additionalProperties: false,\n                description: 'Location information for the target element',\n              },\n            },\n            required: ['thought', 'type', 'param', 'locate'],\n            additionalProperties: false,\n          },\n          description: 'List of actions to be performed',\n        },\n        more_actions_needed_by_instruction: {\n          type: 'boolean',\n          description:\n            'If all the actions described in the instruction have been covered by this action and logs, set this field to false.',\n        },\n        log: {\n          type: 'string',\n          description:\n            'Log what these planned actions do. Do not include further actions that have not been planned.',\n        },\n        error: {\n          type: ['string', 'null'],\n          description: 'Error messages about unexpected situations',\n        },\n      },\n      required: [\n        'actions',\n        'more_actions_needed_by_instruction',\n        'log',\n        'error',\n      ],\n      additionalProperties: false,\n    },\n  },\n};\n"],"names":["commonOutputFields","vlLocateParam","vlMode","bboxDescription","descriptionForAction","action","locatorSchemaTypeDescription","tab","fields","_schema__def","paramLines","schema","isZodObject","shape","getTypeName","field","_actualField__def","unwrapField","f","typeName","actualField","fieldTypeName","ifMidsceneLocatorField","_actualField__def_values","values","option","String","_actualField__def2","options","types","opt","console","getDescription","key","Object","isOptional","keyWithOptional","description","paramLine","line","_schema__def1","schemaTypeName","paramDescription","systemPromptToTaskPlanning","actionSpace","includeBbox","thinkingStrategy","Error","actionDescriptionList","undefined","actionList","logFieldDefinition","exampleLogField","planSchema"],"mappings":";;AASA,MAAMA,qBAAqB,CAAC;+NACmM,CAAC;AAEhO,MAAMC,gBAAgB,CAACC;IACrB,IAAIA,QACF,OAAO,CAAC,6DAA6D,EAAEC,gBAAgBD,SAAS;IAElG,OAAO;AACT;AAEO,MAAME,uBAAuB,CAClCC,QACAC;IAEA,MAAMC,MAAM;IACZ,MAAMC,SAAmB,EAAE;IAG3BA,OAAO,IAAI,CAAC,CAAC,SAAS,EAAEH,OAAO,IAAI,CAAC,CAAC,CAAC;IAGtC,IAAIA,OAAO,WAAW,EAAE;YAKFI;QAJpB,MAAMC,aAAuB,EAAE;QAG/B,MAAMC,SAASN,OAAO,WAAW;QACjC,MAAMO,cAAcH,AAAAA,SAAAA,CAAAA,eAAAA,OAAO,IAAI,AAAD,IAAVA,KAAAA,IAAAA,aAAa,QAAQ,AAAD,MAAM;QAE9C,IAAIG,eAAeD,OAAO,KAAK,EAAE;YAE/B,MAAME,QAAQF,OAAO,KAAK;YAG1B,MAAMG,cAAc,CAACC;oBA4BGC;gBA1BtB,MAAMC,cAAc,CAACC;oBACnB,IAAI,CAACA,EAAE,IAAI,EAAE,OAAOA;oBAEpB,MAAMC,WAAWD,EAAE,IAAI,CAAC,QAAQ;oBAGhC,IACEC,AAAa,kBAAbA,YACAA,AAAa,kBAAbA,YACAA,AAAa,iBAAbA,UAEA,OAAOF,YAAYC,EAAE,IAAI,CAAC,SAAS;oBAIrC,IAAIC,AAAa,iBAAbA,UAEF;wBAAA,IAAID,EAAE,IAAI,CAAC,MAAM,EACf,OAAOD,YAAYC,EAAE,IAAI,CAAC,MAAM;oBAClC;oBAGF,OAAOA;gBACT;gBAEA,MAAME,cAAcH,YAAYF;gBAChC,MAAMM,gBAAgB,QAAAL,CAAAA,oBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,kBAAkB,QAAQ;gBAEhD,IAAIK,AAAkB,gBAAlBA,eAA+B,OAAO;gBAC1C,IAAIA,AAAkB,gBAAlBA,eAA+B,OAAO;gBAC1C,IAAIA,AAAkB,iBAAlBA,eAAgC,OAAO;gBAC3C,IAAIA,AAAkB,eAAlBA,eAA8B,OAAO;gBACzC,IAAIA,AAAkB,gBAAlBA,eAA+B;oBAEjC,IAAIC,uBAAuBF,cACzB,OAAOd;oBAET,OAAO;gBACT;gBACA,IAAIe,AAAkB,cAAlBA,eAA6B;wBAE5BE,0BAAAA;oBADH,MAAMC,SACJ,AAAC,SAAAD,CAAAA,qBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,QAAAA,CAAAA,2BAAAA,mBAAkB,MAAM,AAAD,IAAvBA,KAAAA,IAAAA,yBACG,GAAG,CAAC,CAACE,SAAoBC,OAAO,CAAC,CAAC,EAAED,OAAO,CAAC,CAAC,GAC9C,IAAI,CAAC,KAAI,KAAK;oBAEnB,OAAO,CAAC,KAAK,EAAED,OAAO,CAAC,CAAC;gBAC1B;gBAEA,IAAIH,AAAkB,eAAlBA,eAA8B;wBAChBM;oBAAhB,MAAMC,UAAU,QAAAD,CAAAA,qBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,mBAAkB,OAAO;oBACzC,IAAIC,WAAWA,QAAQ,MAAM,GAAG,GAAG;wBAEjC,MAAMC,QAAQD,QAAQ,GAAG,CAAC,CAACE,MAAahB,YAAYgB;wBACpD,OAAOD,MAAM,IAAI,CAAC;oBACpB;oBACA,OAAO;gBACT;gBAEAE,QAAQ,IAAI,CACV,2EACAX,YAAY,IAAI;gBAElB,OAAOA,YAAY,QAAQ;YAC7B;YAGA,MAAMY,iBAAiB,CAACjB;oBAwClBC;gBAtCJ,MAAMC,cAAc,CAACC;oBACnB,IAAI,CAACA,EAAE,IAAI,EAAE,OAAOA;oBAEpB,MAAMC,WAAWD,EAAE,IAAI,CAAC,QAAQ;oBAGhC,IACEC,AAAa,kBAAbA,YACAA,AAAa,kBAAbA,YACAA,AAAa,iBAAbA,UAEA,OAAOF,YAAYC,EAAE,IAAI,CAAC,SAAS;oBAIrC,IAAIC,AAAa,iBAAbA,UAEF;wBAAA,IAAID,EAAE,IAAI,CAAC,MAAM,EACf,OAAOD,YAAYC,EAAE,IAAI,CAAC,MAAM;oBAClC;oBAGF,OAAOA;gBACT;gBAGA,IAAI,iBAAiBH,OACnB,OAAOA,MAAM,WAAW,IAAI;gBAG9B,MAAMK,cAAcH,YAAYF;gBAGhC,IAAI,iBAAiBK,aACnB,OAAOA,YAAY,WAAW,IAAI;gBAIpC,IAAIJ,AAAAA,SAAAA,CAAAA,oBAAAA,YAAY,IAAI,AAAD,IAAfA,KAAAA,IAAAA,kBAAkB,QAAQ,AAAD,MAAM,aACjC;oBAAA,IAAI,kCAAkCI,YAAY,IAAI,CAAC,KAAK,IAC1D,OAAO;gBACT;gBAGF,OAAO;YACT;YAEA,KAAK,MAAM,CAACa,KAAKlB,MAAM,IAAImB,OAAO,OAAO,CAACrB,OACxC,IAAIE,SAAS,AAAiB,YAAjB,OAAOA,OAAoB;gBAEtC,MAAMoB,aACJ,AAAqC,cAArC,OAAQpB,MAAc,UAAU,IAC/BA,MAAc,UAAU;gBAC3B,MAAMqB,kBAAkBD,aAAa,GAAGF,IAAI,CAAC,CAAC,GAAGA;gBAGjD,MAAMd,WAAWL,YAAYC;gBAG7B,MAAMsB,cAAcL,eAAejB;gBAGnC,IAAIuB,YAAY,GAAGF,gBAAgB,EAAE,EAAEjB,UAAU;gBACjD,IAAIkB,aACFC,aAAa,CAAC,IAAI,EAAED,aAAa;gBAGnC3B,WAAW,IAAI,CAAC4B;YAClB;YAIF,IAAI5B,WAAW,MAAM,GAAG,GAAG;gBACzBF,OAAO,IAAI,CAAC;gBACZE,WAAW,OAAO,CAAC,CAAC6B;oBAClB/B,OAAO,IAAI,CAAC,CAAC,IAAI,EAAE+B,MAAM;gBAC3B;YACF;QACF,OAAO;gBAGkBC;YAAvB,MAAMC,iBAAiB,QAAAD,CAAAA,gBAAAA,OAAO,IAAI,AAAD,IAAVA,KAAAA,IAAAA,cAAa,QAAQ;YAC5C,IAAIrB,WAAW;YAEf,IAAIsB,AAAmB,gBAAnBA,gBAAgCtB,WAAW;iBAC1C,IAAIsB,AAAmB,gBAAnBA,gBAAgCtB,WAAW;iBAC/C,IAAIsB,AAAmB,iBAAnBA,gBAAiCtB,WAAW;YAGrD,MAAMkB,cAAc,iBAAiB1B,SAASA,OAAO,WAAW,GAAG;YAGnE,IAAI+B,mBAAmB,CAAC,SAAS,EAAEvB,UAAU;YAC7C,IAAIkB,aACFK,oBAAoB,CAAC,IAAI,EAAEL,aAAa;YAE1CK,oBAAoB;YAEpBlC,OAAO,IAAI,CAACkC;QACd;IACF;IAEA,OAAO,CAAC,EAAE,EAAErC,OAAO,IAAI,CAAC,EAAE,EAAEA,OAAO,WAAW,IAAI,0BAA0B;AAC9E,EAAEE,MAAMC,OAAO,IAAI,CAAC,CAAC,EAAE,EAAED,KAAK,EAAE;AAChC,CAAC,CAAC,IAAI;AACN;AAEO,eAAeoC,2BAA2B,EAC/CC,WAAW,EACX1C,MAAM,EACN2C,WAAW,EACXC,gBAAgB,EAMjB;IAEC,IAAID,eAAe,CAAC3C,QAClB,MAAM,IAAI6C,MACR;IAIJ,MAAMC,wBAAwBJ,YAAY,GAAG,CAAC,CAACvC,SACtCD,qBACLC,QACAJ,cAAc4C,cAAc3C,SAAS+C;IAGzC,MAAMC,aAAaF,sBAAsB,IAAI,CAAC;IAG9C,MAAMG,qBACJL,AAAqB,UAArBA,mBACI,KACA;IAEN,MAAMM,kBACJN,AAAqB,UAArBA,mBACI,KACA;IAEN,OAAO,CAAC;;;;;;;;;;;AAWV,EAAEI,WAAW;;;;EAIX,EAAEC,qBAAqBnD,mBAAmB;;;;;;;;;;;;;;EAc1C,EAAEoD,gBAAgB;;;;QAIZ,EAAElD,SAAS,kCAAkC,GAAG;;;;;;;AAOxD,CAAC;AACD;AAEO,MAAMmD,aAAuC;IAClD,MAAM;IACN,aAAa;QACX,MAAM;QACN,QAAQ;QACR,QAAQ;YACN,MAAM;YACN,QAAQ;YACR,YAAY;gBACV,SAAS;oBACP,MAAM;oBACN,OAAO;wBACL,MAAM;wBACN,QAAQ;wBACR,YAAY;4BACV,SAAS;gCACP,MAAM;gCACN,aACE;4BACJ;4BACA,MAAM;gCACJ,MAAM;gCACN,aAAa;4BACf;4BACA,OAAO;gCACL,OAAO;oCACL;wCAAE,MAAM;oCAAO;oCACf;wCACE,MAAM;wCACN,sBAAsB;oCACxB;iCACD;gCACD,aAAa;4BACf;4BACA,QAAQ;gCACN,MAAM;oCAAC;oCAAU;iCAAO;gCACxB,YAAY;oCACV,IAAI;wCAAE,MAAM;oCAAS;oCACrB,QAAQ;wCAAE,MAAM;oCAAS;gCAC3B;gCACA,UAAU;oCAAC;oCAAM;iCAAS;gCAC1B,sBAAsB;gCACtB,aAAa;4BACf;wBACF;wBACA,UAAU;4BAAC;4BAAW;4BAAQ;4BAAS;yBAAS;wBAChD,sBAAsB;oBACxB;oBACA,aAAa;gBACf;gBACA,oCAAoC;oBAClC,MAAM;oBACN,aACE;gBACJ;gBACA,KAAK;oBACH,MAAM;oBACN,aACE;gBACJ;gBACA,OAAO;oBACL,MAAM;wBAAC;wBAAU;qBAAO;oBACxB,aAAa;gBACf;YACF;YACA,UAAU;gBACR;gBACA;gBACA;gBACA;aACD;YACD,sBAAsB;QACxB;IACF;AACF"}

package/dist/es/types.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"types.mjs","sources":["webpack://@midscene/core/./src/types.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport type { NodeType } from '@midscene/shared/constants';\nimport type { CreateOpenAIClientFn, TModelConfig } from '@midscene/shared/env';\nimport type {\n  BaseElement,\n  LocateResultElement,\n  Rect,\n  Size,\n} from '@midscene/shared/types';\nimport type { z } from 'zod';\nimport type { TUserPrompt } from './common';\nimport type { DetailedLocateParam, MidsceneYamlFlowItem } from './yaml';\n\nexport type {\n  ElementTreeNode,\n  BaseElement,\n  Rect,\n  Size,\n  Point,\n} from '@midscene/shared/types';\nexport * from './yaml';\n\nexport type AIUsageInfo = Record<string, any> & {\n  prompt_tokens: number | undefined;\n  completion_tokens: number | undefined;\n  total_tokens: number | undefined;\n  time_cost: number | undefined;\n  model_name: string | undefined;\n  model_description: string | undefined;\n  intent: string | undefined;\n};\n\nexport type { LocateResultElement };\n\n/**\n * openai\n *\n */\nexport enum AIResponseFormat {\n  JSON = 'json_object',\n  TEXT = 'text',\n}\n\nexport type AISingleElementResponseById = {\n  id: string;\n  reason?: string;\n  text?: string;\n  xpaths?: string[];\n};\n\nexport type AISingleElementResponseByPosition = {\n  position?: {\n    x: number;\n    y: number;\n  };\n  bbox?: [number, number, number, number];\n  reason: string;\n  text: string;\n};\n\nexport type AISingleElementResponse = AISingleElementResponseById;\n\nexport interface AIElementCoordinatesResponse {\n  bbox: [number, number, number, number];\n  errors?: string[];\n}\n\nexport type AIElementResponse = AIElementCoordinatesResponse;\n\nexport interface AIDataExtractionResponse<DataDemand> {\n  data: DataDemand;\n  errors?: string[];\n  thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n  bbox: [number, number, number, number];\n  references_bbox?: [number, number, number, number][];\n  error?: string;\n}\n\nexport interface AIAssertionResponse {\n  pass: boolean;\n  thought: string;\n}\n\nexport interface AIDescribeElementResponse {\n  description: string;\n  error?: string;\n}\n\nexport interface LocatorValidatorOption {\n  centerDistanceThreshold?: number;\n}\n\nexport interface LocateValidatorResult {\n  pass: boolean;\n  rect: Rect;\n  center: [number, number];\n  centerDistance?: number;\n}\n\nexport interface AgentDescribeElementAtPointResult {\n  prompt: string;\n  deepThink: boolean;\n  verifyResult?: LocateValidatorResult;\n}\n\n/**\n * context\n */\n\nexport abstract class UIContext {\n  abstract screenshotBase64: string;\n\n  abstract size: Size;\n\n  abstract _isFrozen?: boolean;\n}\n\nexport type EnsureObject<T> = { [K in keyof T]: any };\n\nexport type ServiceAction = 'locate' | 'extract' | 'assert' | 'describe';\n\nexport type ServiceExtractParam = string | Record<string, string>;\n\nexport type ElementCacheFeature = Record<string, unknown>;\n\nexport interface LocateResult {\n  element: LocateResultElement | null;\n  rect?: Rect;\n}\n\nexport type ThinkingLevel = 'off' | 'medium' | 'high';\n\nexport interface ServiceTaskInfo {\n  durationMs: number;\n  formatResponse?: string;\n  rawResponse?: string;\n  usage?: AIUsageInfo;\n  searchArea?: Rect;\n  searchAreaRawResponse?: string;\n  searchAreaUsage?: AIUsageInfo;\n}\n\nexport interface DumpMeta {\n  logTime: number;\n}\n\nexport interface ReportDumpWithAttributes {\n  dumpString: string;\n  attributes?: Record<string, any>;\n}\n\nexport interface ServiceDump extends DumpMeta {\n  type: 'locate' | 'extract' | 'assert';\n  logId: string;\n  userQuery: {\n    element?: TUserPrompt;\n    dataDemand?: ServiceExtractParam;\n    assertion?: TUserPrompt;\n  };\n  matchedElement: LocateResultElement[];\n  matchedRect?: Rect;\n  deepThink?: boolean;\n  data: any;\n  assertionPass?: boolean;\n  assertionThought?: string;\n  taskInfo: ServiceTaskInfo;\n  error?: string;\n  output?: any;\n}\n\nexport type PartialServiceDumpFromSDK = Omit<\n  ServiceDump,\n  'logTime' | 'logId' | 'model_name'\n>;\n\nexport interface ServiceResultBase {\n  dump: ServiceDump;\n}\n\nexport type LocateResultWithDump = LocateResult & ServiceResultBase;\n\nexport interface ServiceExtractResult<T> extends ServiceResultBase {\n  data: T;\n  thought?: string;\n  usage?: AIUsageInfo;\n}\n\nexport class ServiceError extends Error {\n  dump: ServiceDump;\n\n  constructor(message: string, dump: ServiceDump) {\n    super(message);\n    this.name = 'ServiceError';\n    this.dump = dump;\n  }\n}\n\n// intermediate variables to optimize the return value by AI\nexport interface LiteUISection {\n  name: string;\n  description: string;\n  sectionCharacteristics: string;\n  textIds: string[];\n}\n\nexport type ElementById = (id: string) => BaseElement | null;\n\nexport type ServiceAssertionResponse = AIAssertionResponse & {\n  usage?: AIUsageInfo;\n};\n\n/**\n * agent\n */\n\nexport type OnTaskStartTip = (tip: string) => Promise<void> | void;\n\nexport interface AgentWaitForOpt {\n  checkIntervalMs?: number;\n  timeoutMs?: number;\n}\n\nexport interface AgentAssertOpt {\n  keepRawResponse?: boolean;\n}\n\n/**\n * planning\n *\n */\n\nexport interface PlanningLocateParam extends DetailedLocateParam {\n  bbox?: [number, number, number, number];\n}\n\nexport interface PlanningAction<ParamType = any> {\n  thought?: string;\n  type: string;\n  param: ParamType;\n}\n\nexport interface RawResponsePlanningAIResponse {\n  action: PlanningAction;\n  more_actions_needed_by_instruction: boolean;\n  log: string;\n  sleep?: number;\n  error?: string;\n}\n\nexport interface PlanningAIResponse\n  extends Omit<RawResponsePlanningAIResponse, 'action'> {\n  actions?: PlanningAction[];\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  yamlString?: string;\n}\n\nexport interface PlanningActionParamSleep {\n  timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n  thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {};\n\nexport interface LongPressParam {\n  duration?: number;\n}\n\nexport interface PullParam {\n  direction: 'up' | 'down';\n  distance?: number;\n  duration?: number;\n}\n/**\n * misc\n */\n\nexport interface Color {\n  name: string;\n  hex: string;\n}\n\nexport interface BaseAgentParserOpt {\n  selector?: string;\n}\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PuppeteerParserOpt extends BaseAgentParserOpt {}\n\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PlaywrightParserOpt extends BaseAgentParserOpt {}\n\n/*\naction\n*/\nexport interface ExecutionTaskProgressOptions {\n  onTaskStart?: (task: ExecutionTask) => Promise<void> | void;\n}\n\nexport interface ExecutionRecorderItem {\n  type: 'screenshot';\n  ts: number;\n  screenshot?: string;\n  timing?: string;\n}\n\nexport type ExecutionTaskType = 'Planning' | 'Insight' | 'Action Space' | 'Log';\n\nexport interface ExecutorContext {\n  task: ExecutionTask;\n  element?: LocateResultElement | null;\n  uiContext?: UIContext;\n}\n\nexport interface ExecutionTaskApply<\n  Type extends ExecutionTaskType = any,\n  TaskParam = any,\n  TaskOutput = any,\n  TaskLog = any,\n> {\n  type: Type;\n  subType?: string;\n  subTask?: boolean;\n  param?: TaskParam;\n  thought?: string;\n  uiContext?: UIContext;\n  executor: (\n    param: TaskParam,\n    context: ExecutorContext,\n  ) => // biome-ignore lint/suspicious/noConfusingVoidType: <explanation>\n    | Promise<ExecutionTaskReturn<TaskOutput, TaskLog> | undefined | void>\n    | undefined\n    | void;\n}\n\nexport interface ExecutionTaskHitBy {\n  from: string;\n  context: Record<string, any>;\n}\n\nexport interface ExecutionTaskReturn<TaskOutput = unknown, TaskLog = unknown> {\n  output?: TaskOutput;\n  log?: TaskLog;\n  recorder?: ExecutionRecorderItem[];\n  hitBy?: ExecutionTaskHitBy;\n}\n\nexport type ExecutionTask<\n  E extends ExecutionTaskApply<any, any, any> = ExecutionTaskApply<\n    any,\n    any,\n    any\n  >,\n> = E &\n  ExecutionTaskReturn<\n    E extends ExecutionTaskApply<any, any, infer TaskOutput, any>\n      ? TaskOutput\n      : unknown,\n    E extends ExecutionTaskApply<any, any, any, infer TaskLog>\n      ? TaskLog\n      : unknown\n  > & {\n    status: 'pending' | 'running' | 'finished' | 'failed' | 'cancelled';\n    error?: Error;\n    errorMessage?: string;\n    errorStack?: string;\n    timing?: {\n      start: number;\n      end?: number;\n      cost?: number;\n    };\n    usage?: AIUsageInfo;\n    searchAreaUsage?: AIUsageInfo;\n  };\n\nexport interface ExecutionDump extends DumpMeta {\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n  aiActionContext?: string;\n}\n\n/*\ntask - service-locate\n*/\nexport type ExecutionTaskInsightLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskInsightLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport type ExecutionTaskInsightDump = ServiceDump;\n\nexport type ExecutionTaskInsightLocateApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightLocateParam,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightLocate =\n  ExecutionTask<ExecutionTaskInsightLocateApply>;\n\n/*\ntask - service-query\n*/\nexport interface ExecutionTaskInsightQueryParam {\n  dataDemand: ServiceExtractParam;\n}\n\nexport interface ExecutionTaskInsightQueryOutput {\n  data: any;\n}\n\nexport type ExecutionTaskInsightQueryApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightQueryParam,\n  any,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightQuery =\n  ExecutionTask<ExecutionTaskInsightQueryApply>;\n\n/*\ntask - assertion\n*/\nexport interface ExecutionTaskInsightAssertionParam {\n  assertion: string;\n}\n\nexport type ExecutionTaskInsightAssertionApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightAssertionParam,\n  ServiceAssertionResponse,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightAssertion =\n  ExecutionTask<ExecutionTaskInsightAssertionApply>;\n\n/*\ntask - action (i.e. interact) \n*/\nexport type ExecutionTaskActionApply<ActionParam = any> = ExecutionTaskApply<\n  'Action Space',\n  ActionParam,\n  void,\n  void\n>;\n\nexport type ExecutionTaskAction = ExecutionTask<ExecutionTaskActionApply>;\n\n/*\ntask - Log\n*/\n\nexport type ExecutionTaskLogApply<\n  LogParam = {\n    content: string;\n  },\n> = ExecutionTaskApply<'Log', LogParam, void, void>;\n\nexport type ExecutionTaskLog = ExecutionTask<ExecutionTaskLogApply>;\n\n/*\ntask - planning\n*/\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n  'Planning',\n  {\n    userInstruction: string;\n    aiActionContext?: string;\n  },\n  PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\ntask - planning-locate\n*/\nexport type ExecutionTaskPlanningLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskPlanningLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport type ExecutionTaskPlanningDump = ServiceDump;\n\nexport type ExecutionTaskPlanningLocateApply = ExecutionTaskApply<\n  'Planning',\n  ExecutionTaskPlanningLocateParam,\n  ExecutionTaskPlanningLocateOutput,\n  ExecutionTaskPlanningDump\n>;\n\nexport type ExecutionTaskPlanningLocate =\n  ExecutionTask<ExecutionTaskPlanningLocateApply>;\n\n/*\nGrouped dump\n*/\nexport interface GroupedActionDump {\n  sdkVersion: string;\n  groupName: string;\n  groupDescription?: string;\n  modelBriefs: string[];\n  executions: ExecutionDump[];\n}\n\nexport type InterfaceType =\n  | 'puppeteer'\n  | 'playwright'\n  | 'static'\n  | 'chrome-extension-proxy'\n  | 'android'\n  | string;\n\nexport interface StreamingCodeGenerationOptions {\n  /** Whether to enable streaming output */\n  stream?: boolean;\n  /** Callback function to handle streaming chunks */\n  onChunk?: StreamingCallback;\n  /** Callback function to handle streaming completion */\n  onComplete?: (finalCode: string) => void;\n  /** Callback function to handle streaming errors */\n  onError?: (error: Error) => void;\n}\n\nexport type StreamingCallback = (chunk: CodeGenerationChunk) => void;\n\nexport interface CodeGenerationChunk {\n  /** The incremental content chunk */\n  content: string;\n  /** The reasoning content */\n  reasoning_content: string;\n  /** The accumulated content so far */\n  accumulated: string;\n  /** Whether this is the final chunk */\n  isComplete: boolean;\n  /** Token usage information if available */\n  usage?: AIUsageInfo;\n}\n\nexport interface StreamingAIResponse {\n  /** The final accumulated content */\n  content: string;\n  /** Token usage information */\n  usage?: AIUsageInfo;\n  /** Whether the response was streamed */\n  isStreamed: boolean;\n}\n\nexport interface DeviceAction<TParam = any, TReturn = any> {\n  name: string;\n  description?: string;\n  interfaceAlias?: string;\n  paramSchema?: z.ZodType<TParam>;\n  call: (param: TParam, context: ExecutorContext) => Promise<TReturn> | TReturn;\n  delayAfterRunner?: number;\n}\n\n/**\n * Type utilities for extracting types from DeviceAction definitions\n */\n\n/**\n * Extract parameter type from a DeviceAction\n */\nexport type ActionParam<Action extends DeviceAction<any, any>> =\n  Action extends DeviceAction<infer P, any> ? P : never;\n\n/**\n * Extract return type from a DeviceAction\n */\nexport type ActionReturn<Action extends DeviceAction<any, any>> =\n  Action extends DeviceAction<any, infer R> ? R : never;\n\n/**\n * Web-specific types\n */\nexport interface WebElementInfo extends BaseElement {\n  id: string;\n  attributes: {\n    nodeType: NodeType;\n    [key: string]: string;\n  };\n}\n\nexport type WebUIContext = UIContext;\n\n/**\n * Agent\n */\n\nexport type CacheConfig = {\n  strategy?: 'read-only' | 'read-write' | 'write-only';\n  id: string;\n};\n\nexport type Cache =\n  | false // No read, no write\n  | true // Will throw error at runtime - deprecated\n  | CacheConfig; // Object configuration (requires explicit id)\n\nexport interface AgentOpt {\n  testId?: string;\n  // @deprecated\n  cacheId?: string; // Keep backward compatibility, but marked as deprecated\n  groupName?: string;\n  groupDescription?: string;\n  /* if auto generate report, default true */\n  generateReport?: boolean;\n  /* if auto print report msg, default true */\n  autoPrintReportMsg?: boolean;\n  onTaskStartTip?: OnTaskStartTip;\n  aiActionContext?: string;\n  /* custom report file name */\n  reportFileName?: string;\n  modelConfig?: TModelConfig;\n  cache?: Cache;\n  replanningCycleLimit?: number;\n\n  /**\n   * Custom OpenAI client factory function\n   *\n   * If provided, this function will be called to create OpenAI client instances\n   * for each AI call, allowing you to:\n   * - Wrap clients with observability tools (langsmith, langfuse)\n   * - Use custom OpenAI-compatible clients\n   * - Apply different configurations based on intent\n   *\n   * @param config - Resolved model configuration\n   * @returns OpenAI client instance (original or wrapped)\n   *\n   * @example\n   * ```typescript\n   * createOpenAIClient: async (openai, opts) => {\n   *   // Wrap with langsmith for planning tasks\n   *   if (opts.baseURL?.includes('planning')) {\n   *     return wrapOpenAI(openai, { metadata: { task: 'planning' } });\n   *   }\n   *\n   *   return openai;\n   * }\n   * ```\n   */\n  createOpenAIClient?: CreateOpenAIClientFn;\n}\n\nexport type TestStatus =\n  | 'passed'\n  | 'failed'\n  | 'timedOut'\n  | 'skipped'\n  | 'interrupted';\n\nexport interface ReportFileWithAttributes {\n  reportFilePath: string;\n  reportAttributes: {\n    testDuration: number;\n    testStatus: TestStatus;\n    testTitle: string;\n    testId: string;\n    testDescription: string;\n  };\n}\n"],"names":["AIResponseFormat","UIContext","ServiceError","Error","message","dump"],"mappings":";AAAqD;;;;;;;;;;AAuC9C,IAAKA,yBAAgBA,WAAAA,GAAAA,SAAhBA,gBAAgB;;;WAAhBA;;AA0EL,MAAeC;AAMtB;AAwEO,MAAMC,qBAAqBC;IAGhC,YAAYC,OAAe,EAAEC,IAAiB,CAAE;QAC9C,KAAK,CAACD,UAHR;QAIE,IAAI,CAAC,IAAI,GAAG;QACZ,IAAI,CAAC,IAAI,GAAGC;IACd;AACF"}
+{"version":3,"file":"types.mjs","sources":["webpack://@midscene/core/./src/types.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport type { NodeType } from '@midscene/shared/constants';\nimport type { CreateOpenAIClientFn, TModelConfig } from '@midscene/shared/env';\nimport type {\n  BaseElement,\n  LocateResultElement,\n  Rect,\n  Size,\n} from '@midscene/shared/types';\nimport type { z } from 'zod';\nimport type { TUserPrompt } from './common';\nimport type { DetailedLocateParam, MidsceneYamlFlowItem } from './yaml';\n\nexport type {\n  ElementTreeNode,\n  BaseElement,\n  Rect,\n  Size,\n  Point,\n} from '@midscene/shared/types';\nexport * from './yaml';\n\nexport type AIUsageInfo = Record<string, any> & {\n  prompt_tokens: number | undefined;\n  completion_tokens: number | undefined;\n  total_tokens: number | undefined;\n  time_cost: number | undefined;\n  model_name: string | undefined;\n  model_description: string | undefined;\n  intent: string | undefined;\n};\n\nexport type { LocateResultElement };\n\n/**\n * openai\n *\n */\nexport enum AIResponseFormat {\n  JSON = 'json_object',\n  TEXT = 'text',\n}\n\nexport type AISingleElementResponseById = {\n  id: string;\n  reason?: string;\n  text?: string;\n  xpaths?: string[];\n};\n\nexport type AISingleElementResponseByPosition = {\n  position?: {\n    x: number;\n    y: number;\n  };\n  bbox?: [number, number, number, number];\n  reason: string;\n  text: string;\n};\n\nexport type AISingleElementResponse = AISingleElementResponseById;\n\nexport interface AIElementCoordinatesResponse {\n  bbox: [number, number, number, number];\n  errors?: string[];\n}\n\nexport type AIElementResponse = AIElementCoordinatesResponse;\n\nexport interface AIDataExtractionResponse<DataDemand> {\n  data: DataDemand;\n  errors?: string[];\n  thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n  bbox: [number, number, number, number];\n  references_bbox?: [number, number, number, number][];\n  error?: string;\n}\n\nexport interface AIAssertionResponse {\n  pass: boolean;\n  thought: string;\n}\n\nexport interface AIDescribeElementResponse {\n  description: string;\n  error?: string;\n}\n\nexport interface LocatorValidatorOption {\n  centerDistanceThreshold?: number;\n}\n\nexport interface LocateValidatorResult {\n  pass: boolean;\n  rect: Rect;\n  center: [number, number];\n  centerDistance?: number;\n}\n\nexport interface AgentDescribeElementAtPointResult {\n  prompt: string;\n  deepThink: boolean;\n  verifyResult?: LocateValidatorResult;\n}\n\n/**\n * context\n */\n\nexport abstract class UIContext {\n  abstract screenshotBase64: string;\n\n  abstract size: Size;\n\n  abstract _isFrozen?: boolean;\n}\n\nexport type EnsureObject<T> = { [K in keyof T]: any };\n\nexport type ServiceAction = 'locate' | 'extract' | 'assert' | 'describe';\n\nexport type ServiceExtractParam = string | Record<string, string>;\n\nexport type ElementCacheFeature = Record<string, unknown>;\n\nexport interface LocateResult {\n  element: LocateResultElement | null;\n  rect?: Rect;\n}\n\nexport type ThinkingLevel = 'off' | 'medium' | 'high';\nexport type ThinkingStrategy = 'off' | 'cot';\n\nexport interface ServiceTaskInfo {\n  durationMs: number;\n  formatResponse?: string;\n  rawResponse?: string;\n  usage?: AIUsageInfo;\n  searchArea?: Rect;\n  searchAreaRawResponse?: string;\n  searchAreaUsage?: AIUsageInfo;\n}\n\nexport interface DumpMeta {\n  logTime: number;\n}\n\nexport interface ReportDumpWithAttributes {\n  dumpString: string;\n  attributes?: Record<string, any>;\n}\n\nexport interface ServiceDump extends DumpMeta {\n  type: 'locate' | 'extract' | 'assert';\n  logId: string;\n  userQuery: {\n    element?: TUserPrompt;\n    dataDemand?: ServiceExtractParam;\n    assertion?: TUserPrompt;\n  };\n  matchedElement: LocateResultElement[];\n  matchedRect?: Rect;\n  deepThink?: boolean;\n  data: any;\n  assertionPass?: boolean;\n  assertionThought?: string;\n  taskInfo: ServiceTaskInfo;\n  error?: string;\n  output?: any;\n}\n\nexport type PartialServiceDumpFromSDK = Omit<\n  ServiceDump,\n  'logTime' | 'logId' | 'model_name'\n>;\n\nexport interface ServiceResultBase {\n  dump: ServiceDump;\n}\n\nexport type LocateResultWithDump = LocateResult & ServiceResultBase;\n\nexport interface ServiceExtractResult<T> extends ServiceResultBase {\n  data: T;\n  thought?: string;\n  usage?: AIUsageInfo;\n}\n\nexport class ServiceError extends Error {\n  dump: ServiceDump;\n\n  constructor(message: string, dump: ServiceDump) {\n    super(message);\n    this.name = 'ServiceError';\n    this.dump = dump;\n  }\n}\n\n// intermediate variables to optimize the return value by AI\nexport interface LiteUISection {\n  name: string;\n  description: string;\n  sectionCharacteristics: string;\n  textIds: string[];\n}\n\nexport type ElementById = (id: string) => BaseElement | null;\n\nexport type ServiceAssertionResponse = AIAssertionResponse & {\n  usage?: AIUsageInfo;\n};\n\n/**\n * agent\n */\n\nexport type OnTaskStartTip = (tip: string) => Promise<void> | void;\n\nexport interface AgentWaitForOpt {\n  checkIntervalMs?: number;\n  timeoutMs?: number;\n}\n\nexport interface AgentAssertOpt {\n  keepRawResponse?: boolean;\n}\n\n/**\n * planning\n *\n */\n\nexport interface PlanningLocateParam extends DetailedLocateParam {\n  bbox?: [number, number, number, number];\n}\n\nexport interface PlanningAction<ParamType = any> {\n  thought?: string;\n  type: string;\n  param: ParamType;\n}\n\nexport interface RawResponsePlanningAIResponse {\n  action: PlanningAction;\n  more_actions_needed_by_instruction: boolean;\n  log: string;\n  sleep?: number;\n  error?: string;\n}\n\nexport interface PlanningAIResponse\n  extends Omit<RawResponsePlanningAIResponse, 'action'> {\n  actions?: PlanningAction[];\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  yamlString?: string;\n}\n\nexport interface PlanningActionParamSleep {\n  timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n  thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {};\n\nexport interface LongPressParam {\n  duration?: number;\n}\n\nexport interface PullParam {\n  direction: 'up' | 'down';\n  distance?: number;\n  duration?: number;\n}\n/**\n * misc\n */\n\nexport interface Color {\n  name: string;\n  hex: string;\n}\n\nexport interface BaseAgentParserOpt {\n  selector?: string;\n}\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PuppeteerParserOpt extends BaseAgentParserOpt {}\n\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PlaywrightParserOpt extends BaseAgentParserOpt {}\n\n/*\naction\n*/\nexport interface ExecutionTaskProgressOptions {\n  onTaskStart?: (task: ExecutionTask) => Promise<void> | void;\n}\n\nexport interface ExecutionRecorderItem {\n  type: 'screenshot';\n  ts: number;\n  screenshot?: string;\n  timing?: string;\n}\n\nexport type ExecutionTaskType = 'Planning' | 'Insight' | 'Action Space' | 'Log';\n\nexport interface ExecutorContext {\n  task: ExecutionTask;\n  element?: LocateResultElement | null;\n  uiContext?: UIContext;\n}\n\nexport interface ExecutionTaskApply<\n  Type extends ExecutionTaskType = any,\n  TaskParam = any,\n  TaskOutput = any,\n  TaskLog = any,\n> {\n  type: Type;\n  subType?: string;\n  subTask?: boolean;\n  param?: TaskParam;\n  thought?: string;\n  uiContext?: UIContext;\n  executor: (\n    param: TaskParam,\n    context: ExecutorContext,\n  ) => // biome-ignore lint/suspicious/noConfusingVoidType: <explanation>\n    | Promise<ExecutionTaskReturn<TaskOutput, TaskLog> | undefined | void>\n    | undefined\n    | void;\n}\n\nexport interface ExecutionTaskHitBy {\n  from: string;\n  context: Record<string, any>;\n}\n\nexport interface ExecutionTaskReturn<TaskOutput = unknown, TaskLog = unknown> {\n  output?: TaskOutput;\n  log?: TaskLog;\n  recorder?: ExecutionRecorderItem[];\n  hitBy?: ExecutionTaskHitBy;\n}\n\nexport type ExecutionTask<\n  E extends ExecutionTaskApply<any, any, any> = ExecutionTaskApply<\n    any,\n    any,\n    any\n  >,\n> = E &\n  ExecutionTaskReturn<\n    E extends ExecutionTaskApply<any, any, infer TaskOutput, any>\n      ? TaskOutput\n      : unknown,\n    E extends ExecutionTaskApply<any, any, any, infer TaskLog>\n      ? TaskLog\n      : unknown\n  > & {\n    status: 'pending' | 'running' | 'finished' | 'failed' | 'cancelled';\n    error?: Error;\n    errorMessage?: string;\n    errorStack?: string;\n    timing?: {\n      start: number;\n      end?: number;\n      cost?: number;\n    };\n    usage?: AIUsageInfo;\n    searchAreaUsage?: AIUsageInfo;\n  };\n\nexport interface ExecutionDump extends DumpMeta {\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n  aiActionContext?: string;\n}\n\n/*\ntask - service-locate\n*/\nexport type ExecutionTaskInsightLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskInsightLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport type ExecutionTaskInsightDump = ServiceDump;\n\nexport type ExecutionTaskInsightLocateApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightLocateParam,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightLocate =\n  ExecutionTask<ExecutionTaskInsightLocateApply>;\n\n/*\ntask - service-query\n*/\nexport interface ExecutionTaskInsightQueryParam {\n  dataDemand: ServiceExtractParam;\n}\n\nexport interface ExecutionTaskInsightQueryOutput {\n  data: any;\n}\n\nexport type ExecutionTaskInsightQueryApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightQueryParam,\n  any,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightQuery =\n  ExecutionTask<ExecutionTaskInsightQueryApply>;\n\n/*\ntask - assertion\n*/\nexport interface ExecutionTaskInsightAssertionParam {\n  assertion: string;\n}\n\nexport type ExecutionTaskInsightAssertionApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightAssertionParam,\n  ServiceAssertionResponse,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightAssertion =\n  ExecutionTask<ExecutionTaskInsightAssertionApply>;\n\n/*\ntask - action (i.e. interact) \n*/\nexport type ExecutionTaskActionApply<ActionParam = any> = ExecutionTaskApply<\n  'Action Space',\n  ActionParam,\n  void,\n  void\n>;\n\nexport type ExecutionTaskAction = ExecutionTask<ExecutionTaskActionApply>;\n\n/*\ntask - Log\n*/\n\nexport type ExecutionTaskLogApply<\n  LogParam = {\n    content: string;\n  },\n> = ExecutionTaskApply<'Log', LogParam, void, void>;\n\nexport type ExecutionTaskLog = ExecutionTask<ExecutionTaskLogApply>;\n\n/*\ntask - planning\n*/\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n  'Planning',\n  {\n    userInstruction: string;\n    aiActionContext?: string;\n  },\n  PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\ntask - planning-locate\n*/\nexport type ExecutionTaskPlanningLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskPlanningLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport type ExecutionTaskPlanningDump = ServiceDump;\n\nexport type ExecutionTaskPlanningLocateApply = ExecutionTaskApply<\n  'Planning',\n  ExecutionTaskPlanningLocateParam,\n  ExecutionTaskPlanningLocateOutput,\n  ExecutionTaskPlanningDump\n>;\n\nexport type ExecutionTaskPlanningLocate =\n  ExecutionTask<ExecutionTaskPlanningLocateApply>;\n\n/*\nGrouped dump\n*/\nexport interface GroupedActionDump {\n  sdkVersion: string;\n  groupName: string;\n  groupDescription?: string;\n  modelBriefs: string[];\n  executions: ExecutionDump[];\n}\n\nexport type InterfaceType =\n  | 'puppeteer'\n  | 'playwright'\n  | 'static'\n  | 'chrome-extension-proxy'\n  | 'android'\n  | string;\n\nexport interface StreamingCodeGenerationOptions {\n  /** Whether to enable streaming output */\n  stream?: boolean;\n  /** Callback function to handle streaming chunks */\n  onChunk?: StreamingCallback;\n  /** Callback function to handle streaming completion */\n  onComplete?: (finalCode: string) => void;\n  /** Callback function to handle streaming errors */\n  onError?: (error: Error) => void;\n}\n\nexport type StreamingCallback = (chunk: CodeGenerationChunk) => void;\n\nexport interface CodeGenerationChunk {\n  /** The incremental content chunk */\n  content: string;\n  /** The reasoning content */\n  reasoning_content: string;\n  /** The accumulated content so far */\n  accumulated: string;\n  /** Whether this is the final chunk */\n  isComplete: boolean;\n  /** Token usage information if available */\n  usage?: AIUsageInfo;\n}\n\nexport interface StreamingAIResponse {\n  /** The final accumulated content */\n  content: string;\n  /** Token usage information */\n  usage?: AIUsageInfo;\n  /** Whether the response was streamed */\n  isStreamed: boolean;\n}\n\nexport interface DeviceAction<TParam = any, TReturn = any> {\n  name: string;\n  description?: string;\n  interfaceAlias?: string;\n  paramSchema?: z.ZodType<TParam>;\n  call: (param: TParam, context: ExecutorContext) => Promise<TReturn> | TReturn;\n  delayAfterRunner?: number;\n}\n\n/**\n * Type utilities for extracting types from DeviceAction definitions\n */\n\n/**\n * Extract parameter type from a DeviceAction\n */\nexport type ActionParam<Action extends DeviceAction<any, any>> =\n  Action extends DeviceAction<infer P, any> ? P : never;\n\n/**\n * Extract return type from a DeviceAction\n */\nexport type ActionReturn<Action extends DeviceAction<any, any>> =\n  Action extends DeviceAction<any, infer R> ? R : never;\n\n/**\n * Web-specific types\n */\nexport interface WebElementInfo extends BaseElement {\n  id: string;\n  attributes: {\n    nodeType: NodeType;\n    [key: string]: string;\n  };\n}\n\nexport type WebUIContext = UIContext;\n\n/**\n * Agent\n */\n\nexport type CacheConfig = {\n  strategy?: 'read-only' | 'read-write' | 'write-only';\n  id: string;\n};\n\nexport type Cache =\n  | false // No read, no write\n  | true // Will throw error at runtime - deprecated\n  | CacheConfig; // Object configuration (requires explicit id)\n\nexport interface AgentOpt {\n  testId?: string;\n  // @deprecated\n  cacheId?: string; // Keep backward compatibility, but marked as deprecated\n  groupName?: string;\n  groupDescription?: string;\n  /* if auto generate report, default true */\n  generateReport?: boolean;\n  /* if auto print report msg, default true */\n  autoPrintReportMsg?: boolean;\n  onTaskStartTip?: OnTaskStartTip;\n  aiActionContext?: string;\n  /* custom report file name */\n  reportFileName?: string;\n  modelConfig?: TModelConfig;\n  cache?: Cache;\n  replanningCycleLimit?: number;\n\n  /**\n   * Custom OpenAI client factory function\n   *\n   * If provided, this function will be called to create OpenAI client instances\n   * for each AI call, allowing you to:\n   * - Wrap clients with observability tools (langsmith, langfuse)\n   * - Use custom OpenAI-compatible clients\n   * - Apply different configurations based on intent\n   *\n   * @param config - Resolved model configuration\n   * @returns OpenAI client instance (original or wrapped)\n   *\n   * @example\n   * ```typescript\n   * createOpenAIClient: async (openai, opts) => {\n   *   // Wrap with langsmith for planning tasks\n   *   if (opts.baseURL?.includes('planning')) {\n   *     return wrapOpenAI(openai, { metadata: { task: 'planning' } });\n   *   }\n   *\n   *   return openai;\n   * }\n   * ```\n   */\n  createOpenAIClient?: CreateOpenAIClientFn;\n}\n\nexport type TestStatus =\n  | 'passed'\n  | 'failed'\n  | 'timedOut'\n  | 'skipped'\n  | 'interrupted';\n\nexport interface ReportFileWithAttributes {\n  reportFilePath: string;\n  reportAttributes: {\n    testDuration: number;\n    testStatus: TestStatus;\n    testTitle: string;\n    testId: string;\n    testDescription: string;\n  };\n}\n"],"names":["AIResponseFormat","UIContext","ServiceError","Error","message","dump"],"mappings":";AAAqD;;;;;;;;;;AAuC9C,IAAKA,yBAAgBA,WAAAA,GAAAA,SAAhBA,gBAAgB;;;WAAhBA;;AA0EL,MAAeC;AAMtB;AAyEO,MAAMC,qBAAqBC;IAGhC,YAAYC,OAAe,EAAEC,IAAiB,CAAE;QAC9C,KAAK,CAACD,UAHR;QAIE,IAAI,CAAC,IAAI,GAAG;QACZ,IAAI,CAAC,IAAI,GAAGC;IACd;AACF"}