@midscene/core 0.30.5-beta-20251017025531.0 → 0.30.5-beta-20251020035347.0

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

@@ -31,7 +31,6 @@ const logger_namespaceObject = require("@midscene/shared/logger");
 const utils_namespaceObject = require("@midscene/shared/utils");
 const external_common_js_namespaceObject = require("./common.js");
 const llm_planning_js_namespaceObject = require("./prompt/llm-planning.js");
-const util_js_namespaceObject = require("./prompt/util.js");
 const index_js_namespaceObject = require("./service-caller/index.js");
 const debug = (0, logger_namespaceObject.getDebug)('planning');
 async function plan(userInstruction, opts) {
@@ -39,9 +38,7 @@ async function plan(userInstruction, opts) {
     const { context, modelConfig, conversationHistory } = opts;
     const { screenshotBase64, size } = context;
     const { modelName, vlMode } = modelConfig;
-    const { description: pageDescription, elementById } = await (0, util_js_namespaceObject.describeUserPage)(context, {
-        vlMode
-    });
+    (0, utils_namespaceObject.assert)(vlMode, 'Planning requires vlMode to be configured.');
     const systemPrompt = await (0, llm_planning_js_namespaceObject.systemPromptToTaskPlanning)({
         actionSpace: opts.actionSpace,
         vlMode: vlMode
@@ -56,11 +53,7 @@ async function plan(userInstruction, opts) {
         imageWidth = paddedResult.width;
         imageHeight = paddedResult.height;
         imagePayload = paddedResult.imageBase64;
-    } else if ('qwen3-vl' === vlMode) ;
-    else if (!vlMode) imagePayload = await (0, external_common_js_namespaceObject.markupImageForLLM)(screenshotBase64, context.tree, {
-        width: imageWidth,
-        height: imageHeight
-    });
+    }
     (0, external_common_js_namespaceObject.warnGPT4oSizeLimit)(size, modelName);
     const historyLog = (null == (_opts_conversationHistory = opts.conversationHistory) ? void 0 : _opts_conversationHistory.snapshot()) || [];
     const knowledgeContext = opts.actionContext ? [
@@ -102,13 +95,7 @@ async function plan(userInstruction, opts) {
                         url: imagePayload,
                         detail: 'high'
                     }
-                },
-                ...vlMode ? [] : [
-                    {
-                        type: 'text',
-                        text: pageDescription
-                    }
-                ]
+                }
             ]
         }
     ];
@@ -134,11 +121,7 @@ async function plan(userInstruction, opts) {
         debug('locateFields', locateFields);
         locateFields.forEach((field)=>{
             const locateResult = action.param[field];
-            if (locateResult) if (vlMode) action.param[field] = (0, external_common_js_namespaceObject.fillBboxParam)(locateResult, imageWidth, imageHeight, rightLimit, bottomLimit, vlMode);
-            else {
-                const element = elementById(locateResult);
-                if (element) action.param[field].id = element.id;
-            }
+            if (locateResult) action.param[field] = (0, external_common_js_namespaceObject.fillBboxParam)(locateResult, imageWidth, imageHeight, rightLimit, bottomLimit, vlMode);
         });
     });
     (0, utils_namespaceObject.assert)(!planFromAI.error, `Failed to plan actions: ${planFromAI.error}`);

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

@@ -1 +1 @@
-{"version":3,"file":"ai-model/llm-planning.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/llm-planning.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n  DeviceAction,\n  InterfaceType,\n  PlanningAIResponse,\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 {\n  ChatCompletionContentPart,\n  ChatCompletionMessageParam,\n} from 'openai/resources/index';\nimport {\n  AIActionType,\n  buildYamlFlowFromPlans,\n  fillBboxParam,\n  findAllMidsceneLocatorField,\n  markupImageForLLM,\n  warnGPT4oSizeLimit,\n} from './common';\nimport type { ConversationHistory } from './conversation-history';\nimport { systemPromptToTaskPlanning } from './prompt/llm-planning';\nimport { describeUserPage } from './prompt/util';\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  },\n): Promise<PlanningAIResponse> {\n  const { context, modelConfig, conversationHistory } = opts;\n  const { screenshotBase64, size } = context;\n\n  const { modelName, vlMode } = modelConfig;\n\n  const { description: pageDescription, elementById } = await describeUserPage(\n    context,\n    { vlMode },\n  );\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    vlMode: vlMode,\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  if (vlMode === 'qwen-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  } else if (vlMode === 'qwen3-vl') {\n    // const paddedResult = await paddingToMatchBlockByBase64(imagePayload, 32);\n    // imageWidth = paddedResult.width;\n    // imageHeight = paddedResult.height;\n    // imagePayload = paddedResult.imageBase64;\n  } else if (!vlMode) {\n    imagePayload = await markupImageForLLM(screenshotBase64, context.tree, {\n      width: imageWidth,\n      height: imageHeight,\n    });\n  }\n\n  warnGPT4oSizeLimit(size, modelName);\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        ...(vlMode\n          ? []\n          : ([\n              {\n                type: 'text',\n                text: pageDescription,\n              },\n            ] as ChatCompletionContentPart[])),\n      ],\n    },\n  ];\n\n  const { content, usage } = await callAIWithObjectResponse<PlanningAIResponse>(\n    msgs,\n    AIActionType.PLAN,\n    modelConfig,\n  );\n  const rawResponse = JSON.stringify(content, undefined, 2);\n  const planFromAI = content;\n\n  const actions =\n    (planFromAI.action?.type ? [planFromAI.action] : planFromAI.actions) || [];\n  const returnValue: PlanningAIResponse = {\n    ...planFromAI,\n    actions,\n    rawResponse,\n    usage,\n    yamlFlow: buildYamlFlowFromPlans(\n      actions,\n      opts.actionSpace,\n      planFromAI.sleep,\n    ),\n  };\n\n  assert(planFromAI, \"can't get plans from AI\");\n\n  // TODO: use zod.parse to parse the action.param, and then fill the bbox param.\n  actions.forEach((action) => {\n    const type = action.type;\n    const actionInActionSpace = opts.actionSpace.find(\n      (action) => action.name === type,\n    );\n\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) {\n        if (vlMode) {\n          action.param[field] = fillBboxParam(\n            locateResult,\n            imageWidth,\n            imageHeight,\n            rightLimit,\n            bottomLimit,\n            vlMode,\n          );\n        } else {\n          const element = elementById(locateResult);\n          if (element) {\n            action.param[field].id = element.id;\n          }\n        }\n      }\n    });\n  });\n  // in Qwen-VL, error means error. In GPT-4o, error may mean more actions are needed.\n  assert(!planFromAI.error, `Failed to plan actions: ${planFromAI.error}`);\n\n  if (\n    actions.length === 0 &&\n    returnValue.more_actions_needed_by_instruction &&\n    !returnValue.sleep\n  ) {\n    console.warn(\n      'No actions planned for the prompt, but model said more actions are needed:',\n      userInstruction,\n    );\n  }\n\n  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":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","plan","userInstruction","opts","_opts_conversationHistory","_planFromAI_action","context","modelConfig","conversationHistory","screenshotBase64","size","modelName","vlMode","pageDescription","elementById","describeUserPage","systemPrompt","systemPromptToTaskPlanning","imagePayload","imageWidth","imageHeight","rightLimit","bottomLimit","paddedResult","paddingToMatchBlockByBase64","markupImageForLLM","warnGPT4oSizeLimit","historyLog","knowledgeContext","instruction","msgs","content","usage","callAIWithObjectResponse","AIActionType","rawResponse","JSON","undefined","planFromAI","actions","returnValue","buildYamlFlowFromPlans","assert","action","type","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam","element","console"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;ACqBA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAEhB,eAAeC,KACpBC,eAAuB,EACvBC,IAOC;QAwCkBC,2BAiEhBC;IAvGH,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAEC,mBAAmB,EAAE,GAAGL;IACtD,MAAM,EAAEM,gBAAgB,EAAEC,IAAI,EAAE,GAAGJ;IAEnC,MAAM,EAAEK,SAAS,EAAEC,MAAM,EAAE,GAAGL;IAE9B,MAAM,EAAE,aAAaM,eAAe,EAAEC,WAAW,EAAE,GAAG,MAAMC,AAAAA,IAAAA,wBAAAA,gBAAAA,AAAAA,EAC1DT,SACA;QAAEM;IAAO;IAEX,MAAMI,eAAe,MAAMC,AAAAA,IAAAA,gCAAAA,0BAAAA,AAAAA,EAA2B;QACpD,aAAad,KAAK,WAAW;QAC7B,QAAQS;IACV;IAEA,IAAIM,eAAeT;IACnB,IAAIU,aAAaT,KAAK,KAAK;IAC3B,IAAIU,cAAcV,KAAK,MAAM;IAC7B,MAAMW,aAAaF;IACnB,MAAMG,cAAcF;IACpB,IAAIR,AAAW,cAAXA,QAAsB;QACxB,MAAMW,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,2BAAAA,AAAAA,EAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC,OAAO,IAAIX,AAAW,eAAXA;SAKJ,IAAI,CAACA,QACVM,eAAe,MAAMO,AAAAA,IAAAA,mCAAAA,iBAAAA,AAAAA,EAAkBhB,kBAAkBH,QAAQ,IAAI,EAAE;QACrE,OAAOa;QACP,QAAQC;IACV;IAGFM,IAAAA,mCAAAA,kBAAAA,AAAAA,EAAmBhB,MAAMC;IAEzB,MAAMgB,aAAavB,AAAAA,SAAAA,CAAAA,4BAAAA,KAAK,mBAAmB,AAAD,IAAvBA,KAAAA,IAAAA,0BAA0B,QAAQ,EAAC,KAAK,EAAE;IAG7D,MAAMwB,mBAAiDzB,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,MAAM0B,cAA4C;QAChD;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,CAAC,kBAAkB,EAAE3B,gBAAgB,mBAAmB,CAAC;gBACjE;aACD;QACH;KACD;IAED,MAAM4B,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASd;QAAa;WACrCY;WACAC;WACAF;QACH;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKT;wBACL,QAAQ;oBACV;gBACF;mBACIN,SACA,EAAE,GACD;oBACC;wBACE,MAAM;wBACN,MAAMC;oBACR;iBACD;aACN;QACH;KACD;IAED,MAAM,EAAEkB,OAAO,EAAEC,KAAK,EAAE,GAAG,MAAMC,AAAAA,IAAAA,yBAAAA,wBAAAA,AAAAA,EAC/BH,MACAI,mCAAAA,YAAAA,CAAAA,IAAiB,EACjB3B;IAEF,MAAM4B,cAAcC,KAAK,SAAS,CAACL,SAASM,QAAW;IACvD,MAAMC,aAAaP;IAEnB,MAAMQ,UACHlC,AAAAA,CAAAA,SAAAA,CAAAA,qBAAAA,WAAW,MAAM,AAAD,IAAhBA,KAAAA,IAAAA,mBAAmB,IAAI,AAAD,IAAI;QAACiC,WAAW,MAAM;KAAC,GAAGA,WAAW,OAAM,KAAM,EAAE;IAC5E,MAAME,cAAkC;QACtC,GAAGF,UAAU;QACbC;QACAJ;QACAH;QACA,UAAUS,AAAAA,IAAAA,mCAAAA,sBAAAA,AAAAA,EACRF,SACApC,KAAK,WAAW,EAChBmC,WAAW,KAAK;IAEpB;IAEAI,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOJ,YAAY;IAGnBC,QAAQ,OAAO,CAAC,CAACI;QACf,MAAMC,OAAOD,OAAO,IAAI;QACxB,MAAME,sBAAsB1C,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACwC,SAAWA,OAAO,IAAI,KAAKC;QAG9B7C,MAAM,+BAA+B8C;QACrC,MAAMC,eAAeD,sBACjBE,AAAAA,IAAAA,mCAAAA,2BAAAA,AAAAA,EAA4BF,oBAAoB,WAAW,IAC3D,EAAE;QAEN9C,MAAM,gBAAgB+C;QAEtBA,aAAa,OAAO,CAAC,CAACE;YACpB,MAAMC,eAAeN,OAAO,KAAK,CAACK,MAAM;YACxC,IAAIC,cACF,IAAIrC,QACF+B,OAAO,KAAK,CAACK,MAAM,GAAGE,AAAAA,IAAAA,mCAAAA,aAAAA,AAAAA,EACpBD,cACA9B,YACAC,aACAC,YACAC,aACAV;iBAEG;gBACL,MAAMuC,UAAUrC,YAAYmC;gBAC5B,IAAIE,SACFR,OAAO,KAAK,CAACK,MAAM,CAAC,EAAE,GAAGG,QAAQ,EAAE;YAEvC;QAEJ;IACF;IAEAT,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,CAACJ,WAAW,KAAK,EAAE,CAAC,wBAAwB,EAAEA,WAAW,KAAK,EAAE;IAEvE,IACEC,AAAmB,MAAnBA,QAAQ,MAAM,IACdC,YAAY,kCAAkC,IAC9C,CAACA,YAAY,KAAK,EAElBY,QAAQ,IAAI,CACV,8EACAlD;IAIJM,QAAAA,uBAAAA,oBAAqB,MAAM,CAAC;QAC1B,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAM2B;YACR;SACD;IACH;IACA3B,QAAAA,uBAAAA,oBAAqB,MAAM,CAAC;QAC1B,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAM;YACR;SACD;IACH;IAEA,OAAOgC;AACT"}
+{"version":3,"file":"ai-model/llm-planning.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/llm-planning.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n  DeviceAction,\n  InterfaceType,\n  PlanningAIResponse,\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  warnGPT4oSizeLimit,\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  },\n): Promise<PlanningAIResponse> {\n  const { context, modelConfig, conversationHistory } = opts;\n  const { screenshotBase64, size } = context;\n\n  const { modelName, vlMode } = modelConfig;\n\n  // Planning requires VL mode (validated by ModelConfigManager.getModelConfig)\n  assert(vlMode, 'Planning requires vlMode to be configured.');\n\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    vlMode: vlMode,\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 === 'qwen-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  } else if (vlMode === 'qwen3-vl') {\n    // Reserved for qwen3-vl specific processing\n    // const paddedResult = await paddingToMatchBlockByBase64(imagePayload, 32);\n    // imageWidth = paddedResult.width;\n    // imageHeight = paddedResult.height;\n    // imagePayload = paddedResult.imageBase64;\n  }\n\n  warnGPT4oSizeLimit(size, modelName);\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, usage } = await callAIWithObjectResponse<PlanningAIResponse>(\n    msgs,\n    AIActionType.PLAN,\n    modelConfig,\n  );\n  const rawResponse = JSON.stringify(content, undefined, 2);\n  const planFromAI = content;\n\n  const actions =\n    (planFromAI.action?.type ? [planFromAI.action] : planFromAI.actions) || [];\n  const returnValue: PlanningAIResponse = {\n    ...planFromAI,\n    actions,\n    rawResponse,\n    usage,\n    yamlFlow: buildYamlFlowFromPlans(\n      actions,\n      opts.actionSpace,\n      planFromAI.sleep,\n    ),\n  };\n\n  assert(planFromAI, \"can't get plans from AI\");\n\n  // TODO: use zod.parse to parse the action.param, and then fill the bbox param.\n  actions.forEach((action) => {\n    const type = action.type;\n    const actionInActionSpace = opts.actionSpace.find(\n      (action) => action.name === type,\n    );\n\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) {\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":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","plan","userInstruction","opts","_opts_conversationHistory","_planFromAI_action","context","modelConfig","conversationHistory","screenshotBase64","size","modelName","vlMode","assert","systemPrompt","systemPromptToTaskPlanning","imagePayload","imageWidth","imageHeight","rightLimit","bottomLimit","paddedResult","paddingToMatchBlockByBase64","warnGPT4oSizeLimit","historyLog","knowledgeContext","instruction","msgs","content","usage","callAIWithObjectResponse","AIActionType","rawResponse","JSON","undefined","planFromAI","actions","returnValue","buildYamlFlowFromPlans","action","type","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam","console"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;ACgBA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAEhB,eAAeC,KACpBC,eAAuB,EACvBC,IAOC;QAqCkBC,2BA0DhBC;IA7FH,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAEC,mBAAmB,EAAE,GAAGL;IACtD,MAAM,EAAEM,gBAAgB,EAAEC,IAAI,EAAE,GAAGJ;IAEnC,MAAM,EAAEK,SAAS,EAAEC,MAAM,EAAE,GAAGL;IAG9BM,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,QAAQ;IAEf,MAAME,eAAe,MAAMC,AAAAA,IAAAA,gCAAAA,0BAAAA,AAAAA,EAA2B;QACpD,aAAaZ,KAAK,WAAW;QAC7B,QAAQS;IACV;IAEA,IAAII,eAAeP;IACnB,IAAIQ,aAAaP,KAAK,KAAK;IAC3B,IAAIQ,cAAcR,KAAK,MAAM;IAC7B,MAAMS,aAAaF;IACnB,MAAMG,cAAcF;IAGpB,IAAIN,AAAW,cAAXA,QAAsB;QACxB,MAAMS,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,2BAAAA,AAAAA,EAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAQAE,IAAAA,mCAAAA,kBAAAA,AAAAA,EAAmBb,MAAMC;IAEzB,MAAMa,aAAapB,AAAAA,SAAAA,CAAAA,4BAAAA,KAAK,mBAAmB,AAAD,IAAvBA,KAAAA,IAAAA,0BAA0B,QAAQ,EAAC,KAAK,EAAE;IAG7D,MAAMqB,mBAAiDtB,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,MAAMuB,cAA4C;QAChD;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,CAAC,kBAAkB,EAAExB,gBAAgB,mBAAmB,CAAC;gBACjE;aACD;QACH;KACD;IAED,MAAMyB,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASb;QAAa;WACrCW;WACAC;WACAF;QACH;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKR;wBACL,QAAQ;oBACV;gBACF;aAED;QACH;KACD;IAED,MAAM,EAAEY,OAAO,EAAEC,KAAK,EAAE,GAAG,MAAMC,AAAAA,IAAAA,yBAAAA,wBAAAA,AAAAA,EAC/BH,MACAI,mCAAAA,YAAAA,CAAAA,IAAiB,EACjBxB;IAEF,MAAMyB,cAAcC,KAAK,SAAS,CAACL,SAASM,QAAW;IACvD,MAAMC,aAAaP;IAEnB,MAAMQ,UACH/B,AAAAA,CAAAA,SAAAA,CAAAA,qBAAAA,WAAW,MAAM,AAAD,IAAhBA,KAAAA,IAAAA,mBAAmB,IAAI,AAAD,IAAI;QAAC8B,WAAW,MAAM;KAAC,GAAGA,WAAW,OAAM,KAAM,EAAE;IAC5E,MAAME,cAAkC;QACtC,GAAGF,UAAU;QACbC;QACAJ;QACAH;QACA,UAAUS,AAAAA,IAAAA,mCAAAA,sBAAAA,AAAAA,EACRF,SACAjC,KAAK,WAAW,EAChBgC,WAAW,KAAK;IAEpB;IAEAtB,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOsB,YAAY;IAGnBC,QAAQ,OAAO,CAAC,CAACG;QACf,MAAMC,OAAOD,OAAO,IAAI;QACxB,MAAME,sBAAsBtC,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACoC,SAAWA,OAAO,IAAI,KAAKC;QAG9BzC,MAAM,+BAA+B0C;QACrC,MAAMC,eAAeD,sBACjBE,AAAAA,IAAAA,mCAAAA,2BAAAA,AAAAA,EAA4BF,oBAAoB,WAAW,IAC3D,EAAE;QAEN1C,MAAM,gBAAgB2C;QAEtBA,aAAa,OAAO,CAAC,CAACE;YACpB,MAAMC,eAAeN,OAAO,KAAK,CAACK,MAAM;YACxC,IAAIC,cAEFN,OAAO,KAAK,CAACK,MAAM,GAAGE,AAAAA,IAAAA,mCAAAA,aAAAA,AAAAA,EACpBD,cACA5B,YACAC,aACAC,YACAC,aACAR;QAGN;IACF;IAEAC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,CAACsB,WAAW,KAAK,EAAE,CAAC,wBAAwB,EAAEA,WAAW,KAAK,EAAE;IAEvE,IACEC,AAAmB,MAAnBA,QAAQ,MAAM,IACdC,YAAY,kCAAkC,IAC9C,CAACA,YAAY,KAAK,EAElBU,QAAQ,IAAI,CACV,8EACA7C;IAIJM,QAAAA,uBAAAA,oBAAqB,MAAM,CAAC;QAC1B,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAMwB;YACR;SACD;IACH;IACAxB,QAAAA,uBAAAA,oBAAqB,MAAM,CAAC;QAC1B,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAM;YACR;SACD;IACH;IAEA,OAAO6B;AACT"}