@midscene/core 1.0.1-beta-20251205093836.0 → 1.0.1-beta-20251208031823.0

package/dist/es/agent/task-builder.mjs

@@ -16,6 +16,9 @@ function _define_property(obj, key, value) {
     return obj;
 }
 const debug = getDebug('agent:task-builder');
+function hasNonEmptyCache(cache) {
+    return null != cache && 'object' == typeof cache && Object.keys(cache).length > 0;
+}
 function locatePlanForLocate(param) {
     const locate = 'string' == typeof param ? {
         prompt: param
@@ -103,19 +106,14 @@ class TaskBuilder {
         const locateFields = action ? findAllMidsceneLocatorField(action.paramSchema) : [];
         const requiredLocateFields = action ? findAllMidsceneLocatorField(action.paramSchema, true) : [];
         locateFields.forEach((field)=>{
-            if (param[field]) if (ifPlanLocateParamIsBbox(param[field])) {
-                debug('plan locate param is bbox, will match element from plan', param[field]);
-                const elementFromPlan = matchElementFromPlan(param[field]);
-                if (elementFromPlan) param[field] = elementFromPlan;
-            } else {
+            if (param[field]) {
                 const locatePlan = locatePlanForLocate(param[field]);
-                debug('will prepend locate param for field', `action.type=${planType}`, `param=${JSON.stringify(param[field])}`, `locatePlan=${JSON.stringify(locatePlan)}`);
+                debug('will prepend locate param for field', `action.type=${planType}`, `param=${JSON.stringify(param[field])}`, `locatePlan=${JSON.stringify(locatePlan)}`, `hasBbox=${ifPlanLocateParamIsBbox(param[field])}`);
                 const locateTask = this.createLocateTask(locatePlan, param[field], context, (result)=>{
                     param[field] = result;
                 });
                 context.tasks.push(locateTask);
-            }
-            else {
+            } else {
                 assert(!requiredLocateFields.includes(field), `Required locate field '${field}' is not provided for action ${planType}`);
                 debug(`field '${field}' is not provided for action ${planType}`);
             }
@@ -185,7 +183,6 @@ class TaskBuilder {
     createLocateTask(plan, detailedLocateParam, context, onResult) {
         const { cacheable, modelConfigForDefaultIntent } = context;
         let locateParam = detailedLocateParam;
-        console.log('modelConfigForDefaultIntent-locateParam', modelConfigForDefaultIntent, locateParam);
         if ('string' == typeof locateParam) locateParam = {
             prompt: locateParam
         };
@@ -216,32 +213,31 @@ class TaskBuilder {
                     task.usage = dump.taskInfo?.usage;
                     if (dump.taskInfo?.searchAreaUsage) task.searchAreaUsage = dump.taskInfo.searchAreaUsage;
                 };
+                const elementFromBbox = ifPlanLocateParamIsBbox(param) ? matchElementFromPlan(param) : void 0;
+                const isPlanHit = !!elementFromBbox;
                 let rectFromXpath;
-                if (param.xpath && this.interface.rectMatchesCacheFeature) try {
+                if (!isPlanHit && param.xpath && this.interface.rectMatchesCacheFeature) try {
                     rectFromXpath = await this.interface.rectMatchesCacheFeature({
                         xpaths: [
                             param.xpath
                         ]
                     });
-                } catch (error) {
-                    rectFromXpath = void 0;
-                }
+                } catch  {}
                 const elementFromXpath = rectFromXpath ? generateElementByPosition({
                     x: rectFromXpath.left + rectFromXpath.width / 2,
                     y: rectFromXpath.top + rectFromXpath.height / 2
                 }, 'string' == typeof param.prompt ? param.prompt : param.prompt?.prompt || '') : void 0;
-                const userExpectedPathHitFlag = !!elementFromXpath;
+                const isXpathHit = !!elementFromXpath;
                 const cachePrompt = param.prompt;
                 const locateCacheRecord = this.taskCache?.matchLocateCache(cachePrompt);
                 const cacheEntry = locateCacheRecord?.cacheContent?.cache;
-                const elementFromCache = userExpectedPathHitFlag ? null : await matchElementFromCache({
+                const elementFromCache = isPlanHit || isXpathHit ? null : await matchElementFromCache({
                     taskCache: this.taskCache,
                     interfaceInstance: this.interface
                 }, cacheEntry, cachePrompt, param.cacheable);
-                const cacheHitFlag = !!elementFromCache;
-                const planHitFlag = false;
+                const isCacheHit = !!elementFromCache;
                 let elementFromAiLocate;
-                if (!userExpectedPathHitFlag && !cacheHitFlag && !planHitFlag) try {
+                if (!isXpathHit && !isCacheHit && !isPlanHit) try {
                     locateResult = await this.service.locate(param, {
                         context: uiContext
                     }, modelConfigForDefaultIntent);
@@ -251,14 +247,15 @@ class TaskBuilder {
                     if (error instanceof ServiceError) applyDump(error.dump);
                     throw error;
                 }
-                const element = elementFromXpath || elementFromCache || elementFromAiLocate;
+                const element = elementFromBbox || elementFromXpath || elementFromCache || elementFromAiLocate;
+                const locateCacheAlreadyExists = hasNonEmptyCache(locateCacheRecord?.cacheContent?.cache);
                 let currentCacheEntry;
-                if (element && this.taskCache && !cacheHitFlag && param?.cacheable !== false) if (this.interface.cacheFeatureForRect) try {
+                if (element && this.taskCache && !isCacheHit && !locateCacheAlreadyExists && param?.cacheable !== false) if (this.interface.cacheFeatureForRect) try {
                     const feature = await this.interface.cacheFeatureForRect(element.rect, {
                         targetDescription: 'string' == typeof param.prompt ? param.prompt : param.prompt?.prompt,
                         modelConfig: modelConfigForDefaultIntent
                     });
-                    if (feature && Object.keys(feature).length > 0) {
+                    if (hasNonEmptyCache(feature)) {
                         debug('update cache, prompt: %s, cache: %o', cachePrompt, feature);
                         currentCacheEntry = feature;
                         this.taskCache.updateOrAppendCacheRecord({
@@ -276,13 +273,19 @@ class TaskBuilder {
                     throw new Error(`Element not found: ${param.prompt}`);
                 }
                 let hitBy;
-                if (userExpectedPathHitFlag) hitBy = {
+                if (isPlanHit) hitBy = {
+                    from: 'Plan',
+                    context: {
+                        bbox: param.bbox
+                    }
+                };
+                else if (isXpathHit) hitBy = {
                     from: 'User expected path',
                     context: {
                         xpath: param.xpath
                     }
                 };
-                else if (cacheHitFlag) hitBy = {
+                else if (isCacheHit) hitBy = {
                     from: 'Cache',
                     context: {
                         cacheEntry,

package/dist/es/agent/task-builder.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"agent/task-builder.mjs","sources":["../../../src/agent/task-builder.ts"],"sourcesContent":["import { findAllMidsceneLocatorField, parseActionParam } from '@/ai-model';\nimport type { AbstractInterface } from '@/device';\nimport type Service from '@/service';\nimport type {\n  DetailedLocateParam,\n  ElementCacheFeature,\n  ExecutionTaskActionApply,\n  ExecutionTaskApply,\n  ExecutionTaskHitBy,\n  ExecutionTaskPlanningLocateApply,\n  LocateResultElement,\n  LocateResultWithDump,\n  PlanningAction,\n  PlanningActionParamSleep,\n  PlanningLocateParam,\n  Rect,\n  ServiceDump,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport { sleep } from '@/utils';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { generateElementByPosition } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TaskCache } from './task-cache';\nimport {\n  ifPlanLocateParamIsBbox,\n  matchElementFromCache,\n  matchElementFromPlan,\n} from './utils';\n\nconst debug = getDebug('agent:task-builder');\n\nexport function locatePlanForLocate(param: string | DetailedLocateParam) {\n  const locate = typeof param === 'string' ? { prompt: param } : param;\n  const locatePlan: PlanningAction<PlanningLocateParam> = {\n    type: 'Locate',\n    param: locate,\n    thought: '',\n  };\n  return locatePlan;\n}\n\ninterface TaskBuilderDeps {\n  interfaceInstance: AbstractInterface;\n  service: Service;\n  taskCache?: TaskCache;\n}\n\ninterface BuildOptions {\n  cacheable?: boolean;\n  subTask?: boolean;\n}\n\ninterface PlanBuildContext {\n  tasks: ExecutionTaskApply[];\n  modelConfigForPlanning: IModelConfig;\n  modelConfigForDefaultIntent: IModelConfig;\n  cacheable?: boolean;\n  subTask: boolean;\n}\n\nexport class TaskBuilder {\n  private readonly interface: AbstractInterface;\n\n  private readonly service: Service;\n\n  private readonly taskCache?: TaskCache;\n\n  constructor({ interfaceInstance, service, taskCache }: TaskBuilderDeps) {\n    this.interface = interfaceInstance;\n    this.service = service;\n    this.taskCache = taskCache;\n  }\n\n  public async build(\n    plans: PlanningAction[],\n    modelConfigForPlanning: IModelConfig,\n    modelConfigForDefaultIntent: IModelConfig,\n    options?: BuildOptions,\n  ): Promise<{ tasks: ExecutionTaskApply[] }> {\n    const tasks: ExecutionTaskApply[] = [];\n    const cacheable = options?.cacheable;\n\n    const context: PlanBuildContext = {\n      tasks,\n      modelConfigForPlanning,\n      modelConfigForDefaultIntent,\n      cacheable,\n      subTask: !!options?.subTask,\n    };\n\n    type PlanHandler = (plan: PlanningAction) => Promise<void> | void;\n\n    const planHandlers = new Map<string, PlanHandler>([\n      [\n        'Locate',\n        (plan) =>\n          this.handleLocatePlan(\n            plan as PlanningAction<PlanningLocateParam>,\n            context,\n          ),\n      ],\n      ['Finished', (plan) => this.handleFinishedPlan(plan, context)],\n      [\n        'Sleep',\n        (plan) =>\n          this.handleSleepPlan(\n            plan as PlanningAction<PlanningActionParamSleep>,\n            context,\n          ),\n      ],\n    ]);\n\n    const defaultHandler: PlanHandler = (plan) =>\n      this.handleActionPlan(plan, context);\n\n    for (const plan of plans) {\n      const handler = planHandlers.get(plan.type) ?? defaultHandler;\n      await handler(plan);\n    }\n\n    return {\n      tasks,\n    };\n  }\n\n  private handleFinishedPlan(\n    plan: PlanningAction,\n    context: PlanBuildContext,\n  ): void {\n    const taskActionFinished: ExecutionTaskActionApply<null> = {\n      type: 'Action Space',\n      subType: 'Finished',\n      param: null,\n      thought: plan.thought,\n      subTask: context.subTask || undefined,\n      executor: async () => {},\n    };\n    context.tasks.push(taskActionFinished);\n  }\n\n  private handleSleepPlan(\n    plan: PlanningAction<PlanningActionParamSleep>,\n    context: PlanBuildContext,\n  ): void {\n    const sleepTask = this.createSleepTask(plan.param, {\n      thought: plan.thought,\n    });\n    if (context.subTask) {\n      sleepTask.subTask = true;\n    }\n    context.tasks.push(sleepTask);\n  }\n\n  public createSleepTask(\n    param: PlanningActionParamSleep,\n    meta?: { thought?: string },\n  ): ExecutionTaskActionApply<PlanningActionParamSleep> {\n    return {\n      type: 'Action Space',\n      subType: 'Sleep',\n      param,\n      thought: meta?.thought,\n      executor: async (taskParam) => {\n        await sleep(taskParam?.timeMs || 3000);\n      },\n    };\n  }\n\n  private async handleLocatePlan(\n    plan: PlanningAction<PlanningLocateParam>,\n    context: PlanBuildContext,\n  ): Promise<void> {\n    const taskLocate = this.createLocateTask(plan, plan.param, context);\n    context.tasks.push(taskLocate);\n  }\n\n  private async handleActionPlan(\n    plan: PlanningAction,\n    context: PlanBuildContext,\n  ): Promise<void> {\n    const planType = plan.type;\n    const actionSpace = await this.interface.actionSpace();\n    const action = actionSpace.find((item) => item.name === planType);\n    const param = plan.param;\n\n    if (!action) {\n      throw new Error(`Action type '${planType}' not found`);\n    }\n\n    const locateFields = action\n      ? findAllMidsceneLocatorField(action.paramSchema)\n      : [];\n\n    const requiredLocateFields = action\n      ? findAllMidsceneLocatorField(action.paramSchema, true)\n      : [];\n\n    locateFields.forEach((field) => {\n      if (param[field]) {\n        if (ifPlanLocateParamIsBbox(param[field])) {\n          debug(\n            'plan locate param is bbox, will match element from plan',\n            param[field],\n          );\n          const elementFromPlan = matchElementFromPlan(param[field]);\n          if (elementFromPlan) {\n            param[field] = elementFromPlan;\n          }\n        } else {\n          const locatePlan = locatePlanForLocate(param[field]);\n          debug(\n            'will prepend locate param for field',\n            `action.type=${planType}`,\n            `param=${JSON.stringify(param[field])}`,\n            `locatePlan=${JSON.stringify(locatePlan)}`,\n          );\n          const locateTask = this.createLocateTask(\n            locatePlan,\n            param[field],\n            context,\n            (result) => {\n              param[field] = result;\n            },\n          );\n          context.tasks.push(locateTask);\n        }\n      } else {\n        assert(\n          !requiredLocateFields.includes(field),\n          `Required locate field '${field}' is not provided for action ${planType}`,\n        );\n        debug(`field '${field}' is not provided for action ${planType}`);\n      }\n    });\n\n    const task: ExecutionTaskApply<\n      'Action Space',\n      any,\n      { success: boolean; action: string; param: any },\n      void\n    > = {\n      type: 'Action Space',\n      subType: planType,\n      thought: plan.thought,\n      param: plan.param,\n      subTask: context.subTask || undefined,\n      executor: async (param, taskContext) => {\n        debug(\n          'executing action',\n          planType,\n          param,\n          `taskContext.element.center: ${taskContext.element?.center}`,\n        );\n\n        const uiContext = taskContext.uiContext;\n        assert(uiContext, 'uiContext is required for Action task');\n\n        requiredLocateFields.forEach((field) => {\n          assert(\n            param[field],\n            `field '${field}' is required for action ${planType} but not provided. Cannot execute action ${planType}.`,\n          );\n        });\n\n        try {\n          await Promise.all([\n            (async () => {\n              if (this.interface.beforeInvokeAction) {\n                debug('will call \"beforeInvokeAction\" for interface');\n                await this.interface.beforeInvokeAction(action.name, param);\n                debug('called \"beforeInvokeAction\" for interface');\n              }\n            })(),\n            sleep(200),\n          ]);\n        } catch (originalError: any) {\n          const originalMessage =\n            originalError?.message || String(originalError);\n          throw new Error(\n            `error in running beforeInvokeAction for ${action.name}: ${originalMessage}`,\n            { cause: originalError },\n          );\n        }\n\n        if (action.paramSchema) {\n          try {\n            param = parseActionParam(param, action.paramSchema);\n          } catch (error: any) {\n            throw new Error(\n              `Invalid parameters for action ${action.name}: ${error.message}\\nParameters: ${JSON.stringify(param)}`,\n              { cause: error },\n            );\n          }\n        }\n\n        debug('calling action', action.name);\n        const actionFn = action.call.bind(this.interface);\n        const actionResult = await actionFn(param, taskContext);\n        debug('called action', action.name, 'result:', actionResult);\n\n        const delayAfterRunner = action.delayAfterRunner ?? 300;\n        if (delayAfterRunner > 0) {\n          await sleep(delayAfterRunner);\n        }\n\n        try {\n          if (this.interface.afterInvokeAction) {\n            debug('will call \"afterInvokeAction\" for interface');\n            await this.interface.afterInvokeAction(action.name, param);\n            debug('called \"afterInvokeAction\" for interface');\n          }\n        } catch (originalError: any) {\n          const originalMessage =\n            originalError?.message || String(originalError);\n          throw new Error(\n            `error in running afterInvokeAction for ${action.name}: ${originalMessage}`,\n            { cause: originalError },\n          );\n        }\n\n        return {\n          output: actionResult,\n        };\n      },\n    };\n\n    context.tasks.push(task);\n  }\n\n  private createLocateTask(\n    plan: PlanningAction<PlanningLocateParam>,\n    detailedLocateParam: DetailedLocateParam | string,\n    context: PlanBuildContext,\n    onResult?: (result: LocateResultElement) => void,\n  ): ExecutionTaskPlanningLocateApply {\n    const { cacheable, modelConfigForDefaultIntent } = context;\n\n    let locateParam = detailedLocateParam;\n    console.log(\n      'modelConfigForDefaultIntent-locateParam',\n      modelConfigForDefaultIntent,\n      locateParam,\n    );\n\n    if (typeof locateParam === 'string') {\n      locateParam = {\n        prompt: locateParam,\n      };\n    }\n\n    if (cacheable !== undefined) {\n      locateParam = {\n        ...locateParam,\n        cacheable,\n      };\n    }\n\n    const taskLocator: ExecutionTaskPlanningLocateApply = {\n      type: 'Planning',\n      subType: 'Locate',\n      subTask: context.subTask || undefined,\n      param: locateParam,\n      thought: plan.thought,\n      executor: async (param, taskContext) => {\n        const { task } = taskContext;\n        let { uiContext } = taskContext;\n\n        assert(\n          param?.prompt || param?.bbox,\n          `No prompt or id or position or bbox to locate, param=${JSON.stringify(\n            param,\n          )}`,\n        );\n\n        if (!uiContext) {\n          uiContext = await this.service.contextRetrieverFn();\n        }\n\n        assert(uiContext, 'uiContext is required for Service task');\n\n        let locateDump: ServiceDump | undefined;\n        let locateResult: LocateResultWithDump | undefined;\n\n        const applyDump = (dump?: ServiceDump) => {\n          if (!dump) {\n            return;\n          }\n          locateDump = dump;\n          task.log = {\n            dump,\n          };\n          task.usage = dump.taskInfo?.usage;\n          if (dump.taskInfo?.searchAreaUsage) {\n            task.searchAreaUsage = dump.taskInfo.searchAreaUsage;\n          }\n        };\n\n        // from xpath\n        let rectFromXpath: Rect | undefined;\n        if (param.xpath && this.interface.rectMatchesCacheFeature) {\n          try {\n            rectFromXpath = await this.interface.rectMatchesCacheFeature({\n              xpaths: [param.xpath],\n            });\n          } catch (error) {\n            // xpath locate failed, allow fallback to cache or AI locate\n            rectFromXpath = undefined;\n          }\n        }\n        const elementFromXpath = rectFromXpath\n          ? generateElementByPosition(\n              {\n                x: rectFromXpath.left + rectFromXpath.width / 2,\n                y: rectFromXpath.top + rectFromXpath.height / 2,\n              },\n              typeof param.prompt === 'string'\n                ? param.prompt\n                : param.prompt?.prompt || '',\n            )\n          : undefined;\n        const userExpectedPathHitFlag = !!elementFromXpath;\n\n        const cachePrompt = param.prompt;\n        const locateCacheRecord = this.taskCache?.matchLocateCache(cachePrompt);\n        const cacheEntry = locateCacheRecord?.cacheContent?.cache;\n\n        const elementFromCache = userExpectedPathHitFlag\n          ? null\n          : await matchElementFromCache(\n              {\n                taskCache: this.taskCache,\n                interfaceInstance: this.interface,\n              },\n              cacheEntry,\n              cachePrompt,\n              param.cacheable,\n            );\n        const cacheHitFlag = !!elementFromCache;\n\n        const planHitFlag = false;\n\n        let elementFromAiLocate: LocateResultElement | null | undefined;\n        if (!userExpectedPathHitFlag && !cacheHitFlag && !planHitFlag) {\n          try {\n            locateResult = await this.service.locate(\n              param,\n              {\n                context: uiContext,\n              },\n              modelConfigForDefaultIntent,\n            );\n            applyDump(locateResult.dump);\n            elementFromAiLocate = locateResult.element;\n          } catch (error) {\n            if (error instanceof ServiceError) {\n              applyDump(error.dump);\n            }\n            throw error;\n          }\n        }\n\n        const element =\n          elementFromXpath || elementFromCache || elementFromAiLocate;\n\n        let currentCacheEntry: ElementCacheFeature | undefined;\n        if (\n          element &&\n          this.taskCache &&\n          !cacheHitFlag &&\n          param?.cacheable !== false\n        ) {\n          if (this.interface.cacheFeatureForRect) {\n            try {\n              const feature = await this.interface.cacheFeatureForRect(\n                element.rect,\n                {\n                  targetDescription:\n                    typeof param.prompt === 'string'\n                      ? param.prompt\n                      : param.prompt?.prompt,\n                  modelConfig: modelConfigForDefaultIntent,\n                },\n              );\n              if (feature && Object.keys(feature).length > 0) {\n                debug(\n                  'update cache, prompt: %s, cache: %o',\n                  cachePrompt,\n                  feature,\n                );\n                currentCacheEntry = feature;\n                this.taskCache.updateOrAppendCacheRecord(\n                  {\n                    type: 'locate',\n                    prompt: cachePrompt,\n                    cache: feature,\n                  },\n                  locateCacheRecord,\n                );\n              } else {\n                debug(\n                  'no cache data returned, skip cache update, prompt: %s',\n                  cachePrompt,\n                );\n              }\n            } catch (error) {\n              debug('cacheFeatureForRect failed: %s', error);\n            }\n          } else {\n            debug('cacheFeatureForRect is not supported, skip cache update');\n          }\n        }\n\n        if (!element) {\n          if (locateDump) {\n            throw new ServiceError(\n              `Element not found : ${param.prompt}`,\n              locateDump,\n            );\n          }\n          throw new Error(`Element not found: ${param.prompt}`);\n        }\n\n        let hitBy: ExecutionTaskHitBy | undefined;\n\n        if (userExpectedPathHitFlag) {\n          hitBy = {\n            from: 'User expected path',\n            context: {\n              xpath: param.xpath,\n            },\n          };\n        } else if (cacheHitFlag) {\n          hitBy = {\n            from: 'Cache',\n            context: {\n              cacheEntry,\n              cacheToSave: currentCacheEntry,\n            },\n          };\n        }\n\n        onResult?.(element);\n\n        return {\n          output: {\n            element,\n          },\n          hitBy,\n        };\n      },\n    };\n\n    return taskLocator;\n  }\n}\n"],"names":["debug","getDebug","locatePlanForLocate","param","locate","locatePlan","TaskBuilder","plans","modelConfigForPlanning","modelConfigForDefaultIntent","options","tasks","cacheable","context","planHandlers","Map","plan","defaultHandler","handler","taskActionFinished","undefined","sleepTask","meta","taskParam","sleep","taskLocate","planType","actionSpace","action","item","Error","locateFields","findAllMidsceneLocatorField","requiredLocateFields","field","ifPlanLocateParamIsBbox","elementFromPlan","matchElementFromPlan","JSON","locateTask","result","assert","task","taskContext","uiContext","Promise","originalError","originalMessage","String","parseActionParam","error","actionFn","actionResult","delayAfterRunner","detailedLocateParam","onResult","locateParam","console","taskLocator","locateDump","locateResult","applyDump","dump","rectFromXpath","elementFromXpath","generateElementByPosition","userExpectedPathHitFlag","cachePrompt","locateCacheRecord","cacheEntry","elementFromCache","matchElementFromCache","cacheHitFlag","planHitFlag","elementFromAiLocate","ServiceError","element","currentCacheEntry","feature","Object","hitBy","interfaceInstance","service","taskCache"],"mappings":";;;;;;;;;;;;;;;;;AA+BA,MAAMA,QAAQC,SAAS;AAEhB,SAASC,oBAAoBC,KAAmC;IACrE,MAAMC,SAAS,AAAiB,YAAjB,OAAOD,QAAqB;QAAE,QAAQA;IAAM,IAAIA;IAC/D,MAAME,aAAkD;QACtD,MAAM;QACN,OAAOD;QACP,SAAS;IACX;IACA,OAAOC;AACT;AAqBO,MAAMC;IAaX,MAAa,MACXC,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACzCC,OAAsB,EACoB;QAC1C,MAAMC,QAA8B,EAAE;QACtC,MAAMC,YAAYF,SAAS;QAE3B,MAAMG,UAA4B;YAChCF;YACAH;YACAC;YACAG;YACA,SAAS,CAAC,CAACF,SAAS;QACtB;QAIA,MAAMI,eAAe,IAAIC,IAAyB;YAChD;gBACE;gBACA,CAACC,OACC,IAAI,CAAC,gBAAgB,CACnBA,MACAH;aAEL;YACD;gBAAC;gBAAY,CAACG,OAAS,IAAI,CAAC,kBAAkB,CAACA,MAAMH;aAAS;YAC9D;gBACE;gBACA,CAACG,OACC,IAAI,CAAC,eAAe,CAClBA,MACAH;aAEL;SACF;QAED,MAAMI,iBAA8B,CAACD,OACnC,IAAI,CAAC,gBAAgB,CAACA,MAAMH;QAE9B,KAAK,MAAMG,QAAQT,MAAO;YACxB,MAAMW,UAAUJ,aAAa,GAAG,CAACE,KAAK,IAAI,KAAKC;YAC/C,MAAMC,QAAQF;QAChB;QAEA,OAAO;YACLL;QACF;IACF;IAEQ,mBACNK,IAAoB,EACpBH,OAAyB,EACnB;QACN,MAAMM,qBAAqD;YACzD,MAAM;YACN,SAAS;YACT,OAAO;YACP,SAASH,KAAK,OAAO;YACrB,SAASH,QAAQ,OAAO,IAAIO;YAC5B,UAAU,WAAa;QACzB;QACAP,QAAQ,KAAK,CAAC,IAAI,CAACM;IACrB;IAEQ,gBACNH,IAA8C,EAC9CH,OAAyB,EACnB;QACN,MAAMQ,YAAY,IAAI,CAAC,eAAe,CAACL,KAAK,KAAK,EAAE;YACjD,SAASA,KAAK,OAAO;QACvB;QACA,IAAIH,QAAQ,OAAO,EACjBQ,UAAU,OAAO,GAAG;QAEtBR,QAAQ,KAAK,CAAC,IAAI,CAACQ;IACrB;IAEO,gBACLlB,KAA+B,EAC/BmB,IAA2B,EACyB;QACpD,OAAO;YACL,MAAM;YACN,SAAS;YACTnB;YACA,SAASmB,MAAM;YACf,UAAU,OAAOC;gBACf,MAAMC,MAAMD,WAAW,UAAU;YACnC;QACF;IACF;IAEA,MAAc,iBACZP,IAAyC,EACzCH,OAAyB,EACV;QACf,MAAMY,aAAa,IAAI,CAAC,gBAAgB,CAACT,MAAMA,KAAK,KAAK,EAAEH;QAC3DA,QAAQ,KAAK,CAAC,IAAI,CAACY;IACrB;IAEA,MAAc,iBACZT,IAAoB,EACpBH,OAAyB,EACV;QACf,MAAMa,WAAWV,KAAK,IAAI;QAC1B,MAAMW,cAAc,MAAM,IAAI,CAAC,SAAS,CAAC,WAAW;QACpD,MAAMC,SAASD,YAAY,IAAI,CAAC,CAACE,OAASA,KAAK,IAAI,KAAKH;QACxD,MAAMvB,QAAQa,KAAK,KAAK;QAExB,IAAI,CAACY,QACH,MAAM,IAAIE,MAAM,CAAC,aAAa,EAAEJ,SAAS,WAAW,CAAC;QAGvD,MAAMK,eAAeH,SACjBI,4BAA4BJ,OAAO,WAAW,IAC9C,EAAE;QAEN,MAAMK,uBAAuBL,SACzBI,4BAA4BJ,OAAO,WAAW,EAAE,QAChD,EAAE;QAENG,aAAa,OAAO,CAAC,CAACG;YACpB,IAAI/B,KAAK,CAAC+B,MAAM,EACd,IAAIC,wBAAwBhC,KAAK,CAAC+B,MAAM,GAAG;gBACzClC,MACE,2DACAG,KAAK,CAAC+B,MAAM;gBAEd,MAAME,kBAAkBC,qBAAqBlC,KAAK,CAAC+B,MAAM;gBACzD,IAAIE,iBACFjC,KAAK,CAAC+B,MAAM,GAAGE;YAEnB,OAAO;gBACL,MAAM/B,aAAaH,oBAAoBC,KAAK,CAAC+B,MAAM;gBACnDlC,MACE,uCACA,CAAC,YAAY,EAAE0B,UAAU,EACzB,CAAC,MAAM,EAAEY,KAAK,SAAS,CAACnC,KAAK,CAAC+B,MAAM,GAAG,EACvC,CAAC,WAAW,EAAEI,KAAK,SAAS,CAACjC,aAAa;gBAE5C,MAAMkC,aAAa,IAAI,CAAC,gBAAgB,CACtClC,YACAF,KAAK,CAAC+B,MAAM,EACZrB,SACA,CAAC2B;oBACCrC,KAAK,CAAC+B,MAAM,GAAGM;gBACjB;gBAEF3B,QAAQ,KAAK,CAAC,IAAI,CAAC0B;YACrB;iBACK;gBACLE,OACE,CAACR,qBAAqB,QAAQ,CAACC,QAC/B,CAAC,uBAAuB,EAAEA,MAAM,6BAA6B,EAAER,UAAU;gBAE3E1B,MAAM,CAAC,OAAO,EAAEkC,MAAM,6BAA6B,EAAER,UAAU;YACjE;QACF;QAEA,MAAMgB,OAKF;YACF,MAAM;YACN,SAAShB;YACT,SAASV,KAAK,OAAO;YACrB,OAAOA,KAAK,KAAK;YACjB,SAASH,QAAQ,OAAO,IAAIO;YAC5B,UAAU,OAAOjB,OAAOwC;gBACtB3C,MACE,oBACA0B,UACAvB,OACA,CAAC,4BAA4B,EAAEwC,YAAY,OAAO,EAAE,QAAQ;gBAG9D,MAAMC,YAAYD,YAAY,SAAS;gBACvCF,OAAOG,WAAW;gBAElBX,qBAAqB,OAAO,CAAC,CAACC;oBAC5BO,OACEtC,KAAK,CAAC+B,MAAM,EACZ,CAAC,OAAO,EAAEA,MAAM,yBAAyB,EAAER,SAAS,yCAAyC,EAAEA,SAAS,CAAC,CAAC;gBAE9G;gBAEA,IAAI;oBACF,MAAMmB,QAAQ,GAAG,CAAC;wBACf;4BACC,IAAI,IAAI,CAAC,SAAS,CAAC,kBAAkB,EAAE;gCACrC7C,MAAM;gCACN,MAAM,IAAI,CAAC,SAAS,CAAC,kBAAkB,CAAC4B,OAAO,IAAI,EAAEzB;gCACrDH,MAAM;4BACR;wBACF;wBACAwB,MAAM;qBACP;gBACH,EAAE,OAAOsB,eAAoB;oBAC3B,MAAMC,kBACJD,eAAe,WAAWE,OAAOF;oBACnC,MAAM,IAAIhB,MACR,CAAC,wCAAwC,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEmB,iBAAiB,EAC5E;wBAAE,OAAOD;oBAAc;gBAE3B;gBAEA,IAAIlB,OAAO,WAAW,EACpB,IAAI;oBACFzB,QAAQ8C,iBAAiB9C,OAAOyB,OAAO,WAAW;gBACpD,EAAE,OAAOsB,OAAY;oBACnB,MAAM,IAAIpB,MACR,CAAC,8BAA8B,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEsB,MAAM,OAAO,CAAC,cAAc,EAAEZ,KAAK,SAAS,CAACnC,QAAQ,EACtG;wBAAE,OAAO+C;oBAAM;gBAEnB;gBAGFlD,MAAM,kBAAkB4B,OAAO,IAAI;gBACnC,MAAMuB,WAAWvB,OAAO,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS;gBAChD,MAAMwB,eAAe,MAAMD,SAAShD,OAAOwC;gBAC3C3C,MAAM,iBAAiB4B,OAAO,IAAI,EAAE,WAAWwB;gBAE/C,MAAMC,mBAAmBzB,OAAO,gBAAgB,IAAI;gBACpD,IAAIyB,mBAAmB,GACrB,MAAM7B,MAAM6B;gBAGd,IAAI;oBACF,IAAI,IAAI,CAAC,SAAS,CAAC,iBAAiB,EAAE;wBACpCrD,MAAM;wBACN,MAAM,IAAI,CAAC,SAAS,CAAC,iBAAiB,CAAC4B,OAAO,IAAI,EAAEzB;wBACpDH,MAAM;oBACR;gBACF,EAAE,OAAO8C,eAAoB;oBAC3B,MAAMC,kBACJD,eAAe,WAAWE,OAAOF;oBACnC,MAAM,IAAIhB,MACR,CAAC,uCAAuC,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEmB,iBAAiB,EAC3E;wBAAE,OAAOD;oBAAc;gBAE3B;gBAEA,OAAO;oBACL,QAAQM;gBACV;YACF;QACF;QAEAvC,QAAQ,KAAK,CAAC,IAAI,CAAC6B;IACrB;IAEQ,iBACN1B,IAAyC,EACzCsC,mBAAiD,EACjDzC,OAAyB,EACzB0C,QAAgD,EACd;QAClC,MAAM,EAAE3C,SAAS,EAAEH,2BAA2B,EAAE,GAAGI;QAEnD,IAAI2C,cAAcF;QAClBG,QAAQ,GAAG,CACT,2CACAhD,6BACA+C;QAGF,IAAI,AAAuB,YAAvB,OAAOA,aACTA,cAAc;YACZ,QAAQA;QACV;QAGF,IAAI5C,AAAcQ,WAAdR,WACF4C,cAAc;YACZ,GAAGA,WAAW;YACd5C;QACF;QAGF,MAAM8C,cAAgD;YACpD,MAAM;YACN,SAAS;YACT,SAAS7C,QAAQ,OAAO,IAAIO;YAC5B,OAAOoC;YACP,SAASxC,KAAK,OAAO;YACrB,UAAU,OAAOb,OAAOwC;gBACtB,MAAM,EAAED,IAAI,EAAE,GAAGC;gBACjB,IAAI,EAAEC,SAAS,EAAE,GAAGD;gBAEpBF,OACEtC,OAAO,UAAUA,OAAO,MACxB,CAAC,qDAAqD,EAAEmC,KAAK,SAAS,CACpEnC,QACC;gBAGL,IAAI,CAACyC,WACHA,YAAY,MAAM,IAAI,CAAC,OAAO,CAAC,kBAAkB;gBAGnDH,OAAOG,WAAW;gBAElB,IAAIe;gBACJ,IAAIC;gBAEJ,MAAMC,YAAY,CAACC;oBACjB,IAAI,CAACA,MACH;oBAEFH,aAAaG;oBACbpB,KAAK,GAAG,GAAG;wBACToB;oBACF;oBACApB,KAAK,KAAK,GAAGoB,KAAK,QAAQ,EAAE;oBAC5B,IAAIA,KAAK,QAAQ,EAAE,iBACjBpB,KAAK,eAAe,GAAGoB,KAAK,QAAQ,CAAC,eAAe;gBAExD;gBAGA,IAAIC;gBACJ,IAAI5D,MAAM,KAAK,IAAI,IAAI,CAAC,SAAS,CAAC,uBAAuB,EACvD,IAAI;oBACF4D,gBAAgB,MAAM,IAAI,CAAC,SAAS,CAAC,uBAAuB,CAAC;wBAC3D,QAAQ;4BAAC5D,MAAM,KAAK;yBAAC;oBACvB;gBACF,EAAE,OAAO+C,OAAO;oBAEda,gBAAgB3C;gBAClB;gBAEF,MAAM4C,mBAAmBD,gBACrBE,0BACE;oBACE,GAAGF,cAAc,IAAI,GAAGA,cAAc,KAAK,GAAG;oBAC9C,GAAGA,cAAc,GAAG,GAAGA,cAAc,MAAM,GAAG;gBAChD,GACA,AAAwB,YAAxB,OAAO5D,MAAM,MAAM,GACfA,MAAM,MAAM,GACZA,MAAM,MAAM,EAAE,UAAU,MAE9BiB;gBACJ,MAAM8C,0BAA0B,CAAC,CAACF;gBAElC,MAAMG,cAAchE,MAAM,MAAM;gBAChC,MAAMiE,oBAAoB,IAAI,CAAC,SAAS,EAAE,iBAAiBD;gBAC3D,MAAME,aAAaD,mBAAmB,cAAc;gBAEpD,MAAME,mBAAmBJ,0BACrB,OACA,MAAMK,sBACJ;oBACE,WAAW,IAAI,CAAC,SAAS;oBACzB,mBAAmB,IAAI,CAAC,SAAS;gBACnC,GACAF,YACAF,aACAhE,MAAM,SAAS;gBAErB,MAAMqE,eAAe,CAAC,CAACF;gBAEvB,MAAMG,cAAc;gBAEpB,IAAIC;gBACJ,IAAI,CAACR,2BAA2B,CAACM,gBAAgB,CAACC,aAChD,IAAI;oBACFb,eAAe,MAAM,IAAI,CAAC,OAAO,CAAC,MAAM,CACtCzD,OACA;wBACE,SAASyC;oBACX,GACAnC;oBAEFoD,UAAUD,aAAa,IAAI;oBAC3Bc,sBAAsBd,aAAa,OAAO;gBAC5C,EAAE,OAAOV,OAAO;oBACd,IAAIA,iBAAiByB,cACnBd,UAAUX,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAGF,MAAM0B,UACJZ,oBAAoBM,oBAAoBI;gBAE1C,IAAIG;gBACJ,IACED,WACA,IAAI,CAAC,SAAS,IACd,CAACJ,gBACDrE,OAAO,cAAc,OAErB,IAAI,IAAI,CAAC,SAAS,CAAC,mBAAmB,EACpC,IAAI;oBACF,MAAM2E,UAAU,MAAM,IAAI,CAAC,SAAS,CAAC,mBAAmB,CACtDF,QAAQ,IAAI,EACZ;wBACE,mBACE,AAAwB,YAAxB,OAAOzE,MAAM,MAAM,GACfA,MAAM,MAAM,GACZA,MAAM,MAAM,EAAE;wBACpB,aAAaM;oBACf;oBAEF,IAAIqE,WAAWC,OAAO,IAAI,CAACD,SAAS,MAAM,GAAG,GAAG;wBAC9C9E,MACE,uCACAmE,aACAW;wBAEFD,oBAAoBC;wBACpB,IAAI,CAAC,SAAS,CAAC,yBAAyB,CACtC;4BACE,MAAM;4BACN,QAAQX;4BACR,OAAOW;wBACT,GACAV;oBAEJ,OACEpE,MACE,yDACAmE;gBAGN,EAAE,OAAOjB,OAAO;oBACdlD,MAAM,kCAAkCkD;gBAC1C;qBAEAlD,MAAM;gBAIV,IAAI,CAAC4E,SAAS;oBACZ,IAAIjB,YACF,MAAM,IAAIgB,aACR,CAAC,oBAAoB,EAAExE,MAAM,MAAM,EAAE,EACrCwD;oBAGJ,MAAM,IAAI7B,MAAM,CAAC,mBAAmB,EAAE3B,MAAM,MAAM,EAAE;gBACtD;gBAEA,IAAI6E;gBAEJ,IAAId,yBACFc,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACP,OAAO7E,MAAM,KAAK;oBACpB;gBACF;qBACK,IAAIqE,cACTQ,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACPX;wBACA,aAAaQ;oBACf;gBACF;gBAGFtB,WAAWqB;gBAEX,OAAO;oBACL,QAAQ;wBACNA;oBACF;oBACAI;gBACF;YACF;QACF;QAEA,OAAOtB;IACT;IAteA,YAAY,EAAEuB,iBAAiB,EAAEC,OAAO,EAAEC,SAAS,EAAmB,CAAE;QANxE,uBAAiB,aAAjB;QAEA,uBAAiB,WAAjB;QAEA,uBAAiB,aAAjB;QAGE,IAAI,CAAC,SAAS,GAAGF;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC;IACnB;AAmeF"}
+{"version":3,"file":"agent/task-builder.mjs","sources":["../../../src/agent/task-builder.ts"],"sourcesContent":["import { findAllMidsceneLocatorField, parseActionParam } from '@/ai-model';\nimport type { AbstractInterface } from '@/device';\nimport type Service from '@/service';\nimport type {\n  DetailedLocateParam,\n  ElementCacheFeature,\n  ExecutionTaskActionApply,\n  ExecutionTaskApply,\n  ExecutionTaskHitBy,\n  ExecutionTaskPlanningLocateApply,\n  LocateResultElement,\n  LocateResultWithDump,\n  PlanningAction,\n  PlanningActionParamSleep,\n  PlanningLocateParam,\n  Rect,\n  ServiceDump,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport { sleep } from '@/utils';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { generateElementByPosition } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TaskCache } from './task-cache';\nimport {\n  ifPlanLocateParamIsBbox,\n  matchElementFromCache,\n  matchElementFromPlan,\n} from './utils';\n\nconst debug = getDebug('agent:task-builder');\n\n/**\n * Check if a cache object is non-empty\n */\nfunction hasNonEmptyCache(cache: unknown): boolean {\n  return (\n    cache !== null &&\n    cache !== undefined &&\n    typeof cache === 'object' &&\n    Object.keys(cache).length > 0\n  );\n}\n\nexport function locatePlanForLocate(param: string | DetailedLocateParam) {\n  const locate = typeof param === 'string' ? { prompt: param } : param;\n  const locatePlan: PlanningAction<PlanningLocateParam> = {\n    type: 'Locate',\n    param: locate,\n    thought: '',\n  };\n  return locatePlan;\n}\n\ninterface TaskBuilderDeps {\n  interfaceInstance: AbstractInterface;\n  service: Service;\n  taskCache?: TaskCache;\n}\n\ninterface BuildOptions {\n  cacheable?: boolean;\n  subTask?: boolean;\n}\n\ninterface PlanBuildContext {\n  tasks: ExecutionTaskApply[];\n  modelConfigForPlanning: IModelConfig;\n  modelConfigForDefaultIntent: IModelConfig;\n  cacheable?: boolean;\n  subTask: boolean;\n}\n\nexport class TaskBuilder {\n  private readonly interface: AbstractInterface;\n\n  private readonly service: Service;\n\n  private readonly taskCache?: TaskCache;\n\n  constructor({ interfaceInstance, service, taskCache }: TaskBuilderDeps) {\n    this.interface = interfaceInstance;\n    this.service = service;\n    this.taskCache = taskCache;\n  }\n\n  public async build(\n    plans: PlanningAction[],\n    modelConfigForPlanning: IModelConfig,\n    modelConfigForDefaultIntent: IModelConfig,\n    options?: BuildOptions,\n  ): Promise<{ tasks: ExecutionTaskApply[] }> {\n    const tasks: ExecutionTaskApply[] = [];\n    const cacheable = options?.cacheable;\n\n    const context: PlanBuildContext = {\n      tasks,\n      modelConfigForPlanning,\n      modelConfigForDefaultIntent,\n      cacheable,\n      subTask: !!options?.subTask,\n    };\n\n    type PlanHandler = (plan: PlanningAction) => Promise<void> | void;\n\n    const planHandlers = new Map<string, PlanHandler>([\n      [\n        'Locate',\n        (plan) =>\n          this.handleLocatePlan(\n            plan as PlanningAction<PlanningLocateParam>,\n            context,\n          ),\n      ],\n      ['Finished', (plan) => this.handleFinishedPlan(plan, context)],\n      [\n        'Sleep',\n        (plan) =>\n          this.handleSleepPlan(\n            plan as PlanningAction<PlanningActionParamSleep>,\n            context,\n          ),\n      ],\n    ]);\n\n    const defaultHandler: PlanHandler = (plan) =>\n      this.handleActionPlan(plan, context);\n\n    for (const plan of plans) {\n      const handler = planHandlers.get(plan.type) ?? defaultHandler;\n      await handler(plan);\n    }\n\n    return {\n      tasks,\n    };\n  }\n\n  private handleFinishedPlan(\n    plan: PlanningAction,\n    context: PlanBuildContext,\n  ): void {\n    const taskActionFinished: ExecutionTaskActionApply<null> = {\n      type: 'Action Space',\n      subType: 'Finished',\n      param: null,\n      thought: plan.thought,\n      subTask: context.subTask || undefined,\n      executor: async () => {},\n    };\n    context.tasks.push(taskActionFinished);\n  }\n\n  private handleSleepPlan(\n    plan: PlanningAction<PlanningActionParamSleep>,\n    context: PlanBuildContext,\n  ): void {\n    const sleepTask = this.createSleepTask(plan.param, {\n      thought: plan.thought,\n    });\n    if (context.subTask) {\n      sleepTask.subTask = true;\n    }\n    context.tasks.push(sleepTask);\n  }\n\n  public createSleepTask(\n    param: PlanningActionParamSleep,\n    meta?: { thought?: string },\n  ): ExecutionTaskActionApply<PlanningActionParamSleep> {\n    return {\n      type: 'Action Space',\n      subType: 'Sleep',\n      param,\n      thought: meta?.thought,\n      executor: async (taskParam) => {\n        await sleep(taskParam?.timeMs || 3000);\n      },\n    };\n  }\n\n  private async handleLocatePlan(\n    plan: PlanningAction<PlanningLocateParam>,\n    context: PlanBuildContext,\n  ): Promise<void> {\n    const taskLocate = this.createLocateTask(plan, plan.param, context);\n    context.tasks.push(taskLocate);\n  }\n\n  private async handleActionPlan(\n    plan: PlanningAction,\n    context: PlanBuildContext,\n  ): Promise<void> {\n    const planType = plan.type;\n    const actionSpace = await this.interface.actionSpace();\n    const action = actionSpace.find((item) => item.name === planType);\n    const param = plan.param;\n\n    if (!action) {\n      throw new Error(`Action type '${planType}' not found`);\n    }\n\n    const locateFields = action\n      ? findAllMidsceneLocatorField(action.paramSchema)\n      : [];\n\n    const requiredLocateFields = action\n      ? findAllMidsceneLocatorField(action.paramSchema, true)\n      : [];\n\n    locateFields.forEach((field) => {\n      if (param[field]) {\n        // Always use createLocateTask for all locate params (including bbox)\n        // This ensures cache writing happens even when bbox is available\n        const locatePlan = locatePlanForLocate(param[field]);\n        debug(\n          'will prepend locate param for field',\n          `action.type=${planType}`,\n          `param=${JSON.stringify(param[field])}`,\n          `locatePlan=${JSON.stringify(locatePlan)}`,\n          `hasBbox=${ifPlanLocateParamIsBbox(param[field])}`,\n        );\n        const locateTask = this.createLocateTask(\n          locatePlan,\n          param[field],\n          context,\n          (result) => {\n            param[field] = result;\n          },\n        );\n        context.tasks.push(locateTask);\n      } else {\n        assert(\n          !requiredLocateFields.includes(field),\n          `Required locate field '${field}' is not provided for action ${planType}`,\n        );\n        debug(`field '${field}' is not provided for action ${planType}`);\n      }\n    });\n\n    const task: ExecutionTaskApply<\n      'Action Space',\n      any,\n      { success: boolean; action: string; param: any },\n      void\n    > = {\n      type: 'Action Space',\n      subType: planType,\n      thought: plan.thought,\n      param: plan.param,\n      subTask: context.subTask || undefined,\n      executor: async (param, taskContext) => {\n        debug(\n          'executing action',\n          planType,\n          param,\n          `taskContext.element.center: ${taskContext.element?.center}`,\n        );\n\n        const uiContext = taskContext.uiContext;\n        assert(uiContext, 'uiContext is required for Action task');\n\n        requiredLocateFields.forEach((field) => {\n          assert(\n            param[field],\n            `field '${field}' is required for action ${planType} but not provided. Cannot execute action ${planType}.`,\n          );\n        });\n\n        try {\n          await Promise.all([\n            (async () => {\n              if (this.interface.beforeInvokeAction) {\n                debug('will call \"beforeInvokeAction\" for interface');\n                await this.interface.beforeInvokeAction(action.name, param);\n                debug('called \"beforeInvokeAction\" for interface');\n              }\n            })(),\n            sleep(200),\n          ]);\n        } catch (originalError: any) {\n          const originalMessage =\n            originalError?.message || String(originalError);\n          throw new Error(\n            `error in running beforeInvokeAction for ${action.name}: ${originalMessage}`,\n            { cause: originalError },\n          );\n        }\n\n        if (action.paramSchema) {\n          try {\n            param = parseActionParam(param, action.paramSchema);\n          } catch (error: any) {\n            throw new Error(\n              `Invalid parameters for action ${action.name}: ${error.message}\\nParameters: ${JSON.stringify(param)}`,\n              { cause: error },\n            );\n          }\n        }\n\n        debug('calling action', action.name);\n        const actionFn = action.call.bind(this.interface);\n        const actionResult = await actionFn(param, taskContext);\n        debug('called action', action.name, 'result:', actionResult);\n\n        const delayAfterRunner = action.delayAfterRunner ?? 300;\n        if (delayAfterRunner > 0) {\n          await sleep(delayAfterRunner);\n        }\n\n        try {\n          if (this.interface.afterInvokeAction) {\n            debug('will call \"afterInvokeAction\" for interface');\n            await this.interface.afterInvokeAction(action.name, param);\n            debug('called \"afterInvokeAction\" for interface');\n          }\n        } catch (originalError: any) {\n          const originalMessage =\n            originalError?.message || String(originalError);\n          throw new Error(\n            `error in running afterInvokeAction for ${action.name}: ${originalMessage}`,\n            { cause: originalError },\n          );\n        }\n\n        return {\n          output: actionResult,\n        };\n      },\n    };\n\n    context.tasks.push(task);\n  }\n\n  private createLocateTask(\n    plan: PlanningAction<PlanningLocateParam>,\n    detailedLocateParam: DetailedLocateParam | string,\n    context: PlanBuildContext,\n    onResult?: (result: LocateResultElement) => void,\n  ): ExecutionTaskPlanningLocateApply {\n    const { cacheable, modelConfigForDefaultIntent } = context;\n\n    let locateParam = detailedLocateParam;\n\n    if (typeof locateParam === 'string') {\n      locateParam = {\n        prompt: locateParam,\n      };\n    }\n\n    if (cacheable !== undefined) {\n      locateParam = {\n        ...locateParam,\n        cacheable,\n      };\n    }\n\n    const taskLocator: ExecutionTaskPlanningLocateApply = {\n      type: 'Planning',\n      subType: 'Locate',\n      subTask: context.subTask || undefined,\n      param: locateParam,\n      thought: plan.thought,\n      executor: async (param, taskContext) => {\n        const { task } = taskContext;\n        let { uiContext } = taskContext;\n\n        assert(\n          param?.prompt || param?.bbox,\n          `No prompt or id or position or bbox to locate, param=${JSON.stringify(\n            param,\n          )}`,\n        );\n\n        if (!uiContext) {\n          uiContext = await this.service.contextRetrieverFn();\n        }\n\n        assert(uiContext, 'uiContext is required for Service task');\n\n        let locateDump: ServiceDump | undefined;\n        let locateResult: LocateResultWithDump | undefined;\n\n        const applyDump = (dump?: ServiceDump) => {\n          if (!dump) {\n            return;\n          }\n          locateDump = dump;\n          task.log = {\n            dump,\n          };\n          task.usage = dump.taskInfo?.usage;\n          if (dump.taskInfo?.searchAreaUsage) {\n            task.searchAreaUsage = dump.taskInfo.searchAreaUsage;\n          }\n        };\n\n        // from bbox (plan hit)\n        const elementFromBbox = ifPlanLocateParamIsBbox(param)\n          ? matchElementFromPlan(param)\n          : undefined;\n        const isPlanHit = !!elementFromBbox;\n\n        // from xpath\n        let rectFromXpath: Rect | undefined;\n        if (\n          !isPlanHit &&\n          param.xpath &&\n          this.interface.rectMatchesCacheFeature\n        ) {\n          try {\n            rectFromXpath = await this.interface.rectMatchesCacheFeature({\n              xpaths: [param.xpath],\n            });\n          } catch {\n            // xpath locate failed, allow fallback to cache or AI locate\n          }\n        }\n        const elementFromXpath = rectFromXpath\n          ? generateElementByPosition(\n              {\n                x: rectFromXpath.left + rectFromXpath.width / 2,\n                y: rectFromXpath.top + rectFromXpath.height / 2,\n              },\n              typeof param.prompt === 'string'\n                ? param.prompt\n                : param.prompt?.prompt || '',\n            )\n          : undefined;\n        const isXpathHit = !!elementFromXpath;\n\n        const cachePrompt = param.prompt;\n        const locateCacheRecord = this.taskCache?.matchLocateCache(cachePrompt);\n        const cacheEntry = locateCacheRecord?.cacheContent?.cache;\n\n        const elementFromCache =\n          isPlanHit || isXpathHit\n            ? null\n            : await matchElementFromCache(\n                {\n                  taskCache: this.taskCache,\n                  interfaceInstance: this.interface,\n                },\n                cacheEntry,\n                cachePrompt,\n                param.cacheable,\n              );\n        const isCacheHit = !!elementFromCache;\n\n        let elementFromAiLocate: LocateResultElement | null | undefined;\n        if (!isXpathHit && !isCacheHit && !isPlanHit) {\n          try {\n            locateResult = await this.service.locate(\n              param,\n              {\n                context: uiContext,\n              },\n              modelConfigForDefaultIntent,\n            );\n            applyDump(locateResult.dump);\n            elementFromAiLocate = locateResult.element;\n          } catch (error) {\n            if (error instanceof ServiceError) {\n              applyDump(error.dump);\n            }\n            throw error;\n          }\n        }\n\n        const element =\n          elementFromBbox ||\n          elementFromXpath ||\n          elementFromCache ||\n          elementFromAiLocate;\n\n        // Check if locate cache already exists (for planHitFlag case)\n        const locateCacheAlreadyExists = hasNonEmptyCache(\n          locateCacheRecord?.cacheContent?.cache,\n        );\n\n        let currentCacheEntry: ElementCacheFeature | undefined;\n        // Write cache if:\n        // 1. element found\n        // 2. taskCache enabled\n        // 3. not a cache hit (otherwise we'd be writing what we just read)\n        // 4. not already cached (for bbox/plan hit case, avoid redundant writes)\n        // 5. cacheable is not explicitly false\n        if (\n          element &&\n          this.taskCache &&\n          !isCacheHit &&\n          !locateCacheAlreadyExists &&\n          param?.cacheable !== false\n        ) {\n          if (this.interface.cacheFeatureForRect) {\n            try {\n              const feature = await this.interface.cacheFeatureForRect(\n                element.rect,\n                {\n                  targetDescription:\n                    typeof param.prompt === 'string'\n                      ? param.prompt\n                      : param.prompt?.prompt,\n                  modelConfig: modelConfigForDefaultIntent,\n                },\n              );\n              if (hasNonEmptyCache(feature)) {\n                debug(\n                  'update cache, prompt: %s, cache: %o',\n                  cachePrompt,\n                  feature,\n                );\n                currentCacheEntry = feature;\n                this.taskCache.updateOrAppendCacheRecord(\n                  {\n                    type: 'locate',\n                    prompt: cachePrompt,\n                    cache: feature,\n                  },\n                  locateCacheRecord,\n                );\n              } else {\n                debug(\n                  'no cache data returned, skip cache update, prompt: %s',\n                  cachePrompt,\n                );\n              }\n            } catch (error) {\n              debug('cacheFeatureForRect failed: %s', error);\n            }\n          } else {\n            debug('cacheFeatureForRect is not supported, skip cache update');\n          }\n        }\n\n        if (!element) {\n          if (locateDump) {\n            throw new ServiceError(\n              `Element not found : ${param.prompt}`,\n              locateDump,\n            );\n          }\n          throw new Error(`Element not found: ${param.prompt}`);\n        }\n\n        let hitBy: ExecutionTaskHitBy | undefined;\n\n        if (isPlanHit) {\n          hitBy = {\n            from: 'Plan',\n            context: {\n              bbox: param.bbox,\n            },\n          };\n        } else if (isXpathHit) {\n          hitBy = {\n            from: 'User expected path',\n            context: {\n              xpath: param.xpath,\n            },\n          };\n        } else if (isCacheHit) {\n          hitBy = {\n            from: 'Cache',\n            context: {\n              cacheEntry,\n              cacheToSave: currentCacheEntry,\n            },\n          };\n        }\n\n        onResult?.(element);\n\n        return {\n          output: {\n            element,\n          },\n          hitBy,\n        };\n      },\n    };\n\n    return taskLocator;\n  }\n}\n"],"names":["debug","getDebug","hasNonEmptyCache","cache","Object","locatePlanForLocate","param","locate","locatePlan","TaskBuilder","plans","modelConfigForPlanning","modelConfigForDefaultIntent","options","tasks","cacheable","context","planHandlers","Map","plan","defaultHandler","handler","taskActionFinished","undefined","sleepTask","meta","taskParam","sleep","taskLocate","planType","actionSpace","action","item","Error","locateFields","findAllMidsceneLocatorField","requiredLocateFields","field","JSON","ifPlanLocateParamIsBbox","locateTask","result","assert","task","taskContext","uiContext","Promise","originalError","originalMessage","String","parseActionParam","error","actionFn","actionResult","delayAfterRunner","detailedLocateParam","onResult","locateParam","taskLocator","locateDump","locateResult","applyDump","dump","elementFromBbox","matchElementFromPlan","isPlanHit","rectFromXpath","elementFromXpath","generateElementByPosition","isXpathHit","cachePrompt","locateCacheRecord","cacheEntry","elementFromCache","matchElementFromCache","isCacheHit","elementFromAiLocate","ServiceError","element","locateCacheAlreadyExists","currentCacheEntry","feature","hitBy","interfaceInstance","service","taskCache"],"mappings":";;;;;;;;;;;;;;;;;AA+BA,MAAMA,QAAQC,SAAS;AAKvB,SAASC,iBAAiBC,KAAc;IACtC,OACEA,QAAAA,SAEA,AAAiB,YAAjB,OAAOA,SACPC,OAAO,IAAI,CAACD,OAAO,MAAM,GAAG;AAEhC;AAEO,SAASE,oBAAoBC,KAAmC;IACrE,MAAMC,SAAS,AAAiB,YAAjB,OAAOD,QAAqB;QAAE,QAAQA;IAAM,IAAIA;IAC/D,MAAME,aAAkD;QACtD,MAAM;QACN,OAAOD;QACP,SAAS;IACX;IACA,OAAOC;AACT;AAqBO,MAAMC;IAaX,MAAa,MACXC,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACzCC,OAAsB,EACoB;QAC1C,MAAMC,QAA8B,EAAE;QACtC,MAAMC,YAAYF,SAAS;QAE3B,MAAMG,UAA4B;YAChCF;YACAH;YACAC;YACAG;YACA,SAAS,CAAC,CAACF,SAAS;QACtB;QAIA,MAAMI,eAAe,IAAIC,IAAyB;YAChD;gBACE;gBACA,CAACC,OACC,IAAI,CAAC,gBAAgB,CACnBA,MACAH;aAEL;YACD;gBAAC;gBAAY,CAACG,OAAS,IAAI,CAAC,kBAAkB,CAACA,MAAMH;aAAS;YAC9D;gBACE;gBACA,CAACG,OACC,IAAI,CAAC,eAAe,CAClBA,MACAH;aAEL;SACF;QAED,MAAMI,iBAA8B,CAACD,OACnC,IAAI,CAAC,gBAAgB,CAACA,MAAMH;QAE9B,KAAK,MAAMG,QAAQT,MAAO;YACxB,MAAMW,UAAUJ,aAAa,GAAG,CAACE,KAAK,IAAI,KAAKC;YAC/C,MAAMC,QAAQF;QAChB;QAEA,OAAO;YACLL;QACF;IACF;IAEQ,mBACNK,IAAoB,EACpBH,OAAyB,EACnB;QACN,MAAMM,qBAAqD;YACzD,MAAM;YACN,SAAS;YACT,OAAO;YACP,SAASH,KAAK,OAAO;YACrB,SAASH,QAAQ,OAAO,IAAIO;YAC5B,UAAU,WAAa;QACzB;QACAP,QAAQ,KAAK,CAAC,IAAI,CAACM;IACrB;IAEQ,gBACNH,IAA8C,EAC9CH,OAAyB,EACnB;QACN,MAAMQ,YAAY,IAAI,CAAC,eAAe,CAACL,KAAK,KAAK,EAAE;YACjD,SAASA,KAAK,OAAO;QACvB;QACA,IAAIH,QAAQ,OAAO,EACjBQ,UAAU,OAAO,GAAG;QAEtBR,QAAQ,KAAK,CAAC,IAAI,CAACQ;IACrB;IAEO,gBACLlB,KAA+B,EAC/BmB,IAA2B,EACyB;QACpD,OAAO;YACL,MAAM;YACN,SAAS;YACTnB;YACA,SAASmB,MAAM;YACf,UAAU,OAAOC;gBACf,MAAMC,MAAMD,WAAW,UAAU;YACnC;QACF;IACF;IAEA,MAAc,iBACZP,IAAyC,EACzCH,OAAyB,EACV;QACf,MAAMY,aAAa,IAAI,CAAC,gBAAgB,CAACT,MAAMA,KAAK,KAAK,EAAEH;QAC3DA,QAAQ,KAAK,CAAC,IAAI,CAACY;IACrB;IAEA,MAAc,iBACZT,IAAoB,EACpBH,OAAyB,EACV;QACf,MAAMa,WAAWV,KAAK,IAAI;QAC1B,MAAMW,cAAc,MAAM,IAAI,CAAC,SAAS,CAAC,WAAW;QACpD,MAAMC,SAASD,YAAY,IAAI,CAAC,CAACE,OAASA,KAAK,IAAI,KAAKH;QACxD,MAAMvB,QAAQa,KAAK,KAAK;QAExB,IAAI,CAACY,QACH,MAAM,IAAIE,MAAM,CAAC,aAAa,EAAEJ,SAAS,WAAW,CAAC;QAGvD,MAAMK,eAAeH,SACjBI,4BAA4BJ,OAAO,WAAW,IAC9C,EAAE;QAEN,MAAMK,uBAAuBL,SACzBI,4BAA4BJ,OAAO,WAAW,EAAE,QAChD,EAAE;QAENG,aAAa,OAAO,CAAC,CAACG;YACpB,IAAI/B,KAAK,CAAC+B,MAAM,EAAE;gBAGhB,MAAM7B,aAAaH,oBAAoBC,KAAK,CAAC+B,MAAM;gBACnDrC,MACE,uCACA,CAAC,YAAY,EAAE6B,UAAU,EACzB,CAAC,MAAM,EAAES,KAAK,SAAS,CAAChC,KAAK,CAAC+B,MAAM,GAAG,EACvC,CAAC,WAAW,EAAEC,KAAK,SAAS,CAAC9B,aAAa,EAC1C,CAAC,QAAQ,EAAE+B,wBAAwBjC,KAAK,CAAC+B,MAAM,GAAG;gBAEpD,MAAMG,aAAa,IAAI,CAAC,gBAAgB,CACtChC,YACAF,KAAK,CAAC+B,MAAM,EACZrB,SACA,CAACyB;oBACCnC,KAAK,CAAC+B,MAAM,GAAGI;gBACjB;gBAEFzB,QAAQ,KAAK,CAAC,IAAI,CAACwB;YACrB,OAAO;gBACLE,OACE,CAACN,qBAAqB,QAAQ,CAACC,QAC/B,CAAC,uBAAuB,EAAEA,MAAM,6BAA6B,EAAER,UAAU;gBAE3E7B,MAAM,CAAC,OAAO,EAAEqC,MAAM,6BAA6B,EAAER,UAAU;YACjE;QACF;QAEA,MAAMc,OAKF;YACF,MAAM;YACN,SAASd;YACT,SAASV,KAAK,OAAO;YACrB,OAAOA,KAAK,KAAK;YACjB,SAASH,QAAQ,OAAO,IAAIO;YAC5B,UAAU,OAAOjB,OAAOsC;gBACtB5C,MACE,oBACA6B,UACAvB,OACA,CAAC,4BAA4B,EAAEsC,YAAY,OAAO,EAAE,QAAQ;gBAG9D,MAAMC,YAAYD,YAAY,SAAS;gBACvCF,OAAOG,WAAW;gBAElBT,qBAAqB,OAAO,CAAC,CAACC;oBAC5BK,OACEpC,KAAK,CAAC+B,MAAM,EACZ,CAAC,OAAO,EAAEA,MAAM,yBAAyB,EAAER,SAAS,yCAAyC,EAAEA,SAAS,CAAC,CAAC;gBAE9G;gBAEA,IAAI;oBACF,MAAMiB,QAAQ,GAAG,CAAC;wBACf;4BACC,IAAI,IAAI,CAAC,SAAS,CAAC,kBAAkB,EAAE;gCACrC9C,MAAM;gCACN,MAAM,IAAI,CAAC,SAAS,CAAC,kBAAkB,CAAC+B,OAAO,IAAI,EAAEzB;gCACrDN,MAAM;4BACR;wBACF;wBACA2B,MAAM;qBACP;gBACH,EAAE,OAAOoB,eAAoB;oBAC3B,MAAMC,kBACJD,eAAe,WAAWE,OAAOF;oBACnC,MAAM,IAAId,MACR,CAAC,wCAAwC,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEiB,iBAAiB,EAC5E;wBAAE,OAAOD;oBAAc;gBAE3B;gBAEA,IAAIhB,OAAO,WAAW,EACpB,IAAI;oBACFzB,QAAQ4C,iBAAiB5C,OAAOyB,OAAO,WAAW;gBACpD,EAAE,OAAOoB,OAAY;oBACnB,MAAM,IAAIlB,MACR,CAAC,8BAA8B,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEoB,MAAM,OAAO,CAAC,cAAc,EAAEb,KAAK,SAAS,CAAChC,QAAQ,EACtG;wBAAE,OAAO6C;oBAAM;gBAEnB;gBAGFnD,MAAM,kBAAkB+B,OAAO,IAAI;gBACnC,MAAMqB,WAAWrB,OAAO,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS;gBAChD,MAAMsB,eAAe,MAAMD,SAAS9C,OAAOsC;gBAC3C5C,MAAM,iBAAiB+B,OAAO,IAAI,EAAE,WAAWsB;gBAE/C,MAAMC,mBAAmBvB,OAAO,gBAAgB,IAAI;gBACpD,IAAIuB,mBAAmB,GACrB,MAAM3B,MAAM2B;gBAGd,IAAI;oBACF,IAAI,IAAI,CAAC,SAAS,CAAC,iBAAiB,EAAE;wBACpCtD,MAAM;wBACN,MAAM,IAAI,CAAC,SAAS,CAAC,iBAAiB,CAAC+B,OAAO,IAAI,EAAEzB;wBACpDN,MAAM;oBACR;gBACF,EAAE,OAAO+C,eAAoB;oBAC3B,MAAMC,kBACJD,eAAe,WAAWE,OAAOF;oBACnC,MAAM,IAAId,MACR,CAAC,uCAAuC,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEiB,iBAAiB,EAC3E;wBAAE,OAAOD;oBAAc;gBAE3B;gBAEA,OAAO;oBACL,QAAQM;gBACV;YACF;QACF;QAEArC,QAAQ,KAAK,CAAC,IAAI,CAAC2B;IACrB;IAEQ,iBACNxB,IAAyC,EACzCoC,mBAAiD,EACjDvC,OAAyB,EACzBwC,QAAgD,EACd;QAClC,MAAM,EAAEzC,SAAS,EAAEH,2BAA2B,EAAE,GAAGI;QAEnD,IAAIyC,cAAcF;QAElB,IAAI,AAAuB,YAAvB,OAAOE,aACTA,cAAc;YACZ,QAAQA;QACV;QAGF,IAAI1C,AAAcQ,WAAdR,WACF0C,cAAc;YACZ,GAAGA,WAAW;YACd1C;QACF;QAGF,MAAM2C,cAAgD;YACpD,MAAM;YACN,SAAS;YACT,SAAS1C,QAAQ,OAAO,IAAIO;YAC5B,OAAOkC;YACP,SAAStC,KAAK,OAAO;YACrB,UAAU,OAAOb,OAAOsC;gBACtB,MAAM,EAAED,IAAI,EAAE,GAAGC;gBACjB,IAAI,EAAEC,SAAS,EAAE,GAAGD;gBAEpBF,OACEpC,OAAO,UAAUA,OAAO,MACxB,CAAC,qDAAqD,EAAEgC,KAAK,SAAS,CACpEhC,QACC;gBAGL,IAAI,CAACuC,WACHA,YAAY,MAAM,IAAI,CAAC,OAAO,CAAC,kBAAkB;gBAGnDH,OAAOG,WAAW;gBAElB,IAAIc;gBACJ,IAAIC;gBAEJ,MAAMC,YAAY,CAACC;oBACjB,IAAI,CAACA,MACH;oBAEFH,aAAaG;oBACbnB,KAAK,GAAG,GAAG;wBACTmB;oBACF;oBACAnB,KAAK,KAAK,GAAGmB,KAAK,QAAQ,EAAE;oBAC5B,IAAIA,KAAK,QAAQ,EAAE,iBACjBnB,KAAK,eAAe,GAAGmB,KAAK,QAAQ,CAAC,eAAe;gBAExD;gBAGA,MAAMC,kBAAkBxB,wBAAwBjC,SAC5C0D,qBAAqB1D,SACrBiB;gBACJ,MAAM0C,YAAY,CAAC,CAACF;gBAGpB,IAAIG;gBACJ,IACE,CAACD,aACD3D,MAAM,KAAK,IACX,IAAI,CAAC,SAAS,CAAC,uBAAuB,EAEtC,IAAI;oBACF4D,gBAAgB,MAAM,IAAI,CAAC,SAAS,CAAC,uBAAuB,CAAC;wBAC3D,QAAQ;4BAAC5D,MAAM,KAAK;yBAAC;oBACvB;gBACF,EAAE,OAAM,CAER;gBAEF,MAAM6D,mBAAmBD,gBACrBE,0BACE;oBACE,GAAGF,cAAc,IAAI,GAAGA,cAAc,KAAK,GAAG;oBAC9C,GAAGA,cAAc,GAAG,GAAGA,cAAc,MAAM,GAAG;gBAChD,GACA,AAAwB,YAAxB,OAAO5D,MAAM,MAAM,GACfA,MAAM,MAAM,GACZA,MAAM,MAAM,EAAE,UAAU,MAE9BiB;gBACJ,MAAM8C,aAAa,CAAC,CAACF;gBAErB,MAAMG,cAAchE,MAAM,MAAM;gBAChC,MAAMiE,oBAAoB,IAAI,CAAC,SAAS,EAAE,iBAAiBD;gBAC3D,MAAME,aAAaD,mBAAmB,cAAc;gBAEpD,MAAME,mBACJR,aAAaI,aACT,OACA,MAAMK,sBACJ;oBACE,WAAW,IAAI,CAAC,SAAS;oBACzB,mBAAmB,IAAI,CAAC,SAAS;gBACnC,GACAF,YACAF,aACAhE,MAAM,SAAS;gBAEvB,MAAMqE,aAAa,CAAC,CAACF;gBAErB,IAAIG;gBACJ,IAAI,CAACP,cAAc,CAACM,cAAc,CAACV,WACjC,IAAI;oBACFL,eAAe,MAAM,IAAI,CAAC,OAAO,CAAC,MAAM,CACtCtD,OACA;wBACE,SAASuC;oBACX,GACAjC;oBAEFiD,UAAUD,aAAa,IAAI;oBAC3BgB,sBAAsBhB,aAAa,OAAO;gBAC5C,EAAE,OAAOT,OAAO;oBACd,IAAIA,iBAAiB0B,cACnBhB,UAAUV,MAAM,IAAI;oBAEtB,MAAMA;gBACR;gBAGF,MAAM2B,UACJf,mBACAI,oBACAM,oBACAG;gBAGF,MAAMG,2BAA2B7E,iBAC/BqE,mBAAmB,cAAc;gBAGnC,IAAIS;gBAOJ,IACEF,WACA,IAAI,CAAC,SAAS,IACd,CAACH,cACD,CAACI,4BACDzE,OAAO,cAAc,OAErB,IAAI,IAAI,CAAC,SAAS,CAAC,mBAAmB,EACpC,IAAI;oBACF,MAAM2E,UAAU,MAAM,IAAI,CAAC,SAAS,CAAC,mBAAmB,CACtDH,QAAQ,IAAI,EACZ;wBACE,mBACE,AAAwB,YAAxB,OAAOxE,MAAM,MAAM,GACfA,MAAM,MAAM,GACZA,MAAM,MAAM,EAAE;wBACpB,aAAaM;oBACf;oBAEF,IAAIV,iBAAiB+E,UAAU;wBAC7BjF,MACE,uCACAsE,aACAW;wBAEFD,oBAAoBC;wBACpB,IAAI,CAAC,SAAS,CAAC,yBAAyB,CACtC;4BACE,MAAM;4BACN,QAAQX;4BACR,OAAOW;wBACT,GACAV;oBAEJ,OACEvE,MACE,yDACAsE;gBAGN,EAAE,OAAOnB,OAAO;oBACdnD,MAAM,kCAAkCmD;gBAC1C;qBAEAnD,MAAM;gBAIV,IAAI,CAAC8E,SAAS;oBACZ,IAAInB,YACF,MAAM,IAAIkB,aACR,CAAC,oBAAoB,EAAEvE,MAAM,MAAM,EAAE,EACrCqD;oBAGJ,MAAM,IAAI1B,MAAM,CAAC,mBAAmB,EAAE3B,MAAM,MAAM,EAAE;gBACtD;gBAEA,IAAI4E;gBAEJ,IAAIjB,WACFiB,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACP,MAAM5E,MAAM,IAAI;oBAClB;gBACF;qBACK,IAAI+D,YACTa,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACP,OAAO5E,MAAM,KAAK;oBACpB;gBACF;qBACK,IAAIqE,YACTO,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACPV;wBACA,aAAaQ;oBACf;gBACF;gBAGFxB,WAAWsB;gBAEX,OAAO;oBACL,QAAQ;wBACNA;oBACF;oBACAI;gBACF;YACF;QACF;QAEA,OAAOxB;IACT;IAvfA,YAAY,EAAEyB,iBAAiB,EAAEC,OAAO,EAAEC,SAAS,EAAmB,CAAE;QANxE,uBAAiB,aAAjB;QAEA,uBAAiB,WAAjB;QAEA,uBAAiB,aAAjB;QAGE,IAAI,CAAC,SAAS,GAAGF;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC;IACnB;AAofF"}

package/dist/es/agent/utils.mjs

@@ -100,7 +100,7 @@ async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable
         return;
     }
 }
-const getMidsceneVersion = ()=>"1.0.1-beta-20251205093836.0";
+const getMidsceneVersion = ()=>"1.0.1-beta-20251208031823.0";
 const parsePrompt = (prompt)=>{
     if ('string' == typeof prompt) return {
         textPrompt: prompt,

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

@@ -134,7 +134,7 @@ Please tell what the next one action is (or null if no action should be done) to
 - Make sure the previous actions are completed successfully before performing the next step
 - If there are some error messages reported by the previous actions, don't give up, try parse a new action to recover. If the error persists for more than 5 times, you should think this is an error and set the "error" field to the error message.
 - If there is nothing to do but waiting, set the "sleep" field to the positive waiting time in milliseconds and null for the "action" field.
-- Assertions are also important steps. When getting the assertion instruction, a solid conclusion is required. You should explicitly state your conclusion in the "log" field like this: "Assert: <condition>. I think <...>, so the assertion is <truthy / falsy>". If the assertion is falsy, think this an fatal error and set the reason into the "error" field.
+- Assertions are also important steps. When getting the assertion instruction, a solid conclusion is required. You should explicitly state your conclusion in the "log" field like this: "Assert: <condition>. I can see <...>, and I think <...>, so the assertion is <truthy / falsy>". If the assertion is falsy, think this an fatal error and set the reason into the "error" field.
 
 ## Supporting actions
 ${actionList}

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

@@ -1 +1 @@
-{"version":3,"file":"ai-model/prompt/llm-planning.mjs","sources":["../../../../src/ai-model/prompt/llm-planning.ts"],"sourcesContent":["import type { DeviceAction } from '@/types';\nimport type { TVlModeTypes } from '@midscene/shared/env';\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  const logFieldInstruction = `\n## About the \\`log\\` field (preamble message)\n\nThe \\`log\\` field is a brief preamble message to the user explaining what you’re about to do. It should follow these principles and examples:\n\n- **Use the same language as the user's instruction**\n- **Keep it concise**: be no more than 1-2 sentences, focused on immediate, tangible next steps. (8–12 words or Chinese characters for quick updates).\n- **Build on prior context**: if this is not the first action to be done, use the preamble message to connect the dots with what’s been done so far and create a sense of momentum and clarity for the user to understand your next actions.\n- **Keep your tone light, friendly and curious**: add small touches of personality in preambles feel collaborative and engaging.\n\n**Examples:**\n- \"Click the login button\"\n- \"Scroll to find the 'Yes' button in popup\"\n- \"Previous actions failed to find the 'Yes' button, i will try again\"\n- \"Go back to find the login button\"\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 according to current situation 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\n## Rules\n\n- Don't give extra actions or plans beyond the instruction. 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- Consider the current screenshot and give the action that is most likely to accomplish the instruction. For example, if the next step is to click a button but it's not visible in the screenshot, you should try to find it first instead of give a click action.\n- Make sure the previous actions are completed successfully before performing the next step\n- If there are some error messages reported by the previous actions, don't give up, try parse a new action to recover. If the error persists for more than 5 times, you should think this is an error and set the \"error\" field to the error message.\n- If there is nothing to do but waiting, set the \"sleep\" field to the positive waiting time in milliseconds and null for the \"action\" field.\n- Assertions are also important steps. When getting the assertion instruction, a solid conclusion is required. You should explicitly state your conclusion in the \"log\" field like this: \"Assert: <condition>. I think <...>, so the assertion is <truthy / falsy>\". If the assertion is falsy, think this an fatal error and set the reason into the \"error\" field.\n\n## Supporting actions\n${actionList}\n\n${logFieldInstruction}\n\n## Return format\n\nReturn in JSON format:\n{\n  \"log\": string, // a brief preamble to the user explaining what you’re about to do\n  ${commonOutputFields}\n  \"action\": \n    {\n      \"type\": string, // the type of the action\n      \"param\"?: { // The parameter of the action, if any\n         // k-v style parameter fields\n      }, \n    } | null,\n  ,\n  \"sleep\"?: number, // The sleep time after the action, in milliseconds.\n}\n\nFor example, if the instruction is to login and the form has already been filled, this is a valid return value:\n\n{\n  \"log\": \"Click the login button\",\n  \"more_actions_needed_by_instruction\": false,\n  \"action\": {\n    \"type\": \"Tap\",\n    \"param\": {\n      \"locate\": { \n        \"prompt\": \"The login button\"${vlMode ? `, \"bbox\": [100, 200, 300, 400]` : ''}\n      }\n    }\n  }\n`;\n}\n"],"names":["commonOutputFields","vlLocateParam","vlMode","bboxDescription","descriptionForAction","action","locatorSchemaTypeDescription","tab","fields","paramLines","schema","isZodObject","shape","getTypeName","field","unwrapField","f","typeName","actualField","fieldTypeName","ifMidsceneLocatorField","values","option","String","options","types","opt","console","getDescription","key","Object","isOptional","keyWithOptional","description","paramLine","line","schemaTypeName","paramDescription","systemPromptToTaskPlanning","actionSpace","includeBbox","Error","actionDescriptionList","undefined","actionList","logFieldInstruction"],"mappings":";;AAQA,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;QACtB,MAAMI,aAAuB,EAAE;QAG/B,MAAMC,SAASL,OAAO,WAAW;QACjC,MAAMM,cAAcD,OAAO,IAAI,EAAE,aAAa;QAE9C,IAAIC,eAAeD,OAAO,KAAK,EAAE;YAE/B,MAAME,QAAQF,OAAO,KAAK;YAG1B,MAAMG,cAAc,CAACC;gBAEnB,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,YAAYD;gBAChC,MAAMK,gBAAgBD,YAAY,IAAI,EAAE;gBAExC,IAAIC,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,OAAOZ;oBAET,OAAO;gBACT;gBACA,IAAIa,AAAkB,cAAlBA,eAA6B;oBAC/B,MAAME,SACHH,YAAY,IAAI,EAAE,QACf,IAAI,CAACI,SAAoBC,OAAO,CAAC,CAAC,EAAED,OAAO,CAAC,CAAC,GAC9C,KAAK,SAAS;oBAEnB,OAAO,CAAC,KAAK,EAAED,OAAO,CAAC,CAAC;gBAC1B;gBAEA,IAAIF,AAAkB,eAAlBA,eAA8B;oBAChC,MAAMK,UAAUN,YAAY,IAAI,EAAE;oBAClC,IAAIM,WAAWA,QAAQ,MAAM,GAAG,GAAG;wBAEjC,MAAMC,QAAQD,QAAQ,GAAG,CAAC,CAACE,MAAab,YAAYa;wBACpD,OAAOD,MAAM,IAAI,CAAC;oBACpB;oBACA,OAAO;gBACT;gBAEAE,QAAQ,IAAI,CACV,2EACAT,YAAY,IAAI;gBAElB,OAAOA,YAAY,QAAQ;YAC7B;YAGA,MAAMU,iBAAiB,CAACd;gBAEtB,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,iBAAiBF,OACnB,OAAOA,MAAM,WAAW,IAAI;gBAG9B,MAAMI,cAAcH,YAAYD;gBAGhC,IAAI,iBAAiBI,aACnB,OAAOA,YAAY,WAAW,IAAI;gBAIpC,IAAIA,YAAY,IAAI,EAAE,aAAa,aACjC;oBAAA,IAAI,kCAAkCA,YAAY,IAAI,CAAC,KAAK,IAC1D,OAAO;gBACT;gBAGF,OAAO;YACT;YAEA,KAAK,MAAM,CAACW,KAAKf,MAAM,IAAIgB,OAAO,OAAO,CAAClB,OACxC,IAAIE,SAAS,AAAiB,YAAjB,OAAOA,OAAoB;gBAEtC,MAAMiB,aACJ,AAAqC,cAArC,OAAQjB,MAAc,UAAU,IAC/BA,MAAc,UAAU;gBAC3B,MAAMkB,kBAAkBD,aAAa,GAAGF,IAAI,CAAC,CAAC,GAAGA;gBAGjD,MAAMZ,WAAWJ,YAAYC;gBAG7B,MAAMmB,cAAcL,eAAed;gBAGnC,IAAIoB,YAAY,GAAGF,gBAAgB,EAAE,EAAEf,UAAU;gBACjD,IAAIgB,aACFC,aAAa,CAAC,IAAI,EAAED,aAAa;gBAGnCxB,WAAW,IAAI,CAACyB;YAClB;YAIF,IAAIzB,WAAW,MAAM,GAAG,GAAG;gBACzBD,OAAO,IAAI,CAAC;gBACZC,WAAW,OAAO,CAAC,CAAC0B;oBAClB3B,OAAO,IAAI,CAAC,CAAC,IAAI,EAAE2B,MAAM;gBAC3B;YACF;QACF,OAAO;YAGL,MAAMC,iBAAiB1B,OAAO,IAAI,EAAE;YACpC,IAAIO,WAAW;YAEf,IAAImB,AAAmB,gBAAnBA,gBAAgCnB,WAAW;iBAC1C,IAAImB,AAAmB,gBAAnBA,gBAAgCnB,WAAW;iBAC/C,IAAImB,AAAmB,iBAAnBA,gBAAiCnB,WAAW;YAGrD,MAAMgB,cAAc,iBAAiBvB,SAASA,OAAO,WAAW,GAAG;YAGnE,IAAI2B,mBAAmB,CAAC,SAAS,EAAEpB,UAAU;YAC7C,IAAIgB,aACFI,oBAAoB,CAAC,IAAI,EAAEJ,aAAa;YAE1CI,oBAAoB;YAEpB7B,OAAO,IAAI,CAAC6B;QACd;IACF;IAEA,OAAO,CAAC,EAAE,EAAEhC,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,eAAe+B,2BAA2B,EAC/CC,WAAW,EACXrC,MAAM,EACNsC,WAAW,EAKZ;IAEC,IAAIA,eAAe,CAACtC,QAClB,MAAM,IAAIuC,MACR;IAIJ,MAAMC,wBAAwBH,YAAY,GAAG,CAAC,CAAClC,SACtCD,qBACLC,QACAJ,cAAcuC,cAActC,SAASyC;IAGzC,MAAMC,aAAaF,sBAAsB,IAAI,CAAC;IAE9C,MAAMG,sBAAsB,CAAC;;;;;;;;;;;;;;;AAe/B,CAAC;IAEC,OAAO,CAAC;;;;;;;;;;;;;;;;AAgBV,EAAED,WAAW;;AAEb,EAAEC,oBAAoB;;;;;;;EAOpB,EAAE7C,mBAAmB;;;;;;;;;;;;;;;;;;;;;oCAqBa,EAAEE,SAAS,mCAAmC,GAAG;;;;AAIrF,CAAC;AACD"}
+{"version":3,"file":"ai-model/prompt/llm-planning.mjs","sources":["../../../../src/ai-model/prompt/llm-planning.ts"],"sourcesContent":["import type { DeviceAction } from '@/types';\nimport type { TVlModeTypes } from '@midscene/shared/env';\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  const logFieldInstruction = `\n## About the \\`log\\` field (preamble message)\n\nThe \\`log\\` field is a brief preamble message to the user explaining what you’re about to do. It should follow these principles and examples:\n\n- **Use the same language as the user's instruction**\n- **Keep it concise**: be no more than 1-2 sentences, focused on immediate, tangible next steps. (8–12 words or Chinese characters for quick updates).\n- **Build on prior context**: if this is not the first action to be done, use the preamble message to connect the dots with what’s been done so far and create a sense of momentum and clarity for the user to understand your next actions.\n- **Keep your tone light, friendly and curious**: add small touches of personality in preambles feel collaborative and engaging.\n\n**Examples:**\n- \"Click the login button\"\n- \"Scroll to find the 'Yes' button in popup\"\n- \"Previous actions failed to find the 'Yes' button, i will try again\"\n- \"Go back to find the login button\"\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 according to current situation 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\n## Rules\n\n- Don't give extra actions or plans beyond the instruction. 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- Consider the current screenshot and give the action that is most likely to accomplish the instruction. For example, if the next step is to click a button but it's not visible in the screenshot, you should try to find it first instead of give a click action.\n- Make sure the previous actions are completed successfully before performing the next step\n- If there are some error messages reported by the previous actions, don't give up, try parse a new action to recover. If the error persists for more than 5 times, you should think this is an error and set the \"error\" field to the error message.\n- If there is nothing to do but waiting, set the \"sleep\" field to the positive waiting time in milliseconds and null for the \"action\" field.\n- Assertions are also important steps. When getting the assertion instruction, a solid conclusion is required. You should explicitly state your conclusion in the \"log\" field like this: \"Assert: <condition>. I can see <...>, and I think <...>, so the assertion is <truthy / falsy>\". If the assertion is falsy, think this an fatal error and set the reason into the \"error\" field.\n\n## Supporting actions\n${actionList}\n\n${logFieldInstruction}\n\n## Return format\n\nReturn in JSON format:\n{\n  \"log\": string, // a brief preamble to the user explaining what you’re about to do\n  ${commonOutputFields}\n  \"action\": \n    {\n      \"type\": string, // the type of the action\n      \"param\"?: { // The parameter of the action, if any\n         // k-v style parameter fields\n      }, \n    } | null,\n  ,\n  \"sleep\"?: number, // The sleep time after the action, in milliseconds.\n}\n\nFor example, if the instruction is to login and the form has already been filled, this is a valid return value:\n\n{\n  \"log\": \"Click the login button\",\n  \"more_actions_needed_by_instruction\": false,\n  \"action\": {\n    \"type\": \"Tap\",\n    \"param\": {\n      \"locate\": { \n        \"prompt\": \"The login button\"${vlMode ? `, \"bbox\": [100, 200, 300, 400]` : ''}\n      }\n    }\n  }\n`;\n}\n"],"names":["commonOutputFields","vlLocateParam","vlMode","bboxDescription","descriptionForAction","action","locatorSchemaTypeDescription","tab","fields","paramLines","schema","isZodObject","shape","getTypeName","field","unwrapField","f","typeName","actualField","fieldTypeName","ifMidsceneLocatorField","values","option","String","options","types","opt","console","getDescription","key","Object","isOptional","keyWithOptional","description","paramLine","line","schemaTypeName","paramDescription","systemPromptToTaskPlanning","actionSpace","includeBbox","Error","actionDescriptionList","undefined","actionList","logFieldInstruction"],"mappings":";;AAQA,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;QACtB,MAAMI,aAAuB,EAAE;QAG/B,MAAMC,SAASL,OAAO,WAAW;QACjC,MAAMM,cAAcD,OAAO,IAAI,EAAE,aAAa;QAE9C,IAAIC,eAAeD,OAAO,KAAK,EAAE;YAE/B,MAAME,QAAQF,OAAO,KAAK;YAG1B,MAAMG,cAAc,CAACC;gBAEnB,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,YAAYD;gBAChC,MAAMK,gBAAgBD,YAAY,IAAI,EAAE;gBAExC,IAAIC,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,OAAOZ;oBAET,OAAO;gBACT;gBACA,IAAIa,AAAkB,cAAlBA,eAA6B;oBAC/B,MAAME,SACHH,YAAY,IAAI,EAAE,QACf,IAAI,CAACI,SAAoBC,OAAO,CAAC,CAAC,EAAED,OAAO,CAAC,CAAC,GAC9C,KAAK,SAAS;oBAEnB,OAAO,CAAC,KAAK,EAAED,OAAO,CAAC,CAAC;gBAC1B;gBAEA,IAAIF,AAAkB,eAAlBA,eAA8B;oBAChC,MAAMK,UAAUN,YAAY,IAAI,EAAE;oBAClC,IAAIM,WAAWA,QAAQ,MAAM,GAAG,GAAG;wBAEjC,MAAMC,QAAQD,QAAQ,GAAG,CAAC,CAACE,MAAab,YAAYa;wBACpD,OAAOD,MAAM,IAAI,CAAC;oBACpB;oBACA,OAAO;gBACT;gBAEAE,QAAQ,IAAI,CACV,2EACAT,YAAY,IAAI;gBAElB,OAAOA,YAAY,QAAQ;YAC7B;YAGA,MAAMU,iBAAiB,CAACd;gBAEtB,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,iBAAiBF,OACnB,OAAOA,MAAM,WAAW,IAAI;gBAG9B,MAAMI,cAAcH,YAAYD;gBAGhC,IAAI,iBAAiBI,aACnB,OAAOA,YAAY,WAAW,IAAI;gBAIpC,IAAIA,YAAY,IAAI,EAAE,aAAa,aACjC;oBAAA,IAAI,kCAAkCA,YAAY,IAAI,CAAC,KAAK,IAC1D,OAAO;gBACT;gBAGF,OAAO;YACT;YAEA,KAAK,MAAM,CAACW,KAAKf,MAAM,IAAIgB,OAAO,OAAO,CAAClB,OACxC,IAAIE,SAAS,AAAiB,YAAjB,OAAOA,OAAoB;gBAEtC,MAAMiB,aACJ,AAAqC,cAArC,OAAQjB,MAAc,UAAU,IAC/BA,MAAc,UAAU;gBAC3B,MAAMkB,kBAAkBD,aAAa,GAAGF,IAAI,CAAC,CAAC,GAAGA;gBAGjD,MAAMZ,WAAWJ,YAAYC;gBAG7B,MAAMmB,cAAcL,eAAed;gBAGnC,IAAIoB,YAAY,GAAGF,gBAAgB,EAAE,EAAEf,UAAU;gBACjD,IAAIgB,aACFC,aAAa,CAAC,IAAI,EAAED,aAAa;gBAGnCxB,WAAW,IAAI,CAACyB;YAClB;YAIF,IAAIzB,WAAW,MAAM,GAAG,GAAG;gBACzBD,OAAO,IAAI,CAAC;gBACZC,WAAW,OAAO,CAAC,CAAC0B;oBAClB3B,OAAO,IAAI,CAAC,CAAC,IAAI,EAAE2B,MAAM;gBAC3B;YACF;QACF,OAAO;YAGL,MAAMC,iBAAiB1B,OAAO,IAAI,EAAE;YACpC,IAAIO,WAAW;YAEf,IAAImB,AAAmB,gBAAnBA,gBAAgCnB,WAAW;iBAC1C,IAAImB,AAAmB,gBAAnBA,gBAAgCnB,WAAW;iBAC/C,IAAImB,AAAmB,iBAAnBA,gBAAiCnB,WAAW;YAGrD,MAAMgB,cAAc,iBAAiBvB,SAASA,OAAO,WAAW,GAAG;YAGnE,IAAI2B,mBAAmB,CAAC,SAAS,EAAEpB,UAAU;YAC7C,IAAIgB,aACFI,oBAAoB,CAAC,IAAI,EAAEJ,aAAa;YAE1CI,oBAAoB;YAEpB7B,OAAO,IAAI,CAAC6B;QACd;IACF;IAEA,OAAO,CAAC,EAAE,EAAEhC,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,eAAe+B,2BAA2B,EAC/CC,WAAW,EACXrC,MAAM,EACNsC,WAAW,EAKZ;IAEC,IAAIA,eAAe,CAACtC,QAClB,MAAM,IAAIuC,MACR;IAIJ,MAAMC,wBAAwBH,YAAY,GAAG,CAAC,CAAClC,SACtCD,qBACLC,QACAJ,cAAcuC,cAActC,SAASyC;IAGzC,MAAMC,aAAaF,sBAAsB,IAAI,CAAC;IAE9C,MAAMG,sBAAsB,CAAC;;;;;;;;;;;;;;;AAe/B,CAAC;IAEC,OAAO,CAAC;;;;;;;;;;;;;;;;AAgBV,EAAED,WAAW;;AAEb,EAAEC,oBAAoB;;;;;;;EAOpB,EAAE7C,mBAAmB;;;;;;;;;;;;;;;;;;;;;oCAqBa,EAAEE,SAAS,mCAAmC,GAAG;;;;AAIrF,CAAC;AACD"}