@midscene/core 1.0.1-beta-20251111072933.0 → 1.0.1-beta-20251111100700.0

package/dist/es/agent/ui-utils.mjs

@@ -22,7 +22,7 @@ function locateParamStr(locate) {
 }
 function scrollParamStr(scrollParam) {
     if (!scrollParam) return '';
-    return `${scrollParam.direction || 'down'}, ${scrollParam.scrollType || 'once'}, ${scrollParam.distance || 'distance-not-set'}`;
+    return `${scrollParam.direction || 'down'}, ${scrollParam.scrollType || 'singleAction'}, ${scrollParam.distance || 'distance-not-set'}`;
 }
 function pullParamStr(pullParam) {
     if (!pullParam) return '';

package/dist/es/agent/ui-utils.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"agent/ui-utils.mjs","sources":["webpack://@midscene/core/./src/agent/ui-utils.ts"],"sourcesContent":["import type {\n  DetailedLocateParam,\n  ExecutionTask,\n  ExecutionTaskAction,\n  ExecutionTaskInsightAssertion,\n  ExecutionTaskInsightQuery,\n  ExecutionTaskPlanning,\n  ExecutionTaskPlanningLocate,\n  PullParam,\n  ScrollParam,\n} from '@/types';\n\nexport function typeStr(task: ExecutionTask) {\n  // // For Action tasks with subType, show \"Action Space / subType\"\n  // if (task.type === 'Action' && task.subType) {\n  //   return `Action Space / ${task.subType}`;\n  // }\n\n  // // For all other cases with subType, show \"type / subType\"\n  // if (task.subType) {\n  //   return `${task.type} / ${task.subType}`;\n  // }\n\n  // No subType, just show type\n  return task.subType || task.type;\n}\n\nexport function locateParamStr(locate?: DetailedLocateParam | string): string {\n  if (!locate) {\n    return '';\n  }\n\n  if (typeof locate === 'string') {\n    return locate;\n  }\n\n  if (typeof locate === 'object') {\n    if (typeof locate.prompt === 'string') {\n      return locate.prompt;\n    }\n\n    if (typeof locate.prompt === 'object' && locate.prompt.prompt) {\n      const prompt = locate.prompt.prompt;\n      const images = locate.prompt.images || [];\n\n      if (images.length === 0) return prompt;\n\n      const imagesStr = images\n        .map((image) => {\n          let url = image.url;\n          if (\n            url.startsWith('data:image/') ||\n            (url.startsWith('data:') && url.includes('base64'))\n          ) {\n            url = `${url.substring(0, 15)}...`;\n          }\n          return `[${image.name}](${url})`;\n        })\n        .join(', ');\n\n      return `${prompt}, ${imagesStr}`;\n    }\n  }\n\n  return '';\n}\n\nexport function scrollParamStr(scrollParam?: ScrollParam) {\n  if (!scrollParam) {\n    return '';\n  }\n  return `${scrollParam.direction || 'down'}, ${scrollParam.scrollType || 'once'}, ${scrollParam.distance || 'distance-not-set'}`;\n}\n\nexport function pullParamStr(pullParam?: PullParam) {\n  if (!pullParam) {\n    return '';\n  }\n  const parts: string[] = [];\n  parts.push(`direction: ${pullParam.direction || 'down'}`);\n  if (pullParam.distance) {\n    parts.push(`distance: ${pullParam.distance}`);\n  }\n  if (pullParam.duration) {\n    parts.push(`duration: ${pullParam.duration}ms`);\n  }\n  return parts.join(', ');\n}\n\nexport function taskTitleStr(\n  type:\n    | 'Tap'\n    | 'Hover'\n    | 'Input'\n    | 'RightClick'\n    | 'KeyboardPress'\n    | 'Scroll'\n    | 'Action'\n    | 'Query'\n    | 'Assert'\n    | 'WaitFor'\n    | 'Locate'\n    | 'Boolean'\n    | 'Number'\n    | 'String',\n  prompt: string,\n) {\n  if (prompt) {\n    return `${type} - ${prompt}`;\n  }\n  return type;\n}\n\nexport function paramStr(task: ExecutionTask) {\n  let value: string | undefined | object;\n  if (task.type === 'Planning') {\n    if (task.subType === 'Locate') {\n      value = locateParamStr((task as ExecutionTaskPlanningLocate)?.param);\n    } else {\n      value = (task as ExecutionTaskPlanning)?.param?.userInstruction;\n    }\n  }\n\n  if (task.type === 'Insight') {\n    value =\n      (task as ExecutionTaskInsightQuery)?.param?.dataDemand ||\n      (task as ExecutionTaskInsightAssertion)?.param?.assertion;\n  }\n\n  if (task.type === 'Action Space') {\n    const locate = (task as ExecutionTaskAction)?.param?.locate;\n    const locateStr = locate ? locateParamStr(locate) : '';\n\n    value = task.thought || '';\n    if (typeof (task as ExecutionTaskAction)?.param?.timeMs === 'number') {\n      value = `${(task as ExecutionTaskAction)?.param?.timeMs}ms`;\n    } else if (\n      typeof (task as ExecutionTaskAction)?.param?.scrollType === 'string'\n    ) {\n      value = scrollParamStr((task as ExecutionTaskAction)?.param);\n    } else if (\n      typeof (task as ExecutionTaskAction)?.param?.direction === 'string' &&\n      (task as ExecutionTaskAction)?.subType === 'AndroidPull'\n    ) {\n      value = pullParamStr((task as ExecutionTaskAction)?.param);\n    } else if (\n      typeof (task as ExecutionTaskAction)?.param?.value !== 'undefined'\n    ) {\n      value = (task as ExecutionTaskAction)?.param?.value;\n    } else if (\n      (task as ExecutionTaskAction)?.param &&\n      typeof (task as ExecutionTaskAction)?.param === 'object' &&\n      Object.keys((task as ExecutionTaskAction)?.param || {}).length > 0\n    ) {\n      // General parameter handling for actions with custom parameters\n      // (e.g., runWdaRequest, runAdbShell)\n      value = (task as ExecutionTaskAction)?.param;\n    }\n\n    if (locateStr) {\n      if (value) {\n        value = `${locateStr} - ${value}`;\n      } else {\n        value = locateStr;\n      }\n    }\n  }\n\n  if (typeof value === 'undefined') return '';\n\n  if (typeof value === 'string') return value;\n\n  if (typeof value === 'object' && locateParamStr(value as any)) {\n    return locateParamStr(value as any);\n  }\n\n  return JSON.stringify(value, undefined, 2);\n}\n"],"names":["typeStr","task","locateParamStr","locate","prompt","images","imagesStr","image","url","scrollParamStr","scrollParam","pullParamStr","pullParam","parts","taskTitleStr","type","paramStr","value","_task_param","_task_param1","_task_param2","_task_param3","_task_param4","_task_param5","_task_param6","_task_param7","locateStr","_task_param8","_task_param9","Object","JSON","undefined"],"mappings":"AAYO,SAASA,QAAQC,IAAmB;IAYzC,OAAOA,KAAK,OAAO,IAAIA,KAAK,IAAI;AAClC;AAEO,SAASC,eAAeC,MAAqC;IAClE,IAAI,CAACA,QACH,OAAO;IAGT,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAGT,IAAI,AAAkB,YAAlB,OAAOA,QAAqB;QAC9B,IAAI,AAAyB,YAAzB,OAAOA,OAAO,MAAM,EACtB,OAAOA,OAAO,MAAM;QAGtB,IAAI,AAAyB,YAAzB,OAAOA,OAAO,MAAM,IAAiBA,OAAO,MAAM,CAAC,MAAM,EAAE;YAC7D,MAAMC,SAASD,OAAO,MAAM,CAAC,MAAM;YACnC,MAAME,SAASF,OAAO,MAAM,CAAC,MAAM,IAAI,EAAE;YAEzC,IAAIE,AAAkB,MAAlBA,OAAO,MAAM,EAAQ,OAAOD;YAEhC,MAAME,YAAYD,OACf,GAAG,CAAC,CAACE;gBACJ,IAAIC,MAAMD,MAAM,GAAG;gBACnB,IACEC,IAAI,UAAU,CAAC,kBACdA,IAAI,UAAU,CAAC,YAAYA,IAAI,QAAQ,CAAC,WAEzCA,MAAM,GAAGA,IAAI,SAAS,CAAC,GAAG,IAAI,GAAG,CAAC;gBAEpC,OAAO,CAAC,CAAC,EAAED,MAAM,IAAI,CAAC,EAAE,EAAEC,IAAI,CAAC,CAAC;YAClC,GACC,IAAI,CAAC;YAER,OAAO,GAAGJ,OAAO,EAAE,EAAEE,WAAW;QAClC;IACF;IAEA,OAAO;AACT;AAEO,SAASG,eAAeC,WAAyB;IACtD,IAAI,CAACA,aACH,OAAO;IAET,OAAO,GAAGA,YAAY,SAAS,IAAI,OAAO,EAAE,EAAEA,YAAY,UAAU,IAAI,OAAO,EAAE,EAAEA,YAAY,QAAQ,IAAI,oBAAoB;AACjI;AAEO,SAASC,aAAaC,SAAqB;IAChD,IAAI,CAACA,WACH,OAAO;IAET,MAAMC,QAAkB,EAAE;IAC1BA,MAAM,IAAI,CAAC,CAAC,WAAW,EAAED,UAAU,SAAS,IAAI,QAAQ;IACxD,IAAIA,UAAU,QAAQ,EACpBC,MAAM,IAAI,CAAC,CAAC,UAAU,EAAED,UAAU,QAAQ,EAAE;IAE9C,IAAIA,UAAU,QAAQ,EACpBC,MAAM,IAAI,CAAC,CAAC,UAAU,EAAED,UAAU,QAAQ,CAAC,EAAE,CAAC;IAEhD,OAAOC,MAAM,IAAI,CAAC;AACpB;AAEO,SAASC,aACdC,IAcY,EACZX,MAAc;IAEd,IAAIA,QACF,OAAO,GAAGW,KAAK,GAAG,EAAEX,QAAQ;IAE9B,OAAOW;AACT;AAEO,SAASC,SAASf,IAAmB;IAC1C,IAAIgB;IACJ,IAAIhB,AAAc,eAAdA,KAAK,IAAI,EACX,IAAIA,AAAiB,aAAjBA,KAAK,OAAO,EACdgB,QAAQf,eAAgBD,QAAAA,OAAAA,KAAAA,IAAAA,KAAsC,KAAK;SAC9D;YACGiB;QAARD,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,QAADiB,CAAAA,cAACjB,KAAgC,KAAK,AAAD,IAArCiB,KAAAA,IAAAA,YAAwC,eAAe;IACjE;IAGF,IAAIjB,AAAc,cAAdA,KAAK,IAAI,EAAgB;YAEzBkB,cACAC;QAFFH,QACEE,AAAClB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADkB,CAAAA,eAAClB,KAAoC,KAAK,AAAD,IAAzCkB,KAAAA,IAAAA,aAA4C,UAAU,AAAD,KACpDlB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADmB,CAAAA,eAACnB,KAAwC,KAAK,AAAD,IAA7CmB,KAAAA,IAAAA,aAAgD,SAAS,AAAD;IAC5D;IAEA,IAAInB,AAAc,mBAAdA,KAAK,IAAI,EAAqB;YACjBoB,cAIJC,cAGFC,cAIAC,cAKAC;QAhBT,MAAMtB,SAAUF,QAAAA,OAAAA,KAAAA,IAAAA,QAADoB,CAAAA,eAACpB,KAA8B,KAAK,AAAD,IAAnCoB,KAAAA,IAAAA,aAAsC,MAAM;QAC3D,MAAMK,YAAYvB,SAASD,eAAeC,UAAU;QAEpDc,QAAQhB,KAAK,OAAO,IAAI;QACxB,IAAI,AAAwD,YAAxD,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADqB,CAAAA,eAACrB,KAA8B,KAAK,AAAD,IAAnCqB,KAAAA,IAAAA,aAAsC,MAAM,AAAD,GAAgB;gBACzDK;YAAXV,QAAQ,GAAIhB,QAAAA,OAAAA,KAAAA,IAAAA,QAAD0B,CAAAA,eAAC1B,KAA8B,KAAK,AAAD,IAAnC0B,KAAAA,IAAAA,aAAsC,MAAM,CAAC,EAAE,CAAC;QAC7D,OAAO,IACL,AAA4D,YAA5D,OAAQ1B,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADsB,CAAAA,eAACtB,KAA8B,KAAK,AAAD,IAAnCsB,KAAAA,IAAAA,aAAsC,UAAU,AAAD,GAEtDN,QAAQR,eAAgBR,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;aACtD,IACL,AAA2D,YAA3D,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADuB,CAAAA,eAACvB,KAA8B,KAAK,AAAD,IAAnCuB,KAAAA,IAAAA,aAAsC,SAAS,AAAD,KACrD,AAACvB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,OAAO,AAAD,MAAM,eAE3CgB,QAAQN,aAAcV,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;aACpD,IACL,AAAuD,WAA/CA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADwB,CAAAA,eAACxB,KAA8B,KAAK,AAAD,IAAnCwB,KAAAA,IAAAA,aAAsC,KAAK,AAAD,GACjD;gBACQG;YAARX,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,QAAD2B,CAAAA,eAAC3B,KAA8B,KAAK,AAAD,IAAnC2B,KAAAA,IAAAA,aAAsC,KAAK;QACrD,OAAO,IACL,AAAC3B,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK,AAAD,KACnC,AAAgD,YAAhD,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK,AAAD,KAC1C4B,OAAO,IAAI,CAAC,AAAC5B,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK,AAAD,KAAK,CAAC,GAAG,MAAM,GAAG,GAIjEgB,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;QAG9C,IAAIyB,WAEAT,QADEA,QACM,GAAGS,UAAU,GAAG,EAAET,OAAO,GAEzBS;IAGd;IAEA,IAAI,AAAiB,WAAVT,OAAuB,OAAO;IAEzC,IAAI,AAAiB,YAAjB,OAAOA,OAAoB,OAAOA;IAEtC,IAAI,AAAiB,YAAjB,OAAOA,SAAsBf,eAAee,QAC9C,OAAOf,eAAee;IAGxB,OAAOa,KAAK,SAAS,CAACb,OAAOc,QAAW;AAC1C"}
+{"version":3,"file":"agent/ui-utils.mjs","sources":["webpack://@midscene/core/./src/agent/ui-utils.ts"],"sourcesContent":["import type {\n  DetailedLocateParam,\n  ExecutionTask,\n  ExecutionTaskAction,\n  ExecutionTaskInsightAssertion,\n  ExecutionTaskInsightQuery,\n  ExecutionTaskPlanning,\n  ExecutionTaskPlanningLocate,\n  PullParam,\n  ScrollParam,\n} from '@/types';\n\nexport function typeStr(task: ExecutionTask) {\n  // // For Action tasks with subType, show \"Action Space / subType\"\n  // if (task.type === 'Action' && task.subType) {\n  //   return `Action Space / ${task.subType}`;\n  // }\n\n  // // For all other cases with subType, show \"type / subType\"\n  // if (task.subType) {\n  //   return `${task.type} / ${task.subType}`;\n  // }\n\n  // No subType, just show type\n  return task.subType || task.type;\n}\n\nexport function locateParamStr(locate?: DetailedLocateParam | string): string {\n  if (!locate) {\n    return '';\n  }\n\n  if (typeof locate === 'string') {\n    return locate;\n  }\n\n  if (typeof locate === 'object') {\n    if (typeof locate.prompt === 'string') {\n      return locate.prompt;\n    }\n\n    if (typeof locate.prompt === 'object' && locate.prompt.prompt) {\n      const prompt = locate.prompt.prompt;\n      const images = locate.prompt.images || [];\n\n      if (images.length === 0) return prompt;\n\n      const imagesStr = images\n        .map((image) => {\n          let url = image.url;\n          if (\n            url.startsWith('data:image/') ||\n            (url.startsWith('data:') && url.includes('base64'))\n          ) {\n            url = `${url.substring(0, 15)}...`;\n          }\n          return `[${image.name}](${url})`;\n        })\n        .join(', ');\n\n      return `${prompt}, ${imagesStr}`;\n    }\n  }\n\n  return '';\n}\n\nexport function scrollParamStr(scrollParam?: ScrollParam) {\n  if (!scrollParam) {\n    return '';\n  }\n  return `${scrollParam.direction || 'down'}, ${scrollParam.scrollType || 'singleAction'}, ${scrollParam.distance || 'distance-not-set'}`;\n}\n\nexport function pullParamStr(pullParam?: PullParam) {\n  if (!pullParam) {\n    return '';\n  }\n  const parts: string[] = [];\n  parts.push(`direction: ${pullParam.direction || 'down'}`);\n  if (pullParam.distance) {\n    parts.push(`distance: ${pullParam.distance}`);\n  }\n  if (pullParam.duration) {\n    parts.push(`duration: ${pullParam.duration}ms`);\n  }\n  return parts.join(', ');\n}\n\nexport function taskTitleStr(\n  type:\n    | 'Tap'\n    | 'Hover'\n    | 'Input'\n    | 'RightClick'\n    | 'KeyboardPress'\n    | 'Scroll'\n    | 'Action'\n    | 'Query'\n    | 'Assert'\n    | 'WaitFor'\n    | 'Locate'\n    | 'Boolean'\n    | 'Number'\n    | 'String',\n  prompt: string,\n) {\n  if (prompt) {\n    return `${type} - ${prompt}`;\n  }\n  return type;\n}\n\nexport function paramStr(task: ExecutionTask) {\n  let value: string | undefined | object;\n  if (task.type === 'Planning') {\n    if (task.subType === 'Locate') {\n      value = locateParamStr((task as ExecutionTaskPlanningLocate)?.param);\n    } else {\n      value = (task as ExecutionTaskPlanning)?.param?.userInstruction;\n    }\n  }\n\n  if (task.type === 'Insight') {\n    value =\n      (task as ExecutionTaskInsightQuery)?.param?.dataDemand ||\n      (task as ExecutionTaskInsightAssertion)?.param?.assertion;\n  }\n\n  if (task.type === 'Action Space') {\n    const locate = (task as ExecutionTaskAction)?.param?.locate;\n    const locateStr = locate ? locateParamStr(locate) : '';\n\n    value = task.thought || '';\n    if (typeof (task as ExecutionTaskAction)?.param?.timeMs === 'number') {\n      value = `${(task as ExecutionTaskAction)?.param?.timeMs}ms`;\n    } else if (\n      typeof (task as ExecutionTaskAction)?.param?.scrollType === 'string'\n    ) {\n      value = scrollParamStr((task as ExecutionTaskAction)?.param);\n    } else if (\n      typeof (task as ExecutionTaskAction)?.param?.direction === 'string' &&\n      (task as ExecutionTaskAction)?.subType === 'AndroidPull'\n    ) {\n      value = pullParamStr((task as ExecutionTaskAction)?.param);\n    } else if (\n      typeof (task as ExecutionTaskAction)?.param?.value !== 'undefined'\n    ) {\n      value = (task as ExecutionTaskAction)?.param?.value;\n    } else if (\n      (task as ExecutionTaskAction)?.param &&\n      typeof (task as ExecutionTaskAction)?.param === 'object' &&\n      Object.keys((task as ExecutionTaskAction)?.param || {}).length > 0\n    ) {\n      // General parameter handling for actions with custom parameters\n      // (e.g., runWdaRequest, runAdbShell)\n      value = (task as ExecutionTaskAction)?.param;\n    }\n\n    if (locateStr) {\n      if (value) {\n        value = `${locateStr} - ${value}`;\n      } else {\n        value = locateStr;\n      }\n    }\n  }\n\n  if (typeof value === 'undefined') return '';\n\n  if (typeof value === 'string') return value;\n\n  if (typeof value === 'object' && locateParamStr(value as any)) {\n    return locateParamStr(value as any);\n  }\n\n  return JSON.stringify(value, undefined, 2);\n}\n"],"names":["typeStr","task","locateParamStr","locate","prompt","images","imagesStr","image","url","scrollParamStr","scrollParam","pullParamStr","pullParam","parts","taskTitleStr","type","paramStr","value","_task_param","_task_param1","_task_param2","_task_param3","_task_param4","_task_param5","_task_param6","_task_param7","locateStr","_task_param8","_task_param9","Object","JSON","undefined"],"mappings":"AAYO,SAASA,QAAQC,IAAmB;IAYzC,OAAOA,KAAK,OAAO,IAAIA,KAAK,IAAI;AAClC;AAEO,SAASC,eAAeC,MAAqC;IAClE,IAAI,CAACA,QACH,OAAO;IAGT,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAGT,IAAI,AAAkB,YAAlB,OAAOA,QAAqB;QAC9B,IAAI,AAAyB,YAAzB,OAAOA,OAAO,MAAM,EACtB,OAAOA,OAAO,MAAM;QAGtB,IAAI,AAAyB,YAAzB,OAAOA,OAAO,MAAM,IAAiBA,OAAO,MAAM,CAAC,MAAM,EAAE;YAC7D,MAAMC,SAASD,OAAO,MAAM,CAAC,MAAM;YACnC,MAAME,SAASF,OAAO,MAAM,CAAC,MAAM,IAAI,EAAE;YAEzC,IAAIE,AAAkB,MAAlBA,OAAO,MAAM,EAAQ,OAAOD;YAEhC,MAAME,YAAYD,OACf,GAAG,CAAC,CAACE;gBACJ,IAAIC,MAAMD,MAAM,GAAG;gBACnB,IACEC,IAAI,UAAU,CAAC,kBACdA,IAAI,UAAU,CAAC,YAAYA,IAAI,QAAQ,CAAC,WAEzCA,MAAM,GAAGA,IAAI,SAAS,CAAC,GAAG,IAAI,GAAG,CAAC;gBAEpC,OAAO,CAAC,CAAC,EAAED,MAAM,IAAI,CAAC,EAAE,EAAEC,IAAI,CAAC,CAAC;YAClC,GACC,IAAI,CAAC;YAER,OAAO,GAAGJ,OAAO,EAAE,EAAEE,WAAW;QAClC;IACF;IAEA,OAAO;AACT;AAEO,SAASG,eAAeC,WAAyB;IACtD,IAAI,CAACA,aACH,OAAO;IAET,OAAO,GAAGA,YAAY,SAAS,IAAI,OAAO,EAAE,EAAEA,YAAY,UAAU,IAAI,eAAe,EAAE,EAAEA,YAAY,QAAQ,IAAI,oBAAoB;AACzI;AAEO,SAASC,aAAaC,SAAqB;IAChD,IAAI,CAACA,WACH,OAAO;IAET,MAAMC,QAAkB,EAAE;IAC1BA,MAAM,IAAI,CAAC,CAAC,WAAW,EAAED,UAAU,SAAS,IAAI,QAAQ;IACxD,IAAIA,UAAU,QAAQ,EACpBC,MAAM,IAAI,CAAC,CAAC,UAAU,EAAED,UAAU,QAAQ,EAAE;IAE9C,IAAIA,UAAU,QAAQ,EACpBC,MAAM,IAAI,CAAC,CAAC,UAAU,EAAED,UAAU,QAAQ,CAAC,EAAE,CAAC;IAEhD,OAAOC,MAAM,IAAI,CAAC;AACpB;AAEO,SAASC,aACdC,IAcY,EACZX,MAAc;IAEd,IAAIA,QACF,OAAO,GAAGW,KAAK,GAAG,EAAEX,QAAQ;IAE9B,OAAOW;AACT;AAEO,SAASC,SAASf,IAAmB;IAC1C,IAAIgB;IACJ,IAAIhB,AAAc,eAAdA,KAAK,IAAI,EACX,IAAIA,AAAiB,aAAjBA,KAAK,OAAO,EACdgB,QAAQf,eAAgBD,QAAAA,OAAAA,KAAAA,IAAAA,KAAsC,KAAK;SAC9D;YACGiB;QAARD,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,QAADiB,CAAAA,cAACjB,KAAgC,KAAK,AAAD,IAArCiB,KAAAA,IAAAA,YAAwC,eAAe;IACjE;IAGF,IAAIjB,AAAc,cAAdA,KAAK,IAAI,EAAgB;YAEzBkB,cACAC;QAFFH,QACEE,AAAClB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADkB,CAAAA,eAAClB,KAAoC,KAAK,AAAD,IAAzCkB,KAAAA,IAAAA,aAA4C,UAAU,AAAD,KACpDlB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADmB,CAAAA,eAACnB,KAAwC,KAAK,AAAD,IAA7CmB,KAAAA,IAAAA,aAAgD,SAAS,AAAD;IAC5D;IAEA,IAAInB,AAAc,mBAAdA,KAAK,IAAI,EAAqB;YACjBoB,cAIJC,cAGFC,cAIAC,cAKAC;QAhBT,MAAMtB,SAAUF,QAAAA,OAAAA,KAAAA,IAAAA,QAADoB,CAAAA,eAACpB,KAA8B,KAAK,AAAD,IAAnCoB,KAAAA,IAAAA,aAAsC,MAAM;QAC3D,MAAMK,YAAYvB,SAASD,eAAeC,UAAU;QAEpDc,QAAQhB,KAAK,OAAO,IAAI;QACxB,IAAI,AAAwD,YAAxD,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADqB,CAAAA,eAACrB,KAA8B,KAAK,AAAD,IAAnCqB,KAAAA,IAAAA,aAAsC,MAAM,AAAD,GAAgB;gBACzDK;YAAXV,QAAQ,GAAIhB,QAAAA,OAAAA,KAAAA,IAAAA,QAAD0B,CAAAA,eAAC1B,KAA8B,KAAK,AAAD,IAAnC0B,KAAAA,IAAAA,aAAsC,MAAM,CAAC,EAAE,CAAC;QAC7D,OAAO,IACL,AAA4D,YAA5D,OAAQ1B,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADsB,CAAAA,eAACtB,KAA8B,KAAK,AAAD,IAAnCsB,KAAAA,IAAAA,aAAsC,UAAU,AAAD,GAEtDN,QAAQR,eAAgBR,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;aACtD,IACL,AAA2D,YAA3D,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADuB,CAAAA,eAACvB,KAA8B,KAAK,AAAD,IAAnCuB,KAAAA,IAAAA,aAAsC,SAAS,AAAD,KACrD,AAACvB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,OAAO,AAAD,MAAM,eAE3CgB,QAAQN,aAAcV,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;aACpD,IACL,AAAuD,WAA/CA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADwB,CAAAA,eAACxB,KAA8B,KAAK,AAAD,IAAnCwB,KAAAA,IAAAA,aAAsC,KAAK,AAAD,GACjD;gBACQG;YAARX,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,QAAD2B,CAAAA,eAAC3B,KAA8B,KAAK,AAAD,IAAnC2B,KAAAA,IAAAA,aAAsC,KAAK;QACrD,OAAO,IACL,AAAC3B,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK,AAAD,KACnC,AAAgD,YAAhD,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK,AAAD,KAC1C4B,OAAO,IAAI,CAAC,AAAC5B,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK,AAAD,KAAK,CAAC,GAAG,MAAM,GAAG,GAIjEgB,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;QAG9C,IAAIyB,WAEAT,QADEA,QACM,GAAGS,UAAU,GAAG,EAAET,OAAO,GAEzBS;IAGd;IAEA,IAAI,AAAiB,WAAVT,OAAuB,OAAO;IAEzC,IAAI,AAAiB,YAAjB,OAAOA,OAAoB,OAAOA;IAEtC,IAAI,AAAiB,YAAjB,OAAOA,SAAsBf,eAAee,QAC9C,OAAOf,eAAee;IAGxB,OAAOa,KAAK,SAAS,CAACb,OAAOc,QAAW;AAC1C"}

package/dist/es/agent/utils.mjs

@@ -103,7 +103,7 @@ async function matchElementFromCache(context, cacheEntry, cachePrompt, cacheable
         return;
     }
 }
-const getMidsceneVersion = ()=>"1.0.1-beta-20251111072933.0";
+const getMidsceneVersion = ()=>"1.0.1-beta-20251111100700.0";
 const parsePrompt = (prompt)=>{
     if ('string' == typeof prompt) return {
         textPrompt: prompt,

package/dist/es/device/index.mjs

@@ -81,14 +81,14 @@ const actionScrollParamSchema = z.object({
         'left'
     ]).default('down').describe('The direction to scroll'),
     scrollType: z["enum"]([
-        'once',
-        'untilBottom',
-        'untilTop',
-        'untilRight',
-        'untilLeft'
-    ]).default('once').describe('The scroll type'),
+        'singleAction',
+        'scrollToBottom',
+        'scrollToTop',
+        'scrollToRight',
+        'scrollToLeft'
+    ]).default('singleAction').describe('The scroll behavior: "singleAction" for a single scroll action, "scrollToBottom" for scrolling to the bottom, "scrollToTop" for scrolling to the top, "scrollToRight" for scrolling to the right, "scrollToLeft" for scrolling to the left'),
     distance: z.number().nullable().optional().describe('The distance in pixels to scroll'),
-    locate: getMidsceneLocationSchema().optional().describe('The element to be scrolled')
+    locate: getMidsceneLocationSchema().optional().describe('The target element to be scrolled')
 });
 const defineActionScroll = (call)=>defineAction({
         name: 'Scroll',

package/dist/es/device/index.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"device/index.mjs","sources":["webpack://@midscene/core/./src/device/index.ts"],"sourcesContent":["import { getMidsceneLocationSchema } from '@/ai-model';\nimport type { DeviceAction, LocateResultElement } from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport type { ElementNode } from '@midscene/shared/extractor';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { z } from 'zod';\nimport type { ElementCacheFeature, Rect, Size, UIContext } from '../types';\n\nexport abstract class AbstractInterface {\n  abstract interfaceType: string;\n\n  abstract screenshotBase64(): Promise<string>;\n  abstract size(): Promise<Size>;\n  abstract actionSpace(): DeviceAction[] | Promise<DeviceAction[]>;\n\n  abstract cacheFeatureForRect?(\n    rect: Rect,\n    options?: {\n      targetDescription?: string;\n      modelConfig?: IModelConfig;\n    },\n  ): Promise<ElementCacheFeature>;\n  abstract rectMatchesCacheFeature?(\n    feature: ElementCacheFeature,\n  ): Promise<Rect>;\n\n  abstract destroy?(): Promise<void>;\n\n  abstract describe?(): string;\n  abstract beforeInvokeAction?(actionName: string, param: any): Promise<void>;\n  abstract afterInvokeAction?(actionName: string, param: any): Promise<void>;\n\n  // @deprecated do NOT extend this method\n  abstract getElementsNodeTree?: () => Promise<ElementNode>;\n\n  // @deprecated do NOT extend this method\n  abstract url?: () => string | Promise<string>;\n\n  // @deprecated do NOT extend this method\n  abstract evaluateJavaScript?<T = any>(script: string): Promise<T>;\n\n  // @deprecated do NOT extend this method\n  abstract getContext?(): Promise<UIContext>;\n}\n\n// Generic function to define actions with proper type inference\n// TRuntime allows specifying a different type for the runtime parameter (after location resolution)\n// TReturn allows specifying the return type of the action\nexport const defineAction = <\n  TSchema extends z.ZodType | undefined = undefined,\n  TRuntime = TSchema extends z.ZodType ? z.infer<TSchema> : undefined,\n  TReturn = any,\n>(\n  config: {\n    name: string;\n    description: string;\n    interfaceAlias?: string;\n    paramSchema?: TSchema;\n    call: (param: TRuntime) => Promise<TReturn> | TReturn;\n  } & Partial<\n    Omit<\n      DeviceAction<TRuntime, TReturn>,\n      'name' | 'description' | 'interfaceAlias' | 'paramSchema' | 'call'\n    >\n  >,\n): DeviceAction<TRuntime, TReturn> => {\n  return config as any; // Type assertion needed because schema validation type differs from runtime type\n};\n\n// Tap\nexport const actionTapParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The element to be tapped'),\n});\n// Override the inferred type to use LocateResultElement for the runtime locate field\nexport type ActionTapParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionTap = (\n  call: (param: ActionTapParam) => Promise<void>,\n): DeviceAction<ActionTapParam> => {\n  return defineAction<typeof actionTapParamSchema, ActionTapParam>({\n    name: 'Tap',\n    description: 'Tap the element',\n    interfaceAlias: 'aiTap',\n    paramSchema: actionTapParamSchema,\n    call,\n  });\n};\n\n// RightClick\nexport const actionRightClickParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be right clicked',\n  ),\n});\nexport type ActionRightClickParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionRightClick = (\n  call: (param: ActionRightClickParam) => Promise<void>,\n): DeviceAction<ActionRightClickParam> => {\n  return defineAction<\n    typeof actionRightClickParamSchema,\n    ActionRightClickParam\n  >({\n    name: 'RightClick',\n    description: 'Right click the element',\n    interfaceAlias: 'aiRightClick',\n    paramSchema: actionRightClickParamSchema,\n    call,\n  });\n};\n\n// DoubleClick\nexport const actionDoubleClickParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be double clicked',\n  ),\n});\nexport type ActionDoubleClickParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionDoubleClick = (\n  call: (param: ActionDoubleClickParam) => Promise<void>,\n): DeviceAction<ActionDoubleClickParam> => {\n  return defineAction<\n    typeof actionDoubleClickParamSchema,\n    ActionDoubleClickParam\n  >({\n    name: 'DoubleClick',\n    description: 'Double click the element',\n    interfaceAlias: 'aiDoubleClick',\n    paramSchema: actionDoubleClickParamSchema,\n    call,\n  });\n};\n\n// Hover\nexport const actionHoverParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The element to be hovered'),\n});\nexport type ActionHoverParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionHover = (\n  call: (param: ActionHoverParam) => Promise<void>,\n): DeviceAction<ActionHoverParam> => {\n  return defineAction<typeof actionHoverParamSchema, ActionHoverParam>({\n    name: 'Hover',\n    description: 'Move the mouse to the element',\n    interfaceAlias: 'aiHover',\n    paramSchema: actionHoverParamSchema,\n    call,\n  });\n};\n\n// Input\nexport const actionInputParamSchema = z.object({\n  value: z\n    .union([z.string(), z.number()])\n    .transform((val) => String(val))\n    .describe(\n      'The text to input. Provide the final content for replace/append modes, or an empty string when using clear mode to remove existing text.',\n    ),\n  locate: getMidsceneLocationSchema()\n    .describe('The element to be input')\n    .optional(),\n  mode: z\n    .enum(['replace', 'clear', 'append'])\n    .default('replace')\n    .optional()\n    .describe(\n      'Input mode: \"replace\" (default) - clear the field and input the value; \"append\" - append the value to existing content; \"clear\" - clear the field without inputting new text.',\n    ),\n});\nexport type ActionInputParam = {\n  value: string;\n  locate?: LocateResultElement;\n  mode?: 'replace' | 'clear' | 'append';\n};\n\nexport const defineActionInput = (\n  call: (param: ActionInputParam) => Promise<void>,\n): DeviceAction<ActionInputParam> => {\n  return defineAction<typeof actionInputParamSchema, ActionInputParam>({\n    name: 'Input',\n    description: 'Input the value into the element',\n    interfaceAlias: 'aiInput',\n    paramSchema: actionInputParamSchema,\n    call,\n  });\n};\n\n// KeyboardPress\nexport const actionKeyboardPressParamSchema = z.object({\n  locate: getMidsceneLocationSchema()\n    .describe('The element to be clicked before pressing the key')\n    .optional(),\n  keyName: z\n    .string()\n    .describe(\n      \"The key to be pressed. Use '+' for key combinations, e.g., 'Control+A', 'Shift+Enter'\",\n    ),\n});\nexport type ActionKeyboardPressParam = {\n  locate?: LocateResultElement;\n  keyName: string;\n};\n\nexport const defineActionKeyboardPress = (\n  call: (param: ActionKeyboardPressParam) => Promise<void>,\n): DeviceAction<ActionKeyboardPressParam> => {\n  return defineAction<\n    typeof actionKeyboardPressParamSchema,\n    ActionKeyboardPressParam\n  >({\n    name: 'KeyboardPress',\n    description:\n      'Press a key or key combination, like \"Enter\", \"Tab\", \"Escape\", or \"Control+A\", \"Shift+Enter\". Do not use this to type text.',\n    interfaceAlias: 'aiKeyboardPress',\n    paramSchema: actionKeyboardPressParamSchema,\n    call,\n  });\n};\n\n// Scroll\nexport const actionScrollParamSchema = z.object({\n  direction: z\n    .enum(['down', 'up', 'right', 'left'])\n    .default('down')\n    .describe('The direction to scroll'),\n  scrollType: z\n    .enum(['once', 'untilBottom', 'untilTop', 'untilRight', 'untilLeft'])\n    .default('once')\n    .describe('The scroll type'),\n  distance: z\n    .number()\n    .nullable()\n    .optional()\n    .describe('The distance in pixels to scroll'),\n  locate: getMidsceneLocationSchema()\n    .optional()\n    .describe('The element to be scrolled'),\n});\nexport type ActionScrollParam = {\n  direction?: 'down' | 'up' | 'right' | 'left';\n  scrollType?: 'once' | 'untilBottom' | 'untilTop' | 'untilRight' | 'untilLeft';\n  distance?: number | null;\n  locate?: LocateResultElement;\n};\n\nexport const defineActionScroll = (\n  call: (param: ActionScrollParam) => Promise<void>,\n): DeviceAction<ActionScrollParam> => {\n  return defineAction<typeof actionScrollParamSchema, ActionScrollParam>({\n    name: 'Scroll',\n    description:\n      'Scroll the page or an element. The direction to scroll, the scroll type, and the distance to scroll. The distance is the number of pixels to scroll. If not specified, use `down` direction, `once` scroll type, and `null` distance.',\n    interfaceAlias: 'aiScroll',\n    paramSchema: actionScrollParamSchema,\n    call,\n  });\n};\n\n// DragAndDrop\nexport const actionDragAndDropParamSchema = z.object({\n  from: getMidsceneLocationSchema().describe('The position to be dragged'),\n  to: getMidsceneLocationSchema().describe('The position to be dropped'),\n});\nexport type ActionDragAndDropParam = {\n  from: LocateResultElement;\n  to: LocateResultElement;\n};\n\nexport const defineActionDragAndDrop = (\n  call: (param: ActionDragAndDropParam) => Promise<void>,\n): DeviceAction<ActionDragAndDropParam> => {\n  return defineAction<\n    typeof actionDragAndDropParamSchema,\n    ActionDragAndDropParam\n  >({\n    name: 'DragAndDrop',\n    description: 'Drag and drop the element',\n    interfaceAlias: 'aiDragAndDrop',\n    paramSchema: actionDragAndDropParamSchema,\n    call,\n  });\n};\n\nexport const ActionLongPressParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be long pressed',\n  ),\n  duration: z\n    .number()\n    .default(500)\n    .optional()\n    .describe('Long press duration in milliseconds'),\n});\n\nexport type ActionLongPressParam = {\n  locate: LocateResultElement;\n  duration?: number;\n};\nexport const defineActionLongPress = (\n  call: (param: ActionLongPressParam) => Promise<void>,\n): DeviceAction<ActionLongPressParam> => {\n  return defineAction<typeof ActionLongPressParamSchema, ActionLongPressParam>({\n    name: 'LongPress',\n    description: 'Long press the element',\n    paramSchema: ActionLongPressParamSchema,\n    call,\n  });\n};\n\nexport const ActionSwipeParamSchema = z.object({\n  start: getMidsceneLocationSchema()\n    .optional()\n    .describe(\n      'Starting point of the swipe gesture, if not specified, the center of the page will be used',\n    ),\n  direction: z\n    .enum(['up', 'down', 'left', 'right'])\n    .optional()\n    .describe(\n      'The direction to swipe (required when using distance). The direction means the direction of the finger swipe.',\n    ),\n  distance: z\n    .number()\n    .optional()\n    .describe('The distance in pixels to swipe (mutually exclusive with end)'),\n  end: getMidsceneLocationSchema()\n    .optional()\n    .describe(\n      'Ending point of the swipe gesture (mutually exclusive with distance)',\n    ),\n  duration: z\n    .number()\n    .default(300)\n    .describe('Duration of the swipe gesture in milliseconds'),\n  repeat: z\n    .number()\n    .optional()\n    .describe(\n      'The number of times to repeat the swipe gesture. 1 for default, 0 for infinite (e.g. endless swipe until the end of the page)',\n    ),\n});\n\nexport type ActionSwipeParam = {\n  start?: LocateResultElement;\n  direction?: 'up' | 'down' | 'left' | 'right';\n  distance?: number;\n  end?: LocateResultElement;\n  duration?: number;\n  repeat?: number;\n};\n\nexport const defineActionSwipe = (\n  call: (param: ActionSwipeParam) => Promise<void>,\n): DeviceAction<ActionSwipeParam> => {\n  return defineAction<typeof ActionSwipeParamSchema, ActionSwipeParam>({\n    name: 'Swipe',\n    description:\n      'Perform a swipe gesture. You must specify either \"end\" (target location) or \"distance\" + \"direction\" - they are mutually exclusive. Use \"end\" for precise location-based swipes, or \"distance\" + \"direction\" for relative movement.',\n    paramSchema: ActionSwipeParamSchema,\n    call,\n  });\n};\n\n// ClearInput\nexport const actionClearInputParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The input field to be cleared'),\n});\nexport type ActionClearInputParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionClearInput = (\n  call: (param: ActionClearInputParam) => Promise<void>,\n): DeviceAction<ActionClearInputParam> => {\n  return defineAction<\n    typeof actionClearInputParamSchema,\n    ActionClearInputParam\n  >({\n    name: 'ClearInput',\n    description: 'Clear the text content of an input field',\n    interfaceAlias: 'aiClearInput',\n    paramSchema: actionClearInputParamSchema,\n    call,\n  });\n};\n\nexport type { DeviceAction } from '../types';\nexport type {\n  AndroidDeviceOpt,\n  AndroidDeviceInputOpt,\n  IOSDeviceOpt,\n  IOSDeviceInputOpt,\n} from './device-options';\n"],"names":["AbstractInterface","defineAction","config","actionTapParamSchema","z","getMidsceneLocationSchema","defineActionTap","call","actionRightClickParamSchema","defineActionRightClick","actionDoubleClickParamSchema","defineActionDoubleClick","actionHoverParamSchema","defineActionHover","actionInputParamSchema","val","String","defineActionInput","actionKeyboardPressParamSchema","defineActionKeyboardPress","actionScrollParamSchema","defineActionScroll","actionDragAndDropParamSchema","defineActionDragAndDrop","ActionLongPressParamSchema","defineActionLongPress","ActionSwipeParamSchema","defineActionSwipe","actionClearInputParamSchema","defineActionClearInput"],"mappings":";;AAQO,MAAeA;AAmCtB;AAKO,MAAMC,eAAe,CAK1BC,SAaOA;AAIF,MAAMC,uBAAuBC,EAAE,MAAM,CAAC;IAC3C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAMO,MAAMC,kBAAkB,CAC7BC,OAEON,aAA0D;QAC/D,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaE;QACbI;IACF;AAIK,MAAMC,8BAA8BJ,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMI,yBAAyB,CACpCF,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaO;QACbD;IACF;AAIK,MAAMG,+BAA+BN,EAAE,MAAM,CAAC;IACnD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMM,0BAA0B,CACrCJ,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaS;QACbH;IACF;AAIK,MAAMK,yBAAyBR,EAAE,MAAM,CAAC;IAC7C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMQ,oBAAoB,CAC/BN,OAEON,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaW;QACbL;IACF;AAIK,MAAMO,yBAAyBV,EAAE,MAAM,CAAC;IAC7C,OAAOA,EAAAA,KACC,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,MAAM;KAAG,EAC9B,SAAS,CAAC,CAACW,MAAQC,OAAOD,MAC1B,QAAQ,CACP;IAEJ,QAAQV,4BACL,QAAQ,CAAC,2BACT,QAAQ;IACX,MAAMD,CAAC,CAADA,OACC,CAAC;QAAC;QAAW;QAAS;KAAS,EACnC,OAAO,CAAC,WACR,QAAQ,GACR,QAAQ,CACP;AAEN;AAOO,MAAMa,oBAAoB,CAC/BV,OAEON,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaa;QACbP;IACF;AAIK,MAAMW,iCAAiCd,EAAE,MAAM,CAAC;IACrD,QAAQC,4BACL,QAAQ,CAAC,qDACT,QAAQ;IACX,SAASD,EAAAA,MACA,GACN,QAAQ,CACP;AAEN;AAMO,MAAMe,4BAA4B,CACvCZ,OAEON,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAaiB;QACbX;IACF;AAIK,MAAMa,0BAA0BhB,EAAE,MAAM,CAAC;IAC9C,WAAWA,CAAC,CAADA,OACJ,CAAC;QAAC;QAAQ;QAAM;QAAS;KAAO,EACpC,OAAO,CAAC,QACR,QAAQ,CAAC;IACZ,YAAYA,CAAC,CAADA,OACL,CAAC;QAAC;QAAQ;QAAe;QAAY;QAAc;KAAY,EACnE,OAAO,CAAC,QACR,QAAQ,CAAC;IACZ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,CAAC;IACZ,QAAQC,4BACL,QAAQ,GACR,QAAQ,CAAC;AACd;AAQO,MAAMgB,qBAAqB,CAChCd,OAEON,aAAgE;QACrE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAamB;QACbb;IACF;AAIK,MAAMe,+BAA+BlB,EAAE,MAAM,CAAC;IACnD,MAAMC,4BAA4B,QAAQ,CAAC;IAC3C,IAAIA,4BAA4B,QAAQ,CAAC;AAC3C;AAMO,MAAMkB,0BAA0B,CACrChB,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaqB;QACbf;IACF;AAGK,MAAMiB,6BAA6BpB,EAAE,MAAM,CAAC;IACjD,QAAQC,4BAA4B,QAAQ,CAC1C;IAEF,UAAUD,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,GACR,QAAQ,CAAC;AACd;AAMO,MAAMqB,wBAAwB,CACnClB,OAEON,aAAsE;QAC3E,MAAM;QACN,aAAa;QACb,aAAauB;QACbjB;IACF;AAGK,MAAMmB,yBAAyBtB,EAAE,MAAM,CAAC;IAC7C,OAAOC,4BACJ,QAAQ,GACR,QAAQ,CACP;IAEJ,WAAWD,CAAC,CAADA,OACJ,CAAC;QAAC;QAAM;QAAQ;QAAQ;KAAQ,EACpC,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,CAAC;IACZ,KAAKC,4BACF,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUD,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,CAAC;IACZ,QAAQA,EAAAA,MACC,GACN,QAAQ,GACR,QAAQ,CACP;AAEN;AAWO,MAAMuB,oBAAoB,CAC/BpB,OAEON,aAA8D;QACnE,MAAM;QACN,aACE;QACF,aAAayB;QACbnB;IACF;AAIK,MAAMqB,8BAA8BxB,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMwB,yBAAyB,CACpCtB,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAa2B;QACbrB;IACF"}
+{"version":3,"file":"device/index.mjs","sources":["webpack://@midscene/core/./src/device/index.ts"],"sourcesContent":["import { getMidsceneLocationSchema } from '@/ai-model';\nimport type { DeviceAction, LocateResultElement } from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport type { ElementNode } from '@midscene/shared/extractor';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { z } from 'zod';\nimport type { ElementCacheFeature, Rect, Size, UIContext } from '../types';\n\nexport abstract class AbstractInterface {\n  abstract interfaceType: string;\n\n  abstract screenshotBase64(): Promise<string>;\n  abstract size(): Promise<Size>;\n  abstract actionSpace(): DeviceAction[] | Promise<DeviceAction[]>;\n\n  abstract cacheFeatureForRect?(\n    rect: Rect,\n    options?: {\n      targetDescription?: string;\n      modelConfig?: IModelConfig;\n    },\n  ): Promise<ElementCacheFeature>;\n  abstract rectMatchesCacheFeature?(\n    feature: ElementCacheFeature,\n  ): Promise<Rect>;\n\n  abstract destroy?(): Promise<void>;\n\n  abstract describe?(): string;\n  abstract beforeInvokeAction?(actionName: string, param: any): Promise<void>;\n  abstract afterInvokeAction?(actionName: string, param: any): Promise<void>;\n\n  // @deprecated do NOT extend this method\n  abstract getElementsNodeTree?: () => Promise<ElementNode>;\n\n  // @deprecated do NOT extend this method\n  abstract url?: () => string | Promise<string>;\n\n  // @deprecated do NOT extend this method\n  abstract evaluateJavaScript?<T = any>(script: string): Promise<T>;\n\n  // @deprecated do NOT extend this method\n  abstract getContext?(): Promise<UIContext>;\n}\n\n// Generic function to define actions with proper type inference\n// TRuntime allows specifying a different type for the runtime parameter (after location resolution)\n// TReturn allows specifying the return type of the action\nexport const defineAction = <\n  TSchema extends z.ZodType | undefined = undefined,\n  TRuntime = TSchema extends z.ZodType ? z.infer<TSchema> : undefined,\n  TReturn = any,\n>(\n  config: {\n    name: string;\n    description: string;\n    interfaceAlias?: string;\n    paramSchema?: TSchema;\n    call: (param: TRuntime) => Promise<TReturn> | TReturn;\n  } & Partial<\n    Omit<\n      DeviceAction<TRuntime, TReturn>,\n      'name' | 'description' | 'interfaceAlias' | 'paramSchema' | 'call'\n    >\n  >,\n): DeviceAction<TRuntime, TReturn> => {\n  return config as any; // Type assertion needed because schema validation type differs from runtime type\n};\n\n// Tap\nexport const actionTapParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The element to be tapped'),\n});\n// Override the inferred type to use LocateResultElement for the runtime locate field\nexport type ActionTapParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionTap = (\n  call: (param: ActionTapParam) => Promise<void>,\n): DeviceAction<ActionTapParam> => {\n  return defineAction<typeof actionTapParamSchema, ActionTapParam>({\n    name: 'Tap',\n    description: 'Tap the element',\n    interfaceAlias: 'aiTap',\n    paramSchema: actionTapParamSchema,\n    call,\n  });\n};\n\n// RightClick\nexport const actionRightClickParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be right clicked',\n  ),\n});\nexport type ActionRightClickParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionRightClick = (\n  call: (param: ActionRightClickParam) => Promise<void>,\n): DeviceAction<ActionRightClickParam> => {\n  return defineAction<\n    typeof actionRightClickParamSchema,\n    ActionRightClickParam\n  >({\n    name: 'RightClick',\n    description: 'Right click the element',\n    interfaceAlias: 'aiRightClick',\n    paramSchema: actionRightClickParamSchema,\n    call,\n  });\n};\n\n// DoubleClick\nexport const actionDoubleClickParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be double clicked',\n  ),\n});\nexport type ActionDoubleClickParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionDoubleClick = (\n  call: (param: ActionDoubleClickParam) => Promise<void>,\n): DeviceAction<ActionDoubleClickParam> => {\n  return defineAction<\n    typeof actionDoubleClickParamSchema,\n    ActionDoubleClickParam\n  >({\n    name: 'DoubleClick',\n    description: 'Double click the element',\n    interfaceAlias: 'aiDoubleClick',\n    paramSchema: actionDoubleClickParamSchema,\n    call,\n  });\n};\n\n// Hover\nexport const actionHoverParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The element to be hovered'),\n});\nexport type ActionHoverParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionHover = (\n  call: (param: ActionHoverParam) => Promise<void>,\n): DeviceAction<ActionHoverParam> => {\n  return defineAction<typeof actionHoverParamSchema, ActionHoverParam>({\n    name: 'Hover',\n    description: 'Move the mouse to the element',\n    interfaceAlias: 'aiHover',\n    paramSchema: actionHoverParamSchema,\n    call,\n  });\n};\n\n// Input\nexport const actionInputParamSchema = z.object({\n  value: z\n    .union([z.string(), z.number()])\n    .transform((val) => String(val))\n    .describe(\n      'The text to input. Provide the final content for replace/append modes, or an empty string when using clear mode to remove existing text.',\n    ),\n  locate: getMidsceneLocationSchema()\n    .describe('The element to be input')\n    .optional(),\n  mode: z\n    .enum(['replace', 'clear', 'append'])\n    .default('replace')\n    .optional()\n    .describe(\n      'Input mode: \"replace\" (default) - clear the field and input the value; \"append\" - append the value to existing content; \"clear\" - clear the field without inputting new text.',\n    ),\n});\nexport type ActionInputParam = {\n  value: string;\n  locate?: LocateResultElement;\n  mode?: 'replace' | 'clear' | 'append';\n};\n\nexport const defineActionInput = (\n  call: (param: ActionInputParam) => Promise<void>,\n): DeviceAction<ActionInputParam> => {\n  return defineAction<typeof actionInputParamSchema, ActionInputParam>({\n    name: 'Input',\n    description: 'Input the value into the element',\n    interfaceAlias: 'aiInput',\n    paramSchema: actionInputParamSchema,\n    call,\n  });\n};\n\n// KeyboardPress\nexport const actionKeyboardPressParamSchema = z.object({\n  locate: getMidsceneLocationSchema()\n    .describe('The element to be clicked before pressing the key')\n    .optional(),\n  keyName: z\n    .string()\n    .describe(\n      \"The key to be pressed. Use '+' for key combinations, e.g., 'Control+A', 'Shift+Enter'\",\n    ),\n});\nexport type ActionKeyboardPressParam = {\n  locate?: LocateResultElement;\n  keyName: string;\n};\n\nexport const defineActionKeyboardPress = (\n  call: (param: ActionKeyboardPressParam) => Promise<void>,\n): DeviceAction<ActionKeyboardPressParam> => {\n  return defineAction<\n    typeof actionKeyboardPressParamSchema,\n    ActionKeyboardPressParam\n  >({\n    name: 'KeyboardPress',\n    description:\n      'Press a key or key combination, like \"Enter\", \"Tab\", \"Escape\", or \"Control+A\", \"Shift+Enter\". Do not use this to type text.',\n    interfaceAlias: 'aiKeyboardPress',\n    paramSchema: actionKeyboardPressParamSchema,\n    call,\n  });\n};\n\n// Scroll\nexport const actionScrollParamSchema = z.object({\n  direction: z\n    .enum(['down', 'up', 'right', 'left'])\n    .default('down')\n    .describe('The direction to scroll'),\n  scrollType: z\n    .enum([\n      'singleAction',\n      'scrollToBottom',\n      'scrollToTop',\n      'scrollToRight',\n      'scrollToLeft',\n    ])\n    .default('singleAction')\n    .describe(\n      'The scroll behavior: \"singleAction\" for a single scroll action, \"scrollToBottom\" for scrolling to the bottom, \"scrollToTop\" for scrolling to the top, \"scrollToRight\" for scrolling to the right, \"scrollToLeft\" for scrolling to the left',\n    ),\n  distance: z\n    .number()\n    .nullable()\n    .optional()\n    .describe('The distance in pixels to scroll'),\n  locate: getMidsceneLocationSchema()\n    .optional()\n    .describe('The target element to be scrolled'),\n});\nexport type ActionScrollParam = {\n  direction?: 'down' | 'up' | 'right' | 'left';\n  scrollType?:\n    | 'singleAction'\n    | 'scrollToBottom'\n    | 'scrollToTop'\n    | 'scrollToRight'\n    | 'scrollToLeft';\n  distance?: number | null;\n  locate?: LocateResultElement;\n};\n\nexport const defineActionScroll = (\n  call: (param: ActionScrollParam) => Promise<void>,\n): DeviceAction<ActionScrollParam> => {\n  return defineAction<typeof actionScrollParamSchema, ActionScrollParam>({\n    name: 'Scroll',\n    description:\n      'Scroll the page or an element. The direction to scroll, the scroll type, and the distance to scroll. The distance is the number of pixels to scroll. If not specified, use `down` direction, `once` scroll type, and `null` distance.',\n    interfaceAlias: 'aiScroll',\n    paramSchema: actionScrollParamSchema,\n    call,\n  });\n};\n\n// DragAndDrop\nexport const actionDragAndDropParamSchema = z.object({\n  from: getMidsceneLocationSchema().describe('The position to be dragged'),\n  to: getMidsceneLocationSchema().describe('The position to be dropped'),\n});\nexport type ActionDragAndDropParam = {\n  from: LocateResultElement;\n  to: LocateResultElement;\n};\n\nexport const defineActionDragAndDrop = (\n  call: (param: ActionDragAndDropParam) => Promise<void>,\n): DeviceAction<ActionDragAndDropParam> => {\n  return defineAction<\n    typeof actionDragAndDropParamSchema,\n    ActionDragAndDropParam\n  >({\n    name: 'DragAndDrop',\n    description: 'Drag and drop the element',\n    interfaceAlias: 'aiDragAndDrop',\n    paramSchema: actionDragAndDropParamSchema,\n    call,\n  });\n};\n\nexport const ActionLongPressParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be long pressed',\n  ),\n  duration: z\n    .number()\n    .default(500)\n    .optional()\n    .describe('Long press duration in milliseconds'),\n});\n\nexport type ActionLongPressParam = {\n  locate: LocateResultElement;\n  duration?: number;\n};\nexport const defineActionLongPress = (\n  call: (param: ActionLongPressParam) => Promise<void>,\n): DeviceAction<ActionLongPressParam> => {\n  return defineAction<typeof ActionLongPressParamSchema, ActionLongPressParam>({\n    name: 'LongPress',\n    description: 'Long press the element',\n    paramSchema: ActionLongPressParamSchema,\n    call,\n  });\n};\n\nexport const ActionSwipeParamSchema = z.object({\n  start: getMidsceneLocationSchema()\n    .optional()\n    .describe(\n      'Starting point of the swipe gesture, if not specified, the center of the page will be used',\n    ),\n  direction: z\n    .enum(['up', 'down', 'left', 'right'])\n    .optional()\n    .describe(\n      'The direction to swipe (required when using distance). The direction means the direction of the finger swipe.',\n    ),\n  distance: z\n    .number()\n    .optional()\n    .describe('The distance in pixels to swipe (mutually exclusive with end)'),\n  end: getMidsceneLocationSchema()\n    .optional()\n    .describe(\n      'Ending point of the swipe gesture (mutually exclusive with distance)',\n    ),\n  duration: z\n    .number()\n    .default(300)\n    .describe('Duration of the swipe gesture in milliseconds'),\n  repeat: z\n    .number()\n    .optional()\n    .describe(\n      'The number of times to repeat the swipe gesture. 1 for default, 0 for infinite (e.g. endless swipe until the end of the page)',\n    ),\n});\n\nexport type ActionSwipeParam = {\n  start?: LocateResultElement;\n  direction?: 'up' | 'down' | 'left' | 'right';\n  distance?: number;\n  end?: LocateResultElement;\n  duration?: number;\n  repeat?: number;\n};\n\nexport const defineActionSwipe = (\n  call: (param: ActionSwipeParam) => Promise<void>,\n): DeviceAction<ActionSwipeParam> => {\n  return defineAction<typeof ActionSwipeParamSchema, ActionSwipeParam>({\n    name: 'Swipe',\n    description:\n      'Perform a swipe gesture. You must specify either \"end\" (target location) or \"distance\" + \"direction\" - they are mutually exclusive. Use \"end\" for precise location-based swipes, or \"distance\" + \"direction\" for relative movement.',\n    paramSchema: ActionSwipeParamSchema,\n    call,\n  });\n};\n\n// ClearInput\nexport const actionClearInputParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The input field to be cleared'),\n});\nexport type ActionClearInputParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionClearInput = (\n  call: (param: ActionClearInputParam) => Promise<void>,\n): DeviceAction<ActionClearInputParam> => {\n  return defineAction<\n    typeof actionClearInputParamSchema,\n    ActionClearInputParam\n  >({\n    name: 'ClearInput',\n    description: 'Clear the text content of an input field',\n    interfaceAlias: 'aiClearInput',\n    paramSchema: actionClearInputParamSchema,\n    call,\n  });\n};\n\nexport type { DeviceAction } from '../types';\nexport type {\n  AndroidDeviceOpt,\n  AndroidDeviceInputOpt,\n  IOSDeviceOpt,\n  IOSDeviceInputOpt,\n} from './device-options';\n"],"names":["AbstractInterface","defineAction","config","actionTapParamSchema","z","getMidsceneLocationSchema","defineActionTap","call","actionRightClickParamSchema","defineActionRightClick","actionDoubleClickParamSchema","defineActionDoubleClick","actionHoverParamSchema","defineActionHover","actionInputParamSchema","val","String","defineActionInput","actionKeyboardPressParamSchema","defineActionKeyboardPress","actionScrollParamSchema","defineActionScroll","actionDragAndDropParamSchema","defineActionDragAndDrop","ActionLongPressParamSchema","defineActionLongPress","ActionSwipeParamSchema","defineActionSwipe","actionClearInputParamSchema","defineActionClearInput"],"mappings":";;AAQO,MAAeA;AAmCtB;AAKO,MAAMC,eAAe,CAK1BC,SAaOA;AAIF,MAAMC,uBAAuBC,EAAE,MAAM,CAAC;IAC3C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAMO,MAAMC,kBAAkB,CAC7BC,OAEON,aAA0D;QAC/D,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaE;QACbI;IACF;AAIK,MAAMC,8BAA8BJ,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMI,yBAAyB,CACpCF,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaO;QACbD;IACF;AAIK,MAAMG,+BAA+BN,EAAE,MAAM,CAAC;IACnD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMM,0BAA0B,CACrCJ,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaS;QACbH;IACF;AAIK,MAAMK,yBAAyBR,EAAE,MAAM,CAAC;IAC7C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMQ,oBAAoB,CAC/BN,OAEON,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaW;QACbL;IACF;AAIK,MAAMO,yBAAyBV,EAAE,MAAM,CAAC;IAC7C,OAAOA,EAAAA,KACC,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,MAAM;KAAG,EAC9B,SAAS,CAAC,CAACW,MAAQC,OAAOD,MAC1B,QAAQ,CACP;IAEJ,QAAQV,4BACL,QAAQ,CAAC,2BACT,QAAQ;IACX,MAAMD,CAAC,CAADA,OACC,CAAC;QAAC;QAAW;QAAS;KAAS,EACnC,OAAO,CAAC,WACR,QAAQ,GACR,QAAQ,CACP;AAEN;AAOO,MAAMa,oBAAoB,CAC/BV,OAEON,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaa;QACbP;IACF;AAIK,MAAMW,iCAAiCd,EAAE,MAAM,CAAC;IACrD,QAAQC,4BACL,QAAQ,CAAC,qDACT,QAAQ;IACX,SAASD,EAAAA,MACA,GACN,QAAQ,CACP;AAEN;AAMO,MAAMe,4BAA4B,CACvCZ,OAEON,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAaiB;QACbX;IACF;AAIK,MAAMa,0BAA0BhB,EAAE,MAAM,CAAC;IAC9C,WAAWA,CAAC,CAADA,OACJ,CAAC;QAAC;QAAQ;QAAM;QAAS;KAAO,EACpC,OAAO,CAAC,QACR,QAAQ,CAAC;IACZ,YAAYA,CAAC,CAADA,OACL,CAAC;QACJ;QACA;QACA;QACA;QACA;KACD,EACA,OAAO,CAAC,gBACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,CAAC;IACZ,QAAQC,4BACL,QAAQ,GACR,QAAQ,CAAC;AACd;AAaO,MAAMgB,qBAAqB,CAChCd,OAEON,aAAgE;QACrE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAamB;QACbb;IACF;AAIK,MAAMe,+BAA+BlB,EAAE,MAAM,CAAC;IACnD,MAAMC,4BAA4B,QAAQ,CAAC;IAC3C,IAAIA,4BAA4B,QAAQ,CAAC;AAC3C;AAMO,MAAMkB,0BAA0B,CACrChB,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaqB;QACbf;IACF;AAGK,MAAMiB,6BAA6BpB,EAAE,MAAM,CAAC;IACjD,QAAQC,4BAA4B,QAAQ,CAC1C;IAEF,UAAUD,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,GACR,QAAQ,CAAC;AACd;AAMO,MAAMqB,wBAAwB,CACnClB,OAEON,aAAsE;QAC3E,MAAM;QACN,aAAa;QACb,aAAauB;QACbjB;IACF;AAGK,MAAMmB,yBAAyBtB,EAAE,MAAM,CAAC;IAC7C,OAAOC,4BACJ,QAAQ,GACR,QAAQ,CACP;IAEJ,WAAWD,CAAC,CAADA,OACJ,CAAC;QAAC;QAAM;QAAQ;QAAQ;KAAQ,EACpC,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,CAAC;IACZ,KAAKC,4BACF,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUD,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,CAAC;IACZ,QAAQA,EAAAA,MACC,GACN,QAAQ,GACR,QAAQ,CACP;AAEN;AAWO,MAAMuB,oBAAoB,CAC/BpB,OAEON,aAA8D;QACnE,MAAM;QACN,aACE;QACF,aAAayB;QACbnB;IACF;AAIK,MAAMqB,8BAA8BxB,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMwB,yBAAyB,CACpCtB,OAEON,aAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAa2B;QACbrB;IACF"}