@midscene/core 0.28.11-beta-20250922065131.0 → 0.28.12-beta-20250923080328.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (41) hide show
  1. package/dist/es/agent/agent.mjs +1 -1
  2. package/dist/es/agent/agent.mjs.map +1 -1
  3. package/dist/es/agent/tasks.mjs +37 -164
  4. package/dist/es/agent/tasks.mjs.map +1 -1
  5. package/dist/es/agent/ui-utils.mjs +27 -13
  6. package/dist/es/agent/ui-utils.mjs.map +1 -1
  7. package/dist/es/agent/utils.mjs +1 -1
  8. package/dist/es/ai-model/conversation-history.mjs +58 -0
  9. package/dist/es/ai-model/conversation-history.mjs.map +1 -0
  10. package/dist/es/ai-model/index.mjs +2 -1
  11. package/dist/es/ai-model/llm-planning.mjs +23 -3
  12. package/dist/es/ai-model/llm-planning.mjs.map +1 -1
  13. package/dist/es/ai-model/ui-tars-planning.mjs +26 -6
  14. package/dist/es/ai-model/ui-tars-planning.mjs.map +1 -1
  15. package/dist/es/utils.mjs +2 -2
  16. package/dist/es/yaml/player.mjs +0 -9
  17. package/dist/es/yaml/player.mjs.map +1 -1
  18. package/dist/lib/agent/agent.js +1 -1
  19. package/dist/lib/agent/agent.js.map +1 -1
  20. package/dist/lib/agent/tasks.js +36 -163
  21. package/dist/lib/agent/tasks.js.map +1 -1
  22. package/dist/lib/agent/ui-utils.js +27 -13
  23. package/dist/lib/agent/ui-utils.js.map +1 -1
  24. package/dist/lib/agent/utils.js +1 -1
  25. package/dist/lib/ai-model/conversation-history.js +92 -0
  26. package/dist/lib/ai-model/conversation-history.js.map +1 -0
  27. package/dist/lib/ai-model/index.js +6 -2
  28. package/dist/lib/ai-model/llm-planning.js +23 -3
  29. package/dist/lib/ai-model/llm-planning.js.map +1 -1
  30. package/dist/lib/ai-model/ui-tars-planning.js +26 -6
  31. package/dist/lib/ai-model/ui-tars-planning.js.map +1 -1
  32. package/dist/lib/utils.js +2 -2
  33. package/dist/lib/yaml/player.js +0 -9
  34. package/dist/lib/yaml/player.js.map +1 -1
  35. package/dist/types/agent/tasks.d.ts +3 -18
  36. package/dist/types/agent/ui-utils.d.ts +1 -1
  37. package/dist/types/ai-model/conversation-history.d.ts +18 -0
  38. package/dist/types/ai-model/index.d.ts +1 -0
  39. package/dist/types/ai-model/llm-planning.d.ts +2 -0
  40. package/dist/types/ai-model/ui-tars-planning.d.ts +6 -18
  41. package/package.json +3 -3
@@ -37,8 +37,20 @@ function typeStr(task) {
37
37
  function locateParamStr(locate) {
38
38
  if (!locate) return '';
39
39
  if ('string' == typeof locate) return locate;
40
- if ('string' == typeof locate.prompt) return locate.prompt;
41
- if ('object' == typeof locate.prompt && locate.prompt.prompt) return locate.prompt.prompt;
40
+ if ('object' == typeof locate) {
41
+ if ('string' == typeof locate.prompt) return locate.prompt;
42
+ if ('object' == typeof locate.prompt && locate.prompt.prompt) {
43
+ const prompt = locate.prompt.prompt;
44
+ const images = locate.prompt.images || [];
45
+ if (0 === images.length) return prompt;
46
+ const imagesStr = images.map((image)=>{
47
+ let url = image.url;
48
+ if (url.startsWith('data:image/') || url.startsWith('data:') && url.includes('base64')) url = `${url.substring(0, 15)}...`;
49
+ return `[${image.name}](${url})`;
50
+ }).join(', ');
51
+ return `${prompt}, ${imagesStr}`;
52
+ }
53
+ }
42
54
  return '';
43
55
  }
44
56
  function scrollParamStr(scrollParam) {
@@ -64,27 +76,29 @@ function paramStr(task) {
64
76
  value = null == task ? void 0 : null == (_task_param = task.param) ? void 0 : _task_param.userInstruction;
65
77
  }
66
78
  if ('Insight' === task.type) {
67
- var _task_param1, _task_param2, _task_param3, _task_param4;
68
- value = (null == task ? void 0 : null == (_task_param1 = task.param) ? void 0 : _task_param1.prompt) || (null == task ? void 0 : null == (_task_param2 = task.param) ? void 0 : _task_param2.id) || (null == task ? void 0 : null == (_task_param3 = task.param) ? void 0 : _task_param3.dataDemand) || (null == task ? void 0 : null == (_task_param4 = task.param) ? void 0 : _task_param4.assertion);
79
+ var _task_param1, _task_param2, _task_param3;
80
+ value = locateParamStr(null == task ? void 0 : task.param) || (null == task ? void 0 : null == (_task_param1 = task.param) ? void 0 : _task_param1.id) || (null == task ? void 0 : null == (_task_param2 = task.param) ? void 0 : _task_param2.dataDemand) || (null == task ? void 0 : null == (_task_param3 = task.param) ? void 0 : _task_param3.assertion);
69
81
  }
70
82
  if ('Action' === task.type) {
71
- var _task_param5, _task_param6, _task_param7, _task_param8;
83
+ var _task_param4, _task_param5, _task_param6, _task_param7;
72
84
  const locate = null == task ? void 0 : task.locate;
73
85
  const locateStr = locate ? locateParamStr(locate) : '';
74
86
  value = task.thought || '';
75
- if ('number' == typeof (null == task ? void 0 : null == (_task_param5 = task.param) ? void 0 : _task_param5.timeMs)) {
87
+ if ('number' == typeof (null == task ? void 0 : null == (_task_param4 = task.param) ? void 0 : _task_param4.timeMs)) {
88
+ var _task_param8;
89
+ value = `${null == task ? void 0 : null == (_task_param8 = task.param) ? void 0 : _task_param8.timeMs}ms`;
90
+ } else if ('string' == typeof (null == task ? void 0 : null == (_task_param5 = task.param) ? void 0 : _task_param5.scrollType)) value = scrollParamStr(null == task ? void 0 : task.param);
91
+ else if ('string' == typeof (null == task ? void 0 : null == (_task_param6 = task.param) ? void 0 : _task_param6.direction) && (null == task ? void 0 : task.subType) === 'AndroidPull') value = pullParamStr(null == task ? void 0 : task.param);
92
+ else if (void 0 !== (null == task ? void 0 : null == (_task_param7 = task.param) ? void 0 : _task_param7.value)) {
76
93
  var _task_param9;
77
- value = `${null == task ? void 0 : null == (_task_param9 = task.param) ? void 0 : _task_param9.timeMs}ms`;
78
- } else if ('string' == typeof (null == task ? void 0 : null == (_task_param6 = task.param) ? void 0 : _task_param6.scrollType)) value = scrollParamStr(null == task ? void 0 : task.param);
79
- else if ('string' == typeof (null == task ? void 0 : null == (_task_param7 = task.param) ? void 0 : _task_param7.direction) && (null == task ? void 0 : task.subType) === 'AndroidPull') value = pullParamStr(null == task ? void 0 : task.param);
80
- else if (void 0 !== (null == task ? void 0 : null == (_task_param8 = task.param) ? void 0 : _task_param8.value)) {
81
- var _task_param10;
82
- value = null == task ? void 0 : null == (_task_param10 = task.param) ? void 0 : _task_param10.value;
94
+ value = null == task ? void 0 : null == (_task_param9 = task.param) ? void 0 : _task_param9.value;
83
95
  }
84
96
  if (locateStr) value = value ? `${locateStr} - ${value}` : locateStr;
85
97
  }
86
98
  if (void 0 === value) return '';
87
- return 'string' == typeof value ? value : JSON.stringify(value, void 0, 2);
99
+ if ('string' == typeof value) return value;
100
+ if ('object' == typeof value && locateParamStr(value)) return locateParamStr(value);
101
+ return JSON.stringify(value, void 0, 2);
88
102
  }
89
103
  exports.locateParamStr = __webpack_exports__.locateParamStr;
90
104
  exports.paramStr = __webpack_exports__.paramStr;
@@ -1 +1 @@
1
- {"version":3,"file":"agent/ui-utils.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/agent/ui-utils.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n AndroidPullParam,\n DetailedLocateParam,\n ExecutionTask,\n ExecutionTaskAction,\n ExecutionTaskInsightAssertion,\n ExecutionTaskInsightLocate,\n ExecutionTaskInsightQuery,\n ExecutionTaskPlanning,\n ScrollParam,\n} from '@/index';\n\nexport function typeStr(task: ExecutionTask) {\n return task.subType && task.subType !== 'Plan'\n ? `${task.type} / ${task.subType || ''}`\n : task.type;\n}\n\nexport function locateParamStr(locate?: DetailedLocateParam) {\n if (!locate) {\n return '';\n }\n\n if (typeof locate === 'string') {\n return locate;\n }\n\n if (typeof locate.prompt === 'string') {\n return locate.prompt;\n }\n\n if (typeof locate.prompt === 'object' && locate.prompt.prompt) {\n return locate.prompt.prompt;\n }\n\n return '';\n}\n\nexport function scrollParamStr(scrollParam?: ScrollParam) {\n if (!scrollParam) {\n return '';\n }\n return `${scrollParam.direction || 'down'}, ${scrollParam.scrollType || 'once'}, ${scrollParam.distance || 'distance-not-set'}`;\n}\n\nexport function pullParamStr(pullParam?: AndroidPullParam) {\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} - ${prompt}`;\n }\n return type;\n}\n\nexport function paramStr(task: ExecutionTask) {\n let value: string | undefined | object;\n if (task.type === 'Planning') {\n value = (task as ExecutionTaskPlanning)?.param?.userInstruction;\n }\n\n if (task.type === 'Insight') {\n value =\n (task as ExecutionTaskInsightLocate)?.param?.prompt ||\n (task as ExecutionTaskInsightLocate)?.param?.id ||\n (task as ExecutionTaskInsightQuery)?.param?.dataDemand ||\n (task as ExecutionTaskInsightAssertion)?.param?.assertion;\n }\n\n if (task.type === 'Action') {\n const locate = (task as ExecutionTaskAction)?.locate;\n const locateStr = locate ? locateParamStr(locate) : '';\n\n value = task.thought || '';\n if (typeof (task as ExecutionTaskAction)?.param?.timeMs === 'number') {\n value = `${(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 }\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 return typeof value === 'string'\n ? value\n : JSON.stringify(value, undefined, 2);\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","typeStr","task","locateParamStr","locate","scrollParamStr","scrollParam","pullParamStr","pullParam","parts","taskTitleStr","type","prompt","paramStr","value","_task_param","_task_param1","_task_param2","_task_param3","_task_param4","_task_param5","_task_param6","_task_param7","_task_param8","locateStr","_task_param9","_task_param10","JSON","undefined"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;ACMO,SAASI,QAAQC,IAAmB;IACzC,OAAOA,KAAK,OAAO,IAAIA,AAAiB,WAAjBA,KAAK,OAAO,GAC/B,GAAGA,KAAK,IAAI,CAAC,GAAG,EAAEA,KAAK,OAAO,IAAI,IAAI,GACtCA,KAAK,IAAI;AACf;AAEO,SAASC,eAAeC,MAA4B;IACzD,IAAI,CAACA,QACH,OAAO;IAGT,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAGT,IAAI,AAAyB,YAAzB,OAAOA,OAAO,MAAM,EACtB,OAAOA,OAAO,MAAM;IAGtB,IAAI,AAAyB,YAAzB,OAAOA,OAAO,MAAM,IAAiBA,OAAO,MAAM,CAAC,MAAM,EAC3D,OAAOA,OAAO,MAAM,CAAC,MAAM;IAG7B,OAAO;AACT;AAEO,SAASC,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,SAA4B;IACvD,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,EACZC,MAAc;IAEd,IAAIA,QACF,OAAO,GAAGD,KAAK,GAAG,EAAEC,QAAQ;IAE9B,OAAOD;AACT;AAEO,SAASE,SAASX,IAAmB;IAC1C,IAAIY;IACJ,IAAIZ,AAAc,eAAdA,KAAK,IAAI,EAAiB;YACpBa;QAARD,QAASZ,QAAAA,OAAAA,KAAAA,IAAAA,QAADa,CAAAA,cAACb,KAAgC,KAAK,AAAD,IAArCa,KAAAA,IAAAA,YAAwC,eAAe;IACjE;IAEA,IAAIb,AAAc,cAAdA,KAAK,IAAI,EAAgB;YAEzBc,cACAC,cACAC,cACAC;QAJFL,QACEE,AAACd,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADc,CAAAA,eAACd,KAAqC,KAAK,AAAD,IAA1Cc,KAAAA,IAAAA,aAA6C,MAAM,AAAD,KACjDd,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADe,CAAAA,eAACf,KAAqC,KAAK,AAAD,IAA1Ce,KAAAA,IAAAA,aAA6C,EAAE,AAAD,KAC7Cf,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADgB,CAAAA,eAAChB,KAAoC,KAAK,AAAD,IAAzCgB,KAAAA,IAAAA,aAA4C,UAAU,AAAD,KACpDhB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADiB,CAAAA,eAACjB,KAAwC,KAAK,AAAD,IAA7CiB,KAAAA,IAAAA,aAAgD,SAAS,AAAD;IAC5D;IAEA,IAAIjB,AAAc,aAAdA,KAAK,IAAI,EAAe;YAKfkB,cAGFC,cAIAC,cAKAC;QAhBT,MAAMnB,SAAUF,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,MAAM;QACpD,MAAMsB,YAAYpB,SAASD,eAAeC,UAAU;QAEpDU,QAAQZ,KAAK,OAAO,IAAI;QACxB,IAAI,AAAwD,YAAxD,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADkB,CAAAA,eAAClB,KAA8B,KAAK,AAAD,IAAnCkB,KAAAA,IAAAA,aAAsC,MAAM,AAAD,GAAgB;gBACzDK;YAAXX,QAAQ,GAAIZ,QAAAA,OAAAA,KAAAA,IAAAA,QAADuB,CAAAA,eAACvB,KAA8B,KAAK,AAAD,IAAnCuB,KAAAA,IAAAA,aAAsC,MAAM,CAAC,EAAE,CAAC;QAC7D,OAAO,IACL,AAA4D,YAA5D,OAAQvB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADmB,CAAAA,eAACnB,KAA8B,KAAK,AAAD,IAAnCmB,KAAAA,IAAAA,aAAsC,UAAU,AAAD,GAEtDP,QAAQT,eAAgBH,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;aACtD,IACL,AAA2D,YAA3D,OAAQA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADoB,CAAAA,eAACpB,KAA8B,KAAK,AAAD,IAAnCoB,KAAAA,IAAAA,aAAsC,SAAS,AAAD,KACrD,AAACpB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,OAAO,AAAD,MAAM,eAE3CY,QAAQP,aAAcL,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,KAAK;aACpD,IACL,AAAuD,WAA/CA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADqB,CAAAA,eAACrB,KAA8B,KAAK,AAAD,IAAnCqB,KAAAA,IAAAA,aAAsC,KAAK,AAAD,GACjD;gBACQG;YAARZ,QAASZ,QAAAA,OAAAA,KAAAA,IAAAA,QAADwB,CAAAA,gBAACxB,KAA8B,KAAK,AAAD,IAAnCwB,KAAAA,IAAAA,cAAsC,KAAK;QACrD;QAEA,IAAIF,WAEAV,QADEA,QACM,GAAGU,UAAU,GAAG,EAAEV,OAAO,GAEzBU;IAGd;IAEA,IAAI,AAAiB,WAAVV,OAAuB,OAAO;IACzC,OAAO,AAAiB,YAAjB,OAAOA,QACVA,QACAa,KAAK,SAAS,CAACb,OAAOc,QAAW;AACvC"}
1
+ {"version":3,"file":"agent/ui-utils.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/agent/ui-utils.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n AndroidPullParam,\n DetailedLocateParam,\n ExecutionTask,\n ExecutionTaskAction,\n ExecutionTaskInsightAssertion,\n ExecutionTaskInsightLocate,\n ExecutionTaskInsightQuery,\n ExecutionTaskPlanning,\n ScrollParam,\n} from '@/index';\n\nexport function typeStr(task: ExecutionTask) {\n return task.subType && task.subType !== 'Plan'\n ? `${task.type} / ${task.subType || ''}`\n : task.type;\n}\n\nexport function locateParamStr(locate?: DetailedLocateParam | string): string {\n if (!locate) {\n return '';\n }\n\n if (typeof locate === 'string') {\n return locate;\n }\n\n if (typeof locate === 'object') {\n if (typeof locate.prompt === 'string') {\n return locate.prompt;\n }\n\n if (typeof locate.prompt === 'object' && locate.prompt.prompt) {\n const prompt = locate.prompt.prompt;\n const images = locate.prompt.images || [];\n\n if (images.length === 0) return prompt;\n\n const imagesStr = images\n .map((image) => {\n let url = image.url;\n if (\n url.startsWith('data:image/') ||\n (url.startsWith('data:') && url.includes('base64'))\n ) {\n url = `${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?: AndroidPullParam) {\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} - ${prompt}`;\n }\n return type;\n}\n\nexport function paramStr(task: ExecutionTask) {\n let value: string | undefined | object;\n if (task.type === 'Planning') {\n value = (task as ExecutionTaskPlanning)?.param?.userInstruction;\n }\n\n if (task.type === 'Insight') {\n value =\n locateParamStr((task as ExecutionTaskInsightLocate)?.param) ||\n (task as ExecutionTaskInsightLocate)?.param?.id ||\n (task as ExecutionTaskInsightQuery)?.param?.dataDemand ||\n (task as ExecutionTaskInsightAssertion)?.param?.assertion;\n }\n\n if (task.type === 'Action') {\n const locate = (task as ExecutionTaskAction)?.locate;\n const locateStr = locate ? locateParamStr(locate) : '';\n\n value = task.thought || '';\n if (typeof (task as ExecutionTaskAction)?.param?.timeMs === 'number') {\n value = `${(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 }\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":["__webpack_require__","definition","key","Object","obj","prop","Symbol","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","JSON","undefined"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;ACMO,SAASI,QAAQC,IAAmB;IACzC,OAAOA,KAAK,OAAO,IAAIA,AAAiB,WAAjBA,KAAK,OAAO,GAC/B,GAAGA,KAAK,IAAI,CAAC,GAAG,EAAEA,KAAK,OAAO,IAAI,IAAI,GACtCA,KAAK,IAAI;AACf;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,SAA4B;IACvD,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,EAAiB;YACpBiB;QAARD,QAAShB,QAAAA,OAAAA,KAAAA,IAAAA,QAADiB,CAAAA,cAACjB,KAAgC,KAAK,AAAD,IAArCiB,KAAAA,IAAAA,YAAwC,eAAe;IACjE;IAEA,IAAIjB,AAAc,cAAdA,KAAK,IAAI,EAAgB;YAGzBkB,cACAC,cACAC;QAJFJ,QACEf,eAAgBD,QAAAA,OAAAA,KAAAA,IAAAA,KAAqC,KAAK,KACzDA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADkB,CAAAA,eAAClB,KAAqC,KAAK,AAAD,IAA1CkB,KAAAA,IAAAA,aAA6C,EAAE,AAAD,KAC7ClB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADmB,CAAAA,eAACnB,KAAoC,KAAK,AAAD,IAAzCmB,KAAAA,IAAAA,aAA4C,UAAU,AAAD,KACpDnB,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,QAADoB,CAAAA,eAACpB,KAAwC,KAAK,AAAD,IAA7CoB,KAAAA,IAAAA,aAAgD,SAAS,AAAD;IAC5D;IAEA,IAAIpB,AAAc,aAAdA,KAAK,IAAI,EAAe;YAKfqB,cAGFC,cAIAC,cAKAC;QAhBT,MAAMtB,SAAUF,QAAAA,OAAAA,KAAAA,IAAAA,KAA8B,MAAM;QACpD,MAAMyB,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;QAEA,IAAIF,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,OAAOY,KAAK,SAAS,CAACZ,OAAOa,QAAW;AAC1C"}
@@ -186,7 +186,7 @@ function trimContextByViewport(execution) {
186
186
  }) : execution.tasks
187
187
  };
188
188
  }
189
- const getMidsceneVersion = ()=>"0.28.11-beta-20250922065131.0";
189
+ const getMidsceneVersion = ()=>"0.28.12-beta-20250923080328.0";
190
190
  const parsePrompt = (prompt)=>{
191
191
  if ('string' == typeof prompt) return {
192
192
  textPrompt: prompt,
@@ -0,0 +1,92 @@
1
+ "use strict";
2
+ var __webpack_require__ = {};
3
+ (()=>{
4
+ __webpack_require__.d = (exports1, definition)=>{
5
+ for(var key in definition)if (__webpack_require__.o(definition, key) && !__webpack_require__.o(exports1, key)) Object.defineProperty(exports1, key, {
6
+ enumerable: true,
7
+ get: definition[key]
8
+ });
9
+ };
10
+ })();
11
+ (()=>{
12
+ __webpack_require__.o = (obj, prop)=>Object.prototype.hasOwnProperty.call(obj, prop);
13
+ })();
14
+ (()=>{
15
+ __webpack_require__.r = (exports1)=>{
16
+ if ('undefined' != typeof Symbol && Symbol.toStringTag) Object.defineProperty(exports1, Symbol.toStringTag, {
17
+ value: 'Module'
18
+ });
19
+ Object.defineProperty(exports1, '__esModule', {
20
+ value: true
21
+ });
22
+ };
23
+ })();
24
+ var __webpack_exports__ = {};
25
+ __webpack_require__.r(__webpack_exports__);
26
+ __webpack_require__.d(__webpack_exports__, {
27
+ ConversationHistory: ()=>ConversationHistory
28
+ });
29
+ function _define_property(obj, key, value) {
30
+ if (key in obj) Object.defineProperty(obj, key, {
31
+ value: value,
32
+ enumerable: true,
33
+ configurable: true,
34
+ writable: true
35
+ });
36
+ else obj[key] = value;
37
+ return obj;
38
+ }
39
+ var _computedKey;
40
+ _computedKey = Symbol.iterator;
41
+ let _computedKey1 = _computedKey;
42
+ class ConversationHistory {
43
+ append(message) {
44
+ if ('user' === message.role) this.pruneOldestUserMessageIfNecessary();
45
+ this.messages.push(message);
46
+ }
47
+ seed(messages) {
48
+ this.reset();
49
+ messages.forEach((message)=>{
50
+ this.append(message);
51
+ });
52
+ }
53
+ reset() {
54
+ this.messages.length = 0;
55
+ }
56
+ snapshot() {
57
+ return [
58
+ ...this.messages
59
+ ];
60
+ }
61
+ get length() {
62
+ return this.messages.length;
63
+ }
64
+ [_computedKey1]() {
65
+ return this.messages[Symbol.iterator]();
66
+ }
67
+ toJSON() {
68
+ return this.snapshot();
69
+ }
70
+ pruneOldestUserMessageIfNecessary() {
71
+ const userMessages = this.messages.filter((item)=>'user' === item.role);
72
+ if (userMessages.length < this.maxUserImageMessages) return;
73
+ const firstUserMessageIndex = this.messages.findIndex((item)=>'user' === item.role);
74
+ if (firstUserMessageIndex >= 0) this.messages.splice(firstUserMessageIndex, 1);
75
+ }
76
+ constructor(options){
77
+ var _options_initialMessages;
78
+ _define_property(this, "maxUserImageMessages", void 0);
79
+ _define_property(this, "messages", []);
80
+ this.maxUserImageMessages = (null == options ? void 0 : options.maxUserImageMessages) ?? 4;
81
+ if (null == options ? void 0 : null == (_options_initialMessages = options.initialMessages) ? void 0 : _options_initialMessages.length) this.seed(options.initialMessages);
82
+ }
83
+ }
84
+ exports.ConversationHistory = __webpack_exports__.ConversationHistory;
85
+ for(var __webpack_i__ in __webpack_exports__)if (-1 === [
86
+ "ConversationHistory"
87
+ ].indexOf(__webpack_i__)) exports[__webpack_i__] = __webpack_exports__[__webpack_i__];
88
+ Object.defineProperty(exports, '__esModule', {
89
+ value: true
90
+ });
91
+
92
+ //# sourceMappingURL=conversation-history.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"ai-model/conversation-history.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/conversation-history.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type { ChatCompletionMessageParam } from 'openai/resources/index';\n\nexport interface ConversationHistoryOptions {\n maxUserImageMessages?: number;\n initialMessages?: ChatCompletionMessageParam[];\n}\n\nexport class ConversationHistory {\n private readonly maxUserImageMessages: number;\n private readonly messages: ChatCompletionMessageParam[] = [];\n\n constructor(options?: ConversationHistoryOptions) {\n this.maxUserImageMessages = options?.maxUserImageMessages ?? 4;\n if (options?.initialMessages?.length) {\n this.seed(options.initialMessages);\n }\n }\n\n append(message: ChatCompletionMessageParam) {\n if (message.role === 'user') {\n this.pruneOldestUserMessageIfNecessary();\n }\n\n this.messages.push(message);\n }\n\n seed(messages: ChatCompletionMessageParam[]) {\n this.reset();\n messages.forEach((message) => {\n this.append(message);\n });\n }\n\n reset() {\n this.messages.length = 0;\n }\n\n snapshot(): ChatCompletionMessageParam[] {\n return [...this.messages];\n }\n\n get length(): number {\n return this.messages.length;\n }\n\n [Symbol.iterator](): IterableIterator<ChatCompletionMessageParam> {\n return this.messages[Symbol.iterator]();\n }\n\n toJSON(): ChatCompletionMessageParam[] {\n return this.snapshot();\n }\n\n private pruneOldestUserMessageIfNecessary() {\n const userMessages = this.messages.filter((item) => item.role === 'user');\n if (userMessages.length < this.maxUserImageMessages) {\n return;\n }\n\n const firstUserMessageIndex = this.messages.findIndex(\n (item) => item.role === 'user',\n );\n\n if (firstUserMessageIndex >= 0) {\n this.messages.splice(firstUserMessageIndex, 1);\n }\n }\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","ConversationHistory","message","messages","userMessages","item","firstUserMessageIndex","options","_options_initialMessages"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;eCuCGG,OAAO,QAAQ;;AAtCX,MAAMC;IAWX,OAAOC,OAAmC,EAAE;QAC1C,IAAIA,AAAiB,WAAjBA,QAAQ,IAAI,EACd,IAAI,CAAC,iCAAiC;QAGxC,IAAI,CAAC,QAAQ,CAAC,IAAI,CAACA;IACrB;IAEA,KAAKC,QAAsC,EAAE;QAC3C,IAAI,CAAC,KAAK;QACVA,SAAS,OAAO,CAAC,CAACD;YAChB,IAAI,CAAC,MAAM,CAACA;QACd;IACF;IAEA,QAAQ;QACN,IAAI,CAAC,QAAQ,CAAC,MAAM,GAAG;IACzB;IAEA,WAAyC;QACvC,OAAO;eAAI,IAAI,CAAC,QAAQ;SAAC;IAC3B;IAEA,IAAI,SAAiB;QACnB,OAAO,IAAI,CAAC,QAAQ,CAAC,MAAM;IAC7B;IAEA,CAAC,cAAD,GAAkE;QAChE,OAAO,IAAI,CAAC,QAAQ,CAACF,OAAO,QAAQ,CAAC;IACvC;IAEA,SAAuC;QACrC,OAAO,IAAI,CAAC,QAAQ;IACtB;IAEQ,oCAAoC;QAC1C,MAAMI,eAAe,IAAI,CAAC,QAAQ,CAAC,MAAM,CAAC,CAACC,OAASA,AAAc,WAAdA,KAAK,IAAI;QAC7D,IAAID,aAAa,MAAM,GAAG,IAAI,CAAC,oBAAoB,EACjD;QAGF,MAAME,wBAAwB,IAAI,CAAC,QAAQ,CAAC,SAAS,CACnD,CAACD,OAASA,AAAc,WAAdA,KAAK,IAAI;QAGrB,IAAIC,yBAAyB,GAC3B,IAAI,CAAC,QAAQ,CAAC,MAAM,CAACA,uBAAuB;IAEhD;IAvDA,YAAYC,OAAoC,CAAE;YAE5CC;QALN,uBAAiB,wBAAjB;QACA,uBAAiB,YAAyC,EAAE;QAG1D,IAAI,CAAC,oBAAoB,GAAGD,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,oBAAoB,AAAD,KAAK;QAC7D,IAAIC,QAAAA,UAAAA,KAAAA,IAAAA,QAAAA,CAAAA,2BAAAA,QAAS,eAAe,AAAD,IAAvBA,KAAAA,IAAAA,yBAA0B,MAAM,EAClC,IAAI,CAAC,IAAI,CAACD,QAAQ,eAAe;IAErC;AAmDF"}
@@ -32,19 +32,20 @@ __webpack_require__.d(__webpack_exports__, {
32
32
  AiLocateSection: ()=>external_inspect_js_namespaceObject.AiLocateSection,
33
33
  AIActionType: ()=>external_common_js_namespaceObject.AIActionType,
34
34
  TUserPromptSchema: ()=>external_common_js_namespaceObject.TUserPromptSchema,
35
+ ConversationHistory: ()=>external_conversation_history_js_namespaceObject.ConversationHistory,
35
36
  callAI: ()=>index_js_namespaceObject.callAI,
36
- findAllMidsceneLocatorField: ()=>external_common_js_namespaceObject.findAllMidsceneLocatorField,
37
37
  AiExtractElementInfo: ()=>external_inspect_js_namespaceObject.AiExtractElementInfo,
38
+ findAllMidsceneLocatorField: ()=>external_common_js_namespaceObject.findAllMidsceneLocatorField,
38
39
  generatePlaywrightTest: ()=>playwright_generator_js_namespaceObject.generatePlaywrightTest,
39
40
  generateYamlTestStream: ()=>yaml_generator_js_namespaceObject.generateYamlTestStream,
40
41
  loadActionParam: ()=>external_common_js_namespaceObject.loadActionParam,
41
42
  resizeImageForUiTars: ()=>external_ui_tars_planning_js_namespaceObject.resizeImageForUiTars,
42
43
  systemPromptToLocateElement: ()=>llm_locator_js_namespaceObject.systemPromptToLocateElement,
43
- vlmPlanning: ()=>external_ui_tars_planning_js_namespaceObject.vlmPlanning,
44
44
  RectSchema: ()=>external_common_js_namespaceObject.RectSchema,
45
45
  PointSchema: ()=>external_common_js_namespaceObject.PointSchema,
46
46
  adaptBboxToRect: ()=>external_common_js_namespaceObject.adaptBboxToRect,
47
47
  callAIWithObjectResponse: ()=>index_js_namespaceObject.callAIWithObjectResponse,
48
+ vlmPlanning: ()=>external_ui_tars_planning_js_namespaceObject.vlmPlanning,
48
49
  AiLocateElement: ()=>external_inspect_js_namespaceObject.AiLocateElement,
49
50
  getMidsceneLocationSchema: ()=>external_common_js_namespaceObject.getMidsceneLocationSchema,
50
51
  callAIWithStringResponse: ()=>index_js_namespaceObject.callAIWithStringResponse,
@@ -61,10 +62,12 @@ const external_inspect_js_namespaceObject = require("./inspect.js");
61
62
  const external_llm_planning_js_namespaceObject = require("./llm-planning.js");
62
63
  const external_common_js_namespaceObject = require("./common.js");
63
64
  const external_ui_tars_planning_js_namespaceObject = require("./ui-tars-planning.js");
65
+ const external_conversation_history_js_namespaceObject = require("./conversation-history.js");
64
66
  exports.AIActionType = __webpack_exports__.AIActionType;
65
67
  exports.AiExtractElementInfo = __webpack_exports__.AiExtractElementInfo;
66
68
  exports.AiLocateElement = __webpack_exports__.AiLocateElement;
67
69
  exports.AiLocateSection = __webpack_exports__.AiLocateSection;
70
+ exports.ConversationHistory = __webpack_exports__.ConversationHistory;
68
71
  exports.PointSchema = __webpack_exports__.PointSchema;
69
72
  exports.RectSchema = __webpack_exports__.RectSchema;
70
73
  exports.SizeSchema = __webpack_exports__.SizeSchema;
@@ -93,6 +96,7 @@ for(var __webpack_i__ in __webpack_exports__)if (-1 === [
93
96
  "AiExtractElementInfo",
94
97
  "AiLocateElement",
95
98
  "AiLocateSection",
99
+ "ConversationHistory",
96
100
  "PointSchema",
97
101
  "RectSchema",
98
102
  "SizeSchema",
@@ -35,8 +35,8 @@ const util_js_namespaceObject = require("./prompt/util.js");
35
35
  const index_js_namespaceObject = require("./service-caller/index.js");
36
36
  const debug = (0, logger_namespaceObject.getDebug)('planning');
37
37
  async function plan(userInstruction, opts) {
38
- var _planFromAI_action;
39
- const { context, modelConfig } = opts;
38
+ var _opts_conversationHistory, _planFromAI_action;
39
+ const { context, modelConfig, conversationHistory } = opts;
40
40
  const { screenshotBase64, size } = context;
41
41
  const { modelName, vlMode } = modelConfig;
42
42
  const { description: pageDescription, elementById } = await (0, util_js_namespaceObject.describeUserPage)(context, {
@@ -46,7 +46,9 @@ async function plan(userInstruction, opts) {
46
46
  actionSpace: opts.actionSpace,
47
47
  vlMode: vlMode
48
48
  });
49
- const taskBackgroundContextText = (0, llm_planning_js_namespaceObject.generateTaskBackgroundContext)(userInstruction, opts.log, opts.actionContext);
49
+ const historyLog = getLogFromConversationHistory(null == (_opts_conversationHistory = opts.conversationHistory) ? void 0 : _opts_conversationHistory.snapshot());
50
+ const logForContext = historyLog ?? opts.log;
51
+ const taskBackgroundContextText = (0, llm_planning_js_namespaceObject.generateTaskBackgroundContext)(userInstruction, logForContext, opts.actionContext);
50
52
  const userInstructionPrompt = await (0, llm_planning_js_namespaceObject.automationUserPrompt)(vlMode).format({
51
53
  pageDescription,
52
54
  taskBackgroundContext: taskBackgroundContextText
@@ -108,8 +110,26 @@ async function plan(userInstruction, opts) {
108
110
  });
109
111
  (0, utils_namespaceObject.assert)(!planFromAI.error, `Failed to plan actions: ${planFromAI.error}`);
110
112
  if (0 === actions.length && returnValue.more_actions_needed_by_instruction && !returnValue.sleep) console.warn('No actions planned for the prompt, but model said more actions are needed:', userInstruction);
113
+ null == conversationHistory || conversationHistory.append({
114
+ role: 'assistant',
115
+ content: returnValue.log
116
+ });
111
117
  return returnValue;
112
118
  }
119
+ function getLogFromConversationHistory(conversationHistory) {
120
+ if (!(null == conversationHistory ? void 0 : conversationHistory.length)) return;
121
+ const assistantMessages = conversationHistory.filter((message)=>'assistant' === message.role).map((message)=>{
122
+ if ('string' == typeof message.content) return message.content.trim();
123
+ if (Array.isArray(message.content)) return message.content.map((part)=>{
124
+ if ('text' === part.type && 'string' == typeof part.text) return part.text.trim();
125
+ if ('string' == typeof part.text) return part.text.trim();
126
+ return '';
127
+ }).filter(Boolean).join('\n');
128
+ return '';
129
+ }).map((content)=>content.trim()).filter(Boolean);
130
+ if (0 === assistantMessages.length) return;
131
+ return `- ${assistantMessages.join('\n- ')}`;
132
+ }
113
133
  exports.plan = __webpack_exports__.plan;
114
134
  for(var __webpack_i__ in __webpack_exports__)if (-1 === [
115
135
  "plan"
@@ -1 +1 @@
1
- {"version":3,"file":"ai-model/llm-planning.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/llm-planning.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n DeviceAction,\n InterfaceType,\n PlanningAIResponse,\n UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport {\n AIActionType,\n type AIArgs,\n buildYamlFlowFromPlans,\n fillBboxParam,\n findAllMidsceneLocatorField,\n markupImageForLLM,\n warnGPT4oSizeLimit,\n} from './common';\nimport {\n automationUserPrompt,\n generateTaskBackgroundContext,\n systemPromptToTaskPlanning,\n} from './prompt/llm-planning';\nimport { describeUserPage } from './prompt/util';\nimport { callAIWithObjectResponse } from './service-caller/index';\n\nconst debug = getDebug('planning');\n\nexport async function plan(\n userInstruction: string,\n opts: {\n context: UIContext;\n interfaceType: InterfaceType;\n actionSpace: DeviceAction<any>[];\n log?: string;\n actionContext?: string;\n modelConfig: IModelConfig;\n },\n): Promise<PlanningAIResponse> {\n const { context, modelConfig } = opts;\n const { screenshotBase64, size } = context;\n\n const { modelName, vlMode } = modelConfig;\n\n const { description: pageDescription, elementById } = await describeUserPage(\n context,\n { vlMode },\n );\n const systemPrompt = await systemPromptToTaskPlanning({\n actionSpace: opts.actionSpace,\n vlMode: vlMode,\n });\n const taskBackgroundContextText = generateTaskBackgroundContext(\n userInstruction,\n opts.log,\n opts.actionContext,\n );\n const userInstructionPrompt = await automationUserPrompt(vlMode).format({\n pageDescription,\n taskBackgroundContext: taskBackgroundContextText,\n });\n\n let imagePayload = screenshotBase64;\n if (vlMode === 'qwen-vl') {\n imagePayload = await paddingToMatchBlockByBase64(imagePayload);\n } else if (!vlMode) {\n imagePayload = await markupImageForLLM(\n screenshotBase64,\n context.tree,\n context.size,\n );\n }\n\n warnGPT4oSizeLimit(size, modelName);\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n detail: 'high',\n },\n },\n {\n type: 'text',\n text: userInstructionPrompt,\n },\n ],\n },\n ];\n\n const { content, usage } = await callAIWithObjectResponse<PlanningAIResponse>(\n msgs,\n AIActionType.PLAN,\n modelConfig,\n );\n const rawResponse = JSON.stringify(content, undefined, 2);\n const planFromAI = content;\n\n const actions =\n (planFromAI.action?.type ? [planFromAI.action] : planFromAI.actions) || [];\n const returnValue: PlanningAIResponse = {\n ...planFromAI,\n actions,\n rawResponse,\n usage,\n yamlFlow: buildYamlFlowFromPlans(\n actions,\n opts.actionSpace,\n planFromAI.sleep,\n ),\n };\n\n assert(planFromAI, \"can't get plans from AI\");\n\n // TODO: use zod.parse to parse the action.param, and then fill the bbox param.\n actions.forEach((action) => {\n const type = action.type;\n const actionInActionSpace = opts.actionSpace.find(\n (action) => action.name === type,\n );\n\n debug('actionInActionSpace matched', actionInActionSpace);\n const locateFields = actionInActionSpace\n ? findAllMidsceneLocatorField(actionInActionSpace.paramSchema)\n : [];\n\n debug('locateFields', locateFields);\n\n locateFields.forEach((field) => {\n const locateResult = action.param[field];\n if (locateResult) {\n if (vlMode) {\n action.param[field] = fillBboxParam(\n locateResult,\n size.width,\n size.height,\n vlMode,\n );\n } else {\n const element = elementById(locateResult);\n if (element) {\n action.param[field].id = element.id;\n }\n }\n }\n });\n });\n // in Qwen-VL, error means error. In GPT-4o, error may mean more actions are needed.\n assert(!planFromAI.error, `Failed to plan actions: ${planFromAI.error}`);\n\n if (\n actions.length === 0 &&\n returnValue.more_actions_needed_by_instruction &&\n !returnValue.sleep\n ) {\n console.warn(\n 'No actions planned for the prompt, but model said more actions are needed:',\n userInstruction,\n );\n }\n\n return returnValue;\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","plan","userInstruction","opts","_planFromAI_action","context","modelConfig","screenshotBase64","size","modelName","vlMode","pageDescription","elementById","describeUserPage","systemPrompt","systemPromptToTaskPlanning","taskBackgroundContextText","generateTaskBackgroundContext","userInstructionPrompt","automationUserPrompt","imagePayload","paddingToMatchBlockByBase64","markupImageForLLM","warnGPT4oSizeLimit","msgs","content","usage","callAIWithObjectResponse","AIActionType","rawResponse","JSON","undefined","planFromAI","actions","returnValue","buildYamlFlowFromPlans","assert","action","type","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam","element","console"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;ACqBA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAEhB,eAAeC,KACpBC,eAAuB,EACvBC,IAOC;QAmEEC;IAjEH,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAE,GAAGH;IACjC,MAAM,EAAEI,gBAAgB,EAAEC,IAAI,EAAE,GAAGH;IAEnC,MAAM,EAAEI,SAAS,EAAEC,MAAM,EAAE,GAAGJ;IAE9B,MAAM,EAAE,aAAaK,eAAe,EAAEC,WAAW,EAAE,GAAG,MAAMC,AAAAA,IAAAA,wBAAAA,gBAAAA,AAAAA,EAC1DR,SACA;QAAEK;IAAO;IAEX,MAAMI,eAAe,MAAMC,AAAAA,IAAAA,gCAAAA,0BAAAA,AAAAA,EAA2B;QACpD,aAAaZ,KAAK,WAAW;QAC7B,QAAQO;IACV;IACA,MAAMM,4BAA4BC,AAAAA,IAAAA,gCAAAA,6BAAAA,AAAAA,EAChCf,iBACAC,KAAK,GAAG,EACRA,KAAK,aAAa;IAEpB,MAAMe,wBAAwB,MAAMC,AAAAA,IAAAA,gCAAAA,oBAAAA,AAAAA,EAAqBT,QAAQ,MAAM,CAAC;QACtEC;QACA,uBAAuBK;IACzB;IAEA,IAAII,eAAeb;IACnB,IAAIG,AAAW,cAAXA,QACFU,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,2BAAAA,AAAAA,EAA4BD;SAC5C,IAAI,CAACV,QACVU,eAAe,MAAME,AAAAA,IAAAA,mCAAAA,iBAAAA,AAAAA,EACnBf,kBACAF,QAAQ,IAAI,EACZA,QAAQ,IAAI;IAIhBkB,IAAAA,mCAAAA,kBAAAA,AAAAA,EAAmBf,MAAMC;IAEzB,MAAMe,OAAe;QACnB;YAAE,MAAM;YAAU,SAASV;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKM;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMF;gBACR;aACD;QACH;KACD;IAED,MAAM,EAAEO,OAAO,EAAEC,KAAK,EAAE,GAAG,MAAMC,AAAAA,IAAAA,yBAAAA,wBAAAA,AAAAA,EAC/BH,MACAI,mCAAAA,YAAAA,CAAAA,IAAiB,EACjBtB;IAEF,MAAMuB,cAAcC,KAAK,SAAS,CAACL,SAASM,QAAW;IACvD,MAAMC,aAAaP;IAEnB,MAAMQ,UACH7B,AAAAA,CAAAA,SAAAA,CAAAA,qBAAAA,WAAW,MAAM,AAAD,IAAhBA,KAAAA,IAAAA,mBAAmB,IAAI,AAAD,IAAI;QAAC4B,WAAW,MAAM;KAAC,GAAGA,WAAW,OAAM,KAAM,EAAE;IAC5E,MAAME,cAAkC;QACtC,GAAGF,UAAU;QACbC;QACAJ;QACAH;QACA,UAAUS,AAAAA,IAAAA,mCAAAA,sBAAAA,AAAAA,EACRF,SACA9B,KAAK,WAAW,EAChB6B,WAAW,KAAK;IAEpB;IAEAI,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOJ,YAAY;IAGnBC,QAAQ,OAAO,CAAC,CAACI;QACf,MAAMC,OAAOD,OAAO,IAAI;QACxB,MAAME,sBAAsBpC,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACkC,SAAWA,OAAO,IAAI,KAAKC;QAG9BvC,MAAM,+BAA+BwC;QACrC,MAAMC,eAAeD,sBACjBE,AAAAA,IAAAA,mCAAAA,2BAAAA,AAAAA,EAA4BF,oBAAoB,WAAW,IAC3D,EAAE;QAENxC,MAAM,gBAAgByC;QAEtBA,aAAa,OAAO,CAAC,CAACE;YACpB,MAAMC,eAAeN,OAAO,KAAK,CAACK,MAAM;YACxC,IAAIC,cACF,IAAIjC,QACF2B,OAAO,KAAK,CAACK,MAAM,GAAGE,AAAAA,IAAAA,mCAAAA,aAAAA,AAAAA,EACpBD,cACAnC,KAAK,KAAK,EACVA,KAAK,MAAM,EACXE;iBAEG;gBACL,MAAMmC,UAAUjC,YAAY+B;gBAC5B,IAAIE,SACFR,OAAO,KAAK,CAACK,MAAM,CAAC,EAAE,GAAGG,QAAQ,EAAE;YAEvC;QAEJ;IACF;IAEAT,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,CAACJ,WAAW,KAAK,EAAE,CAAC,wBAAwB,EAAEA,WAAW,KAAK,EAAE;IAEvE,IACEC,AAAmB,MAAnBA,QAAQ,MAAM,IACdC,YAAY,kCAAkC,IAC9C,CAACA,YAAY,KAAK,EAElBY,QAAQ,IAAI,CACV,8EACA5C;IAIJ,OAAOgC;AACT"}
1
+ {"version":3,"file":"ai-model/llm-planning.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/llm-planning.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n DeviceAction,\n InterfaceType,\n PlanningAIResponse,\n UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport {\n AIActionType,\n type AIArgs,\n buildYamlFlowFromPlans,\n fillBboxParam,\n findAllMidsceneLocatorField,\n markupImageForLLM,\n warnGPT4oSizeLimit,\n} from './common';\nimport type { ConversationHistory } from './conversation-history';\nimport {\n automationUserPrompt,\n generateTaskBackgroundContext,\n systemPromptToTaskPlanning,\n} from './prompt/llm-planning';\nimport { describeUserPage } from './prompt/util';\nimport { callAIWithObjectResponse } from './service-caller/index';\n\nconst debug = getDebug('planning');\n\nexport async function plan(\n userInstruction: string,\n opts: {\n context: UIContext;\n interfaceType: InterfaceType;\n actionSpace: DeviceAction<any>[];\n log?: string;\n actionContext?: string;\n modelConfig: IModelConfig;\n conversationHistory?: ConversationHistory;\n },\n): Promise<PlanningAIResponse> {\n const { context, modelConfig, conversationHistory } = opts;\n const { screenshotBase64, size } = context;\n\n const { modelName, vlMode } = modelConfig;\n\n const { description: pageDescription, elementById } = await describeUserPage(\n context,\n { vlMode },\n );\n const systemPrompt = await systemPromptToTaskPlanning({\n actionSpace: opts.actionSpace,\n vlMode: vlMode,\n });\n const historyLog = getLogFromConversationHistory(\n opts.conversationHistory?.snapshot(),\n );\n const logForContext = historyLog ?? opts.log;\n const taskBackgroundContextText = generateTaskBackgroundContext(\n userInstruction,\n logForContext,\n opts.actionContext,\n );\n const userInstructionPrompt = await automationUserPrompt(vlMode).format({\n pageDescription,\n taskBackgroundContext: taskBackgroundContextText,\n });\n\n let imagePayload = screenshotBase64;\n if (vlMode === 'qwen-vl') {\n imagePayload = await paddingToMatchBlockByBase64(imagePayload);\n } else if (!vlMode) {\n imagePayload = await markupImageForLLM(\n screenshotBase64,\n context.tree,\n context.size,\n );\n }\n\n warnGPT4oSizeLimit(size, modelName);\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n detail: 'high',\n },\n },\n {\n type: 'text',\n text: userInstructionPrompt,\n },\n ],\n },\n ];\n\n const { content, usage } = await callAIWithObjectResponse<PlanningAIResponse>(\n msgs,\n AIActionType.PLAN,\n modelConfig,\n );\n const rawResponse = JSON.stringify(content, undefined, 2);\n const planFromAI = content;\n\n const actions =\n (planFromAI.action?.type ? [planFromAI.action] : planFromAI.actions) || [];\n const returnValue: PlanningAIResponse = {\n ...planFromAI,\n actions,\n rawResponse,\n usage,\n yamlFlow: buildYamlFlowFromPlans(\n actions,\n opts.actionSpace,\n planFromAI.sleep,\n ),\n };\n\n assert(planFromAI, \"can't get plans from AI\");\n\n // TODO: use zod.parse to parse the action.param, and then fill the bbox param.\n actions.forEach((action) => {\n const type = action.type;\n const actionInActionSpace = opts.actionSpace.find(\n (action) => action.name === type,\n );\n\n debug('actionInActionSpace matched', actionInActionSpace);\n const locateFields = actionInActionSpace\n ? findAllMidsceneLocatorField(actionInActionSpace.paramSchema)\n : [];\n\n debug('locateFields', locateFields);\n\n locateFields.forEach((field) => {\n const locateResult = action.param[field];\n if (locateResult) {\n if (vlMode) {\n action.param[field] = fillBboxParam(\n locateResult,\n size.width,\n size.height,\n vlMode,\n );\n } else {\n const element = elementById(locateResult);\n if (element) {\n action.param[field].id = element.id;\n }\n }\n }\n });\n });\n // in Qwen-VL, error means error. In GPT-4o, error may mean more actions are needed.\n assert(!planFromAI.error, `Failed to plan actions: ${planFromAI.error}`);\n\n if (\n actions.length === 0 &&\n returnValue.more_actions_needed_by_instruction &&\n !returnValue.sleep\n ) {\n console.warn(\n 'No actions planned for the prompt, but model said more actions are needed:',\n userInstruction,\n );\n }\n\n conversationHistory?.append({\n role: 'assistant',\n content: returnValue.log,\n });\n\n return returnValue;\n}\n\nfunction getLogFromConversationHistory(\n conversationHistory?: ChatCompletionMessageParam[],\n): string | undefined {\n if (!conversationHistory?.length) {\n return undefined;\n }\n\n const assistantMessages = conversationHistory\n .filter((message) => message.role === 'assistant')\n .map((message) => {\n if (typeof message.content === 'string') {\n return message.content.trim();\n }\n\n if (Array.isArray(message.content)) {\n return message.content\n .map((part) => {\n if (part.type === 'text' && typeof part.text === 'string') {\n return part.text.trim();\n }\n\n if (typeof (part as any).text === 'string') {\n return (part as any).text.trim();\n }\n\n return '';\n })\n .filter(Boolean)\n .join('\\n');\n }\n\n return '';\n })\n .map((content) => content.trim())\n .filter(Boolean);\n\n if (assistantMessages.length === 0) {\n return undefined;\n }\n\n return `- ${assistantMessages.join('\\n- ')}`;\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","plan","userInstruction","opts","_opts_conversationHistory","_planFromAI_action","context","modelConfig","conversationHistory","screenshotBase64","size","modelName","vlMode","pageDescription","elementById","describeUserPage","systemPrompt","systemPromptToTaskPlanning","historyLog","getLogFromConversationHistory","logForContext","taskBackgroundContextText","generateTaskBackgroundContext","userInstructionPrompt","automationUserPrompt","imagePayload","paddingToMatchBlockByBase64","markupImageForLLM","warnGPT4oSizeLimit","msgs","content","usage","callAIWithObjectResponse","AIActionType","rawResponse","JSON","undefined","planFromAI","actions","returnValue","buildYamlFlowFromPlans","assert","action","type","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam","element","console","assistantMessages","message","Array","part","Boolean"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;ACuBA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AAEhB,eAAeC,KACpBC,eAAuB,EACvBC,IAQC;QAgBCC,2BAuDCC;IArEH,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAEC,mBAAmB,EAAE,GAAGL;IACtD,MAAM,EAAEM,gBAAgB,EAAEC,IAAI,EAAE,GAAGJ;IAEnC,MAAM,EAAEK,SAAS,EAAEC,MAAM,EAAE,GAAGL;IAE9B,MAAM,EAAE,aAAaM,eAAe,EAAEC,WAAW,EAAE,GAAG,MAAMC,AAAAA,IAAAA,wBAAAA,gBAAAA,AAAAA,EAC1DT,SACA;QAAEM;IAAO;IAEX,MAAMI,eAAe,MAAMC,AAAAA,IAAAA,gCAAAA,0BAAAA,AAAAA,EAA2B;QACpD,aAAad,KAAK,WAAW;QAC7B,QAAQS;IACV;IACA,MAAMM,aAAaC,8BAA8B,QAC/Cf,CAAAA,4BAAAA,KAAK,mBAAmB,AAAD,IAAvBA,KAAAA,IAAAA,0BAA0B,QAAQ;IAEpC,MAAMgB,gBAAgBF,cAAcf,KAAK,GAAG;IAC5C,MAAMkB,4BAA4BC,AAAAA,IAAAA,gCAAAA,6BAAAA,AAAAA,EAChCpB,iBACAkB,eACAjB,KAAK,aAAa;IAEpB,MAAMoB,wBAAwB,MAAMC,AAAAA,IAAAA,gCAAAA,oBAAAA,AAAAA,EAAqBZ,QAAQ,MAAM,CAAC;QACtEC;QACA,uBAAuBQ;IACzB;IAEA,IAAII,eAAehB;IACnB,IAAIG,AAAW,cAAXA,QACFa,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,2BAAAA,AAAAA,EAA4BD;SAC5C,IAAI,CAACb,QACVa,eAAe,MAAME,AAAAA,IAAAA,mCAAAA,iBAAAA,AAAAA,EACnBlB,kBACAH,QAAQ,IAAI,EACZA,QAAQ,IAAI;IAIhBsB,IAAAA,mCAAAA,kBAAAA,AAAAA,EAAmBlB,MAAMC;IAEzB,MAAMkB,OAAe;QACnB;YAAE,MAAM;YAAU,SAASb;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKS;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMF;gBACR;aACD;QACH;KACD;IAED,MAAM,EAAEO,OAAO,EAAEC,KAAK,EAAE,GAAG,MAAMC,AAAAA,IAAAA,yBAAAA,wBAAAA,AAAAA,EAC/BH,MACAI,mCAAAA,YAAAA,CAAAA,IAAiB,EACjB1B;IAEF,MAAM2B,cAAcC,KAAK,SAAS,CAACL,SAASM,QAAW;IACvD,MAAMC,aAAaP;IAEnB,MAAMQ,UACHjC,AAAAA,CAAAA,SAAAA,CAAAA,qBAAAA,WAAW,MAAM,AAAD,IAAhBA,KAAAA,IAAAA,mBAAmB,IAAI,AAAD,IAAI;QAACgC,WAAW,MAAM;KAAC,GAAGA,WAAW,OAAM,KAAM,EAAE;IAC5E,MAAME,cAAkC;QACtC,GAAGF,UAAU;QACbC;QACAJ;QACAH;QACA,UAAUS,AAAAA,IAAAA,mCAAAA,sBAAAA,AAAAA,EACRF,SACAnC,KAAK,WAAW,EAChBkC,WAAW,KAAK;IAEpB;IAEAI,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOJ,YAAY;IAGnBC,QAAQ,OAAO,CAAC,CAACI;QACf,MAAMC,OAAOD,OAAO,IAAI;QACxB,MAAME,sBAAsBzC,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACuC,SAAWA,OAAO,IAAI,KAAKC;QAG9B5C,MAAM,+BAA+B6C;QACrC,MAAMC,eAAeD,sBACjBE,AAAAA,IAAAA,mCAAAA,2BAAAA,AAAAA,EAA4BF,oBAAoB,WAAW,IAC3D,EAAE;QAEN7C,MAAM,gBAAgB8C;QAEtBA,aAAa,OAAO,CAAC,CAACE;YACpB,MAAMC,eAAeN,OAAO,KAAK,CAACK,MAAM;YACxC,IAAIC,cACF,IAAIpC,QACF8B,OAAO,KAAK,CAACK,MAAM,GAAGE,AAAAA,IAAAA,mCAAAA,aAAAA,AAAAA,EACpBD,cACAtC,KAAK,KAAK,EACVA,KAAK,MAAM,EACXE;iBAEG;gBACL,MAAMsC,UAAUpC,YAAYkC;gBAC5B,IAAIE,SACFR,OAAO,KAAK,CAACK,MAAM,CAAC,EAAE,GAAGG,QAAQ,EAAE;YAEvC;QAEJ;IACF;IAEAT,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,CAACJ,WAAW,KAAK,EAAE,CAAC,wBAAwB,EAAEA,WAAW,KAAK,EAAE;IAEvE,IACEC,AAAmB,MAAnBA,QAAQ,MAAM,IACdC,YAAY,kCAAkC,IAC9C,CAACA,YAAY,KAAK,EAElBY,QAAQ,IAAI,CACV,8EACAjD;IAIJM,QAAAA,uBAAAA,oBAAqB,MAAM,CAAC;QAC1B,MAAM;QACN,SAAS+B,YAAY,GAAG;IAC1B;IAEA,OAAOA;AACT;AAEA,SAASpB,8BACPX,mBAAkD;IAElD,IAAI,CAACA,CAAAA,QAAAA,sBAAAA,KAAAA,IAAAA,oBAAqB,MAAM,AAAD,GAC7B;IAGF,MAAM4C,oBAAoB5C,oBACvB,MAAM,CAAC,CAAC6C,UAAYA,AAAiB,gBAAjBA,QAAQ,IAAI,EAChC,GAAG,CAAC,CAACA;QACJ,IAAI,AAA2B,YAA3B,OAAOA,QAAQ,OAAO,EACxB,OAAOA,QAAQ,OAAO,CAAC,IAAI;QAG7B,IAAIC,MAAM,OAAO,CAACD,QAAQ,OAAO,GAC/B,OAAOA,QAAQ,OAAO,CACnB,GAAG,CAAC,CAACE;YACJ,IAAIA,AAAc,WAAdA,KAAK,IAAI,IAAe,AAAqB,YAArB,OAAOA,KAAK,IAAI,EAC1C,OAAOA,KAAK,IAAI,CAAC,IAAI;YAGvB,IAAI,AAA8B,YAA9B,OAAQA,KAAa,IAAI,EAC3B,OAAQA,KAAa,IAAI,CAAC,IAAI;YAGhC,OAAO;QACT,GACC,MAAM,CAACC,SACP,IAAI,CAAC;QAGV,OAAO;IACT,GACC,GAAG,CAAC,CAAC1B,UAAYA,QAAQ,IAAI,IAC7B,MAAM,CAAC0B;IAEV,IAAIJ,AAA6B,MAA7BA,kBAAkB,MAAM,EAC1B;IAGF,OAAO,CAAC,EAAE,EAAEA,kBAAkB,IAAI,CAAC,SAAS;AAC9C"}
@@ -44,18 +44,31 @@ const pointToBbox = (point, width, height)=>[
44
44
  Math.round(Math.min(point.x + bboxSize / 2, width)),
45
45
  Math.round(Math.min(point.y + bboxSize / 2, height))
46
46
  ];
47
- async function vlmPlanning(options) {
48
- const { conversationHistory, userInstruction, size, modelConfig } = options;
47
+ async function vlmPlanning(userInstruction, options) {
48
+ const { conversationHistory, context, modelConfig } = options;
49
49
  const { uiTarsModelVersion } = modelConfig;
50
50
  const systemPrompt = (0, ui_tars_planning_js_namespaceObject.getUiTarsPlanningPrompt)() + userInstruction;
51
+ const imagePayload = await resizeImageForUiTars(context.screenshotBase64, context.size, uiTarsModelVersion);
52
+ conversationHistory.append({
53
+ role: 'user',
54
+ content: [
55
+ {
56
+ type: 'image_url',
57
+ image_url: {
58
+ url: imagePayload
59
+ }
60
+ }
61
+ ]
62
+ });
51
63
  const res = await (0, index_js_namespaceObject.callAIWithStringResponse)([
52
64
  {
53
65
  role: 'user',
54
66
  content: systemPrompt
55
67
  },
56
- ...conversationHistory
68
+ ...conversationHistory.snapshot()
57
69
  ], external_common_js_namespaceObject.AIActionType.INSPECT_ELEMENT, modelConfig);
58
70
  const convertedText = convertBboxToCoordinates(res.content);
71
+ const { size } = context;
59
72
  const { parsed } = (0, action_parser_namespaceObject.actionParser)({
60
73
  prediction: convertedText,
61
74
  factor: [
@@ -70,6 +83,7 @@ async function vlmPlanning(options) {
70
83
  });
71
84
  debug('ui-tars modelVer', uiTarsModelVersion, ', parsed', JSON.stringify(parsed));
72
85
  const transformActions = [];
86
+ let shouldContinue = true;
73
87
  parsed.forEach((action)=>{
74
88
  if ('click' === action.action_type) {
75
89
  (0, utils_namespaceObject.assert)(action.action_inputs.start_box, 'start_box is required');
@@ -147,6 +161,7 @@ async function vlmPlanning(options) {
147
161
  },
148
162
  thought: action.thought || ''
149
163
  });
164
+ else if ('finished' === action.action_type) shouldContinue = false;
150
165
  });
151
166
  if (0 === transformActions.length) throw new Error(`No actions found, response: ${res.content}`, {
152
167
  cause: {
@@ -155,12 +170,17 @@ async function vlmPlanning(options) {
155
170
  }
156
171
  });
157
172
  debug('transformActions', JSON.stringify(transformActions, null, 2));
173
+ const log = (0, ui_tars_planning_js_namespaceObject.getSummary)(res.content);
174
+ conversationHistory.append({
175
+ role: 'assistant',
176
+ content: log
177
+ });
158
178
  return {
159
179
  actions: transformActions,
160
- actionsFromModel: parsed,
161
- action_summary: (0, ui_tars_planning_js_namespaceObject.getSummary)(res.content),
180
+ log,
162
181
  usage: res.usage,
163
- rawResponse: JSON.stringify(res.content, void 0, 2)
182
+ rawResponse: JSON.stringify(res.content, void 0, 2),
183
+ more_actions_needed_by_instruction: shouldContinue
164
184
  };
165
185
  }
166
186
  function convertBboxToCoordinates(text) {
@@ -1 +1 @@
1
- {"version":3,"file":"ai-model/ui-tars-planning.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/ui-tars-planning.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n AIUsageInfo,\n MidsceneYamlFlowItem,\n PlanningAction,\n Size,\n} from '@/types';\nimport { type IModelConfig, UITarsModelVersion } from '@midscene/shared/env';\nimport { resizeImgBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { transformHotkeyInput } from '@midscene/shared/us-keyboard-layout';\nimport { assert } from '@midscene/shared/utils';\nimport { actionParser } from '@ui-tars/action-parser';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport { AIActionType } from './common';\nimport { getSummary, getUiTarsPlanningPrompt } from './prompt/ui-tars-planning';\nimport { callAIWithStringResponse } from './service-caller/index';\ntype ActionType =\n | 'click'\n | 'drag'\n | 'type'\n | 'hotkey'\n | 'finished'\n | 'scroll'\n | 'wait';\n\nconst debug = getDebug('ui-tars-planning');\nconst bboxSize = 10;\nconst pointToBbox = (\n point: { x: number; y: number },\n width: number,\n height: number,\n): [number, number, number, number] => {\n return [\n Math.round(Math.max(point.x - bboxSize / 2, 0)),\n Math.round(Math.max(point.y - bboxSize / 2, 0)),\n Math.round(Math.min(point.x + bboxSize / 2, width)),\n Math.round(Math.min(point.y + bboxSize / 2, height)),\n ];\n};\n\nexport async function vlmPlanning(options: {\n userInstruction: string;\n conversationHistory: ChatCompletionMessageParam[];\n size: { width: number; height: number };\n modelConfig: IModelConfig;\n}): Promise<{\n actions: PlanningAction<any>[];\n actionsFromModel: ReturnType<typeof actionParser>['parsed'];\n action_summary: string;\n yamlFlow?: MidsceneYamlFlowItem[];\n usage?: AIUsageInfo;\n rawResponse?: string;\n}> {\n const { conversationHistory, userInstruction, size, modelConfig } = options;\n const { uiTarsModelVersion } = modelConfig;\n const systemPrompt = getUiTarsPlanningPrompt() + userInstruction;\n\n const res = await callAIWithStringResponse(\n [\n {\n role: 'user',\n content: systemPrompt,\n },\n ...conversationHistory,\n ],\n AIActionType.INSPECT_ELEMENT,\n modelConfig,\n );\n const convertedText = convertBboxToCoordinates(res.content);\n\n const { parsed } = actionParser({\n prediction: convertedText,\n factor: [1000, 1000],\n screenContext: {\n width: size.width,\n height: size.height,\n },\n modelVer: uiTarsModelVersion,\n });\n\n debug(\n 'ui-tars modelVer',\n uiTarsModelVersion,\n ', parsed',\n JSON.stringify(parsed),\n );\n\n const transformActions: PlanningAction[] = [];\n parsed.forEach((action) => {\n if (action.action_type === 'click') {\n assert(action.action_inputs.start_box, 'start_box is required');\n const point = getPoint(action.action_inputs.start_box, size);\n transformActions.push({\n type: 'Tap',\n param: {\n locate: {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: point[0], y: point[1] },\n size.width,\n size.height,\n ),\n },\n },\n });\n } else if (action.action_type === 'drag') {\n assert(action.action_inputs.start_box, 'start_box is required');\n assert(action.action_inputs.end_box, 'end_box is required');\n const startPoint = getPoint(action.action_inputs.start_box, size);\n const endPoint = getPoint(action.action_inputs.end_box, size);\n transformActions.push({\n type: 'DragAndDrop',\n param: {\n from: {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: startPoint[0], y: startPoint[1] },\n size.width,\n size.height,\n ),\n },\n to: {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: endPoint[0], y: endPoint[1] },\n size.width,\n size.height,\n ),\n },\n },\n thought: action.thought || '',\n });\n } else if (action.action_type === 'type') {\n transformActions.push({\n type: 'Input',\n param: {\n value: action.action_inputs.content,\n },\n thought: action.thought || '',\n });\n } else if (action.action_type === 'scroll') {\n transformActions.push({\n type: 'Scroll',\n param: {\n direction: action.action_inputs.direction,\n },\n thought: action.thought || '',\n });\n } else if (action.action_type === 'finished') {\n transformActions.push({\n type: 'Finished',\n param: {},\n thought: action.thought || '',\n });\n } else if (action.action_type === 'hotkey') {\n if (!action.action_inputs.key) {\n console.warn(\n 'No key found in action: hotkey. Will not perform action.',\n );\n } else {\n const keys = transformHotkeyInput(action.action_inputs.key);\n\n transformActions.push({\n type: 'KeyboardPress',\n param: {\n keyName: keys,\n },\n thought: action.thought || '',\n });\n }\n } else if (action.action_type === 'wait') {\n transformActions.push({\n type: 'Sleep',\n param: {\n timeMs: 1000,\n },\n thought: action.thought || '',\n });\n }\n });\n\n if (transformActions.length === 0) {\n throw new Error(`No actions found, response: ${res.content}`, {\n cause: {\n prediction: res.content,\n parsed,\n },\n });\n }\n\n debug('transformActions', JSON.stringify(transformActions, null, 2));\n\n return {\n actions: transformActions,\n actionsFromModel: parsed,\n action_summary: getSummary(res.content),\n usage: res.usage,\n rawResponse: JSON.stringify(res.content, undefined, 2),\n };\n}\n\n/**\n * Converts bounding box notation to coordinate points\n * @param text - The text containing bbox tags to be converted\n * @returns The text with bbox tags replaced by coordinate points\n */\nfunction convertBboxToCoordinates(text: string): string {\n // Match the four numbers after <bbox>\n const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n function replaceMatch(\n match: string,\n x1: string,\n y1: string,\n x2: string,\n y2: string,\n ): string {\n // Convert strings to numbers and calculate center point\n const x1Num = Number.parseInt(x1, 10);\n const y1Num = Number.parseInt(y1, 10);\n const x2Num = Number.parseInt(x2, 10);\n const y2Num = Number.parseInt(y2, 10);\n\n // Use Math.floor to truncate and calculate center point\n const x = Math.floor((x1Num + x2Num) / 2);\n const y = Math.floor((y1Num + y2Num) / 2);\n\n // Return formatted coordinate string\n return `(${x},${y})`;\n }\n\n // Remove [EOS] and replace <bbox> coordinates\n const cleanedText = text.replace(/\\[EOS\\]/g, '');\n return cleanedText.replace(pattern, replaceMatch).trim();\n}\n\nfunction getPoint(startBox: string, size: { width: number; height: number }) {\n const [x, y] = JSON.parse(startBox);\n return [x * size.width, y * size.height];\n}\n\ninterface BaseAction {\n action_type: ActionType;\n action_inputs: Record<string, any>;\n reflection: string | null;\n thought: string | null;\n}\n\ninterface ClickAction extends BaseAction {\n action_type: 'click';\n action_inputs: {\n start_box: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface DragAction extends BaseAction {\n action_type: 'drag';\n action_inputs: {\n start_box: string; // JSON string of [x, y] coordinates\n end_box: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface WaitAction extends BaseAction {\n action_type: 'wait';\n action_inputs: {\n time: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface TypeAction extends BaseAction {\n action_type: 'type';\n action_inputs: {\n content: string;\n };\n}\n\ninterface HotkeyAction extends BaseAction {\n action_type: 'hotkey';\n action_inputs: {\n key: string;\n };\n}\n\ninterface ScrollAction extends BaseAction {\n action_type: 'scroll';\n action_inputs: {\n direction: 'up' | 'down';\n };\n}\n\ninterface FinishedAction extends BaseAction {\n action_type: 'finished';\n action_inputs: Record<string, never>;\n}\n\nexport type Action =\n | ClickAction\n | DragAction\n | TypeAction\n | HotkeyAction\n | ScrollAction\n | FinishedAction\n | WaitAction;\n\nexport async function resizeImageForUiTars(\n imageBase64: string,\n size: Size,\n uiTarsVersion: UITarsModelVersion | undefined,\n) {\n if (uiTarsVersion === UITarsModelVersion.V1_5) {\n debug('ui-tars-v1.5, will check image size', size);\n const currentPixels = size.width * size.height;\n const maxPixels = 16384 * 28 * 28; //\n if (currentPixels > maxPixels) {\n const resizeFactor = Math.sqrt(maxPixels / currentPixels);\n const newWidth = Math.floor(size.width * resizeFactor);\n const newHeight = Math.floor(size.height * resizeFactor);\n debug(\n 'resize image for ui-tars, new width: %s, new height: %s',\n newWidth,\n newHeight,\n );\n const resizedImage = await resizeImgBase64(imageBase64, {\n width: newWidth,\n height: newHeight,\n });\n return resizedImage;\n }\n }\n return imageBase64;\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","bboxSize","pointToBbox","point","width","height","Math","vlmPlanning","options","conversationHistory","userInstruction","size","modelConfig","uiTarsModelVersion","systemPrompt","getUiTarsPlanningPrompt","res","callAIWithStringResponse","AIActionType","convertedText","convertBboxToCoordinates","parsed","actionParser","JSON","transformActions","action","assert","getPoint","startPoint","endPoint","keys","transformHotkeyInput","console","Error","getSummary","undefined","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","y","cleanedText","startBox","resizeImageForUiTars","imageBase64","uiTarsVersion","UITarsModelVersion","currentPixels","maxPixels","resizeFactor","newWidth","newHeight","resizedImage","resizeImgBase64"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;ACmBA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AACvB,MAAMC,WAAW;AACjB,MAAMC,cAAc,CAClBC,OACAC,OACAC,SAEO;QACLC,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGG;QAC5CE,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGI;KAC7C;AAGI,eAAeE,YAAYC,OAKjC;IAQC,MAAM,EAAEC,mBAAmB,EAAEC,eAAe,EAAEC,IAAI,EAAEC,WAAW,EAAE,GAAGJ;IACpE,MAAM,EAAEK,kBAAkB,EAAE,GAAGD;IAC/B,MAAME,eAAeC,AAAAA,IAAAA,oCAAAA,uBAAAA,AAAAA,MAA4BL;IAEjD,MAAMM,MAAM,MAAMC,AAAAA,IAAAA,yBAAAA,wBAAAA,AAAAA,EAChB;QACE;YACE,MAAM;YACN,SAASH;QACX;WACGL;KACJ,EACDS,mCAAAA,YAAAA,CAAAA,eAA4B,EAC5BN;IAEF,MAAMO,gBAAgBC,yBAAyBJ,IAAI,OAAO;IAE1D,MAAM,EAAEK,MAAM,EAAE,GAAGC,AAAAA,IAAAA,8BAAAA,YAAAA,AAAAA,EAAa;QAC9B,YAAYH;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOR,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUE;IACZ;IAEAd,MACE,oBACAc,oBACA,YACAU,KAAK,SAAS,CAACF;IAGjB,MAAMG,mBAAqC,EAAE;IAC7CH,OAAO,OAAO,CAAC,CAACI;QACd,IAAIA,AAAuB,YAAvBA,OAAO,WAAW,EAAc;YAClCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAMtB,QAAQwB,SAASF,OAAO,aAAa,CAAC,SAAS,EAAEd;YACvDa,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQC,OAAO,OAAO,IAAI;wBAC1B,MAAMvB,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BQ,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;YACF;QACF,OAAO,IAAIc,AAAuB,WAAvBA,OAAO,WAAW,EAAa;YACxCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMG,aAAaD,SAASF,OAAO,aAAa,CAAC,SAAS,EAAEd;YAC5D,MAAMkB,WAAWF,SAASF,OAAO,aAAa,CAAC,OAAO,EAAEd;YACxDa,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,MAAM;wBACJ,QAAQC,OAAO,OAAO,IAAI;wBAC1B,MAAMvB,YACJ;4BAAE,GAAG0B,UAAU,CAAC,EAAE;4BAAE,GAAGA,UAAU,CAAC,EAAE;wBAAC,GACrCjB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;oBACA,IAAI;wBACF,QAAQc,OAAO,OAAO,IAAI;wBAC1B,MAAMvB,YACJ;4BAAE,GAAG2B,QAAQ,CAAC,EAAE;4BAAE,GAAGA,QAAQ,CAAC,EAAE;wBAAC,GACjClB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASc,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIA,AAAuB,WAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,OAAOC,OAAO,aAAa,CAAC,OAAO;YACrC;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,aAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWC,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,eAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO,CAAC;YACR,SAASC,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,aAAvBA,OAAO,WAAW,EAC3B,IAAKA,OAAO,aAAa,CAAC,GAAG,EAItB;YACL,MAAMK,OAAOC,AAAAA,IAAAA,mCAAAA,oBAAAA,AAAAA,EAAqBN,OAAO,aAAa,CAAC,GAAG;YAE1DD,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,SAASM;gBACX;gBACA,SAASL,OAAO,OAAO,IAAI;YAC7B;QACF,OAbEO,QAAQ,IAAI,CACV;aAaC,IAAIP,AAAuB,WAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,SAASC,OAAO,OAAO,IAAI;QAC7B;IAEJ;IAEA,IAAID,AAA4B,MAA5BA,iBAAiB,MAAM,EACzB,MAAM,IAAIS,MAAM,CAAC,4BAA4B,EAAEjB,IAAI,OAAO,EAAE,EAAE;QAC5D,OAAO;YACL,YAAYA,IAAI,OAAO;YACvBK;QACF;IACF;IAGFtB,MAAM,oBAAoBwB,KAAK,SAAS,CAACC,kBAAkB,MAAM;IAEjE,OAAO;QACL,SAASA;QACT,kBAAkBH;QAClB,gBAAgBa,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAWlB,IAAI,OAAO;QACtC,OAAOA,IAAI,KAAK;QAChB,aAAaO,KAAK,SAAS,CAACP,IAAI,OAAO,EAAEmB,QAAW;IACtD;AACF;AAOA,SAASf,yBAAyBgB,IAAY;IAE5C,MAAMC,UAAU;IAEhB,SAASC,aACPC,KAAa,EACbC,EAAU,EACVC,EAAU,EACVC,EAAU,EACVC,EAAU;QAGV,MAAMC,QAAQC,OAAO,QAAQ,CAACL,IAAI;QAClC,MAAMM,QAAQD,OAAO,QAAQ,CAACJ,IAAI;QAClC,MAAMM,QAAQF,OAAO,QAAQ,CAACH,IAAI;QAClC,MAAMM,QAAQH,OAAO,QAAQ,CAACF,IAAI;QAGlC,MAAMM,IAAI3C,KAAK,KAAK,CAAEsC,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMG,IAAI5C,KAAK,KAAK,CAAEwC,AAAAA,CAAAA,QAAQE,KAAI,IAAK;QAGvC,OAAO,CAAC,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC,CAAC;IACtB;IAGA,MAAMC,cAAcf,KAAK,OAAO,CAAC,YAAY;IAC7C,OAAOe,YAAY,OAAO,CAACd,SAASC,cAAc,IAAI;AACxD;AAEA,SAASX,SAASyB,QAAgB,EAAEzC,IAAuC;IACzE,MAAM,CAACsC,GAAGC,EAAE,GAAG3B,KAAK,KAAK,CAAC6B;IAC1B,OAAO;QAACH,IAAItC,KAAK,KAAK;QAAEuC,IAAIvC,KAAK,MAAM;KAAC;AAC1C;AAkEO,eAAe0C,qBACpBC,WAAmB,EACnB3C,IAAU,EACV4C,aAA6C;IAE7C,IAAIA,kBAAkBC,oBAAAA,kBAAAA,CAAAA,IAAuB,EAAE;QAC7CzD,MAAM,uCAAuCY;QAC7C,MAAM8C,gBAAgB9C,KAAK,KAAK,GAAGA,KAAK,MAAM;QAC9C,MAAM+C,YAAY;QAClB,IAAID,gBAAgBC,WAAW;YAC7B,MAAMC,eAAerD,KAAK,IAAI,CAACoD,YAAYD;YAC3C,MAAMG,WAAWtD,KAAK,KAAK,CAACK,KAAK,KAAK,GAAGgD;YACzC,MAAME,YAAYvD,KAAK,KAAK,CAACK,KAAK,MAAM,GAAGgD;YAC3C5D,MACE,2DACA6D,UACAC;YAEF,MAAMC,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,eAAAA,AAAAA,EAAgBT,aAAa;gBACtD,OAAOM;gBACP,QAAQC;YACV;YACA,OAAOC;QACT;IACF;IACA,OAAOR;AACT"}
1
+ {"version":3,"file":"ai-model/ui-tars-planning.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/ui-tars-planning.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n }\n }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n PlanningAIResponse,\n PlanningAction,\n Size,\n UIContext,\n} from '@/types';\nimport { type IModelConfig, UITarsModelVersion } from '@midscene/shared/env';\nimport { resizeImgBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { transformHotkeyInput } from '@midscene/shared/us-keyboard-layout';\nimport { assert } from '@midscene/shared/utils';\nimport { actionParser } from '@ui-tars/action-parser';\nimport { AIActionType } from './common';\nimport type { ConversationHistory } from './conversation-history';\nimport { getSummary, getUiTarsPlanningPrompt } from './prompt/ui-tars-planning';\nimport { callAIWithStringResponse } from './service-caller/index';\ntype ActionType =\n | 'click'\n | 'drag'\n | 'type'\n | 'hotkey'\n | 'finished'\n | 'scroll'\n | 'wait';\n\nconst debug = getDebug('ui-tars-planning');\nconst bboxSize = 10;\nconst pointToBbox = (\n point: { x: number; y: number },\n width: number,\n height: number,\n): [number, number, number, number] => {\n return [\n Math.round(Math.max(point.x - bboxSize / 2, 0)),\n Math.round(Math.max(point.y - bboxSize / 2, 0)),\n Math.round(Math.min(point.x + bboxSize / 2, width)),\n Math.round(Math.min(point.y + bboxSize / 2, height)),\n ];\n};\n\nexport async function vlmPlanning(\n userInstruction: string,\n options: {\n conversationHistory: ConversationHistory;\n context: UIContext;\n modelConfig: IModelConfig;\n },\n): Promise<PlanningAIResponse> {\n const { conversationHistory, context, modelConfig } = options;\n const { uiTarsModelVersion } = modelConfig;\n const systemPrompt = getUiTarsPlanningPrompt() + userInstruction;\n\n const imagePayload = await resizeImageForUiTars(\n context.screenshotBase64,\n context.size,\n uiTarsModelVersion,\n );\n\n conversationHistory.append({\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n },\n },\n ],\n });\n\n const res = await callAIWithStringResponse(\n [\n {\n role: 'user',\n content: systemPrompt,\n },\n ...conversationHistory.snapshot(),\n ],\n AIActionType.INSPECT_ELEMENT,\n modelConfig,\n );\n const convertedText = convertBboxToCoordinates(res.content);\n\n const { size } = context;\n const { parsed } = actionParser({\n prediction: convertedText,\n factor: [1000, 1000],\n screenContext: {\n width: size.width,\n height: size.height,\n },\n modelVer: uiTarsModelVersion,\n });\n\n debug(\n 'ui-tars modelVer',\n uiTarsModelVersion,\n ', parsed',\n JSON.stringify(parsed),\n );\n\n const transformActions: PlanningAction[] = [];\n let shouldContinue = true;\n parsed.forEach((action) => {\n if (action.action_type === 'click') {\n assert(action.action_inputs.start_box, 'start_box is required');\n const point = getPoint(action.action_inputs.start_box, size);\n transformActions.push({\n type: 'Tap',\n param: {\n locate: {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: point[0], y: point[1] },\n size.width,\n size.height,\n ),\n },\n },\n });\n } else if (action.action_type === 'drag') {\n assert(action.action_inputs.start_box, 'start_box is required');\n assert(action.action_inputs.end_box, 'end_box is required');\n const startPoint = getPoint(action.action_inputs.start_box, size);\n const endPoint = getPoint(action.action_inputs.end_box, size);\n transformActions.push({\n type: 'DragAndDrop',\n param: {\n from: {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: startPoint[0], y: startPoint[1] },\n size.width,\n size.height,\n ),\n },\n to: {\n prompt: action.thought || '',\n bbox: pointToBbox(\n { x: endPoint[0], y: endPoint[1] },\n size.width,\n size.height,\n ),\n },\n },\n thought: action.thought || '',\n });\n } else if (action.action_type === 'type') {\n transformActions.push({\n type: 'Input',\n param: {\n value: action.action_inputs.content,\n },\n thought: action.thought || '',\n });\n } else if (action.action_type === 'scroll') {\n transformActions.push({\n type: 'Scroll',\n param: {\n direction: action.action_inputs.direction,\n },\n thought: action.thought || '',\n });\n } else if (action.action_type === 'finished') {\n transformActions.push({\n type: 'Finished',\n param: {},\n thought: action.thought || '',\n });\n } else if (action.action_type === 'hotkey') {\n if (!action.action_inputs.key) {\n console.warn(\n 'No key found in action: hotkey. Will not perform action.',\n );\n } else {\n const keys = transformHotkeyInput(action.action_inputs.key);\n\n transformActions.push({\n type: 'KeyboardPress',\n param: {\n keyName: keys,\n },\n thought: action.thought || '',\n });\n }\n } else if (action.action_type === 'wait') {\n transformActions.push({\n type: 'Sleep',\n param: {\n timeMs: 1000,\n },\n thought: action.thought || '',\n });\n } else if (action.action_type === 'finished') {\n shouldContinue = false;\n }\n });\n\n if (transformActions.length === 0) {\n throw new Error(`No actions found, response: ${res.content}`, {\n cause: {\n prediction: res.content,\n parsed,\n },\n });\n }\n\n debug('transformActions', JSON.stringify(transformActions, null, 2));\n const log = getSummary(res.content);\n\n conversationHistory.append({\n role: 'assistant',\n content: log,\n });\n\n return {\n actions: transformActions,\n log,\n usage: res.usage,\n rawResponse: JSON.stringify(res.content, undefined, 2),\n more_actions_needed_by_instruction: shouldContinue,\n };\n}\n\n/**\n * Converts bounding box notation to coordinate points\n * @param text - The text containing bbox tags to be converted\n * @returns The text with bbox tags replaced by coordinate points\n */\nfunction convertBboxToCoordinates(text: string): string {\n // Match the four numbers after <bbox>\n const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n function replaceMatch(\n match: string,\n x1: string,\n y1: string,\n x2: string,\n y2: string,\n ): string {\n // Convert strings to numbers and calculate center point\n const x1Num = Number.parseInt(x1, 10);\n const y1Num = Number.parseInt(y1, 10);\n const x2Num = Number.parseInt(x2, 10);\n const y2Num = Number.parseInt(y2, 10);\n\n // Use Math.floor to truncate and calculate center point\n const x = Math.floor((x1Num + x2Num) / 2);\n const y = Math.floor((y1Num + y2Num) / 2);\n\n // Return formatted coordinate string\n return `(${x},${y})`;\n }\n\n // Remove [EOS] and replace <bbox> coordinates\n const cleanedText = text.replace(/\\[EOS\\]/g, '');\n return cleanedText.replace(pattern, replaceMatch).trim();\n}\n\nfunction getPoint(startBox: string, size: { width: number; height: number }) {\n const [x, y] = JSON.parse(startBox);\n return [x * size.width, y * size.height];\n}\n\ninterface BaseAction {\n action_type: ActionType;\n action_inputs: Record<string, any>;\n reflection: string | null;\n thought: string | null;\n}\n\ninterface ClickAction extends BaseAction {\n action_type: 'click';\n action_inputs: {\n start_box: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface DragAction extends BaseAction {\n action_type: 'drag';\n action_inputs: {\n start_box: string; // JSON string of [x, y] coordinates\n end_box: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface WaitAction extends BaseAction {\n action_type: 'wait';\n action_inputs: {\n time: string; // JSON string of [x, y] coordinates\n };\n}\n\ninterface TypeAction extends BaseAction {\n action_type: 'type';\n action_inputs: {\n content: string;\n };\n}\n\ninterface HotkeyAction extends BaseAction {\n action_type: 'hotkey';\n action_inputs: {\n key: string;\n };\n}\n\ninterface ScrollAction extends BaseAction {\n action_type: 'scroll';\n action_inputs: {\n direction: 'up' | 'down';\n };\n}\n\ninterface FinishedAction extends BaseAction {\n action_type: 'finished';\n action_inputs: Record<string, never>;\n}\n\nexport type Action =\n | ClickAction\n | DragAction\n | TypeAction\n | HotkeyAction\n | ScrollAction\n | FinishedAction\n | WaitAction;\n\nexport async function resizeImageForUiTars(\n imageBase64: string,\n size: Size,\n uiTarsVersion: UITarsModelVersion | undefined,\n) {\n if (uiTarsVersion === UITarsModelVersion.V1_5) {\n debug('ui-tars-v1.5, will check image size', size);\n const currentPixels = size.width * size.height;\n const maxPixels = 16384 * 28 * 28; //\n if (currentPixels > maxPixels) {\n const resizeFactor = Math.sqrt(maxPixels / currentPixels);\n const newWidth = Math.floor(size.width * resizeFactor);\n const newHeight = Math.floor(size.height * resizeFactor);\n debug(\n 'resize image for ui-tars, new width: %s, new height: %s',\n newWidth,\n newHeight,\n );\n const resizedImage = await resizeImgBase64(imageBase64, {\n width: newWidth,\n height: newHeight,\n });\n return resizedImage;\n }\n }\n return imageBase64;\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","bboxSize","pointToBbox","point","width","height","Math","vlmPlanning","userInstruction","options","conversationHistory","context","modelConfig","uiTarsModelVersion","systemPrompt","getUiTarsPlanningPrompt","imagePayload","resizeImageForUiTars","res","callAIWithStringResponse","AIActionType","convertedText","convertBboxToCoordinates","size","parsed","actionParser","JSON","transformActions","shouldContinue","action","assert","getPoint","startPoint","endPoint","keys","transformHotkeyInput","console","Error","log","getSummary","undefined","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","y","cleanedText","startBox","imageBase64","uiTarsVersion","UITarsModelVersion","currentPixels","maxPixels","resizeFactor","newWidth","newHeight","resizedImage","resizeImgBase64"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;ACmBA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AACvB,MAAMC,WAAW;AACjB,MAAMC,cAAc,CAClBC,OACAC,OACAC,SAEO;QACLC,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGG;QAC5CE,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGI;KAC7C;AAGI,eAAeE,YACpBC,eAAuB,EACvBC,OAIC;IAED,MAAM,EAAEC,mBAAmB,EAAEC,OAAO,EAAEC,WAAW,EAAE,GAAGH;IACtD,MAAM,EAAEI,kBAAkB,EAAE,GAAGD;IAC/B,MAAME,eAAeC,AAAAA,IAAAA,oCAAAA,uBAAAA,AAAAA,MAA4BP;IAEjD,MAAMQ,eAAe,MAAMC,qBACzBN,QAAQ,gBAAgB,EACxBA,QAAQ,IAAI,EACZE;IAGFH,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKM;gBACP;YACF;SACD;IACH;IAEA,MAAME,MAAM,MAAMC,AAAAA,IAAAA,yBAAAA,wBAAAA,AAAAA,EAChB;QACE;YACE,MAAM;YACN,SAASL;QACX;WACGJ,oBAAoB,QAAQ;KAChC,EACDU,mCAAAA,YAAAA,CAAAA,eAA4B,EAC5BR;IAEF,MAAMS,gBAAgBC,yBAAyBJ,IAAI,OAAO;IAE1D,MAAM,EAAEK,IAAI,EAAE,GAAGZ;IACjB,MAAM,EAAEa,MAAM,EAAE,GAAGC,AAAAA,IAAAA,8BAAAA,YAAAA,AAAAA,EAAa;QAC9B,YAAYJ;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOE,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUV;IACZ;IAEAd,MACE,oBACAc,oBACA,YACAa,KAAK,SAAS,CAACF;IAGjB,MAAMG,mBAAqC,EAAE;IAC7C,IAAIC,iBAAiB;IACrBJ,OAAO,OAAO,CAAC,CAACK;QACd,IAAIA,AAAuB,YAAvBA,OAAO,WAAW,EAAc;YAClCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM1B,QAAQ4B,SAASF,OAAO,aAAa,CAAC,SAAS,EAAEN;YACvDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQE,OAAO,OAAO,IAAI;wBAC1B,MAAM3B,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BoB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;YACF;QACF,OAAO,IAAIM,AAAuB,WAAvBA,OAAO,WAAW,EAAa;YACxCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMG,aAAaD,SAASF,OAAO,aAAa,CAAC,SAAS,EAAEN;YAC5D,MAAMU,WAAWF,SAASF,OAAO,aAAa,CAAC,OAAO,EAAEN;YACxDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,MAAM;wBACJ,QAAQE,OAAO,OAAO,IAAI;wBAC1B,MAAM3B,YACJ;4BAAE,GAAG8B,UAAU,CAAC,EAAE;4BAAE,GAAGA,UAAU,CAAC,EAAE;wBAAC,GACrCT,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;oBACA,IAAI;wBACF,QAAQM,OAAO,OAAO,IAAI;wBAC1B,MAAM3B,YACJ;4BAAE,GAAG+B,QAAQ,CAAC,EAAE;4BAAE,GAAGA,QAAQ,CAAC,EAAE;wBAAC,GACjCV,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASM,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIA,AAAuB,WAAvBA,OAAO,WAAW,EAC3BF,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,OAAOE,OAAO,aAAa,CAAC,OAAO;YACrC;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,aAAvBA,OAAO,WAAW,EAC3BF,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWE,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,eAAvBA,OAAO,WAAW,EAC3BF,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO,CAAC;YACR,SAASE,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,aAAvBA,OAAO,WAAW,EAC3B,IAAKA,OAAO,aAAa,CAAC,GAAG,EAItB;YACL,MAAMK,OAAOC,AAAAA,IAAAA,mCAAAA,oBAAAA,AAAAA,EAAqBN,OAAO,aAAa,CAAC,GAAG;YAE1DF,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,SAASO;gBACX;gBACA,SAASL,OAAO,OAAO,IAAI;YAC7B;QACF,OAbEO,QAAQ,IAAI,CACV;aAaC,IAAIP,AAAuB,WAAvBA,OAAO,WAAW,EAC3BF,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,SAASE,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,eAAvBA,OAAO,WAAW,EAC3BD,iBAAiB;IAErB;IAEA,IAAID,AAA4B,MAA5BA,iBAAiB,MAAM,EACzB,MAAM,IAAIU,MAAM,CAAC,4BAA4B,EAAEnB,IAAI,OAAO,EAAE,EAAE;QAC5D,OAAO;YACL,YAAYA,IAAI,OAAO;YACvBM;QACF;IACF;IAGFzB,MAAM,oBAAoB2B,KAAK,SAAS,CAACC,kBAAkB,MAAM;IACjE,MAAMW,MAAMC,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAWrB,IAAI,OAAO;IAElCR,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS4B;IACX;IAEA,OAAO;QACL,SAASX;QACTW;QACA,OAAOpB,IAAI,KAAK;QAChB,aAAaQ,KAAK,SAAS,CAACR,IAAI,OAAO,EAAEsB,QAAW;QACpD,oCAAoCZ;IACtC;AACF;AAOA,SAASN,yBAAyBmB,IAAY;IAE5C,MAAMC,UAAU;IAEhB,SAASC,aACPC,KAAa,EACbC,EAAU,EACVC,EAAU,EACVC,EAAU,EACVC,EAAU;QAGV,MAAMC,QAAQC,OAAO,QAAQ,CAACL,IAAI;QAClC,MAAMM,QAAQD,OAAO,QAAQ,CAACJ,IAAI;QAClC,MAAMM,QAAQF,OAAO,QAAQ,CAACH,IAAI;QAClC,MAAMM,QAAQH,OAAO,QAAQ,CAACF,IAAI;QAGlC,MAAMM,IAAIhD,KAAK,KAAK,CAAE2C,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMG,IAAIjD,KAAK,KAAK,CAAE6C,AAAAA,CAAAA,QAAQE,KAAI,IAAK;QAGvC,OAAO,CAAC,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC,CAAC;IACtB;IAGA,MAAMC,cAAcf,KAAK,OAAO,CAAC,YAAY;IAC7C,OAAOe,YAAY,OAAO,CAACd,SAASC,cAAc,IAAI;AACxD;AAEA,SAASZ,SAAS0B,QAAgB,EAAElC,IAAuC;IACzE,MAAM,CAAC+B,GAAGC,EAAE,GAAG7B,KAAK,KAAK,CAAC+B;IAC1B,OAAO;QAACH,IAAI/B,KAAK,KAAK;QAAEgC,IAAIhC,KAAK,MAAM;KAAC;AAC1C;AAkEO,eAAeN,qBACpByC,WAAmB,EACnBnC,IAAU,EACVoC,aAA6C;IAE7C,IAAIA,kBAAkBC,oBAAAA,kBAAAA,CAAAA,IAAuB,EAAE;QAC7C7D,MAAM,uCAAuCwB;QAC7C,MAAMsC,gBAAgBtC,KAAK,KAAK,GAAGA,KAAK,MAAM;QAC9C,MAAMuC,YAAY;QAClB,IAAID,gBAAgBC,WAAW;YAC7B,MAAMC,eAAezD,KAAK,IAAI,CAACwD,YAAYD;YAC3C,MAAMG,WAAW1D,KAAK,KAAK,CAACiB,KAAK,KAAK,GAAGwC;YACzC,MAAME,YAAY3D,KAAK,KAAK,CAACiB,KAAK,MAAM,GAAGwC;YAC3ChE,MACE,2DACAiE,UACAC;YAEF,MAAMC,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,eAAAA,AAAAA,EAAgBT,aAAa;gBACtD,OAAOM;gBACP,QAAQC;YACV;YACA,OAAOC;QACT;IACF;IACA,OAAOR;AACT"}