@midscene/core 0.30.10 → 1.0.0

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 (203) hide show
  1. package/dist/es/agent/agent.mjs +233 -144
  2. package/dist/es/agent/agent.mjs.map +1 -1
  3. package/dist/es/agent/execution-session.mjs +41 -0
  4. package/dist/es/agent/execution-session.mjs.map +1 -0
  5. package/dist/es/agent/index.mjs +3 -3
  6. package/dist/es/agent/task-builder.mjs +319 -0
  7. package/dist/es/agent/task-builder.mjs.map +1 -0
  8. package/dist/es/agent/task-cache.mjs +4 -4
  9. package/dist/es/agent/task-cache.mjs.map +1 -1
  10. package/dist/es/agent/tasks.mjs +197 -504
  11. package/dist/es/agent/tasks.mjs.map +1 -1
  12. package/dist/es/agent/ui-utils.mjs +54 -35
  13. package/dist/es/agent/ui-utils.mjs.map +1 -1
  14. package/dist/es/agent/utils.mjs +16 -58
  15. package/dist/es/agent/utils.mjs.map +1 -1
  16. package/dist/es/ai-model/conversation-history.mjs +25 -13
  17. package/dist/es/ai-model/conversation-history.mjs.map +1 -1
  18. package/dist/es/ai-model/index.mjs +4 -4
  19. package/dist/es/ai-model/inspect.mjs +45 -54
  20. package/dist/es/ai-model/inspect.mjs.map +1 -1
  21. package/dist/es/ai-model/llm-planning.mjs +47 -65
  22. package/dist/es/ai-model/llm-planning.mjs.map +1 -1
  23. package/dist/es/ai-model/prompt/assertion.mjs.map +1 -1
  24. package/dist/es/ai-model/prompt/common.mjs.map +1 -1
  25. package/dist/es/ai-model/prompt/describe.mjs.map +1 -1
  26. package/dist/es/ai-model/prompt/extraction.mjs.map +1 -1
  27. package/dist/es/ai-model/prompt/llm-locator.mjs +11 -235
  28. package/dist/es/ai-model/prompt/llm-locator.mjs.map +1 -1
  29. package/dist/es/ai-model/prompt/llm-planning.mjs +76 -322
  30. package/dist/es/ai-model/prompt/llm-planning.mjs.map +1 -1
  31. package/dist/es/ai-model/prompt/llm-section-locator.mjs +15 -14
  32. package/dist/es/ai-model/prompt/llm-section-locator.mjs.map +1 -1
  33. package/dist/es/ai-model/prompt/order-sensitive-judge.mjs +35 -0
  34. package/dist/es/ai-model/prompt/order-sensitive-judge.mjs.map +1 -0
  35. package/dist/es/ai-model/prompt/playwright-generator.mjs +2 -2
  36. package/dist/es/ai-model/prompt/playwright-generator.mjs.map +1 -1
  37. package/dist/es/ai-model/prompt/ui-tars-locator.mjs.map +1 -1
  38. package/dist/es/ai-model/prompt/ui-tars-planning.mjs.map +1 -1
  39. package/dist/es/ai-model/prompt/util.mjs +3 -88
  40. package/dist/es/ai-model/prompt/util.mjs.map +1 -1
  41. package/dist/es/ai-model/prompt/yaml-generator.mjs +10 -10
  42. package/dist/es/ai-model/prompt/yaml-generator.mjs.map +1 -1
  43. package/dist/es/ai-model/service-caller/index.mjs +182 -274
  44. package/dist/es/ai-model/service-caller/index.mjs.map +1 -1
  45. package/dist/es/ai-model/ui-tars-planning.mjs +69 -8
  46. package/dist/es/ai-model/ui-tars-planning.mjs.map +1 -1
  47. package/dist/es/{ai-model/common.mjs → common.mjs} +18 -30
  48. package/dist/es/common.mjs.map +1 -0
  49. package/dist/es/device/device-options.mjs +0 -0
  50. package/dist/es/device/index.mjs +29 -12
  51. package/dist/es/device/index.mjs.map +1 -1
  52. package/dist/es/index.mjs +5 -4
  53. package/dist/es/index.mjs.map +1 -1
  54. package/dist/es/report.mjs.map +1 -1
  55. package/dist/es/{insight → service}/index.mjs +38 -51
  56. package/dist/es/service/index.mjs.map +1 -0
  57. package/dist/es/{insight → service}/utils.mjs +3 -3
  58. package/dist/es/service/utils.mjs.map +1 -0
  59. package/dist/es/task-runner.mjs +264 -0
  60. package/dist/es/task-runner.mjs.map +1 -0
  61. package/dist/es/tree.mjs +13 -2
  62. package/dist/es/tree.mjs.map +1 -0
  63. package/dist/es/types.mjs +18 -1
  64. package/dist/es/types.mjs.map +1 -1
  65. package/dist/es/utils.mjs +6 -7
  66. package/dist/es/utils.mjs.map +1 -1
  67. package/dist/es/yaml/builder.mjs.map +1 -1
  68. package/dist/es/yaml/player.mjs +121 -98
  69. package/dist/es/yaml/player.mjs.map +1 -1
  70. package/dist/es/yaml/utils.mjs +1 -1
  71. package/dist/es/yaml/utils.mjs.map +1 -1
  72. package/dist/lib/agent/agent.js +231 -142
  73. package/dist/lib/agent/agent.js.map +1 -1
  74. package/dist/lib/agent/common.js +1 -1
  75. package/dist/lib/agent/execution-session.js +75 -0
  76. package/dist/lib/agent/execution-session.js.map +1 -0
  77. package/dist/lib/agent/index.js +14 -14
  78. package/dist/lib/agent/index.js.map +1 -1
  79. package/dist/lib/agent/task-builder.js +356 -0
  80. package/dist/lib/agent/task-builder.js.map +1 -0
  81. package/dist/lib/agent/task-cache.js +8 -8
  82. package/dist/lib/agent/task-cache.js.map +1 -1
  83. package/dist/lib/agent/tasks.js +202 -506
  84. package/dist/lib/agent/tasks.js.map +1 -1
  85. package/dist/lib/agent/ui-utils.js +58 -36
  86. package/dist/lib/agent/ui-utils.js.map +1 -1
  87. package/dist/lib/agent/utils.js +26 -68
  88. package/dist/lib/agent/utils.js.map +1 -1
  89. package/dist/lib/ai-model/conversation-history.js +27 -15
  90. package/dist/lib/ai-model/conversation-history.js.map +1 -1
  91. package/dist/lib/ai-model/index.js +27 -27
  92. package/dist/lib/ai-model/index.js.map +1 -1
  93. package/dist/lib/ai-model/inspect.js +51 -57
  94. package/dist/lib/ai-model/inspect.js.map +1 -1
  95. package/dist/lib/ai-model/llm-planning.js +49 -67
  96. package/dist/lib/ai-model/llm-planning.js.map +1 -1
  97. package/dist/lib/ai-model/prompt/assertion.js +2 -2
  98. package/dist/lib/ai-model/prompt/assertion.js.map +1 -1
  99. package/dist/lib/ai-model/prompt/common.js +2 -2
  100. package/dist/lib/ai-model/prompt/common.js.map +1 -1
  101. package/dist/lib/ai-model/prompt/describe.js +2 -2
  102. package/dist/lib/ai-model/prompt/describe.js.map +1 -1
  103. package/dist/lib/ai-model/prompt/extraction.js +2 -2
  104. package/dist/lib/ai-model/prompt/extraction.js.map +1 -1
  105. package/dist/lib/ai-model/prompt/llm-locator.js +14 -241
  106. package/dist/lib/ai-model/prompt/llm-locator.js.map +1 -1
  107. package/dist/lib/ai-model/prompt/llm-planning.js +79 -328
  108. package/dist/lib/ai-model/prompt/llm-planning.js.map +1 -1
  109. package/dist/lib/ai-model/prompt/llm-section-locator.js +17 -16
  110. package/dist/lib/ai-model/prompt/llm-section-locator.js.map +1 -1
  111. package/dist/lib/ai-model/prompt/order-sensitive-judge.js +72 -0
  112. package/dist/lib/ai-model/prompt/order-sensitive-judge.js.map +1 -0
  113. package/dist/lib/ai-model/prompt/playwright-generator.js +11 -11
  114. package/dist/lib/ai-model/prompt/playwright-generator.js.map +1 -1
  115. package/dist/lib/ai-model/prompt/ui-tars-locator.js +2 -2
  116. package/dist/lib/ai-model/prompt/ui-tars-locator.js.map +1 -1
  117. package/dist/lib/ai-model/prompt/ui-tars-planning.js +2 -2
  118. package/dist/lib/ai-model/prompt/ui-tars-planning.js.map +1 -1
  119. package/dist/lib/ai-model/prompt/util.js +7 -95
  120. package/dist/lib/ai-model/prompt/util.js.map +1 -1
  121. package/dist/lib/ai-model/prompt/yaml-generator.js +18 -18
  122. package/dist/lib/ai-model/prompt/yaml-generator.js.map +1 -1
  123. package/dist/lib/ai-model/service-caller/index.js +288 -401
  124. package/dist/lib/ai-model/service-caller/index.js.map +1 -1
  125. package/dist/lib/ai-model/ui-tars-planning.js +71 -10
  126. package/dist/lib/ai-model/ui-tars-planning.js.map +1 -1
  127. package/dist/lib/{ai-model/common.js → common.js} +40 -55
  128. package/dist/lib/common.js.map +1 -0
  129. package/dist/lib/device/device-options.js +20 -0
  130. package/dist/lib/device/device-options.js.map +1 -0
  131. package/dist/lib/device/index.js +63 -40
  132. package/dist/lib/device/index.js.map +1 -1
  133. package/dist/lib/image/index.js +5 -5
  134. package/dist/lib/image/index.js.map +1 -1
  135. package/dist/lib/index.js +24 -20
  136. package/dist/lib/index.js.map +1 -1
  137. package/dist/lib/report.js +2 -2
  138. package/dist/lib/report.js.map +1 -1
  139. package/dist/lib/{insight → service}/index.js +41 -54
  140. package/dist/lib/service/index.js.map +1 -0
  141. package/dist/lib/{insight → service}/utils.js +7 -7
  142. package/dist/lib/service/utils.js.map +1 -0
  143. package/dist/lib/task-runner.js +301 -0
  144. package/dist/lib/task-runner.js.map +1 -0
  145. package/dist/lib/tree.js +13 -4
  146. package/dist/lib/tree.js.map +1 -1
  147. package/dist/lib/types.js +31 -12
  148. package/dist/lib/types.js.map +1 -1
  149. package/dist/lib/utils.js +16 -17
  150. package/dist/lib/utils.js.map +1 -1
  151. package/dist/lib/yaml/builder.js +2 -2
  152. package/dist/lib/yaml/builder.js.map +1 -1
  153. package/dist/lib/yaml/index.js +16 -22
  154. package/dist/lib/yaml/index.js.map +1 -1
  155. package/dist/lib/yaml/player.js +123 -100
  156. package/dist/lib/yaml/player.js.map +1 -1
  157. package/dist/lib/yaml/utils.js +6 -6
  158. package/dist/lib/yaml/utils.js.map +1 -1
  159. package/dist/lib/yaml.js +1 -1
  160. package/dist/lib/yaml.js.map +1 -1
  161. package/dist/types/agent/agent.d.ts +62 -17
  162. package/dist/types/agent/execution-session.d.ts +36 -0
  163. package/dist/types/agent/index.d.ts +3 -2
  164. package/dist/types/agent/task-builder.d.ts +35 -0
  165. package/dist/types/agent/tasks.d.ts +32 -23
  166. package/dist/types/agent/ui-utils.d.ts +9 -2
  167. package/dist/types/agent/utils.d.ts +9 -35
  168. package/dist/types/ai-model/conversation-history.d.ts +8 -4
  169. package/dist/types/ai-model/index.d.ts +5 -5
  170. package/dist/types/ai-model/inspect.d.ts +20 -12
  171. package/dist/types/ai-model/llm-planning.d.ts +3 -1
  172. package/dist/types/ai-model/prompt/llm-locator.d.ts +1 -6
  173. package/dist/types/ai-model/prompt/llm-planning.d.ts +2 -3
  174. package/dist/types/ai-model/prompt/llm-section-locator.d.ts +1 -3
  175. package/dist/types/ai-model/prompt/order-sensitive-judge.d.ts +2 -0
  176. package/dist/types/ai-model/prompt/util.d.ts +2 -34
  177. package/dist/types/ai-model/service-caller/index.d.ts +2 -3
  178. package/dist/types/ai-model/ui-tars-planning.d.ts +15 -2
  179. package/dist/types/{ai-model/common.d.ts → common.d.ts} +6 -6
  180. package/dist/types/device/device-options.d.ts +57 -0
  181. package/dist/types/device/index.d.ts +55 -39
  182. package/dist/types/index.d.ts +7 -6
  183. package/dist/types/service/index.d.ts +26 -0
  184. package/dist/types/service/utils.d.ts +2 -0
  185. package/dist/types/task-runner.d.ts +49 -0
  186. package/dist/types/tree.d.ts +4 -1
  187. package/dist/types/types.d.ts +103 -66
  188. package/dist/types/yaml/utils.d.ts +1 -1
  189. package/dist/types/yaml.d.ts +68 -43
  190. package/package.json +9 -12
  191. package/dist/es/ai-model/action-executor.mjs +0 -129
  192. package/dist/es/ai-model/action-executor.mjs.map +0 -1
  193. package/dist/es/ai-model/common.mjs.map +0 -1
  194. package/dist/es/insight/index.mjs.map +0 -1
  195. package/dist/es/insight/utils.mjs.map +0 -1
  196. package/dist/lib/ai-model/action-executor.js +0 -163
  197. package/dist/lib/ai-model/action-executor.js.map +0 -1
  198. package/dist/lib/ai-model/common.js.map +0 -1
  199. package/dist/lib/insight/index.js.map +0 -1
  200. package/dist/lib/insight/utils.js.map +0 -1
  201. package/dist/types/ai-model/action-executor.d.ts +0 -19
  202. package/dist/types/insight/index.d.ts +0 -31
  203. package/dist/types/insight/utils.d.ts +0 -2
@@ -1 +1 @@
1
- {"version":3,"file":"ai-model/prompt/llm-section-locator.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/prompt/llm-section-locator.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 { TVlModeTypes } from '@midscene/shared/env';\nimport { bboxDescription } from './common';\n\nexport function systemPromptToLocateSection(vlMode: TVlModeTypes | undefined) {\n return `\nYou goal is to find out one section containing the target element in the screenshot, put it in the \\`bbox\\` field. If the user describe the target element with some reference elements, you should also find the section containing the reference elements, put it in the \\`references_bbox\\` field.\n\nUsually, it should be approximately an area not more than 300x300px. Changes of the size are allowed if there are many elements to cover.\n\nreturn in this JSON format:\n\\`\\`\\`json\n{\n \"bbox\": [number, number, number, number],\n \"references_bbox\"?: [\n [number, number, number, number],\n [number, number, number, number],\n ...\n ],\n \"error\"?: string\n}\n\\`\\`\\`\n\nIn which, all the numbers in the \\`bbox\\` and \\`references_bbox\\` represent ${bboxDescription(vlMode)}.\n\nFor example, if the user describe the target element as \"the delete button on the second row with title 'Peter'\", you should put the bounding box of the delete button in the \\`bbox\\` field, and the bounding box of the second row in the \\`references_bbox\\` field.\n\nthe return value should be like this:\n\\`\\`\\`json\n{\n \"bbox\": [100, 100, 200, 200],\n \"references_bbox\": [[100, 100, 200, 200]]\n}\n\\`\\`\\`\n`;\n}\n\nexport const sectionLocatorInstruction = ({\n sectionDescription,\n}: {\n sectionDescription: string;\n}) => `Here is the target element user interested in:\n<targetDescription>\n${sectionDescription}\n</targetDescription>\n `;\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","systemPromptToLocateSection","vlMode","bboxDescription","sectionLocatorInstruction","sectionDescription"],"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;;;;;;;;;ACHO,SAASI,4BAA4BC,MAAgC;IAC1E,OAAO,CAAC;;;;;;;;;;;;;;;;;;4EAkBkE,EAAEC,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EAAgBD,QAAQ;;;;;;;;;;;AAWtG,CAAC;AACD;AAEO,MAAME,4BAA4B,CAAC,EACxCC,kBAAkB,EAGnB,GAAK,CAAC;;AAEP,EAAEA,mBAAmB;;EAEnB,CAAC"}
1
+ {"version":3,"file":"ai-model/prompt/llm-section-locator.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../../src/ai-model/prompt/llm-section-locator.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 { TVlModeTypes } from '@midscene/shared/env';\nimport { bboxDescription } from './common';\n\nexport function systemPromptToLocateSection(vlMode: TVlModeTypes | undefined) {\n const bboxFormat = bboxDescription(vlMode);\n return `\n## Role:\nYou are an AI assistant that helps identify UI elements.\n\n## Objective:\n- Find a section containing the target element\n- If the description mentions reference elements, also locate sections containing those references\n\n## Output Format:\n\\`\\`\\`json\n{\n \"bbox\": [number, number, number, number], // ${bboxFormat}\n \"references_bbox\"?: [\n [number, number, number, number],\n ...\n ],\n \"error\"?: string\n}\n\\`\\`\\`\n\nFields:\n* \\`bbox\\` - Bounding box of the section containing the target element\n* \\`references_bbox\\` - Optional array of bounding boxes for reference elements\n* \\`error\\` - Optional error message if the section cannot be found\n\nExample:\nIf the description is \"delete button on the second row with title 'Peter'\", return:\n\\`\\`\\`json\n{\n \"bbox\": [100, 100, 200, 200],\n \"references_bbox\": [[100, 100, 200, 200]]\n}\n\\`\\`\\`\n`;\n}\n\nexport const sectionLocatorInstruction = (sectionDescription: string) =>\n `Find section containing: ${sectionDescription}`;\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","systemPromptToLocateSection","vlMode","bboxFormat","bboxDescription","sectionLocatorInstruction","sectionDescription"],"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;;;;;;;;;ACHO,SAASI,4BAA4BC,MAAgC;IAC1E,MAAMC,aAAaC,AAAAA,IAAAA,mCAAAA,eAAAA,AAAAA,EAAgBF;IACnC,OAAO,CAAC;;;;;;;;;;;gDAWsC,EAAEC,WAAW;;;;;;;;;;;;;;;;;;;;;;AAsB7D,CAAC;AACD;AAEO,MAAME,4BAA4B,CAACC,qBACxC,CAAC,yBAAyB,EAAEA,oBAAoB"}
@@ -0,0 +1,72 @@
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
+ orderSensitiveJudgePrompt: ()=>orderSensitiveJudgePrompt,
28
+ systemPromptToJudgeOrderSensitive: ()=>systemPromptToJudgeOrderSensitive
29
+ });
30
+ function systemPromptToJudgeOrderSensitive() {
31
+ return `
32
+ ## Role:
33
+ You are an AI assistant that analyzes UI element descriptions.
34
+
35
+ ## Objective:
36
+ Determine whether a given element description is order-sensitive.
37
+
38
+ Order-sensitive descriptions contain phrases that specify position or sequence, such as:
39
+ - "the first button"
40
+ - "the second item"
41
+ - "the third row"
42
+ - "the last input"
43
+ - "the 5th element"
44
+
45
+ Order-insensitive descriptions do not specify position:
46
+ - "login button"
47
+ - "search input"
48
+ - "submit button"
49
+ - "user avatar"
50
+
51
+ ## Output Format:
52
+ \`\`\`json
53
+ {
54
+ "isOrderSensitive": boolean
55
+ }
56
+ \`\`\`
57
+
58
+ Return true if the description is order-sensitive, false otherwise.
59
+ `;
60
+ }
61
+ const orderSensitiveJudgePrompt = (description)=>`Analyze this element description: "${description}"`;
62
+ exports.orderSensitiveJudgePrompt = __webpack_exports__.orderSensitiveJudgePrompt;
63
+ exports.systemPromptToJudgeOrderSensitive = __webpack_exports__.systemPromptToJudgeOrderSensitive;
64
+ for(var __rspack_i in __webpack_exports__)if (-1 === [
65
+ "orderSensitiveJudgePrompt",
66
+ "systemPromptToJudgeOrderSensitive"
67
+ ].indexOf(__rspack_i)) exports[__rspack_i] = __webpack_exports__[__rspack_i];
68
+ Object.defineProperty(exports, '__esModule', {
69
+ value: true
70
+ });
71
+
72
+ //# sourceMappingURL=order-sensitive-judge.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"ai-model/prompt/order-sensitive-judge.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../../src/ai-model/prompt/order-sensitive-judge.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};","export function systemPromptToJudgeOrderSensitive() {\n return `\n## Role:\nYou are an AI assistant that analyzes UI element descriptions.\n\n## Objective:\nDetermine whether a given element description is order-sensitive.\n\nOrder-sensitive descriptions contain phrases that specify position or sequence, such as:\n- \"the first button\"\n- \"the second item\"\n- \"the third row\"\n- \"the last input\"\n- \"the 5th element\"\n\nOrder-insensitive descriptions do not specify position:\n- \"login button\"\n- \"search input\"\n- \"submit button\"\n- \"user avatar\"\n\n## Output Format:\n\\`\\`\\`json\n{\n \"isOrderSensitive\": boolean\n}\n\\`\\`\\`\n\nReturn true if the description is order-sensitive, false otherwise.\n`;\n}\n\nexport const orderSensitiveJudgePrompt = (description: string) => {\n return `Analyze this element description: \"${description}\"`;\n};\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","systemPromptToJudgeOrderSensitive","orderSensitiveJudgePrompt","description"],"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;;;;;;;;ACNO,SAASI;IACd,OAAO,CAAC;;;;;;;;;;;;;;;;;;;;;;;;;;;;AA4BV,CAAC;AACD;AAEO,MAAMC,4BAA4B,CAACC,cACjC,CAAC,mCAAmC,EAAEA,YAAY,CAAC,CAAC"}
@@ -24,16 +24,16 @@ var __webpack_require__ = {};
24
24
  var __webpack_exports__ = {};
25
25
  __webpack_require__.r(__webpack_exports__);
26
26
  __webpack_require__.d(__webpack_exports__, {
27
- getScreenshotsForLLM: ()=>external_yaml_generator_js_namespaceObject.getScreenshotsForLLM,
28
- createMessageContent: ()=>external_yaml_generator_js_namespaceObject.createMessageContent,
27
+ createEventCounts: ()=>external_yaml_generator_js_namespaceObject.createEventCounts,
29
28
  filterEventsByType: ()=>external_yaml_generator_js_namespaceObject.filterEventsByType,
30
29
  generatePlaywrightTest: ()=>generatePlaywrightTest,
31
- prepareEventSummary: ()=>external_yaml_generator_js_namespaceObject.prepareEventSummary,
32
- processEventsForLLM: ()=>external_yaml_generator_js_namespaceObject.processEventsForLLM,
33
- extractInputDescriptions: ()=>external_yaml_generator_js_namespaceObject.extractInputDescriptions,
34
- validateEvents: ()=>external_yaml_generator_js_namespaceObject.validateEvents,
35
30
  generatePlaywrightTestStream: ()=>generatePlaywrightTestStream,
36
- createEventCounts: ()=>external_yaml_generator_js_namespaceObject.createEventCounts
31
+ validateEvents: ()=>external_yaml_generator_js_namespaceObject.validateEvents,
32
+ createMessageContent: ()=>external_yaml_generator_js_namespaceObject.createMessageContent,
33
+ getScreenshotsForLLM: ()=>external_yaml_generator_js_namespaceObject.getScreenshotsForLLM,
34
+ extractInputDescriptions: ()=>external_yaml_generator_js_namespaceObject.extractInputDescriptions,
35
+ prepareEventSummary: ()=>external_yaml_generator_js_namespaceObject.prepareEventSummary,
36
+ processEventsForLLM: ()=>external_yaml_generator_js_namespaceObject.processEventsForLLM
37
37
  });
38
38
  const constants_namespaceObject = require("@midscene/shared/constants");
39
39
  const external_index_js_namespaceObject = require("../index.js");
@@ -85,7 +85,7 @@ ${constants_namespaceObject.PLAYWRIGHT_EXAMPLE_CODE}`;
85
85
  }
86
86
  ];
87
87
  const response = await (0, external_index_js_namespaceObject.callAIWithStringResponse)(prompt, external_index_js_namespaceObject.AIActionType.TEXT, modelConfig);
88
- if ((null == response ? void 0 : response.content) && 'string' == typeof response.content) return response.content;
88
+ if (response?.content && 'string' == typeof response.content) return response.content;
89
89
  throw new Error('Failed to generate Playwright test code');
90
90
  };
91
91
  const generatePlaywrightTestStream = async (events, options, modelConfig)=>{
@@ -141,7 +141,7 @@ ${constants_namespaceObject.PLAYWRIGHT_EXAMPLE_CODE}`;
141
141
  });
142
142
  {
143
143
  const response = await (0, external_index_js_namespaceObject.callAIWithStringResponse)(prompt, external_index_js_namespaceObject.AIActionType.TEXT, modelConfig);
144
- if ((null == response ? void 0 : response.content) && 'string' == typeof response.content) return {
144
+ if (response?.content && 'string' == typeof response.content) return {
145
145
  content: response.content,
146
146
  usage: response.usage,
147
147
  isStreamed: false
@@ -159,7 +159,7 @@ exports.getScreenshotsForLLM = __webpack_exports__.getScreenshotsForLLM;
159
159
  exports.prepareEventSummary = __webpack_exports__.prepareEventSummary;
160
160
  exports.processEventsForLLM = __webpack_exports__.processEventsForLLM;
161
161
  exports.validateEvents = __webpack_exports__.validateEvents;
162
- for(var __webpack_i__ in __webpack_exports__)if (-1 === [
162
+ for(var __rspack_i in __webpack_exports__)if (-1 === [
163
163
  "createEventCounts",
164
164
  "createMessageContent",
165
165
  "extractInputDescriptions",
@@ -170,7 +170,7 @@ for(var __webpack_i__ in __webpack_exports__)if (-1 === [
170
170
  "prepareEventSummary",
171
171
  "processEventsForLLM",
172
172
  "validateEvents"
173
- ].indexOf(__webpack_i__)) exports[__webpack_i__] = __webpack_exports__[__webpack_i__];
173
+ ].indexOf(__rspack_i)) exports[__rspack_i] = __webpack_exports__[__rspack_i];
174
174
  Object.defineProperty(exports, '__esModule', {
175
175
  value: true
176
176
  });
@@ -1 +1 @@
1
- {"version":3,"file":"ai-model/prompt/playwright-generator.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/prompt/playwright-generator.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 StreamingAIResponse,\n StreamingCodeGenerationOptions,\n} from '@/types';\nimport { PLAYWRIGHT_EXAMPLE_CODE } from '@midscene/shared/constants';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport { AIActionType, callAI, callAIWithStringResponse } from '../index';\n// Import shared utilities and types from yaml-generator\nimport {\n type ChromeRecordedEvent,\n type EventCounts,\n type EventSummary,\n type InputDescription,\n type ProcessedEvent,\n createEventCounts,\n createMessageContent,\n extractInputDescriptions,\n filterEventsByType,\n getScreenshotsForLLM,\n prepareEventSummary,\n processEventsForLLM,\n validateEvents,\n} from './yaml-generator';\n\n// Playwright-specific interfaces\nexport interface PlaywrightGenerationOptions {\n testName?: string;\n includeScreenshots?: boolean;\n includeTimestamps?: boolean;\n maxScreenshots?: number;\n description?: string;\n viewportSize?: { width: number; height: number };\n waitForNetworkIdle?: boolean;\n waitForNetworkIdleTimeout?: number;\n}\n\n// Re-export shared types for backward compatibility\nexport type {\n ChromeRecordedEvent,\n EventCounts,\n InputDescription,\n ProcessedEvent,\n EventSummary,\n};\n\n// Re-export shared utilities for backward compatibility\nexport {\n getScreenshotsForLLM,\n filterEventsByType,\n createEventCounts,\n extractInputDescriptions,\n processEventsForLLM,\n prepareEventSummary,\n createMessageContent,\n validateEvents,\n};\n\n/**\n * Generates Playwright test code from recorded events\n */\nexport const generatePlaywrightTest = async (\n events: ChromeRecordedEvent[],\n options: PlaywrightGenerationOptions,\n modelConfig: IModelConfig,\n): Promise<string> => {\n // Validate input\n validateEvents(events);\n\n // Prepare event summary using shared utilities\n const summary = prepareEventSummary(events, {\n testName: options.testName,\n maxScreenshots: options.maxScreenshots || 3,\n });\n\n // Add Playwright-specific options to summary\n const playwrightSummary = {\n ...summary,\n waitForNetworkIdle: options.waitForNetworkIdle !== false,\n waitForNetworkIdleTimeout: options.waitForNetworkIdleTimeout || 2000,\n viewportSize: options.viewportSize || { width: 1280, height: 800 },\n };\n\n // Get screenshots for visual context\n const screenshots = getScreenshotsForLLM(events, options.maxScreenshots || 3);\n\n // Create prompt text\n const promptText = `Generate a Playwright test using @midscene/web/playwright that reproduces this recorded browser session. The test should be based on the following events and follow the structure of the example provided. Make the test descriptive with appropriate assertions and validations.\n\nEvent Summary:\n${JSON.stringify(playwrightSummary, null, 2)}\n\nGenerated code should:\n1. Import required dependencies\n2. Set up the test with proper configuration\n3. Include a beforeEach hook to navigate to the starting URL\n4. Implement a test that uses Midscene AI methods (aiTap, aiInput, aiAssert, etc.)\n5. Include appropriate assertions and validations\n6. Follow best practices for Playwright tests\n7. Be ready to execute without further modification\n\nImportant: Return ONLY the raw Playwright test code. Do NOT wrap the response in markdown code blocks (no \\`\\`\\`typescript, \\`\\`\\`javascript or \\`\\`\\`). Start directly with the code content.`;\n\n // Create message content with screenshots\n const messageContent = createMessageContent(\n promptText,\n screenshots,\n options.includeScreenshots !== false,\n );\n\n // Create system prompt\n const systemPrompt = `You are an expert test automation engineer specializing in Playwright and Midscene. \nYour task is to generate a complete, executable Playwright test using @midscene/web/playwright that reproduces a recorded browser session.\n\n${PLAYWRIGHT_EXAMPLE_CODE}`;\n\n // Use LLM to generate the Playwright test code\n const prompt: ChatCompletionMessageParam[] = [\n {\n role: 'system',\n content: systemPrompt,\n },\n {\n role: 'user',\n content: messageContent,\n },\n ];\n\n const response = await callAIWithStringResponse(\n prompt,\n AIActionType.TEXT,\n modelConfig,\n );\n\n if (response?.content && typeof response.content === 'string') {\n return response.content;\n }\n\n throw new Error('Failed to generate Playwright test code');\n};\n\n/**\n * Generates Playwright test code from recorded events with streaming support\n */\nexport const generatePlaywrightTestStream = async (\n events: ChromeRecordedEvent[],\n options: PlaywrightGenerationOptions & StreamingCodeGenerationOptions,\n modelConfig: IModelConfig,\n): Promise<StreamingAIResponse> => {\n // Validate input\n validateEvents(events);\n\n // Prepare event summary using shared utilities\n const summary = prepareEventSummary(events, {\n testName: options.testName,\n maxScreenshots: options.maxScreenshots || 3,\n });\n\n // Add Playwright-specific options to summary\n const playwrightSummary = {\n ...summary,\n waitForNetworkIdle: options.waitForNetworkIdle !== false,\n waitForNetworkIdleTimeout: options.waitForNetworkIdleTimeout || 2000,\n viewportSize: options.viewportSize || { width: 1280, height: 800 },\n };\n\n // Get screenshots for visual context\n const screenshots = getScreenshotsForLLM(events, options.maxScreenshots || 3);\n\n // Create prompt text\n const promptText = `Generate a Playwright test using @midscene/web/playwright that reproduces this recorded browser session. The test should be based on the following events and follow the structure of the example provided. Make the test descriptive with appropriate assertions and validations.\n\nEvent Summary:\n${JSON.stringify(playwrightSummary, null, 2)}\n\nGenerated code should:\n1. Import required dependencies\n2. Set up the test with proper configuration\n3. Include a beforeEach hook to navigate to the starting URL\n4. Implement a test that uses Midscene AI methods (aiTap, aiInput, aiAssert, etc.)\n5. Include appropriate assertions and validations\n6. Follow best practices for Playwright tests\n7. Be ready to execute without further modification\n8. can't wrap this test code in markdown code block\n\nImportant: Return ONLY the raw Playwright test code. Do NOT wrap the response in markdown code blocks (no \\`\\`\\`typescript, \\`\\`\\`javascript or \\`\\`\\`). Start directly with the code content.`;\n\n // Create message content with screenshots\n const messageContent = createMessageContent(\n promptText,\n screenshots,\n options.includeScreenshots !== false,\n );\n\n // Create system prompt\n const systemPrompt = `You are an expert test automation engineer specializing in Playwright and Midscene. \nYour task is to generate a complete, executable Playwright test using @midscene/web/playwright that reproduces a recorded browser session.\n\n${PLAYWRIGHT_EXAMPLE_CODE}`;\n\n // Use LLM to generate the Playwright test code with streaming\n const prompt: ChatCompletionMessageParam[] = [\n {\n role: 'system',\n content: systemPrompt,\n },\n {\n role: 'user',\n content: messageContent,\n },\n ];\n\n if (options.stream && options.onChunk) {\n // Use streaming\n return await callAI(prompt, AIActionType.TEXT, modelConfig, {\n stream: true,\n onChunk: options.onChunk,\n });\n } else {\n // Fallback to non-streaming\n const response = await callAIWithStringResponse(\n prompt,\n AIActionType.TEXT,\n modelConfig,\n );\n\n if (response?.content && typeof response.content === 'string') {\n return {\n content: response.content,\n usage: response.usage,\n isStreamed: false,\n };\n }\n\n throw new Error('Failed to generate Playwright test code');\n }\n};\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","generatePlaywrightTest","events","options","modelConfig","validateEvents","summary","prepareEventSummary","playwrightSummary","screenshots","getScreenshotsForLLM","promptText","JSON","messageContent","createMessageContent","systemPrompt","PLAYWRIGHT_EXAMPLE_CODE","prompt","response","callAIWithStringResponse","AIActionType","Error","generatePlaywrightTestStream","callAI"],"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;;;;;;;;;;;;;;;;;;;ACuDO,MAAMI,yBAAyB,OACpCC,QACAC,SACAC;IAGAC,IAAAA,2CAAAA,cAAAA,AAAAA,EAAeH;IAGf,MAAMI,UAAUC,AAAAA,IAAAA,2CAAAA,mBAAAA,AAAAA,EAAoBL,QAAQ;QAC1C,UAAUC,QAAQ,QAAQ;QAC1B,gBAAgBA,QAAQ,cAAc,IAAI;IAC5C;IAGA,MAAMK,oBAAoB;QACxB,GAAGF,OAAO;QACV,oBAAoBH,AAA+B,UAA/BA,QAAQ,kBAAkB;QAC9C,2BAA2BA,QAAQ,yBAAyB,IAAI;QAChE,cAAcA,QAAQ,YAAY,IAAI;YAAE,OAAO;YAAM,QAAQ;QAAI;IACnE;IAGA,MAAMM,cAAcC,AAAAA,IAAAA,2CAAAA,oBAAAA,AAAAA,EAAqBR,QAAQC,QAAQ,cAAc,IAAI;IAG3E,MAAMQ,aAAa,CAAC;;;AAGtB,EAAEC,KAAK,SAAS,CAACJ,mBAAmB,MAAM,GAAG;;;;;;;;;;;8LAWiJ,CAAC;IAG7L,MAAMK,iBAAiBC,AAAAA,IAAAA,2CAAAA,oBAAAA,AAAAA,EACrBH,YACAF,aACAN,AAA+B,UAA/BA,QAAQ,kBAAkB;IAI5B,MAAMY,eAAe,CAAC;;;AAGxB,EAAEC,0BAAAA,uBAAuBA,EAAE;IAGzB,MAAMC,SAAuC;QAC3C;YACE,MAAM;YACN,SAASF;QACX;QACA;YACE,MAAM;YACN,SAASF;QACX;KACD;IAED,MAAMK,WAAW,MAAMC,AAAAA,IAAAA,kCAAAA,wBAAAA,AAAAA,EACrBF,QACAG,kCAAAA,YAAAA,CAAAA,IAAiB,EACjBhB;IAGF,IAAIc,AAAAA,CAAAA,QAAAA,WAAAA,KAAAA,IAAAA,SAAU,OAAO,AAAD,KAAK,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAOA,SAAS,OAAO;IAGzB,MAAM,IAAIG,MAAM;AAClB;AAKO,MAAMC,+BAA+B,OAC1CpB,QACAC,SACAC;IAGAC,IAAAA,2CAAAA,cAAAA,AAAAA,EAAeH;IAGf,MAAMI,UAAUC,AAAAA,IAAAA,2CAAAA,mBAAAA,AAAAA,EAAoBL,QAAQ;QAC1C,UAAUC,QAAQ,QAAQ;QAC1B,gBAAgBA,QAAQ,cAAc,IAAI;IAC5C;IAGA,MAAMK,oBAAoB;QACxB,GAAGF,OAAO;QACV,oBAAoBH,AAA+B,UAA/BA,QAAQ,kBAAkB;QAC9C,2BAA2BA,QAAQ,yBAAyB,IAAI;QAChE,cAAcA,QAAQ,YAAY,IAAI;YAAE,OAAO;YAAM,QAAQ;QAAI;IACnE;IAGA,MAAMM,cAAcC,AAAAA,IAAAA,2CAAAA,oBAAAA,AAAAA,EAAqBR,QAAQC,QAAQ,cAAc,IAAI;IAG3E,MAAMQ,aAAa,CAAC;;;AAGtB,EAAEC,KAAK,SAAS,CAACJ,mBAAmB,MAAM,GAAG;;;;;;;;;;;;8LAYiJ,CAAC;IAG7L,MAAMK,iBAAiBC,AAAAA,IAAAA,2CAAAA,oBAAAA,AAAAA,EACrBH,YACAF,aACAN,AAA+B,UAA/BA,QAAQ,kBAAkB;IAI5B,MAAMY,eAAe,CAAC;;;AAGxB,EAAEC,0BAAAA,uBAAuBA,EAAE;IAGzB,MAAMC,SAAuC;QAC3C;YACE,MAAM;YACN,SAASF;QACX;QACA;YACE,MAAM;YACN,SAASF;QACX;KACD;IAED,IAAIV,QAAQ,MAAM,IAAIA,QAAQ,OAAO,EAEnC,OAAO,MAAMoB,AAAAA,IAAAA,kCAAAA,MAAAA,AAAAA,EAAON,QAAQG,kCAAAA,YAAAA,CAAAA,IAAiB,EAAEhB,aAAa;QAC1D,QAAQ;QACR,SAASD,QAAQ,OAAO;IAC1B;IACK;QAEL,MAAMe,WAAW,MAAMC,AAAAA,IAAAA,kCAAAA,wBAAAA,AAAAA,EACrBF,QACAG,kCAAAA,YAAAA,CAAAA,IAAiB,EACjBhB;QAGF,IAAIc,AAAAA,CAAAA,QAAAA,WAAAA,KAAAA,IAAAA,SAAU,OAAO,AAAD,KAAK,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAO;YACL,SAASA,SAAS,OAAO;YACzB,OAAOA,SAAS,KAAK;YACrB,YAAY;QACd;QAGF,MAAM,IAAIG,MAAM;IAClB;AACF"}
1
+ {"version":3,"file":"ai-model/prompt/playwright-generator.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../../src/ai-model/prompt/playwright-generator.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 StreamingAIResponse,\n StreamingCodeGenerationOptions,\n} from '@/types';\nimport { PLAYWRIGHT_EXAMPLE_CODE } from '@midscene/shared/constants';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport { AIActionType, callAI, callAIWithStringResponse } from '../index';\n// Import shared utilities and types from yaml-generator\nimport {\n type ChromeRecordedEvent,\n type EventCounts,\n type EventSummary,\n type InputDescription,\n type ProcessedEvent,\n createEventCounts,\n createMessageContent,\n extractInputDescriptions,\n filterEventsByType,\n getScreenshotsForLLM,\n prepareEventSummary,\n processEventsForLLM,\n validateEvents,\n} from './yaml-generator';\n\n// Playwright-specific interfaces\nexport interface PlaywrightGenerationOptions {\n testName?: string;\n includeScreenshots?: boolean;\n includeTimestamps?: boolean;\n maxScreenshots?: number;\n description?: string;\n viewportSize?: { width: number; height: number };\n waitForNetworkIdle?: boolean;\n waitForNetworkIdleTimeout?: number;\n}\n\n// Re-export shared types for backward compatibility\nexport type {\n ChromeRecordedEvent,\n EventCounts,\n InputDescription,\n ProcessedEvent,\n EventSummary,\n};\n\n// Re-export shared utilities for backward compatibility\nexport {\n getScreenshotsForLLM,\n filterEventsByType,\n createEventCounts,\n extractInputDescriptions,\n processEventsForLLM,\n prepareEventSummary,\n createMessageContent,\n validateEvents,\n};\n\n/**\n * Generates Playwright test code from recorded events\n */\nexport const generatePlaywrightTest = async (\n events: ChromeRecordedEvent[],\n options: PlaywrightGenerationOptions,\n modelConfig: IModelConfig,\n): Promise<string> => {\n // Validate input\n validateEvents(events);\n\n // Prepare event summary using shared utilities\n const summary = prepareEventSummary(events, {\n testName: options.testName,\n maxScreenshots: options.maxScreenshots || 3,\n });\n\n // Add Playwright-specific options to summary\n const playwrightSummary = {\n ...summary,\n waitForNetworkIdle: options.waitForNetworkIdle !== false,\n waitForNetworkIdleTimeout: options.waitForNetworkIdleTimeout || 2000,\n viewportSize: options.viewportSize || { width: 1280, height: 800 },\n };\n\n // Get screenshots for visual context\n const screenshots = getScreenshotsForLLM(events, options.maxScreenshots || 3);\n\n // Create prompt text\n const promptText = `Generate a Playwright test using @midscene/web/playwright that reproduces this recorded browser session. The test should be based on the following events and follow the structure of the example provided. Make the test descriptive with appropriate assertions and validations.\n\nEvent Summary:\n${JSON.stringify(playwrightSummary, null, 2)}\n\nGenerated code should:\n1. Import required dependencies\n2. Set up the test with proper configuration\n3. Include a beforeEach hook to navigate to the starting URL\n4. Implement a test that uses Midscene AI methods (aiTap, aiInput, aiAssert, etc.)\n5. Include appropriate assertions and validations\n6. Follow best practices for Playwright tests\n7. Be ready to execute without further modification\n\nImportant: Return ONLY the raw Playwright test code. Do NOT wrap the response in markdown code blocks (no \\`\\`\\`typescript, \\`\\`\\`javascript or \\`\\`\\`). Start directly with the code content.`;\n\n // Create message content with screenshots\n const messageContent = createMessageContent(\n promptText,\n screenshots,\n options.includeScreenshots !== false,\n );\n\n // Create system prompt\n const systemPrompt = `You are an expert test automation engineer specializing in Playwright and Midscene. \nYour task is to generate a complete, executable Playwright test using @midscene/web/playwright that reproduces a recorded browser session.\n\n${PLAYWRIGHT_EXAMPLE_CODE}`;\n\n // Use LLM to generate the Playwright test code\n const prompt: ChatCompletionMessageParam[] = [\n {\n role: 'system',\n content: systemPrompt,\n },\n {\n role: 'user',\n content: messageContent,\n },\n ];\n\n const response = await callAIWithStringResponse(\n prompt,\n AIActionType.TEXT,\n modelConfig,\n );\n\n if (response?.content && typeof response.content === 'string') {\n return response.content;\n }\n\n throw new Error('Failed to generate Playwright test code');\n};\n\n/**\n * Generates Playwright test code from recorded events with streaming support\n */\nexport const generatePlaywrightTestStream = async (\n events: ChromeRecordedEvent[],\n options: PlaywrightGenerationOptions & StreamingCodeGenerationOptions,\n modelConfig: IModelConfig,\n): Promise<StreamingAIResponse> => {\n // Validate input\n validateEvents(events);\n\n // Prepare event summary using shared utilities\n const summary = prepareEventSummary(events, {\n testName: options.testName,\n maxScreenshots: options.maxScreenshots || 3,\n });\n\n // Add Playwright-specific options to summary\n const playwrightSummary = {\n ...summary,\n waitForNetworkIdle: options.waitForNetworkIdle !== false,\n waitForNetworkIdleTimeout: options.waitForNetworkIdleTimeout || 2000,\n viewportSize: options.viewportSize || { width: 1280, height: 800 },\n };\n\n // Get screenshots for visual context\n const screenshots = getScreenshotsForLLM(events, options.maxScreenshots || 3);\n\n // Create prompt text\n const promptText = `Generate a Playwright test using @midscene/web/playwright that reproduces this recorded browser session. The test should be based on the following events and follow the structure of the example provided. Make the test descriptive with appropriate assertions and validations.\n\nEvent Summary:\n${JSON.stringify(playwrightSummary, null, 2)}\n\nGenerated code should:\n1. Import required dependencies\n2. Set up the test with proper configuration\n3. Include a beforeEach hook to navigate to the starting URL\n4. Implement a test that uses Midscene AI methods (aiTap, aiInput, aiAssert, etc.)\n5. Include appropriate assertions and validations\n6. Follow best practices for Playwright tests\n7. Be ready to execute without further modification\n8. can't wrap this test code in markdown code block\n\nImportant: Return ONLY the raw Playwright test code. Do NOT wrap the response in markdown code blocks (no \\`\\`\\`typescript, \\`\\`\\`javascript or \\`\\`\\`). Start directly with the code content.`;\n\n // Create message content with screenshots\n const messageContent = createMessageContent(\n promptText,\n screenshots,\n options.includeScreenshots !== false,\n );\n\n // Create system prompt\n const systemPrompt = `You are an expert test automation engineer specializing in Playwright and Midscene. \nYour task is to generate a complete, executable Playwright test using @midscene/web/playwright that reproduces a recorded browser session.\n\n${PLAYWRIGHT_EXAMPLE_CODE}`;\n\n // Use LLM to generate the Playwright test code with streaming\n const prompt: ChatCompletionMessageParam[] = [\n {\n role: 'system',\n content: systemPrompt,\n },\n {\n role: 'user',\n content: messageContent,\n },\n ];\n\n if (options.stream && options.onChunk) {\n // Use streaming\n return await callAI(prompt, AIActionType.TEXT, modelConfig, {\n stream: true,\n onChunk: options.onChunk,\n });\n } else {\n // Fallback to non-streaming\n const response = await callAIWithStringResponse(\n prompt,\n AIActionType.TEXT,\n modelConfig,\n );\n\n if (response?.content && typeof response.content === 'string') {\n return {\n content: response.content,\n usage: response.usage,\n isStreamed: false,\n };\n }\n\n throw new Error('Failed to generate Playwright test code');\n }\n};\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","generatePlaywrightTest","events","options","modelConfig","validateEvents","summary","prepareEventSummary","playwrightSummary","screenshots","getScreenshotsForLLM","promptText","JSON","messageContent","createMessageContent","systemPrompt","PLAYWRIGHT_EXAMPLE_CODE","prompt","response","callAIWithStringResponse","AIActionType","Error","generatePlaywrightTestStream","callAI"],"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;;;;;;;;;;;;;;;;;;;ACuDO,MAAMI,yBAAyB,OACpCC,QACAC,SACAC;IAGAC,IAAAA,2CAAAA,cAAAA,AAAAA,EAAeH;IAGf,MAAMI,UAAUC,AAAAA,IAAAA,2CAAAA,mBAAAA,AAAAA,EAAoBL,QAAQ;QAC1C,UAAUC,QAAQ,QAAQ;QAC1B,gBAAgBA,QAAQ,cAAc,IAAI;IAC5C;IAGA,MAAMK,oBAAoB;QACxB,GAAGF,OAAO;QACV,oBAAoBH,AAA+B,UAA/BA,QAAQ,kBAAkB;QAC9C,2BAA2BA,QAAQ,yBAAyB,IAAI;QAChE,cAAcA,QAAQ,YAAY,IAAI;YAAE,OAAO;YAAM,QAAQ;QAAI;IACnE;IAGA,MAAMM,cAAcC,AAAAA,IAAAA,2CAAAA,oBAAAA,AAAAA,EAAqBR,QAAQC,QAAQ,cAAc,IAAI;IAG3E,MAAMQ,aAAa,CAAC;;;AAGtB,EAAEC,KAAK,SAAS,CAACJ,mBAAmB,MAAM,GAAG;;;;;;;;;;;8LAWiJ,CAAC;IAG7L,MAAMK,iBAAiBC,AAAAA,IAAAA,2CAAAA,oBAAAA,AAAAA,EACrBH,YACAF,aACAN,AAA+B,UAA/BA,QAAQ,kBAAkB;IAI5B,MAAMY,eAAe,CAAC;;;AAGxB,EAAEC,0BAAAA,uBAAuBA,EAAE;IAGzB,MAAMC,SAAuC;QAC3C;YACE,MAAM;YACN,SAASF;QACX;QACA;YACE,MAAM;YACN,SAASF;QACX;KACD;IAED,MAAMK,WAAW,MAAMC,AAAAA,IAAAA,kCAAAA,wBAAAA,AAAAA,EACrBF,QACAG,kCAAAA,YAAAA,CAAAA,IAAiB,EACjBhB;IAGF,IAAIc,UAAU,WAAW,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAOA,SAAS,OAAO;IAGzB,MAAM,IAAIG,MAAM;AAClB;AAKO,MAAMC,+BAA+B,OAC1CpB,QACAC,SACAC;IAGAC,IAAAA,2CAAAA,cAAAA,AAAAA,EAAeH;IAGf,MAAMI,UAAUC,AAAAA,IAAAA,2CAAAA,mBAAAA,AAAAA,EAAoBL,QAAQ;QAC1C,UAAUC,QAAQ,QAAQ;QAC1B,gBAAgBA,QAAQ,cAAc,IAAI;IAC5C;IAGA,MAAMK,oBAAoB;QACxB,GAAGF,OAAO;QACV,oBAAoBH,AAA+B,UAA/BA,QAAQ,kBAAkB;QAC9C,2BAA2BA,QAAQ,yBAAyB,IAAI;QAChE,cAAcA,QAAQ,YAAY,IAAI;YAAE,OAAO;YAAM,QAAQ;QAAI;IACnE;IAGA,MAAMM,cAAcC,AAAAA,IAAAA,2CAAAA,oBAAAA,AAAAA,EAAqBR,QAAQC,QAAQ,cAAc,IAAI;IAG3E,MAAMQ,aAAa,CAAC;;;AAGtB,EAAEC,KAAK,SAAS,CAACJ,mBAAmB,MAAM,GAAG;;;;;;;;;;;;8LAYiJ,CAAC;IAG7L,MAAMK,iBAAiBC,AAAAA,IAAAA,2CAAAA,oBAAAA,AAAAA,EACrBH,YACAF,aACAN,AAA+B,UAA/BA,QAAQ,kBAAkB;IAI5B,MAAMY,eAAe,CAAC;;;AAGxB,EAAEC,0BAAAA,uBAAuBA,EAAE;IAGzB,MAAMC,SAAuC;QAC3C;YACE,MAAM;YACN,SAASF;QACX;QACA;YACE,MAAM;YACN,SAASF;QACX;KACD;IAED,IAAIV,QAAQ,MAAM,IAAIA,QAAQ,OAAO,EAEnC,OAAO,MAAMoB,AAAAA,IAAAA,kCAAAA,MAAAA,AAAAA,EAAON,QAAQG,kCAAAA,YAAAA,CAAAA,IAAiB,EAAEhB,aAAa;QAC1D,QAAQ;QACR,SAASD,QAAQ,OAAO;IAC1B;IACK;QAEL,MAAMe,WAAW,MAAMC,AAAAA,IAAAA,kCAAAA,wBAAAA,AAAAA,EACrBF,QACAG,kCAAAA,YAAAA,CAAAA,IAAiB,EACjBhB;QAGF,IAAIc,UAAU,WAAW,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAO;YACL,SAASA,SAAS,OAAO;YACzB,OAAOA,SAAS,KAAK;YACrB,YAAY;QACd;QAGF,MAAM,IAAIG,MAAM;IAClB;AACF"}
@@ -58,9 +58,9 @@ call_user() # Submit the task and call the user when the task is unsolvable, or
58
58
  `;
59
59
  }
60
60
  exports.systemPromptToLocateElementPosition = __webpack_exports__.systemPromptToLocateElementPosition;
61
- for(var __webpack_i__ in __webpack_exports__)if (-1 === [
61
+ for(var __rspack_i in __webpack_exports__)if (-1 === [
62
62
  "systemPromptToLocateElementPosition"
63
- ].indexOf(__webpack_i__)) exports[__webpack_i__] = __webpack_exports__[__webpack_i__];
63
+ ].indexOf(__rspack_i)) exports[__rspack_i] = __webpack_exports__[__rspack_i];
64
64
  Object.defineProperty(exports, '__esModule', {
65
65
  value: true
66
66
  });
@@ -1 +1 @@
1
- {"version":3,"file":"ai-model/prompt/ui-tars-locator.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/prompt/ui-tars-locator.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 { getPreferredLanguage } from '@midscene/shared/env';\n\n// claude 3.5 sonnet computer The ability to understand the content of the image is better, Does not provide element snapshot effect\nexport function systemPromptToLocateElementPosition() {\n const preferredLanguage = getPreferredLanguage();\n\n return `\nYou are a GUI agent. You are given a task and your action history, with screenshots. You need to perform the next action to complete the task. \n\n## Output Format\n\\`\\`\\`\nThought: ...\nAction: ...\n\\`\\`\\`\n\n## Action Space\nclick(start_box='[x1, y1, x2, y2]')\nleft_double(start_box='[x1, y1, x2, y2]')\nright_single(start_box='[x1, y1, x2, y2]')\ndrag(start_box='[x1, y1, x2, y2]', end_box='[x3, y3, x4, y4]')\nhotkey(key='')\ntype(content='') #If you want to submit your input, use \"\\\\n\" at the end of \\`content\\`.\nscroll(start_box='[x1, y1, x2, y2]', direction='down or up or right or left')\nwait() #Sleep for 5s and take a screenshot to check for any changes.\nfinished()\ncall_user() # Submit the task and call the user when the task is unsolvable, or when you need the user's help.\n\n## Note\n- Use ${preferredLanguage} in \\`Thought\\` part.\n- Write a small plan and finally summarize your next action (with its target element) in one sentence in \\`Thought\\` part.\n\n## User Instruction\n `;\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","systemPromptToLocateElementPosition","preferredLanguage","getPreferredLanguage"],"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;;;;;;;;ACHO,SAASI;IACd,MAAMC,oBAAoBC,AAAAA,IAAAA,oBAAAA,oBAAAA,AAAAA;IAE1B,OAAO,CAAC;;;;;;;;;;;;;;;;;;;;;;MAsBJ,EAAED,kBAAkB;;;;IAItB,CAAC;AACL"}
1
+ {"version":3,"file":"ai-model/prompt/ui-tars-locator.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../../src/ai-model/prompt/ui-tars-locator.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 { getPreferredLanguage } from '@midscene/shared/env';\n\n// claude 3.5 sonnet computer The ability to understand the content of the image is better, Does not provide element snapshot effect\nexport function systemPromptToLocateElementPosition() {\n const preferredLanguage = getPreferredLanguage();\n\n return `\nYou are a GUI agent. You are given a task and your action history, with screenshots. You need to perform the next action to complete the task. \n\n## Output Format\n\\`\\`\\`\nThought: ...\nAction: ...\n\\`\\`\\`\n\n## Action Space\nclick(start_box='[x1, y1, x2, y2]')\nleft_double(start_box='[x1, y1, x2, y2]')\nright_single(start_box='[x1, y1, x2, y2]')\ndrag(start_box='[x1, y1, x2, y2]', end_box='[x3, y3, x4, y4]')\nhotkey(key='')\ntype(content='') #If you want to submit your input, use \"\\\\n\" at the end of \\`content\\`.\nscroll(start_box='[x1, y1, x2, y2]', direction='down or up or right or left')\nwait() #Sleep for 5s and take a screenshot to check for any changes.\nfinished()\ncall_user() # Submit the task and call the user when the task is unsolvable, or when you need the user's help.\n\n## Note\n- Use ${preferredLanguage} in \\`Thought\\` part.\n- Write a small plan and finally summarize your next action (with its target element) in one sentence in \\`Thought\\` part.\n\n## User Instruction\n `;\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","systemPromptToLocateElementPosition","preferredLanguage","getPreferredLanguage"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;ACHO,SAASI;IACd,MAAMC,oBAAoBC,AAAAA,IAAAA,oBAAAA,oBAAAA,AAAAA;IAE1B,OAAO,CAAC;;;;;;;;;;;;;;;;;;;;;;MAsBJ,EAAED,kBAAkB;;;;IAItB,CAAC;AACL"}
@@ -62,10 +62,10 @@ finished(content='xxx') # Use escape characters \\', \\", and \\n in content par
62
62
  const getSummary = (prediction)=>prediction.replace(/Reflection:[\s\S]*?(?=Action_Summary:|Action:|$)/g, '').trim();
63
63
  exports.getSummary = __webpack_exports__.getSummary;
64
64
  exports.getUiTarsPlanningPrompt = __webpack_exports__.getUiTarsPlanningPrompt;
65
- for(var __webpack_i__ in __webpack_exports__)if (-1 === [
65
+ for(var __rspack_i in __webpack_exports__)if (-1 === [
66
66
  "getSummary",
67
67
  "getUiTarsPlanningPrompt"
68
- ].indexOf(__webpack_i__)) exports[__webpack_i__] = __webpack_exports__[__webpack_i__];
68
+ ].indexOf(__rspack_i)) exports[__rspack_i] = __webpack_exports__[__rspack_i];
69
69
  Object.defineProperty(exports, '__esModule', {
70
70
  value: true
71
71
  });
@@ -1 +1 @@
1
- {"version":3,"file":"ai-model/prompt/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/prompt/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 { getPreferredLanguage } from '@midscene/shared/env';\n\nexport function getUiTarsPlanningPrompt(): string {\n const preferredLanguage = getPreferredLanguage();\n\n return `\nYou are a GUI agent. You are given a task and your action history, with screenshots. You need to perform the next action to complete the task. \n\n## Output Format\n\\`\\`\\`\nThought: ...\nAction: ...\n\\`\\`\\`\n\n## Action Space\n\nclick(start_box='[x1, y1, x2, y2]')\nleft_double(start_box='[x1, y1, x2, y2]')\nright_single(start_box='[x1, y1, x2, y2]')\ndrag(start_box='[x1, y1, x2, y2]', end_box='[x3, y3, x4, y4]')\nhotkey(key='')\ntype(content='xxx') # Use escape characters \\\\', \\\\\\\", and \\\\n in content part to ensure we can parse the content in normal python string format. If you want to submit your input, use \\\\n at the end of content. \nscroll(start_box='[x1, y1, x2, y2]', direction='down or up or right or left')\nwait() #Sleep for 5s and take a screenshot to check for any changes.\nfinished(content='xxx') # Use escape characters \\\\', \\\\\", and \\\\n in content part to ensure we can parse the content in normal python string format.\n\n\n## Note\n- Use ${preferredLanguage} in \\`Thought\\` part.\n- Write a small plan and finally summarize your next action (with its target element) in one sentence in \\`Thought\\` part.\n\n## User Instruction\n`;\n}\n\nexport const getSummary = (prediction: string) =>\n prediction\n .replace(/Reflection:[\\s\\S]*?(?=Action_Summary:|Action:|$)/g, '')\n .trim();\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","getUiTarsPlanningPrompt","preferredLanguage","getPreferredLanguage","getSummary","prediction"],"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;;;;;;;;;ACJO,SAASI;IACd,MAAMC,oBAAoBC,AAAAA,IAAAA,oBAAAA,oBAAAA,AAAAA;IAE1B,OAAO,CAAC;;;;;;;;;;;;;;;;;;;;;;;MAuBJ,EAAED,kBAAkB;;;;AAI1B,CAAC;AACD;AAEO,MAAME,aAAa,CAACC,aACzBA,WACG,OAAO,CAAC,qDAAqD,IAC7D,IAAI"}
1
+ {"version":3,"file":"ai-model/prompt/ui-tars-planning.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../../src/ai-model/prompt/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 { getPreferredLanguage } from '@midscene/shared/env';\n\nexport function getUiTarsPlanningPrompt(): string {\n const preferredLanguage = getPreferredLanguage();\n\n return `\nYou are a GUI agent. You are given a task and your action history, with screenshots. You need to perform the next action to complete the task. \n\n## Output Format\n\\`\\`\\`\nThought: ...\nAction: ...\n\\`\\`\\`\n\n## Action Space\n\nclick(start_box='[x1, y1, x2, y2]')\nleft_double(start_box='[x1, y1, x2, y2]')\nright_single(start_box='[x1, y1, x2, y2]')\ndrag(start_box='[x1, y1, x2, y2]', end_box='[x3, y3, x4, y4]')\nhotkey(key='')\ntype(content='xxx') # Use escape characters \\\\', \\\\\\\", and \\\\n in content part to ensure we can parse the content in normal python string format. If you want to submit your input, use \\\\n at the end of content. \nscroll(start_box='[x1, y1, x2, y2]', direction='down or up or right or left')\nwait() #Sleep for 5s and take a screenshot to check for any changes.\nfinished(content='xxx') # Use escape characters \\\\', \\\\\", and \\\\n in content part to ensure we can parse the content in normal python string format.\n\n\n## Note\n- Use ${preferredLanguage} in \\`Thought\\` part.\n- Write a small plan and finally summarize your next action (with its target element) in one sentence in \\`Thought\\` part.\n\n## User Instruction\n`;\n}\n\nexport const getSummary = (prediction: string) =>\n prediction\n .replace(/Reflection:[\\s\\S]*?(?=Action_Summary:|Action:|$)/g, '')\n .trim();\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","getUiTarsPlanningPrompt","preferredLanguage","getPreferredLanguage","getSummary","prediction"],"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;;;;;;;;;ACJO,SAASI;IACd,MAAMC,oBAAoBC,AAAAA,IAAAA,oBAAAA,oBAAAA,AAAAA;IAE1B,OAAO,CAAC;;;;;;;;;;;;;;;;;;;;;;;MAuBJ,EAAED,kBAAkB;;;;AAI1B,CAAC;AACD;AAEO,MAAME,aAAa,CAACC,aACzBA,WACG,OAAO,CAAC,qDAAqD,IAC7D,IAAI"}
@@ -24,18 +24,13 @@ var __webpack_require__ = {};
24
24
  var __webpack_exports__ = {};
25
25
  __webpack_require__.r(__webpack_exports__);
26
26
  __webpack_require__.d(__webpack_exports__, {
27
- distance: ()=>distance,
27
+ describeElement: ()=>describeElement,
28
+ describeSize: ()=>describeSize,
28
29
  describeUserPage: ()=>describeUserPage,
30
+ distance: ()=>distance,
29
31
  distanceThreshold: ()=>distanceThreshold,
30
- describeSize: ()=>describeSize,
31
- describeElement: ()=>describeElement,
32
- elementByPositionWithElementInfo: ()=>elementByPositionWithElementInfo,
33
32
  samplePageDescription: ()=>samplePageDescription
34
33
  });
35
- const index_js_namespaceObject = require("../../image/index.js");
36
- const constants_namespaceObject = require("@midscene/shared/constants");
37
- const extractor_namespaceObject = require("@midscene/shared/extractor");
38
- const utils_namespaceObject = require("@midscene/shared/utils");
39
34
  function describeSize(size) {
40
35
  return `${size.width} x ${size.height}`;
41
36
  }
@@ -51,35 +46,6 @@ function describeElement(elements) {
51
46
  ].join(', ')).join('\n');
52
47
  }
53
48
  const distanceThreshold = 16;
54
- function elementByPositionWithElementInfo(treeRoot, position, options) {
55
- const requireStrictDistance = (null == options ? void 0 : options.requireStrictDistance) ?? true;
56
- const filterPositionElements = (null == options ? void 0 : options.filterPositionElements) ?? false;
57
- (0, utils_namespaceObject.assert)(void 0 !== position, 'position is required for query');
58
- const matchingElements = [];
59
- function dfs(node) {
60
- if (null == node ? void 0 : node.node) {
61
- const item = node.node;
62
- if (item.rect.left <= position.x && position.x <= item.rect.left + item.rect.width && item.rect.top <= position.y && position.y <= item.rect.top + item.rect.height) {
63
- var _item_attributes;
64
- if (!(filterPositionElements && (null == (_item_attributes = item.attributes) ? void 0 : _item_attributes.nodeType) === constants_namespaceObject.NodeType.POSITION) && item.isVisible) matchingElements.push(item);
65
- }
66
- }
67
- for (const child of node.children)dfs(child);
68
- }
69
- dfs(treeRoot);
70
- if (0 === matchingElements.length) return;
71
- const element = matchingElements.reduce((smallest, current)=>{
72
- const smallestArea = smallest.rect.width * smallest.rect.height;
73
- const currentArea = current.rect.width * current.rect.height;
74
- return currentArea < smallestArea ? current : smallest;
75
- });
76
- const distanceToCenter = distance({
77
- x: element.center[0],
78
- y: element.center[1]
79
- }, position);
80
- if (requireStrictDistance) return distanceToCenter <= distanceThreshold ? element : void 0;
81
- return element;
82
- }
83
49
  function distance(point1, point2) {
84
50
  return Math.sqrt((point1.x - point2.x) ** 2 + (point1.y - point2.y) ** 2);
85
51
  }
@@ -98,77 +64,23 @@ Description of all the elements in screenshot:
98
64
  </div>
99
65
  ====================
100
66
  `;
101
- async function describeUserPage(context, opt) {
102
- const { screenshotBase64 } = context;
103
- let width;
104
- let height;
105
- if (context.size) ({ width, height } = context.size);
106
- else {
107
- const imgSize = await (0, index_js_namespaceObject.imageInfoOfBase64)(screenshotBase64);
108
- ({ width, height } = imgSize);
109
- }
110
- const treeRoot = context.tree;
111
- const idElementMap = {};
112
- const flatElements = (0, extractor_namespaceObject.treeToList)(treeRoot);
113
- if ((null == opt ? void 0 : opt.domIncluded) === true && flatElements.length >= 5000) console.warn('The number of elements is too large, it may cause the prompt to be too long, please use domIncluded: "visible-only" to reduce the number of elements');
114
- flatElements.forEach((element)=>{
115
- idElementMap[element.id] = element;
116
- if (void 0 !== element.indexId) idElementMap[`${element.indexId}`] = element;
117
- });
118
- let pageDescription = '';
119
- const visibleOnly = (null == opt ? void 0 : opt.domIncluded) === 'visible-only' ? true : (null == opt ? void 0 : opt.visibleOnly) ?? false;
120
- const resolvedVlMode = null == opt ? void 0 : opt.vlMode;
121
- const shouldIncludeDOM = (null == opt ? void 0 : opt.domIncluded) || !resolvedVlMode;
122
- if (shouldIncludeDOM) {
123
- const contentTree = await (0, extractor_namespaceObject.descriptionOfTree)(treeRoot, null == opt ? void 0 : opt.truncateTextLength, null == opt ? void 0 : opt.filterNonTextContent, visibleOnly);
124
- const sizeDescription = describeSize({
125
- width,
126
- height
127
- });
128
- pageDescription = `The size of the page: ${sizeDescription} \n The page elements tree:\n${contentTree}`;
129
- }
130
- return {
131
- description: pageDescription,
132
- elementById (idOrIndexId) {
133
- (0, utils_namespaceObject.assert)(void 0 !== idOrIndexId, 'id is required for query');
134
- const item = idElementMap[`${idOrIndexId}`];
135
- return item;
136
- },
137
- elementByPosition (position, size) {
138
- return elementByPositionWithElementInfo(treeRoot, position);
139
- },
140
- insertElementByPosition (position) {
141
- const element = (0, extractor_namespaceObject.generateElementByPosition)(position);
142
- treeRoot.children.push({
143
- node: element,
144
- children: []
145
- });
146
- flatElements.push(element);
147
- idElementMap[element.id] = element;
148
- return element;
149
- },
150
- size: {
151
- width,
152
- height
153
- }
154
- };
67
+ async function describeUserPage(context) {
68
+ return `The size of the page: ${describeSize(context.size)}`;
155
69
  }
156
70
  exports.describeElement = __webpack_exports__.describeElement;
157
71
  exports.describeSize = __webpack_exports__.describeSize;
158
72
  exports.describeUserPage = __webpack_exports__.describeUserPage;
159
73
  exports.distance = __webpack_exports__.distance;
160
74
  exports.distanceThreshold = __webpack_exports__.distanceThreshold;
161
- exports.elementByPositionWithElementInfo = __webpack_exports__.elementByPositionWithElementInfo;
162
75
  exports.samplePageDescription = __webpack_exports__.samplePageDescription;
163
- for(var __webpack_i__ in __webpack_exports__)if (-1 === [
76
+ for(var __rspack_i in __webpack_exports__)if (-1 === [
164
77
  "describeElement",
165
78
  "describeSize",
166
79
  "describeUserPage",
167
80
  "distance",
168
81
  "distanceThreshold",
169
- "elementByPositionWithElementInfo",
170
82
  "samplePageDescription"
171
- ].indexOf(__webpack_i__)) exports[__webpack_i__] = __webpack_exports__[__webpack_i__];
83
+ ].indexOf(__rspack_i)) exports[__rspack_i] = __webpack_exports__[__rspack_i];
172
84
  Object.defineProperty(exports, '__esModule', {
173
85
  value: true
174
86
  });
@@ -1 +1 @@
1
- {"version":3,"file":"ai-model/prompt/util.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/prompt/util.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 { imageInfoOfBase64 } from '@/image/index';\nimport type { BaseElement, ElementTreeNode, Size, UIContext } from '@/types';\nimport { NodeType } from '@midscene/shared/constants';\nimport type { TVlModeTypes } from '@midscene/shared/env';\nimport {\n descriptionOfTree,\n generateElementByPosition,\n treeToList,\n} from '@midscene/shared/extractor';\nimport { assert } from '@midscene/shared/utils';\n\nexport function describeSize(size: Size) {\n return `${size.width} x ${size.height}`;\n}\n\nexport function describeElement(\n elements: (Pick<BaseElement, 'rect' | 'content'> & { id: string })[],\n) {\n const sliceLength = 80;\n return elements\n .map((item) =>\n [\n item.id,\n item.rect.left,\n item.rect.top,\n item.rect.left + item.rect.width,\n item.rect.top + item.rect.height,\n item.content.length > sliceLength\n ? `${item.content.slice(0, sliceLength)}...`\n : item.content,\n ].join(', '),\n )\n .join('\\n');\n}\nexport const distanceThreshold = 16;\n\nexport function elementByPositionWithElementInfo(\n treeRoot: ElementTreeNode<BaseElement>,\n position: {\n x: number;\n y: number;\n },\n options?: {\n requireStrictDistance?: boolean;\n filterPositionElements?: boolean;\n },\n) {\n const requireStrictDistance = options?.requireStrictDistance ?? true;\n const filterPositionElements = options?.filterPositionElements ?? false;\n\n assert(typeof position !== 'undefined', 'position is required for query');\n\n const matchingElements: BaseElement[] = [];\n\n function dfs(node: ElementTreeNode<BaseElement>) {\n if (node?.node) {\n const item = node.node;\n if (\n item.rect.left <= position.x &&\n position.x <= item.rect.left + item.rect.width &&\n item.rect.top <= position.y &&\n position.y <= item.rect.top + item.rect.height\n ) {\n if (\n !(\n filterPositionElements &&\n item.attributes?.nodeType === NodeType.POSITION\n ) &&\n item.isVisible\n ) {\n matchingElements.push(item);\n }\n }\n }\n\n for (const child of node.children) {\n dfs(child);\n }\n }\n\n dfs(treeRoot);\n\n if (matchingElements.length === 0) {\n return undefined;\n }\n\n // Find the smallest element by area\n const element = matchingElements.reduce((smallest, current) => {\n const smallestArea = smallest.rect.width * smallest.rect.height;\n const currentArea = current.rect.width * current.rect.height;\n return currentArea < smallestArea ? current : smallest;\n });\n\n const distanceToCenter = distance(\n { x: element.center[0], y: element.center[1] },\n position,\n );\n\n if (requireStrictDistance) {\n return distanceToCenter <= distanceThreshold ? element : undefined;\n }\n\n return element;\n}\n\nexport function distance(\n point1: { x: number; y: number },\n point2: { x: number; y: number },\n) {\n return Math.sqrt((point1.x - point2.x) ** 2 + (point1.y - point2.y) ** 2);\n}\n\nexport const samplePageDescription = `\nAnd the page is described as follows:\n====================\nThe size of the page: 1280 x 720\nSome of the elements are marked with a rectangle in the screenshot corresponding to the markerId, some are not.\n\nDescription of all the elements in screenshot:\n<div id=\"969f1637\" markerId=\"1\" left=\"100\" top=\"100\" width=\"100\" height=\"100\"> // The markerId indicated by the rectangle label in the screenshot\n <h4 id=\"b211ecb2\" markerId=\"5\" left=\"150\" top=\"150\" width=\"90\" height=\"60\">\n The username is accepted\n </h4>\n ...many more\n</div>\n====================\n`;\n\nexport async function describeUserPage<\n ElementType extends BaseElement = BaseElement,\n>(\n context: Omit<UIContext<ElementType>, 'describer'>,\n opt: {\n truncateTextLength?: number;\n filterNonTextContent?: boolean;\n domIncluded?: boolean | 'visible-only';\n visibleOnly?: boolean;\n vlMode: TVlModeTypes | undefined;\n },\n) {\n const { screenshotBase64 } = context;\n let width: number;\n let height: number;\n\n if (context.size) {\n ({ width, height } = context.size);\n } else {\n const imgSize = await imageInfoOfBase64(screenshotBase64);\n ({ width, height } = imgSize);\n }\n\n const treeRoot = context.tree;\n // dfs tree, save the id and element info\n const idElementMap: Record<string, ElementType> = {};\n const flatElements: ElementType[] = treeToList(treeRoot);\n\n if (opt?.domIncluded === true && flatElements.length >= 5000) {\n console.warn(\n 'The number of elements is too large, it may cause the prompt to be too long, please use domIncluded: \"visible-only\" to reduce the number of elements',\n );\n }\n\n flatElements.forEach((element) => {\n idElementMap[element.id] = element;\n if (typeof element.indexId !== 'undefined') {\n idElementMap[`${element.indexId}`] = element;\n }\n });\n\n let pageDescription = '';\n\n const visibleOnly =\n opt?.domIncluded === 'visible-only' ? true : (opt?.visibleOnly ?? false);\n const resolvedVlMode = opt?.vlMode;\n const shouldIncludeDOM = opt?.domIncluded || !resolvedVlMode;\n\n if (shouldIncludeDOM) {\n // non-vl mode must provide the page description\n const contentTree = await descriptionOfTree(\n treeRoot,\n opt?.truncateTextLength,\n opt?.filterNonTextContent,\n visibleOnly,\n );\n\n // if match by position, don't need to provide the page description\n const sizeDescription = describeSize({ width, height });\n pageDescription = `The size of the page: ${sizeDescription} \\n The page elements tree:\\n${contentTree}`;\n }\n\n return {\n description: pageDescription,\n elementById(idOrIndexId: string) {\n assert(typeof idOrIndexId !== 'undefined', 'id is required for query');\n const item = idElementMap[`${idOrIndexId}`];\n return item;\n },\n elementByPosition(\n position: { x: number; y: number },\n size: { width: number; height: number },\n ) {\n return elementByPositionWithElementInfo(treeRoot, position);\n },\n insertElementByPosition(position: { x: number; y: number }) {\n const element = generateElementByPosition(position) as ElementType;\n\n treeRoot.children.push({\n node: element,\n children: [],\n });\n flatElements.push(element);\n idElementMap[element.id] = element;\n return element;\n },\n size: { width, height },\n };\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","describeSize","size","describeElement","elements","sliceLength","item","distanceThreshold","elementByPositionWithElementInfo","treeRoot","position","options","requireStrictDistance","filterPositionElements","assert","matchingElements","dfs","node","_item_attributes","NodeType","child","element","smallest","current","smallestArea","currentArea","distanceToCenter","distance","undefined","point1","point2","Math","samplePageDescription","describeUserPage","context","opt","screenshotBase64","width","height","imgSize","imageInfoOfBase64","idElementMap","flatElements","treeToList","console","pageDescription","visibleOnly","resolvedVlMode","shouldIncludeDOM","contentTree","descriptionOfTree","sizeDescription","idOrIndexId","generateElementByPosition"],"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;;;;;;;;;;;;;;;;;ACKO,SAASI,aAAaC,IAAU;IACrC,OAAO,GAAGA,KAAK,KAAK,CAAC,GAAG,EAAEA,KAAK,MAAM,EAAE;AACzC;AAEO,SAASC,gBACdC,QAAoE;IAEpE,MAAMC,cAAc;IACpB,OAAOD,SACJ,GAAG,CAAC,CAACE,OACJ;YACEA,KAAK,EAAE;YACPA,KAAK,IAAI,CAAC,IAAI;YACdA,KAAK,IAAI,CAAC,GAAG;YACbA,KAAK,IAAI,CAAC,IAAI,GAAGA,KAAK,IAAI,CAAC,KAAK;YAChCA,KAAK,IAAI,CAAC,GAAG,GAAGA,KAAK,IAAI,CAAC,MAAM;YAChCA,KAAK,OAAO,CAAC,MAAM,GAAGD,cAClB,GAAGC,KAAK,OAAO,CAAC,KAAK,CAAC,GAAGD,aAAa,GAAG,CAAC,GAC1CC,KAAK,OAAO;SACjB,CAAC,IAAI,CAAC,OAER,IAAI,CAAC;AACV;AACO,MAAMC,oBAAoB;AAE1B,SAASC,iCACdC,QAAsC,EACtCC,QAGC,EACDC,OAGC;IAED,MAAMC,wBAAwBD,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,qBAAqB,AAAD,KAAK;IAChE,MAAME,yBAAyBF,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,sBAAsB,AAAD,KAAK;IAElEG,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,AAAoB,WAAbJ,UAA0B;IAExC,MAAMK,mBAAkC,EAAE;IAE1C,SAASC,IAAIC,IAAkC;QAC7C,IAAIA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,IAAI,EAAE;YACd,MAAMX,OAAOW,KAAK,IAAI;YACtB,IACEX,KAAK,IAAI,CAAC,IAAI,IAAII,SAAS,CAAC,IAC5BA,SAAS,CAAC,IAAIJ,KAAK,IAAI,CAAC,IAAI,GAAGA,KAAK,IAAI,CAAC,KAAK,IAC9CA,KAAK,IAAI,CAAC,GAAG,IAAII,SAAS,CAAC,IAC3BA,SAAS,CAAC,IAAIJ,KAAK,IAAI,CAAC,GAAG,GAAGA,KAAK,IAAI,CAAC,MAAM,EAC9C;oBAIIY;gBAHJ,IACE,CACEL,CAAAA,0BACAK,AAAAA,SAAAA,CAAAA,mBAAAA,KAAK,UAAU,AAAD,IAAdA,KAAAA,IAAAA,iBAAiB,QAAQ,AAAD,MAAMC,0BAAAA,QAAAA,CAAAA,QAAgB,AAAhBA,KAEhCb,KAAK,SAAS,EAEdS,iBAAiB,IAAI,CAACT;YAE1B;QACF;QAEA,KAAK,MAAMc,SAASH,KAAK,QAAQ,CAC/BD,IAAII;IAER;IAEAJ,IAAIP;IAEJ,IAAIM,AAA4B,MAA5BA,iBAAiB,MAAM,EACzB;IAIF,MAAMM,UAAUN,iBAAiB,MAAM,CAAC,CAACO,UAAUC;QACjD,MAAMC,eAAeF,SAAS,IAAI,CAAC,KAAK,GAAGA,SAAS,IAAI,CAAC,MAAM;QAC/D,MAAMG,cAAcF,QAAQ,IAAI,CAAC,KAAK,GAAGA,QAAQ,IAAI,CAAC,MAAM;QAC5D,OAAOE,cAAcD,eAAeD,UAAUD;IAChD;IAEA,MAAMI,mBAAmBC,SACvB;QAAE,GAAGN,QAAQ,MAAM,CAAC,EAAE;QAAE,GAAGA,QAAQ,MAAM,CAAC,EAAE;IAAC,GAC7CX;IAGF,IAAIE,uBACF,OAAOc,oBAAoBnB,oBAAoBc,UAAUO;IAG3D,OAAOP;AACT;AAEO,SAASM,SACdE,MAAgC,EAChCC,MAAgC;IAEhC,OAAOC,KAAK,IAAI,CAAEF,AAAAA,CAAAA,OAAO,CAAC,GAAGC,OAAO,CAAC,AAAD,KAAM,IAAKD,AAAAA,CAAAA,OAAO,CAAC,GAAGC,OAAO,CAAC,AAAD,KAAM;AACzE;AAEO,MAAME,wBAAwB,CAAC;;;;;;;;;;;;;;AActC,CAAC;AAEM,eAAeC,iBAGpBC,OAAkD,EAClDC,GAMC;IAED,MAAM,EAAEC,gBAAgB,EAAE,GAAGF;IAC7B,IAAIG;IACJ,IAAIC;IAEJ,IAAIJ,QAAQ,IAAI,EACb,GAAEG,KAAK,EAAEC,MAAM,EAAE,GAAGJ,QAAQ,IAAG;SAC3B;QACL,MAAMK,UAAU,MAAMC,AAAAA,IAAAA,yBAAAA,iBAAAA,AAAAA,EAAkBJ;QACvC,GAAEC,KAAK,EAAEC,MAAM,EAAE,GAAGC,OAAM;IAC7B;IAEA,MAAM9B,WAAWyB,QAAQ,IAAI;IAE7B,MAAMO,eAA4C,CAAC;IACnD,MAAMC,eAA8BC,AAAAA,IAAAA,0BAAAA,UAAAA,AAAAA,EAAWlC;IAE/C,IAAI0B,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,MAAM,QAAQO,aAAa,MAAM,IAAI,MACtDE,QAAQ,IAAI,CACV;IAIJF,aAAa,OAAO,CAAC,CAACrB;QACpBoB,YAAY,CAACpB,QAAQ,EAAE,CAAC,GAAGA;QAC3B,IAAI,AAA2B,WAApBA,QAAQ,OAAO,EACxBoB,YAAY,CAAC,GAAGpB,QAAQ,OAAO,EAAE,CAAC,GAAGA;IAEzC;IAEA,IAAIwB,kBAAkB;IAEtB,MAAMC,cACJX,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,MAAM,iBAAiB,OAAQA,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAK;IACpE,MAAMY,iBAAiBZ,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,MAAM;IAClC,MAAMa,mBAAmBb,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,WAAW,AAAD,KAAK,CAACY;IAE9C,IAAIC,kBAAkB;QAEpB,MAAMC,cAAc,MAAMC,AAAAA,IAAAA,0BAAAA,iBAAAA,AAAAA,EACxBzC,UACA0B,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,kBAAkB,EACvBA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,oBAAoB,EACzBW;QAIF,MAAMK,kBAAkBlD,aAAa;YAAEoC;YAAOC;QAAO;QACrDO,kBAAkB,CAAC,sBAAsB,EAAEM,gBAAgB,6BAA6B,EAAEF,aAAa;IACzG;IAEA,OAAO;QACL,aAAaJ;QACb,aAAYO,WAAmB;YAC7BtC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,AAAuB,WAAhBsC,aAA6B;YAC3C,MAAM9C,OAAOmC,YAAY,CAAC,GAAGW,aAAa,CAAC;YAC3C,OAAO9C;QACT;QACA,mBACEI,QAAkC,EAClCR,IAAuC;YAEvC,OAAOM,iCAAiCC,UAAUC;QACpD;QACA,yBAAwBA,QAAkC;YACxD,MAAMW,UAAUgC,AAAAA,IAAAA,0BAAAA,yBAAAA,AAAAA,EAA0B3C;YAE1CD,SAAS,QAAQ,CAAC,IAAI,CAAC;gBACrB,MAAMY;gBACN,UAAU,EAAE;YACd;YACAqB,aAAa,IAAI,CAACrB;YAClBoB,YAAY,CAACpB,QAAQ,EAAE,CAAC,GAAGA;YAC3B,OAAOA;QACT;QACA,MAAM;YAAEgB;YAAOC;QAAO;IACxB;AACF"}
1
+ {"version":3,"file":"ai-model/prompt/util.js","sources":["webpack/runtime/define_property_getters","webpack/runtime/has_own_property","webpack/runtime/make_namespace_object","../../../../src/ai-model/prompt/util.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 { BaseElement, ElementTreeNode, Size, UIContext } from '@/types';\nimport { NodeType } from '@midscene/shared/constants';\nimport { assert } from '@midscene/shared/utils';\n\nexport function describeSize(size: Size) {\n return `${size.width} x ${size.height}`;\n}\n\nexport function describeElement(\n elements: (Pick<BaseElement, 'rect' | 'content'> & { id: string })[],\n) {\n const sliceLength = 80;\n return elements\n .map((item) =>\n [\n item.id,\n item.rect.left,\n item.rect.top,\n item.rect.left + item.rect.width,\n item.rect.top + item.rect.height,\n item.content.length > sliceLength\n ? `${item.content.slice(0, sliceLength)}...`\n : item.content,\n ].join(', '),\n )\n .join('\\n');\n}\nexport const distanceThreshold = 16;\n\n// export function elementByPositionWithElementInfo(\n// treeRoot: ElementTreeNode<BaseElement>,\n// position: {\n// x: number;\n// y: number;\n// },\n// options?: {\n// requireStrictDistance?: boolean;\n// filterPositionElements?: boolean;\n// },\n// ) {\n// const requireStrictDistance = options?.requireStrictDistance ?? true;\n// const filterPositionElements = options?.filterPositionElements ?? false;\n\n// assert(typeof position !== 'undefined', 'position is required for query');\n\n// const matchingElements: BaseElement[] = [];\n\n// function dfs(node: ElementTreeNode<BaseElement>) {\n// if (node?.node) {\n// const item = node.node;\n// if (\n// item.rect.left <= position.x &&\n// position.x <= item.rect.left + item.rect.width &&\n// item.rect.top <= position.y &&\n// position.y <= item.rect.top + item.rect.height\n// ) {\n// if (\n// !(\n// filterPositionElements &&\n// item.attributes?.nodeType === NodeType.POSITION\n// ) &&\n// item.isVisible\n// ) {\n// matchingElements.push(item);\n// }\n// }\n// }\n\n// for (const child of node.children) {\n// dfs(child);\n// }\n// }\n\n// dfs(treeRoot);\n\n// if (matchingElements.length === 0) {\n// return undefined;\n// }\n\n// // Find the smallest element by area\n// const element = matchingElements.reduce((smallest, current) => {\n// const smallestArea = smallest.rect.width * smallest.rect.height;\n// const currentArea = current.rect.width * current.rect.height;\n// return currentArea < smallestArea ? current : smallest;\n// });\n\n// const distanceToCenter = distance(\n// { x: element.center[0], y: element.center[1] },\n// position,\n// );\n\n// if (requireStrictDistance) {\n// return distanceToCenter <= distanceThreshold ? element : undefined;\n// }\n\n// return element;\n// }\n\nexport function distance(\n point1: { x: number; y: number },\n point2: { x: number; y: number },\n) {\n return Math.sqrt((point1.x - point2.x) ** 2 + (point1.y - point2.y) ** 2);\n}\n\nexport const samplePageDescription = `\nAnd the page is described as follows:\n====================\nThe size of the page: 1280 x 720\nSome of the elements are marked with a rectangle in the screenshot corresponding to the markerId, some are not.\n\nDescription of all the elements in screenshot:\n<div id=\"969f1637\" markerId=\"1\" left=\"100\" top=\"100\" width=\"100\" height=\"100\"> // The markerId indicated by the rectangle label in the screenshot\n <h4 id=\"b211ecb2\" markerId=\"5\" left=\"150\" top=\"150\" width=\"90\" height=\"60\">\n The username is accepted\n </h4>\n ...many more\n</div>\n====================\n`;\n\nexport async function describeUserPage(context: UIContext) {\n return `The size of the page: ${describeSize(context.size)}`;\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","describeSize","size","describeElement","elements","sliceLength","item","distanceThreshold","distance","point1","point2","Math","samplePageDescription","describeUserPage","context"],"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;;;;;;;;;;;;ACFO,SAASI,aAAaC,IAAU;IACrC,OAAO,GAAGA,KAAK,KAAK,CAAC,GAAG,EAAEA,KAAK,MAAM,EAAE;AACzC;AAEO,SAASC,gBACdC,QAAoE;IAEpE,MAAMC,cAAc;IACpB,OAAOD,SACJ,GAAG,CAAC,CAACE,OACJ;YACEA,KAAK,EAAE;YACPA,KAAK,IAAI,CAAC,IAAI;YACdA,KAAK,IAAI,CAAC,GAAG;YACbA,KAAK,IAAI,CAAC,IAAI,GAAGA,KAAK,IAAI,CAAC,KAAK;YAChCA,KAAK,IAAI,CAAC,GAAG,GAAGA,KAAK,IAAI,CAAC,MAAM;YAChCA,KAAK,OAAO,CAAC,MAAM,GAAGD,cAClB,GAAGC,KAAK,OAAO,CAAC,KAAK,CAAC,GAAGD,aAAa,GAAG,CAAC,GAC1CC,KAAK,OAAO;SACjB,CAAC,IAAI,CAAC,OAER,IAAI,CAAC;AACV;AACO,MAAMC,oBAAoB;AAuE1B,SAASC,SACdC,MAAgC,EAChCC,MAAgC;IAEhC,OAAOC,KAAK,IAAI,CAAEF,AAAAA,CAAAA,OAAO,CAAC,GAAGC,OAAO,CAAC,AAAD,KAAM,IAAKD,AAAAA,CAAAA,OAAO,CAAC,GAAGC,OAAO,CAAC,AAAD,KAAM;AACzE;AAEO,MAAME,wBAAwB,CAAC;;;;;;;;;;;;;;AActC,CAAC;AAEM,eAAeC,iBAAiBC,OAAkB;IACvD,OAAO,CAAC,sBAAsB,EAAEb,aAAaa,QAAQ,IAAI,GAAG;AAC9D"}
@@ -24,16 +24,16 @@ var __webpack_require__ = {};
24
24
  var __webpack_exports__ = {};
25
25
  __webpack_require__.r(__webpack_exports__);
26
26
  __webpack_require__.d(__webpack_exports__, {
27
- getScreenshotsForLLM: ()=>getScreenshotsForLLM,
28
- generateYamlTest: ()=>generateYamlTest,
29
- createMessageContent: ()=>createMessageContent,
27
+ createEventCounts: ()=>createEventCounts,
30
28
  filterEventsByType: ()=>filterEventsByType,
31
- prepareEventSummary: ()=>prepareEventSummary,
32
29
  processEventsForLLM: ()=>processEventsForLLM,
30
+ validateEvents: ()=>validateEvents,
31
+ createMessageContent: ()=>createMessageContent,
33
32
  generateYamlTestStream: ()=>generateYamlTestStream,
33
+ getScreenshotsForLLM: ()=>getScreenshotsForLLM,
34
34
  extractInputDescriptions: ()=>extractInputDescriptions,
35
- validateEvents: ()=>validateEvents,
36
- createEventCounts: ()=>createEventCounts
35
+ prepareEventSummary: ()=>prepareEventSummary,
36
+ generateYamlTest: ()=>generateYamlTest
37
37
  });
38
38
  const constants_namespaceObject = require("@midscene/shared/constants");
39
39
  const external_index_js_namespaceObject = require("../index.js");
@@ -154,10 +154,10 @@ Event Summary:
154
154
  ${JSON.stringify(yamlSummary, null, 2)}
155
155
 
156
156
  Convert events:
157
- - navigation \u{2192} target.url
158
- - click \u{2192} aiTap with element description
159
- - input \u{2192} aiInput with value and locate
160
- - scroll \u{2192} aiScroll with appropriate direction
157
+ - navigation target.url
158
+ - click aiTap with element description
159
+ - input aiInput with value and locate
160
+ - scroll aiScroll with appropriate direction
161
161
  - Add aiAssert for important state changes
162
162
 
163
163
  Important: Return ONLY the raw YAML content. Do NOT wrap the response in markdown code blocks (no \`\`\`yaml or \`\`\`). Start directly with the YAML content.`
@@ -179,7 +179,7 @@ Important: Return ONLY the raw YAML content. Do NOT wrap the response in markdow
179
179
  });
180
180
  }
181
181
  const response = await (0, external_index_js_namespaceObject.callAIWithStringResponse)(prompt, external_index_js_namespaceObject.AIActionType.TEXT, modelConfig);
182
- if ((null == response ? void 0 : response.content) && 'string' == typeof response.content) return response.content;
182
+ if (response?.content && 'string' == typeof response.content) return response.content;
183
183
  throw new Error('Failed to generate YAML test configuration');
184
184
  } catch (error) {
185
185
  throw new Error(`Failed to generate YAML test: ${error}`);
@@ -210,10 +210,10 @@ Event Summary:
210
210
  ${JSON.stringify(yamlSummary, null, 2)}
211
211
 
212
212
  Convert events:
213
- - navigation \u{2192} target.url
214
- - click \u{2192} aiTap with element description
215
- - input \u{2192} aiInput with value and locate
216
- - scroll \u{2192} aiScroll with appropriate direction
213
+ - navigation target.url
214
+ - click aiTap with element description
215
+ - input aiInput with value and locate
216
+ - scroll aiScroll with appropriate direction
217
217
  - Add aiAssert for important state changes
218
218
 
219
219
  Important: Return ONLY the raw YAML content. Do NOT wrap the response in markdown code blocks (no \`\`\`yaml or \`\`\`). Start directly with the YAML content.`
@@ -240,7 +240,7 @@ Important: Return ONLY the raw YAML content. Do NOT wrap the response in markdow
240
240
  });
241
241
  {
242
242
  const response = await (0, external_index_js_namespaceObject.callAIWithStringResponse)(prompt, external_index_js_namespaceObject.AIActionType.TEXT, modelConfig);
243
- if ((null == response ? void 0 : response.content) && 'string' == typeof response.content) return {
243
+ if (response?.content && 'string' == typeof response.content) return {
244
244
  content: response.content,
245
245
  usage: response.usage,
246
246
  isStreamed: false
@@ -261,7 +261,7 @@ exports.getScreenshotsForLLM = __webpack_exports__.getScreenshotsForLLM;
261
261
  exports.prepareEventSummary = __webpack_exports__.prepareEventSummary;
262
262
  exports.processEventsForLLM = __webpack_exports__.processEventsForLLM;
263
263
  exports.validateEvents = __webpack_exports__.validateEvents;
264
- for(var __webpack_i__ in __webpack_exports__)if (-1 === [
264
+ for(var __rspack_i in __webpack_exports__)if (-1 === [
265
265
  "createEventCounts",
266
266
  "createMessageContent",
267
267
  "extractInputDescriptions",
@@ -272,7 +272,7 @@ for(var __webpack_i__ in __webpack_exports__)if (-1 === [
272
272
  "prepareEventSummary",
273
273
  "processEventsForLLM",
274
274
  "validateEvents"
275
- ].indexOf(__webpack_i__)) exports[__webpack_i__] = __webpack_exports__[__webpack_i__];
275
+ ].indexOf(__rspack_i)) exports[__rspack_i] = __webpack_exports__[__rspack_i];
276
276
  Object.defineProperty(exports, '__esModule', {
277
277
  value: true
278
278
  });