@midscene/core 1.8.10 → 1.9.0

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 (304) hide show
  1. package/dist/es/agent/agent.mjs +40 -50
  2. package/dist/es/agent/agent.mjs.map +1 -1
  3. package/dist/es/agent/task-builder.mjs +39 -19
  4. package/dist/es/agent/task-builder.mjs.map +1 -1
  5. package/dist/es/agent/tasks.mjs +24 -22
  6. package/dist/es/agent/tasks.mjs.map +1 -1
  7. package/dist/es/agent/utils.mjs +11 -14
  8. package/dist/es/agent/utils.mjs.map +1 -1
  9. package/dist/es/ai-model/connectivity.mjs +7 -3
  10. package/dist/es/ai-model/connectivity.mjs.map +1 -1
  11. package/dist/es/ai-model/errors.mjs +9 -0
  12. package/dist/es/ai-model/errors.mjs.map +1 -0
  13. package/dist/es/ai-model/index.mjs +3 -4
  14. package/dist/es/ai-model/inspect.mjs +132 -144
  15. package/dist/es/ai-model/inspect.mjs.map +1 -1
  16. package/dist/es/ai-model/llm-planning.mjs +46 -28
  17. package/dist/es/ai-model/llm-planning.mjs.map +1 -1
  18. package/dist/es/ai-model/{auto-glm → models/auto-glm}/actions.mjs +22 -44
  19. package/dist/es/ai-model/models/auto-glm/actions.mjs.map +1 -0
  20. package/dist/es/ai-model/models/auto-glm/adapter.mjs +45 -0
  21. package/dist/es/ai-model/models/auto-glm/adapter.mjs.map +1 -0
  22. package/dist/es/ai-model/models/auto-glm/locate.mjs +112 -0
  23. package/dist/es/ai-model/models/auto-glm/locate.mjs.map +1 -0
  24. package/dist/es/ai-model/models/auto-glm/parser.mjs.map +1 -0
  25. package/dist/es/ai-model/{auto-glm → models/auto-glm}/planning.mjs +6 -7
  26. package/dist/es/ai-model/models/auto-glm/planning.mjs.map +1 -0
  27. package/dist/es/ai-model/{auto-glm → models/auto-glm}/prompt.mjs +3 -11
  28. package/dist/es/ai-model/models/auto-glm/prompt.mjs.map +1 -0
  29. package/dist/es/ai-model/models/default.mjs +12 -0
  30. package/dist/es/ai-model/models/default.mjs.map +1 -0
  31. package/dist/es/ai-model/models/doubao.mjs +138 -0
  32. package/dist/es/ai-model/models/doubao.mjs.map +1 -0
  33. package/dist/es/ai-model/models/gemini.mjs +34 -0
  34. package/dist/es/ai-model/models/gemini.mjs.map +1 -0
  35. package/dist/es/ai-model/models/glm.mjs +37 -0
  36. package/dist/es/ai-model/models/glm.mjs.map +1 -0
  37. package/dist/es/ai-model/models/gpt.mjs +31 -0
  38. package/dist/es/ai-model/models/gpt.mjs.map +1 -0
  39. package/dist/es/ai-model/models/index.mjs +2 -0
  40. package/dist/es/ai-model/models/qwen.mjs +113 -0
  41. package/dist/es/ai-model/models/qwen.mjs.map +1 -0
  42. package/dist/es/ai-model/models/registry.mjs +45 -0
  43. package/dist/es/ai-model/models/registry.mjs.map +1 -0
  44. package/dist/es/ai-model/models/resolved.mjs +104 -0
  45. package/dist/es/ai-model/models/resolved.mjs.map +1 -0
  46. package/dist/es/ai-model/models/types.mjs +0 -0
  47. package/dist/es/ai-model/models/ui-tars/adapter.mjs +142 -0
  48. package/dist/es/ai-model/models/ui-tars/adapter.mjs.map +1 -0
  49. package/dist/es/ai-model/{ui-tars-planning.mjs → models/ui-tars/planning.mjs} +44 -62
  50. package/dist/es/ai-model/models/ui-tars/planning.mjs.map +1 -0
  51. package/dist/es/ai-model/prompt/extraction.mjs +3 -3
  52. package/dist/es/ai-model/prompt/extraction.mjs.map +1 -1
  53. package/dist/es/ai-model/prompt/llm-locator.mjs +11 -11
  54. package/dist/es/ai-model/prompt/llm-locator.mjs.map +1 -1
  55. package/dist/es/ai-model/prompt/llm-planning.mjs +25 -60
  56. package/dist/es/ai-model/prompt/llm-planning.mjs.map +1 -1
  57. package/dist/es/ai-model/prompt/llm-section-locator.mjs +15 -10
  58. package/dist/es/ai-model/prompt/llm-section-locator.mjs.map +1 -1
  59. package/dist/es/ai-model/prompt/locate-grounding-rules.mjs +9 -0
  60. package/dist/es/ai-model/prompt/locate-grounding-rules.mjs.map +1 -0
  61. package/dist/es/ai-model/prompt/locate-param-example.mjs +15 -0
  62. package/dist/es/ai-model/prompt/locate-param-example.mjs.map +1 -0
  63. package/dist/es/ai-model/prompt/playwright-generator.mjs +5 -5
  64. package/dist/es/ai-model/prompt/playwright-generator.mjs.map +1 -1
  65. package/dist/es/ai-model/prompt/yaml-generator.mjs +5 -5
  66. package/dist/es/ai-model/prompt/yaml-generator.mjs.map +1 -1
  67. package/dist/es/ai-model/prompts/locate-result-coordinates.mjs +107 -0
  68. package/dist/es/ai-model/prompts/locate-result-coordinates.mjs.map +1 -0
  69. package/dist/es/ai-model/service-caller/index.mjs +59 -190
  70. package/dist/es/ai-model/service-caller/index.mjs.map +1 -1
  71. package/dist/es/ai-model/service-caller/json.mjs +60 -0
  72. package/dist/es/ai-model/service-caller/json.mjs.map +1 -0
  73. package/dist/es/ai-model/shared/model-locate-result/bbox.mjs +68 -0
  74. package/dist/es/ai-model/shared/model-locate-result/bbox.mjs.map +1 -0
  75. package/dist/es/ai-model/shared/model-locate-result/factory.mjs +96 -0
  76. package/dist/es/ai-model/shared/model-locate-result/factory.mjs.map +1 -0
  77. package/dist/es/ai-model/shared/model-locate-result/index.mjs +3 -0
  78. package/dist/es/ai-model/shared/model-locate-result/parse.mjs +41 -0
  79. package/dist/es/ai-model/shared/model-locate-result/parse.mjs.map +1 -0
  80. package/dist/es/ai-model/shared/model-locate-result/pixel-bbox-mapper.mjs +64 -0
  81. package/dist/es/ai-model/shared/model-locate-result/pixel-bbox-mapper.mjs.map +1 -0
  82. package/dist/es/ai-model/shared/model-locate-result/types.mjs +0 -0
  83. package/dist/es/ai-model/types.mjs +0 -0
  84. package/dist/es/ai-model/workflows/image-preprocess.mjs +27 -0
  85. package/dist/es/ai-model/workflows/image-preprocess.mjs.map +1 -0
  86. package/dist/es/ai-model/workflows/inspect/index.mjs +2 -0
  87. package/dist/es/ai-model/workflows/inspect/locate-result-rect.mjs +23 -0
  88. package/dist/es/ai-model/workflows/inspect/locate-result-rect.mjs.map +1 -0
  89. package/dist/es/ai-model/workflows/inspect/search-area-mapping.mjs +18 -0
  90. package/dist/es/ai-model/workflows/inspect/search-area-mapping.mjs.map +1 -0
  91. package/dist/es/ai-model/workflows/inspect/types.mjs +0 -0
  92. package/dist/es/ai-model/workflows/planning/index.mjs +5 -0
  93. package/dist/es/ai-model/workflows/planning/index.mjs.map +1 -0
  94. package/dist/es/ai-model/workflows/planning/types.mjs +0 -0
  95. package/dist/es/common.mjs +2 -174
  96. package/dist/es/common.mjs.map +1 -1
  97. package/dist/es/device/index.mjs.map +1 -1
  98. package/dist/es/service/index.mjs +96 -69
  99. package/dist/es/service/index.mjs.map +1 -1
  100. package/dist/es/types.mjs.map +1 -1
  101. package/dist/es/utils.mjs +2 -2
  102. package/dist/es/yaml/player.mjs +4 -3
  103. package/dist/es/yaml/player.mjs.map +1 -1
  104. package/dist/lib/agent/agent.js +43 -53
  105. package/dist/lib/agent/agent.js.map +1 -1
  106. package/dist/lib/agent/task-builder.js +38 -18
  107. package/dist/lib/agent/task-builder.js.map +1 -1
  108. package/dist/lib/agent/tasks.js +23 -21
  109. package/dist/lib/agent/tasks.js.map +1 -1
  110. package/dist/lib/agent/utils.js +17 -17
  111. package/dist/lib/agent/utils.js.map +1 -1
  112. package/dist/lib/ai-model/connectivity.js +7 -3
  113. package/dist/lib/ai-model/connectivity.js.map +1 -1
  114. package/dist/lib/ai-model/errors.js +46 -0
  115. package/dist/lib/ai-model/errors.js.map +1 -0
  116. package/dist/lib/ai-model/index.js +7 -14
  117. package/dist/lib/ai-model/inspect.js +141 -144
  118. package/dist/lib/ai-model/inspect.js.map +1 -1
  119. package/dist/lib/ai-model/llm-planning.js +44 -26
  120. package/dist/lib/ai-model/llm-planning.js.map +1 -1
  121. package/dist/lib/ai-model/{auto-glm → models/auto-glm}/actions.js +22 -44
  122. package/dist/lib/ai-model/models/auto-glm/actions.js.map +1 -0
  123. package/dist/lib/ai-model/models/auto-glm/adapter.js +79 -0
  124. package/dist/lib/ai-model/models/auto-glm/adapter.js.map +1 -0
  125. package/dist/lib/ai-model/models/auto-glm/locate.js +146 -0
  126. package/dist/lib/ai-model/models/auto-glm/locate.js.map +1 -0
  127. package/dist/lib/ai-model/models/auto-glm/parser.js.map +1 -0
  128. package/dist/lib/ai-model/{auto-glm → models/auto-glm}/planning.js +8 -9
  129. package/dist/lib/ai-model/models/auto-glm/planning.js.map +1 -0
  130. package/dist/lib/ai-model/{auto-glm → models/auto-glm}/prompt.js +14 -16
  131. package/dist/lib/ai-model/models/auto-glm/prompt.js.map +1 -0
  132. package/dist/lib/ai-model/{auto-glm/util.js → models/default.js} +13 -13
  133. package/dist/lib/ai-model/models/default.js.map +1 -0
  134. package/dist/lib/ai-model/models/doubao.js +184 -0
  135. package/dist/lib/ai-model/models/doubao.js.map +1 -0
  136. package/dist/lib/ai-model/models/gemini.js +68 -0
  137. package/dist/lib/ai-model/models/gemini.js.map +1 -0
  138. package/dist/lib/ai-model/models/glm.js +71 -0
  139. package/dist/lib/ai-model/models/glm.js.map +1 -0
  140. package/dist/lib/ai-model/models/gpt.js +65 -0
  141. package/dist/lib/ai-model/models/gpt.js.map +1 -0
  142. package/dist/lib/ai-model/{service-caller/image-detail.js → models/index.js} +8 -7
  143. package/dist/lib/ai-model/models/index.js.map +1 -0
  144. package/dist/lib/ai-model/models/qwen.js +147 -0
  145. package/dist/lib/ai-model/models/qwen.js.map +1 -0
  146. package/dist/lib/ai-model/models/registry.js +85 -0
  147. package/dist/lib/ai-model/models/registry.js.map +1 -0
  148. package/dist/lib/ai-model/models/resolved.js +138 -0
  149. package/dist/lib/ai-model/models/resolved.js.map +1 -0
  150. package/dist/lib/ai-model/models/types.js +20 -0
  151. package/dist/lib/ai-model/models/types.js.map +1 -0
  152. package/dist/lib/ai-model/models/ui-tars/adapter.js +176 -0
  153. package/dist/lib/ai-model/models/ui-tars/adapter.js.map +1 -0
  154. package/dist/lib/ai-model/{ui-tars-planning.js → models/ui-tars/planning.js} +44 -62
  155. package/dist/lib/ai-model/models/ui-tars/planning.js.map +1 -0
  156. package/dist/lib/ai-model/prompt/extraction.js +3 -3
  157. package/dist/lib/ai-model/prompt/extraction.js.map +1 -1
  158. package/dist/lib/ai-model/prompt/llm-locator.js +11 -11
  159. package/dist/lib/ai-model/prompt/llm-locator.js.map +1 -1
  160. package/dist/lib/ai-model/prompt/llm-planning.js +25 -60
  161. package/dist/lib/ai-model/prompt/llm-planning.js.map +1 -1
  162. package/dist/lib/ai-model/prompt/llm-section-locator.js +15 -10
  163. package/dist/lib/ai-model/prompt/llm-section-locator.js.map +1 -1
  164. package/dist/lib/ai-model/prompt/locate-grounding-rules.js +43 -0
  165. package/dist/lib/ai-model/prompt/locate-grounding-rules.js.map +1 -0
  166. package/dist/lib/ai-model/prompt/locate-param-example.js +52 -0
  167. package/dist/lib/ai-model/prompt/locate-param-example.js.map +1 -0
  168. package/dist/lib/ai-model/prompt/playwright-generator.js +5 -5
  169. package/dist/lib/ai-model/prompt/playwright-generator.js.map +1 -1
  170. package/dist/lib/ai-model/prompt/yaml-generator.js +5 -5
  171. package/dist/lib/ai-model/prompt/yaml-generator.js.map +1 -1
  172. package/dist/lib/ai-model/prompts/locate-result-coordinates.js +150 -0
  173. package/dist/lib/ai-model/prompts/locate-result-coordinates.js.map +1 -0
  174. package/dist/lib/ai-model/service-caller/index.js +68 -199
  175. package/dist/lib/ai-model/service-caller/index.js.map +1 -1
  176. package/dist/lib/ai-model/service-caller/json.js +100 -0
  177. package/dist/lib/ai-model/service-caller/json.js.map +1 -0
  178. package/dist/lib/ai-model/shared/model-locate-result/bbox.js +117 -0
  179. package/dist/lib/ai-model/shared/model-locate-result/bbox.js.map +1 -0
  180. package/dist/lib/ai-model/shared/model-locate-result/factory.js +130 -0
  181. package/dist/lib/ai-model/shared/model-locate-result/factory.js.map +1 -0
  182. package/dist/lib/ai-model/{prompt/common.js → shared/model-locate-result/index.js} +9 -9
  183. package/dist/lib/ai-model/shared/model-locate-result/index.js.map +1 -0
  184. package/dist/lib/ai-model/shared/model-locate-result/parse.js +78 -0
  185. package/dist/lib/ai-model/shared/model-locate-result/parse.js.map +1 -0
  186. package/dist/lib/ai-model/shared/model-locate-result/pixel-bbox-mapper.js +98 -0
  187. package/dist/lib/ai-model/shared/model-locate-result/pixel-bbox-mapper.js.map +1 -0
  188. package/dist/lib/ai-model/shared/model-locate-result/types.js +20 -0
  189. package/dist/lib/ai-model/shared/model-locate-result/types.js.map +1 -0
  190. package/dist/lib/ai-model/types.js +20 -0
  191. package/dist/lib/ai-model/types.js.map +1 -0
  192. package/dist/lib/ai-model/workflows/image-preprocess.js +61 -0
  193. package/dist/lib/ai-model/workflows/image-preprocess.js.map +1 -0
  194. package/dist/lib/ai-model/workflows/inspect/index.js +50 -0
  195. package/dist/lib/ai-model/workflows/inspect/index.js.map +1 -0
  196. package/dist/lib/ai-model/workflows/inspect/locate-result-rect.js +60 -0
  197. package/dist/lib/ai-model/workflows/inspect/locate-result-rect.js.map +1 -0
  198. package/dist/lib/ai-model/workflows/inspect/search-area-mapping.js +52 -0
  199. package/dist/lib/ai-model/workflows/inspect/search-area-mapping.js.map +1 -0
  200. package/dist/lib/ai-model/workflows/inspect/types.js +20 -0
  201. package/dist/lib/ai-model/workflows/inspect/types.js.map +1 -0
  202. package/dist/lib/ai-model/{model-family.js → workflows/planning/index.js} +6 -7
  203. package/dist/lib/ai-model/workflows/planning/index.js.map +1 -0
  204. package/dist/lib/ai-model/workflows/planning/types.js +20 -0
  205. package/dist/lib/ai-model/workflows/planning/types.js.map +1 -0
  206. package/dist/lib/common.js +4 -206
  207. package/dist/lib/common.js.map +1 -1
  208. package/dist/lib/device/index.js.map +1 -1
  209. package/dist/lib/service/index.js +96 -69
  210. package/dist/lib/service/index.js.map +1 -1
  211. package/dist/lib/types.js.map +1 -1
  212. package/dist/lib/utils.js +2 -2
  213. package/dist/lib/yaml/player.js +4 -3
  214. package/dist/lib/yaml/player.js.map +1 -1
  215. package/dist/types/agent/agent.d.ts +14 -6
  216. package/dist/types/agent/task-builder.d.ts +2 -2
  217. package/dist/types/agent/tasks.d.ts +6 -6
  218. package/dist/types/agent/utils.d.ts +8 -5
  219. package/dist/types/ai-model/errors.d.ts +2 -0
  220. package/dist/types/ai-model/index.d.ts +2 -4
  221. package/dist/types/ai-model/inspect.d.ts +13 -33
  222. package/dist/types/ai-model/llm-planning.d.ts +6 -17
  223. package/dist/types/ai-model/{auto-glm → models/auto-glm}/actions.d.ts +2 -2
  224. package/dist/types/ai-model/models/auto-glm/adapter.d.ts +5 -0
  225. package/dist/types/ai-model/models/auto-glm/locate.d.ts +3 -0
  226. package/dist/types/ai-model/models/auto-glm/planning.d.ts +3 -0
  227. package/dist/types/ai-model/models/auto-glm/prompt.d.ts +4 -0
  228. package/dist/types/ai-model/models/default.d.ts +2 -0
  229. package/dist/types/ai-model/models/doubao.d.ts +10 -0
  230. package/dist/types/ai-model/models/gemini.d.ts +18 -0
  231. package/dist/types/ai-model/models/glm.d.ts +18 -0
  232. package/dist/types/ai-model/models/gpt.d.ts +18 -0
  233. package/dist/types/ai-model/models/index.d.ts +2 -0
  234. package/dist/types/ai-model/models/qwen.d.ts +30 -0
  235. package/dist/types/ai-model/models/registry.d.ts +81 -0
  236. package/dist/types/ai-model/models/resolved.d.ts +9 -0
  237. package/dist/types/ai-model/models/types.d.ts +102 -0
  238. package/dist/types/ai-model/models/ui-tars/adapter.d.ts +6 -0
  239. package/dist/types/ai-model/{ui-tars-planning.d.ts → models/ui-tars/planning.d.ts} +7 -11
  240. package/dist/types/ai-model/prompt/llm-locator.d.ts +2 -2
  241. package/dist/types/ai-model/prompt/llm-planning.d.ts +5 -5
  242. package/dist/types/ai-model/prompt/llm-section-locator.d.ts +2 -2
  243. package/dist/types/ai-model/prompt/locate-grounding-rules.d.ts +1 -0
  244. package/dist/types/ai-model/prompt/locate-param-example.d.ts +3 -0
  245. package/dist/types/ai-model/prompt/playwright-generator.d.ts +3 -3
  246. package/dist/types/ai-model/prompt/yaml-generator.d.ts +3 -3
  247. package/dist/types/ai-model/prompts/locate-result-coordinates.d.ts +6 -0
  248. package/dist/types/ai-model/service-caller/index.d.ts +19 -27
  249. package/dist/types/ai-model/service-caller/json.d.ts +9 -0
  250. package/dist/types/ai-model/shared/model-locate-result/bbox.d.ts +7 -0
  251. package/dist/types/ai-model/shared/model-locate-result/factory.d.ts +2 -0
  252. package/dist/types/ai-model/shared/model-locate-result/index.d.ts +3 -0
  253. package/dist/types/ai-model/shared/model-locate-result/parse.d.ts +5 -0
  254. package/dist/types/ai-model/shared/model-locate-result/pixel-bbox-mapper.d.ts +7 -0
  255. package/dist/types/ai-model/shared/model-locate-result/types.d.ts +157 -0
  256. package/dist/types/ai-model/types.d.ts +2 -0
  257. package/dist/types/ai-model/workflows/image-preprocess.d.ts +30 -0
  258. package/dist/types/ai-model/workflows/inspect/index.d.ts +1 -0
  259. package/dist/types/ai-model/workflows/inspect/locate-result-rect.d.ts +4 -0
  260. package/dist/types/ai-model/workflows/inspect/search-area-mapping.d.ts +3 -0
  261. package/dist/types/ai-model/workflows/inspect/types.d.ts +37 -0
  262. package/dist/types/ai-model/workflows/planning/index.d.ts +2 -0
  263. package/dist/types/ai-model/workflows/planning/types.d.ts +15 -0
  264. package/dist/types/common.d.ts +0 -30
  265. package/dist/types/device/index.d.ts +22 -22
  266. package/dist/types/service/index.d.ts +5 -4
  267. package/dist/types/types.d.ts +21 -9
  268. package/dist/types/yaml.d.ts +8 -2
  269. package/package.json +2 -2
  270. package/dist/es/ai-model/auto-glm/actions.mjs.map +0 -1
  271. package/dist/es/ai-model/auto-glm/index.mjs +0 -6
  272. package/dist/es/ai-model/auto-glm/parser.mjs.map +0 -1
  273. package/dist/es/ai-model/auto-glm/planning.mjs.map +0 -1
  274. package/dist/es/ai-model/auto-glm/prompt.mjs.map +0 -1
  275. package/dist/es/ai-model/auto-glm/util.mjs +0 -9
  276. package/dist/es/ai-model/auto-glm/util.mjs.map +0 -1
  277. package/dist/es/ai-model/model-family.mjs +0 -6
  278. package/dist/es/ai-model/model-family.mjs.map +0 -1
  279. package/dist/es/ai-model/prompt/common.mjs +0 -8
  280. package/dist/es/ai-model/prompt/common.mjs.map +0 -1
  281. package/dist/es/ai-model/service-caller/image-detail.mjs +0 -6
  282. package/dist/es/ai-model/service-caller/image-detail.mjs.map +0 -1
  283. package/dist/es/ai-model/ui-tars-planning.mjs.map +0 -1
  284. package/dist/lib/ai-model/auto-glm/actions.js.map +0 -1
  285. package/dist/lib/ai-model/auto-glm/index.js +0 -66
  286. package/dist/lib/ai-model/auto-glm/index.js.map +0 -1
  287. package/dist/lib/ai-model/auto-glm/parser.js.map +0 -1
  288. package/dist/lib/ai-model/auto-glm/planning.js.map +0 -1
  289. package/dist/lib/ai-model/auto-glm/prompt.js.map +0 -1
  290. package/dist/lib/ai-model/auto-glm/util.js.map +0 -1
  291. package/dist/lib/ai-model/model-family.js.map +0 -1
  292. package/dist/lib/ai-model/prompt/common.js.map +0 -1
  293. package/dist/lib/ai-model/service-caller/image-detail.js.map +0 -1
  294. package/dist/lib/ai-model/ui-tars-planning.js.map +0 -1
  295. package/dist/types/ai-model/auto-glm/index.d.ts +0 -6
  296. package/dist/types/ai-model/auto-glm/planning.d.ts +0 -12
  297. package/dist/types/ai-model/auto-glm/prompt.d.ts +0 -27
  298. package/dist/types/ai-model/auto-glm/util.d.ts +0 -13
  299. package/dist/types/ai-model/model-family.d.ts +0 -7
  300. package/dist/types/ai-model/prompt/common.d.ts +0 -2
  301. package/dist/types/ai-model/service-caller/image-detail.d.ts +0 -2
  302. /package/dist/es/ai-model/{auto-glm → models/auto-glm}/parser.mjs +0 -0
  303. /package/dist/lib/ai-model/{auto-glm → models/auto-glm}/parser.js +0 -0
  304. /package/dist/types/ai-model/{auto-glm → models/auto-glm}/parser.d.ts +0 -0
@@ -6,7 +6,7 @@ import { generateElementByRect } from "@midscene/shared/extractor";
6
6
  import { getDebug } from "@midscene/shared/logger";
7
7
  import { assert } from "@midscene/shared/utils";
8
8
  import { withUsageIntent } from "./usage-intent.mjs";
9
- import { ifPlanLocateParamIsBbox, matchElementFromCache, matchElementFromPlan, transformLogicalElementToScreenshot, transformLogicalRectToScreenshotRect } from "./utils.mjs";
9
+ import { ifPlanLocateParamHasLocatedPixelBbox, matchElementFromCache, matchElementFromPlan, transformLogicalElementToScreenshot, transformLogicalRectToScreenshotRect } from "./utils.mjs";
10
10
  function _define_property(obj, key, value) {
11
11
  if (key in obj) Object.defineProperty(obj, key, {
12
12
  value: value,
@@ -21,6 +21,18 @@ const debug = getDebug('agent:task-builder');
21
21
  function hasNonEmptyCache(cache) {
22
22
  return null != cache && 'object' == typeof cache && Object.keys(cache).length > 0;
23
23
  }
24
+ function invalidLocateElementReason(element) {
25
+ const values = [
26
+ element.center?.[0],
27
+ element.center?.[1],
28
+ element.rect?.left,
29
+ element.rect?.top,
30
+ element.rect?.width,
31
+ element.rect?.height
32
+ ];
33
+ if (values.some((value)=>'number' != typeof value || !Number.isFinite(value))) return `Invalid locate result coordinates: ${JSON.stringify(element)}`;
34
+ if (element.rect.width <= 0 || element.rect.height <= 0) return `Invalid locate result rect size: ${JSON.stringify(element)}`;
35
+ }
24
36
  function locatePlanForLocate(param) {
25
37
  const locate = 'string' == typeof param ? {
26
38
  prompt: param
@@ -33,13 +45,13 @@ function locatePlanForLocate(param) {
33
45
  return locatePlan;
34
46
  }
35
47
  class TaskBuilder {
36
- async build(plans, modelConfigForPlanning, modelConfigForDefaultIntent, options) {
48
+ async build(plans, planningModel, defaultModel, options) {
37
49
  const tasks = [];
38
50
  const cacheable = options?.cacheable;
39
51
  const context = {
40
52
  tasks,
41
- modelConfigForPlanning,
42
- modelConfigForDefaultIntent,
53
+ planningModel,
54
+ defaultModel,
43
55
  cacheable,
44
56
  deepLocate: options?.deepLocate,
45
57
  abortSignal: options?.abortSignal
@@ -88,7 +100,7 @@ class TaskBuilder {
88
100
  locateFields.forEach((field)=>{
89
101
  if (param[field]) {
90
102
  const locatePlan = locatePlanForLocate(param[field]);
91
- debug('will prepend locate param for field', `action.type=${planType}`, `param=${JSON.stringify(param[field])}`, `locatePlan=${JSON.stringify(locatePlan)}`, `hasBbox=${ifPlanLocateParamIsBbox(param[field])}`);
103
+ debug('will prepend locate param for field', `action.type=${planType}`, `param=${JSON.stringify(param[field])}`, `locatePlan=${JSON.stringify(locatePlan)}`, `hasLocatedPixelBbox=${ifPlanLocateParamHasLocatedPixelBbox(param[field])}`);
92
104
  const locateTask = this.createLocateTask(locatePlan, param[field], context, (result)=>{
93
105
  param[field] = result;
94
106
  });
@@ -172,7 +184,7 @@ class TaskBuilder {
172
184
  context.tasks.push(task);
173
185
  }
174
186
  createLocateTask(plan, detailedLocateParam, context, onResult) {
175
- const { cacheable, modelConfigForDefaultIntent, deepLocate, abortSignal } = context;
187
+ const { cacheable, defaultModel, deepLocate, abortSignal } = context;
176
188
  let locateParam = detailedLocateParam;
177
189
  if ('string' == typeof locateParam) locateParam = {
178
190
  prompt: locateParam
@@ -193,7 +205,8 @@ class TaskBuilder {
193
205
  executor: async (param, taskContext)=>{
194
206
  const { task } = taskContext;
195
207
  let { uiContext } = taskContext;
196
- assert(param?.prompt || param?.bbox, `No prompt or id or position or bbox to locate, param=${JSON.stringify(param)}`);
208
+ const paramWithLocatedPixelBbox = ifPlanLocateParamHasLocatedPixelBbox(param) ? param : void 0;
209
+ assert(param?.prompt || paramWithLocatedPixelBbox, `No prompt or id or position or locatedPixelBbox to locate, param=${JSON.stringify(param)}`);
197
210
  if (!uiContext) uiContext = await this.service.contextRetrieverFn();
198
211
  assert(uiContext, 'uiContext is required for Service task');
199
212
  const { shrunkShotToLogicalRatio } = uiContext;
@@ -211,11 +224,11 @@ class TaskBuilder {
211
224
  if (dump.taskInfo?.searchAreaUsage) task.searchAreaUsage = dump.taskInfo.searchAreaUsage;
212
225
  if (dump.taskInfo?.reasoning_content) task.reasoning_content = dump.taskInfo.reasoning_content;
213
226
  };
214
- const planLocatedElement = ifPlanLocateParamIsBbox(param) ? matchElementFromPlan(param) : void 0;
215
- const elementFromBbox = param.deepLocate ? void 0 : planLocatedElement;
216
- const isPlanHit = !!elementFromBbox;
227
+ const planLocatedElement = paramWithLocatedPixelBbox ? matchElementFromPlan(paramWithLocatedPixelBbox) : void 0;
228
+ const elementFromPlan = param.deepLocate ? void 0 : planLocatedElement;
229
+ const isPlanDirectHit = !!elementFromPlan;
217
230
  let rectFromXpath;
218
- if (!isPlanHit && param.xpath && this.interface.rectMatchesCacheFeature) try {
231
+ if (!isPlanDirectHit && param.xpath && this.interface.rectMatchesCacheFeature) try {
219
232
  rectFromXpath = await this.interface.rectMatchesCacheFeature({
220
233
  xpaths: [
221
234
  param.xpath
@@ -227,7 +240,7 @@ class TaskBuilder {
227
240
  const cachePrompt = param.prompt;
228
241
  const locateCacheRecord = this.taskCache?.matchLocateCache(cachePrompt);
229
242
  const cacheEntry = locateCacheRecord?.cacheContent?.cache;
230
- const elementFromCacheResult = isPlanHit || isXpathHit ? null : await matchElementFromCache({
243
+ const elementFromCacheResult = isPlanDirectHit || isXpathHit ? null : await matchElementFromCache({
231
244
  taskCache: this.taskCache,
232
245
  interfaceInstance: this.interface
233
246
  }, cacheEntry, cachePrompt, param.cacheable);
@@ -235,12 +248,12 @@ class TaskBuilder {
235
248
  const isCacheHit = !!elementFromCache;
236
249
  let elementFromAiLocate;
237
250
  const timing = taskContext.task.timing;
238
- if (!isXpathHit && !isCacheHit && !isPlanHit) try {
251
+ if (!isXpathHit && !isCacheHit && !isPlanDirectHit) try {
239
252
  setTimingFieldOnce(timing, 'callAiStart');
240
253
  locateResult = await this.service.locate(param, {
241
254
  context: uiContext,
242
255
  planLocatedElement
243
- }, modelConfigForDefaultIntent, abortSignal);
256
+ }, defaultModel, abortSignal);
244
257
  applyDump(locateResult.dump);
245
258
  elementFromAiLocate = locateResult.element;
246
259
  } catch (error) {
@@ -249,10 +262,17 @@ class TaskBuilder {
249
262
  } finally{
250
263
  setTimingFieldOnce(timing, 'callAiEnd');
251
264
  }
252
- const element = elementFromBbox || elementFromXpath || elementFromCache || elementFromAiLocate;
265
+ const element = elementFromPlan || elementFromXpath || elementFromCache || elementFromAiLocate;
266
+ if (element) {
267
+ const invalidElementReason = invalidLocateElementReason(element);
268
+ if (invalidElementReason) {
269
+ if (locateDump) throw new ServiceError(invalidElementReason, locateDump);
270
+ throw new Error(invalidElementReason);
271
+ }
272
+ }
253
273
  const locateCacheAlreadyExists = hasNonEmptyCache(locateCacheRecord?.cacheContent?.cache);
254
274
  let currentCacheEntry;
255
- if (element && this.taskCache && !isCacheHit && (!isPlanHit || !locateCacheAlreadyExists) && param?.cacheable !== false) if (this.interface.cacheFeatureForPoint) try {
275
+ if (element && this.taskCache && !isCacheHit && (!isPlanDirectHit || !locateCacheAlreadyExists) && param?.cacheable !== false) if (this.interface.cacheFeatureForPoint) try {
256
276
  let pointForCache = element.center;
257
277
  if (1 !== shrunkShotToLogicalRatio) {
258
278
  pointForCache = [
@@ -263,7 +283,7 @@ class TaskBuilder {
263
283
  }
264
284
  const feature = await this.interface.cacheFeatureForPoint(pointForCache, {
265
285
  targetDescription: 'string' == typeof param.prompt ? param.prompt : param.prompt?.prompt,
266
- modelConfig: modelConfigForDefaultIntent
286
+ modelRuntime: defaultModel
267
287
  });
268
288
  if (hasNonEmptyCache(feature)) {
269
289
  debug('update cache, prompt: %s, cache: %o', cachePrompt, feature);
@@ -283,10 +303,10 @@ class TaskBuilder {
283
303
  throw new Error(`Element not found: ${param.prompt}`);
284
304
  }
285
305
  let hitBy;
286
- if (isPlanHit) hitBy = {
306
+ if (isPlanDirectHit && paramWithLocatedPixelBbox) hitBy = {
287
307
  from: 'Plan',
288
308
  context: {
289
- bbox: param.bbox
309
+ locatedPixelBbox: paramWithLocatedPixelBbox.locatedPixelBbox
290
310
  }
291
311
  };
292
312
  else if (isXpathHit) hitBy = {
@@ -1 +1 @@
1
- {"version":3,"file":"agent/task-builder.mjs","sources":["../../../src/agent/task-builder.ts"],"sourcesContent":["import { findAllMidsceneLocatorField, parseActionParam } from '@/ai-model';\nimport type { AbstractInterface } from '@/device';\nimport type Service from '@/service';\nimport { setTimingFieldOnce } from '@/task-timing';\nimport type {\n AIUsageInfo,\n DetailedLocateParam,\n DeviceAction,\n ElementCacheFeature,\n ExecutionTaskActionApply,\n ExecutionTaskApply,\n ExecutionTaskHitBy,\n ExecutionTaskPlanningLocateApply,\n LocateResultElement,\n LocateResultWithDump,\n PlanningAction,\n PlanningLocateParam,\n Rect,\n ServiceDump,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport { sleep } from '@/utils';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { generateElementByRect } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TaskCache } from './task-cache';\nimport { withUsageIntent } from './usage-intent';\nimport {\n ifPlanLocateParamIsBbox,\n matchElementFromCache,\n matchElementFromPlan,\n transformLogicalElementToScreenshot,\n transformLogicalRectToScreenshotRect,\n} from './utils';\n\nconst debug = getDebug('agent:task-builder');\n\n/**\n * Check if a cache object is non-empty\n */\nfunction hasNonEmptyCache(cache: unknown): boolean {\n return (\n cache !== null &&\n cache !== undefined &&\n typeof cache === 'object' &&\n Object.keys(cache).length > 0\n );\n}\n\nexport function locatePlanForLocate(param: string | DetailedLocateParam) {\n const locate = typeof param === 'string' ? { prompt: param } : param;\n const locatePlan: PlanningAction<PlanningLocateParam> = {\n type: 'Locate',\n param: locate,\n thought: '',\n };\n return locatePlan;\n}\n\ninterface TaskBuilderDeps {\n interfaceInstance: AbstractInterface;\n service: Service;\n taskCache?: TaskCache;\n actionSpace: DeviceAction[];\n waitAfterAction?: number;\n}\n\ninterface BuildOptions {\n cacheable?: boolean;\n deepLocate?: boolean;\n abortSignal?: AbortSignal;\n}\n\ninterface PlanBuildContext {\n tasks: ExecutionTaskApply[];\n modelConfigForPlanning: IModelConfig;\n modelConfigForDefaultIntent: IModelConfig;\n cacheable?: boolean;\n deepLocate?: boolean;\n abortSignal?: AbortSignal;\n}\n\nexport class TaskBuilder {\n private readonly interface: AbstractInterface;\n\n private readonly service: Service;\n\n private readonly taskCache?: TaskCache;\n\n private readonly actionSpace: DeviceAction[];\n\n private readonly waitAfterAction?: number;\n\n constructor({\n interfaceInstance,\n service,\n taskCache,\n actionSpace,\n waitAfterAction,\n }: TaskBuilderDeps) {\n this.interface = interfaceInstance;\n this.service = service;\n this.taskCache = taskCache;\n this.actionSpace = actionSpace;\n this.waitAfterAction = waitAfterAction;\n }\n\n public async build(\n plans: PlanningAction[],\n modelConfigForPlanning: IModelConfig,\n modelConfigForDefaultIntent: IModelConfig,\n options?: BuildOptions,\n ): Promise<{ tasks: ExecutionTaskApply[] }> {\n const tasks: ExecutionTaskApply[] = [];\n const cacheable = options?.cacheable;\n\n const context: PlanBuildContext = {\n tasks,\n modelConfigForPlanning,\n modelConfigForDefaultIntent,\n cacheable,\n deepLocate: options?.deepLocate,\n abortSignal: options?.abortSignal,\n };\n\n type PlanHandler = (plan: PlanningAction) => Promise<void> | void;\n\n const planHandlers = new Map<string, PlanHandler>([\n [\n 'Locate',\n (plan) =>\n this.handleLocatePlan(\n plan as PlanningAction<PlanningLocateParam>,\n context,\n ),\n ],\n ['Finished', (plan) => this.handleFinishedPlan(plan, context)],\n ]);\n\n const defaultHandler: PlanHandler = (plan) =>\n this.handleActionPlan(plan, context);\n\n for (const plan of plans) {\n const handler = planHandlers.get(plan.type) ?? defaultHandler;\n await handler(plan);\n }\n\n return {\n tasks,\n };\n }\n\n private handleFinishedPlan(\n plan: PlanningAction,\n context: PlanBuildContext,\n ): void {\n const taskActionFinished: ExecutionTaskActionApply<null> = {\n type: 'Action Space',\n subType: 'Finished',\n param: null,\n thought: plan.thought,\n executor: async () => {},\n };\n context.tasks.push(taskActionFinished);\n }\n\n private async handleLocatePlan(\n plan: PlanningAction<PlanningLocateParam>,\n context: PlanBuildContext,\n ): Promise<void> {\n const taskLocate = this.createLocateTask(plan, plan.param, context);\n context.tasks.push(taskLocate);\n }\n\n private async handleActionPlan(\n plan: PlanningAction,\n context: PlanBuildContext,\n ): Promise<void> {\n const planType = plan.type;\n const actionSpace = this.actionSpace;\n const action = actionSpace.find((item) => item.name === planType);\n const param = plan.param;\n\n if (!action) {\n throw new Error(`Action type '${planType}' not found`);\n }\n\n const locateFields = action\n ? findAllMidsceneLocatorField(action.paramSchema)\n : [];\n\n const requiredLocateFields = action\n ? findAllMidsceneLocatorField(action.paramSchema, true)\n : [];\n\n locateFields.forEach((field) => {\n if (param[field]) {\n // Always use createLocateTask for all locate params (including bbox)\n // This ensures cache writing happens even when bbox is available\n const locatePlan = locatePlanForLocate(param[field]);\n debug(\n 'will prepend locate param for field',\n `action.type=${planType}`,\n `param=${JSON.stringify(param[field])}`,\n `locatePlan=${JSON.stringify(locatePlan)}`,\n `hasBbox=${ifPlanLocateParamIsBbox(param[field])}`,\n );\n const locateTask = this.createLocateTask(\n locatePlan,\n param[field],\n context,\n (result) => {\n param[field] = result;\n },\n );\n context.tasks.push(locateTask);\n } else {\n assert(\n !requiredLocateFields.includes(field),\n `Required locate field '${field}' is not provided for action ${planType}`,\n );\n debug(`field '${field}' is not provided for action ${planType}`);\n }\n });\n\n const task: ExecutionTaskApply<\n 'Action Space',\n any,\n { success: boolean; action: string; param: any },\n void\n > = {\n type: 'Action Space',\n subType: planType,\n thought: plan.thought,\n param: plan.param,\n executor: async (param, taskContext) => {\n const timing = taskContext.task.timing;\n\n debug(\n 'executing action',\n planType,\n param,\n `taskContext.element.center: ${taskContext.element?.center}`,\n );\n\n const uiContext = taskContext.uiContext;\n assert(uiContext, 'uiContext is required for Action task');\n\n requiredLocateFields.forEach((field) => {\n assert(\n param[field],\n `field '${field}' is required for action ${planType} but not provided. Cannot execute action ${planType}.`,\n );\n });\n\n setTimingFieldOnce(timing, 'beforeInvokeActionHookStart');\n const delayBeforeRunner = action.delayBeforeRunner ?? 200;\n try {\n await Promise.all([\n (async () => {\n if (this.interface.beforeInvokeAction) {\n debug(\n `will call \"beforeInvokeAction\" for interface with action name ${action.name}`,\n );\n await this.interface.beforeInvokeAction(action.name, param);\n debug(\n `called \"beforeInvokeAction\" for interface with action name ${action.name}`,\n );\n }\n })(),\n delayBeforeRunner > 0\n ? sleep(delayBeforeRunner)\n : Promise.resolve(),\n ]);\n } catch (originalError: any) {\n const originalMessage =\n originalError?.message || String(originalError);\n throw new Error(\n `error in running beforeInvokeAction for ${action.name}: ${originalMessage}`,\n { cause: originalError },\n );\n }\n setTimingFieldOnce(timing, 'beforeInvokeActionHookEnd');\n\n const { shrunkShotToLogicalRatio } = uiContext;\n if (shrunkShotToLogicalRatio === undefined) {\n throw new Error(\n 'shrunkShotToLogicalRatio is not defined in Action task',\n );\n }\n\n if (action.paramSchema) {\n try {\n param = parseActionParam(param, action.paramSchema, {\n shrunkShotToLogicalRatio,\n });\n } catch (error: any) {\n throw new Error(\n `Invalid parameters for action ${action.name}: ${error.message}\\nParameters: ${JSON.stringify(param)}`,\n { cause: error },\n );\n }\n }\n\n setTimingFieldOnce(timing, 'callActionStart');\n\n debug('calling action', action.name);\n const actionFn = action.call.bind(this.interface);\n const actionResult = await actionFn(param, taskContext);\n setTimingFieldOnce(timing, 'callActionEnd');\n debug('called action', action.name, 'result:', actionResult);\n\n setTimingFieldOnce(timing, 'afterInvokeActionHookStart');\n\n const delayAfterRunner =\n action.delayAfterRunner ?? this.waitAfterAction ?? 300;\n if (delayAfterRunner > 0) {\n await sleep(delayAfterRunner);\n }\n\n try {\n if (this.interface.afterInvokeAction) {\n debug(\n `will call \"afterInvokeAction\" for interface with action name ${action.name}`,\n );\n await this.interface.afterInvokeAction(action.name, param);\n debug(\n `called \"afterInvokeAction\" for interface with action name ${action.name}`,\n );\n }\n } catch (originalError: any) {\n const originalMessage =\n originalError?.message || String(originalError);\n throw new Error(\n `error in running afterInvokeAction for ${action.name}: ${originalMessage}`,\n { cause: originalError },\n );\n }\n\n setTimingFieldOnce(timing, 'afterInvokeActionHookEnd');\n\n return {\n output: actionResult,\n };\n },\n };\n\n context.tasks.push(task);\n }\n\n private createLocateTask(\n plan: PlanningAction<PlanningLocateParam>,\n detailedLocateParam: DetailedLocateParam | string,\n context: PlanBuildContext,\n onResult?: (result: LocateResultElement) => void,\n ): ExecutionTaskPlanningLocateApply {\n const { cacheable, modelConfigForDefaultIntent, deepLocate, abortSignal } =\n context;\n\n let locateParam = detailedLocateParam;\n\n if (typeof locateParam === 'string') {\n locateParam = {\n prompt: locateParam,\n };\n }\n\n if (cacheable !== undefined) {\n locateParam = {\n ...locateParam,\n cacheable,\n };\n }\n\n if (deepLocate && !locateParam.deepLocate) {\n locateParam = {\n ...locateParam,\n deepLocate: true,\n };\n }\n\n const taskLocator: ExecutionTaskPlanningLocateApply = {\n type: 'Planning',\n subType: 'Locate',\n param: locateParam,\n thought: plan.thought,\n executor: async (param, taskContext) => {\n const { task } = taskContext;\n let { uiContext } = taskContext;\n\n assert(\n param?.prompt || param?.bbox,\n `No prompt or id or position or bbox to locate, param=${JSON.stringify(\n param,\n )}`,\n );\n\n if (!uiContext) {\n uiContext = await this.service.contextRetrieverFn();\n }\n\n assert(uiContext, 'uiContext is required for Service task');\n\n const { shrunkShotToLogicalRatio } = uiContext;\n\n if (shrunkShotToLogicalRatio === undefined) {\n throw new Error(\n 'shrunkShotToLogicalRatio is not defined in locate task',\n );\n }\n\n let locateDump: ServiceDump | undefined;\n let locateResult: LocateResultWithDump | undefined;\n\n const applyDump = (dump?: ServiceDump) => {\n if (!dump) {\n return;\n }\n locateDump = dump;\n task.log = {\n dump,\n rawResponse: dump.taskInfo?.rawResponse,\n };\n task.usage = withUsageIntent(dump.taskInfo?.usage, 'default');\n if (dump.taskInfo?.searchAreaUsage) {\n task.searchAreaUsage = dump.taskInfo.searchAreaUsage;\n }\n if (dump.taskInfo?.reasoning_content) {\n task.reasoning_content = dump.taskInfo.reasoning_content;\n }\n };\n\n const planLocatedElement = ifPlanLocateParamIsBbox(param)\n ? matchElementFromPlan(param)\n : undefined;\n\n // from bbox (plan hit)\n // when deepLocate is enabled, bbox should be used as search area hint,\n // not as a final direct hit\n const elementFromBbox = param.deepLocate\n ? undefined\n : planLocatedElement;\n const isPlanHit = !!elementFromBbox;\n\n // from xpath\n let rectFromXpath: Rect | undefined;\n if (\n !isPlanHit &&\n param.xpath &&\n this.interface.rectMatchesCacheFeature\n ) {\n try {\n rectFromXpath = await this.interface.rectMatchesCacheFeature({\n xpaths: [param.xpath],\n });\n } catch {\n // xpath locate failed, allow fallback to cache or AI locate\n }\n }\n\n const elementFromXpath = rectFromXpath\n ? generateElementByRect(\n // rectFromXpath is in logical coordinates, which should be transformed to screenshot coordinates;\n transformLogicalRectToScreenshotRect(\n rectFromXpath,\n shrunkShotToLogicalRatio,\n ),\n typeof param.prompt === 'string'\n ? param.prompt\n : param.prompt?.prompt || '',\n )\n : undefined;\n\n const isXpathHit = !!elementFromXpath;\n\n const cachePrompt = param.prompt;\n const locateCacheRecord = this.taskCache?.matchLocateCache(cachePrompt);\n const cacheEntry = locateCacheRecord?.cacheContent?.cache;\n\n const elementFromCacheResult =\n isPlanHit || isXpathHit\n ? null\n : await matchElementFromCache(\n {\n taskCache: this.taskCache,\n interfaceInstance: this.interface,\n },\n cacheEntry,\n cachePrompt,\n param.cacheable,\n );\n\n // elementFromCacheResult is in logical coordinates, which should be transformed to screenshot coordinates;\n const elementFromCache = elementFromCacheResult\n ? transformLogicalElementToScreenshot(\n elementFromCacheResult,\n shrunkShotToLogicalRatio,\n )\n : undefined;\n\n const isCacheHit = !!elementFromCache;\n\n let elementFromAiLocate: LocateResultElement | null | undefined;\n const timing = taskContext.task.timing;\n if (!isXpathHit && !isCacheHit && !isPlanHit) {\n try {\n setTimingFieldOnce(timing, 'callAiStart');\n locateResult = await this.service.locate(\n param,\n {\n context: uiContext,\n planLocatedElement,\n },\n modelConfigForDefaultIntent,\n abortSignal,\n );\n applyDump(locateResult.dump);\n elementFromAiLocate = locateResult.element;\n } catch (error) {\n if (error instanceof ServiceError) {\n applyDump(error.dump);\n }\n throw error;\n } finally {\n setTimingFieldOnce(timing, 'callAiEnd');\n }\n }\n\n const element =\n elementFromBbox ||\n elementFromXpath ||\n elementFromCache ||\n elementFromAiLocate;\n\n // Check if locate cache already exists (for planHitFlag case)\n const locateCacheAlreadyExists = hasNonEmptyCache(\n locateCacheRecord?.cacheContent?.cache,\n );\n\n let currentCacheEntry: ElementCacheFeature | undefined;\n // Write cache if:\n // 1. element found\n // 2. taskCache enabled\n // 3. not a cache hit (otherwise we'd be writing what we just read)\n // 4. not already cached for plan hit case (avoid redundant writes), OR allow update if cache validation failed\n // 5. cacheable is not explicitly false\n if (\n element &&\n this.taskCache &&\n !isCacheHit &&\n (!isPlanHit || !locateCacheAlreadyExists) &&\n param?.cacheable !== false\n ) {\n if (this.interface.cacheFeatureForPoint) {\n try {\n // Transform coordinates to logical space for cacheFeatureForPoint\n // cacheFeatureForPoint needs logical coordinates to locate elements in DOM\n let pointForCache: [number, number] = element.center;\n if (shrunkShotToLogicalRatio !== 1) {\n pointForCache = [\n Math.round(element.center[0] / shrunkShotToLogicalRatio),\n Math.round(element.center[1] / shrunkShotToLogicalRatio),\n ];\n debug(\n 'Transformed coordinates for cacheFeatureForPoint: %o -> %o',\n element.center,\n pointForCache,\n );\n }\n\n const feature = await this.interface.cacheFeatureForPoint(\n pointForCache,\n {\n targetDescription:\n typeof param.prompt === 'string'\n ? param.prompt\n : param.prompt?.prompt,\n modelConfig: modelConfigForDefaultIntent,\n },\n );\n if (hasNonEmptyCache(feature)) {\n debug(\n 'update cache, prompt: %s, cache: %o',\n cachePrompt,\n feature,\n );\n currentCacheEntry = feature;\n this.taskCache.updateOrAppendCacheRecord(\n {\n type: 'locate',\n prompt: cachePrompt,\n cache: feature,\n },\n locateCacheRecord,\n );\n } else {\n debug(\n 'no cache data returned, skip cache update, prompt: %s',\n cachePrompt,\n );\n }\n } catch (error) {\n debug('cacheFeatureForPoint failed: %s', error);\n }\n } else {\n debug('cacheFeatureForPoint is not supported, skip cache update');\n }\n }\n\n if (!element) {\n if (locateDump) {\n throw new ServiceError(\n `Element not found : ${param.prompt}`,\n locateDump,\n );\n }\n throw new Error(`Element not found: ${param.prompt}`);\n }\n\n let hitBy: ExecutionTaskHitBy | undefined;\n\n if (isPlanHit) {\n hitBy = {\n from: 'Plan',\n context: {\n bbox: param.bbox,\n },\n };\n } else if (isXpathHit) {\n hitBy = {\n from: 'User expected path',\n context: {\n xpath: param.xpath,\n },\n };\n } else if (isCacheHit) {\n hitBy = {\n from: 'Cache',\n context: {\n cacheEntry,\n cacheToSave: currentCacheEntry,\n },\n };\n }\n\n onResult?.(element);\n\n return {\n output: {\n element: {\n ...element,\n // backward compatibility for aiLocate, which return value needs a dpr field\n dpr: uiContext.deprecatedDpr,\n },\n },\n hitBy,\n };\n },\n };\n\n return taskLocator;\n }\n}\n"],"names":["debug","getDebug","hasNonEmptyCache","cache","Object","locatePlanForLocate","param","locate","locatePlan","TaskBuilder","plans","modelConfigForPlanning","modelConfigForDefaultIntent","options","tasks","cacheable","context","planHandlers","Map","plan","defaultHandler","handler","taskActionFinished","taskLocate","planType","actionSpace","action","item","Error","locateFields","findAllMidsceneLocatorField","requiredLocateFields","field","JSON","ifPlanLocateParamIsBbox","locateTask","result","assert","task","taskContext","timing","uiContext","setTimingFieldOnce","delayBeforeRunner","Promise","sleep","originalError","originalMessage","String","shrunkShotToLogicalRatio","undefined","parseActionParam","error","actionFn","actionResult","delayAfterRunner","detailedLocateParam","onResult","deepLocate","abortSignal","locateParam","taskLocator","locateDump","locateResult","applyDump","dump","withUsageIntent","planLocatedElement","matchElementFromPlan","elementFromBbox","isPlanHit","rectFromXpath","elementFromXpath","generateElementByRect","transformLogicalRectToScreenshotRect","isXpathHit","cachePrompt","locateCacheRecord","cacheEntry","elementFromCacheResult","matchElementFromCache","elementFromCache","transformLogicalElementToScreenshot","isCacheHit","elementFromAiLocate","ServiceError","element","locateCacheAlreadyExists","currentCacheEntry","pointForCache","Math","feature","hitBy","interfaceInstance","service","taskCache","waitAfterAction"],"mappings":";;;;;;;;;;;;;;;;;;;AAoCA,MAAMA,QAAQC,SAAS;AAKvB,SAASC,iBAAiBC,KAAc;IACtC,OACEA,QAAAA,SAEA,AAAiB,YAAjB,OAAOA,SACPC,OAAO,IAAI,CAACD,OAAO,MAAM,GAAG;AAEhC;AAEO,SAASE,oBAAoBC,KAAmC;IACrE,MAAMC,SAAS,AAAiB,YAAjB,OAAOD,QAAqB;QAAE,QAAQA;IAAM,IAAIA;IAC/D,MAAME,aAAkD;QACtD,MAAM;QACN,OAAOD;QACP,SAAS;IACX;IACA,OAAOC;AACT;AAyBO,MAAMC;IAyBX,MAAa,MACXC,KAAuB,EACvBC,sBAAoC,EACpCC,2BAAyC,EACzCC,OAAsB,EACoB;QAC1C,MAAMC,QAA8B,EAAE;QACtC,MAAMC,YAAYF,SAAS;QAE3B,MAAMG,UAA4B;YAChCF;YACAH;YACAC;YACAG;YACA,YAAYF,SAAS;YACrB,aAAaA,SAAS;QACxB;QAIA,MAAMI,eAAe,IAAIC,IAAyB;YAChD;gBACE;gBACA,CAACC,OACC,IAAI,CAAC,gBAAgB,CACnBA,MACAH;aAEL;YACD;gBAAC;gBAAY,CAACG,OAAS,IAAI,CAAC,kBAAkB,CAACA,MAAMH;aAAS;SAC/D;QAED,MAAMI,iBAA8B,CAACD,OACnC,IAAI,CAAC,gBAAgB,CAACA,MAAMH;QAE9B,KAAK,MAAMG,QAAQT,MAAO;YACxB,MAAMW,UAAUJ,aAAa,GAAG,CAACE,KAAK,IAAI,KAAKC;YAC/C,MAAMC,QAAQF;QAChB;QAEA,OAAO;YACLL;QACF;IACF;IAEQ,mBACNK,IAAoB,EACpBH,OAAyB,EACnB;QACN,MAAMM,qBAAqD;YACzD,MAAM;YACN,SAAS;YACT,OAAO;YACP,SAASH,KAAK,OAAO;YACrB,UAAU,WAAa;QACzB;QACAH,QAAQ,KAAK,CAAC,IAAI,CAACM;IACrB;IAEA,MAAc,iBACZH,IAAyC,EACzCH,OAAyB,EACV;QACf,MAAMO,aAAa,IAAI,CAAC,gBAAgB,CAACJ,MAAMA,KAAK,KAAK,EAAEH;QAC3DA,QAAQ,KAAK,CAAC,IAAI,CAACO;IACrB;IAEA,MAAc,iBACZJ,IAAoB,EACpBH,OAAyB,EACV;QACf,MAAMQ,WAAWL,KAAK,IAAI;QAC1B,MAAMM,cAAc,IAAI,CAAC,WAAW;QACpC,MAAMC,SAASD,YAAY,IAAI,CAAC,CAACE,OAASA,KAAK,IAAI,KAAKH;QACxD,MAAMlB,QAAQa,KAAK,KAAK;QAExB,IAAI,CAACO,QACH,MAAM,IAAIE,MAAM,CAAC,aAAa,EAAEJ,SAAS,WAAW,CAAC;QAGvD,MAAMK,eAAeH,SACjBI,4BAA4BJ,OAAO,WAAW,IAC9C,EAAE;QAEN,MAAMK,uBAAuBL,SACzBI,4BAA4BJ,OAAO,WAAW,EAAE,QAChD,EAAE;QAENG,aAAa,OAAO,CAAC,CAACG;YACpB,IAAI1B,KAAK,CAAC0B,MAAM,EAAE;gBAGhB,MAAMxB,aAAaH,oBAAoBC,KAAK,CAAC0B,MAAM;gBACnDhC,MACE,uCACA,CAAC,YAAY,EAAEwB,UAAU,EACzB,CAAC,MAAM,EAAES,KAAK,SAAS,CAAC3B,KAAK,CAAC0B,MAAM,GAAG,EACvC,CAAC,WAAW,EAAEC,KAAK,SAAS,CAACzB,aAAa,EAC1C,CAAC,QAAQ,EAAE0B,wBAAwB5B,KAAK,CAAC0B,MAAM,GAAG;gBAEpD,MAAMG,aAAa,IAAI,CAAC,gBAAgB,CACtC3B,YACAF,KAAK,CAAC0B,MAAM,EACZhB,SACA,CAACoB;oBACC9B,KAAK,CAAC0B,MAAM,GAAGI;gBACjB;gBAEFpB,QAAQ,KAAK,CAAC,IAAI,CAACmB;YACrB,OAAO;gBACLE,OACE,CAACN,qBAAqB,QAAQ,CAACC,QAC/B,CAAC,uBAAuB,EAAEA,MAAM,6BAA6B,EAAER,UAAU;gBAE3ExB,MAAM,CAAC,OAAO,EAAEgC,MAAM,6BAA6B,EAAER,UAAU;YACjE;QACF;QAEA,MAAMc,OAKF;YACF,MAAM;YACN,SAASd;YACT,SAASL,KAAK,OAAO;YACrB,OAAOA,KAAK,KAAK;YACjB,UAAU,OAAOb,OAAOiC;gBACtB,MAAMC,SAASD,YAAY,IAAI,CAAC,MAAM;gBAEtCvC,MACE,oBACAwB,UACAlB,OACA,CAAC,4BAA4B,EAAEiC,YAAY,OAAO,EAAE,QAAQ;gBAG9D,MAAME,YAAYF,YAAY,SAAS;gBACvCF,OAAOI,WAAW;gBAElBV,qBAAqB,OAAO,CAAC,CAACC;oBAC5BK,OACE/B,KAAK,CAAC0B,MAAM,EACZ,CAAC,OAAO,EAAEA,MAAM,yBAAyB,EAAER,SAAS,yCAAyC,EAAEA,SAAS,CAAC,CAAC;gBAE9G;gBAEAkB,mBAAmBF,QAAQ;gBAC3B,MAAMG,oBAAoBjB,OAAO,iBAAiB,IAAI;gBACtD,IAAI;oBACF,MAAMkB,QAAQ,GAAG,CAAC;wBACf;4BACC,IAAI,IAAI,CAAC,SAAS,CAAC,kBAAkB,EAAE;gCACrC5C,MACE,CAAC,8DAA8D,EAAE0B,OAAO,IAAI,EAAE;gCAEhF,MAAM,IAAI,CAAC,SAAS,CAAC,kBAAkB,CAACA,OAAO,IAAI,EAAEpB;gCACrDN,MACE,CAAC,2DAA2D,EAAE0B,OAAO,IAAI,EAAE;4BAE/E;wBACF;wBACAiB,oBAAoB,IAChBE,MAAMF,qBACNC,QAAQ,OAAO;qBACpB;gBACH,EAAE,OAAOE,eAAoB;oBAC3B,MAAMC,kBACJD,eAAe,WAAWE,OAAOF;oBACnC,MAAM,IAAIlB,MACR,CAAC,wCAAwC,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEqB,iBAAiB,EAC5E;wBAAE,OAAOD;oBAAc;gBAE3B;gBACAJ,mBAAmBF,QAAQ;gBAE3B,MAAM,EAAES,wBAAwB,EAAE,GAAGR;gBACrC,IAAIQ,AAA6BC,WAA7BD,0BACF,MAAM,IAAIrB,MACR;gBAIJ,IAAIF,OAAO,WAAW,EACpB,IAAI;oBACFpB,QAAQ6C,iBAAiB7C,OAAOoB,OAAO,WAAW,EAAE;wBAClDuB;oBACF;gBACF,EAAE,OAAOG,OAAY;oBACnB,MAAM,IAAIxB,MACR,CAAC,8BAA8B,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAE0B,MAAM,OAAO,CAAC,cAAc,EAAEnB,KAAK,SAAS,CAAC3B,QAAQ,EACtG;wBAAE,OAAO8C;oBAAM;gBAEnB;gBAGFV,mBAAmBF,QAAQ;gBAE3BxC,MAAM,kBAAkB0B,OAAO,IAAI;gBACnC,MAAM2B,WAAW3B,OAAO,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS;gBAChD,MAAM4B,eAAe,MAAMD,SAAS/C,OAAOiC;gBAC3CG,mBAAmBF,QAAQ;gBAC3BxC,MAAM,iBAAiB0B,OAAO,IAAI,EAAE,WAAW4B;gBAE/CZ,mBAAmBF,QAAQ;gBAE3B,MAAMe,mBACJ7B,OAAO,gBAAgB,IAAI,IAAI,CAAC,eAAe,IAAI;gBACrD,IAAI6B,mBAAmB,GACrB,MAAMV,MAAMU;gBAGd,IAAI;oBACF,IAAI,IAAI,CAAC,SAAS,CAAC,iBAAiB,EAAE;wBACpCvD,MACE,CAAC,6DAA6D,EAAE0B,OAAO,IAAI,EAAE;wBAE/E,MAAM,IAAI,CAAC,SAAS,CAAC,iBAAiB,CAACA,OAAO,IAAI,EAAEpB;wBACpDN,MACE,CAAC,0DAA0D,EAAE0B,OAAO,IAAI,EAAE;oBAE9E;gBACF,EAAE,OAAOoB,eAAoB;oBAC3B,MAAMC,kBACJD,eAAe,WAAWE,OAAOF;oBACnC,MAAM,IAAIlB,MACR,CAAC,uCAAuC,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEqB,iBAAiB,EAC3E;wBAAE,OAAOD;oBAAc;gBAE3B;gBAEAJ,mBAAmBF,QAAQ;gBAE3B,OAAO;oBACL,QAAQc;gBACV;YACF;QACF;QAEAtC,QAAQ,KAAK,CAAC,IAAI,CAACsB;IACrB;IAEQ,iBACNnB,IAAyC,EACzCqC,mBAAiD,EACjDxC,OAAyB,EACzByC,QAAgD,EACd;QAClC,MAAM,EAAE1C,SAAS,EAAEH,2BAA2B,EAAE8C,UAAU,EAAEC,WAAW,EAAE,GACvE3C;QAEF,IAAI4C,cAAcJ;QAElB,IAAI,AAAuB,YAAvB,OAAOI,aACTA,cAAc;YACZ,QAAQA;QACV;QAGF,IAAI7C,AAAcmC,WAAdnC,WACF6C,cAAc;YACZ,GAAGA,WAAW;YACd7C;QACF;QAGF,IAAI2C,cAAc,CAACE,YAAY,UAAU,EACvCA,cAAc;YACZ,GAAGA,WAAW;YACd,YAAY;QACd;QAGF,MAAMC,cAAgD;YACpD,MAAM;YACN,SAAS;YACT,OAAOD;YACP,SAASzC,KAAK,OAAO;YACrB,UAAU,OAAOb,OAAOiC;gBACtB,MAAM,EAAED,IAAI,EAAE,GAAGC;gBACjB,IAAI,EAAEE,SAAS,EAAE,GAAGF;gBAEpBF,OACE/B,OAAO,UAAUA,OAAO,MACxB,CAAC,qDAAqD,EAAE2B,KAAK,SAAS,CACpE3B,QACC;gBAGL,IAAI,CAACmC,WACHA,YAAY,MAAM,IAAI,CAAC,OAAO,CAAC,kBAAkB;gBAGnDJ,OAAOI,WAAW;gBAElB,MAAM,EAAEQ,wBAAwB,EAAE,GAAGR;gBAErC,IAAIQ,AAA6BC,WAA7BD,0BACF,MAAM,IAAIrB,MACR;gBAIJ,IAAIkC;gBACJ,IAAIC;gBAEJ,MAAMC,YAAY,CAACC;oBACjB,IAAI,CAACA,MACH;oBAEFH,aAAaG;oBACb3B,KAAK,GAAG,GAAG;wBACT2B;wBACA,aAAaA,KAAK,QAAQ,EAAE;oBAC9B;oBACA3B,KAAK,KAAK,GAAG4B,gBAAgBD,KAAK,QAAQ,EAAE,OAAO;oBACnD,IAAIA,KAAK,QAAQ,EAAE,iBACjB3B,KAAK,eAAe,GAAG2B,KAAK,QAAQ,CAAC,eAAe;oBAEtD,IAAIA,KAAK,QAAQ,EAAE,mBACjB3B,KAAK,iBAAiB,GAAG2B,KAAK,QAAQ,CAAC,iBAAiB;gBAE5D;gBAEA,MAAME,qBAAqBjC,wBAAwB5B,SAC/C8D,qBAAqB9D,SACrB4C;gBAKJ,MAAMmB,kBAAkB/D,MAAM,UAAU,GACpC4C,SACAiB;gBACJ,MAAMG,YAAY,CAAC,CAACD;gBAGpB,IAAIE;gBACJ,IACE,CAACD,aACDhE,MAAM,KAAK,IACX,IAAI,CAAC,SAAS,CAAC,uBAAuB,EAEtC,IAAI;oBACFiE,gBAAgB,MAAM,IAAI,CAAC,SAAS,CAAC,uBAAuB,CAAC;wBAC3D,QAAQ;4BAACjE,MAAM,KAAK;yBAAC;oBACvB;gBACF,EAAE,OAAM,CAER;gBAGF,MAAMkE,mBAAmBD,gBACrBE,sBAEEC,qCACEH,eACAtB,2BAEF,AAAwB,YAAxB,OAAO3C,MAAM,MAAM,GACfA,MAAM,MAAM,GACZA,MAAM,MAAM,EAAE,UAAU,MAE9B4C;gBAEJ,MAAMyB,aAAa,CAAC,CAACH;gBAErB,MAAMI,cAActE,MAAM,MAAM;gBAChC,MAAMuE,oBAAoB,IAAI,CAAC,SAAS,EAAE,iBAAiBD;gBAC3D,MAAME,aAAaD,mBAAmB,cAAc;gBAEpD,MAAME,yBACJT,aAAaK,aACT,OACA,MAAMK,sBACJ;oBACE,WAAW,IAAI,CAAC,SAAS;oBACzB,mBAAmB,IAAI,CAAC,SAAS;gBACnC,GACAF,YACAF,aACAtE,MAAM,SAAS;gBAIvB,MAAM2E,mBAAmBF,yBACrBG,oCACEH,wBACA9B,4BAEFC;gBAEJ,MAAMiC,aAAa,CAAC,CAACF;gBAErB,IAAIG;gBACJ,MAAM5C,SAASD,YAAY,IAAI,CAAC,MAAM;gBACtC,IAAI,CAACoC,cAAc,CAACQ,cAAc,CAACb,WACjC,IAAI;oBACF5B,mBAAmBF,QAAQ;oBAC3BuB,eAAe,MAAM,IAAI,CAAC,OAAO,CAAC,MAAM,CACtCzD,OACA;wBACE,SAASmC;wBACT0B;oBACF,GACAvD,6BACA+C;oBAEFK,UAAUD,aAAa,IAAI;oBAC3BqB,sBAAsBrB,aAAa,OAAO;gBAC5C,EAAE,OAAOX,OAAO;oBACd,IAAIA,iBAAiBiC,cACnBrB,UAAUZ,MAAM,IAAI;oBAEtB,MAAMA;gBACR,SAAU;oBACRV,mBAAmBF,QAAQ;gBAC7B;gBAGF,MAAM8C,UACJjB,mBACAG,oBACAS,oBACAG;gBAGF,MAAMG,2BAA2BrF,iBAC/B2E,mBAAmB,cAAc;gBAGnC,IAAIW;gBAOJ,IACEF,WACA,IAAI,CAAC,SAAS,IACd,CAACH,cACA,EAACb,aAAa,CAACiB,wBAAuB,KACvCjF,OAAO,cAAc,OAErB,IAAI,IAAI,CAAC,SAAS,CAAC,oBAAoB,EACrC,IAAI;oBAGF,IAAImF,gBAAkCH,QAAQ,MAAM;oBACpD,IAAIrC,AAA6B,MAA7BA,0BAAgC;wBAClCwC,gBAAgB;4BACdC,KAAK,KAAK,CAACJ,QAAQ,MAAM,CAAC,EAAE,GAAGrC;4BAC/ByC,KAAK,KAAK,CAACJ,QAAQ,MAAM,CAAC,EAAE,GAAGrC;yBAChC;wBACDjD,MACE,8DACAsF,QAAQ,MAAM,EACdG;oBAEJ;oBAEA,MAAME,UAAU,MAAM,IAAI,CAAC,SAAS,CAAC,oBAAoB,CACvDF,eACA;wBACE,mBACE,AAAwB,YAAxB,OAAOnF,MAAM,MAAM,GACfA,MAAM,MAAM,GACZA,MAAM,MAAM,EAAE;wBACpB,aAAaM;oBACf;oBAEF,IAAIV,iBAAiByF,UAAU;wBAC7B3F,MACE,uCACA4E,aACAe;wBAEFH,oBAAoBG;wBACpB,IAAI,CAAC,SAAS,CAAC,yBAAyB,CACtC;4BACE,MAAM;4BACN,QAAQf;4BACR,OAAOe;wBACT,GACAd;oBAEJ,OACE7E,MACE,yDACA4E;gBAGN,EAAE,OAAOxB,OAAO;oBACdpD,MAAM,mCAAmCoD;gBAC3C;qBAEApD,MAAM;gBAIV,IAAI,CAACsF,SAAS;oBACZ,IAAIxB,YACF,MAAM,IAAIuB,aACR,CAAC,oBAAoB,EAAE/E,MAAM,MAAM,EAAE,EACrCwD;oBAGJ,MAAM,IAAIlC,MAAM,CAAC,mBAAmB,EAAEtB,MAAM,MAAM,EAAE;gBACtD;gBAEA,IAAIsF;gBAEJ,IAAItB,WACFsB,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACP,MAAMtF,MAAM,IAAI;oBAClB;gBACF;qBACK,IAAIqE,YACTiB,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACP,OAAOtF,MAAM,KAAK;oBACpB;gBACF;qBACK,IAAI6E,YACTS,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACPd;wBACA,aAAaU;oBACf;gBACF;gBAGF/B,WAAW6B;gBAEX,OAAO;oBACL,QAAQ;wBACN,SAAS;4BACP,GAAGA,OAAO;4BAEV,KAAK7C,UAAU,aAAa;wBAC9B;oBACF;oBACAmD;gBACF;YACF;QACF;QAEA,OAAO/B;IACT;IAxjBA,YAAY,EACVgC,iBAAiB,EACjBC,OAAO,EACPC,SAAS,EACTtE,WAAW,EACXuE,eAAe,EACC,CAAE;QAhBpB,uBAAiB,aAAjB;QAEA,uBAAiB,WAAjB;QAEA,uBAAiB,aAAjB;QAEA,uBAAiB,eAAjB;QAEA,uBAAiB,mBAAjB;QASE,IAAI,CAAC,SAAS,GAAGH;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC;QACjB,IAAI,CAAC,WAAW,GAAGtE;QACnB,IAAI,CAAC,eAAe,GAAGuE;IACzB;AA6iBF"}
1
+ {"version":3,"file":"agent/task-builder.mjs","sources":["../../../src/agent/task-builder.ts"],"sourcesContent":["import { findAllMidsceneLocatorField, parseActionParam } from '@/ai-model';\nimport type { ModelRuntime } from '@/ai-model/models';\nimport type { AbstractInterface } from '@/device';\nimport type Service from '@/service';\nimport { setTimingFieldOnce } from '@/task-timing';\nimport type {\n AIUsageInfo,\n DetailedLocateParam,\n DeviceAction,\n ElementCacheFeature,\n ExecutionTaskActionApply,\n ExecutionTaskApply,\n ExecutionTaskHitBy,\n ExecutionTaskPlanningLocateApply,\n LocateResultElement,\n LocateResultWithDump,\n PlanningAction,\n PlanningLocateParam,\n Rect,\n ServiceDump,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport { sleep } from '@/utils';\nimport { generateElementByRect } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TaskCache } from './task-cache';\nimport { withUsageIntent } from './usage-intent';\nimport {\n ifPlanLocateParamHasLocatedPixelBbox,\n matchElementFromCache,\n matchElementFromPlan,\n transformLogicalElementToScreenshot,\n transformLogicalRectToScreenshotRect,\n} from './utils';\n\nconst debug = getDebug('agent:task-builder');\n\n/**\n * Check if a cache object is non-empty\n */\nfunction hasNonEmptyCache(cache: unknown): boolean {\n return (\n cache !== null &&\n cache !== undefined &&\n typeof cache === 'object' &&\n Object.keys(cache).length > 0\n );\n}\n\nfunction invalidLocateElementReason(\n element: LocateResultElement,\n): string | undefined {\n const values = [\n element.center?.[0],\n element.center?.[1],\n element.rect?.left,\n element.rect?.top,\n element.rect?.width,\n element.rect?.height,\n ];\n if (\n values.some((value) => typeof value !== 'number' || !Number.isFinite(value))\n ) {\n return `Invalid locate result coordinates: ${JSON.stringify(element)}`;\n }\n if (element.rect.width <= 0 || element.rect.height <= 0) {\n return `Invalid locate result rect size: ${JSON.stringify(element)}`;\n }\n return undefined;\n}\n\nexport function locatePlanForLocate(param: string | DetailedLocateParam) {\n const locate = typeof param === 'string' ? { prompt: param } : param;\n const locatePlan: PlanningAction<PlanningLocateParam> = {\n type: 'Locate',\n param: locate,\n thought: '',\n };\n return locatePlan;\n}\n\ninterface TaskBuilderDeps {\n interfaceInstance: AbstractInterface;\n service: Service;\n taskCache?: TaskCache;\n actionSpace: DeviceAction[];\n waitAfterAction?: number;\n}\n\ninterface BuildOptions {\n cacheable?: boolean;\n deepLocate?: boolean;\n abortSignal?: AbortSignal;\n}\n\ninterface PlanBuildContext {\n tasks: ExecutionTaskApply[];\n planningModel: ModelRuntime;\n defaultModel: ModelRuntime;\n cacheable?: boolean;\n deepLocate?: boolean;\n abortSignal?: AbortSignal;\n}\n\nexport class TaskBuilder {\n private readonly interface: AbstractInterface;\n\n private readonly service: Service;\n\n private readonly taskCache?: TaskCache;\n\n private readonly actionSpace: DeviceAction[];\n\n private readonly waitAfterAction?: number;\n\n constructor({\n interfaceInstance,\n service,\n taskCache,\n actionSpace,\n waitAfterAction,\n }: TaskBuilderDeps) {\n this.interface = interfaceInstance;\n this.service = service;\n this.taskCache = taskCache;\n this.actionSpace = actionSpace;\n this.waitAfterAction = waitAfterAction;\n }\n\n public async build(\n plans: PlanningAction[],\n planningModel: ModelRuntime,\n defaultModel: ModelRuntime,\n options?: BuildOptions,\n ): Promise<{ tasks: ExecutionTaskApply[] }> {\n const tasks: ExecutionTaskApply[] = [];\n const cacheable = options?.cacheable;\n\n const context: PlanBuildContext = {\n tasks,\n planningModel,\n defaultModel,\n cacheable,\n deepLocate: options?.deepLocate,\n abortSignal: options?.abortSignal,\n };\n\n type PlanHandler = (plan: PlanningAction) => Promise<void> | void;\n\n const planHandlers = new Map<string, PlanHandler>([\n [\n 'Locate',\n (plan) =>\n this.handleLocatePlan(\n plan as PlanningAction<PlanningLocateParam>,\n context,\n ),\n ],\n ['Finished', (plan) => this.handleFinishedPlan(plan, context)],\n ]);\n\n const defaultHandler: PlanHandler = (plan) =>\n this.handleActionPlan(plan, context);\n\n for (const plan of plans) {\n const handler = planHandlers.get(plan.type) ?? defaultHandler;\n await handler(plan);\n }\n\n return {\n tasks,\n };\n }\n\n private handleFinishedPlan(\n plan: PlanningAction,\n context: PlanBuildContext,\n ): void {\n const taskActionFinished: ExecutionTaskActionApply<null> = {\n type: 'Action Space',\n subType: 'Finished',\n param: null,\n thought: plan.thought,\n executor: async () => {},\n };\n context.tasks.push(taskActionFinished);\n }\n\n private async handleLocatePlan(\n plan: PlanningAction<PlanningLocateParam>,\n context: PlanBuildContext,\n ): Promise<void> {\n const taskLocate = this.createLocateTask(plan, plan.param, context);\n context.tasks.push(taskLocate);\n }\n\n private async handleActionPlan(\n plan: PlanningAction,\n context: PlanBuildContext,\n ): Promise<void> {\n const planType = plan.type;\n const actionSpace = this.actionSpace;\n const action = actionSpace.find((item) => item.name === planType);\n const param = plan.param;\n\n if (!action) {\n throw new Error(`Action type '${planType}' not found`);\n }\n\n const locateFields = action\n ? findAllMidsceneLocatorField(action.paramSchema)\n : [];\n\n const requiredLocateFields = action\n ? findAllMidsceneLocatorField(action.paramSchema, true)\n : [];\n\n locateFields.forEach((field) => {\n if (param[field]) {\n // Always use createLocateTask for all locate params.\n // This ensures cache writing happens even when locatedPixelBbox is available\n const locatePlan = locatePlanForLocate(param[field]);\n debug(\n 'will prepend locate param for field',\n `action.type=${planType}`,\n `param=${JSON.stringify(param[field])}`,\n `locatePlan=${JSON.stringify(locatePlan)}`,\n `hasLocatedPixelBbox=${ifPlanLocateParamHasLocatedPixelBbox(param[field])}`,\n );\n const locateTask = this.createLocateTask(\n locatePlan,\n param[field],\n context,\n (result) => {\n param[field] = result;\n },\n );\n context.tasks.push(locateTask);\n } else {\n assert(\n !requiredLocateFields.includes(field),\n `Required locate field '${field}' is not provided for action ${planType}`,\n );\n debug(`field '${field}' is not provided for action ${planType}`);\n }\n });\n\n const task: ExecutionTaskApply<\n 'Action Space',\n any,\n { success: boolean; action: string; param: any },\n void\n > = {\n type: 'Action Space',\n subType: planType,\n thought: plan.thought,\n param: plan.param,\n executor: async (param, taskContext) => {\n const timing = taskContext.task.timing;\n\n debug(\n 'executing action',\n planType,\n param,\n `taskContext.element.center: ${taskContext.element?.center}`,\n );\n\n const uiContext = taskContext.uiContext;\n assert(uiContext, 'uiContext is required for Action task');\n\n requiredLocateFields.forEach((field) => {\n assert(\n param[field],\n `field '${field}' is required for action ${planType} but not provided. Cannot execute action ${planType}.`,\n );\n });\n\n setTimingFieldOnce(timing, 'beforeInvokeActionHookStart');\n const delayBeforeRunner = action.delayBeforeRunner ?? 200;\n try {\n await Promise.all([\n (async () => {\n if (this.interface.beforeInvokeAction) {\n debug(\n `will call \"beforeInvokeAction\" for interface with action name ${action.name}`,\n );\n await this.interface.beforeInvokeAction(action.name, param);\n debug(\n `called \"beforeInvokeAction\" for interface with action name ${action.name}`,\n );\n }\n })(),\n delayBeforeRunner > 0\n ? sleep(delayBeforeRunner)\n : Promise.resolve(),\n ]);\n } catch (originalError: any) {\n const originalMessage =\n originalError?.message || String(originalError);\n throw new Error(\n `error in running beforeInvokeAction for ${action.name}: ${originalMessage}`,\n { cause: originalError },\n );\n }\n setTimingFieldOnce(timing, 'beforeInvokeActionHookEnd');\n\n const { shrunkShotToLogicalRatio } = uiContext;\n if (shrunkShotToLogicalRatio === undefined) {\n throw new Error(\n 'shrunkShotToLogicalRatio is not defined in Action task',\n );\n }\n\n if (action.paramSchema) {\n try {\n param = parseActionParam(param, action.paramSchema, {\n shrunkShotToLogicalRatio,\n });\n } catch (error: any) {\n throw new Error(\n `Invalid parameters for action ${action.name}: ${error.message}\\nParameters: ${JSON.stringify(param)}`,\n { cause: error },\n );\n }\n }\n\n setTimingFieldOnce(timing, 'callActionStart');\n\n debug('calling action', action.name);\n const actionFn = action.call.bind(this.interface);\n const actionResult = await actionFn(param, taskContext);\n setTimingFieldOnce(timing, 'callActionEnd');\n debug('called action', action.name, 'result:', actionResult);\n\n setTimingFieldOnce(timing, 'afterInvokeActionHookStart');\n\n const delayAfterRunner =\n action.delayAfterRunner ?? this.waitAfterAction ?? 300;\n if (delayAfterRunner > 0) {\n await sleep(delayAfterRunner);\n }\n\n try {\n if (this.interface.afterInvokeAction) {\n debug(\n `will call \"afterInvokeAction\" for interface with action name ${action.name}`,\n );\n await this.interface.afterInvokeAction(action.name, param);\n debug(\n `called \"afterInvokeAction\" for interface with action name ${action.name}`,\n );\n }\n } catch (originalError: any) {\n const originalMessage =\n originalError?.message || String(originalError);\n throw new Error(\n `error in running afterInvokeAction for ${action.name}: ${originalMessage}`,\n { cause: originalError },\n );\n }\n\n setTimingFieldOnce(timing, 'afterInvokeActionHookEnd');\n\n return {\n output: actionResult,\n };\n },\n };\n\n context.tasks.push(task);\n }\n\n private createLocateTask(\n plan: PlanningAction<PlanningLocateParam>,\n detailedLocateParam: DetailedLocateParam | string,\n context: PlanBuildContext,\n onResult?: (result: LocateResultElement) => void,\n ): ExecutionTaskPlanningLocateApply {\n const { cacheable, defaultModel, deepLocate, abortSignal } = context;\n\n let locateParam = detailedLocateParam;\n\n if (typeof locateParam === 'string') {\n locateParam = {\n prompt: locateParam,\n };\n }\n\n if (cacheable !== undefined) {\n locateParam = {\n ...locateParam,\n cacheable,\n };\n }\n\n if (deepLocate && !locateParam.deepLocate) {\n locateParam = {\n ...locateParam,\n deepLocate: true,\n };\n }\n\n const taskLocator: ExecutionTaskPlanningLocateApply = {\n type: 'Planning',\n subType: 'Locate',\n param: locateParam,\n thought: plan.thought,\n executor: async (param, taskContext) => {\n const { task } = taskContext;\n let { uiContext } = taskContext;\n const paramWithLocatedPixelBbox = ifPlanLocateParamHasLocatedPixelBbox(\n param,\n )\n ? param\n : undefined;\n\n assert(\n param?.prompt || paramWithLocatedPixelBbox,\n `No prompt or id or position or locatedPixelBbox to locate, param=${JSON.stringify(\n param,\n )}`,\n );\n\n if (!uiContext) {\n uiContext = await this.service.contextRetrieverFn();\n }\n\n assert(uiContext, 'uiContext is required for Service task');\n\n const { shrunkShotToLogicalRatio } = uiContext;\n\n if (shrunkShotToLogicalRatio === undefined) {\n throw new Error(\n 'shrunkShotToLogicalRatio is not defined in locate task',\n );\n }\n\n let locateDump: ServiceDump | undefined;\n let locateResult: LocateResultWithDump | undefined;\n\n const applyDump = (dump?: ServiceDump) => {\n if (!dump) {\n return;\n }\n locateDump = dump;\n task.log = {\n dump,\n rawResponse: dump.taskInfo?.rawResponse,\n };\n task.usage = withUsageIntent(dump.taskInfo?.usage, 'default');\n if (dump.taskInfo?.searchAreaUsage) {\n task.searchAreaUsage = dump.taskInfo.searchAreaUsage;\n }\n if (dump.taskInfo?.reasoning_content) {\n task.reasoning_content = dump.taskInfo.reasoning_content;\n }\n };\n\n const planLocatedElement = paramWithLocatedPixelBbox\n ? matchElementFromPlan(paramWithLocatedPixelBbox)\n : undefined;\n\n // from locatedPixelBbox (direct plan hit)\n // when deepLocate is enabled, locatedPixelBbox should be used as search\n // area hint, not as a final direct hit\n const elementFromPlan = param.deepLocate\n ? undefined\n : planLocatedElement;\n const isPlanDirectHit = !!elementFromPlan;\n\n // from xpath\n let rectFromXpath: Rect | undefined;\n if (\n !isPlanDirectHit &&\n param.xpath &&\n this.interface.rectMatchesCacheFeature\n ) {\n try {\n rectFromXpath = await this.interface.rectMatchesCacheFeature({\n xpaths: [param.xpath],\n });\n } catch {\n // xpath locate failed, allow fallback to cache or AI locate\n }\n }\n\n const elementFromXpath = rectFromXpath\n ? generateElementByRect(\n // rectFromXpath is in logical coordinates, which should be transformed to screenshot coordinates;\n transformLogicalRectToScreenshotRect(\n rectFromXpath,\n shrunkShotToLogicalRatio,\n ),\n typeof param.prompt === 'string'\n ? param.prompt\n : param.prompt?.prompt || '',\n )\n : undefined;\n\n const isXpathHit = !!elementFromXpath;\n\n const cachePrompt = param.prompt;\n const locateCacheRecord = this.taskCache?.matchLocateCache(cachePrompt);\n const cacheEntry = locateCacheRecord?.cacheContent?.cache;\n\n const elementFromCacheResult =\n isPlanDirectHit || isXpathHit\n ? null\n : await matchElementFromCache(\n {\n taskCache: this.taskCache,\n interfaceInstance: this.interface,\n },\n cacheEntry,\n cachePrompt,\n param.cacheable,\n );\n\n // elementFromCacheResult is in logical coordinates, which should be transformed to screenshot coordinates;\n const elementFromCache = elementFromCacheResult\n ? transformLogicalElementToScreenshot(\n elementFromCacheResult,\n shrunkShotToLogicalRatio,\n )\n : undefined;\n\n const isCacheHit = !!elementFromCache;\n\n let elementFromAiLocate: LocateResultElement | null | undefined;\n const timing = taskContext.task.timing;\n if (!isXpathHit && !isCacheHit && !isPlanDirectHit) {\n try {\n setTimingFieldOnce(timing, 'callAiStart');\n locateResult = await this.service.locate(\n param,\n {\n context: uiContext,\n planLocatedElement,\n },\n defaultModel,\n abortSignal,\n );\n applyDump(locateResult.dump);\n elementFromAiLocate = locateResult.element;\n } catch (error) {\n if (error instanceof ServiceError) {\n applyDump(error.dump);\n }\n throw error;\n } finally {\n setTimingFieldOnce(timing, 'callAiEnd');\n }\n }\n\n const element =\n elementFromPlan ||\n elementFromXpath ||\n elementFromCache ||\n elementFromAiLocate;\n\n if (element) {\n const invalidElementReason = invalidLocateElementReason(element);\n if (invalidElementReason) {\n if (locateDump) {\n throw new ServiceError(invalidElementReason, locateDump);\n }\n throw new Error(invalidElementReason);\n }\n }\n\n // Check if locate cache already exists (for planHitFlag case)\n const locateCacheAlreadyExists = hasNonEmptyCache(\n locateCacheRecord?.cacheContent?.cache,\n );\n\n let currentCacheEntry: ElementCacheFeature | undefined;\n // Write cache if:\n // 1. element found\n // 2. taskCache enabled\n // 3. not a cache hit (otherwise we'd be writing what we just read)\n // 4. not already cached for plan hit case (avoid redundant writes), OR allow update if cache validation failed\n // 5. cacheable is not explicitly false\n if (\n element &&\n this.taskCache &&\n !isCacheHit &&\n (!isPlanDirectHit || !locateCacheAlreadyExists) &&\n param?.cacheable !== false\n ) {\n if (this.interface.cacheFeatureForPoint) {\n try {\n // Transform coordinates to logical space for cacheFeatureForPoint\n // cacheFeatureForPoint needs logical coordinates to locate elements in DOM\n let pointForCache: [number, number] = element.center;\n if (shrunkShotToLogicalRatio !== 1) {\n pointForCache = [\n Math.round(element.center[0] / shrunkShotToLogicalRatio),\n Math.round(element.center[1] / shrunkShotToLogicalRatio),\n ];\n debug(\n 'Transformed coordinates for cacheFeatureForPoint: %o -> %o',\n element.center,\n pointForCache,\n );\n }\n\n const feature = await this.interface.cacheFeatureForPoint(\n pointForCache,\n {\n targetDescription:\n typeof param.prompt === 'string'\n ? param.prompt\n : param.prompt?.prompt,\n modelRuntime: defaultModel,\n },\n );\n if (hasNonEmptyCache(feature)) {\n debug(\n 'update cache, prompt: %s, cache: %o',\n cachePrompt,\n feature,\n );\n currentCacheEntry = feature;\n this.taskCache.updateOrAppendCacheRecord(\n {\n type: 'locate',\n prompt: cachePrompt,\n cache: feature,\n },\n locateCacheRecord,\n );\n } else {\n debug(\n 'no cache data returned, skip cache update, prompt: %s',\n cachePrompt,\n );\n }\n } catch (error) {\n debug('cacheFeatureForPoint failed: %s', error);\n }\n } else {\n debug('cacheFeatureForPoint is not supported, skip cache update');\n }\n }\n\n if (!element) {\n if (locateDump) {\n throw new ServiceError(\n `Element not found : ${param.prompt}`,\n locateDump,\n );\n }\n throw new Error(`Element not found: ${param.prompt}`);\n }\n\n let hitBy: ExecutionTaskHitBy | undefined;\n\n if (isPlanDirectHit && paramWithLocatedPixelBbox) {\n hitBy = {\n from: 'Plan',\n context: {\n locatedPixelBbox: paramWithLocatedPixelBbox.locatedPixelBbox,\n },\n };\n } else if (isXpathHit) {\n hitBy = {\n from: 'User expected path',\n context: {\n xpath: param.xpath,\n },\n };\n } else if (isCacheHit) {\n hitBy = {\n from: 'Cache',\n context: {\n cacheEntry,\n cacheToSave: currentCacheEntry,\n },\n };\n }\n\n onResult?.(element);\n\n return {\n output: {\n element: {\n ...element,\n // backward compatibility for aiLocate, which return value needs a dpr field\n dpr: uiContext.deprecatedDpr,\n },\n },\n hitBy,\n };\n },\n };\n\n return taskLocator;\n }\n}\n"],"names":["debug","getDebug","hasNonEmptyCache","cache","Object","invalidLocateElementReason","element","values","value","Number","JSON","locatePlanForLocate","param","locate","locatePlan","TaskBuilder","plans","planningModel","defaultModel","options","tasks","cacheable","context","planHandlers","Map","plan","defaultHandler","handler","taskActionFinished","taskLocate","planType","actionSpace","action","item","Error","locateFields","findAllMidsceneLocatorField","requiredLocateFields","field","ifPlanLocateParamHasLocatedPixelBbox","locateTask","result","assert","task","taskContext","timing","uiContext","setTimingFieldOnce","delayBeforeRunner","Promise","sleep","originalError","originalMessage","String","shrunkShotToLogicalRatio","undefined","parseActionParam","error","actionFn","actionResult","delayAfterRunner","detailedLocateParam","onResult","deepLocate","abortSignal","locateParam","taskLocator","paramWithLocatedPixelBbox","locateDump","locateResult","applyDump","dump","withUsageIntent","planLocatedElement","matchElementFromPlan","elementFromPlan","isPlanDirectHit","rectFromXpath","elementFromXpath","generateElementByRect","transformLogicalRectToScreenshotRect","isXpathHit","cachePrompt","locateCacheRecord","cacheEntry","elementFromCacheResult","matchElementFromCache","elementFromCache","transformLogicalElementToScreenshot","isCacheHit","elementFromAiLocate","ServiceError","invalidElementReason","locateCacheAlreadyExists","currentCacheEntry","pointForCache","Math","feature","hitBy","interfaceInstance","service","taskCache","waitAfterAction"],"mappings":";;;;;;;;;;;;;;;;;;;AAoCA,MAAMA,QAAQC,SAAS;AAKvB,SAASC,iBAAiBC,KAAc;IACtC,OACEA,QAAAA,SAEA,AAAiB,YAAjB,OAAOA,SACPC,OAAO,IAAI,CAACD,OAAO,MAAM,GAAG;AAEhC;AAEA,SAASE,2BACPC,OAA4B;IAE5B,MAAMC,SAAS;QACbD,QAAQ,MAAM,EAAE,CAAC,EAAE;QACnBA,QAAQ,MAAM,EAAE,CAAC,EAAE;QACnBA,QAAQ,IAAI,EAAE;QACdA,QAAQ,IAAI,EAAE;QACdA,QAAQ,IAAI,EAAE;QACdA,QAAQ,IAAI,EAAE;KACf;IACD,IACEC,OAAO,IAAI,CAAC,CAACC,QAAU,AAAiB,YAAjB,OAAOA,SAAsB,CAACC,OAAO,QAAQ,CAACD,SAErE,OAAO,CAAC,mCAAmC,EAAEE,KAAK,SAAS,CAACJ,UAAU;IAExE,IAAIA,QAAQ,IAAI,CAAC,KAAK,IAAI,KAAKA,QAAQ,IAAI,CAAC,MAAM,IAAI,GACpD,OAAO,CAAC,iCAAiC,EAAEI,KAAK,SAAS,CAACJ,UAAU;AAGxE;AAEO,SAASK,oBAAoBC,KAAmC;IACrE,MAAMC,SAAS,AAAiB,YAAjB,OAAOD,QAAqB;QAAE,QAAQA;IAAM,IAAIA;IAC/D,MAAME,aAAkD;QACtD,MAAM;QACN,OAAOD;QACP,SAAS;IACX;IACA,OAAOC;AACT;AAyBO,MAAMC;IAyBX,MAAa,MACXC,KAAuB,EACvBC,aAA2B,EAC3BC,YAA0B,EAC1BC,OAAsB,EACoB;QAC1C,MAAMC,QAA8B,EAAE;QACtC,MAAMC,YAAYF,SAAS;QAE3B,MAAMG,UAA4B;YAChCF;YACAH;YACAC;YACAG;YACA,YAAYF,SAAS;YACrB,aAAaA,SAAS;QACxB;QAIA,MAAMI,eAAe,IAAIC,IAAyB;YAChD;gBACE;gBACA,CAACC,OACC,IAAI,CAAC,gBAAgB,CACnBA,MACAH;aAEL;YACD;gBAAC;gBAAY,CAACG,OAAS,IAAI,CAAC,kBAAkB,CAACA,MAAMH;aAAS;SAC/D;QAED,MAAMI,iBAA8B,CAACD,OACnC,IAAI,CAAC,gBAAgB,CAACA,MAAMH;QAE9B,KAAK,MAAMG,QAAQT,MAAO;YACxB,MAAMW,UAAUJ,aAAa,GAAG,CAACE,KAAK,IAAI,KAAKC;YAC/C,MAAMC,QAAQF;QAChB;QAEA,OAAO;YACLL;QACF;IACF;IAEQ,mBACNK,IAAoB,EACpBH,OAAyB,EACnB;QACN,MAAMM,qBAAqD;YACzD,MAAM;YACN,SAAS;YACT,OAAO;YACP,SAASH,KAAK,OAAO;YACrB,UAAU,WAAa;QACzB;QACAH,QAAQ,KAAK,CAAC,IAAI,CAACM;IACrB;IAEA,MAAc,iBACZH,IAAyC,EACzCH,OAAyB,EACV;QACf,MAAMO,aAAa,IAAI,CAAC,gBAAgB,CAACJ,MAAMA,KAAK,KAAK,EAAEH;QAC3DA,QAAQ,KAAK,CAAC,IAAI,CAACO;IACrB;IAEA,MAAc,iBACZJ,IAAoB,EACpBH,OAAyB,EACV;QACf,MAAMQ,WAAWL,KAAK,IAAI;QAC1B,MAAMM,cAAc,IAAI,CAAC,WAAW;QACpC,MAAMC,SAASD,YAAY,IAAI,CAAC,CAACE,OAASA,KAAK,IAAI,KAAKH;QACxD,MAAMlB,QAAQa,KAAK,KAAK;QAExB,IAAI,CAACO,QACH,MAAM,IAAIE,MAAM,CAAC,aAAa,EAAEJ,SAAS,WAAW,CAAC;QAGvD,MAAMK,eAAeH,SACjBI,4BAA4BJ,OAAO,WAAW,IAC9C,EAAE;QAEN,MAAMK,uBAAuBL,SACzBI,4BAA4BJ,OAAO,WAAW,EAAE,QAChD,EAAE;QAENG,aAAa,OAAO,CAAC,CAACG;YACpB,IAAI1B,KAAK,CAAC0B,MAAM,EAAE;gBAGhB,MAAMxB,aAAaH,oBAAoBC,KAAK,CAAC0B,MAAM;gBACnDtC,MACE,uCACA,CAAC,YAAY,EAAE8B,UAAU,EACzB,CAAC,MAAM,EAAEpB,KAAK,SAAS,CAACE,KAAK,CAAC0B,MAAM,GAAG,EACvC,CAAC,WAAW,EAAE5B,KAAK,SAAS,CAACI,aAAa,EAC1C,CAAC,oBAAoB,EAAEyB,qCAAqC3B,KAAK,CAAC0B,MAAM,GAAG;gBAE7E,MAAME,aAAa,IAAI,CAAC,gBAAgB,CACtC1B,YACAF,KAAK,CAAC0B,MAAM,EACZhB,SACA,CAACmB;oBACC7B,KAAK,CAAC0B,MAAM,GAAGG;gBACjB;gBAEFnB,QAAQ,KAAK,CAAC,IAAI,CAACkB;YACrB,OAAO;gBACLE,OACE,CAACL,qBAAqB,QAAQ,CAACC,QAC/B,CAAC,uBAAuB,EAAEA,MAAM,6BAA6B,EAAER,UAAU;gBAE3E9B,MAAM,CAAC,OAAO,EAAEsC,MAAM,6BAA6B,EAAER,UAAU;YACjE;QACF;QAEA,MAAMa,OAKF;YACF,MAAM;YACN,SAASb;YACT,SAASL,KAAK,OAAO;YACrB,OAAOA,KAAK,KAAK;YACjB,UAAU,OAAOb,OAAOgC;gBACtB,MAAMC,SAASD,YAAY,IAAI,CAAC,MAAM;gBAEtC5C,MACE,oBACA8B,UACAlB,OACA,CAAC,4BAA4B,EAAEgC,YAAY,OAAO,EAAE,QAAQ;gBAG9D,MAAME,YAAYF,YAAY,SAAS;gBACvCF,OAAOI,WAAW;gBAElBT,qBAAqB,OAAO,CAAC,CAACC;oBAC5BI,OACE9B,KAAK,CAAC0B,MAAM,EACZ,CAAC,OAAO,EAAEA,MAAM,yBAAyB,EAAER,SAAS,yCAAyC,EAAEA,SAAS,CAAC,CAAC;gBAE9G;gBAEAiB,mBAAmBF,QAAQ;gBAC3B,MAAMG,oBAAoBhB,OAAO,iBAAiB,IAAI;gBACtD,IAAI;oBACF,MAAMiB,QAAQ,GAAG,CAAC;wBACf;4BACC,IAAI,IAAI,CAAC,SAAS,CAAC,kBAAkB,EAAE;gCACrCjD,MACE,CAAC,8DAA8D,EAAEgC,OAAO,IAAI,EAAE;gCAEhF,MAAM,IAAI,CAAC,SAAS,CAAC,kBAAkB,CAACA,OAAO,IAAI,EAAEpB;gCACrDZ,MACE,CAAC,2DAA2D,EAAEgC,OAAO,IAAI,EAAE;4BAE/E;wBACF;wBACAgB,oBAAoB,IAChBE,MAAMF,qBACNC,QAAQ,OAAO;qBACpB;gBACH,EAAE,OAAOE,eAAoB;oBAC3B,MAAMC,kBACJD,eAAe,WAAWE,OAAOF;oBACnC,MAAM,IAAIjB,MACR,CAAC,wCAAwC,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEoB,iBAAiB,EAC5E;wBAAE,OAAOD;oBAAc;gBAE3B;gBACAJ,mBAAmBF,QAAQ;gBAE3B,MAAM,EAAES,wBAAwB,EAAE,GAAGR;gBACrC,IAAIQ,AAA6BC,WAA7BD,0BACF,MAAM,IAAIpB,MACR;gBAIJ,IAAIF,OAAO,WAAW,EACpB,IAAI;oBACFpB,QAAQ4C,iBAAiB5C,OAAOoB,OAAO,WAAW,EAAE;wBAClDsB;oBACF;gBACF,EAAE,OAAOG,OAAY;oBACnB,MAAM,IAAIvB,MACR,CAAC,8BAA8B,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEyB,MAAM,OAAO,CAAC,cAAc,EAAE/C,KAAK,SAAS,CAACE,QAAQ,EACtG;wBAAE,OAAO6C;oBAAM;gBAEnB;gBAGFV,mBAAmBF,QAAQ;gBAE3B7C,MAAM,kBAAkBgC,OAAO,IAAI;gBACnC,MAAM0B,WAAW1B,OAAO,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS;gBAChD,MAAM2B,eAAe,MAAMD,SAAS9C,OAAOgC;gBAC3CG,mBAAmBF,QAAQ;gBAC3B7C,MAAM,iBAAiBgC,OAAO,IAAI,EAAE,WAAW2B;gBAE/CZ,mBAAmBF,QAAQ;gBAE3B,MAAMe,mBACJ5B,OAAO,gBAAgB,IAAI,IAAI,CAAC,eAAe,IAAI;gBACrD,IAAI4B,mBAAmB,GACrB,MAAMV,MAAMU;gBAGd,IAAI;oBACF,IAAI,IAAI,CAAC,SAAS,CAAC,iBAAiB,EAAE;wBACpC5D,MACE,CAAC,6DAA6D,EAAEgC,OAAO,IAAI,EAAE;wBAE/E,MAAM,IAAI,CAAC,SAAS,CAAC,iBAAiB,CAACA,OAAO,IAAI,EAAEpB;wBACpDZ,MACE,CAAC,0DAA0D,EAAEgC,OAAO,IAAI,EAAE;oBAE9E;gBACF,EAAE,OAAOmB,eAAoB;oBAC3B,MAAMC,kBACJD,eAAe,WAAWE,OAAOF;oBACnC,MAAM,IAAIjB,MACR,CAAC,uCAAuC,EAAEF,OAAO,IAAI,CAAC,EAAE,EAAEoB,iBAAiB,EAC3E;wBAAE,OAAOD;oBAAc;gBAE3B;gBAEAJ,mBAAmBF,QAAQ;gBAE3B,OAAO;oBACL,QAAQc;gBACV;YACF;QACF;QAEArC,QAAQ,KAAK,CAAC,IAAI,CAACqB;IACrB;IAEQ,iBACNlB,IAAyC,EACzCoC,mBAAiD,EACjDvC,OAAyB,EACzBwC,QAAgD,EACd;QAClC,MAAM,EAAEzC,SAAS,EAAEH,YAAY,EAAE6C,UAAU,EAAEC,WAAW,EAAE,GAAG1C;QAE7D,IAAI2C,cAAcJ;QAElB,IAAI,AAAuB,YAAvB,OAAOI,aACTA,cAAc;YACZ,QAAQA;QACV;QAGF,IAAI5C,AAAckC,WAAdlC,WACF4C,cAAc;YACZ,GAAGA,WAAW;YACd5C;QACF;QAGF,IAAI0C,cAAc,CAACE,YAAY,UAAU,EACvCA,cAAc;YACZ,GAAGA,WAAW;YACd,YAAY;QACd;QAGF,MAAMC,cAAgD;YACpD,MAAM;YACN,SAAS;YACT,OAAOD;YACP,SAASxC,KAAK,OAAO;YACrB,UAAU,OAAOb,OAAOgC;gBACtB,MAAM,EAAED,IAAI,EAAE,GAAGC;gBACjB,IAAI,EAAEE,SAAS,EAAE,GAAGF;gBACpB,MAAMuB,4BAA4B5B,qCAChC3B,SAEEA,QACA2C;gBAEJb,OACE9B,OAAO,UAAUuD,2BACjB,CAAC,iEAAiE,EAAEzD,KAAK,SAAS,CAChFE,QACC;gBAGL,IAAI,CAACkC,WACHA,YAAY,MAAM,IAAI,CAAC,OAAO,CAAC,kBAAkB;gBAGnDJ,OAAOI,WAAW;gBAElB,MAAM,EAAEQ,wBAAwB,EAAE,GAAGR;gBAErC,IAAIQ,AAA6BC,WAA7BD,0BACF,MAAM,IAAIpB,MACR;gBAIJ,IAAIkC;gBACJ,IAAIC;gBAEJ,MAAMC,YAAY,CAACC;oBACjB,IAAI,CAACA,MACH;oBAEFH,aAAaG;oBACb5B,KAAK,GAAG,GAAG;wBACT4B;wBACA,aAAaA,KAAK,QAAQ,EAAE;oBAC9B;oBACA5B,KAAK,KAAK,GAAG6B,gBAAgBD,KAAK,QAAQ,EAAE,OAAO;oBACnD,IAAIA,KAAK,QAAQ,EAAE,iBACjB5B,KAAK,eAAe,GAAG4B,KAAK,QAAQ,CAAC,eAAe;oBAEtD,IAAIA,KAAK,QAAQ,EAAE,mBACjB5B,KAAK,iBAAiB,GAAG4B,KAAK,QAAQ,CAAC,iBAAiB;gBAE5D;gBAEA,MAAME,qBAAqBN,4BACvBO,qBAAqBP,6BACrBZ;gBAKJ,MAAMoB,kBAAkB/D,MAAM,UAAU,GACpC2C,SACAkB;gBACJ,MAAMG,kBAAkB,CAAC,CAACD;gBAG1B,IAAIE;gBACJ,IACE,CAACD,mBACDhE,MAAM,KAAK,IACX,IAAI,CAAC,SAAS,CAAC,uBAAuB,EAEtC,IAAI;oBACFiE,gBAAgB,MAAM,IAAI,CAAC,SAAS,CAAC,uBAAuB,CAAC;wBAC3D,QAAQ;4BAACjE,MAAM,KAAK;yBAAC;oBACvB;gBACF,EAAE,OAAM,CAER;gBAGF,MAAMkE,mBAAmBD,gBACrBE,sBAEEC,qCACEH,eACAvB,2BAEF,AAAwB,YAAxB,OAAO1C,MAAM,MAAM,GACfA,MAAM,MAAM,GACZA,MAAM,MAAM,EAAE,UAAU,MAE9B2C;gBAEJ,MAAM0B,aAAa,CAAC,CAACH;gBAErB,MAAMI,cAActE,MAAM,MAAM;gBAChC,MAAMuE,oBAAoB,IAAI,CAAC,SAAS,EAAE,iBAAiBD;gBAC3D,MAAME,aAAaD,mBAAmB,cAAc;gBAEpD,MAAME,yBACJT,mBAAmBK,aACf,OACA,MAAMK,sBACJ;oBACE,WAAW,IAAI,CAAC,SAAS;oBACzB,mBAAmB,IAAI,CAAC,SAAS;gBACnC,GACAF,YACAF,aACAtE,MAAM,SAAS;gBAIvB,MAAM2E,mBAAmBF,yBACrBG,oCACEH,wBACA/B,4BAEFC;gBAEJ,MAAMkC,aAAa,CAAC,CAACF;gBAErB,IAAIG;gBACJ,MAAM7C,SAASD,YAAY,IAAI,CAAC,MAAM;gBACtC,IAAI,CAACqC,cAAc,CAACQ,cAAc,CAACb,iBACjC,IAAI;oBACF7B,mBAAmBF,QAAQ;oBAC3BwB,eAAe,MAAM,IAAI,CAAC,OAAO,CAAC,MAAM,CACtCzD,OACA;wBACE,SAASkC;wBACT2B;oBACF,GACAvD,cACA8C;oBAEFM,UAAUD,aAAa,IAAI;oBAC3BqB,sBAAsBrB,aAAa,OAAO;gBAC5C,EAAE,OAAOZ,OAAO;oBACd,IAAIA,iBAAiBkC,cACnBrB,UAAUb,MAAM,IAAI;oBAEtB,MAAMA;gBACR,SAAU;oBACRV,mBAAmBF,QAAQ;gBAC7B;gBAGF,MAAMvC,UACJqE,mBACAG,oBACAS,oBACAG;gBAEF,IAAIpF,SAAS;oBACX,MAAMsF,uBAAuBvF,2BAA2BC;oBACxD,IAAIsF,sBAAsB;wBACxB,IAAIxB,YACF,MAAM,IAAIuB,aAAaC,sBAAsBxB;wBAE/C,MAAM,IAAIlC,MAAM0D;oBAClB;gBACF;gBAGA,MAAMC,2BAA2B3F,iBAC/BiF,mBAAmB,cAAc;gBAGnC,IAAIW;gBAOJ,IACExF,WACA,IAAI,CAAC,SAAS,IACd,CAACmF,cACA,EAACb,mBAAmB,CAACiB,wBAAuB,KAC7CjF,OAAO,cAAc,OAErB,IAAI,IAAI,CAAC,SAAS,CAAC,oBAAoB,EACrC,IAAI;oBAGF,IAAImF,gBAAkCzF,QAAQ,MAAM;oBACpD,IAAIgD,AAA6B,MAA7BA,0BAAgC;wBAClCyC,gBAAgB;4BACdC,KAAK,KAAK,CAAC1F,QAAQ,MAAM,CAAC,EAAE,GAAGgD;4BAC/B0C,KAAK,KAAK,CAAC1F,QAAQ,MAAM,CAAC,EAAE,GAAGgD;yBAChC;wBACDtD,MACE,8DACAM,QAAQ,MAAM,EACdyF;oBAEJ;oBAEA,MAAME,UAAU,MAAM,IAAI,CAAC,SAAS,CAAC,oBAAoB,CACvDF,eACA;wBACE,mBACE,AAAwB,YAAxB,OAAOnF,MAAM,MAAM,GACfA,MAAM,MAAM,GACZA,MAAM,MAAM,EAAE;wBACpB,cAAcM;oBAChB;oBAEF,IAAIhB,iBAAiB+F,UAAU;wBAC7BjG,MACE,uCACAkF,aACAe;wBAEFH,oBAAoBG;wBACpB,IAAI,CAAC,SAAS,CAAC,yBAAyB,CACtC;4BACE,MAAM;4BACN,QAAQf;4BACR,OAAOe;wBACT,GACAd;oBAEJ,OACEnF,MACE,yDACAkF;gBAGN,EAAE,OAAOzB,OAAO;oBACdzD,MAAM,mCAAmCyD;gBAC3C;qBAEAzD,MAAM;gBAIV,IAAI,CAACM,SAAS;oBACZ,IAAI8D,YACF,MAAM,IAAIuB,aACR,CAAC,oBAAoB,EAAE/E,MAAM,MAAM,EAAE,EACrCwD;oBAGJ,MAAM,IAAIlC,MAAM,CAAC,mBAAmB,EAAEtB,MAAM,MAAM,EAAE;gBACtD;gBAEA,IAAIsF;gBAEJ,IAAItB,mBAAmBT,2BACrB+B,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACP,kBAAkB/B,0BAA0B,gBAAgB;oBAC9D;gBACF;qBACK,IAAIc,YACTiB,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACP,OAAOtF,MAAM,KAAK;oBACpB;gBACF;qBACK,IAAI6E,YACTS,QAAQ;oBACN,MAAM;oBACN,SAAS;wBACPd;wBACA,aAAaU;oBACf;gBACF;gBAGFhC,WAAWxD;gBAEX,OAAO;oBACL,QAAQ;wBACN,SAAS;4BACP,GAAGA,OAAO;4BAEV,KAAKwC,UAAU,aAAa;wBAC9B;oBACF;oBACAoD;gBACF;YACF;QACF;QAEA,OAAOhC;IACT;IAtkBA,YAAY,EACViC,iBAAiB,EACjBC,OAAO,EACPC,SAAS,EACTtE,WAAW,EACXuE,eAAe,EACC,CAAE;QAhBpB,uBAAiB,aAAjB;QAEA,uBAAiB,WAAjB;QAEA,uBAAiB,aAAjB;QAEA,uBAAiB,eAAjB;QAEA,uBAAiB,mBAAjB;QASE,IAAI,CAAC,SAAS,GAAGH;QACjB,IAAI,CAAC,OAAO,GAAGC;QACf,IAAI,CAAC,SAAS,GAAGC;QACjB,IAAI,CAAC,WAAW,GAAGtE;QACnB,IAAI,CAAC,eAAe,GAAGuE;IACzB;AA2jBF"}
@@ -1,6 +1,6 @@
1
- import { AIResponseParseError, ConversationHistory, autoGLMPlanning, plan, uiTarsPlanning } from "../ai-model/index.mjs";
2
- import { isAutoGLM, isUITars } from "../ai-model/auto-glm/util.mjs";
1
+ import { AIResponseParseError, ConversationHistory } from "../ai-model/index.mjs";
3
2
  import { buildTypeQueryDemandValue } from "../ai-model/prompt/extraction.mjs";
3
+ import { genericXmlPlan } from "../ai-model/workflows/planning/index.mjs";
4
4
  import { getReadableTimeString } from "../common.mjs";
5
5
  import { TaskExecutionError } from "../task-runner.mjs";
6
6
  import { ServiceError } from "../types.mjs";
@@ -51,8 +51,8 @@ class TaskExecutor {
51
51
  else warnLog('useDeviceTime is enabled but getDeviceLocalTimeString is not implemented, falling back to runtime time.');
52
52
  return getReadableTimeString(format);
53
53
  }
54
- async convertPlanToExecutable(plans, modelConfigForPlanning, modelConfigForDefaultIntent, options) {
55
- return this.taskBuilder.build(plans, modelConfigForPlanning, modelConfigForDefaultIntent, options);
54
+ async convertPlanToExecutable(plans, planningModel, defaultModel, options) {
55
+ return this.taskBuilder.build(plans, planningModel, defaultModel, options);
56
56
  }
57
57
  async loadYamlFlowAsPlanning(userInstruction, yamlString) {
58
58
  const session = this.createExecutionSession(taskTitleStr('Act', userInstruction));
@@ -90,9 +90,9 @@ class TaskExecutor {
90
90
  runner
91
91
  };
92
92
  }
93
- async runPlans(title, plans, modelConfigForPlanning, modelConfigForDefaultIntent) {
93
+ async runPlans(title, plans, planningModel, defaultModel) {
94
94
  const session = this.createExecutionSession(title);
95
- const { tasks } = await this.convertPlanToExecutable(plans, modelConfigForPlanning, modelConfigForDefaultIntent);
95
+ const { tasks } = await this.convertPlanToExecutable(plans, planningModel, defaultModel);
96
96
  const runner = session.getRunner();
97
97
  const result = await session.appendAndRun(tasks);
98
98
  const { output } = result ?? {};
@@ -101,10 +101,14 @@ class TaskExecutor {
101
101
  runner
102
102
  };
103
103
  }
104
- async action(userPrompt, modelConfigForPlanning, modelConfigForDefaultIntent, includeBboxInPlanning, aiActContext, cacheable, replanningCycleLimitOverride, imagesIncludeCount, deepThink, fileChooserAccept, deepLocate, abortSignal) {
105
- return withFileChooser(this.interface, fileChooserAccept, async ()=>this.runAction(userPrompt, modelConfigForPlanning, modelConfigForDefaultIntent, includeBboxInPlanning, aiActContext, cacheable, replanningCycleLimitOverride, imagesIncludeCount, deepThink, deepLocate, abortSignal));
104
+ async action(userPrompt, planningModel, defaultModel, includeLocateInPlanning, aiActContext, cacheable, replanningCycleLimitOverride, imagesIncludeCount, deepThink, fileChooserAccept, deepLocate, abortSignal) {
105
+ return withFileChooser(this.interface, fileChooserAccept, async ()=>this.runAction(userPrompt, planningModel, defaultModel, includeLocateInPlanning, aiActContext, cacheable, replanningCycleLimitOverride, imagesIncludeCount, deepThink, deepLocate, abortSignal));
106
106
  }
107
- async runAction(userPrompt, modelConfigForPlanning, modelConfigForDefaultIntent, includeBboxInPlanning, aiActContext, cacheable, replanningCycleLimitOverride, imagesIncludeCount, deepThink, deepLocate, abortSignal) {
107
+ async runAction(userPrompt, planningModel, defaultModel, includeLocateInPlanning, aiActContext, cacheable, replanningCycleLimitOverride, imagesIncludeCount, deepThink, deepLocate, abortSignal) {
108
+ if (deepLocate && !planningModel.adapter.planning.supportsActionDeepLocate) {
109
+ warnLog(`The "deepLocate" option is not supported for aiAct with the current planning adapter (modelFamily: ${planningModel.config.modelFamily ?? 'unknown'}). It will be ignored.`);
110
+ deepLocate = false;
111
+ }
108
112
  const conversationHistory = new ConversationHistory();
109
113
  const session = this.createExecutionSession(taskTitleStr('Act', userPrompt));
110
114
  const runner = session.getRunner();
@@ -136,24 +140,22 @@ class TaskExecutor {
136
140
  executor: async (param, executorContext)=>{
137
141
  const { uiContext } = executorContext;
138
142
  assert(uiContext, 'uiContext is required for Planning task');
139
- const { modelFamily } = modelConfigForPlanning;
140
143
  const timing = executorContext.task.timing;
141
144
  const actionSpace = this.getActionSpace();
142
145
  debug('actionSpace for this interface is:', actionSpace.map((action)=>action.name).join(', '));
143
146
  assert(Array.isArray(actionSpace), 'actionSpace must be an array');
144
147
  if (0 === actionSpace.length) console.warn(`ActionSpace for ${this.interface.interfaceType} is empty. This may lead to unexpected behavior.`);
145
- const planImpl = isUITars(modelFamily) ? uiTarsPlanning : isAutoGLM(modelFamily) ? autoGLMPlanning : plan;
148
+ const planImpl = 'custom' === planningModel.adapter.planning.kind ? planningModel.adapter.planning.planFn : genericXmlPlan;
146
149
  let planResult;
147
150
  try {
148
151
  setTimingFieldOnce(timing, 'callAiStart');
149
152
  planResult = await planImpl(param.userInstruction, {
150
153
  context: uiContext,
151
154
  actionContext: param.aiActContext,
152
- interfaceType: this.interface.interfaceType,
153
155
  actionSpace,
154
- modelConfig: modelConfigForPlanning,
156
+ modelRuntime: planningModel,
155
157
  conversationHistory,
156
- includeBbox: includeBboxInPlanning,
158
+ includeLocateInPlanning,
157
159
  imagesIncludeCount,
158
160
  deepThink,
159
161
  abortSignal
@@ -207,7 +209,7 @@ class TaskExecutor {
207
209
  yamlFlow.push(...planResult?.yamlFlow || []);
208
210
  let executables;
209
211
  try {
210
- executables = await this.convertPlanToExecutable(plans, modelConfigForPlanning, modelConfigForDefaultIntent, {
212
+ executables = await this.convertPlanToExecutable(plans, planningModel, defaultModel, {
211
213
  cacheable,
212
214
  deepLocate,
213
215
  abortSignal
@@ -247,7 +249,7 @@ class TaskExecutor {
247
249
  runner
248
250
  };
249
251
  }
250
- createTypeQueryTask(type, demand, modelConfig, opt, multimodalPrompt) {
252
+ createTypeQueryTask(type, demand, modelRuntime, opt, multimodalPrompt) {
251
253
  const queryTask = {
252
254
  type: 'Insight',
253
255
  subType: type,
@@ -294,7 +296,7 @@ class TaskExecutor {
294
296
  extraPageDescription = await descriptionOfTree(tree, 200, false, opt?.domIncluded === 'visible-only');
295
297
  }
296
298
  try {
297
- extractResult = await this.service.extract(demandInput, modelConfig, opt, extraPageDescription, multimodalPrompt, uiContext);
299
+ extractResult = await this.service.extract(demandInput, modelRuntime, opt, extraPageDescription, multimodalPrompt, uiContext);
298
300
  } catch (error) {
299
301
  if (error instanceof ServiceError) applyDump(error.dump);
300
302
  throw error;
@@ -321,9 +323,9 @@ class TaskExecutor {
321
323
  };
322
324
  return queryTask;
323
325
  }
324
- async createTypeQueryExecution(type, demand, modelConfig, opt, multimodalPrompt) {
326
+ async createTypeQueryExecution(type, demand, modelRuntime, opt, multimodalPrompt) {
325
327
  const session = this.createExecutionSession(taskTitleStr(type, 'string' == typeof demand ? demand : JSON.stringify(demand)));
326
- const queryTask = await this.createTypeQueryTask(type, demand, modelConfig, opt, multimodalPrompt);
328
+ const queryTask = await this.createTypeQueryTask(type, demand, modelRuntime, opt, multimodalPrompt);
327
329
  const runner = session.getRunner();
328
330
  const result = await session.appendAndRun(queryTask);
329
331
  if (!result) throw new Error('result of taskExecutor.flush() is undefined in function createTypeQueryTask');
@@ -334,7 +336,7 @@ class TaskExecutor {
334
336
  runner
335
337
  };
336
338
  }
337
- async waitFor(assertion, opt, modelConfig) {
339
+ async waitFor(assertion, opt, modelRuntime) {
338
340
  const { textPrompt, multimodalPrompt } = parsePrompt(assertion);
339
341
  const description = `waitFor: ${textPrompt}`;
340
342
  const session = this.createExecutionSession(taskTitleStr('WaitFor', description));
@@ -355,7 +357,7 @@ class TaskExecutor {
355
357
  while(lastCheckStart - overallStartTime <= timeoutMs){
356
358
  const currentCheckStart = Date.now();
357
359
  lastCheckStart = currentCheckStart;
358
- const queryTask = await this.createTypeQueryTask('WaitFor', textPrompt, modelConfig, serviceExtractOpt, multimodalPrompt);
360
+ const queryTask = await this.createTypeQueryTask('WaitFor', textPrompt, modelRuntime, serviceExtractOpt, multimodalPrompt);
359
361
  const result = await session.appendAndRun(queryTask);
360
362
  if (result?.output) return {
361
363
  output: void 0,
@@ -375,7 +377,7 @@ class TaskExecutor {
375
377
  },
376
378
  thought
377
379
  }
378
- ], modelConfig, modelConfig);
380
+ ], modelRuntime, modelRuntime);
379
381
  if (sleepTasks[0]) await session.appendAndRun(sleepTasks[0]);
380
382
  }
381
383
  }