@midscene/core 1.9.8 → 1.10.0

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 (256) hide show
  1. package/dist/es/agent/agent.mjs +26 -94
  2. package/dist/es/agent/agent.mjs.map +1 -1
  3. package/dist/es/agent/prompt-context.mjs +14 -0
  4. package/dist/es/agent/prompt-context.mjs.map +1 -0
  5. package/dist/es/agent/run-gherkin-scenario.mjs +114 -0
  6. package/dist/es/agent/run-gherkin-scenario.mjs.map +1 -0
  7. package/dist/es/agent/task-builder.mjs +4 -4
  8. package/dist/es/agent/task-builder.mjs.map +1 -1
  9. package/dist/es/agent/tasks.mjs +4 -8
  10. package/dist/es/agent/tasks.mjs.map +1 -1
  11. package/dist/es/agent/utils.mjs +15 -3
  12. package/dist/es/agent/utils.mjs.map +1 -1
  13. package/dist/es/ai-model/index.mjs +1 -2
  14. package/dist/es/ai-model/inspect.mjs +77 -50
  15. package/dist/es/ai-model/inspect.mjs.map +1 -1
  16. package/dist/es/ai-model/llm-planning.mjs +10 -26
  17. package/dist/es/ai-model/llm-planning.mjs.map +1 -1
  18. package/dist/es/ai-model/model-adapter/chat-completion.mjs +46 -0
  19. package/dist/es/ai-model/model-adapter/chat-completion.mjs.map +1 -0
  20. package/dist/es/ai-model/model-adapter/index.mjs +2 -0
  21. package/dist/es/ai-model/model-adapter/locate.mjs +33 -0
  22. package/dist/es/ai-model/model-adapter/locate.mjs.map +1 -0
  23. package/dist/es/ai-model/model-adapter/planning.mjs +44 -0
  24. package/dist/es/ai-model/model-adapter/planning.mjs.map +1 -0
  25. package/dist/es/ai-model/model-adapter/resolve.mjs +43 -0
  26. package/dist/es/ai-model/model-adapter/resolve.mjs.map +1 -0
  27. package/dist/es/ai-model/model-adapter/types.mjs +0 -0
  28. package/dist/es/ai-model/models/auto-glm/actions.mjs +19 -31
  29. package/dist/es/ai-model/models/auto-glm/actions.mjs.map +1 -1
  30. package/dist/es/ai-model/models/auto-glm/adapter.mjs +8 -16
  31. package/dist/es/ai-model/models/auto-glm/adapter.mjs.map +1 -1
  32. package/dist/es/ai-model/models/auto-glm/locate.mjs +6 -108
  33. package/dist/es/ai-model/models/auto-glm/locate.mjs.map +1 -1
  34. package/dist/es/ai-model/models/auto-glm/parser.mjs +19 -41
  35. package/dist/es/ai-model/models/auto-glm/parser.mjs.map +1 -1
  36. package/dist/es/ai-model/models/auto-glm/planning.mjs +27 -70
  37. package/dist/es/ai-model/models/auto-glm/planning.mjs.map +1 -1
  38. package/dist/es/ai-model/models/default.mjs.map +1 -1
  39. package/dist/es/ai-model/models/doubao.mjs.map +1 -1
  40. package/dist/es/ai-model/models/gemini.mjs +1 -1
  41. package/dist/es/ai-model/models/gemini.mjs.map +1 -1
  42. package/dist/es/ai-model/models/glm.mjs.map +1 -1
  43. package/dist/es/ai-model/models/gpt.mjs.map +1 -1
  44. package/dist/es/ai-model/models/kimi.mjs.map +1 -1
  45. package/dist/es/ai-model/models/mimo.mjs.map +1 -1
  46. package/dist/es/ai-model/models/qwen.mjs.map +1 -1
  47. package/dist/es/ai-model/models/registry.mjs +1 -1
  48. package/dist/es/ai-model/models/registry.mjs.map +1 -1
  49. package/dist/es/ai-model/models/ui-tars/actions.mjs +147 -0
  50. package/dist/es/ai-model/models/ui-tars/actions.mjs.map +1 -0
  51. package/dist/es/ai-model/models/ui-tars/adapter.mjs +2 -2
  52. package/dist/es/ai-model/models/ui-tars/adapter.mjs.map +1 -1
  53. package/dist/es/ai-model/models/ui-tars/parser.mjs +40 -0
  54. package/dist/es/ai-model/models/ui-tars/parser.mjs.map +1 -0
  55. package/dist/es/ai-model/models/ui-tars/planning.mjs +19 -230
  56. package/dist/es/ai-model/models/ui-tars/planning.mjs.map +1 -1
  57. package/dist/es/ai-model/{prompt/ui-tars-planning.mjs → models/ui-tars/prompt.mjs} +1 -1
  58. package/dist/es/ai-model/models/ui-tars/prompt.mjs.map +1 -0
  59. package/dist/es/ai-model/service-caller/index.mjs +1 -3
  60. package/dist/es/ai-model/service-caller/index.mjs.map +1 -1
  61. package/dist/es/ai-model/shared/model-locate-result/coordinate-distance.mjs +10 -0
  62. package/dist/es/ai-model/shared/model-locate-result/coordinate-distance.mjs.map +1 -0
  63. package/dist/es/ai-model/shared/model-locate-result/factory.mjs +1 -3
  64. package/dist/es/ai-model/shared/model-locate-result/factory.mjs.map +1 -1
  65. package/dist/es/ai-model/shared/model-locate-result/index.mjs +3 -2
  66. package/dist/es/ai-model/shared/planning-action.mjs +6 -0
  67. package/dist/es/ai-model/shared/planning-action.mjs.map +1 -0
  68. package/dist/es/ai-model/workflows/inspect/planning-action-locate.mjs +95 -0
  69. package/dist/es/ai-model/workflows/inspect/planning-action-locate.mjs.map +1 -0
  70. package/dist/es/ai-model/workflows/planning/custom-planning.mjs +126 -0
  71. package/dist/es/ai-model/workflows/planning/custom-planning.mjs.map +1 -0
  72. package/dist/es/ai-model/workflows/planning/locate-normalization.mjs +31 -0
  73. package/dist/es/ai-model/workflows/planning/locate-normalization.mjs.map +1 -0
  74. package/dist/es/common.mjs +10 -6
  75. package/dist/es/common.mjs.map +1 -1
  76. package/dist/es/element-describer.mjs +180 -0
  77. package/dist/es/element-describer.mjs.map +1 -0
  78. package/dist/es/index.mjs +2 -1
  79. package/dist/es/index.mjs.map +1 -1
  80. package/dist/es/report-generator.mjs +3 -5
  81. package/dist/es/report-generator.mjs.map +1 -1
  82. package/dist/es/service/index.mjs +3 -2
  83. package/dist/es/service/index.mjs.map +1 -1
  84. package/dist/es/skill/index.mjs +1 -1
  85. package/dist/es/skill/index.mjs.map +1 -1
  86. package/dist/es/types.mjs.map +1 -1
  87. package/dist/es/utils.mjs +2 -2
  88. package/dist/es/yaml/player.mjs +7 -0
  89. package/dist/es/yaml/player.mjs.map +1 -1
  90. package/dist/lib/agent/agent.js +25 -93
  91. package/dist/lib/agent/agent.js.map +1 -1
  92. package/dist/lib/{ai-model/models/chat-content.js → agent/prompt-context.js} +13 -13
  93. package/dist/lib/agent/prompt-context.js.map +1 -0
  94. package/dist/lib/agent/run-gherkin-scenario.js +151 -0
  95. package/dist/lib/agent/run-gherkin-scenario.js.map +1 -0
  96. package/dist/lib/agent/task-builder.js +4 -4
  97. package/dist/lib/agent/task-builder.js.map +1 -1
  98. package/dist/lib/agent/tasks.js +4 -8
  99. package/dist/lib/agent/tasks.js.map +1 -1
  100. package/dist/lib/agent/utils.js +14 -2
  101. package/dist/lib/agent/utils.js.map +1 -1
  102. package/dist/lib/ai-model/index.js +0 -10
  103. package/dist/lib/ai-model/inspect.js +77 -50
  104. package/dist/lib/ai-model/inspect.js.map +1 -1
  105. package/dist/lib/ai-model/llm-planning.js +9 -25
  106. package/dist/lib/ai-model/llm-planning.js.map +1 -1
  107. package/dist/lib/ai-model/model-adapter/chat-completion.js +83 -0
  108. package/dist/lib/ai-model/model-adapter/chat-completion.js.map +1 -0
  109. package/dist/lib/ai-model/model-adapter/custom-planning-types.js +20 -0
  110. package/dist/lib/ai-model/model-adapter/custom-planning-types.js.map +1 -0
  111. package/dist/lib/ai-model/model-adapter/index.js +38 -0
  112. package/dist/lib/ai-model/model-adapter/index.js.map +1 -0
  113. package/dist/lib/ai-model/model-adapter/locate.js +67 -0
  114. package/dist/lib/ai-model/model-adapter/locate.js.map +1 -0
  115. package/dist/lib/ai-model/model-adapter/planning.js +81 -0
  116. package/dist/lib/ai-model/model-adapter/planning.js.map +1 -0
  117. package/dist/lib/ai-model/model-adapter/resolve.js +77 -0
  118. package/dist/lib/ai-model/model-adapter/resolve.js.map +1 -0
  119. package/dist/lib/ai-model/model-adapter/types.js.map +1 -0
  120. package/dist/lib/ai-model/models/auto-glm/actions.js +19 -31
  121. package/dist/lib/ai-model/models/auto-glm/actions.js.map +1 -1
  122. package/dist/lib/ai-model/models/auto-glm/adapter.js +6 -14
  123. package/dist/lib/ai-model/models/auto-glm/adapter.js.map +1 -1
  124. package/dist/lib/ai-model/models/auto-glm/locate.js +8 -110
  125. package/dist/lib/ai-model/models/auto-glm/locate.js.map +1 -1
  126. package/dist/lib/ai-model/models/auto-glm/parser.js +24 -46
  127. package/dist/lib/ai-model/models/auto-glm/parser.js.map +1 -1
  128. package/dist/lib/ai-model/models/auto-glm/planning.js +28 -71
  129. package/dist/lib/ai-model/models/auto-glm/planning.js.map +1 -1
  130. package/dist/lib/ai-model/models/default.js.map +1 -1
  131. package/dist/lib/ai-model/models/doubao.js.map +1 -1
  132. package/dist/lib/ai-model/models/gemini.js +2 -2
  133. package/dist/lib/ai-model/models/gemini.js.map +1 -1
  134. package/dist/lib/ai-model/models/glm.js.map +1 -1
  135. package/dist/lib/ai-model/models/gpt.js.map +1 -1
  136. package/dist/lib/ai-model/models/kimi.js.map +1 -1
  137. package/dist/lib/ai-model/models/mimo.js.map +1 -1
  138. package/dist/lib/ai-model/models/qwen.js.map +1 -1
  139. package/dist/lib/ai-model/models/registry.js +2 -2
  140. package/dist/lib/ai-model/models/registry.js.map +1 -1
  141. package/dist/lib/ai-model/models/ui-tars/actions.js +181 -0
  142. package/dist/lib/ai-model/models/ui-tars/actions.js.map +1 -0
  143. package/dist/lib/ai-model/models/ui-tars/adapter.js +1 -1
  144. package/dist/lib/ai-model/models/ui-tars/adapter.js.map +1 -1
  145. package/dist/lib/ai-model/models/ui-tars/parser.js +74 -0
  146. package/dist/lib/ai-model/models/ui-tars/parser.js.map +1 -0
  147. package/dist/lib/ai-model/models/ui-tars/planning.js +21 -232
  148. package/dist/lib/ai-model/models/ui-tars/planning.js.map +1 -1
  149. package/dist/lib/ai-model/{prompt/ui-tars-planning.js → models/ui-tars/prompt.js} +1 -1
  150. package/dist/lib/ai-model/models/ui-tars/prompt.js.map +1 -0
  151. package/dist/lib/ai-model/service-caller/index.js +1 -3
  152. package/dist/lib/ai-model/service-caller/index.js.map +1 -1
  153. package/dist/lib/ai-model/shared/model-locate-result/coordinate-distance.js +44 -0
  154. package/dist/lib/ai-model/shared/model-locate-result/coordinate-distance.js.map +1 -0
  155. package/dist/lib/ai-model/shared/model-locate-result/factory.js +5 -4
  156. package/dist/lib/ai-model/shared/model-locate-result/factory.js.map +1 -1
  157. package/dist/lib/ai-model/shared/model-locate-result/index.js +7 -0
  158. package/dist/lib/ai-model/shared/planning-action.js +40 -0
  159. package/dist/lib/ai-model/shared/planning-action.js.map +1 -0
  160. package/dist/lib/ai-model/workflows/inspect/planning-action-locate.js +129 -0
  161. package/dist/lib/ai-model/workflows/inspect/planning-action-locate.js.map +1 -0
  162. package/dist/lib/ai-model/workflows/planning/custom-planning.js +163 -0
  163. package/dist/lib/ai-model/workflows/planning/custom-planning.js.map +1 -0
  164. package/dist/lib/ai-model/workflows/planning/locate-normalization.js +65 -0
  165. package/dist/lib/ai-model/workflows/planning/locate-normalization.js.map +1 -0
  166. package/dist/lib/common.js +13 -6
  167. package/dist/lib/common.js.map +1 -1
  168. package/dist/lib/element-describer.js +223 -0
  169. package/dist/lib/element-describer.js.map +1 -0
  170. package/dist/lib/index.js +14 -1
  171. package/dist/lib/index.js.map +1 -1
  172. package/dist/lib/report-generator.js +4 -6
  173. package/dist/lib/report-generator.js.map +1 -1
  174. package/dist/lib/service/index.js +3 -2
  175. package/dist/lib/service/index.js.map +1 -1
  176. package/dist/lib/skill/index.js +1 -1
  177. package/dist/lib/skill/index.js.map +1 -1
  178. package/dist/lib/types.js.map +1 -1
  179. package/dist/lib/utils.js +2 -2
  180. package/dist/lib/yaml/player.js +7 -0
  181. package/dist/lib/yaml/player.js.map +1 -1
  182. package/dist/types/agent/agent.d.ts +4 -18
  183. package/dist/types/agent/index.d.ts +3 -2
  184. package/dist/types/agent/prompt-context.d.ts +2 -0
  185. package/dist/types/agent/run-gherkin-scenario.d.ts +26 -0
  186. package/dist/types/agent/tasks.d.ts +3 -1
  187. package/dist/types/ai-model/index.d.ts +0 -2
  188. package/dist/types/ai-model/inspect.d.ts +3 -2
  189. package/dist/types/ai-model/model-adapter/chat-completion.d.ts +3 -0
  190. package/dist/types/ai-model/model-adapter/custom-planning-types.d.ts +32 -0
  191. package/dist/types/ai-model/model-adapter/index.d.ts +2 -0
  192. package/dist/types/ai-model/model-adapter/locate.d.ts +3 -0
  193. package/dist/types/ai-model/model-adapter/planning.d.ts +4 -0
  194. package/dist/types/ai-model/{models → model-adapter}/types.d.ts +39 -4
  195. package/dist/types/ai-model/models/auto-glm/actions.d.ts +8 -5
  196. package/dist/types/ai-model/models/auto-glm/adapter.d.ts +1 -1
  197. package/dist/types/ai-model/models/auto-glm/locate.d.ts +2 -3
  198. package/dist/types/ai-model/models/auto-glm/parser.d.ts +6 -10
  199. package/dist/types/ai-model/models/auto-glm/planning.d.ts +5 -4
  200. package/dist/types/ai-model/models/default.d.ts +1 -1
  201. package/dist/types/ai-model/models/doubao.d.ts +1 -1
  202. package/dist/types/ai-model/models/gemini.d.ts +1 -1
  203. package/dist/types/ai-model/models/glm.d.ts +1 -1
  204. package/dist/types/ai-model/models/gpt.d.ts +1 -1
  205. package/dist/types/ai-model/models/index.d.ts +1 -1
  206. package/dist/types/ai-model/models/kimi.d.ts +1 -1
  207. package/dist/types/ai-model/models/mimo.d.ts +1 -1
  208. package/dist/types/ai-model/models/qwen.d.ts +1 -1
  209. package/dist/types/ai-model/models/registry.d.ts +10 -10
  210. package/dist/types/ai-model/models/ui-tars/actions.d.ts +3 -0
  211. package/dist/types/ai-model/models/ui-tars/adapter.d.ts +1 -1
  212. package/dist/types/ai-model/models/ui-tars/parser.d.ts +72 -0
  213. package/dist/types/ai-model/models/ui-tars/planning.d.ts +3 -68
  214. package/dist/types/ai-model/service-caller/index.d.ts +4 -5
  215. package/dist/types/ai-model/shared/model-locate-result/coordinate-distance.d.ts +4 -0
  216. package/dist/types/ai-model/shared/model-locate-result/factory.d.ts +2 -1
  217. package/dist/types/ai-model/shared/model-locate-result/index.d.ts +4 -2
  218. package/dist/types/ai-model/shared/model-locate-result/types.d.ts +0 -1
  219. package/dist/types/ai-model/shared/planning-action.d.ts +20 -0
  220. package/dist/types/ai-model/workflows/inspect/planning-action-locate.d.ts +4 -0
  221. package/dist/types/ai-model/workflows/inspect/types.d.ts +21 -2
  222. package/dist/types/ai-model/workflows/planning/custom-planning.d.ts +7 -0
  223. package/dist/types/ai-model/workflows/planning/locate-normalization.d.ts +9 -0
  224. package/dist/types/common.d.ts +1 -0
  225. package/dist/types/device/index.d.ts +17 -17
  226. package/dist/types/element-describer.d.ts +46 -0
  227. package/dist/types/index.d.ts +2 -1
  228. package/dist/types/service/index.d.ts +3 -1
  229. package/dist/types/types.d.ts +5 -0
  230. package/dist/types/yaml.d.ts +4 -1
  231. package/package.json +3 -3
  232. package/dist/es/ai-model/models/chat-content.mjs +0 -14
  233. package/dist/es/ai-model/models/chat-content.mjs.map +0 -1
  234. package/dist/es/ai-model/models/resolved.mjs +0 -107
  235. package/dist/es/ai-model/models/resolved.mjs.map +0 -1
  236. package/dist/es/ai-model/prompt/recorder-ui-describer.mjs +0 -56
  237. package/dist/es/ai-model/prompt/recorder-ui-describer.mjs.map +0 -1
  238. package/dist/es/ai-model/prompt/ui-tars-planning.mjs.map +0 -1
  239. package/dist/es/recorder-ui-describer.mjs +0 -379
  240. package/dist/es/recorder-ui-describer.mjs.map +0 -1
  241. package/dist/lib/ai-model/models/chat-content.js.map +0 -1
  242. package/dist/lib/ai-model/models/resolved.js +0 -141
  243. package/dist/lib/ai-model/models/resolved.js.map +0 -1
  244. package/dist/lib/ai-model/models/types.js.map +0 -1
  245. package/dist/lib/ai-model/prompt/recorder-ui-describer.js +0 -90
  246. package/dist/lib/ai-model/prompt/recorder-ui-describer.js.map +0 -1
  247. package/dist/lib/ai-model/prompt/ui-tars-planning.js.map +0 -1
  248. package/dist/lib/recorder-ui-describer.js +0 -419
  249. package/dist/lib/recorder-ui-describer.js.map +0 -1
  250. package/dist/types/ai-model/models/chat-content.d.ts +0 -2
  251. package/dist/types/ai-model/prompt/recorder-ui-describer.d.ts +0 -1
  252. package/dist/types/recorder-ui-describer.d.ts +0 -20
  253. /package/dist/es/ai-model/{models/types.mjs → model-adapter/custom-planning-types.mjs} +0 -0
  254. /package/dist/lib/ai-model/{models → model-adapter}/types.js +0 -0
  255. /package/dist/types/ai-model/{models/resolved.d.ts → model-adapter/resolve.d.ts} +0 -0
  256. /package/dist/types/ai-model/{prompt/ui-tars-planning.d.ts → models/ui-tars/prompt.d.ts} +0 -0
@@ -1 +1 @@
1
- {"version":3,"file":"ai-model/models/mimo.mjs","sources":["../../../../src/ai-model/models/mimo.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type {\n ChatCompletionCallContext,\n ChatCompletionParamsResult,\n ModelAdapterDefinition,\n} from './types';\n\nconst buildMimoChatCompletionParams = (\n input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n const { intent, midsceneDefaults, userConfig } = input;\n const { reasoningEnabled } = userConfig;\n const commonOverrideConfig: Record<string, unknown> = {};\n\n // https://platform.xiaomimimo.com/docs/zh-CN/api/chat/openai-api\n // Observed with thinking disabled: Mimo needs json_object to return JSON.\n commonOverrideConfig.response_format = {\n type: intent === 'default' ? 'json_object' : 'text',\n };\n\n if (userConfig.temperature !== undefined) {\n commonOverrideConfig.temperature = userConfig.temperature;\n }\n\n const modelSpecificConfig: Record<string, unknown> = {\n thinking: {\n type: (reasoningEnabled ?? false) ? 'enabled' : 'disabled',\n },\n };\n\n return {\n config: {\n ...midsceneDefaults,\n ...commonOverrideConfig,\n ...modelSpecificConfig,\n },\n };\n};\n\nexport const mimoAdapters = {\n 'xiaomi-mimo': {\n chatCompletion: {\n unsupportedUserConfig: ['reasoningEffort', 'reasoningBudget'],\n buildChatCompletionParams: buildMimoChatCompletionParams,\n },\n },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'xiaomi-mimo'>;\n"],"names":["buildMimoChatCompletionParams","input","intent","midsceneDefaults","userConfig","reasoningEnabled","commonOverrideConfig","undefined","modelSpecificConfig","mimoAdapters"],"mappings":"AAOA,MAAMA,gCAAgC,CACpCC;IAEA,MAAM,EAAEC,MAAM,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAGH;IACjD,MAAM,EAAEI,gBAAgB,EAAE,GAAGD;IAC7B,MAAME,uBAAgD,CAAC;IAIvDA,qBAAqB,eAAe,GAAG;QACrC,MAAMJ,AAAW,cAAXA,SAAuB,gBAAgB;IAC/C;IAEA,IAAIE,AAA2BG,WAA3BH,WAAW,WAAW,EACxBE,qBAAqB,WAAW,GAAGF,WAAW,WAAW;IAG3D,MAAMI,sBAA+C;QACnD,UAAU;YACR,MAAOH,oBAAoB,QAAS,YAAY;QAClD;IACF;IAEA,OAAO;QACL,QAAQ;YACN,GAAGF,gBAAgB;YACnB,GAAGG,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEO,MAAMC,eAAe;IAC1B,eAAe;QACb,gBAAgB;YACd,uBAAuB;gBAAC;gBAAmB;aAAkB;YAC7D,2BAA2BT;QAC7B;IACF;AACF"}
1
+ {"version":3,"file":"ai-model/models/mimo.mjs","sources":["../../../../src/ai-model/models/mimo.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type {\n ChatCompletionCallContext,\n ChatCompletionParamsResult,\n ModelAdapterDefinition,\n} from '../model-adapter/types';\n\nconst buildMimoChatCompletionParams = (\n input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n const { intent, midsceneDefaults, userConfig } = input;\n const { reasoningEnabled } = userConfig;\n const commonOverrideConfig: Record<string, unknown> = {};\n\n // https://platform.xiaomimimo.com/docs/zh-CN/api/chat/openai-api\n // Observed with thinking disabled: Mimo needs json_object to return JSON.\n commonOverrideConfig.response_format = {\n type: intent === 'default' ? 'json_object' : 'text',\n };\n\n if (userConfig.temperature !== undefined) {\n commonOverrideConfig.temperature = userConfig.temperature;\n }\n\n const modelSpecificConfig: Record<string, unknown> = {\n thinking: {\n type: (reasoningEnabled ?? false) ? 'enabled' : 'disabled',\n },\n };\n\n return {\n config: {\n ...midsceneDefaults,\n ...commonOverrideConfig,\n ...modelSpecificConfig,\n },\n };\n};\n\nexport const mimoAdapters = {\n 'xiaomi-mimo': {\n chatCompletion: {\n unsupportedUserConfig: ['reasoningEffort', 'reasoningBudget'],\n buildChatCompletionParams: buildMimoChatCompletionParams,\n },\n },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'xiaomi-mimo'>;\n"],"names":["buildMimoChatCompletionParams","input","intent","midsceneDefaults","userConfig","reasoningEnabled","commonOverrideConfig","undefined","modelSpecificConfig","mimoAdapters"],"mappings":"AAOA,MAAMA,gCAAgC,CACpCC;IAEA,MAAM,EAAEC,MAAM,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAGH;IACjD,MAAM,EAAEI,gBAAgB,EAAE,GAAGD;IAC7B,MAAME,uBAAgD,CAAC;IAIvDA,qBAAqB,eAAe,GAAG;QACrC,MAAMJ,AAAW,cAAXA,SAAuB,gBAAgB;IAC/C;IAEA,IAAIE,AAA2BG,WAA3BH,WAAW,WAAW,EACxBE,qBAAqB,WAAW,GAAGF,WAAW,WAAW;IAG3D,MAAMI,sBAA+C;QACnD,UAAU;YACR,MAAOH,oBAAoB,QAAS,YAAY;QAClD;IACF;IAEA,OAAO;QACL,QAAQ;YACN,GAAGF,gBAAgB;YACnB,GAAGG,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEO,MAAMC,eAAe;IAC1B,eAAe;QACb,gBAAgB;YACd,uBAAuB;gBAAC;gBAAmB;aAAkB;YAC7D,2BAA2BT;QAC7B;IACF;AACF"}
@@ -1 +1 @@
1
- {"version":3,"file":"ai-model/models/qwen.mjs","sources":["../../../../src/ai-model/models/qwen.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport {\n type LocateResultValue,\n type PixelBbox,\n unwrapCoordinateListLikeInput,\n} from '../shared/model-locate-result';\nimport type {\n ChatCompletionCallContext,\n ChatCompletionParamsResult,\n ModelAdapterDefinition,\n} from './types';\n\nconst defaultBboxSize = 20;\n\nfunction topLeftPointToPixelBbox(x: number, y: number): PixelBbox {\n return [\n Math.round(x),\n Math.round(y),\n Math.round(x + defaultBboxSize),\n Math.round(y + defaultBboxSize),\n ];\n}\n\nfunction parseQwen25RawLocateValue(input: unknown): LocateResultValue {\n const bbox = unwrapCoordinateListLikeInput(input as any) as number[];\n if (bbox.length < 2) {\n const msg = `invalid bbox data for qwen-vl mode: ${JSON.stringify(bbox)} `;\n throw new Error(msg);\n }\n\n if (typeof bbox[2] === 'number' && typeof bbox[3] === 'number') {\n return {\n type: 'bbox',\n coordinates: [bbox[0], bbox[1], bbox[2], bbox[3]],\n };\n }\n\n return { type: 'point', coordinates: [bbox[0], bbox[1]] };\n}\n\nfunction normalizeQwen25ResultToPixelBbox(\n result: LocateResultValue,\n): PixelBbox {\n if (result.type === 'bbox') {\n return [\n Math.round(result.coordinates[0]),\n Math.round(result.coordinates[1]),\n Math.round(result.coordinates[2]),\n Math.round(result.coordinates[3]),\n ];\n }\n\n return topLeftPointToPixelBbox(result.coordinates[0], result.coordinates[1]);\n}\n\nconst buildQwenChatCompletionParams = (\n input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n const { midsceneDefaults, userConfig } = input;\n const { reasoningEnabled, reasoningBudget } = userConfig;\n\n const commonOverrideConfig: Record<string, unknown> = {};\n\n if (userConfig.temperature !== undefined) {\n commonOverrideConfig.temperature = userConfig.temperature;\n }\n\n const modelSpecificConfig: Record<string, unknown> = {};\n\n if (reasoningEnabled !== 'default') {\n modelSpecificConfig.enable_thinking = reasoningEnabled ?? false;\n if (reasoningBudget !== undefined) {\n modelSpecificConfig.thinking_budget = reasoningBudget;\n }\n }\n\n return {\n config: {\n ...midsceneDefaults,\n ...commonOverrideConfig,\n ...modelSpecificConfig,\n },\n };\n};\n\nconst buildQwen25ChatCompletionParams = (\n input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n const { midsceneDefaults, userConfig } = input;\n const commonOverrideConfig: Record<string, unknown> = {};\n\n if (userConfig.temperature !== undefined) {\n commonOverrideConfig.temperature = userConfig.temperature;\n }\n\n return {\n config: {\n ...midsceneDefaults,\n ...commonOverrideConfig,\n vl_high_resolution_images: true,\n },\n };\n};\n\nconst qwen3Adapter: ModelAdapterDefinition = {\n chatCompletion: {\n unsupportedUserConfig: ['reasoningEffort'],\n buildChatCompletionParams: buildQwenChatCompletionParams,\n },\n locate: {\n resultAdapter: {\n coordinates: { shape: 'bbox', order: 'xy', normalizedBy: 1000 },\n },\n },\n};\n\nexport const qwenAdapters = {\n 'qwen2.5-vl': {\n chatCompletion: {\n unsupportedUserConfig: [\n 'reasoningEnabled',\n 'reasoningEffort',\n 'reasoningBudget',\n ],\n buildChatCompletionParams: buildQwen25ChatCompletionParams,\n },\n imagePreprocess: {\n padBlockSize: 28,\n },\n locate: {\n resultAdapter: {\n coordinates: { shape: 'bbox', order: 'xy' },\n parseRawLocateValue: parseQwen25RawLocateValue,\n mapLocateResultToPixelBbox: normalizeQwen25ResultToPixelBbox,\n },\n },\n },\n 'qwen3-vl': qwen3Adapter,\n qwen3: qwen3Adapter,\n 'qwen3.5': qwen3Adapter,\n 'qwen3.6': qwen3Adapter,\n} satisfies Pick<\n Record<TModelFamily, ModelAdapterDefinition>,\n 'qwen2.5-vl' | 'qwen3-vl' | 'qwen3' | 'qwen3.5' | 'qwen3.6'\n>;\n"],"names":["defaultBboxSize","topLeftPointToPixelBbox","x","y","Math","parseQwen25RawLocateValue","input","bbox","unwrapCoordinateListLikeInput","msg","JSON","Error","normalizeQwen25ResultToPixelBbox","result","buildQwenChatCompletionParams","midsceneDefaults","userConfig","reasoningEnabled","reasoningBudget","commonOverrideConfig","undefined","modelSpecificConfig","buildQwen25ChatCompletionParams","qwen3Adapter","qwenAdapters"],"mappings":";AAYA,MAAMA,kBAAkB;AAExB,SAASC,wBAAwBC,CAAS,EAAEC,CAAS;IACnD,OAAO;QACLC,KAAK,KAAK,CAACF;QACXE,KAAK,KAAK,CAACD;QACXC,KAAK,KAAK,CAACF,IAAIF;QACfI,KAAK,KAAK,CAACD,IAAIH;KAChB;AACH;AAEA,SAASK,0BAA0BC,KAAc;IAC/C,MAAMC,OAAOC,8BAA8BF;IAC3C,IAAIC,KAAK,MAAM,GAAG,GAAG;QACnB,MAAME,MAAM,CAAC,oCAAoC,EAAEC,KAAK,SAAS,CAACH,MAAM,CAAC,CAAC;QAC1E,MAAM,IAAII,MAAMF;IAClB;IAEA,IAAI,AAAmB,YAAnB,OAAOF,IAAI,CAAC,EAAE,IAAiB,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EAC/C,OAAO;QACL,MAAM;QACN,aAAa;YAACA,IAAI,CAAC,EAAE;YAAEA,IAAI,CAAC,EAAE;YAAEA,IAAI,CAAC,EAAE;YAAEA,IAAI,CAAC,EAAE;SAAC;IACnD;IAGF,OAAO;QAAE,MAAM;QAAS,aAAa;YAACA,IAAI,CAAC,EAAE;YAAEA,IAAI,CAAC,EAAE;SAAC;IAAC;AAC1D;AAEA,SAASK,iCACPC,MAAyB;IAEzB,IAAIA,AAAgB,WAAhBA,OAAO,IAAI,EACb,OAAO;QACLT,KAAK,KAAK,CAACS,OAAO,WAAW,CAAC,EAAE;QAChCT,KAAK,KAAK,CAACS,OAAO,WAAW,CAAC,EAAE;QAChCT,KAAK,KAAK,CAACS,OAAO,WAAW,CAAC,EAAE;QAChCT,KAAK,KAAK,CAACS,OAAO,WAAW,CAAC,EAAE;KACjC;IAGH,OAAOZ,wBAAwBY,OAAO,WAAW,CAAC,EAAE,EAAEA,OAAO,WAAW,CAAC,EAAE;AAC7E;AAEA,MAAMC,gCAAgC,CACpCR;IAEA,MAAM,EAAES,gBAAgB,EAAEC,UAAU,EAAE,GAAGV;IACzC,MAAM,EAAEW,gBAAgB,EAAEC,eAAe,EAAE,GAAGF;IAE9C,MAAMG,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BI,WAA3BJ,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAG3D,MAAMK,sBAA+C,CAAC;IAEtD,IAAIJ,AAAqB,cAArBA,kBAAgC;QAClCI,oBAAoB,eAAe,GAAGJ,oBAAoB;QAC1D,IAAIC,AAAoBE,WAApBF,iBACFG,oBAAoB,eAAe,GAAGH;IAE1C;IAEA,OAAO;QACL,QAAQ;YACN,GAAGH,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEA,MAAMC,kCAAkC,CACtChB;IAEA,MAAM,EAAES,gBAAgB,EAAEC,UAAU,EAAE,GAAGV;IACzC,MAAMa,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BI,WAA3BJ,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAG3D,OAAO;QACL,QAAQ;YACN,GAAGD,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,2BAA2B;QAC7B;IACF;AACF;AAEA,MAAMI,eAAuC;IAC3C,gBAAgB;QACd,uBAAuB;YAAC;SAAkB;QAC1C,2BAA2BT;IAC7B;IACA,QAAQ;QACN,eAAe;YACb,aAAa;gBAAE,OAAO;gBAAQ,OAAO;gBAAM,cAAc;YAAK;QAChE;IACF;AACF;AAEO,MAAMU,eAAe;IAC1B,cAAc;QACZ,gBAAgB;YACd,uBAAuB;gBACrB;gBACA;gBACA;aACD;YACD,2BAA2BF;QAC7B;QACA,iBAAiB;YACf,cAAc;QAChB;QACA,QAAQ;YACN,eAAe;gBACb,aAAa;oBAAE,OAAO;oBAAQ,OAAO;gBAAK;gBAC1C,qBAAqBjB;gBACrB,4BAA4BO;YAC9B;QACF;IACF;IACA,YAAYW;IACZ,OAAOA;IACP,WAAWA;IACX,WAAWA;AACb"}
1
+ {"version":3,"file":"ai-model/models/qwen.mjs","sources":["../../../../src/ai-model/models/qwen.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type {\n ChatCompletionCallContext,\n ChatCompletionParamsResult,\n ModelAdapterDefinition,\n} from '../model-adapter/types';\nimport {\n type LocateResultValue,\n type PixelBbox,\n unwrapCoordinateListLikeInput,\n} from '../shared/model-locate-result';\n\nconst defaultBboxSize = 20;\n\nfunction topLeftPointToPixelBbox(x: number, y: number): PixelBbox {\n return [\n Math.round(x),\n Math.round(y),\n Math.round(x + defaultBboxSize),\n Math.round(y + defaultBboxSize),\n ];\n}\n\nfunction parseQwen25RawLocateValue(input: unknown): LocateResultValue {\n const bbox = unwrapCoordinateListLikeInput(input as any) as number[];\n if (bbox.length < 2) {\n const msg = `invalid bbox data for qwen-vl mode: ${JSON.stringify(bbox)} `;\n throw new Error(msg);\n }\n\n if (typeof bbox[2] === 'number' && typeof bbox[3] === 'number') {\n return {\n type: 'bbox',\n coordinates: [bbox[0], bbox[1], bbox[2], bbox[3]],\n };\n }\n\n return { type: 'point', coordinates: [bbox[0], bbox[1]] };\n}\n\nfunction normalizeQwen25ResultToPixelBbox(\n result: LocateResultValue,\n): PixelBbox {\n if (result.type === 'bbox') {\n return [\n Math.round(result.coordinates[0]),\n Math.round(result.coordinates[1]),\n Math.round(result.coordinates[2]),\n Math.round(result.coordinates[3]),\n ];\n }\n\n return topLeftPointToPixelBbox(result.coordinates[0], result.coordinates[1]);\n}\n\nconst buildQwenChatCompletionParams = (\n input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n const { midsceneDefaults, userConfig } = input;\n const { reasoningEnabled, reasoningBudget } = userConfig;\n\n const commonOverrideConfig: Record<string, unknown> = {};\n\n if (userConfig.temperature !== undefined) {\n commonOverrideConfig.temperature = userConfig.temperature;\n }\n\n const modelSpecificConfig: Record<string, unknown> = {};\n\n if (reasoningEnabled !== 'default') {\n modelSpecificConfig.enable_thinking = reasoningEnabled ?? false;\n if (reasoningBudget !== undefined) {\n modelSpecificConfig.thinking_budget = reasoningBudget;\n }\n }\n\n return {\n config: {\n ...midsceneDefaults,\n ...commonOverrideConfig,\n ...modelSpecificConfig,\n },\n };\n};\n\nconst buildQwen25ChatCompletionParams = (\n input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n const { midsceneDefaults, userConfig } = input;\n const commonOverrideConfig: Record<string, unknown> = {};\n\n if (userConfig.temperature !== undefined) {\n commonOverrideConfig.temperature = userConfig.temperature;\n }\n\n return {\n config: {\n ...midsceneDefaults,\n ...commonOverrideConfig,\n vl_high_resolution_images: true,\n },\n };\n};\n\nconst qwen3Adapter: ModelAdapterDefinition = {\n chatCompletion: {\n unsupportedUserConfig: ['reasoningEffort'],\n buildChatCompletionParams: buildQwenChatCompletionParams,\n },\n locate: {\n resultAdapter: {\n coordinates: { shape: 'bbox', order: 'xy', normalizedBy: 1000 },\n },\n },\n};\n\nexport const qwenAdapters = {\n 'qwen2.5-vl': {\n chatCompletion: {\n unsupportedUserConfig: [\n 'reasoningEnabled',\n 'reasoningEffort',\n 'reasoningBudget',\n ],\n buildChatCompletionParams: buildQwen25ChatCompletionParams,\n },\n imagePreprocess: {\n padBlockSize: 28,\n },\n locate: {\n resultAdapter: {\n coordinates: { shape: 'bbox', order: 'xy' },\n parseRawLocateValue: parseQwen25RawLocateValue,\n mapLocateResultToPixelBbox: normalizeQwen25ResultToPixelBbox,\n },\n },\n },\n 'qwen3-vl': qwen3Adapter,\n qwen3: qwen3Adapter,\n 'qwen3.5': qwen3Adapter,\n 'qwen3.6': qwen3Adapter,\n} satisfies Pick<\n Record<TModelFamily, ModelAdapterDefinition>,\n 'qwen2.5-vl' | 'qwen3-vl' | 'qwen3' | 'qwen3.5' | 'qwen3.6'\n>;\n"],"names":["defaultBboxSize","topLeftPointToPixelBbox","x","y","Math","parseQwen25RawLocateValue","input","bbox","unwrapCoordinateListLikeInput","msg","JSON","Error","normalizeQwen25ResultToPixelBbox","result","buildQwenChatCompletionParams","midsceneDefaults","userConfig","reasoningEnabled","reasoningBudget","commonOverrideConfig","undefined","modelSpecificConfig","buildQwen25ChatCompletionParams","qwen3Adapter","qwenAdapters"],"mappings":";AAYA,MAAMA,kBAAkB;AAExB,SAASC,wBAAwBC,CAAS,EAAEC,CAAS;IACnD,OAAO;QACLC,KAAK,KAAK,CAACF;QACXE,KAAK,KAAK,CAACD;QACXC,KAAK,KAAK,CAACF,IAAIF;QACfI,KAAK,KAAK,CAACD,IAAIH;KAChB;AACH;AAEA,SAASK,0BAA0BC,KAAc;IAC/C,MAAMC,OAAOC,8BAA8BF;IAC3C,IAAIC,KAAK,MAAM,GAAG,GAAG;QACnB,MAAME,MAAM,CAAC,oCAAoC,EAAEC,KAAK,SAAS,CAACH,MAAM,CAAC,CAAC;QAC1E,MAAM,IAAII,MAAMF;IAClB;IAEA,IAAI,AAAmB,YAAnB,OAAOF,IAAI,CAAC,EAAE,IAAiB,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EAC/C,OAAO;QACL,MAAM;QACN,aAAa;YAACA,IAAI,CAAC,EAAE;YAAEA,IAAI,CAAC,EAAE;YAAEA,IAAI,CAAC,EAAE;YAAEA,IAAI,CAAC,EAAE;SAAC;IACnD;IAGF,OAAO;QAAE,MAAM;QAAS,aAAa;YAACA,IAAI,CAAC,EAAE;YAAEA,IAAI,CAAC,EAAE;SAAC;IAAC;AAC1D;AAEA,SAASK,iCACPC,MAAyB;IAEzB,IAAIA,AAAgB,WAAhBA,OAAO,IAAI,EACb,OAAO;QACLT,KAAK,KAAK,CAACS,OAAO,WAAW,CAAC,EAAE;QAChCT,KAAK,KAAK,CAACS,OAAO,WAAW,CAAC,EAAE;QAChCT,KAAK,KAAK,CAACS,OAAO,WAAW,CAAC,EAAE;QAChCT,KAAK,KAAK,CAACS,OAAO,WAAW,CAAC,EAAE;KACjC;IAGH,OAAOZ,wBAAwBY,OAAO,WAAW,CAAC,EAAE,EAAEA,OAAO,WAAW,CAAC,EAAE;AAC7E;AAEA,MAAMC,gCAAgC,CACpCR;IAEA,MAAM,EAAES,gBAAgB,EAAEC,UAAU,EAAE,GAAGV;IACzC,MAAM,EAAEW,gBAAgB,EAAEC,eAAe,EAAE,GAAGF;IAE9C,MAAMG,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BI,WAA3BJ,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAG3D,MAAMK,sBAA+C,CAAC;IAEtD,IAAIJ,AAAqB,cAArBA,kBAAgC;QAClCI,oBAAoB,eAAe,GAAGJ,oBAAoB;QAC1D,IAAIC,AAAoBE,WAApBF,iBACFG,oBAAoB,eAAe,GAAGH;IAE1C;IAEA,OAAO;QACL,QAAQ;YACN,GAAGH,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEA,MAAMC,kCAAkC,CACtChB;IAEA,MAAM,EAAES,gBAAgB,EAAEC,UAAU,EAAE,GAAGV;IACzC,MAAMa,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BI,WAA3BJ,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAG3D,OAAO;QACL,QAAQ;YACN,GAAGD,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,2BAA2B;QAC7B;IACF;AACF;AAEA,MAAMI,eAAuC;IAC3C,gBAAgB;QACd,uBAAuB;YAAC;SAAkB;QAC1C,2BAA2BT;IAC7B;IACA,QAAQ;QACN,eAAe;YACb,aAAa;gBAAE,OAAO;gBAAQ,OAAO;gBAAM,cAAc;YAAK;QAChE;IACF;AACF;AAEO,MAAMU,eAAe;IAC1B,cAAc;QACZ,gBAAgB;YACd,uBAAuB;gBACrB;gBACA;gBACA;aACD;YACD,2BAA2BF;QAC7B;QACA,iBAAiB;YACf,cAAc;QAChB;QACA,QAAQ;YACN,eAAe;gBACb,aAAa;oBAAE,OAAO;oBAAQ,OAAO;gBAAK;gBAC1C,qBAAqBjB;gBACrB,4BAA4BO;YAC9B;QACF;IACF;IACA,YAAYW;IACZ,OAAOA;IACP,WAAWA;IACX,WAAWA;AACb"}
@@ -1,4 +1,5 @@
1
1
  import { getDebug } from "@midscene/shared/logger";
2
+ import { ResolvedModelAdapter } from "../model-adapter/resolve.mjs";
2
3
  import { autoGlmAdapters } from "./auto-glm/adapter.mjs";
3
4
  import { defaultOpenAICompatibleAdapterConfig } from "./default.mjs";
4
5
  import { doubaoAdapters } from "./doubao.mjs";
@@ -8,7 +9,6 @@ import { gptAdapters } from "./gpt.mjs";
8
9
  import { kimiAdapters } from "./kimi.mjs";
9
10
  import { mimoAdapters } from "./mimo.mjs";
10
11
  import { qwenAdapters } from "./qwen.mjs";
11
- import { ResolvedModelAdapter } from "./resolved.mjs";
12
12
  import { uiTarsAdapters } from "./ui-tars/adapter.mjs";
13
13
  const MODEL_ADAPTER_CONFIGS = {
14
14
  ...qwenAdapters,
@@ -1 +1 @@
1
- {"version":3,"file":"ai-model/models/registry.mjs","sources":["../../../../src/ai-model/models/registry.ts"],"sourcesContent":["import type { IModelConfig, TModelFamily } from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { autoGlmAdapters } from './auto-glm/adapter';\nimport { defaultOpenAICompatibleAdapterConfig } from './default';\nimport { doubaoAdapters } from './doubao';\nimport { geminiAdapters } from './gemini';\nimport { glmAdapters } from './glm';\nimport { gptAdapters } from './gpt';\nimport { kimiAdapters } from './kimi';\nimport { mimoAdapters } from './mimo';\nimport { qwenAdapters } from './qwen';\nimport { ResolvedModelAdapter } from './resolved';\nimport type {\n ModelAdapter,\n ModelAdapterDefinition,\n ModelRuntime,\n} from './types';\nimport { uiTarsAdapters } from './ui-tars/adapter';\n\nexport const MODEL_ADAPTER_CONFIGS = {\n ...qwenAdapters,\n ...doubaoAdapters,\n ...geminiAdapters,\n ...uiTarsAdapters,\n ...glmAdapters,\n ...autoGlmAdapters,\n ...gptAdapters,\n ...kimiAdapters,\n ...mimoAdapters,\n} satisfies Record<TModelFamily, ModelAdapterDefinition>;\n\ntype ModelAdapterCacheKey = TModelFamily | 'default';\n\nconst modelAdapterCache = new Map<ModelAdapterCacheKey, ModelAdapter>();\nconst debugModelAdapter = getDebug('ai:model-adapter');\n\nfunction debugAdapterUnsupportedUserConfig(\n modelFamily: ModelAdapterCacheKey,\n adapter: ModelAdapter,\n): void {\n if (adapter.chatCompletion.unsupportedUserConfig.length === 0) {\n return;\n }\n\n debugModelAdapter(\n `model adapter \"${modelFamily}\" unsupportedUserConfig: ${JSON.stringify(\n adapter.chatCompletion.unsupportedUserConfig,\n )}`,\n );\n}\n\nexport function getModelAdapter(modelFamily?: TModelFamily): ModelAdapter {\n const cacheKey: ModelAdapterCacheKey = modelFamily ?? 'default';\n let adapter = modelAdapterCache.get(cacheKey);\n if (adapter) {\n return adapter;\n }\n\n const config = modelFamily\n ? MODEL_ADAPTER_CONFIGS[modelFamily]\n : defaultOpenAICompatibleAdapterConfig;\n if (!config) {\n throw new Error(\n `No model adapter registered for modelFamily: ${modelFamily}`,\n );\n }\n\n adapter = new ResolvedModelAdapter(config, cacheKey);\n modelAdapterCache.set(cacheKey, adapter);\n debugAdapterUnsupportedUserConfig(cacheKey, adapter);\n\n return adapter;\n}\n\nexport function getModelRuntime(config: IModelConfig): ModelRuntime {\n return {\n config,\n adapter: getModelAdapter(config.modelFamily),\n };\n}\n"],"names":["MODEL_ADAPTER_CONFIGS","qwenAdapters","doubaoAdapters","geminiAdapters","uiTarsAdapters","glmAdapters","autoGlmAdapters","gptAdapters","kimiAdapters","mimoAdapters","modelAdapterCache","Map","debugModelAdapter","getDebug","debugAdapterUnsupportedUserConfig","modelFamily","adapter","JSON","getModelAdapter","cacheKey","config","defaultOpenAICompatibleAdapterConfig","Error","ResolvedModelAdapter","getModelRuntime"],"mappings":";;;;;;;;;;;;AAmBO,MAAMA,wBAAwB;IACnC,GAAGC,YAAY;IACf,GAAGC,cAAc;IACjB,GAAGC,cAAc;IACjB,GAAGC,cAAc;IACjB,GAAGC,WAAW;IACd,GAAGC,eAAe;IAClB,GAAGC,WAAW;IACd,GAAGC,YAAY;IACf,GAAGC,YAAY;AACjB;AAIA,MAAMC,oBAAoB,IAAIC;AAC9B,MAAMC,oBAAoBC,SAAS;AAEnC,SAASC,kCACPC,WAAiC,EACjCC,OAAqB;IAErB,IAAIA,AAAwD,MAAxDA,QAAQ,cAAc,CAAC,qBAAqB,CAAC,MAAM,EACrD;IAGFJ,kBACE,CAAC,eAAe,EAAEG,YAAY,yBAAyB,EAAEE,KAAK,SAAS,CACrED,QAAQ,cAAc,CAAC,qBAAqB,GAC3C;AAEP;AAEO,SAASE,gBAAgBH,WAA0B;IACxD,MAAMI,WAAiCJ,eAAe;IACtD,IAAIC,UAAUN,kBAAkB,GAAG,CAACS;IACpC,IAAIH,SACF,OAAOA;IAGT,MAAMI,SAASL,cACXf,qBAAqB,CAACe,YAAY,GAClCM;IACJ,IAAI,CAACD,QACH,MAAM,IAAIE,MACR,CAAC,6CAA6C,EAAEP,aAAa;IAIjEC,UAAU,IAAIO,qBAAqBH,QAAQD;IAC3CT,kBAAkB,GAAG,CAACS,UAAUH;IAChCF,kCAAkCK,UAAUH;IAE5C,OAAOA;AACT;AAEO,SAASQ,gBAAgBJ,MAAoB;IAClD,OAAO;QACLA;QACA,SAASF,gBAAgBE,OAAO,WAAW;IAC7C;AACF"}
1
+ {"version":3,"file":"ai-model/models/registry.mjs","sources":["../../../../src/ai-model/models/registry.ts"],"sourcesContent":["import type { IModelConfig, TModelFamily } from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { ResolvedModelAdapter } from '../model-adapter/resolve';\nimport type {\n ModelAdapter,\n ModelAdapterDefinition,\n ModelRuntime,\n} from '../model-adapter/types';\nimport { autoGlmAdapters } from './auto-glm/adapter';\nimport { defaultOpenAICompatibleAdapterConfig } from './default';\nimport { doubaoAdapters } from './doubao';\nimport { geminiAdapters } from './gemini';\nimport { glmAdapters } from './glm';\nimport { gptAdapters } from './gpt';\nimport { kimiAdapters } from './kimi';\nimport { mimoAdapters } from './mimo';\nimport { qwenAdapters } from './qwen';\nimport { uiTarsAdapters } from './ui-tars/adapter';\n\nexport const MODEL_ADAPTER_CONFIGS = {\n ...qwenAdapters,\n ...doubaoAdapters,\n ...geminiAdapters,\n ...uiTarsAdapters,\n ...glmAdapters,\n ...autoGlmAdapters,\n ...gptAdapters,\n ...kimiAdapters,\n ...mimoAdapters,\n} satisfies Record<TModelFamily, ModelAdapterDefinition>;\n\ntype ModelAdapterCacheKey = TModelFamily | 'default';\n\nconst modelAdapterCache = new Map<ModelAdapterCacheKey, ModelAdapter>();\nconst debugModelAdapter = getDebug('ai:model-adapter');\n\nfunction debugAdapterUnsupportedUserConfig(\n modelFamily: ModelAdapterCacheKey,\n adapter: ModelAdapter,\n): void {\n if (adapter.chatCompletion.unsupportedUserConfig.length === 0) {\n return;\n }\n\n debugModelAdapter(\n `model adapter \"${modelFamily}\" unsupportedUserConfig: ${JSON.stringify(\n adapter.chatCompletion.unsupportedUserConfig,\n )}`,\n );\n}\n\nexport function getModelAdapter(modelFamily?: TModelFamily): ModelAdapter {\n const cacheKey: ModelAdapterCacheKey = modelFamily ?? 'default';\n let adapter = modelAdapterCache.get(cacheKey);\n if (adapter) {\n return adapter;\n }\n\n const config = modelFamily\n ? MODEL_ADAPTER_CONFIGS[modelFamily]\n : defaultOpenAICompatibleAdapterConfig;\n if (!config) {\n throw new Error(\n `No model adapter registered for modelFamily: ${modelFamily}`,\n );\n }\n\n adapter = new ResolvedModelAdapter(config, cacheKey);\n modelAdapterCache.set(cacheKey, adapter);\n debugAdapterUnsupportedUserConfig(cacheKey, adapter);\n\n return adapter;\n}\n\nexport function getModelRuntime(config: IModelConfig): ModelRuntime {\n return {\n config,\n adapter: getModelAdapter(config.modelFamily),\n };\n}\n"],"names":["MODEL_ADAPTER_CONFIGS","qwenAdapters","doubaoAdapters","geminiAdapters","uiTarsAdapters","glmAdapters","autoGlmAdapters","gptAdapters","kimiAdapters","mimoAdapters","modelAdapterCache","Map","debugModelAdapter","getDebug","debugAdapterUnsupportedUserConfig","modelFamily","adapter","JSON","getModelAdapter","cacheKey","config","defaultOpenAICompatibleAdapterConfig","Error","ResolvedModelAdapter","getModelRuntime"],"mappings":";;;;;;;;;;;;AAmBO,MAAMA,wBAAwB;IACnC,GAAGC,YAAY;IACf,GAAGC,cAAc;IACjB,GAAGC,cAAc;IACjB,GAAGC,cAAc;IACjB,GAAGC,WAAW;IACd,GAAGC,eAAe;IAClB,GAAGC,WAAW;IACd,GAAGC,YAAY;IACf,GAAGC,YAAY;AACjB;AAIA,MAAMC,oBAAoB,IAAIC;AAC9B,MAAMC,oBAAoBC,SAAS;AAEnC,SAASC,kCACPC,WAAiC,EACjCC,OAAqB;IAErB,IAAIA,AAAwD,MAAxDA,QAAQ,cAAc,CAAC,qBAAqB,CAAC,MAAM,EACrD;IAGFJ,kBACE,CAAC,eAAe,EAAEG,YAAY,yBAAyB,EAAEE,KAAK,SAAS,CACrED,QAAQ,cAAc,CAAC,qBAAqB,GAC3C;AAEP;AAEO,SAASE,gBAAgBH,WAA0B;IACxD,MAAMI,WAAiCJ,eAAe;IACtD,IAAIC,UAAUN,kBAAkB,GAAG,CAACS;IACpC,IAAIH,SACF,OAAOA;IAGT,MAAMI,SAASL,cACXf,qBAAqB,CAACe,YAAY,GAClCM;IACJ,IAAI,CAACD,QACH,MAAM,IAAIE,MACR,CAAC,6CAA6C,EAAEP,aAAa;IAIjEC,UAAU,IAAIO,qBAAqBH,QAAQD;IAC3CT,kBAAkB,GAAG,CAACS,UAAUH;IAChCF,kCAAkCK,UAAUH;IAE5C,OAAOA;AACT;AAEO,SAASQ,gBAAgBJ,MAAoB;IAClD,OAAO;QACLA;QACA,SAASF,gBAAgBE,OAAO,WAAW;IAC7C;AACF"}
@@ -0,0 +1,147 @@
1
+ import { getDebug } from "@midscene/shared/logger";
2
+ import { transformHotkeyInput } from "@midscene/shared/us-keyboard-layout";
3
+ import { assert } from "@midscene/shared/utils";
4
+ const debug = getDebug('ui-tars-planning');
5
+ const warnLog = getDebug('ui-tars-planning', {
6
+ console: true
7
+ });
8
+ function transformUiTarsActions(parsedPlanningResponse) {
9
+ const transformActions = [];
10
+ const unhandledActions = [];
11
+ parsedPlanningResponse.actions.forEach((action)=>{
12
+ const actionType = (action.action_type || '').toLowerCase();
13
+ if ('click' === actionType) {
14
+ assert(action.action_inputs.start_box, 'start_box is required');
15
+ const point = getPoint(action.action_inputs.start_box);
16
+ const locate = {
17
+ point,
18
+ prompt: action.thought || ''
19
+ };
20
+ transformActions.push({
21
+ type: 'Tap',
22
+ param: {
23
+ locate
24
+ }
25
+ });
26
+ } else if ('left_double' === actionType) {
27
+ assert(action.action_inputs.start_box, 'start_box is required');
28
+ const point = getPoint(action.action_inputs.start_box);
29
+ const locate = {
30
+ point,
31
+ prompt: action.thought || ''
32
+ };
33
+ transformActions.push({
34
+ type: 'DoubleClick',
35
+ param: {
36
+ locate
37
+ },
38
+ thought: action.thought || ''
39
+ });
40
+ } else if ('right_single' === actionType) {
41
+ assert(action.action_inputs.start_box, 'start_box is required');
42
+ const point = getPoint(action.action_inputs.start_box);
43
+ const locate = {
44
+ point,
45
+ prompt: action.thought || ''
46
+ };
47
+ transformActions.push({
48
+ type: 'RightClick',
49
+ param: {
50
+ locate
51
+ },
52
+ thought: action.thought || ''
53
+ });
54
+ } else if ('drag' === actionType) {
55
+ assert(action.action_inputs.start_box, 'start_box is required');
56
+ assert(action.action_inputs.end_box, 'end_box is required');
57
+ const startPoint = getPoint(action.action_inputs.start_box);
58
+ const endPoint = getPoint(action.action_inputs.end_box);
59
+ transformActions.push({
60
+ type: 'DragAndDrop',
61
+ param: {
62
+ from: {
63
+ point: startPoint,
64
+ prompt: action.thought || ''
65
+ },
66
+ to: {
67
+ point: endPoint,
68
+ prompt: action.thought || ''
69
+ }
70
+ },
71
+ thought: action.thought || ''
72
+ });
73
+ } else if ('type' === actionType) transformActions.push({
74
+ type: 'Input',
75
+ param: {
76
+ value: action.action_inputs.content
77
+ },
78
+ thought: action.thought || ''
79
+ });
80
+ else if ('scroll' === actionType) transformActions.push({
81
+ type: 'Scroll',
82
+ param: {
83
+ direction: action.action_inputs.direction
84
+ },
85
+ thought: action.thought || ''
86
+ });
87
+ else if ('finished' === actionType) transformActions.push({
88
+ type: 'Finished',
89
+ param: {},
90
+ thought: action.action_inputs.content || action.thought || ''
91
+ });
92
+ else if ('hotkey' === actionType) if (action.action_inputs.key) {
93
+ const keys = transformHotkeyInput(action.action_inputs.key);
94
+ transformActions.push({
95
+ type: 'KeyboardPress',
96
+ param: {
97
+ keyName: keys.join('+')
98
+ },
99
+ thought: action.thought || ''
100
+ });
101
+ } else warnLog('No key found in action: hotkey. Will not perform action.');
102
+ else if ('wait' === actionType) transformActions.push({
103
+ type: 'Sleep',
104
+ param: {
105
+ timeMs: 1000
106
+ },
107
+ thought: action.thought || ''
108
+ });
109
+ else if (actionType) {
110
+ unhandledActions.push({
111
+ type: actionType,
112
+ thought: action.thought || ''
113
+ });
114
+ debug('Unhandled action type:', actionType, 'thought:', action.thought);
115
+ }
116
+ });
117
+ if (0 === transformActions.length) throw new Error(buildNoUiTarsActionsError(parsedPlanningResponse.rawResponse, parsedPlanningResponse.actions, unhandledActions));
118
+ debug('transformActions', JSON.stringify(transformActions, null, 2));
119
+ return transformActions;
120
+ }
121
+ function getPoint(startBox) {
122
+ const [x, y] = JSON.parse(startBox);
123
+ assert('number' == typeof x && Number.isFinite(x) && 'number' == typeof y && Number.isFinite(y), `invalid point data for ui-tars planning: ${startBox}`);
124
+ return [
125
+ x,
126
+ y
127
+ ];
128
+ }
129
+ function buildNoUiTarsActionsError(rawResponse, actions, unhandledActions) {
130
+ const errorDetails = [];
131
+ if (0 === actions.length) {
132
+ errorDetails.push('Action parser returned no actions');
133
+ if (rawResponse.includes('Thought:') && !rawResponse.includes('Action:')) errorDetails.push('Response contains "Thought:" but missing "Action:" line');
134
+ else errorDetails.push('Response may be malformed or empty');
135
+ }
136
+ if (unhandledActions.length > 0) {
137
+ const types = unhandledActions.map((a)=>a.type).join(', ');
138
+ errorDetails.push(`Unhandled action types: ${types}`);
139
+ }
140
+ return [
141
+ 'No actions found in UI-TARS response.',
142
+ ...errorDetails
143
+ ].join('\n');
144
+ }
145
+ export { transformUiTarsActions };
146
+
147
+ //# sourceMappingURL=actions.mjs.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"ai-model/models/ui-tars/actions.mjs","sources":["../../../../../src/ai-model/models/ui-tars/actions.ts"],"sourcesContent":["import type { PlanningAction } from '@/types';\nimport { getDebug } from '@midscene/shared/logger';\nimport { transformHotkeyInput } from '@midscene/shared/us-keyboard-layout';\nimport { assert } from '@midscene/shared/utils';\nimport type {\n DragAndDropPlanningAction,\n LocatePlanningAction,\n} from '../../shared/planning-action';\nimport type { UiTarsParsedPlanningResponse } from './parser';\n\nconst debug = getDebug('ui-tars-planning');\nconst warnLog = getDebug('ui-tars-planning', { console: true });\n\nexport function transformUiTarsActions(\n parsedPlanningResponse: UiTarsParsedPlanningResponse,\n): PlanningAction[] {\n const transformActions: PlanningAction[] = [];\n const unhandledActions: Array<{ type: string; thought: string }> = [];\n\n parsedPlanningResponse.actions.forEach((action) => {\n const actionType = (action.action_type || '').toLowerCase();\n if (actionType === 'click') {\n assert(action.action_inputs.start_box, 'start_box is required');\n const point = getPoint(action.action_inputs.start_box);\n\n const locate = {\n point,\n prompt: action.thought || '',\n };\n\n transformActions.push({\n type: 'Tap',\n param: {\n locate,\n },\n } satisfies LocatePlanningAction<'Tap'>);\n } else if (actionType === 'left_double') {\n assert(action.action_inputs.start_box, 'start_box is required');\n const point = getPoint(action.action_inputs.start_box);\n\n const locate = {\n point,\n prompt: action.thought || '',\n };\n\n transformActions.push({\n type: 'DoubleClick',\n param: {\n locate,\n },\n thought: action.thought || '',\n } satisfies LocatePlanningAction<'DoubleClick'>);\n } else if (actionType === 'right_single') {\n assert(action.action_inputs.start_box, 'start_box is required');\n const point = getPoint(action.action_inputs.start_box);\n\n const locate = {\n point,\n prompt: action.thought || '',\n };\n\n transformActions.push({\n type: 'RightClick',\n param: {\n locate,\n },\n thought: action.thought || '',\n } satisfies LocatePlanningAction<'RightClick'>);\n } else if (actionType === 'drag') {\n assert(action.action_inputs.start_box, 'start_box is required');\n assert(action.action_inputs.end_box, 'end_box is required');\n const startPoint = getPoint(action.action_inputs.start_box);\n const endPoint = getPoint(action.action_inputs.end_box);\n transformActions.push({\n type: 'DragAndDrop',\n param: {\n from: {\n point: startPoint,\n prompt: action.thought || '',\n },\n to: {\n point: endPoint,\n prompt: action.thought || '',\n },\n },\n thought: action.thought || '',\n } satisfies DragAndDropPlanningAction);\n } else if (actionType === 'type') {\n transformActions.push({\n type: 'Input',\n param: {\n value: action.action_inputs.content,\n },\n thought: action.thought || '',\n });\n } else if (actionType === 'scroll') {\n transformActions.push({\n type: 'Scroll',\n param: {\n direction: action.action_inputs.direction,\n },\n thought: action.thought || '',\n });\n } else if (actionType === 'finished') {\n transformActions.push({\n type: 'Finished',\n param: {},\n thought: action.action_inputs.content || action.thought || '',\n });\n } else if (actionType === 'hotkey') {\n if (!action.action_inputs.key) {\n warnLog('No key found in action: hotkey. Will not perform action.');\n } else {\n const keys = transformHotkeyInput(action.action_inputs.key);\n\n transformActions.push({\n type: 'KeyboardPress',\n param: {\n keyName: keys.join('+'),\n },\n thought: action.thought || '',\n });\n }\n } else if (actionType === 'wait') {\n transformActions.push({\n type: 'Sleep',\n param: {\n timeMs: 1000,\n },\n thought: action.thought || '',\n });\n } else if (actionType) {\n unhandledActions.push({\n type: actionType,\n thought: action.thought || '',\n });\n debug('Unhandled action type:', actionType, 'thought:', action.thought);\n }\n });\n\n if (transformActions.length === 0) {\n throw new Error(\n buildNoUiTarsActionsError(\n parsedPlanningResponse.rawResponse,\n parsedPlanningResponse.actions,\n unhandledActions,\n ),\n );\n }\n\n debug('transformActions', JSON.stringify(transformActions, null, 2));\n return transformActions;\n}\n\nfunction getPoint(startBox: string): [number, number] {\n const [x, y] = JSON.parse(startBox);\n assert(\n typeof x === 'number' &&\n Number.isFinite(x) &&\n typeof y === 'number' &&\n Number.isFinite(y),\n `invalid point data for ui-tars planning: ${startBox}`,\n );\n return [x, y];\n}\n\nfunction buildNoUiTarsActionsError(\n rawResponse: string,\n actions: UiTarsParsedPlanningResponse['actions'],\n unhandledActions: Array<{ type: string; thought: string }>,\n): string {\n const errorDetails: string[] = [];\n\n if (actions.length === 0) {\n errorDetails.push('Action parser returned no actions');\n\n if (rawResponse.includes('Thought:') && !rawResponse.includes('Action:')) {\n errorDetails.push(\n 'Response contains \"Thought:\" but missing \"Action:\" line',\n );\n } else {\n errorDetails.push('Response may be malformed or empty');\n }\n }\n\n if (unhandledActions.length > 0) {\n const types = unhandledActions.map((a) => a.type).join(', ');\n errorDetails.push(`Unhandled action types: ${types}`);\n }\n\n return ['No actions found in UI-TARS response.', ...errorDetails].join('\\n');\n}\n"],"names":["debug","getDebug","warnLog","transformUiTarsActions","parsedPlanningResponse","transformActions","unhandledActions","action","actionType","assert","point","getPoint","locate","startPoint","endPoint","keys","transformHotkeyInput","Error","buildNoUiTarsActionsError","JSON","startBox","x","y","Number","rawResponse","actions","errorDetails","types","a"],"mappings":";;;AAUA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,UAAUD,SAAS,oBAAoB;IAAE,SAAS;AAAK;AAEtD,SAASE,uBACdC,sBAAoD;IAEpD,MAAMC,mBAAqC,EAAE;IAC7C,MAAMC,mBAA6D,EAAE;IAErEF,uBAAuB,OAAO,CAAC,OAAO,CAAC,CAACG;QACtC,MAAMC,aAAcD,AAAAA,CAAAA,OAAO,WAAW,IAAI,EAAC,EAAG,WAAW;QACzD,IAAIC,AAAe,YAAfA,YAAwB;YAC1BC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAMG,QAAQC,SAASJ,OAAO,aAAa,CAAC,SAAS;YAErD,MAAMK,SAAS;gBACbF;gBACA,QAAQH,OAAO,OAAO,IAAI;YAC5B;YAEAF,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACLO;gBACF;YACF;QACF,OAAO,IAAIJ,AAAe,kBAAfA,YAA8B;YACvCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAMG,QAAQC,SAASJ,OAAO,aAAa,CAAC,SAAS;YAErD,MAAMK,SAAS;gBACbF;gBACA,QAAQH,OAAO,OAAO,IAAI;YAC5B;YAEAF,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACLO;gBACF;gBACA,SAASL,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,mBAAfA,YAA+B;YACxCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAMG,QAAQC,SAASJ,OAAO,aAAa,CAAC,SAAS;YAErD,MAAMK,SAAS;gBACbF;gBACA,QAAQH,OAAO,OAAO,IAAI;YAC5B;YAEAF,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACLO;gBACF;gBACA,SAASL,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YAAuB;YAChCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCE,OAAOF,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMM,aAAaF,SAASJ,OAAO,aAAa,CAAC,SAAS;YAC1D,MAAMO,WAAWH,SAASJ,OAAO,aAAa,CAAC,OAAO;YACtDF,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,MAAM;wBACJ,OAAOQ;wBACP,QAAQN,OAAO,OAAO,IAAI;oBAC5B;oBACA,IAAI;wBACF,OAAOO;wBACP,QAAQP,OAAO,OAAO,IAAI;oBAC5B;gBACF;gBACA,SAASA,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YACTH,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,OAAOE,OAAO,aAAa,CAAC,OAAO;YACrC;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,aAAfA,YACTH,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWE,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,eAAfA,YACTH,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO,CAAC;YACR,SAASE,OAAO,aAAa,CAAC,OAAO,IAAIA,OAAO,OAAO,IAAI;QAC7D;aACK,IAAIC,AAAe,aAAfA,YACT,IAAKD,OAAO,aAAa,CAAC,GAAG,EAEtB;YACL,MAAMQ,OAAOC,qBAAqBT,OAAO,aAAa,CAAC,GAAG;YAE1DF,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,SAASU,KAAK,IAAI,CAAC;gBACrB;gBACA,SAASR,OAAO,OAAO,IAAI;YAC7B;QACF,OAXEL,QAAQ;aAYL,IAAIM,AAAe,WAAfA,YACTH,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,SAASE,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,YAAY;YACrBF,iBAAiB,IAAI,CAAC;gBACpB,MAAME;gBACN,SAASD,OAAO,OAAO,IAAI;YAC7B;YACAP,MAAM,0BAA0BQ,YAAY,YAAYD,OAAO,OAAO;QACxE;IACF;IAEA,IAAIF,AAA4B,MAA5BA,iBAAiB,MAAM,EACzB,MAAM,IAAIY,MACRC,0BACEd,uBAAuB,WAAW,EAClCA,uBAAuB,OAAO,EAC9BE;IAKNN,MAAM,oBAAoBmB,KAAK,SAAS,CAACd,kBAAkB,MAAM;IACjE,OAAOA;AACT;AAEA,SAASM,SAASS,QAAgB;IAChC,MAAM,CAACC,GAAGC,EAAE,GAAGH,KAAK,KAAK,CAACC;IAC1BX,OACE,AAAa,YAAb,OAAOY,KACLE,OAAO,QAAQ,CAACF,MAChB,AAAa,YAAb,OAAOC,KACPC,OAAO,QAAQ,CAACD,IAClB,CAAC,yCAAyC,EAAEF,UAAU;IAExD,OAAO;QAACC;QAAGC;KAAE;AACf;AAEA,SAASJ,0BACPM,WAAmB,EACnBC,OAAgD,EAChDnB,gBAA0D;IAE1D,MAAMoB,eAAyB,EAAE;IAEjC,IAAID,AAAmB,MAAnBA,QAAQ,MAAM,EAAQ;QACxBC,aAAa,IAAI,CAAC;QAElB,IAAIF,YAAY,QAAQ,CAAC,eAAe,CAACA,YAAY,QAAQ,CAAC,YAC5DE,aAAa,IAAI,CACf;aAGFA,aAAa,IAAI,CAAC;IAEtB;IAEA,IAAIpB,iBAAiB,MAAM,GAAG,GAAG;QAC/B,MAAMqB,QAAQrB,iBAAiB,GAAG,CAAC,CAACsB,IAAMA,EAAE,IAAI,EAAE,IAAI,CAAC;QACvDF,aAAa,IAAI,CAAC,CAAC,wBAAwB,EAAEC,OAAO;IACtD;IAEA,OAAO;QAAC;WAA4CD;KAAa,CAAC,IAAI,CAAC;AACzE"}
@@ -3,7 +3,7 @@ import { assert } from "@midscene/shared/utils";
3
3
  import { jsonrepair } from "jsonrepair";
4
4
  import { extractJSONFromCodeBlock, safeParseJson } from "../../service-caller/json.mjs";
5
5
  import { unwrapCoordinateListLikeInput } from "../../shared/model-locate-result/index.mjs";
6
- import { uiTarsPlanning } from "./planning.mjs";
6
+ import { createUiTarsPlanner } from "./planning.mjs";
7
7
  const defaultVlmUiTarsReplanningCycleLimit = 40;
8
8
  function normalizeJsonObject(obj, context = {}) {
9
9
  if (null == obj) return obj;
@@ -122,7 +122,7 @@ function createUiTarsAdapter(uiTarsModelVersion) {
122
122
  kind: 'custom',
123
123
  cacheEnabled: false,
124
124
  defaultReplanningCycleLimit: defaultVlmUiTarsReplanningCycleLimit,
125
- planFn: (userInstruction, options)=>uiTarsPlanning(userInstruction, options, uiTarsModelVersion)
125
+ planner: createUiTarsPlanner(uiTarsModelVersion)
126
126
  },
127
127
  locate: {
128
128
  resultAdapter: {
@@ -1 +1 @@
1
- {"version":3,"file":"ai-model/models/ui-tars/adapter.mjs","sources":["../../../../../src/ai-model/models/ui-tars/adapter.ts"],"sourcesContent":["import { type TModelFamily, UITarsModelVersion } from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\nimport { jsonrepair } from 'jsonrepair';\nimport {\n extractJSONFromCodeBlock,\n safeParseJson,\n} from '../../service-caller/json';\nimport {\n type LocateResultValue,\n unwrapCoordinateListLikeInput,\n} from '../../shared/model-locate-result';\nimport type {\n JsonParserContext,\n JsonParserSource,\n ModelAdapterDefinition,\n} from '../types';\nimport { uiTarsPlanning } from './planning';\n\nconst defaultVlmUiTarsReplanningCycleLimit = 40;\n\nfunction normalizeJsonObject(\n obj: any,\n context: Pick<JsonParserContext, 'preserveStringValueKeys'> = {},\n): any {\n if (obj === null || obj === undefined) {\n return obj;\n }\n\n if (Array.isArray(obj)) {\n return obj.map((item) => normalizeJsonObject(item, context));\n }\n\n if (typeof obj === 'object') {\n const normalized: any = {};\n for (const [key, value] of Object.entries(obj)) {\n const trimmedKey = key.trim();\n const preserveStringValue =\n context.preserveStringValueKeys?.includes(trimmedKey) ?? false;\n const normalizedValue =\n typeof value === 'string'\n ? preserveStringValue\n ? value\n : value.trim()\n : normalizeJsonObject(value, context);\n normalized[trimmedKey] = normalizedValue;\n }\n return normalized;\n }\n\n return typeof obj === 'string' ? obj.trim() : obj;\n}\n\nfunction shouldRepairUiTarsLocateJson(source: JsonParserSource) {\n return (\n source === 'locate' ||\n source === 'section-locator' ||\n source === 'planning-action-param'\n );\n}\n\nfunction preprocessUiTarsLocateJson(input: string) {\n if (input.includes('bbox')) {\n while (/\\d+\\s+\\d+/.test(input)) {\n input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n }\n }\n return input;\n}\n\nconst uiTarsJsonParser: ModelAdapterDefinition['jsonParser'] = (\n raw,\n context = { source: 'generic-object' },\n) => {\n const { source } = context;\n try {\n return safeParseJson(raw, context);\n } catch (firstError) {\n if (!shouldRepairUiTarsLocateJson(source)) {\n throw firstError;\n }\n\n const jsonString = preprocessUiTarsLocateJson(\n extractJSONFromCodeBlock(raw),\n );\n try {\n return normalizeJsonObject(JSON.parse(jsonrepair(jsonString)), context);\n } catch (error) {\n throw Error(\n `failed to parse LLM response into JSON. Error - ${String(\n error ?? firstError ?? 'unknown error',\n )}. Response - \\n ${raw}`,\n );\n }\n }\n};\n\n// UI-TARS has not received active updates for a long time, so this parser is\n// intentionally kept separate from Doubao even though the current logic is the\n// same. This avoids coupling UI-TARS behavior to future Doubao adapter changes.\nfunction parseUiTarsRawLocateValue(input: unknown): LocateResultValue {\n const bbox = unwrapCoordinateListLikeInput(input as any);\n if (typeof bbox === 'string') {\n assert(\n /^(\\d+)\\s(\\d+)\\s(\\d+)\\s(\\d+)$/.test(bbox.trim()),\n `invalid bbox data string for ui-tars mode: ${bbox}`,\n );\n const splitted = bbox.split(' ');\n if (splitted.length === 4) {\n return {\n type: 'bbox',\n coordinates: [\n Number(splitted[0]),\n Number(splitted[1]),\n Number(splitted[2]),\n Number(splitted[3]),\n ],\n };\n }\n throw new Error(`invalid bbox data string for ui-tars mode: ${bbox}`);\n }\n\n let bboxList: number[] = [];\n if (Array.isArray(bbox) && typeof bbox[0] === 'string') {\n bbox.forEach((item) => {\n if (typeof item === 'string' && item.includes(',')) {\n const [x, y] = item.split(',');\n bboxList.push(Number(x.trim()), Number(y.trim()));\n } else if (typeof item === 'string' && item.includes(' ')) {\n const [x, y] = item.split(' ');\n bboxList.push(Number(x.trim()), Number(y.trim()));\n } else {\n bboxList.push(Number(item));\n }\n });\n } else {\n bboxList = bbox as number[];\n }\n\n if (bboxList.length === 4 || bboxList.length === 5) {\n return {\n type: 'bbox',\n coordinates: [bboxList[0], bboxList[1], bboxList[2], bboxList[3]],\n };\n }\n\n if (\n bboxList.length === 6 ||\n bboxList.length === 2 ||\n bboxList.length === 3 ||\n bboxList.length === 7\n ) {\n return { type: 'point', coordinates: [bboxList[0], bboxList[1]] };\n }\n\n if (bbox.length === 8) {\n return {\n type: 'bbox',\n coordinates: [bboxList[0], bboxList[1], bboxList[4], bboxList[5]],\n };\n }\n\n const msg = `invalid bbox data for ui-tars mode: ${JSON.stringify(bbox)} `;\n throw new Error(msg);\n}\n\nfunction createUiTarsAdapter(\n uiTarsModelVersion: UITarsModelVersion,\n): ModelAdapterDefinition {\n return {\n jsonParser: uiTarsJsonParser,\n chatCompletion: {\n unsupportedUserConfig: [\n 'reasoningEnabled',\n 'reasoningEffort',\n 'reasoningBudget',\n ],\n buildChatCompletionParams: ({ midsceneDefaults, userConfig }) => {\n const commonOverrideConfig: Record<string, unknown> = {};\n\n if (userConfig.temperature !== undefined) {\n commonOverrideConfig.temperature = userConfig.temperature;\n }\n\n return {\n config: {\n ...midsceneDefaults,\n ...commonOverrideConfig,\n },\n };\n },\n },\n planning: {\n kind: 'custom',\n cacheEnabled: false,\n defaultReplanningCycleLimit: defaultVlmUiTarsReplanningCycleLimit,\n planFn: (userInstruction, options) =>\n uiTarsPlanning(userInstruction, options, uiTarsModelVersion),\n },\n locate: {\n resultAdapter: {\n coordinates: { shape: 'bbox', order: 'xy', normalizedBy: 1000 },\n parseRawLocateValue: parseUiTarsRawLocateValue,\n },\n },\n };\n}\n\nconst uiTarsDoubao15Adapter = createUiTarsAdapter(\n UITarsModelVersion.DOUBAO_1_5_20B,\n);\n\nexport const uiTarsAdapters = {\n 'vlm-ui-tars': createUiTarsAdapter(UITarsModelVersion.V1_0),\n 'vlm-ui-tars-doubao': uiTarsDoubao15Adapter,\n 'vlm-ui-tars-doubao-1.5': uiTarsDoubao15Adapter,\n} satisfies Pick<\n Record<TModelFamily, ModelAdapterDefinition>,\n 'vlm-ui-tars' | 'vlm-ui-tars-doubao' | 'vlm-ui-tars-doubao-1.5'\n>;\n"],"names":["defaultVlmUiTarsReplanningCycleLimit","normalizeJsonObject","obj","context","Array","item","normalized","key","value","Object","trimmedKey","preserveStringValue","normalizedValue","shouldRepairUiTarsLocateJson","source","preprocessUiTarsLocateJson","input","uiTarsJsonParser","raw","safeParseJson","firstError","jsonString","extractJSONFromCodeBlock","JSON","jsonrepair","error","Error","String","parseUiTarsRawLocateValue","bbox","unwrapCoordinateListLikeInput","assert","splitted","Number","bboxList","x","y","msg","createUiTarsAdapter","uiTarsModelVersion","midsceneDefaults","userConfig","commonOverrideConfig","undefined","userInstruction","options","uiTarsPlanning","uiTarsDoubao15Adapter","UITarsModelVersion","uiTarsAdapters"],"mappings":";;;;;;AAkBA,MAAMA,uCAAuC;AAE7C,SAASC,oBACPC,GAAQ,EACRC,UAA8D,CAAC,CAAC;IAEhE,IAAID,QAAAA,KACF,OAAOA;IAGT,IAAIE,MAAM,OAAO,CAACF,MAChB,OAAOA,IAAI,GAAG,CAAC,CAACG,OAASJ,oBAAoBI,MAAMF;IAGrD,IAAI,AAAe,YAAf,OAAOD,KAAkB;QAC3B,MAAMI,aAAkB,CAAC;QACzB,KAAK,MAAM,CAACC,KAAKC,MAAM,IAAIC,OAAO,OAAO,CAACP,KAAM;YAC9C,MAAMQ,aAAaH,IAAI,IAAI;YAC3B,MAAMI,sBACJR,QAAQ,uBAAuB,EAAE,SAASO,eAAe;YAC3D,MAAME,kBACJ,AAAiB,YAAjB,OAAOJ,QACHG,sBACEH,QACAA,MAAM,IAAI,KACZP,oBAAoBO,OAAOL;YACjCG,UAAU,CAACI,WAAW,GAAGE;QAC3B;QACA,OAAON;IACT;IAEA,OAAO,AAAe,YAAf,OAAOJ,MAAmBA,IAAI,IAAI,KAAKA;AAChD;AAEA,SAASW,6BAA6BC,MAAwB;IAC5D,OACEA,AAAW,aAAXA,UACAA,AAAW,sBAAXA,UACAA,AAAW,4BAAXA;AAEJ;AAEA,SAASC,2BAA2BC,KAAa;IAC/C,IAAIA,MAAM,QAAQ,CAAC,SACjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;IAG5C,OAAOA;AACT;AAEA,MAAMC,mBAAyD,CAC7DC,KACAf,UAAU;IAAE,QAAQ;AAAiB,CAAC;IAEtC,MAAM,EAAEW,MAAM,EAAE,GAAGX;IACnB,IAAI;QACF,OAAOgB,cAAcD,KAAKf;IAC5B,EAAE,OAAOiB,YAAY;QACnB,IAAI,CAACP,6BAA6BC,SAChC,MAAMM;QAGR,MAAMC,aAAaN,2BACjBO,yBAAyBJ;QAE3B,IAAI;YACF,OAAOjB,oBAAoBsB,KAAK,KAAK,CAACC,WAAWH,cAAclB;QACjE,EAAE,OAAOsB,OAAO;YACd,MAAMC,MACJ,CAAC,gDAAgD,EAAEC,OACjDF,SAASL,cAAc,iBACvB,gBAAgB,EAAEF,KAAK;QAE7B;IACF;AACF;AAKA,SAASU,0BAA0BZ,KAAc;IAC/C,MAAMa,OAAOC,8BAA8Bd;IAC3C,IAAI,AAAgB,YAAhB,OAAOa,MAAmB;QAC5BE,OACE,+BAA+B,IAAI,CAACF,KAAK,IAAI,KAC7C,CAAC,2CAA2C,EAAEA,MAAM;QAEtD,MAAMG,WAAWH,KAAK,KAAK,CAAC;QAC5B,IAAIG,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,MAAM;YACN,aAAa;gBACXC,OAAOD,QAAQ,CAAC,EAAE;gBAClBC,OAAOD,QAAQ,CAAC,EAAE;gBAClBC,OAAOD,QAAQ,CAAC,EAAE;gBAClBC,OAAOD,QAAQ,CAAC,EAAE;aACnB;QACH;QAEF,MAAM,IAAIN,MAAM,CAAC,2CAA2C,EAAEG,MAAM;IACtE;IAEA,IAAIK,WAAqB,EAAE;IAC3B,IAAI9B,MAAM,OAAO,CAACyB,SAAS,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EACvCA,KAAK,OAAO,CAAC,CAACxB;QACZ,IAAI,AAAgB,YAAhB,OAAOA,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YAClD,MAAM,CAAC8B,GAAGC,EAAE,GAAG/B,KAAK,KAAK,CAAC;YAC1B6B,SAAS,IAAI,CAACD,OAAOE,EAAE,IAAI,KAAKF,OAAOG,EAAE,IAAI;QAC/C,OAAO,IAAI,AAAgB,YAAhB,OAAO/B,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YACzD,MAAM,CAAC8B,GAAGC,EAAE,GAAG/B,KAAK,KAAK,CAAC;YAC1B6B,SAAS,IAAI,CAACD,OAAOE,EAAE,IAAI,KAAKF,OAAOG,EAAE,IAAI;QAC/C,OACEF,SAAS,IAAI,CAACD,OAAO5B;IAEzB;SAEA6B,WAAWL;IAGb,IAAIK,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAO;QACL,MAAM;QACN,aAAa;YAACA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;SAAC;IACnE;IAGF,IACEA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAO;QAAE,MAAM;QAAS,aAAa;YAACA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;SAAC;IAAC;IAGlE,IAAIL,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAO;QACL,MAAM;QACN,aAAa;YAACK,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;SAAC;IACnE;IAGF,MAAMG,MAAM,CAAC,oCAAoC,EAAEd,KAAK,SAAS,CAACM,MAAM,CAAC,CAAC;IAC1E,MAAM,IAAIH,MAAMW;AAClB;AAEA,SAASC,oBACPC,kBAAsC;IAEtC,OAAO;QACL,YAAYtB;QACZ,gBAAgB;YACd,uBAAuB;gBACrB;gBACA;gBACA;aACD;YACD,2BAA2B,CAAC,EAAEuB,gBAAgB,EAAEC,UAAU,EAAE;gBAC1D,MAAMC,uBAAgD,CAAC;gBAEvD,IAAID,AAA2BE,WAA3BF,WAAW,WAAW,EACxBC,qBAAqB,WAAW,GAAGD,WAAW,WAAW;gBAG3D,OAAO;oBACL,QAAQ;wBACN,GAAGD,gBAAgB;wBACnB,GAAGE,oBAAoB;oBACzB;gBACF;YACF;QACF;QACA,UAAU;YACR,MAAM;YACN,cAAc;YACd,6BAA6B1C;YAC7B,QAAQ,CAAC4C,iBAAiBC,UACxBC,eAAeF,iBAAiBC,SAASN;QAC7C;QACA,QAAQ;YACN,eAAe;gBACb,aAAa;oBAAE,OAAO;oBAAQ,OAAO;oBAAM,cAAc;gBAAK;gBAC9D,qBAAqBX;YACvB;QACF;IACF;AACF;AAEA,MAAMmB,wBAAwBT,oBAC5BU,mBAAmB,cAAc;AAG5B,MAAMC,iBAAiB;IAC5B,eAAeX,oBAAoBU,mBAAmB,IAAI;IAC1D,sBAAsBD;IACtB,0BAA0BA;AAC5B"}
1
+ {"version":3,"file":"ai-model/models/ui-tars/adapter.mjs","sources":["../../../../../src/ai-model/models/ui-tars/adapter.ts"],"sourcesContent":["import { type TModelFamily, UITarsModelVersion } from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\nimport { jsonrepair } from 'jsonrepair';\nimport type {\n JsonParserContext,\n JsonParserSource,\n ModelAdapterDefinition,\n} from '../../model-adapter/types';\nimport {\n extractJSONFromCodeBlock,\n safeParseJson,\n} from '../../service-caller/json';\nimport {\n type LocateResultValue,\n unwrapCoordinateListLikeInput,\n} from '../../shared/model-locate-result';\nimport { createUiTarsPlanner } from './planning';\n\nconst defaultVlmUiTarsReplanningCycleLimit = 40;\n\nfunction normalizeJsonObject(\n obj: any,\n context: Pick<JsonParserContext, 'preserveStringValueKeys'> = {},\n): any {\n if (obj === null || obj === undefined) {\n return obj;\n }\n\n if (Array.isArray(obj)) {\n return obj.map((item) => normalizeJsonObject(item, context));\n }\n\n if (typeof obj === 'object') {\n const normalized: any = {};\n for (const [key, value] of Object.entries(obj)) {\n const trimmedKey = key.trim();\n const preserveStringValue =\n context.preserveStringValueKeys?.includes(trimmedKey) ?? false;\n const normalizedValue =\n typeof value === 'string'\n ? preserveStringValue\n ? value\n : value.trim()\n : normalizeJsonObject(value, context);\n normalized[trimmedKey] = normalizedValue;\n }\n return normalized;\n }\n\n return typeof obj === 'string' ? obj.trim() : obj;\n}\n\nfunction shouldRepairUiTarsLocateJson(source: JsonParserSource) {\n return (\n source === 'locate' ||\n source === 'section-locator' ||\n source === 'planning-action-param'\n );\n}\n\nfunction preprocessUiTarsLocateJson(input: string) {\n if (input.includes('bbox')) {\n while (/\\d+\\s+\\d+/.test(input)) {\n input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n }\n }\n return input;\n}\n\nconst uiTarsJsonParser: ModelAdapterDefinition['jsonParser'] = (\n raw,\n context = { source: 'generic-object' },\n) => {\n const { source } = context;\n try {\n return safeParseJson(raw, context);\n } catch (firstError) {\n if (!shouldRepairUiTarsLocateJson(source)) {\n throw firstError;\n }\n\n const jsonString = preprocessUiTarsLocateJson(\n extractJSONFromCodeBlock(raw),\n );\n try {\n return normalizeJsonObject(JSON.parse(jsonrepair(jsonString)), context);\n } catch (error) {\n throw Error(\n `failed to parse LLM response into JSON. Error - ${String(\n error ?? firstError ?? 'unknown error',\n )}. Response - \\n ${raw}`,\n );\n }\n }\n};\n\n// UI-TARS has not received active updates for a long time, so this parser is\n// intentionally kept separate from Doubao even though the current logic is the\n// same. This avoids coupling UI-TARS behavior to future Doubao adapter changes.\nfunction parseUiTarsRawLocateValue(input: unknown): LocateResultValue {\n const bbox = unwrapCoordinateListLikeInput(input as any);\n if (typeof bbox === 'string') {\n assert(\n /^(\\d+)\\s(\\d+)\\s(\\d+)\\s(\\d+)$/.test(bbox.trim()),\n `invalid bbox data string for ui-tars mode: ${bbox}`,\n );\n const splitted = bbox.split(' ');\n if (splitted.length === 4) {\n return {\n type: 'bbox',\n coordinates: [\n Number(splitted[0]),\n Number(splitted[1]),\n Number(splitted[2]),\n Number(splitted[3]),\n ],\n };\n }\n throw new Error(`invalid bbox data string for ui-tars mode: ${bbox}`);\n }\n\n let bboxList: number[] = [];\n if (Array.isArray(bbox) && typeof bbox[0] === 'string') {\n bbox.forEach((item) => {\n if (typeof item === 'string' && item.includes(',')) {\n const [x, y] = item.split(',');\n bboxList.push(Number(x.trim()), Number(y.trim()));\n } else if (typeof item === 'string' && item.includes(' ')) {\n const [x, y] = item.split(' ');\n bboxList.push(Number(x.trim()), Number(y.trim()));\n } else {\n bboxList.push(Number(item));\n }\n });\n } else {\n bboxList = bbox as number[];\n }\n\n if (bboxList.length === 4 || bboxList.length === 5) {\n return {\n type: 'bbox',\n coordinates: [bboxList[0], bboxList[1], bboxList[2], bboxList[3]],\n };\n }\n\n if (\n bboxList.length === 6 ||\n bboxList.length === 2 ||\n bboxList.length === 3 ||\n bboxList.length === 7\n ) {\n return { type: 'point', coordinates: [bboxList[0], bboxList[1]] };\n }\n\n if (bbox.length === 8) {\n return {\n type: 'bbox',\n coordinates: [bboxList[0], bboxList[1], bboxList[4], bboxList[5]],\n };\n }\n\n const msg = `invalid bbox data for ui-tars mode: ${JSON.stringify(bbox)} `;\n throw new Error(msg);\n}\n\nfunction createUiTarsAdapter(\n uiTarsModelVersion: UITarsModelVersion,\n): ModelAdapterDefinition {\n return {\n jsonParser: uiTarsJsonParser,\n chatCompletion: {\n unsupportedUserConfig: [\n 'reasoningEnabled',\n 'reasoningEffort',\n 'reasoningBudget',\n ],\n buildChatCompletionParams: ({ midsceneDefaults, userConfig }) => {\n const commonOverrideConfig: Record<string, unknown> = {};\n\n if (userConfig.temperature !== undefined) {\n commonOverrideConfig.temperature = userConfig.temperature;\n }\n\n return {\n config: {\n ...midsceneDefaults,\n ...commonOverrideConfig,\n },\n };\n },\n },\n planning: {\n kind: 'custom',\n cacheEnabled: false,\n defaultReplanningCycleLimit: defaultVlmUiTarsReplanningCycleLimit,\n planner: createUiTarsPlanner(uiTarsModelVersion),\n },\n locate: {\n resultAdapter: {\n coordinates: { shape: 'bbox', order: 'xy', normalizedBy: 1000 },\n parseRawLocateValue: parseUiTarsRawLocateValue,\n },\n },\n };\n}\n\nconst uiTarsDoubao15Adapter = createUiTarsAdapter(\n UITarsModelVersion.DOUBAO_1_5_20B,\n);\n\nexport const uiTarsAdapters = {\n 'vlm-ui-tars': createUiTarsAdapter(UITarsModelVersion.V1_0),\n 'vlm-ui-tars-doubao': uiTarsDoubao15Adapter,\n 'vlm-ui-tars-doubao-1.5': uiTarsDoubao15Adapter,\n} satisfies Pick<\n Record<TModelFamily, ModelAdapterDefinition>,\n 'vlm-ui-tars' | 'vlm-ui-tars-doubao' | 'vlm-ui-tars-doubao-1.5'\n>;\n"],"names":["defaultVlmUiTarsReplanningCycleLimit","normalizeJsonObject","obj","context","Array","item","normalized","key","value","Object","trimmedKey","preserveStringValue","normalizedValue","shouldRepairUiTarsLocateJson","source","preprocessUiTarsLocateJson","input","uiTarsJsonParser","raw","safeParseJson","firstError","jsonString","extractJSONFromCodeBlock","JSON","jsonrepair","error","Error","String","parseUiTarsRawLocateValue","bbox","unwrapCoordinateListLikeInput","assert","splitted","Number","bboxList","x","y","msg","createUiTarsAdapter","uiTarsModelVersion","midsceneDefaults","userConfig","commonOverrideConfig","undefined","createUiTarsPlanner","uiTarsDoubao15Adapter","UITarsModelVersion","uiTarsAdapters"],"mappings":";;;;;;AAkBA,MAAMA,uCAAuC;AAE7C,SAASC,oBACPC,GAAQ,EACRC,UAA8D,CAAC,CAAC;IAEhE,IAAID,QAAAA,KACF,OAAOA;IAGT,IAAIE,MAAM,OAAO,CAACF,MAChB,OAAOA,IAAI,GAAG,CAAC,CAACG,OAASJ,oBAAoBI,MAAMF;IAGrD,IAAI,AAAe,YAAf,OAAOD,KAAkB;QAC3B,MAAMI,aAAkB,CAAC;QACzB,KAAK,MAAM,CAACC,KAAKC,MAAM,IAAIC,OAAO,OAAO,CAACP,KAAM;YAC9C,MAAMQ,aAAaH,IAAI,IAAI;YAC3B,MAAMI,sBACJR,QAAQ,uBAAuB,EAAE,SAASO,eAAe;YAC3D,MAAME,kBACJ,AAAiB,YAAjB,OAAOJ,QACHG,sBACEH,QACAA,MAAM,IAAI,KACZP,oBAAoBO,OAAOL;YACjCG,UAAU,CAACI,WAAW,GAAGE;QAC3B;QACA,OAAON;IACT;IAEA,OAAO,AAAe,YAAf,OAAOJ,MAAmBA,IAAI,IAAI,KAAKA;AAChD;AAEA,SAASW,6BAA6BC,MAAwB;IAC5D,OACEA,AAAW,aAAXA,UACAA,AAAW,sBAAXA,UACAA,AAAW,4BAAXA;AAEJ;AAEA,SAASC,2BAA2BC,KAAa;IAC/C,IAAIA,MAAM,QAAQ,CAAC,SACjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;IAG5C,OAAOA;AACT;AAEA,MAAMC,mBAAyD,CAC7DC,KACAf,UAAU;IAAE,QAAQ;AAAiB,CAAC;IAEtC,MAAM,EAAEW,MAAM,EAAE,GAAGX;IACnB,IAAI;QACF,OAAOgB,cAAcD,KAAKf;IAC5B,EAAE,OAAOiB,YAAY;QACnB,IAAI,CAACP,6BAA6BC,SAChC,MAAMM;QAGR,MAAMC,aAAaN,2BACjBO,yBAAyBJ;QAE3B,IAAI;YACF,OAAOjB,oBAAoBsB,KAAK,KAAK,CAACC,WAAWH,cAAclB;QACjE,EAAE,OAAOsB,OAAO;YACd,MAAMC,MACJ,CAAC,gDAAgD,EAAEC,OACjDF,SAASL,cAAc,iBACvB,gBAAgB,EAAEF,KAAK;QAE7B;IACF;AACF;AAKA,SAASU,0BAA0BZ,KAAc;IAC/C,MAAMa,OAAOC,8BAA8Bd;IAC3C,IAAI,AAAgB,YAAhB,OAAOa,MAAmB;QAC5BE,OACE,+BAA+B,IAAI,CAACF,KAAK,IAAI,KAC7C,CAAC,2CAA2C,EAAEA,MAAM;QAEtD,MAAMG,WAAWH,KAAK,KAAK,CAAC;QAC5B,IAAIG,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,MAAM;YACN,aAAa;gBACXC,OAAOD,QAAQ,CAAC,EAAE;gBAClBC,OAAOD,QAAQ,CAAC,EAAE;gBAClBC,OAAOD,QAAQ,CAAC,EAAE;gBAClBC,OAAOD,QAAQ,CAAC,EAAE;aACnB;QACH;QAEF,MAAM,IAAIN,MAAM,CAAC,2CAA2C,EAAEG,MAAM;IACtE;IAEA,IAAIK,WAAqB,EAAE;IAC3B,IAAI9B,MAAM,OAAO,CAACyB,SAAS,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EACvCA,KAAK,OAAO,CAAC,CAACxB;QACZ,IAAI,AAAgB,YAAhB,OAAOA,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YAClD,MAAM,CAAC8B,GAAGC,EAAE,GAAG/B,KAAK,KAAK,CAAC;YAC1B6B,SAAS,IAAI,CAACD,OAAOE,EAAE,IAAI,KAAKF,OAAOG,EAAE,IAAI;QAC/C,OAAO,IAAI,AAAgB,YAAhB,OAAO/B,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YACzD,MAAM,CAAC8B,GAAGC,EAAE,GAAG/B,KAAK,KAAK,CAAC;YAC1B6B,SAAS,IAAI,CAACD,OAAOE,EAAE,IAAI,KAAKF,OAAOG,EAAE,IAAI;QAC/C,OACEF,SAAS,IAAI,CAACD,OAAO5B;IAEzB;SAEA6B,WAAWL;IAGb,IAAIK,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAO;QACL,MAAM;QACN,aAAa;YAACA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;SAAC;IACnE;IAGF,IACEA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAO;QAAE,MAAM;QAAS,aAAa;YAACA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;SAAC;IAAC;IAGlE,IAAIL,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAO;QACL,MAAM;QACN,aAAa;YAACK,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;SAAC;IACnE;IAGF,MAAMG,MAAM,CAAC,oCAAoC,EAAEd,KAAK,SAAS,CAACM,MAAM,CAAC,CAAC;IAC1E,MAAM,IAAIH,MAAMW;AAClB;AAEA,SAASC,oBACPC,kBAAsC;IAEtC,OAAO;QACL,YAAYtB;QACZ,gBAAgB;YACd,uBAAuB;gBACrB;gBACA;gBACA;aACD;YACD,2BAA2B,CAAC,EAAEuB,gBAAgB,EAAEC,UAAU,EAAE;gBAC1D,MAAMC,uBAAgD,CAAC;gBAEvD,IAAID,AAA2BE,WAA3BF,WAAW,WAAW,EACxBC,qBAAqB,WAAW,GAAGD,WAAW,WAAW;gBAG3D,OAAO;oBACL,QAAQ;wBACN,GAAGD,gBAAgB;wBACnB,GAAGE,oBAAoB;oBACzB;gBACF;YACF;QACF;QACA,UAAU;YACR,MAAM;YACN,cAAc;YACd,6BAA6B1C;YAC7B,SAAS4C,oBAAoBL;QAC/B;QACA,QAAQ;YACN,eAAe;gBACb,aAAa;oBAAE,OAAO;oBAAQ,OAAO;oBAAM,cAAc;gBAAK;gBAC9D,qBAAqBX;YACvB;QACF;IACF;AACF;AAEA,MAAMiB,wBAAwBP,oBAC5BQ,mBAAmB,cAAc;AAG5B,MAAMC,iBAAiB;IAC5B,eAAeT,oBAAoBQ,mBAAmB,IAAI;IAC1D,sBAAsBD;IACtB,0BAA0BA;AAC5B"}
@@ -0,0 +1,40 @@
1
+ import { getDebug } from "@midscene/shared/logger";
2
+ import { actionParser } from "@ui-tars/action-parser";
3
+ const debug = getDebug('ui-tars-planning');
4
+ function parseUiTarsPlanningResponse(rawResponse, shotSize, uiTarsModelVersion) {
5
+ const convertedText = convertBboxToCoordinates(rawResponse);
6
+ const parseResult = actionParser({
7
+ prediction: convertedText,
8
+ factor: [
9
+ 1000,
10
+ 1000
11
+ ],
12
+ screenContext: {
13
+ width: shotSize.width,
14
+ height: shotSize.height
15
+ },
16
+ modelVer: uiTarsModelVersion
17
+ });
18
+ debug('ui-tars modelVer', uiTarsModelVersion, ', parsed', JSON.stringify(parseResult.parsed));
19
+ return {
20
+ rawResponse,
21
+ actions: parseResult.parsed
22
+ };
23
+ }
24
+ function convertBboxToCoordinates(text) {
25
+ const pattern = /<bbox>(\d+)\s+(\d+)\s+(\d+)\s+(\d+)<\/bbox>/g;
26
+ function replaceMatch(match, x1, y1, x2, y2) {
27
+ const x1Num = Number.parseInt(x1, 10);
28
+ const y1Num = Number.parseInt(y1, 10);
29
+ const x2Num = Number.parseInt(x2, 10);
30
+ const y2Num = Number.parseInt(y2, 10);
31
+ const x = Math.floor((x1Num + x2Num) / 2);
32
+ const y = Math.floor((y1Num + y2Num) / 2);
33
+ return `(${x},${y})`;
34
+ }
35
+ const cleanedText = text.replace(/\[EOS\]/g, '').replace(/```(?:[a-zA-Z0-9_-]+)?/g, '');
36
+ return cleanedText.replace(pattern, replaceMatch).trim();
37
+ }
38
+ export { parseUiTarsPlanningResponse };
39
+
40
+ //# sourceMappingURL=parser.mjs.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"ai-model/models/ui-tars/parser.mjs","sources":["../../../../../src/ai-model/models/ui-tars/parser.ts"],"sourcesContent":["import type { Size } from '@/types';\nimport type { UITarsModelVersion } from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { actionParser } from '@ui-tars/action-parser';\n\ntype ActionType =\n | 'click'\n | 'left_double'\n | 'right_single'\n | 'drag'\n | 'type'\n | 'hotkey'\n | 'finished'\n | 'scroll'\n | 'wait';\n\nconst debug = getDebug('ui-tars-planning');\n\nexport interface UiTarsParsedPlanningResponse {\n rawResponse: string;\n actions: ReturnType<typeof actionParser>['parsed'];\n}\n\nexport function parseUiTarsPlanningResponse(\n rawResponse: string,\n shotSize: Size,\n uiTarsModelVersion: UITarsModelVersion,\n): UiTarsParsedPlanningResponse {\n const convertedText = convertBboxToCoordinates(rawResponse);\n const parseResult = actionParser({\n prediction: convertedText,\n factor: [1000, 1000],\n screenContext: {\n width: shotSize.width,\n height: shotSize.height,\n },\n modelVer: uiTarsModelVersion,\n });\n\n debug(\n 'ui-tars modelVer',\n uiTarsModelVersion,\n ', parsed',\n JSON.stringify(parseResult.parsed),\n );\n\n return {\n rawResponse,\n actions: parseResult.parsed,\n };\n}\n\n/**\n * Converts bounding box notation to coordinate points.\n */\nfunction convertBboxToCoordinates(text: string): string {\n const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n function replaceMatch(\n match: string,\n x1: string,\n y1: string,\n x2: string,\n y2: string,\n ): string {\n const x1Num = Number.parseInt(x1, 10);\n const y1Num = Number.parseInt(y1, 10);\n const x2Num = Number.parseInt(x2, 10);\n const y2Num = Number.parseInt(y2, 10);\n\n const x = Math.floor((x1Num + x2Num) / 2);\n const y = Math.floor((y1Num + y2Num) / 2);\n\n return `(${x},${y})`;\n }\n\n const cleanedText = text\n .replace(/\\[EOS\\]/g, '')\n .replace(/```(?:[a-zA-Z0-9_-]+)?/g, '');\n return cleanedText.replace(pattern, replaceMatch).trim();\n}\n\ninterface BaseAction {\n action_type: ActionType;\n action_inputs: Record<string, any>;\n reflection: string | null;\n thought: string | null;\n}\n\ninterface ClickAction extends BaseAction {\n action_type: 'click';\n action_inputs: {\n start_box: string;\n };\n}\n\ninterface DragAction extends BaseAction {\n action_type: 'drag';\n action_inputs: {\n start_box: string;\n end_box: string;\n };\n}\n\ninterface WaitAction extends BaseAction {\n action_type: 'wait';\n action_inputs: {\n time: string;\n };\n}\n\ninterface LeftDoubleAction extends BaseAction {\n action_type: 'left_double';\n action_inputs: {\n start_box: string;\n };\n}\n\ninterface RightSingleAction extends BaseAction {\n action_type: 'right_single';\n action_inputs: {\n start_box: string;\n };\n}\n\ninterface TypeAction extends BaseAction {\n action_type: 'type';\n action_inputs: {\n content: string;\n };\n}\n\ninterface HotkeyAction extends BaseAction {\n action_type: 'hotkey';\n action_inputs: {\n key: string;\n };\n}\n\ninterface ScrollAction extends BaseAction {\n action_type: 'scroll';\n action_inputs: {\n direction: 'up' | 'down';\n };\n}\n\ninterface FinishedAction extends BaseAction {\n action_type: 'finished';\n action_inputs: {\n content?: string;\n };\n}\n\nexport type Action =\n | ClickAction\n | LeftDoubleAction\n | RightSingleAction\n | DragAction\n | TypeAction\n | HotkeyAction\n | ScrollAction\n | FinishedAction\n | WaitAction;\n"],"names":["debug","getDebug","parseUiTarsPlanningResponse","rawResponse","shotSize","uiTarsModelVersion","convertedText","convertBboxToCoordinates","parseResult","actionParser","JSON","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","Math","y","cleanedText"],"mappings":";;AAgBA,MAAMA,QAAQC,SAAS;AAOhB,SAASC,4BACdC,WAAmB,EACnBC,QAAc,EACdC,kBAAsC;IAEtC,MAAMC,gBAAgBC,yBAAyBJ;IAC/C,MAAMK,cAAcC,aAAa;QAC/B,YAAYH;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOF,SAAS,KAAK;YACrB,QAAQA,SAAS,MAAM;QACzB;QACA,UAAUC;IACZ;IAEAL,MACE,oBACAK,oBACA,YACAK,KAAK,SAAS,CAACF,YAAY,MAAM;IAGnC,OAAO;QACLL;QACA,SAASK,YAAY,MAAM;IAC7B;AACF;AAKA,SAASD,yBAAyBI,IAAY;IAC5C,MAAMC,UAAU;IAEhB,SAASC,aACPC,KAAa,EACbC,EAAU,EACVC,EAAU,EACVC,EAAU,EACVC,EAAU;QAEV,MAAMC,QAAQC,OAAO,QAAQ,CAACL,IAAI;QAClC,MAAMM,QAAQD,OAAO,QAAQ,CAACJ,IAAI;QAClC,MAAMM,QAAQF,OAAO,QAAQ,CAACH,IAAI;QAClC,MAAMM,QAAQH,OAAO,QAAQ,CAACF,IAAI;QAElC,MAAMM,IAAIC,KAAK,KAAK,CAAEN,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMI,IAAID,KAAK,KAAK,CAAEJ,AAAAA,CAAAA,QAAQE,KAAI,IAAK;QAEvC,OAAO,CAAC,CAAC,EAAEC,EAAE,CAAC,EAAEE,EAAE,CAAC,CAAC;IACtB;IAEA,MAAMC,cAAchB,KACjB,OAAO,CAAC,YAAY,IACpB,OAAO,CAAC,2BAA2B;IACtC,OAAOgB,YAAY,OAAO,CAACf,SAASC,cAAc,IAAI;AACxD"}