@midscene/core 1.2.3-beta-20260127070952.0 → 1.3.0

package/dist/es/ai-model/inspect.mjs

@@ -10,7 +10,7 @@ import { extractDataQueryPrompt, parseXMLExtractionResponse, systemPromptToExtra
 import { findElementPrompt, systemPromptToLocateElement } from "./prompt/llm-locator.mjs";
 import { sectionLocatorInstruction, systemPromptToLocateSection } from "./prompt/llm-section-locator.mjs";
 import { orderSensitiveJudgePrompt, systemPromptToJudgeOrderSensitive } from "./prompt/order-sensitive-judge.mjs";
-import { callAI, callAIWithObjectResponse, callAIWithStringResponse } from "./service-caller/index.mjs";
+import { AIResponseParseError, callAI, callAIWithObjectResponse, callAIWithStringResponse } from "./service-caller/index.mjs";
 const debugInspect = getDebug('ai:inspect');
 const debugSection = getDebug('ai:section');
 const extraTextFromUserPrompt = (prompt)=>{
@@ -174,7 +174,26 @@ async function AiLocateElement(options) {
             reasoning_content: parsed.think
         };
     }
-    const res = await callAIFn(msgs, modelConfig);
+    let res;
+    try {
+        res = await callAIFn(msgs, modelConfig);
+    } catch (callError) {
+        const errorMessage = callError instanceof Error ? callError.message : String(callError);
+        const rawResponse = callError instanceof AIResponseParseError ? callError.rawResponse : errorMessage;
+        const usage = callError instanceof AIResponseParseError ? callError.usage : void 0;
+        return {
+            rect: void 0,
+            parseResult: {
+                elements: [],
+                errors: [
+                    `AI call error: ${errorMessage}`
+                ]
+            },
+            rawResponse,
+            usage,
+            reasoning_content: void 0
+        };
+    }
     const rawResponse = JSON.stringify(res.content);
     let resRect;
     let matchedElements = [];
@@ -246,7 +265,21 @@ async function AiLocateSection(options) {
         });
         msgs.push(...addOns);
     }
-    const result = await callAIWithObjectResponse(msgs, modelConfig);
+    let result;
+    try {
+        result = await callAIWithObjectResponse(msgs, modelConfig);
+    } catch (callError) {
+        const errorMessage = callError instanceof Error ? callError.message : String(callError);
+        const rawResponse = callError instanceof AIResponseParseError ? callError.rawResponse : errorMessage;
+        const usage = callError instanceof AIResponseParseError ? callError.usage : void 0;
+        return {
+            rect: void 0,
+            imageBase64: void 0,
+            error: `AI call error: ${errorMessage}`,
+            rawResponse,
+            usage
+        };
+    }
     let sectionRect;
     const sectionBbox = result.content.bbox;
     if (sectionBbox) {
@@ -314,9 +347,16 @@ async function AiExtractElementInfo(options) {
         msgs.push(...addOns);
     }
     const { content: rawResponse, usage, reasoning_content } = await callAI(msgs, modelConfig);
-    const parseResult = parseXMLExtractionResponse(rawResponse);
+    let parseResult;
+    try {
+        parseResult = parseXMLExtractionResponse(rawResponse);
+    } catch (parseError) {
+        const errorMessage = parseError instanceof Error ? parseError.message : String(parseError);
+        throw new AIResponseParseError(`XML parse error: ${errorMessage}`, rawResponse, usage);
+    }
     return {
         parseResult,
+        rawResponse,
         usage,
         reasoning_content
     };

package/dist/es/ai-model/inspect.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/inspect.mjs","sources":["../../../src/ai-model/inspect.ts"],"sourcesContent":["import type {\n  AIElementResponse,\n  AISectionLocatorResponse,\n  AIUsageInfo,\n  Rect,\n  ServiceExtractOption,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { generateElementByPosition } from '@midscene/shared/extractor/dom-util';\nimport {\n  cropByRect,\n  paddingToMatchBlockByBase64,\n  preProcessImageUrl,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { LocateResultElement } from '@midscene/shared/types';\nimport { assert } from '@midscene/shared/utils';\nimport type {\n  ChatCompletionSystemMessageParam,\n  ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport type { TMultimodalPrompt, TUserPrompt } from '../common';\nimport { adaptBboxToRect, expandSearchArea, mergeRects } from '../common';\nimport { parseAutoGLMLocateResponse } from './auto-glm/parser';\nimport { getAutoGLMLocatePrompt } from './auto-glm/prompt';\nimport { isAutoGLM } from './auto-glm/util';\nimport {\n  extractDataQueryPrompt,\n  parseXMLExtractionResponse,\n  systemPromptToExtract,\n} from './prompt/extraction';\nimport {\n  findElementPrompt,\n  systemPromptToLocateElement,\n} from './prompt/llm-locator';\nimport {\n  sectionLocatorInstruction,\n  systemPromptToLocateSection,\n} from './prompt/llm-section-locator';\nimport {\n  orderSensitiveJudgePrompt,\n  systemPromptToJudgeOrderSensitive,\n} from './prompt/order-sensitive-judge';\nimport {\n  callAI,\n  callAIWithObjectResponse,\n  callAIWithStringResponse,\n} from './service-caller/index';\n\nexport type AIArgs = [\n  ChatCompletionSystemMessageParam,\n  ...ChatCompletionUserMessageParam[],\n];\n\nconst debugInspect = getDebug('ai:inspect');\nconst debugSection = getDebug('ai:section');\n\nconst extraTextFromUserPrompt = (prompt: TUserPrompt): string => {\n  if (typeof prompt === 'string') {\n    return prompt;\n  } else {\n    return prompt.prompt;\n  }\n};\n\nconst promptsToChatParam = async (\n  multimodalPrompt: TMultimodalPrompt,\n): Promise<ChatCompletionUserMessageParam[]> => {\n  const msgs: ChatCompletionUserMessageParam[] = [];\n  if (multimodalPrompt?.images?.length) {\n    msgs.push({\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: 'Next, I will provide all the reference images.',\n        },\n      ],\n    });\n\n    for (const item of multimodalPrompt.images) {\n      const base64 = await preProcessImageUrl(\n        item.url,\n        !!multimodalPrompt.convertHttpImage2Base64,\n      );\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'text',\n            text: `this is the reference image named '${item.name}':`,\n          },\n        ],\n      });\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: base64,\n              detail: 'high',\n            },\n          },\n        ],\n      });\n    }\n  }\n  return msgs;\n};\n\nexport async function AiLocateElement(options: {\n  context: UIContext;\n  targetElementDescription: TUserPrompt;\n  callAIFn: typeof callAIWithObjectResponse<\n    AIElementResponse | [number, number]\n  >;\n  searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n  modelConfig: IModelConfig;\n}): Promise<{\n  parseResult: {\n    elements: LocateResultElement[];\n    errors?: string[];\n  };\n  rect?: Rect;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}> {\n  const { context, targetElementDescription, callAIFn, modelConfig } = options;\n  const { modelFamily } = modelConfig;\n  const screenshotBase64 = context.screenshot.base64;\n\n  assert(\n    targetElementDescription,\n    'cannot find the target element description',\n  );\n  const targetElementDescriptionText = extraTextFromUserPrompt(\n    targetElementDescription,\n  );\n  const userInstructionPrompt = findElementPrompt(targetElementDescriptionText);\n  const systemPrompt = isAutoGLM(modelFamily)\n    ? getAutoGLMLocatePrompt(modelFamily)\n    : systemPromptToLocateElement(modelFamily);\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = context.size.width;\n  let imageHeight = context.size.height;\n  let originalImageWidth = imageWidth;\n  let originalImageHeight = imageHeight;\n\n  if (options.searchConfig) {\n    assert(\n      options.searchConfig.rect,\n      'searchArea is provided but its rect cannot be found. Failed to locate element',\n    );\n    assert(\n      options.searchConfig.imageBase64,\n      'searchArea is provided but its imageBase64 cannot be found. Failed to locate element',\n    );\n\n    imagePayload = options.searchConfig.imageBase64;\n    imageWidth = options.searchConfig.rect?.width;\n    imageHeight = options.searchConfig.rect?.height;\n    originalImageWidth = imageWidth;\n    originalImageHeight = imageHeight;\n  } else if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: isAutoGLM(modelFamily)\n            ? `Tap: ${userInstructionPrompt}`\n            : userInstructionPrompt,\n        },\n      ],\n    },\n  ];\n\n  if (typeof targetElementDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: targetElementDescription.images,\n      convertHttpImage2Base64: targetElementDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  if (isAutoGLM(modelFamily)) {\n    const { content: rawResponseContent, usage } =\n      await callAIWithStringResponse(msgs, modelConfig);\n\n    debugInspect('auto-glm rawResponse:', rawResponseContent);\n\n    const parsed = parseAutoGLMLocateResponse(rawResponseContent);\n\n    debugInspect('auto-glm thinking:', parsed.think);\n    debugInspect('auto-glm coordinates:', parsed.coordinates);\n\n    let resRect: Rect | undefined;\n    let matchedElements: LocateResultElement[] = [];\n    let errors: string[] = [];\n\n    if (parsed.error || !parsed.coordinates) {\n      errors = [parsed.error || 'Failed to parse auto-glm response'];\n      debugInspect('auto-glm parse error:', errors[0]);\n    } else {\n      const { x, y } = parsed.coordinates;\n\n      debugInspect('auto-glm coordinates [0-999]:', { x, y });\n\n      // Convert auto-glm coordinates [0,999] to pixel bbox\n      // Map from [0,999] to pixel coordinates\n      const pixelX = Math.round((x * imageWidth) / 1000);\n      const pixelY = Math.round((y * imageHeight) / 1000);\n\n      debugInspect('auto-glm pixel coordinates:', { pixelX, pixelY });\n\n      // Create a small bbox around the point\n      const bboxSize = 10;\n      const x1 = Math.max(pixelX - bboxSize / 2, 0);\n      const y1 = Math.max(pixelY - bboxSize / 2, 0);\n      const x2 = Math.min(pixelX + bboxSize / 2, imageWidth);\n      const y2 = Math.min(pixelY + bboxSize / 2, imageHeight);\n\n      // Convert to Rect format\n      resRect = {\n        left: x1,\n        top: y1,\n        width: x2 - x1,\n        height: y2 - y1,\n      };\n\n      // Apply offset if searching in a cropped area\n      if (options.searchConfig?.rect) {\n        resRect.left += options.searchConfig.rect.left;\n        resRect.top += options.searchConfig.rect.top;\n      }\n\n      debugInspect('auto-glm resRect:', resRect);\n\n      const rectCenter = {\n        x: resRect.left + resRect.width / 2,\n        y: resRect.top + resRect.height / 2,\n      };\n\n      const element: LocateResultElement = generateElementByPosition(\n        rectCenter,\n        targetElementDescriptionText as string,\n      );\n\n      if (element) {\n        matchedElements = [element];\n      }\n    }\n\n    return {\n      rect: resRect,\n      parseResult: {\n        elements: matchedElements,\n        errors,\n      },\n      rawResponse: rawResponseContent,\n      usage,\n      reasoning_content: parsed.think,\n    };\n  }\n\n  const res = await callAIFn(msgs, modelConfig);\n\n  const rawResponse = JSON.stringify(res.content);\n\n  let resRect: Rect | undefined;\n  let matchedElements: LocateResultElement[] = [];\n  let errors: string[] | undefined =\n    'errors' in res.content ? res.content.errors : [];\n  try {\n    if (\n      'bbox' in res.content &&\n      Array.isArray(res.content.bbox) &&\n      res.content.bbox.length >= 1\n    ) {\n      resRect = adaptBboxToRect(\n        res.content.bbox,\n        imageWidth,\n        imageHeight,\n        options.searchConfig?.rect?.left,\n        options.searchConfig?.rect?.top,\n        originalImageWidth,\n        originalImageHeight,\n        modelFamily,\n      );\n\n      debugInspect('resRect', resRect);\n\n      const rectCenter = {\n        x: resRect.left + resRect.width / 2,\n        y: resRect.top + resRect.height / 2,\n      };\n\n      const element: LocateResultElement = generateElementByPosition(\n        rectCenter,\n        targetElementDescriptionText as string,\n      );\n      errors = [];\n\n      if (element) {\n        matchedElements = [element];\n      }\n    }\n  } catch (e) {\n    const msg =\n      e instanceof Error\n        ? `Failed to parse bbox: ${e.message}`\n        : 'unknown error in locate';\n    if (!errors || errors?.length === 0) {\n      errors = [msg];\n    } else {\n      errors.push(`(${msg})`);\n    }\n  }\n\n  return {\n    rect: resRect,\n    parseResult: {\n      elements: matchedElements as LocateResultElement[],\n      errors: errors as string[],\n    },\n    rawResponse,\n    usage: res.usage,\n    reasoning_content: res.reasoning_content,\n  };\n}\n\nexport async function AiLocateSection(options: {\n  context: UIContext;\n  sectionDescription: TUserPrompt;\n  modelConfig: IModelConfig;\n}): Promise<{\n  rect?: Rect;\n  imageBase64?: string;\n  error?: string;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n}> {\n  const { context, sectionDescription, modelConfig } = options;\n  const { modelFamily } = modelConfig;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const systemPrompt = systemPromptToLocateSection(modelFamily);\n  const sectionLocatorInstructionText = sectionLocatorInstruction(\n    extraTextFromUserPrompt(sectionDescription),\n  );\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: screenshotBase64,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: sectionLocatorInstructionText,\n        },\n      ],\n    },\n  ];\n\n  if (typeof sectionDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: sectionDescription.images,\n      convertHttpImage2Base64: sectionDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const result = await callAIWithObjectResponse<AISectionLocatorResponse>(\n    msgs,\n    modelConfig,\n  );\n\n  let sectionRect: Rect | undefined;\n  const sectionBbox = result.content.bbox;\n  if (sectionBbox) {\n    const targetRect = adaptBboxToRect(\n      sectionBbox,\n      context.size.width,\n      context.size.height,\n      0,\n      0,\n      context.size.width,\n      context.size.height,\n      modelFamily,\n    );\n    debugSection('original targetRect %j', targetRect);\n\n    const referenceBboxList = result.content.references_bbox || [];\n    debugSection('referenceBboxList %j', referenceBboxList);\n\n    const referenceRects = referenceBboxList\n      .filter((bbox) => Array.isArray(bbox))\n      .map((bbox) => {\n        return adaptBboxToRect(\n          bbox,\n          context.size.width,\n          context.size.height,\n          0,\n          0,\n          context.size.width,\n          context.size.height,\n          modelFamily,\n        );\n      });\n    debugSection('referenceRects %j', referenceRects);\n\n    // merge the sectionRect and referenceRects\n    const mergedRect = mergeRects([targetRect, ...referenceRects]);\n    debugSection('mergedRect %j', mergedRect);\n\n    // expand search area to at least 200 x 200\n    sectionRect = expandSearchArea(mergedRect, context.size, modelFamily);\n    debugSection('expanded sectionRect %j', sectionRect);\n  }\n\n  let imageBase64 = screenshotBase64;\n  if (sectionRect) {\n    const croppedResult = await cropByRect(\n      screenshotBase64,\n      sectionRect,\n      modelFamily === 'qwen2.5-vl',\n    );\n    imageBase64 = croppedResult.imageBase64;\n    sectionRect.width = croppedResult.width;\n    sectionRect.height = croppedResult.height;\n  }\n\n  return {\n    rect: sectionRect,\n    imageBase64,\n    error: result.content.error,\n    rawResponse: JSON.stringify(result.content),\n    usage: result.usage,\n  };\n}\n\nexport async function AiExtractElementInfo<T>(options: {\n  dataQuery: string | Record<string, string>;\n  multimodalPrompt?: TMultimodalPrompt;\n  context: UIContext;\n  pageDescription?: string;\n  extractOption?: ServiceExtractOption;\n  modelConfig: IModelConfig;\n}) {\n  const { dataQuery, context, extractOption, multimodalPrompt, modelConfig } =\n    options;\n  const systemPrompt = systemPromptToExtract();\n  const screenshotBase64 = context.screenshot.base64;\n\n  const extractDataPromptText = extractDataQueryPrompt(\n    options.pageDescription || '',\n    dataQuery,\n  );\n\n  const userContent: ChatCompletionUserMessageParam['content'] = [];\n\n  if (extractOption?.screenshotIncluded !== false) {\n    userContent.push({\n      type: 'image_url',\n      image_url: {\n        url: screenshotBase64,\n        detail: 'high',\n      },\n    });\n  }\n\n  userContent.push({\n    type: 'text',\n    text: extractDataPromptText,\n  });\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userContent,\n    },\n  ];\n\n  if (multimodalPrompt) {\n    const addOns = await promptsToChatParam({\n      images: multimodalPrompt.images,\n      convertHttpImage2Base64: multimodalPrompt.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const {\n    content: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAI(msgs, modelConfig);\n\n  // Parse XML response to JSON object\n  const parseResult = parseXMLExtractionResponse<T>(rawResponse);\n\n  return {\n    parseResult,\n    usage,\n    reasoning_content,\n  };\n}\n\nexport async function AiJudgeOrderSensitive(\n  description: string,\n  callAIFn: typeof callAIWithObjectResponse<{ isOrderSensitive: boolean }>,\n  modelConfig: IModelConfig,\n): Promise<{\n  isOrderSensitive: boolean;\n  usage?: AIUsageInfo;\n}> {\n  const systemPrompt = systemPromptToJudgeOrderSensitive();\n  const userPrompt = orderSensitiveJudgePrompt(description);\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userPrompt,\n    },\n  ];\n\n  const result = await callAIFn(msgs, modelConfig);\n\n  return {\n    isOrderSensitive: result.content.isOrderSensitive ?? false,\n    usage: result.usage,\n  };\n}\n"],"names":["debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","callAIFn","modelConfig","modelFamily","screenshotBase64","assert","targetElementDescriptionText","userInstructionPrompt","findElementPrompt","systemPrompt","isAutoGLM","getAutoGLMLocatePrompt","systemPromptToLocateElement","imagePayload","imageWidth","imageHeight","originalImageWidth","originalImageHeight","paddedResult","paddingToMatchBlockByBase64","addOns","rawResponseContent","usage","callAIWithStringResponse","parsed","parseAutoGLMLocateResponse","resRect","matchedElements","errors","x","y","pixelX","Math","pixelY","bboxSize","x1","y1","x2","y2","rectCenter","element","generateElementByPosition","res","rawResponse","JSON","Array","adaptBboxToRect","e","msg","Error","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","callAIWithObjectResponse","sectionRect","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandSearchArea","imageBase64","croppedResult","cropByRect","AiExtractElementInfo","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent","reasoning_content","callAI","parseResult","parseXMLExtractionResponse","AiJudgeOrderSensitive","description","systemPromptToJudgeOrderSensitive","userPrompt","orderSensitiveJudgePrompt"],"mappings":";;;;;;;;;;;;;AAuDA,MAAMA,eAAeC,SAAS;AAC9B,MAAMC,eAAeD,SAAS;AAE9B,MAAME,0BAA0B,CAACC;IAC/B,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAEP,OAAOA,OAAO,MAAM;AAExB;AAEA,MAAMC,qBAAqB,OACzBC;IAEA,MAAMC,OAAyC,EAAE;IACjD,IAAID,kBAAkB,QAAQ,QAAQ;QACpCC,KAAK,IAAI,CAAC;YACR,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM;gBACR;aACD;QACH;QAEA,KAAK,MAAMC,QAAQF,iBAAiB,MAAM,CAAE;YAC1C,MAAMG,SAAS,MAAMC,mBACnBF,KAAK,GAAG,EACR,CAAC,CAACF,iBAAiB,uBAAuB;YAG5CC,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,MAAM,CAAC,mCAAmC,EAAEC,KAAK,IAAI,CAAC,EAAE,CAAC;oBAC3D;iBACD;YACH;YAEAD,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKE;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;QACF;IACF;IACA,OAAOF;AACT;AAEO,eAAeI,gBAAgBC,OAQrC;IAUC,MAAM,EAAEC,OAAO,EAAEC,wBAAwB,EAAEC,QAAQ,EAAEC,WAAW,EAAE,GAAGJ;IACrE,MAAM,EAAEK,WAAW,EAAE,GAAGD;IACxB,MAAME,mBAAmBL,QAAQ,UAAU,CAAC,MAAM;IAElDM,OACEL,0BACA;IAEF,MAAMM,+BAA+BjB,wBACnCW;IAEF,MAAMO,wBAAwBC,kBAAkBF;IAChD,MAAMG,eAAeC,UAAUP,eAC3BQ,uBAAuBR,eACvBS,4BAA4BT;IAEhC,IAAIU,eAAeT;IACnB,IAAIU,aAAaf,QAAQ,IAAI,CAAC,KAAK;IACnC,IAAIgB,cAAchB,QAAQ,IAAI,CAAC,MAAM;IACrC,IAAIiB,qBAAqBF;IACzB,IAAIG,sBAAsBF;IAE1B,IAAIjB,QAAQ,YAAY,EAAE;QACxBO,OACEP,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFO,OACEP,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGFe,eAAef,QAAQ,YAAY,CAAC,WAAW;QAC/CgB,aAAahB,QAAQ,YAAY,CAAC,IAAI,EAAE;QACxCiB,cAAcjB,QAAQ,YAAY,CAAC,IAAI,EAAE;QACzCkB,qBAAqBF;QACrBG,sBAAsBF;IACxB,OAAO,IAAIZ,AAAgB,iBAAhBA,aAA8B;QACvC,MAAMe,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAMzB,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKI;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMH,UAAUP,eACZ,CAAC,KAAK,EAAEI,uBAAuB,GAC/BA;gBACN;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAOP,0BAAuC;QAChD,MAAMoB,SAAS,MAAM7B,mBAAmB;YACtC,QAAQS,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAP,KAAK,IAAI,IAAI2B;IACf;IAEA,IAAIV,UAAUP,cAAc;QAC1B,MAAM,EAAE,SAASkB,kBAAkB,EAAEC,KAAK,EAAE,GAC1C,MAAMC,yBAAyB9B,MAAMS;QAEvChB,aAAa,yBAAyBmC;QAEtC,MAAMG,SAASC,2BAA2BJ;QAE1CnC,aAAa,sBAAsBsC,OAAO,KAAK;QAC/CtC,aAAa,yBAAyBsC,OAAO,WAAW;QAExD,IAAIE;QACJ,IAAIC,kBAAyC,EAAE;QAC/C,IAAIC,SAAmB,EAAE;QAEzB,IAAIJ,OAAO,KAAK,IAAI,CAACA,OAAO,WAAW,EAAE;YACvCI,SAAS;gBAACJ,OAAO,KAAK,IAAI;aAAoC;YAC9DtC,aAAa,yBAAyB0C,MAAM,CAAC,EAAE;QACjD,OAAO;YACL,MAAM,EAAEC,CAAC,EAAEC,CAAC,EAAE,GAAGN,OAAO,WAAW;YAEnCtC,aAAa,iCAAiC;gBAAE2C;gBAAGC;YAAE;YAIrD,MAAMC,SAASC,KAAK,KAAK,CAAEH,IAAIf,aAAc;YAC7C,MAAMmB,SAASD,KAAK,KAAK,CAAEF,IAAIf,cAAe;YAE9C7B,aAAa,+BAA+B;gBAAE6C;gBAAQE;YAAO;YAG7D,MAAMC,WAAW;YACjB,MAAMC,KAAKH,KAAK,GAAG,CAACD,SAASG,WAAW,GAAG;YAC3C,MAAME,KAAKJ,KAAK,GAAG,CAACC,SAASC,WAAW,GAAG;YAC3C,MAAMG,KAAKL,KAAK,GAAG,CAACD,SAASG,WAAW,GAAGpB;YAC3C,MAAMwB,KAAKN,KAAK,GAAG,CAACC,SAASC,WAAW,GAAGnB;YAG3CW,UAAU;gBACR,MAAMS;gBACN,KAAKC;gBACL,OAAOC,KAAKF;gBACZ,QAAQG,KAAKF;YACf;YAGA,IAAItC,QAAQ,YAAY,EAAE,MAAM;gBAC9B4B,QAAQ,IAAI,IAAI5B,QAAQ,YAAY,CAAC,IAAI,CAAC,IAAI;gBAC9C4B,QAAQ,GAAG,IAAI5B,QAAQ,YAAY,CAAC,IAAI,CAAC,GAAG;YAC9C;YAEAZ,aAAa,qBAAqBwC;YAElC,MAAMa,aAAa;gBACjB,GAAGb,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YAEA,MAAMc,UAA+BC,0BACnCF,YACAjC;YAGF,IAAIkC,SACFb,kBAAkB;gBAACa;aAAQ;QAE/B;QAEA,OAAO;YACL,MAAMd;YACN,aAAa;gBACX,UAAUC;gBACVC;YACF;YACA,aAAaP;YACbC;YACA,mBAAmBE,OAAO,KAAK;QACjC;IACF;IAEA,MAAMkB,MAAM,MAAMzC,SAASR,MAAMS;IAEjC,MAAMyC,cAAcC,KAAK,SAAS,CAACF,IAAI,OAAO;IAE9C,IAAIhB;IACJ,IAAIC,kBAAyC,EAAE;IAC/C,IAAIC,SACF,YAAYc,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IACE,UAAUA,IAAI,OAAO,IACrBG,MAAM,OAAO,CAACH,IAAI,OAAO,CAAC,IAAI,KAC9BA,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,IAAI,GAC3B;YACAhB,UAAUoB,gBACRJ,IAAI,OAAO,CAAC,IAAI,EAChB5B,YACAC,aACAjB,QAAQ,YAAY,EAAE,MAAM,MAC5BA,QAAQ,YAAY,EAAE,MAAM,KAC5BkB,oBACAC,qBACAd;YAGFjB,aAAa,WAAWwC;YAExB,MAAMa,aAAa;gBACjB,GAAGb,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YAEA,MAAMc,UAA+BC,0BACnCF,YACAjC;YAEFsB,SAAS,EAAE;YAEX,IAAIY,SACFb,kBAAkB;gBAACa;aAAQ;QAE/B;IACF,EAAE,OAAOO,GAAG;QACV,MAAMC,MACJD,aAAaE,QACT,CAAC,sBAAsB,EAAEF,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACnB,UAAUA,QAAQ,WAAW,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAEoB,IAAI,CAAC,CAAC;aAFtBpB,SAAS;YAACoB;SAAI;IAIlB;IAEA,OAAO;QACL,MAAMtB;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAe;QACA,OAAOD,IAAI,KAAK;QAChB,mBAAmBA,IAAI,iBAAiB;IAC1C;AACF;AAEO,eAAeQ,gBAAgBpD,OAIrC;IAOC,MAAM,EAAEC,OAAO,EAAEoD,kBAAkB,EAAEjD,WAAW,EAAE,GAAGJ;IACrD,MAAM,EAAEK,WAAW,EAAE,GAAGD;IACxB,MAAME,mBAAmBL,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAMU,eAAe2C,4BAA4BjD;IACjD,MAAMkD,gCAAgCC,0BACpCjE,wBAAwB8D;IAE1B,MAAM1D,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKL;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMiD;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAM/B,SAAS,MAAM7B,mBAAmB;YACtC,QAAQ4D,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACA1D,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAMmC,SAAS,MAAMC,yBACnB/D,MACAS;IAGF,IAAIuD;IACJ,MAAMC,cAAcH,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIG,aAAa;QACf,MAAMC,aAAab,gBACjBY,aACA3D,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAEFf,aAAa,0BAA0BuE;QAEvC,MAAMC,oBAAoBL,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9DnE,aAAa,wBAAwBwE;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAASjB,MAAM,OAAO,CAACiB,OAC/B,GAAG,CAAC,CAACA,OACGhB,gBACLgB,MACA/D,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAGNf,aAAa,qBAAqByE;QAGlC,MAAME,aAAaC,WAAW;YAACL;eAAeE;SAAe;QAC7DzE,aAAa,iBAAiB2E;QAG9BN,cAAcQ,iBAAiBF,YAAYhE,QAAQ,IAAI,EAAEI;QACzDf,aAAa,2BAA2BqE;IAC1C;IAEA,IAAIS,cAAc9D;IAClB,IAAIqD,aAAa;QACf,MAAMU,gBAAgB,MAAMC,WAC1BhE,kBACAqD,aACAtD,AAAgB,iBAAhBA;QAEF+D,cAAcC,cAAc,WAAW;QACvCV,YAAY,KAAK,GAAGU,cAAc,KAAK;QACvCV,YAAY,MAAM,GAAGU,cAAc,MAAM;IAC3C;IAEA,OAAO;QACL,MAAMV;QACNS;QACA,OAAOX,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAaX,KAAK,SAAS,CAACW,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAec,qBAAwBvE,OAO7C;IACC,MAAM,EAAEwE,SAAS,EAAEvE,OAAO,EAAEwE,aAAa,EAAE/E,gBAAgB,EAAEU,WAAW,EAAE,GACxEJ;IACF,MAAMW,eAAe+D;IACrB,MAAMpE,mBAAmBL,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM0E,wBAAwBC,uBAC5B5E,QAAQ,eAAe,IAAI,IAC3BwE;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,eAAe,uBAAuB,OACxCI,YAAY,IAAI,CAAC;QACf,MAAM;QACN,WAAW;YACT,KAAKvE;YACL,QAAQ;QACV;IACF;IAGFuE,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAMhF,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASkE;QACX;KACD;IAED,IAAInF,kBAAkB;QACpB,MAAM4B,SAAS,MAAM7B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAC,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAM,EACJ,SAASuB,WAAW,EACpBrB,KAAK,EACLsD,iBAAiB,EAClB,GAAG,MAAMC,OAAOpF,MAAMS;IAGvB,MAAM4E,cAAcC,2BAA8BpC;IAElD,OAAO;QACLmC;QACAxD;QACAsD;IACF;AACF;AAEO,eAAeI,sBACpBC,WAAmB,EACnBhF,QAAwE,EACxEC,WAAyB;IAKzB,MAAMO,eAAeyE;IACrB,MAAMC,aAAaC,0BAA0BH;IAE7C,MAAMxF,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS0E;QACX;KACD;IAED,MAAM5B,SAAS,MAAMtD,SAASR,MAAMS;IAEpC,OAAO;QACL,kBAAkBqD,OAAO,OAAO,CAAC,gBAAgB,IAAI;QACrD,OAAOA,OAAO,KAAK;IACrB;AACF"}
+{"version":3,"file":"ai-model/inspect.mjs","sources":["../../../src/ai-model/inspect.ts"],"sourcesContent":["import type {\n  AIDataExtractionResponse,\n  AIElementResponse,\n  AISectionLocatorResponse,\n  AIUsageInfo,\n  Rect,\n  ServiceExtractOption,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { generateElementByPosition } from '@midscene/shared/extractor/dom-util';\nimport {\n  cropByRect,\n  paddingToMatchBlockByBase64,\n  preProcessImageUrl,\n} from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { LocateResultElement } from '@midscene/shared/types';\nimport { assert } from '@midscene/shared/utils';\nimport type {\n  ChatCompletionSystemMessageParam,\n  ChatCompletionUserMessageParam,\n} from 'openai/resources/index';\nimport type { TMultimodalPrompt, TUserPrompt } from '../common';\nimport { adaptBboxToRect, expandSearchArea, mergeRects } from '../common';\nimport { parseAutoGLMLocateResponse } from './auto-glm/parser';\nimport { getAutoGLMLocatePrompt } from './auto-glm/prompt';\nimport { isAutoGLM } from './auto-glm/util';\nimport {\n  extractDataQueryPrompt,\n  parseXMLExtractionResponse,\n  systemPromptToExtract,\n} from './prompt/extraction';\nimport {\n  findElementPrompt,\n  systemPromptToLocateElement,\n} from './prompt/llm-locator';\nimport {\n  sectionLocatorInstruction,\n  systemPromptToLocateSection,\n} from './prompt/llm-section-locator';\nimport {\n  orderSensitiveJudgePrompt,\n  systemPromptToJudgeOrderSensitive,\n} from './prompt/order-sensitive-judge';\nimport {\n  AIResponseParseError,\n  callAI,\n  callAIWithObjectResponse,\n  callAIWithStringResponse,\n} from './service-caller/index';\n\nexport type AIArgs = [\n  ChatCompletionSystemMessageParam,\n  ...ChatCompletionUserMessageParam[],\n];\n\nconst debugInspect = getDebug('ai:inspect');\nconst debugSection = getDebug('ai:section');\n\nconst extraTextFromUserPrompt = (prompt: TUserPrompt): string => {\n  if (typeof prompt === 'string') {\n    return prompt;\n  } else {\n    return prompt.prompt;\n  }\n};\n\nconst promptsToChatParam = async (\n  multimodalPrompt: TMultimodalPrompt,\n): Promise<ChatCompletionUserMessageParam[]> => {\n  const msgs: ChatCompletionUserMessageParam[] = [];\n  if (multimodalPrompt?.images?.length) {\n    msgs.push({\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: 'Next, I will provide all the reference images.',\n        },\n      ],\n    });\n\n    for (const item of multimodalPrompt.images) {\n      const base64 = await preProcessImageUrl(\n        item.url,\n        !!multimodalPrompt.convertHttpImage2Base64,\n      );\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'text',\n            text: `this is the reference image named '${item.name}':`,\n          },\n        ],\n      });\n\n      msgs.push({\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: base64,\n              detail: 'high',\n            },\n          },\n        ],\n      });\n    }\n  }\n  return msgs;\n};\n\nexport async function AiLocateElement(options: {\n  context: UIContext;\n  targetElementDescription: TUserPrompt;\n  callAIFn: typeof callAIWithObjectResponse<\n    AIElementResponse | [number, number]\n  >;\n  searchConfig?: Awaited<ReturnType<typeof AiLocateSection>>;\n  modelConfig: IModelConfig;\n}): Promise<{\n  parseResult: {\n    elements: LocateResultElement[];\n    errors?: string[];\n  };\n  rect?: Rect;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}> {\n  const { context, targetElementDescription, callAIFn, modelConfig } = options;\n  const { modelFamily } = modelConfig;\n  const screenshotBase64 = context.screenshot.base64;\n\n  assert(\n    targetElementDescription,\n    'cannot find the target element description',\n  );\n  const targetElementDescriptionText = extraTextFromUserPrompt(\n    targetElementDescription,\n  );\n  const userInstructionPrompt = findElementPrompt(targetElementDescriptionText);\n  const systemPrompt = isAutoGLM(modelFamily)\n    ? getAutoGLMLocatePrompt(modelFamily)\n    : systemPromptToLocateElement(modelFamily);\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = context.size.width;\n  let imageHeight = context.size.height;\n  let originalImageWidth = imageWidth;\n  let originalImageHeight = imageHeight;\n\n  if (options.searchConfig) {\n    assert(\n      options.searchConfig.rect,\n      'searchArea is provided but its rect cannot be found. Failed to locate element',\n    );\n    assert(\n      options.searchConfig.imageBase64,\n      'searchArea is provided but its imageBase64 cannot be found. Failed to locate element',\n    );\n\n    imagePayload = options.searchConfig.imageBase64;\n    imageWidth = options.searchConfig.rect?.width;\n    imageHeight = options.searchConfig.rect?.height;\n    originalImageWidth = imageWidth;\n    originalImageHeight = imageHeight;\n  } else if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: isAutoGLM(modelFamily)\n            ? `Tap: ${userInstructionPrompt}`\n            : userInstructionPrompt,\n        },\n      ],\n    },\n  ];\n\n  if (typeof targetElementDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: targetElementDescription.images,\n      convertHttpImage2Base64: targetElementDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  if (isAutoGLM(modelFamily)) {\n    const { content: rawResponseContent, usage } =\n      await callAIWithStringResponse(msgs, modelConfig);\n\n    debugInspect('auto-glm rawResponse:', rawResponseContent);\n\n    const parsed = parseAutoGLMLocateResponse(rawResponseContent);\n\n    debugInspect('auto-glm thinking:', parsed.think);\n    debugInspect('auto-glm coordinates:', parsed.coordinates);\n\n    let resRect: Rect | undefined;\n    let matchedElements: LocateResultElement[] = [];\n    let errors: string[] = [];\n\n    if (parsed.error || !parsed.coordinates) {\n      errors = [parsed.error || 'Failed to parse auto-glm response'];\n      debugInspect('auto-glm parse error:', errors[0]);\n    } else {\n      const { x, y } = parsed.coordinates;\n\n      debugInspect('auto-glm coordinates [0-999]:', { x, y });\n\n      // Convert auto-glm coordinates [0,999] to pixel bbox\n      // Map from [0,999] to pixel coordinates\n      const pixelX = Math.round((x * imageWidth) / 1000);\n      const pixelY = Math.round((y * imageHeight) / 1000);\n\n      debugInspect('auto-glm pixel coordinates:', { pixelX, pixelY });\n\n      // Create a small bbox around the point\n      const bboxSize = 10;\n      const x1 = Math.max(pixelX - bboxSize / 2, 0);\n      const y1 = Math.max(pixelY - bboxSize / 2, 0);\n      const x2 = Math.min(pixelX + bboxSize / 2, imageWidth);\n      const y2 = Math.min(pixelY + bboxSize / 2, imageHeight);\n\n      // Convert to Rect format\n      resRect = {\n        left: x1,\n        top: y1,\n        width: x2 - x1,\n        height: y2 - y1,\n      };\n\n      // Apply offset if searching in a cropped area\n      if (options.searchConfig?.rect) {\n        resRect.left += options.searchConfig.rect.left;\n        resRect.top += options.searchConfig.rect.top;\n      }\n\n      debugInspect('auto-glm resRect:', resRect);\n\n      const rectCenter = {\n        x: resRect.left + resRect.width / 2,\n        y: resRect.top + resRect.height / 2,\n      };\n\n      const element: LocateResultElement = generateElementByPosition(\n        rectCenter,\n        targetElementDescriptionText as string,\n      );\n\n      if (element) {\n        matchedElements = [element];\n      }\n    }\n\n    return {\n      rect: resRect,\n      parseResult: {\n        elements: matchedElements,\n        errors,\n      },\n      rawResponse: rawResponseContent,\n      usage,\n      reasoning_content: parsed.think,\n    };\n  }\n\n  let res: Awaited<ReturnType<typeof callAIFn>>;\n  try {\n    res = await callAIFn(msgs, modelConfig);\n  } catch (callError) {\n    // Return error with usage and rawResponse if available\n    const errorMessage =\n      callError instanceof Error ? callError.message : String(callError);\n    const rawResponse =\n      callError instanceof AIResponseParseError\n        ? callError.rawResponse\n        : errorMessage;\n    const usage =\n      callError instanceof AIResponseParseError ? callError.usage : undefined;\n    return {\n      rect: undefined,\n      parseResult: {\n        elements: [],\n        errors: [`AI call error: ${errorMessage}`],\n      },\n      rawResponse,\n      usage,\n      reasoning_content: undefined,\n    };\n  }\n\n  const rawResponse = JSON.stringify(res.content);\n\n  let resRect: Rect | undefined;\n  let matchedElements: LocateResultElement[] = [];\n  let errors: string[] | undefined =\n    'errors' in res.content ? res.content.errors : [];\n  try {\n    if (\n      'bbox' in res.content &&\n      Array.isArray(res.content.bbox) &&\n      res.content.bbox.length >= 1\n    ) {\n      resRect = adaptBboxToRect(\n        res.content.bbox,\n        imageWidth,\n        imageHeight,\n        options.searchConfig?.rect?.left,\n        options.searchConfig?.rect?.top,\n        originalImageWidth,\n        originalImageHeight,\n        modelFamily,\n      );\n\n      debugInspect('resRect', resRect);\n\n      const rectCenter = {\n        x: resRect.left + resRect.width / 2,\n        y: resRect.top + resRect.height / 2,\n      };\n\n      const element: LocateResultElement = generateElementByPosition(\n        rectCenter,\n        targetElementDescriptionText as string,\n      );\n      errors = [];\n\n      if (element) {\n        matchedElements = [element];\n      }\n    }\n  } catch (e) {\n    const msg =\n      e instanceof Error\n        ? `Failed to parse bbox: ${e.message}`\n        : 'unknown error in locate';\n    if (!errors || errors?.length === 0) {\n      errors = [msg];\n    } else {\n      errors.push(`(${msg})`);\n    }\n  }\n\n  return {\n    rect: resRect,\n    parseResult: {\n      elements: matchedElements as LocateResultElement[],\n      errors: errors as string[],\n    },\n    rawResponse,\n    usage: res.usage,\n    reasoning_content: res.reasoning_content,\n  };\n}\n\nexport async function AiLocateSection(options: {\n  context: UIContext;\n  sectionDescription: TUserPrompt;\n  modelConfig: IModelConfig;\n}): Promise<{\n  rect?: Rect;\n  imageBase64?: string;\n  error?: string;\n  rawResponse: string;\n  usage?: AIUsageInfo;\n}> {\n  const { context, sectionDescription, modelConfig } = options;\n  const { modelFamily } = modelConfig;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const systemPrompt = systemPromptToLocateSection(modelFamily);\n  const sectionLocatorInstructionText = sectionLocatorInstruction(\n    extraTextFromUserPrompt(sectionDescription),\n  );\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: screenshotBase64,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: sectionLocatorInstructionText,\n        },\n      ],\n    },\n  ];\n\n  if (typeof sectionDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: sectionDescription.images,\n      convertHttpImage2Base64: sectionDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  let result: Awaited<\n    ReturnType<typeof callAIWithObjectResponse<AISectionLocatorResponse>>\n  >;\n  try {\n    result = await callAIWithObjectResponse<AISectionLocatorResponse>(\n      msgs,\n      modelConfig,\n    );\n  } catch (callError) {\n    // Return error with usage and rawResponse if available\n    const errorMessage =\n      callError instanceof Error ? callError.message : String(callError);\n    const rawResponse =\n      callError instanceof AIResponseParseError\n        ? callError.rawResponse\n        : errorMessage;\n    const usage =\n      callError instanceof AIResponseParseError ? callError.usage : undefined;\n    return {\n      rect: undefined,\n      imageBase64: undefined,\n      error: `AI call error: ${errorMessage}`,\n      rawResponse,\n      usage,\n    };\n  }\n\n  let sectionRect: Rect | undefined;\n  const sectionBbox = result.content.bbox;\n  if (sectionBbox) {\n    const targetRect = adaptBboxToRect(\n      sectionBbox,\n      context.size.width,\n      context.size.height,\n      0,\n      0,\n      context.size.width,\n      context.size.height,\n      modelFamily,\n    );\n    debugSection('original targetRect %j', targetRect);\n\n    const referenceBboxList = result.content.references_bbox || [];\n    debugSection('referenceBboxList %j', referenceBboxList);\n\n    const referenceRects = referenceBboxList\n      .filter((bbox) => Array.isArray(bbox))\n      .map((bbox) => {\n        return adaptBboxToRect(\n          bbox,\n          context.size.width,\n          context.size.height,\n          0,\n          0,\n          context.size.width,\n          context.size.height,\n          modelFamily,\n        );\n      });\n    debugSection('referenceRects %j', referenceRects);\n\n    // merge the sectionRect and referenceRects\n    const mergedRect = mergeRects([targetRect, ...referenceRects]);\n    debugSection('mergedRect %j', mergedRect);\n\n    // expand search area to at least 200 x 200\n    sectionRect = expandSearchArea(mergedRect, context.size, modelFamily);\n    debugSection('expanded sectionRect %j', sectionRect);\n  }\n\n  let imageBase64 = screenshotBase64;\n  if (sectionRect) {\n    const croppedResult = await cropByRect(\n      screenshotBase64,\n      sectionRect,\n      modelFamily === 'qwen2.5-vl',\n    );\n    imageBase64 = croppedResult.imageBase64;\n    sectionRect.width = croppedResult.width;\n    sectionRect.height = croppedResult.height;\n  }\n\n  return {\n    rect: sectionRect,\n    imageBase64,\n    error: result.content.error,\n    rawResponse: JSON.stringify(result.content),\n    usage: result.usage,\n  };\n}\n\nexport async function AiExtractElementInfo<T>(options: {\n  dataQuery: string | Record<string, string>;\n  multimodalPrompt?: TMultimodalPrompt;\n  context: UIContext;\n  pageDescription?: string;\n  extractOption?: ServiceExtractOption;\n  modelConfig: IModelConfig;\n}) {\n  const { dataQuery, context, extractOption, multimodalPrompt, modelConfig } =\n    options;\n  const systemPrompt = systemPromptToExtract();\n  const screenshotBase64 = context.screenshot.base64;\n\n  const extractDataPromptText = extractDataQueryPrompt(\n    options.pageDescription || '',\n    dataQuery,\n  );\n\n  const userContent: ChatCompletionUserMessageParam['content'] = [];\n\n  if (extractOption?.screenshotIncluded !== false) {\n    userContent.push({\n      type: 'image_url',\n      image_url: {\n        url: screenshotBase64,\n        detail: 'high',\n      },\n    });\n  }\n\n  userContent.push({\n    type: 'text',\n    text: extractDataPromptText,\n  });\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userContent,\n    },\n  ];\n\n  if (multimodalPrompt) {\n    const addOns = await promptsToChatParam({\n      images: multimodalPrompt.images,\n      convertHttpImage2Base64: multimodalPrompt.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const {\n    content: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAI(msgs, modelConfig);\n\n  // Parse XML response to JSON object\n  let parseResult: AIDataExtractionResponse<T>;\n  try {\n    parseResult = parseXMLExtractionResponse<T>(rawResponse);\n  } catch (parseError) {\n    // Throw AIResponseParseError with usage and rawResponse preserved\n    const errorMessage =\n      parseError instanceof Error ? parseError.message : String(parseError);\n    throw new AIResponseParseError(\n      `XML parse error: ${errorMessage}`,\n      rawResponse,\n      usage,\n    );\n  }\n\n  return {\n    parseResult,\n    rawResponse,\n    usage,\n    reasoning_content,\n  };\n}\n\nexport async function AiJudgeOrderSensitive(\n  description: string,\n  callAIFn: typeof callAIWithObjectResponse<{ isOrderSensitive: boolean }>,\n  modelConfig: IModelConfig,\n): Promise<{\n  isOrderSensitive: boolean;\n  usage?: AIUsageInfo;\n}> {\n  const systemPrompt = systemPromptToJudgeOrderSensitive();\n  const userPrompt = orderSensitiveJudgePrompt(description);\n\n  const msgs: AIArgs = [\n    { role: 'system', content: systemPrompt },\n    {\n      role: 'user',\n      content: userPrompt,\n    },\n  ];\n\n  const result = await callAIFn(msgs, modelConfig);\n\n  return {\n    isOrderSensitive: result.content.isOrderSensitive ?? false,\n    usage: result.usage,\n  };\n}\n"],"names":["debugInspect","getDebug","debugSection","extraTextFromUserPrompt","prompt","promptsToChatParam","multimodalPrompt","msgs","item","base64","preProcessImageUrl","AiLocateElement","options","context","targetElementDescription","callAIFn","modelConfig","modelFamily","screenshotBase64","assert","targetElementDescriptionText","userInstructionPrompt","findElementPrompt","systemPrompt","isAutoGLM","getAutoGLMLocatePrompt","systemPromptToLocateElement","imagePayload","imageWidth","imageHeight","originalImageWidth","originalImageHeight","paddedResult","paddingToMatchBlockByBase64","addOns","rawResponseContent","usage","callAIWithStringResponse","parsed","parseAutoGLMLocateResponse","resRect","matchedElements","errors","x","y","pixelX","Math","pixelY","bboxSize","x1","y1","x2","y2","rectCenter","element","generateElementByPosition","res","callError","errorMessage","Error","String","rawResponse","AIResponseParseError","undefined","JSON","Array","adaptBboxToRect","e","msg","AiLocateSection","sectionDescription","systemPromptToLocateSection","sectionLocatorInstructionText","sectionLocatorInstruction","result","callAIWithObjectResponse","sectionRect","sectionBbox","targetRect","referenceBboxList","referenceRects","bbox","mergedRect","mergeRects","expandSearchArea","imageBase64","croppedResult","cropByRect","AiExtractElementInfo","dataQuery","extractOption","systemPromptToExtract","extractDataPromptText","extractDataQueryPrompt","userContent","reasoning_content","callAI","parseResult","parseXMLExtractionResponse","parseError","AiJudgeOrderSensitive","description","systemPromptToJudgeOrderSensitive","userPrompt","orderSensitiveJudgePrompt"],"mappings":";;;;;;;;;;;;;AAyDA,MAAMA,eAAeC,SAAS;AAC9B,MAAMC,eAAeD,SAAS;AAE9B,MAAME,0BAA0B,CAACC;IAC/B,IAAI,AAAkB,YAAlB,OAAOA,QACT,OAAOA;IAEP,OAAOA,OAAO,MAAM;AAExB;AAEA,MAAMC,qBAAqB,OACzBC;IAEA,MAAMC,OAAyC,EAAE;IACjD,IAAID,kBAAkB,QAAQ,QAAQ;QACpCC,KAAK,IAAI,CAAC;YACR,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM;gBACR;aACD;QACH;QAEA,KAAK,MAAMC,QAAQF,iBAAiB,MAAM,CAAE;YAC1C,MAAMG,SAAS,MAAMC,mBACnBF,KAAK,GAAG,EACR,CAAC,CAACF,iBAAiB,uBAAuB;YAG5CC,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,MAAM,CAAC,mCAAmC,EAAEC,KAAK,IAAI,CAAC,EAAE,CAAC;oBAC3D;iBACD;YACH;YAEAD,KAAK,IAAI,CAAC;gBACR,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKE;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;QACF;IACF;IACA,OAAOF;AACT;AAEO,eAAeI,gBAAgBC,OAQrC;IAUC,MAAM,EAAEC,OAAO,EAAEC,wBAAwB,EAAEC,QAAQ,EAAEC,WAAW,EAAE,GAAGJ;IACrE,MAAM,EAAEK,WAAW,EAAE,GAAGD;IACxB,MAAME,mBAAmBL,QAAQ,UAAU,CAAC,MAAM;IAElDM,OACEL,0BACA;IAEF,MAAMM,+BAA+BjB,wBACnCW;IAEF,MAAMO,wBAAwBC,kBAAkBF;IAChD,MAAMG,eAAeC,UAAUP,eAC3BQ,uBAAuBR,eACvBS,4BAA4BT;IAEhC,IAAIU,eAAeT;IACnB,IAAIU,aAAaf,QAAQ,IAAI,CAAC,KAAK;IACnC,IAAIgB,cAAchB,QAAQ,IAAI,CAAC,MAAM;IACrC,IAAIiB,qBAAqBF;IACzB,IAAIG,sBAAsBF;IAE1B,IAAIjB,QAAQ,YAAY,EAAE;QACxBO,OACEP,QAAQ,YAAY,CAAC,IAAI,EACzB;QAEFO,OACEP,QAAQ,YAAY,CAAC,WAAW,EAChC;QAGFe,eAAef,QAAQ,YAAY,CAAC,WAAW;QAC/CgB,aAAahB,QAAQ,YAAY,CAAC,IAAI,EAAE;QACxCiB,cAAcjB,QAAQ,YAAY,CAAC,IAAI,EAAE;QACzCkB,qBAAqBF;QACrBG,sBAAsBF;IACxB,OAAO,IAAIZ,AAAgB,iBAAhBA,aAA8B;QACvC,MAAMe,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAMzB,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKI;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMH,UAAUP,eACZ,CAAC,KAAK,EAAEI,uBAAuB,GAC/BA;gBACN;aACD;QACH;KACD;IAED,IAAI,AAAoC,YAApC,OAAOP,0BAAuC;QAChD,MAAMoB,SAAS,MAAM7B,mBAAmB;YACtC,QAAQS,yBAAyB,MAAM;YACvC,yBAAyBA,yBAAyB,uBAAuB;QAC3E;QACAP,KAAK,IAAI,IAAI2B;IACf;IAEA,IAAIV,UAAUP,cAAc;QAC1B,MAAM,EAAE,SAASkB,kBAAkB,EAAEC,KAAK,EAAE,GAC1C,MAAMC,yBAAyB9B,MAAMS;QAEvChB,aAAa,yBAAyBmC;QAEtC,MAAMG,SAASC,2BAA2BJ;QAE1CnC,aAAa,sBAAsBsC,OAAO,KAAK;QAC/CtC,aAAa,yBAAyBsC,OAAO,WAAW;QAExD,IAAIE;QACJ,IAAIC,kBAAyC,EAAE;QAC/C,IAAIC,SAAmB,EAAE;QAEzB,IAAIJ,OAAO,KAAK,IAAI,CAACA,OAAO,WAAW,EAAE;YACvCI,SAAS;gBAACJ,OAAO,KAAK,IAAI;aAAoC;YAC9DtC,aAAa,yBAAyB0C,MAAM,CAAC,EAAE;QACjD,OAAO;YACL,MAAM,EAAEC,CAAC,EAAEC,CAAC,EAAE,GAAGN,OAAO,WAAW;YAEnCtC,aAAa,iCAAiC;gBAAE2C;gBAAGC;YAAE;YAIrD,MAAMC,SAASC,KAAK,KAAK,CAAEH,IAAIf,aAAc;YAC7C,MAAMmB,SAASD,KAAK,KAAK,CAAEF,IAAIf,cAAe;YAE9C7B,aAAa,+BAA+B;gBAAE6C;gBAAQE;YAAO;YAG7D,MAAMC,WAAW;YACjB,MAAMC,KAAKH,KAAK,GAAG,CAACD,SAASG,WAAW,GAAG;YAC3C,MAAME,KAAKJ,KAAK,GAAG,CAACC,SAASC,WAAW,GAAG;YAC3C,MAAMG,KAAKL,KAAK,GAAG,CAACD,SAASG,WAAW,GAAGpB;YAC3C,MAAMwB,KAAKN,KAAK,GAAG,CAACC,SAASC,WAAW,GAAGnB;YAG3CW,UAAU;gBACR,MAAMS;gBACN,KAAKC;gBACL,OAAOC,KAAKF;gBACZ,QAAQG,KAAKF;YACf;YAGA,IAAItC,QAAQ,YAAY,EAAE,MAAM;gBAC9B4B,QAAQ,IAAI,IAAI5B,QAAQ,YAAY,CAAC,IAAI,CAAC,IAAI;gBAC9C4B,QAAQ,GAAG,IAAI5B,QAAQ,YAAY,CAAC,IAAI,CAAC,GAAG;YAC9C;YAEAZ,aAAa,qBAAqBwC;YAElC,MAAMa,aAAa;gBACjB,GAAGb,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YAEA,MAAMc,UAA+BC,0BACnCF,YACAjC;YAGF,IAAIkC,SACFb,kBAAkB;gBAACa;aAAQ;QAE/B;QAEA,OAAO;YACL,MAAMd;YACN,aAAa;gBACX,UAAUC;gBACVC;YACF;YACA,aAAaP;YACbC;YACA,mBAAmBE,OAAO,KAAK;QACjC;IACF;IAEA,IAAIkB;IACJ,IAAI;QACFA,MAAM,MAAMzC,SAASR,MAAMS;IAC7B,EAAE,OAAOyC,WAAW;QAElB,MAAMC,eACJD,qBAAqBE,QAAQF,UAAU,OAAO,GAAGG,OAAOH;QAC1D,MAAMI,cACJJ,qBAAqBK,uBACjBL,UAAU,WAAW,GACrBC;QACN,MAAMtB,QACJqB,qBAAqBK,uBAAuBL,UAAU,KAAK,GAAGM;QAChE,OAAO;YACL,MAAMA;YACN,aAAa;gBACX,UAAU,EAAE;gBACZ,QAAQ;oBAAC,CAAC,eAAe,EAAEL,cAAc;iBAAC;YAC5C;YACAG;YACAzB;YACA,mBAAmB2B;QACrB;IACF;IAEA,MAAMF,cAAcG,KAAK,SAAS,CAACR,IAAI,OAAO;IAE9C,IAAIhB;IACJ,IAAIC,kBAAyC,EAAE;IAC/C,IAAIC,SACF,YAAYc,IAAI,OAAO,GAAGA,IAAI,OAAO,CAAC,MAAM,GAAG,EAAE;IACnD,IAAI;QACF,IACE,UAAUA,IAAI,OAAO,IACrBS,MAAM,OAAO,CAACT,IAAI,OAAO,CAAC,IAAI,KAC9BA,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,IAAI,GAC3B;YACAhB,UAAU0B,gBACRV,IAAI,OAAO,CAAC,IAAI,EAChB5B,YACAC,aACAjB,QAAQ,YAAY,EAAE,MAAM,MAC5BA,QAAQ,YAAY,EAAE,MAAM,KAC5BkB,oBACAC,qBACAd;YAGFjB,aAAa,WAAWwC;YAExB,MAAMa,aAAa;gBACjB,GAAGb,QAAQ,IAAI,GAAGA,QAAQ,KAAK,GAAG;gBAClC,GAAGA,QAAQ,GAAG,GAAGA,QAAQ,MAAM,GAAG;YACpC;YAEA,MAAMc,UAA+BC,0BACnCF,YACAjC;YAEFsB,SAAS,EAAE;YAEX,IAAIY,SACFb,kBAAkB;gBAACa;aAAQ;QAE/B;IACF,EAAE,OAAOa,GAAG;QACV,MAAMC,MACJD,aAAaR,QACT,CAAC,sBAAsB,EAAEQ,EAAE,OAAO,EAAE,GACpC;QACN,IAAI,AAACzB,UAAUA,QAAQ,WAAW,GAGhCA,OAAO,IAAI,CAAC,CAAC,CAAC,EAAE0B,IAAI,CAAC,CAAC;aAFtB1B,SAAS;YAAC0B;SAAI;IAIlB;IAEA,OAAO;QACL,MAAM5B;QACN,aAAa;YACX,UAAUC;YACV,QAAQC;QACV;QACAmB;QACA,OAAOL,IAAI,KAAK;QAChB,mBAAmBA,IAAI,iBAAiB;IAC1C;AACF;AAEO,eAAea,gBAAgBzD,OAIrC;IAOC,MAAM,EAAEC,OAAO,EAAEyD,kBAAkB,EAAEtD,WAAW,EAAE,GAAGJ;IACrD,MAAM,EAAEK,WAAW,EAAE,GAAGD;IACxB,MAAME,mBAAmBL,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAMU,eAAegD,4BAA4BtD;IACjD,MAAMuD,gCAAgCC,0BACpCtE,wBAAwBmE;IAE1B,MAAM/D,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKL;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAMsD;gBACR;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOF,oBAAiC;QAC1C,MAAMpC,SAAS,MAAM7B,mBAAmB;YACtC,QAAQiE,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACA/D,KAAK,IAAI,IAAI2B;IACf;IAEA,IAAIwC;IAGJ,IAAI;QACFA,SAAS,MAAMC,yBACbpE,MACAS;IAEJ,EAAE,OAAOyC,WAAW;QAElB,MAAMC,eACJD,qBAAqBE,QAAQF,UAAU,OAAO,GAAGG,OAAOH;QAC1D,MAAMI,cACJJ,qBAAqBK,uBACjBL,UAAU,WAAW,GACrBC;QACN,MAAMtB,QACJqB,qBAAqBK,uBAAuBL,UAAU,KAAK,GAAGM;QAChE,OAAO;YACL,MAAMA;YACN,aAAaA;YACb,OAAO,CAAC,eAAe,EAAEL,cAAc;YACvCG;YACAzB;QACF;IACF;IAEA,IAAIwC;IACJ,MAAMC,cAAcH,OAAO,OAAO,CAAC,IAAI;IACvC,IAAIG,aAAa;QACf,MAAMC,aAAaZ,gBACjBW,aACAhE,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAEFf,aAAa,0BAA0B4E;QAEvC,MAAMC,oBAAoBL,OAAO,OAAO,CAAC,eAAe,IAAI,EAAE;QAC9DxE,aAAa,wBAAwB6E;QAErC,MAAMC,iBAAiBD,kBACpB,MAAM,CAAC,CAACE,OAAShB,MAAM,OAAO,CAACgB,OAC/B,GAAG,CAAC,CAACA,OACGf,gBACLe,MACApE,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnB,GACA,GACAA,QAAQ,IAAI,CAAC,KAAK,EAClBA,QAAQ,IAAI,CAAC,MAAM,EACnBI;QAGNf,aAAa,qBAAqB8E;QAGlC,MAAME,aAAaC,WAAW;YAACL;eAAeE;SAAe;QAC7D9E,aAAa,iBAAiBgF;QAG9BN,cAAcQ,iBAAiBF,YAAYrE,QAAQ,IAAI,EAAEI;QACzDf,aAAa,2BAA2B0E;IAC1C;IAEA,IAAIS,cAAcnE;IAClB,IAAI0D,aAAa;QACf,MAAMU,gBAAgB,MAAMC,WAC1BrE,kBACA0D,aACA3D,AAAgB,iBAAhBA;QAEFoE,cAAcC,cAAc,WAAW;QACvCV,YAAY,KAAK,GAAGU,cAAc,KAAK;QACvCV,YAAY,MAAM,GAAGU,cAAc,MAAM;IAC3C;IAEA,OAAO;QACL,MAAMV;QACNS;QACA,OAAOX,OAAO,OAAO,CAAC,KAAK;QAC3B,aAAaV,KAAK,SAAS,CAACU,OAAO,OAAO;QAC1C,OAAOA,OAAO,KAAK;IACrB;AACF;AAEO,eAAec,qBAAwB5E,OAO7C;IACC,MAAM,EAAE6E,SAAS,EAAE5E,OAAO,EAAE6E,aAAa,EAAEpF,gBAAgB,EAAEU,WAAW,EAAE,GACxEJ;IACF,MAAMW,eAAeoE;IACrB,MAAMzE,mBAAmBL,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM+E,wBAAwBC,uBAC5BjF,QAAQ,eAAe,IAAI,IAC3B6E;IAGF,MAAMK,cAAyD,EAAE;IAEjE,IAAIJ,eAAe,uBAAuB,OACxCI,YAAY,IAAI,CAAC;QACf,MAAM;QACN,WAAW;YACT,KAAK5E;YACL,QAAQ;QACV;IACF;IAGF4E,YAAY,IAAI,CAAC;QACf,MAAM;QACN,MAAMF;IACR;IAEA,MAAMrF,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASuE;QACX;KACD;IAED,IAAIxF,kBAAkB;QACpB,MAAM4B,SAAS,MAAM7B,mBAAmB;YACtC,QAAQC,iBAAiB,MAAM;YAC/B,yBAAyBA,iBAAiB,uBAAuB;QACnE;QACAC,KAAK,IAAI,IAAI2B;IACf;IAEA,MAAM,EACJ,SAAS2B,WAAW,EACpBzB,KAAK,EACL2D,iBAAiB,EAClB,GAAG,MAAMC,OAAOzF,MAAMS;IAGvB,IAAIiF;IACJ,IAAI;QACFA,cAAcC,2BAA8BrC;IAC9C,EAAE,OAAOsC,YAAY;QAEnB,MAAMzC,eACJyC,sBAAsBxC,QAAQwC,WAAW,OAAO,GAAGvC,OAAOuC;QAC5D,MAAM,IAAIrC,qBACR,CAAC,iBAAiB,EAAEJ,cAAc,EAClCG,aACAzB;IAEJ;IAEA,OAAO;QACL6D;QACApC;QACAzB;QACA2D;IACF;AACF;AAEO,eAAeK,sBACpBC,WAAmB,EACnBtF,QAAwE,EACxEC,WAAyB;IAKzB,MAAMO,eAAe+E;IACrB,MAAMC,aAAaC,0BAA0BH;IAE7C,MAAM9F,OAAe;QACnB;YAAE,MAAM;YAAU,SAASgB;QAAa;QACxC;YACE,MAAM;YACN,SAASgF;QACX;KACD;IAED,MAAM7B,SAAS,MAAM3D,SAASR,MAAMS;IAEpC,OAAO;QACL,kBAAkB0D,OAAO,OAAO,CAAC,gBAAgB,IAAI;QACrD,OAAOA,OAAO,KAAK;IACrB;AACF"}

package/dist/es/ai-model/llm-planning.mjs

@@ -4,7 +4,7 @@ import { assert } from "@midscene/shared/utils";
 import { buildYamlFlowFromPlans, fillBboxParam, findAllMidsceneLocatorField } from "../common.mjs";
 import { systemPromptToTaskPlanning } from "./prompt/llm-planning.mjs";
 import { extractXMLTag, parseMarkFinishedIndexes, parseSubGoalsFromXML } from "./prompt/util.mjs";
-import { callAI, safeParseJson } from "./service-caller/index.mjs";
+import { AIResponseParseError, callAI, safeParseJson } from "./service-caller/index.mjs";
 const debug = getDebug('planning');
 function parseXMLPlanningResponse(xmlString, modelFamily) {
     const thought = extractXMLTag(xmlString, 'thought');
@@ -156,7 +156,13 @@ async function plan(userInstruction, opts) {
     const { content: rawResponse, usage, reasoning_content } = await callAI(msgs, modelConfig, {
         deepThink: 'unset' === opts.deepThink ? void 0 : opts.deepThink
     });
-    const planFromAI = parseXMLPlanningResponse(rawResponse, modelFamily);
+    let planFromAI;
+    try {
+        planFromAI = parseXMLPlanningResponse(rawResponse, modelFamily);
+    } catch (parseError) {
+        const errorMessage = parseError instanceof Error ? parseError.message : String(parseError);
+        throw new AIResponseParseError(`XML parse error: ${errorMessage}`, rawResponse, usage);
+    }
     if (planFromAI.action && void 0 !== planFromAI.finalizeSuccess) {
         console.warn('Planning response included both an action and complete-goal; ignoring complete-goal output.');
         planFromAI.finalizeMessage = void 0;

package/dist/es/ai-model/llm-planning.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/llm-planning.mjs","sources":["../../../src/ai-model/llm-planning.ts"],"sourcesContent":["import type {\n  DeepThinkOption,\n  DeviceAction,\n  InterfaceType,\n  PlanningAIResponse,\n  RawResponsePlanningAIResponse,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig, TModelFamily } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport {\n  buildYamlFlowFromPlans,\n  fillBboxParam,\n  findAllMidsceneLocatorField,\n} from '../common';\nimport type { ConversationHistory } from './conversation-history';\nimport { systemPromptToTaskPlanning } from './prompt/llm-planning';\nimport {\n  extractXMLTag,\n  parseMarkFinishedIndexes,\n  parseSubGoalsFromXML,\n} from './prompt/util';\nimport { callAI } from './service-caller/index';\nimport { safeParseJson } from './service-caller/index';\n\nconst debug = getDebug('planning');\n\n/**\n * Parse XML response from LLM and convert to RawResponsePlanningAIResponse\n */\nexport function parseXMLPlanningResponse(\n  xmlString: string,\n  modelFamily: TModelFamily | undefined,\n): RawResponsePlanningAIResponse {\n  const thought = extractXMLTag(xmlString, 'thought');\n  const note = extractXMLTag(xmlString, 'note');\n  const log = extractXMLTag(xmlString, 'log') || '';\n  const error = extractXMLTag(xmlString, 'error');\n  const actionType = extractXMLTag(xmlString, 'action-type');\n  const actionParamStr = extractXMLTag(xmlString, 'action-param-json');\n\n  // Parse complete-goal tag with success attribute\n  const completeGoalRegex =\n    /<complete-goal\\s+success=\"(true|false)\">([\\s\\S]*?)<\\/complete-goal>/i;\n  const completeGoalMatch = xmlString.match(completeGoalRegex);\n  let finalizeMessage: string | undefined;\n  let finalizeSuccess: boolean | undefined;\n\n  if (completeGoalMatch) {\n    finalizeSuccess = completeGoalMatch[1] === 'true';\n    finalizeMessage = completeGoalMatch[2]?.trim() || undefined;\n  }\n\n  // Parse sub-goal related tags\n  const updatePlanContent = extractXMLTag(xmlString, 'update-plan-content');\n  const markSubGoalDone = extractXMLTag(xmlString, 'mark-sub-goal-done');\n\n  const updateSubGoals = updatePlanContent\n    ? parseSubGoalsFromXML(updatePlanContent)\n    : undefined;\n  const markFinishedIndexes = markSubGoalDone\n    ? parseMarkFinishedIndexes(markSubGoalDone)\n    : undefined;\n\n  // Parse action\n  let action: any = null;\n  if (actionType && actionType.toLowerCase() !== 'null') {\n    const type = actionType.trim();\n    let param: any = undefined;\n\n    if (actionParamStr) {\n      try {\n        // Parse the JSON string in action-param-json\n        param = safeParseJson(actionParamStr, modelFamily);\n      } catch (e) {\n        throw new Error(`Failed to parse action-param-json: ${e}`);\n      }\n    }\n\n    action = {\n      type,\n      ...(param !== undefined ? { param } : {}),\n    };\n  }\n\n  return {\n    ...(thought ? { thought } : {}),\n    ...(note ? { note } : {}),\n    log,\n    ...(error ? { error } : {}),\n    action,\n    ...(finalizeMessage !== undefined ? { finalizeMessage } : {}),\n    ...(finalizeSuccess !== undefined ? { finalizeSuccess } : {}),\n    ...(updateSubGoals?.length ? { updateSubGoals } : {}),\n    ...(markFinishedIndexes?.length ? { markFinishedIndexes } : {}),\n  };\n}\n\nexport async function plan(\n  userInstruction: string,\n  opts: {\n    context: UIContext;\n    interfaceType: InterfaceType;\n    actionSpace: DeviceAction<any>[];\n    actionContext?: string;\n    modelConfig: IModelConfig;\n    conversationHistory: ConversationHistory;\n    includeBbox: boolean;\n    imagesIncludeCount?: number;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<PlanningAIResponse> {\n  const { context, modelConfig, conversationHistory } = opts;\n  const { size } = context;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const { modelFamily } = modelConfig;\n\n  // Only enable sub-goals when deepThink is true\n  const includeSubGoals = opts.deepThink === true;\n\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    modelFamily,\n    includeBbox: opts.includeBbox,\n    includeThought: true, // always include thought\n    includeSubGoals,\n  });\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = size.width;\n  let imageHeight = size.height;\n  const rightLimit = imageWidth;\n  const bottomLimit = imageHeight;\n\n  // Process image based on VL mode requirements\n  if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const actionContext = opts.actionContext\n    ? `<high_priority_knowledge>${opts.actionContext}</high_priority_knowledge>\\n`\n    : '';\n\n  const instruction: ChatCompletionMessageParam[] = [\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${actionContext}<user_instruction>${userInstruction}</user_instruction>`,\n        },\n      ],\n    },\n  ];\n\n  let latestFeedbackMessage: ChatCompletionMessageParam;\n\n  // Build sub-goal status text to include in the message (only when deepThink is enabled)\n  const subGoalsText = includeSubGoals\n    ? conversationHistory.subGoalsToText()\n    : '';\n  const subGoalsSection = subGoalsText ? `\\n\\n${subGoalsText}` : '';\n\n  // Build notes text to include in the message\n  const notesText = conversationHistory.notesToText();\n  const notesSection = notesText ? `\\n\\n${notesText}` : '';\n\n  if (conversationHistory.pendingFeedbackMessage) {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${conversationHistory.pendingFeedbackMessage}. The last screenshot is attached. Please going on according to the instruction.${notesSection}${subGoalsSection}`,\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n\n    conversationHistory.resetPendingFeedbackMessageIfExists();\n  } else {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `this is the latest screenshot${notesSection}${subGoalsSection}`,\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n  }\n  conversationHistory.append(latestFeedbackMessage);\n\n  // Compress history if it exceeds the threshold to avoid context overflow\n  conversationHistory.compressHistory(50, 20);\n\n  const historyLog = conversationHistory.snapshot(opts.imagesIncludeCount);\n\n  const msgs: ChatCompletionMessageParam[] = [\n    { role: 'system', content: systemPrompt },\n    ...instruction,\n    ...historyLog,\n  ];\n\n  const {\n    content: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAI(msgs, modelConfig, {\n    deepThink: opts.deepThink === 'unset' ? undefined : opts.deepThink,\n  });\n\n  // Parse XML response to JSON object\n  const planFromAI = parseXMLPlanningResponse(rawResponse, modelFamily);\n\n  if (planFromAI.action && planFromAI.finalizeSuccess !== undefined) {\n    console.warn(\n      'Planning response included both an action and complete-goal; ignoring complete-goal output.',\n    );\n    planFromAI.finalizeMessage = undefined;\n    planFromAI.finalizeSuccess = undefined;\n  }\n\n  const actions = planFromAI.action ? [planFromAI.action] : [];\n  let shouldContinuePlanning = true;\n\n  // Check if goal is completed via complete-goal tag\n  if (planFromAI.finalizeSuccess !== undefined) {\n    debug('goal completed via complete-goal tag, stop planning');\n    shouldContinuePlanning = false;\n    // Mark all sub-goals as finished when goal is completed (only when deepThink is enabled)\n    if (includeSubGoals) {\n      conversationHistory.markAllSubGoalsFinished();\n    }\n  }\n\n  const returnValue: PlanningAIResponse = {\n    ...planFromAI,\n    actions,\n    rawResponse,\n    usage,\n    reasoning_content,\n    yamlFlow: buildYamlFlowFromPlans(actions, opts.actionSpace),\n    shouldContinuePlanning,\n  };\n\n  assert(planFromAI, \"can't get plans from AI\");\n\n  actions.forEach((action) => {\n    const type = action.type;\n    const actionInActionSpace = opts.actionSpace.find(\n      (action) => action.name === type,\n    );\n\n    debug('actionInActionSpace matched', actionInActionSpace);\n    const locateFields = actionInActionSpace\n      ? findAllMidsceneLocatorField(actionInActionSpace.paramSchema)\n      : [];\n\n    debug('locateFields', locateFields);\n\n    locateFields.forEach((field) => {\n      const locateResult = action.param[field];\n      if (locateResult && modelFamily !== undefined) {\n        // Always use model family to fill bbox parameters\n        action.param[field] = fillBboxParam(\n          locateResult,\n          imageWidth,\n          imageHeight,\n          rightLimit,\n          bottomLimit,\n          modelFamily,\n        );\n      }\n    });\n  });\n\n  // Update sub-goals in conversation history based on response (only when deepThink is enabled)\n  if (includeSubGoals) {\n    if (planFromAI.updateSubGoals?.length) {\n      conversationHistory.setSubGoals(planFromAI.updateSubGoals);\n    }\n    if (planFromAI.markFinishedIndexes?.length) {\n      for (const index of planFromAI.markFinishedIndexes) {\n        conversationHistory.markSubGoalFinished(index);\n      }\n    }\n  }\n\n  // Append note to conversation history if present\n  if (planFromAI.note) {\n    conversationHistory.appendNote(planFromAI.note);\n  }\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: [\n      {\n        type: 'text',\n        text: rawResponse,\n      },\n    ],\n  });\n\n  return returnValue;\n}\n"],"names":["debug","getDebug","parseXMLPlanningResponse","xmlString","modelFamily","thought","extractXMLTag","note","log","error","actionType","actionParamStr","completeGoalRegex","completeGoalMatch","finalizeMessage","finalizeSuccess","undefined","updatePlanContent","markSubGoalDone","updateSubGoals","parseSubGoalsFromXML","markFinishedIndexes","parseMarkFinishedIndexes","action","type","param","safeParseJson","e","Error","plan","userInstruction","opts","context","modelConfig","conversationHistory","size","screenshotBase64","includeSubGoals","systemPrompt","systemPromptToTaskPlanning","imagePayload","imageWidth","imageHeight","rightLimit","bottomLimit","paddedResult","paddingToMatchBlockByBase64","actionContext","instruction","latestFeedbackMessage","subGoalsText","subGoalsSection","notesText","notesSection","historyLog","msgs","rawResponse","usage","reasoning_content","callAI","planFromAI","console","actions","shouldContinuePlanning","returnValue","buildYamlFlowFromPlans","assert","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam","index"],"mappings":";;;;;;;AA4BA,MAAMA,QAAQC,SAAS;AAKhB,SAASC,yBACdC,SAAiB,EACjBC,WAAqC;IAErC,MAAMC,UAAUC,cAAcH,WAAW;IACzC,MAAMI,OAAOD,cAAcH,WAAW;IACtC,MAAMK,MAAMF,cAAcH,WAAW,UAAU;IAC/C,MAAMM,QAAQH,cAAcH,WAAW;IACvC,MAAMO,aAAaJ,cAAcH,WAAW;IAC5C,MAAMQ,iBAAiBL,cAAcH,WAAW;IAGhD,MAAMS,oBACJ;IACF,MAAMC,oBAAoBV,UAAU,KAAK,CAACS;IAC1C,IAAIE;IACJ,IAAIC;IAEJ,IAAIF,mBAAmB;QACrBE,kBAAkBF,AAAyB,WAAzBA,iBAAiB,CAAC,EAAE;QACtCC,kBAAkBD,iBAAiB,CAAC,EAAE,EAAE,UAAUG;IACpD;IAGA,MAAMC,oBAAoBX,cAAcH,WAAW;IACnD,MAAMe,kBAAkBZ,cAAcH,WAAW;IAEjD,MAAMgB,iBAAiBF,oBACnBG,qBAAqBH,qBACrBD;IACJ,MAAMK,sBAAsBH,kBACxBI,yBAAyBJ,mBACzBF;IAGJ,IAAIO,SAAc;IAClB,IAAIb,cAAcA,AAA6B,WAA7BA,WAAW,WAAW,IAAe;QACrD,MAAMc,OAAOd,WAAW,IAAI;QAC5B,IAAIe;QAEJ,IAAId,gBACF,IAAI;YAEFc,QAAQC,cAAcf,gBAAgBP;QACxC,EAAE,OAAOuB,GAAG;YACV,MAAM,IAAIC,MAAM,CAAC,mCAAmC,EAAED,GAAG;QAC3D;QAGFJ,SAAS;YACPC;YACA,GAAIC,AAAUT,WAAVS,QAAsB;gBAAEA;YAAM,IAAI,CAAC,CAAC;QAC1C;IACF;IAEA,OAAO;QACL,GAAIpB,UAAU;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAC9B,GAAIE,OAAO;YAAEA;QAAK,IAAI,CAAC,CAAC;QACxBC;QACA,GAAIC,QAAQ;YAAEA;QAAM,IAAI,CAAC,CAAC;QAC1Bc;QACA,GAAIT,AAAoBE,WAApBF,kBAAgC;YAAEA;QAAgB,IAAI,CAAC,CAAC;QAC5D,GAAIC,AAAoBC,WAApBD,kBAAgC;YAAEA;QAAgB,IAAI,CAAC,CAAC;QAC5D,GAAII,gBAAgB,SAAS;YAAEA;QAAe,IAAI,CAAC,CAAC;QACpD,GAAIE,qBAAqB,SAAS;YAAEA;QAAoB,IAAI,CAAC,CAAC;IAChE;AACF;AAEO,eAAeQ,KACpBC,eAAuB,EACvBC,IAUC;IAED,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAEC,mBAAmB,EAAE,GAAGH;IACtD,MAAM,EAAEI,IAAI,EAAE,GAAGH;IACjB,MAAMI,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM,EAAE5B,WAAW,EAAE,GAAG6B;IAGxB,MAAMI,kBAAkBN,AAAmB,SAAnBA,KAAK,SAAS;IAEtC,MAAMO,eAAe,MAAMC,2BAA2B;QACpD,aAAaR,KAAK,WAAW;QAC7B3B;QACA,aAAa2B,KAAK,WAAW;QAC7B,gBAAgB;QAChBM;IACF;IAEA,IAAIG,eAAeJ;IACnB,IAAIK,aAAaN,KAAK,KAAK;IAC3B,IAAIO,cAAcP,KAAK,MAAM;IAC7B,MAAMQ,aAAaF;IACnB,MAAMG,cAAcF;IAGpB,IAAItC,AAAgB,iBAAhBA,aAA8B;QAChC,MAAMyC,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAME,gBAAgBhB,KAAK,aAAa,GACpC,CAAC,yBAAyB,EAAEA,KAAK,aAAa,CAAC,4BAA4B,CAAC,GAC5E;IAEJ,MAAMiB,cAA4C;QAChD;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGD,cAAc,kBAAkB,EAAEjB,gBAAgB,mBAAmB,CAAC;gBACjF;aACD;QACH;KACD;IAED,IAAImB;IAGJ,MAAMC,eAAeb,kBACjBH,oBAAoB,cAAc,KAClC;IACJ,MAAMiB,kBAAkBD,eAAe,CAAC,IAAI,EAAEA,cAAc,GAAG;IAG/D,MAAME,YAAYlB,oBAAoB,WAAW;IACjD,MAAMmB,eAAeD,YAAY,CAAC,IAAI,EAAEA,WAAW,GAAG;IAEtD,IAAIlB,oBAAoB,sBAAsB,EAAE;QAC9Ce,wBAAwB;YACtB,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGf,oBAAoB,sBAAsB,CAAC,gFAAgF,EAAEmB,eAAeF,iBAAiB;gBACxK;gBACA;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKX;wBACL,QAAQ;oBACV;gBACF;aACD;QACH;QAEAN,oBAAoB,mCAAmC;IACzD,OACEe,wBAAwB;QACtB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAM,CAAC,6BAA6B,EAAEI,eAAeF,iBAAiB;YACxE;YACA;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKX;oBACL,QAAQ;gBACV;YACF;SACD;IACH;IAEFN,oBAAoB,MAAM,CAACe;IAG3Bf,oBAAoB,eAAe,CAAC,IAAI;IAExC,MAAMoB,aAAapB,oBAAoB,QAAQ,CAACH,KAAK,kBAAkB;IAEvE,MAAMwB,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASjB;QAAa;WACrCU;WACAM;KACJ;IAED,MAAM,EACJ,SAASE,WAAW,EACpBC,KAAK,EACLC,iBAAiB,EAClB,GAAG,MAAMC,OAAOJ,MAAMtB,aAAa;QAClC,WAAWF,AAAmB,YAAnBA,KAAK,SAAS,GAAef,SAAYe,KAAK,SAAS;IACpE;IAGA,MAAM6B,aAAa1D,yBAAyBsD,aAAapD;IAEzD,IAAIwD,WAAW,MAAM,IAAIA,AAA+B5C,WAA/B4C,WAAW,eAAe,EAAgB;QACjEC,QAAQ,IAAI,CACV;QAEFD,WAAW,eAAe,GAAG5C;QAC7B4C,WAAW,eAAe,GAAG5C;IAC/B;IAEA,MAAM8C,UAAUF,WAAW,MAAM,GAAG;QAACA,WAAW,MAAM;KAAC,GAAG,EAAE;IAC5D,IAAIG,yBAAyB;IAG7B,IAAIH,AAA+B5C,WAA/B4C,WAAW,eAAe,EAAgB;QAC5C5D,MAAM;QACN+D,yBAAyB;QAEzB,IAAI1B,iBACFH,oBAAoB,uBAAuB;IAE/C;IAEA,MAAM8B,cAAkC;QACtC,GAAGJ,UAAU;QACbE;QACAN;QACAC;QACAC;QACA,UAAUO,uBAAuBH,SAAS/B,KAAK,WAAW;QAC1DgC;IACF;IAEAG,OAAON,YAAY;IAEnBE,QAAQ,OAAO,CAAC,CAACvC;QACf,MAAMC,OAAOD,OAAO,IAAI;QACxB,MAAM4C,sBAAsBpC,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACR,SAAWA,OAAO,IAAI,KAAKC;QAG9BxB,MAAM,+BAA+BmE;QACrC,MAAMC,eAAeD,sBACjBE,4BAA4BF,oBAAoB,WAAW,IAC3D,EAAE;QAENnE,MAAM,gBAAgBoE;QAEtBA,aAAa,OAAO,CAAC,CAACE;YACpB,MAAMC,eAAehD,OAAO,KAAK,CAAC+C,MAAM;YACxC,IAAIC,gBAAgBnE,AAAgBY,WAAhBZ,aAElBmB,OAAO,KAAK,CAAC+C,MAAM,GAAGE,cACpBD,cACA9B,YACAC,aACAC,YACAC,aACAxC;QAGN;IACF;IAGA,IAAIiC,iBAAiB;QACnB,IAAIuB,WAAW,cAAc,EAAE,QAC7B1B,oBAAoB,WAAW,CAAC0B,WAAW,cAAc;QAE3D,IAAIA,WAAW,mBAAmB,EAAE,QAClC,KAAK,MAAMa,SAASb,WAAW,mBAAmB,CAChD1B,oBAAoB,mBAAmB,CAACuC;IAG9C;IAGA,IAAIb,WAAW,IAAI,EACjB1B,oBAAoB,UAAU,CAAC0B,WAAW,IAAI;IAGhD1B,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAMsB;YACR;SACD;IACH;IAEA,OAAOQ;AACT"}
+{"version":3,"file":"ai-model/llm-planning.mjs","sources":["../../../src/ai-model/llm-planning.ts"],"sourcesContent":["import type {\n  DeepThinkOption,\n  DeviceAction,\n  InterfaceType,\n  PlanningAIResponse,\n  RawResponsePlanningAIResponse,\n  UIContext,\n} from '@/types';\nimport type { IModelConfig, TModelFamily } from '@midscene/shared/env';\nimport { paddingToMatchBlockByBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport {\n  buildYamlFlowFromPlans,\n  fillBboxParam,\n  findAllMidsceneLocatorField,\n} from '../common';\nimport type { ConversationHistory } from './conversation-history';\nimport { systemPromptToTaskPlanning } from './prompt/llm-planning';\nimport {\n  extractXMLTag,\n  parseMarkFinishedIndexes,\n  parseSubGoalsFromXML,\n} from './prompt/util';\nimport {\n  AIResponseParseError,\n  callAI,\n  safeParseJson,\n} from './service-caller/index';\n\nconst debug = getDebug('planning');\n\n/**\n * Parse XML response from LLM and convert to RawResponsePlanningAIResponse\n */\nexport function parseXMLPlanningResponse(\n  xmlString: string,\n  modelFamily: TModelFamily | undefined,\n): RawResponsePlanningAIResponse {\n  const thought = extractXMLTag(xmlString, 'thought');\n  const note = extractXMLTag(xmlString, 'note');\n  const log = extractXMLTag(xmlString, 'log') || '';\n  const error = extractXMLTag(xmlString, 'error');\n  const actionType = extractXMLTag(xmlString, 'action-type');\n  const actionParamStr = extractXMLTag(xmlString, 'action-param-json');\n\n  // Parse complete-goal tag with success attribute\n  const completeGoalRegex =\n    /<complete-goal\\s+success=\"(true|false)\">([\\s\\S]*?)<\\/complete-goal>/i;\n  const completeGoalMatch = xmlString.match(completeGoalRegex);\n  let finalizeMessage: string | undefined;\n  let finalizeSuccess: boolean | undefined;\n\n  if (completeGoalMatch) {\n    finalizeSuccess = completeGoalMatch[1] === 'true';\n    finalizeMessage = completeGoalMatch[2]?.trim() || undefined;\n  }\n\n  // Parse sub-goal related tags\n  const updatePlanContent = extractXMLTag(xmlString, 'update-plan-content');\n  const markSubGoalDone = extractXMLTag(xmlString, 'mark-sub-goal-done');\n\n  const updateSubGoals = updatePlanContent\n    ? parseSubGoalsFromXML(updatePlanContent)\n    : undefined;\n  const markFinishedIndexes = markSubGoalDone\n    ? parseMarkFinishedIndexes(markSubGoalDone)\n    : undefined;\n\n  // Parse action\n  let action: any = null;\n  if (actionType && actionType.toLowerCase() !== 'null') {\n    const type = actionType.trim();\n    let param: any = undefined;\n\n    if (actionParamStr) {\n      try {\n        // Parse the JSON string in action-param-json\n        param = safeParseJson(actionParamStr, modelFamily);\n      } catch (e) {\n        throw new Error(`Failed to parse action-param-json: ${e}`);\n      }\n    }\n\n    action = {\n      type,\n      ...(param !== undefined ? { param } : {}),\n    };\n  }\n\n  return {\n    ...(thought ? { thought } : {}),\n    ...(note ? { note } : {}),\n    log,\n    ...(error ? { error } : {}),\n    action,\n    ...(finalizeMessage !== undefined ? { finalizeMessage } : {}),\n    ...(finalizeSuccess !== undefined ? { finalizeSuccess } : {}),\n    ...(updateSubGoals?.length ? { updateSubGoals } : {}),\n    ...(markFinishedIndexes?.length ? { markFinishedIndexes } : {}),\n  };\n}\n\nexport async function plan(\n  userInstruction: string,\n  opts: {\n    context: UIContext;\n    interfaceType: InterfaceType;\n    actionSpace: DeviceAction<any>[];\n    actionContext?: string;\n    modelConfig: IModelConfig;\n    conversationHistory: ConversationHistory;\n    includeBbox: boolean;\n    imagesIncludeCount?: number;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<PlanningAIResponse> {\n  const { context, modelConfig, conversationHistory } = opts;\n  const { size } = context;\n  const screenshotBase64 = context.screenshot.base64;\n\n  const { modelFamily } = modelConfig;\n\n  // Only enable sub-goals when deepThink is true\n  const includeSubGoals = opts.deepThink === true;\n\n  const systemPrompt = await systemPromptToTaskPlanning({\n    actionSpace: opts.actionSpace,\n    modelFamily,\n    includeBbox: opts.includeBbox,\n    includeThought: true, // always include thought\n    includeSubGoals,\n  });\n\n  let imagePayload = screenshotBase64;\n  let imageWidth = size.width;\n  let imageHeight = size.height;\n  const rightLimit = imageWidth;\n  const bottomLimit = imageHeight;\n\n  // Process image based on VL mode requirements\n  if (modelFamily === 'qwen2.5-vl') {\n    const paddedResult = await paddingToMatchBlockByBase64(imagePayload);\n    imageWidth = paddedResult.width;\n    imageHeight = paddedResult.height;\n    imagePayload = paddedResult.imageBase64;\n  }\n\n  const actionContext = opts.actionContext\n    ? `<high_priority_knowledge>${opts.actionContext}</high_priority_knowledge>\\n`\n    : '';\n\n  const instruction: ChatCompletionMessageParam[] = [\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${actionContext}<user_instruction>${userInstruction}</user_instruction>`,\n        },\n      ],\n    },\n  ];\n\n  let latestFeedbackMessage: ChatCompletionMessageParam;\n\n  // Build sub-goal status text to include in the message (only when deepThink is enabled)\n  const subGoalsText = includeSubGoals\n    ? conversationHistory.subGoalsToText()\n    : '';\n  const subGoalsSection = subGoalsText ? `\\n\\n${subGoalsText}` : '';\n\n  // Build notes text to include in the message\n  const notesText = conversationHistory.notesToText();\n  const notesSection = notesText ? `\\n\\n${notesText}` : '';\n\n  if (conversationHistory.pendingFeedbackMessage) {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `${conversationHistory.pendingFeedbackMessage}. The last screenshot is attached. Please going on according to the instruction.${notesSection}${subGoalsSection}`,\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n\n    conversationHistory.resetPendingFeedbackMessageIfExists();\n  } else {\n    latestFeedbackMessage = {\n      role: 'user',\n      content: [\n        {\n          type: 'text',\n          text: `this is the latest screenshot${notesSection}${subGoalsSection}`,\n        },\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n      ],\n    };\n  }\n  conversationHistory.append(latestFeedbackMessage);\n\n  // Compress history if it exceeds the threshold to avoid context overflow\n  conversationHistory.compressHistory(50, 20);\n\n  const historyLog = conversationHistory.snapshot(opts.imagesIncludeCount);\n\n  const msgs: ChatCompletionMessageParam[] = [\n    { role: 'system', content: systemPrompt },\n    ...instruction,\n    ...historyLog,\n  ];\n\n  const {\n    content: rawResponse,\n    usage,\n    reasoning_content,\n  } = await callAI(msgs, modelConfig, {\n    deepThink: opts.deepThink === 'unset' ? undefined : opts.deepThink,\n  });\n\n  // Parse XML response to JSON object, capture parsing errors\n  let planFromAI: RawResponsePlanningAIResponse;\n  try {\n    planFromAI = parseXMLPlanningResponse(rawResponse, modelFamily);\n  } catch (parseError) {\n    // Throw AIResponseParseError with usage and rawResponse preserved\n    const errorMessage =\n      parseError instanceof Error ? parseError.message : String(parseError);\n    throw new AIResponseParseError(\n      `XML parse error: ${errorMessage}`,\n      rawResponse,\n      usage,\n    );\n  }\n\n  if (planFromAI.action && planFromAI.finalizeSuccess !== undefined) {\n    console.warn(\n      'Planning response included both an action and complete-goal; ignoring complete-goal output.',\n    );\n    planFromAI.finalizeMessage = undefined;\n    planFromAI.finalizeSuccess = undefined;\n  }\n\n  const actions = planFromAI.action ? [planFromAI.action] : [];\n  let shouldContinuePlanning = true;\n\n  // Check if goal is completed via complete-goal tag\n  if (planFromAI.finalizeSuccess !== undefined) {\n    debug('goal completed via complete-goal tag, stop planning');\n    shouldContinuePlanning = false;\n    // Mark all sub-goals as finished when goal is completed (only when deepThink is enabled)\n    if (includeSubGoals) {\n      conversationHistory.markAllSubGoalsFinished();\n    }\n  }\n\n  const returnValue: PlanningAIResponse = {\n    ...planFromAI,\n    actions,\n    rawResponse,\n    usage,\n    reasoning_content,\n    yamlFlow: buildYamlFlowFromPlans(actions, opts.actionSpace),\n    shouldContinuePlanning,\n  };\n\n  assert(planFromAI, \"can't get plans from AI\");\n\n  actions.forEach((action) => {\n    const type = action.type;\n    const actionInActionSpace = opts.actionSpace.find(\n      (action) => action.name === type,\n    );\n\n    debug('actionInActionSpace matched', actionInActionSpace);\n    const locateFields = actionInActionSpace\n      ? findAllMidsceneLocatorField(actionInActionSpace.paramSchema)\n      : [];\n\n    debug('locateFields', locateFields);\n\n    locateFields.forEach((field) => {\n      const locateResult = action.param[field];\n      if (locateResult && modelFamily !== undefined) {\n        // Always use model family to fill bbox parameters\n        action.param[field] = fillBboxParam(\n          locateResult,\n          imageWidth,\n          imageHeight,\n          rightLimit,\n          bottomLimit,\n          modelFamily,\n        );\n      }\n    });\n  });\n\n  // Update sub-goals in conversation history based on response (only when deepThink is enabled)\n  if (includeSubGoals) {\n    if (planFromAI.updateSubGoals?.length) {\n      conversationHistory.setSubGoals(planFromAI.updateSubGoals);\n    }\n    if (planFromAI.markFinishedIndexes?.length) {\n      for (const index of planFromAI.markFinishedIndexes) {\n        conversationHistory.markSubGoalFinished(index);\n      }\n    }\n  }\n\n  // Append note to conversation history if present\n  if (planFromAI.note) {\n    conversationHistory.appendNote(planFromAI.note);\n  }\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: [\n      {\n        type: 'text',\n        text: rawResponse,\n      },\n    ],\n  });\n\n  return returnValue;\n}\n"],"names":["debug","getDebug","parseXMLPlanningResponse","xmlString","modelFamily","thought","extractXMLTag","note","log","error","actionType","actionParamStr","completeGoalRegex","completeGoalMatch","finalizeMessage","finalizeSuccess","undefined","updatePlanContent","markSubGoalDone","updateSubGoals","parseSubGoalsFromXML","markFinishedIndexes","parseMarkFinishedIndexes","action","type","param","safeParseJson","e","Error","plan","userInstruction","opts","context","modelConfig","conversationHistory","size","screenshotBase64","includeSubGoals","systemPrompt","systemPromptToTaskPlanning","imagePayload","imageWidth","imageHeight","rightLimit","bottomLimit","paddedResult","paddingToMatchBlockByBase64","actionContext","instruction","latestFeedbackMessage","subGoalsText","subGoalsSection","notesText","notesSection","historyLog","msgs","rawResponse","usage","reasoning_content","callAI","planFromAI","parseError","errorMessage","String","AIResponseParseError","console","actions","shouldContinuePlanning","returnValue","buildYamlFlowFromPlans","assert","actionInActionSpace","locateFields","findAllMidsceneLocatorField","field","locateResult","fillBboxParam","index"],"mappings":";;;;;;;AA+BA,MAAMA,QAAQC,SAAS;AAKhB,SAASC,yBACdC,SAAiB,EACjBC,WAAqC;IAErC,MAAMC,UAAUC,cAAcH,WAAW;IACzC,MAAMI,OAAOD,cAAcH,WAAW;IACtC,MAAMK,MAAMF,cAAcH,WAAW,UAAU;IAC/C,MAAMM,QAAQH,cAAcH,WAAW;IACvC,MAAMO,aAAaJ,cAAcH,WAAW;IAC5C,MAAMQ,iBAAiBL,cAAcH,WAAW;IAGhD,MAAMS,oBACJ;IACF,MAAMC,oBAAoBV,UAAU,KAAK,CAACS;IAC1C,IAAIE;IACJ,IAAIC;IAEJ,IAAIF,mBAAmB;QACrBE,kBAAkBF,AAAyB,WAAzBA,iBAAiB,CAAC,EAAE;QACtCC,kBAAkBD,iBAAiB,CAAC,EAAE,EAAE,UAAUG;IACpD;IAGA,MAAMC,oBAAoBX,cAAcH,WAAW;IACnD,MAAMe,kBAAkBZ,cAAcH,WAAW;IAEjD,MAAMgB,iBAAiBF,oBACnBG,qBAAqBH,qBACrBD;IACJ,MAAMK,sBAAsBH,kBACxBI,yBAAyBJ,mBACzBF;IAGJ,IAAIO,SAAc;IAClB,IAAIb,cAAcA,AAA6B,WAA7BA,WAAW,WAAW,IAAe;QACrD,MAAMc,OAAOd,WAAW,IAAI;QAC5B,IAAIe;QAEJ,IAAId,gBACF,IAAI;YAEFc,QAAQC,cAAcf,gBAAgBP;QACxC,EAAE,OAAOuB,GAAG;YACV,MAAM,IAAIC,MAAM,CAAC,mCAAmC,EAAED,GAAG;QAC3D;QAGFJ,SAAS;YACPC;YACA,GAAIC,AAAUT,WAAVS,QAAsB;gBAAEA;YAAM,IAAI,CAAC,CAAC;QAC1C;IACF;IAEA,OAAO;QACL,GAAIpB,UAAU;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAC9B,GAAIE,OAAO;YAAEA;QAAK,IAAI,CAAC,CAAC;QACxBC;QACA,GAAIC,QAAQ;YAAEA;QAAM,IAAI,CAAC,CAAC;QAC1Bc;QACA,GAAIT,AAAoBE,WAApBF,kBAAgC;YAAEA;QAAgB,IAAI,CAAC,CAAC;QAC5D,GAAIC,AAAoBC,WAApBD,kBAAgC;YAAEA;QAAgB,IAAI,CAAC,CAAC;QAC5D,GAAII,gBAAgB,SAAS;YAAEA;QAAe,IAAI,CAAC,CAAC;QACpD,GAAIE,qBAAqB,SAAS;YAAEA;QAAoB,IAAI,CAAC,CAAC;IAChE;AACF;AAEO,eAAeQ,KACpBC,eAAuB,EACvBC,IAUC;IAED,MAAM,EAAEC,OAAO,EAAEC,WAAW,EAAEC,mBAAmB,EAAE,GAAGH;IACtD,MAAM,EAAEI,IAAI,EAAE,GAAGH;IACjB,MAAMI,mBAAmBJ,QAAQ,UAAU,CAAC,MAAM;IAElD,MAAM,EAAE5B,WAAW,EAAE,GAAG6B;IAGxB,MAAMI,kBAAkBN,AAAmB,SAAnBA,KAAK,SAAS;IAEtC,MAAMO,eAAe,MAAMC,2BAA2B;QACpD,aAAaR,KAAK,WAAW;QAC7B3B;QACA,aAAa2B,KAAK,WAAW;QAC7B,gBAAgB;QAChBM;IACF;IAEA,IAAIG,eAAeJ;IACnB,IAAIK,aAAaN,KAAK,KAAK;IAC3B,IAAIO,cAAcP,KAAK,MAAM;IAC7B,MAAMQ,aAAaF;IACnB,MAAMG,cAAcF;IAGpB,IAAItC,AAAgB,iBAAhBA,aAA8B;QAChC,MAAMyC,eAAe,MAAMC,4BAA4BN;QACvDC,aAAaI,aAAa,KAAK;QAC/BH,cAAcG,aAAa,MAAM;QACjCL,eAAeK,aAAa,WAAW;IACzC;IAEA,MAAME,gBAAgBhB,KAAK,aAAa,GACpC,CAAC,yBAAyB,EAAEA,KAAK,aAAa,CAAC,4BAA4B,CAAC,GAC5E;IAEJ,MAAMiB,cAA4C;QAChD;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGD,cAAc,kBAAkB,EAAEjB,gBAAgB,mBAAmB,CAAC;gBACjF;aACD;QACH;KACD;IAED,IAAImB;IAGJ,MAAMC,eAAeb,kBACjBH,oBAAoB,cAAc,KAClC;IACJ,MAAMiB,kBAAkBD,eAAe,CAAC,IAAI,EAAEA,cAAc,GAAG;IAG/D,MAAME,YAAYlB,oBAAoB,WAAW;IACjD,MAAMmB,eAAeD,YAAY,CAAC,IAAI,EAAEA,WAAW,GAAG;IAEtD,IAAIlB,oBAAoB,sBAAsB,EAAE;QAC9Ce,wBAAwB;YACtB,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,MAAM,GAAGf,oBAAoB,sBAAsB,CAAC,gFAAgF,EAAEmB,eAAeF,iBAAiB;gBACxK;gBACA;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKX;wBACL,QAAQ;oBACV;gBACF;aACD;QACH;QAEAN,oBAAoB,mCAAmC;IACzD,OACEe,wBAAwB;QACtB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAM,CAAC,6BAA6B,EAAEI,eAAeF,iBAAiB;YACxE;YACA;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKX;oBACL,QAAQ;gBACV;YACF;SACD;IACH;IAEFN,oBAAoB,MAAM,CAACe;IAG3Bf,oBAAoB,eAAe,CAAC,IAAI;IAExC,MAAMoB,aAAapB,oBAAoB,QAAQ,CAACH,KAAK,kBAAkB;IAEvE,MAAMwB,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASjB;QAAa;WACrCU;WACAM;KACJ;IAED,MAAM,EACJ,SAASE,WAAW,EACpBC,KAAK,EACLC,iBAAiB,EAClB,GAAG,MAAMC,OAAOJ,MAAMtB,aAAa;QAClC,WAAWF,AAAmB,YAAnBA,KAAK,SAAS,GAAef,SAAYe,KAAK,SAAS;IACpE;IAGA,IAAI6B;IACJ,IAAI;QACFA,aAAa1D,yBAAyBsD,aAAapD;IACrD,EAAE,OAAOyD,YAAY;QAEnB,MAAMC,eACJD,sBAAsBjC,QAAQiC,WAAW,OAAO,GAAGE,OAAOF;QAC5D,MAAM,IAAIG,qBACR,CAAC,iBAAiB,EAAEF,cAAc,EAClCN,aACAC;IAEJ;IAEA,IAAIG,WAAW,MAAM,IAAIA,AAA+B5C,WAA/B4C,WAAW,eAAe,EAAgB;QACjEK,QAAQ,IAAI,CACV;QAEFL,WAAW,eAAe,GAAG5C;QAC7B4C,WAAW,eAAe,GAAG5C;IAC/B;IAEA,MAAMkD,UAAUN,WAAW,MAAM,GAAG;QAACA,WAAW,MAAM;KAAC,GAAG,EAAE;IAC5D,IAAIO,yBAAyB;IAG7B,IAAIP,AAA+B5C,WAA/B4C,WAAW,eAAe,EAAgB;QAC5C5D,MAAM;QACNmE,yBAAyB;QAEzB,IAAI9B,iBACFH,oBAAoB,uBAAuB;IAE/C;IAEA,MAAMkC,cAAkC;QACtC,GAAGR,UAAU;QACbM;QACAV;QACAC;QACAC;QACA,UAAUW,uBAAuBH,SAASnC,KAAK,WAAW;QAC1DoC;IACF;IAEAG,OAAOV,YAAY;IAEnBM,QAAQ,OAAO,CAAC,CAAC3C;QACf,MAAMC,OAAOD,OAAO,IAAI;QACxB,MAAMgD,sBAAsBxC,KAAK,WAAW,CAAC,IAAI,CAC/C,CAACR,SAAWA,OAAO,IAAI,KAAKC;QAG9BxB,MAAM,+BAA+BuE;QACrC,MAAMC,eAAeD,sBACjBE,4BAA4BF,oBAAoB,WAAW,IAC3D,EAAE;QAENvE,MAAM,gBAAgBwE;QAEtBA,aAAa,OAAO,CAAC,CAACE;YACpB,MAAMC,eAAepD,OAAO,KAAK,CAACmD,MAAM;YACxC,IAAIC,gBAAgBvE,AAAgBY,WAAhBZ,aAElBmB,OAAO,KAAK,CAACmD,MAAM,GAAGE,cACpBD,cACAlC,YACAC,aACAC,YACAC,aACAxC;QAGN;IACF;IAGA,IAAIiC,iBAAiB;QACnB,IAAIuB,WAAW,cAAc,EAAE,QAC7B1B,oBAAoB,WAAW,CAAC0B,WAAW,cAAc;QAE3D,IAAIA,WAAW,mBAAmB,EAAE,QAClC,KAAK,MAAMiB,SAASjB,WAAW,mBAAmB,CAChD1B,oBAAoB,mBAAmB,CAAC2C;IAG9C;IAGA,IAAIjB,WAAW,IAAI,EACjB1B,oBAAoB,UAAU,CAAC0B,WAAW,IAAI;IAGhD1B,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,MAAMsB;YACR;SACD;IACH;IAEA,OAAOY;AACT"}

package/dist/es/ai-model/service-caller/index.mjs

@@ -4,6 +4,24 @@ import { assert, ifInBrowser } from "@midscene/shared/utils";
 import { jsonrepair } from "jsonrepair";
 import openai_0 from "openai";
 import { isAutoGLM, isUITars } from "../auto-glm/util.mjs";
+function _define_property(obj, key, value) {
+    if (key in obj) Object.defineProperty(obj, key, {
+        value: value,
+        enumerable: true,
+        configurable: true,
+        writable: true
+    });
+    else obj[key] = value;
+    return obj;
+}
+class AIResponseParseError extends Error {
+    constructor(message, rawResponse, usage){
+        super(message), _define_property(this, "usage", void 0), _define_property(this, "rawResponse", void 0);
+        this.name = 'AIResponseParseError';
+        this.rawResponse = rawResponse;
+        this.usage = usage;
+    }
+}
 async function createChatClient({ modelConfig }) {
     const { socksProxy, httpProxy, modelName, openaiBaseURL, openaiApiKey, openaiExtraConfig, modelDescription, uiTarsModelVersion, modelFamily, createOpenAIClient, timeout } = modelConfig;
     let proxyAgent;
@@ -265,7 +283,7 @@ async function callAIWithObjectResponse(messages, modelConfig, options) {
     assert(response, 'empty response');
     const modelFamily = modelConfig.modelFamily;
     const jsonContent = safeParseJson(response.content, modelFamily);
-    assert('object' == typeof jsonContent, `failed to parse json response from model (${modelConfig.modelName}): ${response.content}`);
+    if ('object' != typeof jsonContent) throw new AIResponseParseError(`failed to parse json response from model (${modelConfig.modelName}): ${response.content}`, response.content, response.usage);
     return {
         content: jsonContent,
         contentString: response.content,
@@ -385,6 +403,6 @@ function safeParseJson(input, modelFamily) {
     }
     throw Error(`failed to parse LLM response into JSON. Error - ${String(lastError ?? 'unknown error')}. Response - \n ${input}`);
 }
-export { callAI, callAIWithObjectResponse, callAIWithStringResponse, extractJSONFromCodeBlock, preprocessDoubaoBboxJson, resolveDeepThinkConfig, safeParseJson };
+export { AIResponseParseError, callAI, callAIWithObjectResponse, callAIWithStringResponse, extractJSONFromCodeBlock, preprocessDoubaoBboxJson, resolveDeepThinkConfig, safeParseJson };
 
 //# sourceMappingURL=index.mjs.map

package/dist/es/ai-model/service-caller/index.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/service-caller/index.mjs","sources":["../../../../src/ai-model/service-caller/index.ts"],"sourcesContent":["import type { AIUsageInfo, DeepThinkOption } from '@/types';\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\nimport {\n  type IModelConfig,\n  MIDSCENE_LANGFUSE_DEBUG,\n  MIDSCENE_LANGSMITH_DEBUG,\n  MIDSCENE_MODEL_MAX_TOKENS,\n  OPENAI_MAX_TOKENS,\n  type TModelFamily,\n  type UITarsModelVersion,\n  globalConfigManager,\n} from '@midscene/shared/env';\n\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert, ifInBrowser } from '@midscene/shared/utils';\nimport { jsonrepair } from 'jsonrepair';\nimport OpenAI from 'openai';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { Stream } from 'openai/streaming';\nimport type { AIArgs } from '../../common';\nimport { isAutoGLM, isUITars } from '../auto-glm/util';\n\nasync function createChatClient({\n  modelConfig,\n}: {\n  modelConfig: IModelConfig;\n}): Promise<{\n  completion: OpenAI.Chat.Completions;\n  modelName: string;\n  modelDescription: string;\n  uiTarsModelVersion?: UITarsModelVersion;\n  modelFamily: TModelFamily | undefined;\n}> {\n  const {\n    socksProxy,\n    httpProxy,\n    modelName,\n    openaiBaseURL,\n    openaiApiKey,\n    openaiExtraConfig,\n    modelDescription,\n    uiTarsModelVersion,\n    modelFamily,\n    createOpenAIClient,\n    timeout,\n  } = modelConfig;\n\n  let proxyAgent: any = undefined;\n  const debugProxy = getDebug('ai:call:proxy');\n\n  // Helper function to sanitize proxy URL for logging (remove credentials)\n  // Uses URL API instead of regex to avoid ReDoS vulnerabilities\n  const sanitizeProxyUrl = (url: string): string => {\n    try {\n      const parsed = new URL(url);\n      if (parsed.username) {\n        // Keep username for debugging, hide password for security\n        parsed.password = '****';\n        return parsed.href;\n      }\n      return url;\n    } catch {\n      // If URL parsing fails, return original URL (will be caught later)\n      return url;\n    }\n  };\n\n  if (httpProxy) {\n    debugProxy('using http proxy', sanitizeProxyUrl(httpProxy));\n    if (ifInBrowser) {\n      console.warn(\n        'HTTP proxy is configured but not supported in browser environment',\n      );\n    } else {\n      // Dynamic import with variable to avoid bundler static analysis\n      const moduleName = 'undici';\n      const { ProxyAgent } = await import(moduleName);\n      proxyAgent = new ProxyAgent({\n        uri: httpProxy,\n        // Note: authentication is handled via the URI (e.g., http://user:pass@proxy.com:8080)\n      });\n    }\n  } else if (socksProxy) {\n    debugProxy('using socks proxy', sanitizeProxyUrl(socksProxy));\n    if (ifInBrowser) {\n      console.warn(\n        'SOCKS proxy is configured but not supported in browser environment',\n      );\n    } else {\n      try {\n        // Dynamic import with variable to avoid bundler static analysis\n        const moduleName = 'fetch-socks';\n        const { socksDispatcher } = await import(moduleName);\n        // Parse SOCKS proxy URL (e.g., socks5://127.0.0.1:1080)\n        const proxyUrl = new URL(socksProxy);\n\n        // Validate hostname\n        if (!proxyUrl.hostname) {\n          throw new Error('SOCKS proxy URL must include a valid hostname');\n        }\n\n        // Validate and parse port\n        const port = Number.parseInt(proxyUrl.port, 10);\n        if (!proxyUrl.port || Number.isNaN(port)) {\n          throw new Error('SOCKS proxy URL must include a valid port');\n        }\n\n        // Parse SOCKS version from protocol\n        const protocol = proxyUrl.protocol.replace(':', '');\n        const socksType =\n          protocol === 'socks4' ? 4 : protocol === 'socks5' ? 5 : 5;\n\n        proxyAgent = socksDispatcher({\n          type: socksType,\n          host: proxyUrl.hostname,\n          port,\n          ...(proxyUrl.username\n            ? {\n                userId: decodeURIComponent(proxyUrl.username),\n                password: decodeURIComponent(proxyUrl.password || ''),\n              }\n            : {}),\n        });\n        debugProxy('socks proxy configured successfully', {\n          type: socksType,\n          host: proxyUrl.hostname,\n          port: port,\n        });\n      } catch (error) {\n        console.error('Failed to configure SOCKS proxy:', error);\n        throw new Error(\n          `Invalid SOCKS proxy URL: ${socksProxy}. Expected format: socks4://host:port, socks5://host:port, or with authentication: socks5://user:pass@host:port`,\n        );\n      }\n    }\n  }\n\n  const openAIOptions = {\n    baseURL: openaiBaseURL,\n    apiKey: openaiApiKey,\n    // Use fetchOptions.dispatcher for fetch-based SDK instead of httpAgent\n    // Note: Type assertion needed due to undici version mismatch between dependencies\n    ...(proxyAgent ? { fetchOptions: { dispatcher: proxyAgent as any } } : {}),\n    ...openaiExtraConfig,\n    ...(typeof timeout === 'number' ? { timeout } : {}),\n    dangerouslyAllowBrowser: true,\n  };\n\n  const baseOpenAI = new OpenAI(openAIOptions);\n\n  let openai: OpenAI = baseOpenAI;\n\n  // LangSmith wrapper\n  if (\n    openai &&\n    globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)\n  ) {\n    if (ifInBrowser) {\n      throw new Error('langsmith is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langsmith wrapper enabled');\n    // Use variable to prevent static analysis by bundlers\n    const langsmithModule = 'langsmith/wrappers';\n    const { wrapOpenAI } = await import(langsmithModule);\n    openai = wrapOpenAI(openai);\n  }\n\n  // Langfuse wrapper\n  if (\n    openai &&\n    globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGFUSE_DEBUG)\n  ) {\n    if (ifInBrowser) {\n      throw new Error('langfuse is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langfuse wrapper enabled');\n    // Use variable to prevent static analysis by bundlers\n    const langfuseModule = '@langfuse/openai';\n    const { observeOpenAI } = await import(langfuseModule);\n    openai = observeOpenAI(openai);\n  }\n\n  if (createOpenAIClient) {\n    const wrappedClient = await createOpenAIClient(baseOpenAI, openAIOptions);\n\n    if (wrappedClient) {\n      openai = wrappedClient as OpenAI;\n    }\n  }\n\n  return {\n    completion: openai.chat.completions,\n    modelName,\n    modelDescription,\n    uiTarsModelVersion,\n    modelFamily,\n  };\n}\n\nexport async function callAI(\n  messages: ChatCompletionMessageParam[],\n  modelConfig: IModelConfig,\n  options?: {\n    stream?: boolean;\n    onChunk?: StreamingCallback;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<{\n  content: string;\n  reasoning_content?: string;\n  usage?: AIUsageInfo;\n  isStreamed: boolean;\n}> {\n  const {\n    completion,\n    modelName,\n    modelDescription,\n    uiTarsModelVersion,\n    modelFamily,\n  } = await createChatClient({\n    modelConfig,\n  });\n\n  const maxTokens =\n    globalConfigManager.getEnvConfigValueAsNumber(MIDSCENE_MODEL_MAX_TOKENS) ??\n    globalConfigManager.getEnvConfigValueAsNumber(OPENAI_MAX_TOKENS);\n  const debugCall = getDebug('ai:call');\n  const debugProfileStats = getDebug('ai:profile:stats');\n  const debugProfileDetail = getDebug('ai:profile:detail');\n\n  const startTime = Date.now();\n  const temperature = modelConfig.temperature ?? 0;\n\n  const isStreaming = options?.stream && options?.onChunk;\n  let content: string | undefined;\n  let accumulated = '';\n  let accumulatedReasoning = '';\n  let usage: OpenAI.CompletionUsage | undefined;\n  let timeCost: number | undefined;\n\n  const buildUsageInfo = (usageData?: OpenAI.CompletionUsage) => {\n    if (!usageData) return undefined;\n\n    const cachedInputTokens = (\n      usageData as { prompt_tokens_details?: { cached_tokens?: number } }\n    )?.prompt_tokens_details?.cached_tokens;\n\n    return {\n      prompt_tokens: usageData.prompt_tokens ?? 0,\n      completion_tokens: usageData.completion_tokens ?? 0,\n      total_tokens: usageData.total_tokens ?? 0,\n      cached_input: cachedInputTokens ?? 0,\n      time_cost: timeCost ?? 0,\n      model_name: modelName,\n      model_description: modelDescription,\n      intent: modelConfig.intent,\n    } satisfies AIUsageInfo;\n  };\n\n  const commonConfig = {\n    temperature,\n    stream: !!isStreaming,\n    max_tokens: maxTokens,\n    ...(modelFamily === 'qwen2.5-vl' // qwen vl v2 specific config\n      ? {\n          vl_high_resolution_images: true,\n        }\n      : {}),\n  };\n\n  if (isAutoGLM(modelFamily)) {\n    (commonConfig as unknown as Record<string, number>).top_p = 0.85;\n    (commonConfig as unknown as Record<string, number>).frequency_penalty = 0.2;\n  }\n\n  const {\n    config: deepThinkConfig,\n    debugMessage,\n    warningMessage,\n  } = resolveDeepThinkConfig({\n    deepThink: options?.deepThink,\n    modelFamily,\n  });\n  if (debugMessage) {\n    debugCall(debugMessage);\n  }\n  if (warningMessage) {\n    debugCall(warningMessage);\n    console.warn(warningMessage);\n  }\n\n  try {\n    debugCall(\n      `sending ${isStreaming ? 'streaming ' : ''}request to ${modelName}`,\n    );\n\n    if (isStreaming) {\n      const stream = (await completion.create(\n        {\n          model: modelName,\n          messages,\n          ...commonConfig,\n          ...deepThinkConfig,\n        },\n        {\n          stream: true,\n        },\n      )) as Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {\n        _request_id?: string | null;\n      };\n\n      for await (const chunk of stream) {\n        const content = chunk.choices?.[0]?.delta?.content || '';\n        const reasoning_content =\n          (chunk.choices?.[0]?.delta as any)?.reasoning_content || '';\n\n        // Check for usage info in any chunk (OpenAI provides usage in separate chunks)\n        if (chunk.usage) {\n          usage = chunk.usage;\n        }\n\n        if (content || reasoning_content) {\n          accumulated += content;\n          accumulatedReasoning += reasoning_content;\n          const chunkData: CodeGenerationChunk = {\n            content,\n            reasoning_content,\n            accumulated,\n            isComplete: false,\n            usage: undefined,\n          };\n          options.onChunk!(chunkData);\n        }\n\n        // Check if stream is complete\n        if (chunk.choices?.[0]?.finish_reason) {\n          timeCost = Date.now() - startTime;\n\n          // If usage is not available from the stream, provide a basic usage info\n          if (!usage) {\n            // Estimate token counts based on content length (rough approximation)\n            const estimatedTokens = Math.max(\n              1,\n              Math.floor(accumulated.length / 4),\n            );\n            usage = {\n              prompt_tokens: estimatedTokens,\n              completion_tokens: estimatedTokens,\n              total_tokens: estimatedTokens * 2,\n            };\n          }\n\n          // Send final chunk\n          const finalChunk: CodeGenerationChunk = {\n            content: '',\n            accumulated,\n            reasoning_content: '',\n            isComplete: true,\n            usage: buildUsageInfo(usage),\n          };\n          options.onChunk!(finalChunk);\n          break;\n        }\n      }\n      content = accumulated;\n      debugProfileStats(\n        `streaming model, ${modelName}, mode, ${modelFamily || 'default'}, cost-ms, ${timeCost}, temperature, ${temperature ?? ''}`,\n      );\n    } else {\n      // Non-streaming with retry logic\n      const retryCount = modelConfig.retryCount ?? 1;\n      const retryInterval = modelConfig.retryInterval ?? 2000;\n      const maxAttempts = retryCount + 1; // retryCount=1 means 2 total attempts (1 initial + 1 retry)\n\n      let lastError: Error | undefined;\n\n      for (let attempt = 1; attempt <= maxAttempts; attempt++) {\n        try {\n          const result = await completion.create({\n            model: modelName,\n            messages,\n            ...commonConfig,\n            ...deepThinkConfig,\n          } as any);\n\n          timeCost = Date.now() - startTime;\n\n          debugProfileStats(\n            `model, ${modelName}, mode, ${modelFamily || 'default'}, ui-tars-version, ${uiTarsModelVersion}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}, temperature, ${temperature ?? ''}`,\n          );\n\n          debugProfileDetail(\n            `model usage detail: ${JSON.stringify(result.usage)}`,\n          );\n\n          if (!result.choices) {\n            throw new Error(\n              `invalid response from LLM service: ${JSON.stringify(result)}`,\n            );\n          }\n\n          content = result.choices[0].message.content!;\n          if (!content) {\n            throw new Error('empty content from AI model');\n          }\n\n          accumulatedReasoning =\n            (result.choices[0].message as any)?.reasoning_content || '';\n          usage = result.usage;\n          break; // Success, exit retry loop\n        } catch (error) {\n          lastError = error as Error;\n          if (attempt < maxAttempts) {\n            console.warn(\n              `[Midscene] AI call failed (attempt ${attempt}/${maxAttempts}), retrying in ${retryInterval}ms... Error: ${lastError.message}`,\n            );\n            await new Promise((resolve) => setTimeout(resolve, retryInterval));\n          }\n        }\n      }\n\n      if (!content) {\n        throw lastError;\n      }\n    }\n\n    debugCall(`response reasoning content: ${accumulatedReasoning}`);\n    debugCall(`response content: ${content}`);\n\n    // Ensure we always have usage info for streaming responses\n    if (isStreaming && !usage) {\n      // Estimate token counts based on content length (rough approximation)\n      const estimatedTokens = Math.max(\n        1,\n        Math.floor((content || '').length / 4),\n      );\n      usage = {\n        prompt_tokens: estimatedTokens,\n        completion_tokens: estimatedTokens,\n        total_tokens: estimatedTokens * 2,\n      } as OpenAI.CompletionUsage;\n    }\n\n    return {\n      content: content || '',\n      reasoning_content: accumulatedReasoning || undefined,\n      usage: buildUsageInfo(usage),\n      isStreamed: !!isStreaming,\n    };\n  } catch (e: any) {\n    console.error(' call AI error', e);\n    const newError = new Error(\n      `failed to call ${isStreaming ? 'streaming ' : ''}AI model service (${modelName}): ${e.message}\\nTrouble shooting: https://midscenejs.com/model-provider.html`,\n      {\n        cause: e,\n      },\n    );\n    throw newError;\n  }\n}\n\nexport async function callAIWithObjectResponse<T>(\n  messages: ChatCompletionMessageParam[],\n  modelConfig: IModelConfig,\n  options?: {\n    deepThink?: DeepThinkOption;\n  },\n): Promise<{\n  content: T;\n  contentString: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}> {\n  const response = await callAI(messages, modelConfig, {\n    deepThink: options?.deepThink,\n  });\n  assert(response, 'empty response');\n  const modelFamily = modelConfig.modelFamily;\n  const jsonContent = safeParseJson(response.content, modelFamily);\n  assert(\n    typeof jsonContent === 'object',\n    `failed to parse json response from model (${modelConfig.modelName}): ${response.content}`,\n  );\n  return {\n    content: jsonContent,\n    contentString: response.content,\n    usage: response.usage,\n    reasoning_content: response.reasoning_content,\n  };\n}\n\nexport async function callAIWithStringResponse(\n  msgs: AIArgs,\n  modelConfig: IModelConfig,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n  const { content, usage } = await callAI(msgs, modelConfig);\n  return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n  try {\n    // First, try to match a JSON object directly in the response\n    const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\n    if (jsonMatch) {\n      return jsonMatch[1];\n    }\n\n    // If no direct JSON object is found, try to extract JSON from a code block\n    const codeBlockMatch = response.match(\n      /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\n    );\n    if (codeBlockMatch) {\n      return codeBlockMatch[1];\n    }\n\n    // If no code block is found, try to find a JSON-like structure in the text\n    const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\n    if (jsonLikeMatch) {\n      return jsonLikeMatch[0];\n    }\n  } catch {}\n  // If no JSON-like structure is found, return the original response\n  return response;\n}\n\nexport function preprocessDoubaoBboxJson(input: string) {\n  if (input.includes('bbox')) {\n    // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\n    while (/\\d+\\s+\\d+/.test(input)) {\n      input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n    }\n  }\n  return input;\n}\n\nexport function resolveDeepThinkConfig({\n  deepThink,\n  modelFamily,\n}: {\n  deepThink?: DeepThinkOption;\n  modelFamily?: TModelFamily;\n}): {\n  config: Record<string, unknown>;\n  debugMessage?: string;\n  warningMessage?: string;\n} {\n  const normalizedDeepThink = deepThink === 'unset' ? undefined : deepThink;\n\n  if (normalizedDeepThink === undefined) {\n    return { config: {}, debugMessage: undefined };\n  }\n\n  if (modelFamily === 'qwen3-vl') {\n    return {\n      config: { enable_thinking: normalizedDeepThink },\n      debugMessage: `deepThink mapped to enable_thinking=${normalizedDeepThink} for qwen3-vl`,\n    };\n  }\n\n  if (modelFamily === 'doubao-vision') {\n    return {\n      config: {\n        thinking: { type: normalizedDeepThink ? 'enabled' : 'disabled' },\n      },\n      debugMessage: `deepThink mapped to thinking.type=${normalizedDeepThink ? 'enabled' : 'disabled'} for doubao-vision`,\n    };\n  }\n\n  if (modelFamily === 'glm-v') {\n    return {\n      config: {\n        thinking: { type: normalizedDeepThink ? 'enabled' : 'disabled' },\n      },\n      debugMessage: `deepThink mapped to thinking.type=${normalizedDeepThink ? 'enabled' : 'disabled'} for glm-v`,\n    };\n  }\n\n  if (modelFamily === 'gpt-5') {\n    return {\n      config: normalizedDeepThink\n        ? {\n            reasoning: { effort: 'high' },\n          }\n        : {\n            reasoning: { effort: 'low' },\n          },\n      debugMessage: normalizedDeepThink\n        ? 'deepThink mapped to reasoning.effort=high for gpt-5'\n        : 'deepThink disabled for gpt-5',\n    };\n  }\n\n  return {\n    config: {},\n    debugMessage: `deepThink ignored: unsupported model_family \"${modelFamily ?? 'default'}\"`,\n    warningMessage: `The \"deepThink\" option is not supported for model_family \"${modelFamily ?? 'default'}\".`,\n  };\n}\n\n/**\n * Normalize a parsed JSON object by trimming whitespace from:\n * 1. All object keys (e.g., \" prompt \" -> \"prompt\")\n * 2. All string values (e.g., \" Tap \" -> \"Tap\")\n * This handles LLM output that may include leading/trailing spaces.\n */\nfunction normalizeJsonObject(obj: any): any {\n  // Handle null and undefined\n  if (obj === null || obj === undefined) {\n    return obj;\n  }\n\n  // Handle arrays - recursively normalize each element\n  if (Array.isArray(obj)) {\n    return obj.map((item) => normalizeJsonObject(item));\n  }\n\n  // Handle objects\n  if (typeof obj === 'object') {\n    const normalized: any = {};\n\n    for (const [key, value] of Object.entries(obj)) {\n      // Trim the key to remove leading/trailing spaces\n      const trimmedKey = key.trim();\n\n      // Recursively normalize the value\n      let normalizedValue = normalizeJsonObject(value);\n\n      // Trim all string values\n      if (typeof normalizedValue === 'string') {\n        normalizedValue = normalizedValue.trim();\n      }\n\n      normalized[trimmedKey] = normalizedValue;\n    }\n\n    return normalized;\n  }\n\n  // Handle primitive strings\n  if (typeof obj === 'string') {\n    return obj.trim();\n  }\n\n  // Return other primitives as-is\n  return obj;\n}\n\nexport function safeParseJson(\n  input: string,\n  modelFamily: TModelFamily | undefined,\n) {\n  const cleanJsonString = extractJSONFromCodeBlock(input);\n  // match the point\n  if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\n    return cleanJsonString\n      .match(/\\((\\d+),(\\d+)\\)/)\n      ?.slice(1)\n      .map(Number);\n  }\n\n  let parsed: any;\n  let lastError: unknown;\n  try {\n    parsed = JSON.parse(cleanJsonString);\n    return normalizeJsonObject(parsed);\n  } catch (error) {\n    lastError = error;\n  }\n  try {\n    parsed = JSON.parse(jsonrepair(cleanJsonString));\n    return normalizeJsonObject(parsed);\n  } catch (error) {\n    lastError = error;\n  }\n\n  if (modelFamily === 'doubao-vision' || isUITars(modelFamily)) {\n    const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\n    try {\n      parsed = JSON.parse(jsonrepair(jsonString));\n      return normalizeJsonObject(parsed);\n    } catch (error) {\n      lastError = error;\n    }\n  }\n  throw Error(\n    `failed to parse LLM response into JSON. Error - ${String(\n      lastError ?? 'unknown error',\n    )}. Response - \\n ${input}`,\n  );\n}\n"],"names":["createChatClient","modelConfig","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","modelDescription","uiTarsModelVersion","modelFamily","createOpenAIClient","timeout","proxyAgent","debugProxy","getDebug","sanitizeProxyUrl","url","parsed","URL","ifInBrowser","console","moduleName","ProxyAgent","socksDispatcher","proxyUrl","Error","port","Number","protocol","socksType","decodeURIComponent","error","openAIOptions","baseOpenAI","OpenAI","openai","globalConfigManager","MIDSCENE_LANGSMITH_DEBUG","langsmithModule","wrapOpenAI","MIDSCENE_LANGFUSE_DEBUG","langfuseModule","observeOpenAI","wrappedClient","callAI","messages","options","completion","maxTokens","MIDSCENE_MODEL_MAX_TOKENS","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","temperature","isStreaming","content","accumulated","accumulatedReasoning","usage","timeCost","buildUsageInfo","usageData","cachedInputTokens","commonConfig","isAutoGLM","deepThinkConfig","debugMessage","warningMessage","resolveDeepThinkConfig","stream","chunk","reasoning_content","chunkData","undefined","estimatedTokens","Math","finalChunk","retryCount","retryInterval","maxAttempts","lastError","attempt","result","JSON","Promise","resolve","setTimeout","e","newError","callAIWithObjectResponse","response","assert","jsonContent","safeParseJson","callAIWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","deepThink","normalizedDeepThink","normalizeJsonObject","obj","Array","item","normalized","key","value","Object","trimmedKey","normalizedValue","cleanJsonString","jsonrepair","isUITars","jsonString","String"],"mappings":";;;;;;AAsBA,eAAeA,iBAAiB,EAC9BC,WAAW,EAGZ;IAOC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,gBAAgB,EAChBC,kBAAkB,EAClBC,WAAW,EACXC,kBAAkB,EAClBC,OAAO,EACR,GAAGX;IAEJ,IAAIY;IACJ,MAAMC,aAAaC,SAAS;IAI5B,MAAMC,mBAAmB,CAACC;QACxB,IAAI;YACF,MAAMC,SAAS,IAAIC,IAAIF;YACvB,IAAIC,OAAO,QAAQ,EAAE;gBAEnBA,OAAO,QAAQ,GAAG;gBAClB,OAAOA,OAAO,IAAI;YACpB;YACA,OAAOD;QACT,EAAE,OAAM;YAEN,OAAOA;QACT;IACF;IAEA,IAAId,WAAW;QACbW,WAAW,oBAAoBE,iBAAiBb;QAChD,IAAIiB,aACFC,QAAQ,IAAI,CACV;aAEG;YAEL,MAAMC,aAAa;YACnB,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAACD;YACpCT,aAAa,IAAIU,WAAW;gBAC1B,KAAKpB;YAEP;QACF;IACF,OAAO,IAAID,YAAY;QACrBY,WAAW,qBAAqBE,iBAAiBd;QACjD,IAAIkB,aACFC,QAAQ,IAAI,CACV;aAGF,IAAI;YAEF,MAAMC,aAAa;YACnB,MAAM,EAAEE,eAAe,EAAE,GAAG,MAAM,MAAM,CAACF;YAEzC,MAAMG,WAAW,IAAIN,IAAIjB;YAGzB,IAAI,CAACuB,SAAS,QAAQ,EACpB,MAAM,IAAIC,MAAM;YAIlB,MAAMC,OAAOC,OAAO,QAAQ,CAACH,SAAS,IAAI,EAAE;YAC5C,IAAI,CAACA,SAAS,IAAI,IAAIG,OAAO,KAAK,CAACD,OACjC,MAAM,IAAID,MAAM;YAIlB,MAAMG,WAAWJ,SAAS,QAAQ,CAAC,OAAO,CAAC,KAAK;YAChD,MAAMK,YACJD,AAAa,aAAbA,WAAwB,IAAIA,AAAa,aAAbA,WAAwB,IAAI;YAE1DhB,aAAaW,gBAAgB;gBAC3B,MAAMM;gBACN,MAAML,SAAS,QAAQ;gBACvBE;gBACA,GAAIF,SAAS,QAAQ,GACjB;oBACE,QAAQM,mBAAmBN,SAAS,QAAQ;oBAC5C,UAAUM,mBAAmBN,SAAS,QAAQ,IAAI;gBACpD,IACA,CAAC,CAAC;YACR;YACAX,WAAW,uCAAuC;gBAChD,MAAMgB;gBACN,MAAML,SAAS,QAAQ;gBACvB,MAAME;YACR;QACF,EAAE,OAAOK,OAAO;YACdX,QAAQ,KAAK,CAAC,oCAAoCW;YAClD,MAAM,IAAIN,MACR,CAAC,yBAAyB,EAAExB,WAAW,+GAA+G,CAAC;QAE3J;IAEJ;IAEA,MAAM+B,gBAAgB;QACpB,SAAS5B;QACT,QAAQC;QAGR,GAAIO,aAAa;YAAE,cAAc;gBAAE,YAAYA;YAAkB;QAAE,IAAI,CAAC,CAAC;QACzE,GAAGN,iBAAiB;QACpB,GAAI,AAAmB,YAAnB,OAAOK,UAAuB;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAClD,yBAAyB;IAC3B;IAEA,MAAMsB,aAAa,IAAIC,SAAOF;IAE9B,IAAIG,SAAiBF;IAGrB,IACEE,UACAC,oBAAoB,qBAAqB,CAACC,2BAC1C;QACA,IAAIlB,aACF,MAAM,IAAIM,MAAM;QAElBL,QAAQ,GAAG,CAAC;QAEZ,MAAMkB,kBAAkB;QACxB,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAACD;QACpCH,SAASI,WAAWJ;IACtB;IAGA,IACEA,UACAC,oBAAoB,qBAAqB,CAACI,0BAC1C;QACA,IAAIrB,aACF,MAAM,IAAIM,MAAM;QAElBL,QAAQ,GAAG,CAAC;QAEZ,MAAMqB,iBAAiB;QACvB,MAAM,EAAEC,aAAa,EAAE,GAAG,MAAM,MAAM,CAACD;QACvCN,SAASO,cAAcP;IACzB;IAEA,IAAIzB,oBAAoB;QACtB,MAAMiC,gBAAgB,MAAMjC,mBAAmBuB,YAAYD;QAE3D,IAAIW,eACFR,SAASQ;IAEb;IAEA,OAAO;QACL,YAAYR,OAAO,IAAI,CAAC,WAAW;QACnChC;QACAI;QACAC;QACAC;IACF;AACF;AAEO,eAAemC,OACpBC,QAAsC,EACtC7C,WAAyB,EACzB8C,OAIC;IAOD,MAAM,EACJC,UAAU,EACV5C,SAAS,EACTI,gBAAgB,EAChBC,kBAAkB,EAClBC,WAAW,EACZ,GAAG,MAAMV,iBAAiB;QACzBC;IACF;IAEA,MAAMgD,YACJZ,oBAAoB,yBAAyB,CAACa,8BAC9Cb,oBAAoB,yBAAyB,CAACc;IAChD,MAAMC,YAAYrC,SAAS;IAC3B,MAAMsC,oBAAoBtC,SAAS;IACnC,MAAMuC,qBAAqBvC,SAAS;IAEpC,MAAMwC,YAAYC,KAAK,GAAG;IAC1B,MAAMC,cAAcxD,YAAY,WAAW,IAAI;IAE/C,MAAMyD,cAAcX,SAAS,UAAUA,SAAS;IAChD,IAAIY;IACJ,IAAIC,cAAc;IAClB,IAAIC,uBAAuB;IAC3B,IAAIC;IACJ,IAAIC;IAEJ,MAAMC,iBAAiB,CAACC;QACtB,IAAI,CAACA,WAAW;QAEhB,MAAMC,oBACJD,WACC,uBAAuB;QAE1B,OAAO;YACL,eAAeA,UAAU,aAAa,IAAI;YAC1C,mBAAmBA,UAAU,iBAAiB,IAAI;YAClD,cAAcA,UAAU,YAAY,IAAI;YACxC,cAAcC,qBAAqB;YACnC,WAAWH,YAAY;YACvB,YAAY3D;YACZ,mBAAmBI;YACnB,QAAQP,YAAY,MAAM;QAC5B;IACF;IAEA,MAAMkE,eAAe;QACnBV;QACA,QAAQ,CAAC,CAACC;QACV,YAAYT;QACZ,GAAIvC,AAAgB,iBAAhBA,cACA;YACE,2BAA2B;QAC7B,IACA,CAAC,CAAC;IACR;IAEA,IAAI0D,UAAU1D,cAAc;QACzByD,aAAmD,KAAK,GAAG;QAC3DA,aAAmD,iBAAiB,GAAG;IAC1E;IAEA,MAAM,EACJ,QAAQE,eAAe,EACvBC,YAAY,EACZC,cAAc,EACf,GAAGC,uBAAuB;QACzB,WAAWzB,SAAS;QACpBrC;IACF;IACA,IAAI4D,cACFlB,UAAUkB;IAEZ,IAAIC,gBAAgB;QAClBnB,UAAUmB;QACVlD,QAAQ,IAAI,CAACkD;IACf;IAEA,IAAI;QACFnB,UACE,CAAC,QAAQ,EAAEM,cAAc,eAAe,GAAG,WAAW,EAAEtD,WAAW;QAGrE,IAAIsD,aAAa;YACf,MAAMe,SAAU,MAAMzB,WAAW,MAAM,CACrC;gBACE,OAAO5C;gBACP0C;gBACA,GAAGqB,YAAY;gBACf,GAAGE,eAAe;YACpB,GACA;gBACE,QAAQ;YACV;YAKF,WAAW,MAAMK,SAASD,OAAQ;gBAChC,MAAMd,UAAUe,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,OAAO,WAAW;gBACtD,MAAMC,oBACHD,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,OAAe,qBAAqB;gBAG3D,IAAIA,MAAM,KAAK,EACbZ,QAAQY,MAAM,KAAK;gBAGrB,IAAIf,WAAWgB,mBAAmB;oBAChCf,eAAeD;oBACfE,wBAAwBc;oBACxB,MAAMC,YAAiC;wBACrCjB;wBACAgB;wBACAf;wBACA,YAAY;wBACZ,OAAOiB;oBACT;oBACA9B,QAAQ,OAAO,CAAE6B;gBACnB;gBAGA,IAAIF,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,eAAe;oBACrCX,WAAWP,KAAK,GAAG,KAAKD;oBAGxB,IAAI,CAACO,OAAO;wBAEV,MAAMgB,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAACnB,YAAY,MAAM,GAAG;wBAElCE,QAAQ;4BACN,eAAegB;4BACf,mBAAmBA;4BACnB,cAAcA,AAAkB,IAAlBA;wBAChB;oBACF;oBAGA,MAAME,aAAkC;wBACtC,SAAS;wBACTpB;wBACA,mBAAmB;wBACnB,YAAY;wBACZ,OAAOI,eAAeF;oBACxB;oBACAf,QAAQ,OAAO,CAAEiC;oBACjB;gBACF;YACF;YACArB,UAAUC;YACVP,kBACE,CAAC,iBAAiB,EAAEjD,UAAU,QAAQ,EAAEM,eAAe,UAAU,WAAW,EAAEqD,SAAS,eAAe,EAAEN,eAAe,IAAI;QAE/H,OAAO;YAEL,MAAMwB,aAAahF,YAAY,UAAU,IAAI;YAC7C,MAAMiF,gBAAgBjF,YAAY,aAAa,IAAI;YACnD,MAAMkF,cAAcF,aAAa;YAEjC,IAAIG;YAEJ,IAAK,IAAIC,UAAU,GAAGA,WAAWF,aAAaE,UAC5C,IAAI;gBACF,MAAMC,SAAS,MAAMtC,WAAW,MAAM,CAAC;oBACrC,OAAO5C;oBACP0C;oBACA,GAAGqB,YAAY;oBACf,GAAGE,eAAe;gBACpB;gBAEAN,WAAWP,KAAK,GAAG,KAAKD;gBAExBF,kBACE,CAAC,OAAO,EAAEjD,UAAU,QAAQ,EAAEM,eAAe,UAAU,mBAAmB,EAAED,mBAAmB,iBAAiB,EAAE6E,OAAO,KAAK,EAAE,iBAAiB,GAAG,qBAAqB,EAAEA,OAAO,KAAK,EAAE,qBAAqB,GAAG,gBAAgB,EAAEA,OAAO,KAAK,EAAE,gBAAgB,GAAG,WAAW,EAAEvB,SAAS,aAAa,EAAEuB,OAAO,WAAW,IAAI,GAAG,eAAe,EAAE7B,eAAe,IAAI;gBAGxWH,mBACE,CAAC,oBAAoB,EAAEiC,KAAK,SAAS,CAACD,OAAO,KAAK,GAAG;gBAGvD,IAAI,CAACA,OAAO,OAAO,EACjB,MAAM,IAAI5D,MACR,CAAC,mCAAmC,EAAE6D,KAAK,SAAS,CAACD,SAAS;gBAIlE3B,UAAU2B,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;gBAC3C,IAAI,CAAC3B,SACH,MAAM,IAAIjC,MAAM;gBAGlBmC,uBACGyB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,EAAU,qBAAqB;gBAC3DxB,QAAQwB,OAAO,KAAK;gBACpB;YACF,EAAE,OAAOtD,OAAO;gBACdoD,YAAYpD;gBACZ,IAAIqD,UAAUF,aAAa;oBACzB9D,QAAQ,IAAI,CACV,CAAC,mCAAmC,EAAEgE,QAAQ,CAAC,EAAEF,YAAY,eAAe,EAAED,cAAc,aAAa,EAAEE,UAAU,OAAO,EAAE;oBAEhI,MAAM,IAAII,QAAQ,CAACC,UAAYC,WAAWD,SAASP;gBACrD;YACF;YAGF,IAAI,CAACvB,SACH,MAAMyB;QAEV;QAEAhC,UAAU,CAAC,4BAA4B,EAAES,sBAAsB;QAC/DT,UAAU,CAAC,kBAAkB,EAAEO,SAAS;QAGxC,IAAID,eAAe,CAACI,OAAO;YAEzB,MAAMgB,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEpB,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;YAEtCG,QAAQ;gBACN,eAAegB;gBACf,mBAAmBA;gBACnB,cAAcA,AAAkB,IAAlBA;YAChB;QACF;QAEA,OAAO;YACL,SAASnB,WAAW;YACpB,mBAAmBE,wBAAwBgB;YAC3C,OAAOb,eAAeF;YACtB,YAAY,CAAC,CAACJ;QAChB;IACF,EAAE,OAAOiC,GAAQ;QACftE,QAAQ,KAAK,CAAC,kBAAkBsE;QAChC,MAAMC,WAAW,IAAIlE,MACnB,CAAC,eAAe,EAAEgC,cAAc,eAAe,GAAG,kBAAkB,EAAEtD,UAAU,GAAG,EAAEuF,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC9J;YACE,OAAOA;QACT;QAEF,MAAMC;IACR;AACF;AAEO,eAAeC,yBACpB/C,QAAsC,EACtC7C,WAAyB,EACzB8C,OAEC;IAOD,MAAM+C,WAAW,MAAMjD,OAAOC,UAAU7C,aAAa;QACnD,WAAW8C,SAAS;IACtB;IACAgD,OAAOD,UAAU;IACjB,MAAMpF,cAAcT,YAAY,WAAW;IAC3C,MAAM+F,cAAcC,cAAcH,SAAS,OAAO,EAAEpF;IACpDqF,OACE,AAAuB,YAAvB,OAAOC,aACP,CAAC,0CAA0C,EAAE/F,YAAY,SAAS,CAAC,GAAG,EAAE6F,SAAS,OAAO,EAAE;IAE5F,OAAO;QACL,SAASE;QACT,eAAeF,SAAS,OAAO;QAC/B,OAAOA,SAAS,KAAK;QACrB,mBAAmBA,SAAS,iBAAiB;IAC/C;AACF;AAEO,eAAeI,yBACpBC,IAAY,EACZlG,WAAyB;IAEzB,MAAM,EAAE0D,OAAO,EAAEG,KAAK,EAAE,GAAG,MAAMjB,OAAOsD,MAAMlG;IAC9C,OAAO;QAAE0D;QAASG;IAAM;AAC1B;AAEO,SAASsC,yBAAyBN,QAAgB;IACvD,IAAI;QAEF,MAAMO,YAAYP,SAAS,KAAK,CAAC;QACjC,IAAIO,WACF,OAAOA,SAAS,CAAC,EAAE;QAIrB,MAAMC,iBAAiBR,SAAS,KAAK,CACnC;QAEF,IAAIQ,gBACF,OAAOA,cAAc,CAAC,EAAE;QAI1B,MAAMC,gBAAgBT,SAAS,KAAK,CAAC;QACrC,IAAIS,eACF,OAAOA,aAAa,CAAC,EAAE;IAE3B,EAAE,OAAM,CAAC;IAET,OAAOT;AACT;AAEO,SAASU,yBAAyBC,KAAa;IACpD,IAAIA,MAAM,QAAQ,CAAC,SAEjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;IAG5C,OAAOA;AACT;AAEO,SAASjC,uBAAuB,EACrCkC,SAAS,EACThG,WAAW,EAIZ;IAKC,MAAMiG,sBAAsBD,AAAc,YAAdA,YAAwB7B,SAAY6B;IAEhE,IAAIC,AAAwB9B,WAAxB8B,qBACF,OAAO;QAAE,QAAQ,CAAC;QAAG,cAAc9B;IAAU;IAG/C,IAAInE,AAAgB,eAAhBA,aACF,OAAO;QACL,QAAQ;YAAE,iBAAiBiG;QAAoB;QAC/C,cAAc,CAAC,oCAAoC,EAAEA,oBAAoB,aAAa,CAAC;IACzF;IAGF,IAAIjG,AAAgB,oBAAhBA,aACF,OAAO;QACL,QAAQ;YACN,UAAU;gBAAE,MAAMiG,sBAAsB,YAAY;YAAW;QACjE;QACA,cAAc,CAAC,kCAAkC,EAAEA,sBAAsB,YAAY,WAAW,kBAAkB,CAAC;IACrH;IAGF,IAAIjG,AAAgB,YAAhBA,aACF,OAAO;QACL,QAAQ;YACN,UAAU;gBAAE,MAAMiG,sBAAsB,YAAY;YAAW;QACjE;QACA,cAAc,CAAC,kCAAkC,EAAEA,sBAAsB,YAAY,WAAW,UAAU,CAAC;IAC7G;IAGF,IAAIjG,AAAgB,YAAhBA,aACF,OAAO;QACL,QAAQiG,sBACJ;YACE,WAAW;gBAAE,QAAQ;YAAO;QAC9B,IACA;YACE,WAAW;gBAAE,QAAQ;YAAM;QAC7B;QACJ,cAAcA,sBACV,wDACA;IACN;IAGF,OAAO;QACL,QAAQ,CAAC;QACT,cAAc,CAAC,6CAA6C,EAAEjG,eAAe,UAAU,CAAC,CAAC;QACzF,gBAAgB,CAAC,0DAA0D,EAAEA,eAAe,UAAU,EAAE,CAAC;IAC3G;AACF;AAQA,SAASkG,oBAAoBC,GAAQ;IAEnC,IAAIA,QAAAA,KACF,OAAOA;IAIT,IAAIC,MAAM,OAAO,CAACD,MAChB,OAAOA,IAAI,GAAG,CAAC,CAACE,OAASH,oBAAoBG;IAI/C,IAAI,AAAe,YAAf,OAAOF,KAAkB;QAC3B,MAAMG,aAAkB,CAAC;QAEzB,KAAK,MAAM,CAACC,KAAKC,MAAM,IAAIC,OAAO,OAAO,CAACN,KAAM;YAE9C,MAAMO,aAAaH,IAAI,IAAI;YAG3B,IAAII,kBAAkBT,oBAAoBM;YAG1C,IAAI,AAA2B,YAA3B,OAAOG,iBACTA,kBAAkBA,gBAAgB,IAAI;YAGxCL,UAAU,CAACI,WAAW,GAAGC;QAC3B;QAEA,OAAOL;IACT;IAGA,IAAI,AAAe,YAAf,OAAOH,KACT,OAAOA,IAAI,IAAI;IAIjB,OAAOA;AACT;AAEO,SAASZ,cACdQ,KAAa,EACb/F,WAAqC;IAErC,MAAM4G,kBAAkBlB,yBAAyBK;IAEjD,IAAIa,iBAAiB,MAAM,oBACzB,OAAOA,gBACJ,KAAK,CAAC,oBACL,MAAM,GACP,IAAI1F;IAGT,IAAIV;IACJ,IAAIkE;IACJ,IAAI;QACFlE,SAASqE,KAAK,KAAK,CAAC+B;QACpB,OAAOV,oBAAoB1F;IAC7B,EAAE,OAAOc,OAAO;QACdoD,YAAYpD;IACd;IACA,IAAI;QACFd,SAASqE,KAAK,KAAK,CAACgC,WAAWD;QAC/B,OAAOV,oBAAoB1F;IAC7B,EAAE,OAAOc,OAAO;QACdoD,YAAYpD;IACd;IAEA,IAAItB,AAAgB,oBAAhBA,eAAmC8G,SAAS9G,cAAc;QAC5D,MAAM+G,aAAajB,yBAAyBc;QAC5C,IAAI;YACFpG,SAASqE,KAAK,KAAK,CAACgC,WAAWE;YAC/B,OAAOb,oBAAoB1F;QAC7B,EAAE,OAAOc,OAAO;YACdoD,YAAYpD;QACd;IACF;IACA,MAAMN,MACJ,CAAC,gDAAgD,EAAEgG,OACjDtC,aAAa,iBACb,gBAAgB,EAAEqB,OAAO;AAE/B"}
+{"version":3,"file":"ai-model/service-caller/index.mjs","sources":["../../../../src/ai-model/service-caller/index.ts"],"sourcesContent":["import type { AIUsageInfo, DeepThinkOption } from '@/types';\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\n\n// Error class that preserves usage and rawResponse when AI call parsing fails\nexport class AIResponseParseError extends Error {\n  usage?: AIUsageInfo;\n  rawResponse: string;\n\n  constructor(message: string, rawResponse: string, usage?: AIUsageInfo) {\n    super(message);\n    this.name = 'AIResponseParseError';\n    this.rawResponse = rawResponse;\n    this.usage = usage;\n  }\n}\nimport {\n  type IModelConfig,\n  MIDSCENE_LANGFUSE_DEBUG,\n  MIDSCENE_LANGSMITH_DEBUG,\n  MIDSCENE_MODEL_MAX_TOKENS,\n  OPENAI_MAX_TOKENS,\n  type TModelFamily,\n  type UITarsModelVersion,\n  globalConfigManager,\n} from '@midscene/shared/env';\n\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert, ifInBrowser } from '@midscene/shared/utils';\nimport { jsonrepair } from 'jsonrepair';\nimport OpenAI from 'openai';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { Stream } from 'openai/streaming';\nimport type { AIArgs } from '../../common';\nimport { isAutoGLM, isUITars } from '../auto-glm/util';\n\nasync function createChatClient({\n  modelConfig,\n}: {\n  modelConfig: IModelConfig;\n}): Promise<{\n  completion: OpenAI.Chat.Completions;\n  modelName: string;\n  modelDescription: string;\n  uiTarsModelVersion?: UITarsModelVersion;\n  modelFamily: TModelFamily | undefined;\n}> {\n  const {\n    socksProxy,\n    httpProxy,\n    modelName,\n    openaiBaseURL,\n    openaiApiKey,\n    openaiExtraConfig,\n    modelDescription,\n    uiTarsModelVersion,\n    modelFamily,\n    createOpenAIClient,\n    timeout,\n  } = modelConfig;\n\n  let proxyAgent: any = undefined;\n  const debugProxy = getDebug('ai:call:proxy');\n\n  // Helper function to sanitize proxy URL for logging (remove credentials)\n  // Uses URL API instead of regex to avoid ReDoS vulnerabilities\n  const sanitizeProxyUrl = (url: string): string => {\n    try {\n      const parsed = new URL(url);\n      if (parsed.username) {\n        // Keep username for debugging, hide password for security\n        parsed.password = '****';\n        return parsed.href;\n      }\n      return url;\n    } catch {\n      // If URL parsing fails, return original URL (will be caught later)\n      return url;\n    }\n  };\n\n  if (httpProxy) {\n    debugProxy('using http proxy', sanitizeProxyUrl(httpProxy));\n    if (ifInBrowser) {\n      console.warn(\n        'HTTP proxy is configured but not supported in browser environment',\n      );\n    } else {\n      // Dynamic import with variable to avoid bundler static analysis\n      const moduleName = 'undici';\n      const { ProxyAgent } = await import(moduleName);\n      proxyAgent = new ProxyAgent({\n        uri: httpProxy,\n        // Note: authentication is handled via the URI (e.g., http://user:pass@proxy.com:8080)\n      });\n    }\n  } else if (socksProxy) {\n    debugProxy('using socks proxy', sanitizeProxyUrl(socksProxy));\n    if (ifInBrowser) {\n      console.warn(\n        'SOCKS proxy is configured but not supported in browser environment',\n      );\n    } else {\n      try {\n        // Dynamic import with variable to avoid bundler static analysis\n        const moduleName = 'fetch-socks';\n        const { socksDispatcher } = await import(moduleName);\n        // Parse SOCKS proxy URL (e.g., socks5://127.0.0.1:1080)\n        const proxyUrl = new URL(socksProxy);\n\n        // Validate hostname\n        if (!proxyUrl.hostname) {\n          throw new Error('SOCKS proxy URL must include a valid hostname');\n        }\n\n        // Validate and parse port\n        const port = Number.parseInt(proxyUrl.port, 10);\n        if (!proxyUrl.port || Number.isNaN(port)) {\n          throw new Error('SOCKS proxy URL must include a valid port');\n        }\n\n        // Parse SOCKS version from protocol\n        const protocol = proxyUrl.protocol.replace(':', '');\n        const socksType =\n          protocol === 'socks4' ? 4 : protocol === 'socks5' ? 5 : 5;\n\n        proxyAgent = socksDispatcher({\n          type: socksType,\n          host: proxyUrl.hostname,\n          port,\n          ...(proxyUrl.username\n            ? {\n                userId: decodeURIComponent(proxyUrl.username),\n                password: decodeURIComponent(proxyUrl.password || ''),\n              }\n            : {}),\n        });\n        debugProxy('socks proxy configured successfully', {\n          type: socksType,\n          host: proxyUrl.hostname,\n          port: port,\n        });\n      } catch (error) {\n        console.error('Failed to configure SOCKS proxy:', error);\n        throw new Error(\n          `Invalid SOCKS proxy URL: ${socksProxy}. Expected format: socks4://host:port, socks5://host:port, or with authentication: socks5://user:pass@host:port`,\n        );\n      }\n    }\n  }\n\n  const openAIOptions = {\n    baseURL: openaiBaseURL,\n    apiKey: openaiApiKey,\n    // Use fetchOptions.dispatcher for fetch-based SDK instead of httpAgent\n    // Note: Type assertion needed due to undici version mismatch between dependencies\n    ...(proxyAgent ? { fetchOptions: { dispatcher: proxyAgent as any } } : {}),\n    ...openaiExtraConfig,\n    ...(typeof timeout === 'number' ? { timeout } : {}),\n    dangerouslyAllowBrowser: true,\n  };\n\n  const baseOpenAI = new OpenAI(openAIOptions);\n\n  let openai: OpenAI = baseOpenAI;\n\n  // LangSmith wrapper\n  if (\n    openai &&\n    globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)\n  ) {\n    if (ifInBrowser) {\n      throw new Error('langsmith is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langsmith wrapper enabled');\n    // Use variable to prevent static analysis by bundlers\n    const langsmithModule = 'langsmith/wrappers';\n    const { wrapOpenAI } = await import(langsmithModule);\n    openai = wrapOpenAI(openai);\n  }\n\n  // Langfuse wrapper\n  if (\n    openai &&\n    globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGFUSE_DEBUG)\n  ) {\n    if (ifInBrowser) {\n      throw new Error('langfuse is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langfuse wrapper enabled');\n    // Use variable to prevent static analysis by bundlers\n    const langfuseModule = '@langfuse/openai';\n    const { observeOpenAI } = await import(langfuseModule);\n    openai = observeOpenAI(openai);\n  }\n\n  if (createOpenAIClient) {\n    const wrappedClient = await createOpenAIClient(baseOpenAI, openAIOptions);\n\n    if (wrappedClient) {\n      openai = wrappedClient as OpenAI;\n    }\n  }\n\n  return {\n    completion: openai.chat.completions,\n    modelName,\n    modelDescription,\n    uiTarsModelVersion,\n    modelFamily,\n  };\n}\n\nexport async function callAI(\n  messages: ChatCompletionMessageParam[],\n  modelConfig: IModelConfig,\n  options?: {\n    stream?: boolean;\n    onChunk?: StreamingCallback;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<{\n  content: string;\n  reasoning_content?: string;\n  usage?: AIUsageInfo;\n  isStreamed: boolean;\n}> {\n  const {\n    completion,\n    modelName,\n    modelDescription,\n    uiTarsModelVersion,\n    modelFamily,\n  } = await createChatClient({\n    modelConfig,\n  });\n\n  const maxTokens =\n    globalConfigManager.getEnvConfigValueAsNumber(MIDSCENE_MODEL_MAX_TOKENS) ??\n    globalConfigManager.getEnvConfigValueAsNumber(OPENAI_MAX_TOKENS);\n  const debugCall = getDebug('ai:call');\n  const debugProfileStats = getDebug('ai:profile:stats');\n  const debugProfileDetail = getDebug('ai:profile:detail');\n\n  const startTime = Date.now();\n  const temperature = modelConfig.temperature ?? 0;\n\n  const isStreaming = options?.stream && options?.onChunk;\n  let content: string | undefined;\n  let accumulated = '';\n  let accumulatedReasoning = '';\n  let usage: OpenAI.CompletionUsage | undefined;\n  let timeCost: number | undefined;\n\n  const buildUsageInfo = (usageData?: OpenAI.CompletionUsage) => {\n    if (!usageData) return undefined;\n\n    const cachedInputTokens = (\n      usageData as { prompt_tokens_details?: { cached_tokens?: number } }\n    )?.prompt_tokens_details?.cached_tokens;\n\n    return {\n      prompt_tokens: usageData.prompt_tokens ?? 0,\n      completion_tokens: usageData.completion_tokens ?? 0,\n      total_tokens: usageData.total_tokens ?? 0,\n      cached_input: cachedInputTokens ?? 0,\n      time_cost: timeCost ?? 0,\n      model_name: modelName,\n      model_description: modelDescription,\n      intent: modelConfig.intent,\n    } satisfies AIUsageInfo;\n  };\n\n  const commonConfig = {\n    temperature,\n    stream: !!isStreaming,\n    max_tokens: maxTokens,\n    ...(modelFamily === 'qwen2.5-vl' // qwen vl v2 specific config\n      ? {\n          vl_high_resolution_images: true,\n        }\n      : {}),\n  };\n\n  if (isAutoGLM(modelFamily)) {\n    (commonConfig as unknown as Record<string, number>).top_p = 0.85;\n    (commonConfig as unknown as Record<string, number>).frequency_penalty = 0.2;\n  }\n\n  const {\n    config: deepThinkConfig,\n    debugMessage,\n    warningMessage,\n  } = resolveDeepThinkConfig({\n    deepThink: options?.deepThink,\n    modelFamily,\n  });\n  if (debugMessage) {\n    debugCall(debugMessage);\n  }\n  if (warningMessage) {\n    debugCall(warningMessage);\n    console.warn(warningMessage);\n  }\n\n  try {\n    debugCall(\n      `sending ${isStreaming ? 'streaming ' : ''}request to ${modelName}`,\n    );\n\n    if (isStreaming) {\n      const stream = (await completion.create(\n        {\n          model: modelName,\n          messages,\n          ...commonConfig,\n          ...deepThinkConfig,\n        },\n        {\n          stream: true,\n        },\n      )) as Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {\n        _request_id?: string | null;\n      };\n\n      for await (const chunk of stream) {\n        const content = chunk.choices?.[0]?.delta?.content || '';\n        const reasoning_content =\n          (chunk.choices?.[0]?.delta as any)?.reasoning_content || '';\n\n        // Check for usage info in any chunk (OpenAI provides usage in separate chunks)\n        if (chunk.usage) {\n          usage = chunk.usage;\n        }\n\n        if (content || reasoning_content) {\n          accumulated += content;\n          accumulatedReasoning += reasoning_content;\n          const chunkData: CodeGenerationChunk = {\n            content,\n            reasoning_content,\n            accumulated,\n            isComplete: false,\n            usage: undefined,\n          };\n          options.onChunk!(chunkData);\n        }\n\n        // Check if stream is complete\n        if (chunk.choices?.[0]?.finish_reason) {\n          timeCost = Date.now() - startTime;\n\n          // If usage is not available from the stream, provide a basic usage info\n          if (!usage) {\n            // Estimate token counts based on content length (rough approximation)\n            const estimatedTokens = Math.max(\n              1,\n              Math.floor(accumulated.length / 4),\n            );\n            usage = {\n              prompt_tokens: estimatedTokens,\n              completion_tokens: estimatedTokens,\n              total_tokens: estimatedTokens * 2,\n            };\n          }\n\n          // Send final chunk\n          const finalChunk: CodeGenerationChunk = {\n            content: '',\n            accumulated,\n            reasoning_content: '',\n            isComplete: true,\n            usage: buildUsageInfo(usage),\n          };\n          options.onChunk!(finalChunk);\n          break;\n        }\n      }\n      content = accumulated;\n      debugProfileStats(\n        `streaming model, ${modelName}, mode, ${modelFamily || 'default'}, cost-ms, ${timeCost}, temperature, ${temperature ?? ''}`,\n      );\n    } else {\n      // Non-streaming with retry logic\n      const retryCount = modelConfig.retryCount ?? 1;\n      const retryInterval = modelConfig.retryInterval ?? 2000;\n      const maxAttempts = retryCount + 1; // retryCount=1 means 2 total attempts (1 initial + 1 retry)\n\n      let lastError: Error | undefined;\n\n      for (let attempt = 1; attempt <= maxAttempts; attempt++) {\n        try {\n          const result = await completion.create({\n            model: modelName,\n            messages,\n            ...commonConfig,\n            ...deepThinkConfig,\n          } as any);\n\n          timeCost = Date.now() - startTime;\n\n          debugProfileStats(\n            `model, ${modelName}, mode, ${modelFamily || 'default'}, ui-tars-version, ${uiTarsModelVersion}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}, temperature, ${temperature ?? ''}`,\n          );\n\n          debugProfileDetail(\n            `model usage detail: ${JSON.stringify(result.usage)}`,\n          );\n\n          if (!result.choices) {\n            throw new Error(\n              `invalid response from LLM service: ${JSON.stringify(result)}`,\n            );\n          }\n\n          content = result.choices[0].message.content!;\n          if (!content) {\n            throw new Error('empty content from AI model');\n          }\n\n          accumulatedReasoning =\n            (result.choices[0].message as any)?.reasoning_content || '';\n          usage = result.usage;\n          break; // Success, exit retry loop\n        } catch (error) {\n          lastError = error as Error;\n          if (attempt < maxAttempts) {\n            console.warn(\n              `[Midscene] AI call failed (attempt ${attempt}/${maxAttempts}), retrying in ${retryInterval}ms... Error: ${lastError.message}`,\n            );\n            await new Promise((resolve) => setTimeout(resolve, retryInterval));\n          }\n        }\n      }\n\n      if (!content) {\n        throw lastError;\n      }\n    }\n\n    debugCall(`response reasoning content: ${accumulatedReasoning}`);\n    debugCall(`response content: ${content}`);\n\n    // Ensure we always have usage info for streaming responses\n    if (isStreaming && !usage) {\n      // Estimate token counts based on content length (rough approximation)\n      const estimatedTokens = Math.max(\n        1,\n        Math.floor((content || '').length / 4),\n      );\n      usage = {\n        prompt_tokens: estimatedTokens,\n        completion_tokens: estimatedTokens,\n        total_tokens: estimatedTokens * 2,\n      } as OpenAI.CompletionUsage;\n    }\n\n    return {\n      content: content || '',\n      reasoning_content: accumulatedReasoning || undefined,\n      usage: buildUsageInfo(usage),\n      isStreamed: !!isStreaming,\n    };\n  } catch (e: any) {\n    console.error(' call AI error', e);\n    const newError = new Error(\n      `failed to call ${isStreaming ? 'streaming ' : ''}AI model service (${modelName}): ${e.message}\\nTrouble shooting: https://midscenejs.com/model-provider.html`,\n      {\n        cause: e,\n      },\n    );\n    throw newError;\n  }\n}\n\nexport async function callAIWithObjectResponse<T>(\n  messages: ChatCompletionMessageParam[],\n  modelConfig: IModelConfig,\n  options?: {\n    deepThink?: DeepThinkOption;\n  },\n): Promise<{\n  content: T;\n  contentString: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}> {\n  const response = await callAI(messages, modelConfig, {\n    deepThink: options?.deepThink,\n  });\n  assert(response, 'empty response');\n  const modelFamily = modelConfig.modelFamily;\n  const jsonContent = safeParseJson(response.content, modelFamily);\n  if (typeof jsonContent !== 'object') {\n    throw new AIResponseParseError(\n      `failed to parse json response from model (${modelConfig.modelName}): ${response.content}`,\n      response.content,\n      response.usage,\n    );\n  }\n  return {\n    content: jsonContent,\n    contentString: response.content,\n    usage: response.usage,\n    reasoning_content: response.reasoning_content,\n  };\n}\n\nexport async function callAIWithStringResponse(\n  msgs: AIArgs,\n  modelConfig: IModelConfig,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n  const { content, usage } = await callAI(msgs, modelConfig);\n  return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n  try {\n    // First, try to match a JSON object directly in the response\n    const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\n    if (jsonMatch) {\n      return jsonMatch[1];\n    }\n\n    // If no direct JSON object is found, try to extract JSON from a code block\n    const codeBlockMatch = response.match(\n      /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\n    );\n    if (codeBlockMatch) {\n      return codeBlockMatch[1];\n    }\n\n    // If no code block is found, try to find a JSON-like structure in the text\n    const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\n    if (jsonLikeMatch) {\n      return jsonLikeMatch[0];\n    }\n  } catch {}\n  // If no JSON-like structure is found, return the original response\n  return response;\n}\n\nexport function preprocessDoubaoBboxJson(input: string) {\n  if (input.includes('bbox')) {\n    // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\n    while (/\\d+\\s+\\d+/.test(input)) {\n      input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n    }\n  }\n  return input;\n}\n\nexport function resolveDeepThinkConfig({\n  deepThink,\n  modelFamily,\n}: {\n  deepThink?: DeepThinkOption;\n  modelFamily?: TModelFamily;\n}): {\n  config: Record<string, unknown>;\n  debugMessage?: string;\n  warningMessage?: string;\n} {\n  const normalizedDeepThink = deepThink === 'unset' ? undefined : deepThink;\n\n  if (normalizedDeepThink === undefined) {\n    return { config: {}, debugMessage: undefined };\n  }\n\n  if (modelFamily === 'qwen3-vl') {\n    return {\n      config: { enable_thinking: normalizedDeepThink },\n      debugMessage: `deepThink mapped to enable_thinking=${normalizedDeepThink} for qwen3-vl`,\n    };\n  }\n\n  if (modelFamily === 'doubao-vision') {\n    return {\n      config: {\n        thinking: { type: normalizedDeepThink ? 'enabled' : 'disabled' },\n      },\n      debugMessage: `deepThink mapped to thinking.type=${normalizedDeepThink ? 'enabled' : 'disabled'} for doubao-vision`,\n    };\n  }\n\n  if (modelFamily === 'glm-v') {\n    return {\n      config: {\n        thinking: { type: normalizedDeepThink ? 'enabled' : 'disabled' },\n      },\n      debugMessage: `deepThink mapped to thinking.type=${normalizedDeepThink ? 'enabled' : 'disabled'} for glm-v`,\n    };\n  }\n\n  if (modelFamily === 'gpt-5') {\n    return {\n      config: normalizedDeepThink\n        ? {\n            reasoning: { effort: 'high' },\n          }\n        : {\n            reasoning: { effort: 'low' },\n          },\n      debugMessage: normalizedDeepThink\n        ? 'deepThink mapped to reasoning.effort=high for gpt-5'\n        : 'deepThink disabled for gpt-5',\n    };\n  }\n\n  return {\n    config: {},\n    debugMessage: `deepThink ignored: unsupported model_family \"${modelFamily ?? 'default'}\"`,\n    warningMessage: `The \"deepThink\" option is not supported for model_family \"${modelFamily ?? 'default'}\".`,\n  };\n}\n\n/**\n * Normalize a parsed JSON object by trimming whitespace from:\n * 1. All object keys (e.g., \" prompt \" -> \"prompt\")\n * 2. All string values (e.g., \" Tap \" -> \"Tap\")\n * This handles LLM output that may include leading/trailing spaces.\n */\nfunction normalizeJsonObject(obj: any): any {\n  // Handle null and undefined\n  if (obj === null || obj === undefined) {\n    return obj;\n  }\n\n  // Handle arrays - recursively normalize each element\n  if (Array.isArray(obj)) {\n    return obj.map((item) => normalizeJsonObject(item));\n  }\n\n  // Handle objects\n  if (typeof obj === 'object') {\n    const normalized: any = {};\n\n    for (const [key, value] of Object.entries(obj)) {\n      // Trim the key to remove leading/trailing spaces\n      const trimmedKey = key.trim();\n\n      // Recursively normalize the value\n      let normalizedValue = normalizeJsonObject(value);\n\n      // Trim all string values\n      if (typeof normalizedValue === 'string') {\n        normalizedValue = normalizedValue.trim();\n      }\n\n      normalized[trimmedKey] = normalizedValue;\n    }\n\n    return normalized;\n  }\n\n  // Handle primitive strings\n  if (typeof obj === 'string') {\n    return obj.trim();\n  }\n\n  // Return other primitives as-is\n  return obj;\n}\n\nexport function safeParseJson(\n  input: string,\n  modelFamily: TModelFamily | undefined,\n) {\n  const cleanJsonString = extractJSONFromCodeBlock(input);\n  // match the point\n  if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\n    return cleanJsonString\n      .match(/\\((\\d+),(\\d+)\\)/)\n      ?.slice(1)\n      .map(Number);\n  }\n\n  let parsed: any;\n  let lastError: unknown;\n  try {\n    parsed = JSON.parse(cleanJsonString);\n    return normalizeJsonObject(parsed);\n  } catch (error) {\n    lastError = error;\n  }\n  try {\n    parsed = JSON.parse(jsonrepair(cleanJsonString));\n    return normalizeJsonObject(parsed);\n  } catch (error) {\n    lastError = error;\n  }\n\n  if (modelFamily === 'doubao-vision' || isUITars(modelFamily)) {\n    const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\n    try {\n      parsed = JSON.parse(jsonrepair(jsonString));\n      return normalizeJsonObject(parsed);\n    } catch (error) {\n      lastError = error;\n    }\n  }\n  throw Error(\n    `failed to parse LLM response into JSON. Error - ${String(\n      lastError ?? 'unknown error',\n    )}. Response - \\n ${input}`,\n  );\n}\n"],"names":["AIResponseParseError","Error","message","rawResponse","usage","createChatClient","modelConfig","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","modelDescription","uiTarsModelVersion","modelFamily","createOpenAIClient","timeout","proxyAgent","debugProxy","getDebug","sanitizeProxyUrl","url","parsed","URL","ifInBrowser","console","moduleName","ProxyAgent","socksDispatcher","proxyUrl","port","Number","protocol","socksType","decodeURIComponent","error","openAIOptions","baseOpenAI","OpenAI","openai","globalConfigManager","MIDSCENE_LANGSMITH_DEBUG","langsmithModule","wrapOpenAI","MIDSCENE_LANGFUSE_DEBUG","langfuseModule","observeOpenAI","wrappedClient","callAI","messages","options","completion","maxTokens","MIDSCENE_MODEL_MAX_TOKENS","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","temperature","isStreaming","content","accumulated","accumulatedReasoning","timeCost","buildUsageInfo","usageData","cachedInputTokens","commonConfig","isAutoGLM","deepThinkConfig","debugMessage","warningMessage","resolveDeepThinkConfig","stream","chunk","reasoning_content","chunkData","undefined","estimatedTokens","Math","finalChunk","retryCount","retryInterval","maxAttempts","lastError","attempt","result","JSON","Promise","resolve","setTimeout","e","newError","callAIWithObjectResponse","response","assert","jsonContent","safeParseJson","callAIWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","deepThink","normalizedDeepThink","normalizeJsonObject","obj","Array","item","normalized","key","value","Object","trimmedKey","normalizedValue","cleanJsonString","jsonrepair","isUITars","jsonString","String"],"mappings":";;;;;;;;;;;;;;;;AAIO,MAAMA,6BAA6BC;IAIxC,YAAYC,OAAe,EAAEC,WAAmB,EAAEC,KAAmB,CAAE;QACrE,KAAK,CAACF,UAJR,yCACA;QAIE,IAAI,CAAC,IAAI,GAAG;QACZ,IAAI,CAAC,WAAW,GAAGC;QACnB,IAAI,CAAC,KAAK,GAAGC;IACf;AACF;AAqBA,eAAeC,iBAAiB,EAC9BC,WAAW,EAGZ;IAOC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,gBAAgB,EAChBC,kBAAkB,EAClBC,WAAW,EACXC,kBAAkB,EAClBC,OAAO,EACR,GAAGX;IAEJ,IAAIY;IACJ,MAAMC,aAAaC,SAAS;IAI5B,MAAMC,mBAAmB,CAACC;QACxB,IAAI;YACF,MAAMC,SAAS,IAAIC,IAAIF;YACvB,IAAIC,OAAO,QAAQ,EAAE;gBAEnBA,OAAO,QAAQ,GAAG;gBAClB,OAAOA,OAAO,IAAI;YACpB;YACA,OAAOD;QACT,EAAE,OAAM;YAEN,OAAOA;QACT;IACF;IAEA,IAAId,WAAW;QACbW,WAAW,oBAAoBE,iBAAiBb;QAChD,IAAIiB,aACFC,QAAQ,IAAI,CACV;aAEG;YAEL,MAAMC,aAAa;YACnB,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAACD;YACpCT,aAAa,IAAIU,WAAW;gBAC1B,KAAKpB;YAEP;QACF;IACF,OAAO,IAAID,YAAY;QACrBY,WAAW,qBAAqBE,iBAAiBd;QACjD,IAAIkB,aACFC,QAAQ,IAAI,CACV;aAGF,IAAI;YAEF,MAAMC,aAAa;YACnB,MAAM,EAAEE,eAAe,EAAE,GAAG,MAAM,MAAM,CAACF;YAEzC,MAAMG,WAAW,IAAIN,IAAIjB;YAGzB,IAAI,CAACuB,SAAS,QAAQ,EACpB,MAAM,IAAI7B,MAAM;YAIlB,MAAM8B,OAAOC,OAAO,QAAQ,CAACF,SAAS,IAAI,EAAE;YAC5C,IAAI,CAACA,SAAS,IAAI,IAAIE,OAAO,KAAK,CAACD,OACjC,MAAM,IAAI9B,MAAM;YAIlB,MAAMgC,WAAWH,SAAS,QAAQ,CAAC,OAAO,CAAC,KAAK;YAChD,MAAMI,YACJD,AAAa,aAAbA,WAAwB,IAAIA,AAAa,aAAbA,WAAwB,IAAI;YAE1Df,aAAaW,gBAAgB;gBAC3B,MAAMK;gBACN,MAAMJ,SAAS,QAAQ;gBACvBC;gBACA,GAAID,SAAS,QAAQ,GACjB;oBACE,QAAQK,mBAAmBL,SAAS,QAAQ;oBAC5C,UAAUK,mBAAmBL,SAAS,QAAQ,IAAI;gBACpD,IACA,CAAC,CAAC;YACR;YACAX,WAAW,uCAAuC;gBAChD,MAAMe;gBACN,MAAMJ,SAAS,QAAQ;gBACvB,MAAMC;YACR;QACF,EAAE,OAAOK,OAAO;YACdV,QAAQ,KAAK,CAAC,oCAAoCU;YAClD,MAAM,IAAInC,MACR,CAAC,yBAAyB,EAAEM,WAAW,+GAA+G,CAAC;QAE3J;IAEJ;IAEA,MAAM8B,gBAAgB;QACpB,SAAS3B;QACT,QAAQC;QAGR,GAAIO,aAAa;YAAE,cAAc;gBAAE,YAAYA;YAAkB;QAAE,IAAI,CAAC,CAAC;QACzE,GAAGN,iBAAiB;QACpB,GAAI,AAAmB,YAAnB,OAAOK,UAAuB;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAClD,yBAAyB;IAC3B;IAEA,MAAMqB,aAAa,IAAIC,SAAOF;IAE9B,IAAIG,SAAiBF;IAGrB,IACEE,UACAC,oBAAoB,qBAAqB,CAACC,2BAC1C;QACA,IAAIjB,aACF,MAAM,IAAIxB,MAAM;QAElByB,QAAQ,GAAG,CAAC;QAEZ,MAAMiB,kBAAkB;QACxB,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAACD;QACpCH,SAASI,WAAWJ;IACtB;IAGA,IACEA,UACAC,oBAAoB,qBAAqB,CAACI,0BAC1C;QACA,IAAIpB,aACF,MAAM,IAAIxB,MAAM;QAElByB,QAAQ,GAAG,CAAC;QAEZ,MAAMoB,iBAAiB;QACvB,MAAM,EAAEC,aAAa,EAAE,GAAG,MAAM,MAAM,CAACD;QACvCN,SAASO,cAAcP;IACzB;IAEA,IAAIxB,oBAAoB;QACtB,MAAMgC,gBAAgB,MAAMhC,mBAAmBsB,YAAYD;QAE3D,IAAIW,eACFR,SAASQ;IAEb;IAEA,OAAO;QACL,YAAYR,OAAO,IAAI,CAAC,WAAW;QACnC/B;QACAI;QACAC;QACAC;IACF;AACF;AAEO,eAAekC,OACpBC,QAAsC,EACtC5C,WAAyB,EACzB6C,OAIC;IAOD,MAAM,EACJC,UAAU,EACV3C,SAAS,EACTI,gBAAgB,EAChBC,kBAAkB,EAClBC,WAAW,EACZ,GAAG,MAAMV,iBAAiB;QACzBC;IACF;IAEA,MAAM+C,YACJZ,oBAAoB,yBAAyB,CAACa,8BAC9Cb,oBAAoB,yBAAyB,CAACc;IAChD,MAAMC,YAAYpC,SAAS;IAC3B,MAAMqC,oBAAoBrC,SAAS;IACnC,MAAMsC,qBAAqBtC,SAAS;IAEpC,MAAMuC,YAAYC,KAAK,GAAG;IAC1B,MAAMC,cAAcvD,YAAY,WAAW,IAAI;IAE/C,MAAMwD,cAAcX,SAAS,UAAUA,SAAS;IAChD,IAAIY;IACJ,IAAIC,cAAc;IAClB,IAAIC,uBAAuB;IAC3B,IAAI7D;IACJ,IAAI8D;IAEJ,MAAMC,iBAAiB,CAACC;QACtB,IAAI,CAACA,WAAW;QAEhB,MAAMC,oBACJD,WACC,uBAAuB;QAE1B,OAAO;YACL,eAAeA,UAAU,aAAa,IAAI;YAC1C,mBAAmBA,UAAU,iBAAiB,IAAI;YAClD,cAAcA,UAAU,YAAY,IAAI;YACxC,cAAcC,qBAAqB;YACnC,WAAWH,YAAY;YACvB,YAAYzD;YACZ,mBAAmBI;YACnB,QAAQP,YAAY,MAAM;QAC5B;IACF;IAEA,MAAMgE,eAAe;QACnBT;QACA,QAAQ,CAAC,CAACC;QACV,YAAYT;QACZ,GAAItC,AAAgB,iBAAhBA,cACA;YACE,2BAA2B;QAC7B,IACA,CAAC,CAAC;IACR;IAEA,IAAIwD,UAAUxD,cAAc;QACzBuD,aAAmD,KAAK,GAAG;QAC3DA,aAAmD,iBAAiB,GAAG;IAC1E;IAEA,MAAM,EACJ,QAAQE,eAAe,EACvBC,YAAY,EACZC,cAAc,EACf,GAAGC,uBAAuB;QACzB,WAAWxB,SAAS;QACpBpC;IACF;IACA,IAAI0D,cACFjB,UAAUiB;IAEZ,IAAIC,gBAAgB;QAClBlB,UAAUkB;QACVhD,QAAQ,IAAI,CAACgD;IACf;IAEA,IAAI;QACFlB,UACE,CAAC,QAAQ,EAAEM,cAAc,eAAe,GAAG,WAAW,EAAErD,WAAW;QAGrE,IAAIqD,aAAa;YACf,MAAMc,SAAU,MAAMxB,WAAW,MAAM,CACrC;gBACE,OAAO3C;gBACPyC;gBACA,GAAGoB,YAAY;gBACf,GAAGE,eAAe;YACpB,GACA;gBACE,QAAQ;YACV;YAKF,WAAW,MAAMK,SAASD,OAAQ;gBAChC,MAAMb,UAAUc,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,OAAO,WAAW;gBACtD,MAAMC,oBACHD,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,OAAe,qBAAqB;gBAG3D,IAAIA,MAAM,KAAK,EACbzE,QAAQyE,MAAM,KAAK;gBAGrB,IAAId,WAAWe,mBAAmB;oBAChCd,eAAeD;oBACfE,wBAAwBa;oBACxB,MAAMC,YAAiC;wBACrChB;wBACAe;wBACAd;wBACA,YAAY;wBACZ,OAAOgB;oBACT;oBACA7B,QAAQ,OAAO,CAAE4B;gBACnB;gBAGA,IAAIF,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,eAAe;oBACrCX,WAAWN,KAAK,GAAG,KAAKD;oBAGxB,IAAI,CAACvD,OAAO;wBAEV,MAAM6E,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAClB,YAAY,MAAM,GAAG;wBAElC5D,QAAQ;4BACN,eAAe6E;4BACf,mBAAmBA;4BACnB,cAAcA,AAAkB,IAAlBA;wBAChB;oBACF;oBAGA,MAAME,aAAkC;wBACtC,SAAS;wBACTnB;wBACA,mBAAmB;wBACnB,YAAY;wBACZ,OAAOG,eAAe/D;oBACxB;oBACA+C,QAAQ,OAAO,CAAEgC;oBACjB;gBACF;YACF;YACApB,UAAUC;YACVP,kBACE,CAAC,iBAAiB,EAAEhD,UAAU,QAAQ,EAAEM,eAAe,UAAU,WAAW,EAAEmD,SAAS,eAAe,EAAEL,eAAe,IAAI;QAE/H,OAAO;YAEL,MAAMuB,aAAa9E,YAAY,UAAU,IAAI;YAC7C,MAAM+E,gBAAgB/E,YAAY,aAAa,IAAI;YACnD,MAAMgF,cAAcF,aAAa;YAEjC,IAAIG;YAEJ,IAAK,IAAIC,UAAU,GAAGA,WAAWF,aAAaE,UAC5C,IAAI;gBACF,MAAMC,SAAS,MAAMrC,WAAW,MAAM,CAAC;oBACrC,OAAO3C;oBACPyC;oBACA,GAAGoB,YAAY;oBACf,GAAGE,eAAe;gBACpB;gBAEAN,WAAWN,KAAK,GAAG,KAAKD;gBAExBF,kBACE,CAAC,OAAO,EAAEhD,UAAU,QAAQ,EAAEM,eAAe,UAAU,mBAAmB,EAAED,mBAAmB,iBAAiB,EAAE2E,OAAO,KAAK,EAAE,iBAAiB,GAAG,qBAAqB,EAAEA,OAAO,KAAK,EAAE,qBAAqB,GAAG,gBAAgB,EAAEA,OAAO,KAAK,EAAE,gBAAgB,GAAG,WAAW,EAAEvB,SAAS,aAAa,EAAEuB,OAAO,WAAW,IAAI,GAAG,eAAe,EAAE5B,eAAe,IAAI;gBAGxWH,mBACE,CAAC,oBAAoB,EAAEgC,KAAK,SAAS,CAACD,OAAO,KAAK,GAAG;gBAGvD,IAAI,CAACA,OAAO,OAAO,EACjB,MAAM,IAAIxF,MACR,CAAC,mCAAmC,EAAEyF,KAAK,SAAS,CAACD,SAAS;gBAIlE1B,UAAU0B,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;gBAC3C,IAAI,CAAC1B,SACH,MAAM,IAAI9D,MAAM;gBAGlBgE,uBACGwB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,EAAU,qBAAqB;gBAC3DrF,QAAQqF,OAAO,KAAK;gBACpB;YACF,EAAE,OAAOrD,OAAO;gBACdmD,YAAYnD;gBACZ,IAAIoD,UAAUF,aAAa;oBACzB5D,QAAQ,IAAI,CACV,CAAC,mCAAmC,EAAE8D,QAAQ,CAAC,EAAEF,YAAY,eAAe,EAAED,cAAc,aAAa,EAAEE,UAAU,OAAO,EAAE;oBAEhI,MAAM,IAAII,QAAQ,CAACC,UAAYC,WAAWD,SAASP;gBACrD;YACF;YAGF,IAAI,CAACtB,SACH,MAAMwB;QAEV;QAEA/B,UAAU,CAAC,4BAA4B,EAAES,sBAAsB;QAC/DT,UAAU,CAAC,kBAAkB,EAAEO,SAAS;QAGxC,IAAID,eAAe,CAAC1D,OAAO;YAEzB,MAAM6E,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEnB,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;YAEtC3D,QAAQ;gBACN,eAAe6E;gBACf,mBAAmBA;gBACnB,cAAcA,AAAkB,IAAlBA;YAChB;QACF;QAEA,OAAO;YACL,SAASlB,WAAW;YACpB,mBAAmBE,wBAAwBe;YAC3C,OAAOb,eAAe/D;YACtB,YAAY,CAAC,CAAC0D;QAChB;IACF,EAAE,OAAOgC,GAAQ;QACfpE,QAAQ,KAAK,CAAC,kBAAkBoE;QAChC,MAAMC,WAAW,IAAI9F,MACnB,CAAC,eAAe,EAAE6D,cAAc,eAAe,GAAG,kBAAkB,EAAErD,UAAU,GAAG,EAAEqF,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC9J;YACE,OAAOA;QACT;QAEF,MAAMC;IACR;AACF;AAEO,eAAeC,yBACpB9C,QAAsC,EACtC5C,WAAyB,EACzB6C,OAEC;IAOD,MAAM8C,WAAW,MAAMhD,OAAOC,UAAU5C,aAAa;QACnD,WAAW6C,SAAS;IACtB;IACA+C,OAAOD,UAAU;IACjB,MAAMlF,cAAcT,YAAY,WAAW;IAC3C,MAAM6F,cAAcC,cAAcH,SAAS,OAAO,EAAElF;IACpD,IAAI,AAAuB,YAAvB,OAAOoF,aACT,MAAM,IAAInG,qBACR,CAAC,0CAA0C,EAAEM,YAAY,SAAS,CAAC,GAAG,EAAE2F,SAAS,OAAO,EAAE,EAC1FA,SAAS,OAAO,EAChBA,SAAS,KAAK;IAGlB,OAAO;QACL,SAASE;QACT,eAAeF,SAAS,OAAO;QAC/B,OAAOA,SAAS,KAAK;QACrB,mBAAmBA,SAAS,iBAAiB;IAC/C;AACF;AAEO,eAAeI,yBACpBC,IAAY,EACZhG,WAAyB;IAEzB,MAAM,EAAEyD,OAAO,EAAE3D,KAAK,EAAE,GAAG,MAAM6C,OAAOqD,MAAMhG;IAC9C,OAAO;QAAEyD;QAAS3D;IAAM;AAC1B;AAEO,SAASmG,yBAAyBN,QAAgB;IACvD,IAAI;QAEF,MAAMO,YAAYP,SAAS,KAAK,CAAC;QACjC,IAAIO,WACF,OAAOA,SAAS,CAAC,EAAE;QAIrB,MAAMC,iBAAiBR,SAAS,KAAK,CACnC;QAEF,IAAIQ,gBACF,OAAOA,cAAc,CAAC,EAAE;QAI1B,MAAMC,gBAAgBT,SAAS,KAAK,CAAC;QACrC,IAAIS,eACF,OAAOA,aAAa,CAAC,EAAE;IAE3B,EAAE,OAAM,CAAC;IAET,OAAOT;AACT;AAEO,SAASU,yBAAyBC,KAAa;IACpD,IAAIA,MAAM,QAAQ,CAAC,SAEjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;IAG5C,OAAOA;AACT;AAEO,SAASjC,uBAAuB,EACrCkC,SAAS,EACT9F,WAAW,EAIZ;IAKC,MAAM+F,sBAAsBD,AAAc,YAAdA,YAAwB7B,SAAY6B;IAEhE,IAAIC,AAAwB9B,WAAxB8B,qBACF,OAAO;QAAE,QAAQ,CAAC;QAAG,cAAc9B;IAAU;IAG/C,IAAIjE,AAAgB,eAAhBA,aACF,OAAO;QACL,QAAQ;YAAE,iBAAiB+F;QAAoB;QAC/C,cAAc,CAAC,oCAAoC,EAAEA,oBAAoB,aAAa,CAAC;IACzF;IAGF,IAAI/F,AAAgB,oBAAhBA,aACF,OAAO;QACL,QAAQ;YACN,UAAU;gBAAE,MAAM+F,sBAAsB,YAAY;YAAW;QACjE;QACA,cAAc,CAAC,kCAAkC,EAAEA,sBAAsB,YAAY,WAAW,kBAAkB,CAAC;IACrH;IAGF,IAAI/F,AAAgB,YAAhBA,aACF,OAAO;QACL,QAAQ;YACN,UAAU;gBAAE,MAAM+F,sBAAsB,YAAY;YAAW;QACjE;QACA,cAAc,CAAC,kCAAkC,EAAEA,sBAAsB,YAAY,WAAW,UAAU,CAAC;IAC7G;IAGF,IAAI/F,AAAgB,YAAhBA,aACF,OAAO;QACL,QAAQ+F,sBACJ;YACE,WAAW;gBAAE,QAAQ;YAAO;QAC9B,IACA;YACE,WAAW;gBAAE,QAAQ;YAAM;QAC7B;QACJ,cAAcA,sBACV,wDACA;IACN;IAGF,OAAO;QACL,QAAQ,CAAC;QACT,cAAc,CAAC,6CAA6C,EAAE/F,eAAe,UAAU,CAAC,CAAC;QACzF,gBAAgB,CAAC,0DAA0D,EAAEA,eAAe,UAAU,EAAE,CAAC;IAC3G;AACF;AAQA,SAASgG,oBAAoBC,GAAQ;IAEnC,IAAIA,QAAAA,KACF,OAAOA;IAIT,IAAIC,MAAM,OAAO,CAACD,MAChB,OAAOA,IAAI,GAAG,CAAC,CAACE,OAASH,oBAAoBG;IAI/C,IAAI,AAAe,YAAf,OAAOF,KAAkB;QAC3B,MAAMG,aAAkB,CAAC;QAEzB,KAAK,MAAM,CAACC,KAAKC,MAAM,IAAIC,OAAO,OAAO,CAACN,KAAM;YAE9C,MAAMO,aAAaH,IAAI,IAAI;YAG3B,IAAII,kBAAkBT,oBAAoBM;YAG1C,IAAI,AAA2B,YAA3B,OAAOG,iBACTA,kBAAkBA,gBAAgB,IAAI;YAGxCL,UAAU,CAACI,WAAW,GAAGC;QAC3B;QAEA,OAAOL;IACT;IAGA,IAAI,AAAe,YAAf,OAAOH,KACT,OAAOA,IAAI,IAAI;IAIjB,OAAOA;AACT;AAEO,SAASZ,cACdQ,KAAa,EACb7F,WAAqC;IAErC,MAAM0G,kBAAkBlB,yBAAyBK;IAEjD,IAAIa,iBAAiB,MAAM,oBACzB,OAAOA,gBACJ,KAAK,CAAC,oBACL,MAAM,GACP,IAAIzF;IAGT,IAAIT;IACJ,IAAIgE;IACJ,IAAI;QACFhE,SAASmE,KAAK,KAAK,CAAC+B;QACpB,OAAOV,oBAAoBxF;IAC7B,EAAE,OAAOa,OAAO;QACdmD,YAAYnD;IACd;IACA,IAAI;QACFb,SAASmE,KAAK,KAAK,CAACgC,WAAWD;QAC/B,OAAOV,oBAAoBxF;IAC7B,EAAE,OAAOa,OAAO;QACdmD,YAAYnD;IACd;IAEA,IAAIrB,AAAgB,oBAAhBA,eAAmC4G,SAAS5G,cAAc;QAC5D,MAAM6G,aAAajB,yBAAyBc;QAC5C,IAAI;YACFlG,SAASmE,KAAK,KAAK,CAACgC,WAAWE;YAC/B,OAAOb,oBAAoBxF;QAC7B,EAAE,OAAOa,OAAO;YACdmD,YAAYnD;QACd;IACF;IACA,MAAMnC,MACJ,CAAC,gDAAgD,EAAE4H,OACjDtC,aAAa,iBACb,gBAAgB,EAAEqB,OAAO;AAE/B"}

package/dist/es/ai-model/ui-tars-planning.mjs

@@ -3,7 +3,7 @@ import { transformHotkeyInput } from "@midscene/shared/us-keyboard-layout";
 import { assert } from "@midscene/shared/utils";
 import { actionParser } from "@ui-tars/action-parser";
 import { getSummary, getUiTarsPlanningPrompt } from "./prompt/ui-tars-planning.mjs";
-import { callAIWithStringResponse } from "./service-caller/index.mjs";
+import { AIResponseParseError, callAIWithStringResponse } from "./service-caller/index.mjs";
 const debug = getDebug('ui-tars-planning');
 const bboxSize = 10;
 const pointToBbox = (point, width, height)=>[
@@ -37,20 +37,29 @@ async function uiTarsPlanning(userInstruction, options) {
         },
         ...conversationHistory.snapshot()
     ], modelConfig);
-    const convertedText = convertBboxToCoordinates(res.content);
+    let convertedText;
+    let parsed;
+    try {
+        convertedText = convertBboxToCoordinates(res.content);
+        const { size } = context;
+        const parseResult = actionParser({
+            prediction: convertedText,
+            factor: [
+                1000,
+                1000
+            ],
+            screenContext: {
+                width: size.width,
+                height: size.height
+            },
+            modelVer: uiTarsModelVersion
+        });
+        parsed = parseResult.parsed;
+    } catch (parseError) {
+        const errorMessage = parseError instanceof Error ? parseError.message : String(parseError);
+        throw new AIResponseParseError(`Parse error: ${errorMessage}`, JSON.stringify(res.content, void 0, 2), res.usage);
+    }
     const { size } = context;
-    const { parsed } = actionParser({
-        prediction: convertedText,
-        factor: [
-            1000,
-            1000
-        ],
-        screenContext: {
-            width: size.width,
-            height: size.height
-        },
-        modelVer: uiTarsModelVersion
-    });
     debug('ui-tars modelVer', uiTarsModelVersion, ', parsed', JSON.stringify(parsed));
     const transformActions = [];
     const unhandledActions = [];
@@ -191,17 +200,9 @@ async function uiTarsPlanning(userInstruction, options) {
         }
         const errorMessage = [
             'No actions found in UI-TARS response.',
-            ...errorDetails,
-            `\nRaw response: ${res.content}`
+            ...errorDetails
         ].join('\n');
-        throw new Error(errorMessage, {
-            cause: {
-                prediction: res.content,
-                parsed,
-                unhandledActions,
-                convertedText
-            }
-        });
+        throw new AIResponseParseError(errorMessage, JSON.stringify(res.content, void 0, 2), res.usage);
     }
     debug('transformActions', JSON.stringify(transformActions, null, 2));
     const log = getSummary(res.content);