@midscene/core 1.8.11 → 1.9.0

package/dist/es/service/index.mjs

@@ -1,4 +1,4 @@
-import { isAutoGLM } from "../ai-model/auto-glm/util.mjs";
+import { defaultModelFamilyRequiredForLocateMessage } from "../ai-model/errors.mjs";
 import { AiExtractElementInfo, AiLocateElement, AiLocateSection, buildSearchAreaConfig } from "../ai-model/inspect.mjs";
 import { elementDescriberInstruction } from "../ai-model/prompt/describe.mjs";
 import { AIResponseParseError, callAIWithObjectResponse } from "../ai-model/service-caller/index.mjs";
@@ -20,62 +20,27 @@ function _define_property(obj, key, value) {
 }
 const debug = getDebug('ai:service');
 class Service {
-    async locate(query, opt, modelConfig, abortSignal) {
+    async locate(query, opt, modelRuntime, abortSignal) {
+        const { config: modelConfig } = modelRuntime;
         const queryPrompt = 'string' == typeof query ? query : query.prompt;
         assert(queryPrompt, 'query is required for locate');
         assert('object' == typeof query, 'query should be an object for locate');
-        const hasPlanLocatedElement = !!opt?.planLocatedElement?.rect;
-        let searchAreaPrompt;
-        if (query.deepLocate && !hasPlanLocatedElement) searchAreaPrompt = query.prompt;
-        const { modelFamily } = modelConfig;
-        if (searchAreaPrompt && !modelFamily) {
-            console.warn('The "deepLocate" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/model-config');
-            searchAreaPrompt = void 0;
-        }
-        if (searchAreaPrompt && isAutoGLM(modelFamily)) {
-            console.warn('The "deepLocate" feature is not supported with AutoGLM.');
-            searchAreaPrompt = void 0;
-        }
+        if (!modelConfig.modelFamily) throw new Error(defaultModelFamilyRequiredForLocateMessage);
         const context = opt?.context || await this.contextRetrieverFn();
-        let searchArea;
-        let searchAreaRawResponse;
-        let searchAreaUsage;
-        let searchAreaResponse;
-        if (query.deepLocate && hasPlanLocatedElement) {
-            const searchAreaConfig = await buildSearchAreaConfig({
-                context,
-                baseRect: opt.planLocatedElement.rect,
-                modelFamily
-            });
-            searchArea = searchAreaConfig.rect;
-            searchAreaRawResponse = JSON.stringify({
-                source: 'plan-located-element',
-                rect: opt.planLocatedElement.rect
-            });
-            searchAreaResponse = {
-                rect: searchArea,
-                imageBase64: searchAreaConfig.imageBase64,
-                scale: searchAreaConfig.scale,
-                rawResponse: searchAreaRawResponse
-            };
-        } else if (searchAreaPrompt) {
-            searchAreaResponse = await AiLocateSection({
-                context,
-                sectionDescription: searchAreaPrompt,
-                modelConfig,
-                abortSignal
-            });
-            assert(searchAreaResponse.rect, `cannot find search area for "${searchAreaPrompt}"${searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''}`);
-            searchAreaRawResponse = searchAreaResponse.rawResponse;
-            searchAreaUsage = searchAreaResponse.usage;
-            searchArea = searchAreaResponse.rect;
-        }
+        const searchArea = await this.resolveLocateSearchArea({
+            query,
+            queryPrompt,
+            opt,
+            context,
+            modelRuntime,
+            abortSignal
+        });
         const startTime = Date.now();
         const { parseResult, rect, rawResponse, usage, reasoning_content } = await AiLocateElement({
             context,
             targetElementDescription: queryPrompt,
-            searchConfig: searchAreaResponse,
-            modelConfig,
+            searchConfig: searchArea.config,
+            modelRuntime,
             abortSignal
         });
         const timeCost = Date.now() - startTime;
@@ -85,9 +50,9 @@ class Service {
             rawResponse: JSON.stringify(rawResponse),
             formatResponse: JSON.stringify(parseResult),
             usage,
-            searchArea,
-            searchAreaRawResponse,
-            searchAreaUsage,
+            searchArea: searchArea.trace.sourceRect,
+            searchAreaRawResponse: searchArea.trace.rawResponse,
+            searchAreaUsage: searchArea.trace.usage,
             reasoning_content
         };
         let errorLog;
@@ -97,25 +62,25 @@ class Service {
             userQuery: {
                 element: queryPrompt
             },
-            matchedElement: [],
             matchedRect: rect,
             data: null,
             taskInfo,
-            deepLocate: !!searchArea,
+            deepLocate: !!searchArea.trace.sourceRect,
             error: errorLog
         };
-        const elements = parseResult.elements || [];
+        const element = parseResult.element;
         const dump = createServiceDump({
             ...dumpData,
-            matchedElement: elements
+            matchedElement: element ? [
+                element
+            ] : []
         });
         if (errorLog) throw new ServiceError(errorLog, dump);
-        if (elements.length > 1) throw new ServiceError(`locate: multiple elements found, length = ${elements.length}`, dump);
-        if (1 === elements.length) return {
+        if (element) return {
             element: {
-                center: elements[0].center,
-                rect: elements[0].rect,
-                description: elements[0].description
+                center: element.center,
+                rect: element.rect,
+                description: element.description
             },
             rect,
             dump
@@ -126,7 +91,72 @@ class Service {
             dump
         };
     }
-    async extract(dataDemand, modelConfig, opt, pageDescription, multimodalPrompt, context) {
+    async resolveLocateSearchArea(options) {
+        const { query, queryPrompt, opt, context, modelRuntime, abortSignal } = options;
+        const { adapter } = modelRuntime;
+        const hasPlanLocatedElement = !!opt?.planLocatedElement?.rect;
+        if (!query.deepLocate) return {
+            trace: {}
+        };
+        if (hasPlanLocatedElement) {
+            const config = await buildSearchAreaConfig({
+                context,
+                baseRect: opt.planLocatedElement.rect
+            });
+            return {
+                config,
+                trace: {
+                    sourceRect: config.sourceRect,
+                    rawResponse: JSON.stringify({
+                        source: 'plan-located-element',
+                        rect: opt.planLocatedElement.rect
+                    })
+                }
+            };
+        }
+        if (adapter.locate.supportsSearchArea) {
+            const searchAreaResponse = await AiLocateSection({
+                context,
+                sectionDescription: queryPrompt,
+                modelRuntime,
+                abortSignal
+            });
+            const { searchAreaConfig } = searchAreaResponse;
+            assert(searchAreaConfig, `cannot find search area for "${queryPrompt}"${searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''}`);
+            return {
+                config: searchAreaConfig,
+                trace: {
+                    sourceRect: searchAreaConfig.sourceRect,
+                    rawResponse: searchAreaResponse.rawResponse,
+                    usage: searchAreaResponse.usage
+                }
+            };
+        }
+        const firstPassLocateResult = await AiLocateElement({
+            context,
+            targetElementDescription: queryPrompt,
+            modelRuntime,
+            abortSignal
+        });
+        assert(firstPassLocateResult.rect, `cannot find search area for "${queryPrompt}"${firstPassLocateResult.parseResult.errors?.length ? `: ${firstPassLocateResult.parseResult.errors.join('\n')}` : ''}`);
+        const config = await buildSearchAreaConfig({
+            context,
+            baseRect: firstPassLocateResult.rect
+        });
+        return {
+            config,
+            trace: {
+                sourceRect: config.sourceRect,
+                rawResponse: JSON.stringify({
+                    source: 'deep-locate-first-pass',
+                    rect: firstPassLocateResult.rect,
+                    rawResponse: firstPassLocateResult.rawResponse
+                }),
+                usage: firstPassLocateResult.usage
+            }
+        };
+    }
+    async extract(dataDemand, modelRuntime, opt, pageDescription, multimodalPrompt, context) {
         assert(context, 'context is required for extract');
         assert('object' == typeof dataDemand || 'string' == typeof dataDemand, `dataDemand should be object or string, but get ${typeof dataDemand}`);
         const startTime = Date.now();
@@ -140,7 +170,7 @@ class Service {
                 dataQuery: dataDemand,
                 multimodalPrompt,
                 extractOption: opt,
-                modelConfig,
+                modelRuntime,
                 pageDescription
             });
             parseResult = result.parseResult;
@@ -161,7 +191,6 @@ class Service {
                     userQuery: {
                         dataDemand
                     },
-                    matchedElement: [],
                     data: null,
                     taskInfo,
                     error: error.message
@@ -186,7 +215,6 @@ class Service {
             userQuery: {
                 dataDemand
             },
-            matchedElement: [],
             data: null,
             taskInfo,
             error: errorLog
@@ -205,13 +233,12 @@ class Service {
             dump
         };
     }
-    async describe(target, modelConfig, opt) {
+    async describe(target, modelRuntime, opt) {
         assert(target, 'target is required for service.describe');
         const context = await this.contextRetrieverFn();
         const { shotSize } = context;
         const screenshotBase64 = context.screenshot.base64;
         assert(screenshotBase64, 'screenshot is required for service.describe');
-        const { modelFamily } = modelConfig;
         const systemPrompt = elementDescriberInstruction();
         const defaultRectSize = 30;
         const targetRect = Array.isArray(target) ? {
@@ -233,7 +260,7 @@ class Service {
         if (opt?.deepLocate) {
             const searchArea = expandSearchArea(targetRect, shotSize);
             debug('describe: cropping to searchArea', searchArea);
-            const croppedResult = await cropByRect(imagePayload, searchArea, 'qwen2.5-vl' === modelFamily);
+            const croppedResult = await cropByRect(imagePayload, searchArea);
             imagePayload = croppedResult.imageBase64;
         }
         const msgs = [
@@ -254,7 +281,7 @@ class Service {
                 ]
             }
         ];
-        const res = await callAIWithObjectResponse(msgs, modelConfig);
+        const res = await callAIWithObjectResponse(msgs, modelRuntime);
         const { content } = res;
         assert(!content.error, `describe failed: ${content.error}`);
         assert(content.description, 'failed to describe the element');

package/dist/es/service/index.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"service/index.mjs","sources":["../../../src/service/index.ts"],"sourcesContent":["import { isAutoGLM, isUITars } from '@/ai-model/auto-glm/util';\nimport {\n  AiExtractElementInfo,\n  AiLocateElement,\n  AiLocateSection,\n  buildSearchAreaConfig,\n} from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport {\n  AIResponseParseError,\n  callAIWithObjectResponse,\n} from '@/ai-model/service-caller';\nimport { type AIArgs, expandSearchArea } from '@/common';\nimport type {\n  AIDescribeElementResponse,\n  AIUsageInfo,\n  DetailedLocateParam,\n  LocateResultElement,\n  LocateResultWithDump,\n  PartialServiceDumpFromSDK,\n  PlanningLocateParam,\n  Rect,\n  ServiceExtractOption,\n  ServiceExtractParam,\n  ServiceExtractResult,\n  ServiceTaskInfo,\n  UIContext,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TMultimodalPrompt } from '../common';\nimport { createServiceDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext;\n  planLocatedElement?: LocateResultElement;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\ninterface ServiceOptions {\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n}\n\nconst debug = getDebug('ai:service');\nexport default class Service {\n  contextRetrieverFn: () => Promise<UIContext> | UIContext;\n\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n\n  constructor(\n    context: UIContext | (() => Promise<UIContext> | UIContext),\n    opt?: ServiceOptions,\n  ) {\n    assert(context, 'context is required for Service');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    if (typeof opt?.taskInfo !== 'undefined') {\n      this.taskInfo = opt.taskInfo;\n    }\n  }\n\n  async locate(\n    query: PlanningLocateParam,\n    opt: LocateOpts,\n    modelConfig: IModelConfig,\n    abortSignal?: AbortSignal,\n  ): Promise<LocateResultWithDump> {\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    const hasPlanLocatedElement = !!opt?.planLocatedElement?.rect;\n\n    let searchAreaPrompt;\n    if (query.deepLocate && !hasPlanLocatedElement) {\n      searchAreaPrompt = query.prompt;\n    }\n\n    const { modelFamily } = modelConfig;\n\n    if (searchAreaPrompt && !modelFamily) {\n      console.warn(\n        'The \"deepLocate\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/model-config',\n      );\n      searchAreaPrompt = undefined;\n    }\n\n    if (searchAreaPrompt && isAutoGLM(modelFamily)) {\n      console.warn('The \"deepLocate\" feature is not supported with AutoGLM.');\n      searchAreaPrompt = undefined;\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn());\n\n    let searchArea: Rect | undefined = undefined;\n    let searchAreaRawResponse: string | undefined = undefined;\n    let searchAreaUsage: AIUsageInfo | undefined = undefined;\n    let searchAreaResponse:\n      | Awaited<ReturnType<typeof AiLocateSection>>\n      | undefined = undefined;\n    if (query.deepLocate && hasPlanLocatedElement) {\n      const searchAreaConfig = await buildSearchAreaConfig({\n        context,\n        baseRect: opt.planLocatedElement!.rect,\n        modelFamily,\n      });\n      searchArea = searchAreaConfig.rect;\n\n      searchAreaRawResponse = JSON.stringify({\n        source: 'plan-located-element',\n        rect: opt.planLocatedElement!.rect,\n      });\n      searchAreaResponse = {\n        rect: searchArea,\n        imageBase64: searchAreaConfig.imageBase64,\n        scale: searchAreaConfig.scale,\n        rawResponse: searchAreaRawResponse,\n      };\n    } else if (searchAreaPrompt) {\n      searchAreaResponse = await AiLocateSection({\n        context,\n        sectionDescription: searchAreaPrompt,\n        modelConfig,\n        abortSignal,\n      });\n      assert(\n        searchAreaResponse.rect,\n        `cannot find search area for \"${searchAreaPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const { parseResult, rect, rawResponse, usage, reasoning_content } =\n      await AiLocateElement({\n        context,\n        targetElementDescription: queryPrompt,\n        searchConfig: searchAreaResponse,\n        modelConfig,\n        abortSignal,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `failed to locate element: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedElement: [],\n      matchedRect: rect,\n      data: null,\n      taskInfo,\n      deepLocate: !!searchArea,\n      error: errorLog,\n    };\n\n    const elements = parseResult.elements || [];\n\n    const dump = createServiceDump({\n      ...dumpData,\n      matchedElement: elements,\n    });\n\n    if (errorLog) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    if (elements.length > 1) {\n      throw new ServiceError(\n        `locate: multiple elements found, length = ${elements.length}`,\n        dump,\n      );\n    }\n\n    if (elements.length === 1) {\n      return {\n        element: {\n          center: elements[0]!.center,\n          rect: elements[0]!.rect,\n          description: elements[0]!.description,\n        },\n        rect,\n        dump,\n      };\n    }\n\n    return {\n      element: null,\n      rect,\n      dump,\n    };\n  }\n\n  async extract<T>(\n    dataDemand: ServiceExtractParam,\n    modelConfig: IModelConfig,\n    opt?: ServiceExtractOption,\n    pageDescription?: string,\n    multimodalPrompt?: TMultimodalPrompt,\n    context?: UIContext,\n  ): Promise<ServiceExtractResult<T>> {\n    assert(context, 'context is required for extract');\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n\n    const startTime = Date.now();\n\n    let parseResult: Awaited<\n      ReturnType<typeof AiExtractElementInfo<T>>\n    >['parseResult'];\n    let rawResponse: string;\n    let usage: Awaited<ReturnType<typeof AiExtractElementInfo<T>>>['usage'];\n    let reasoning_content: string | undefined;\n\n    try {\n      const result = await AiExtractElementInfo<T>({\n        context,\n        dataQuery: dataDemand,\n        multimodalPrompt,\n        extractOption: opt,\n        modelConfig,\n        pageDescription,\n      });\n      parseResult = result.parseResult;\n      rawResponse = result.rawResponse;\n      usage = result.usage;\n      reasoning_content = result.reasoning_content;\n    } catch (error) {\n      if (error instanceof AIResponseParseError) {\n        // Create dump with usage and rawResponse from the error\n        const timeCost = Date.now() - startTime;\n        const taskInfo: ServiceTaskInfo = {\n          ...(this.taskInfo ? this.taskInfo : {}),\n          durationMs: timeCost,\n          rawResponse: error.rawResponse,\n          usage: error.usage,\n        };\n        const dump = createServiceDump({\n          type: 'extract',\n          userQuery: { dataDemand },\n          matchedElement: [],\n          data: null,\n          taskInfo,\n          error: error.message,\n        });\n        throw new ServiceError(error.message, dump);\n      }\n      throw error;\n    }\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse,\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data, thought } = parseResult || {};\n\n    const dump = createServiceDump({\n      ...dumpData,\n      data,\n    });\n\n    if (errorLog && !data) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    return {\n      data,\n      thought,\n      usage,\n      reasoning_content,\n      dump,\n    };\n  }\n\n  async describe(\n    target: Rect | [number, number],\n    modelConfig: IModelConfig,\n    opt?: {\n      deepLocate?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for service.describe');\n    const context = await this.contextRetrieverFn();\n    const { shotSize } = context;\n    const screenshotBase64 = context.screenshot.base64;\n    assert(screenshotBase64, 'screenshot is required for service.describe');\n    // The result of the \"describe\" function will be used for positioning, so essentially it is a form of grounding.\n    const { modelFamily } = modelConfig;\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      size: shotSize,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepLocate) {\n      const searchArea = expandSearchArea(targetRect, shotSize);\n      // Always crop in describe mode. Unlike locate's deepLocate (where\n      // cropping too small loses context for finding elements), describe's\n      // deepLocate intentionally zooms in so the model produces a more\n      // precise description from a focused view. expandSearchArea already\n      // guarantees a minimum 400x400 area with surrounding context.\n      debug('describe: cropping to searchArea', searchArea);\n      const croppedResult = await cropByRect(\n        imagePayload,\n        searchArea,\n        modelFamily === 'qwen2.5-vl',\n      );\n      imagePayload = croppedResult.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      },\n    ];\n\n    const res = await callAIWithObjectResponse<AIDescribeElementResponse>(\n      msgs,\n      modelConfig,\n    );\n\n    const { content } = res;\n    assert(!content.error, `describe failed: ${content.error}`);\n    assert(content.description, 'failed to describe the element');\n    return content;\n  }\n}\n"],"names":["debug","getDebug","Service","query","opt","modelConfig","abortSignal","queryPrompt","assert","hasPlanLocatedElement","searchAreaPrompt","modelFamily","console","undefined","isAutoGLM","context","searchArea","searchAreaRawResponse","searchAreaUsage","searchAreaResponse","searchAreaConfig","buildSearchAreaConfig","JSON","AiLocateSection","startTime","Date","parseResult","rect","rawResponse","usage","reasoning_content","AiLocateElement","timeCost","taskInfo","errorLog","dumpData","elements","dump","createServiceDump","ServiceError","dataDemand","pageDescription","multimodalPrompt","result","AiExtractElementInfo","error","AIResponseParseError","data","thought","target","shotSize","screenshotBase64","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","croppedResult","cropByRect","msgs","res","callAIWithObjectResponse","content","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;;AAiDA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IAqBnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,WAAyB,EACzBC,WAAyB,EACM;QAC/B,MAAMC,cAAc,AAAiB,YAAjB,OAAOJ,QAAqBA,QAAQA,MAAM,MAAM;QACpEK,OAAOD,aAAa;QAEpBC,OAAO,AAAiB,YAAjB,OAAOL,OAAoB;QAElC,MAAMM,wBAAwB,CAAC,CAACL,KAAK,oBAAoB;QAEzD,IAAIM;QACJ,IAAIP,MAAM,UAAU,IAAI,CAACM,uBACvBC,mBAAmBP,MAAM,MAAM;QAGjC,MAAM,EAAEQ,WAAW,EAAE,GAAGN;QAExB,IAAIK,oBAAoB,CAACC,aAAa;YACpCC,QAAQ,IAAI,CACV;YAEFF,mBAAmBG;QACrB;QAEA,IAAIH,oBAAoBI,UAAUH,cAAc;YAC9CC,QAAQ,IAAI,CAAC;YACbF,mBAAmBG;QACrB;QAEA,MAAME,UAAUX,KAAK,WAAY,MAAM,IAAI,CAAC,kBAAkB;QAE9D,IAAIY;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAGJ,IAAIhB,MAAM,UAAU,IAAIM,uBAAuB;YAC7C,MAAMW,mBAAmB,MAAMC,sBAAsB;gBACnDN;gBACA,UAAUX,IAAI,kBAAkB,CAAE,IAAI;gBACtCO;YACF;YACAK,aAAaI,iBAAiB,IAAI;YAElCH,wBAAwBK,KAAK,SAAS,CAAC;gBACrC,QAAQ;gBACR,MAAMlB,IAAI,kBAAkB,CAAE,IAAI;YACpC;YACAe,qBAAqB;gBACnB,MAAMH;gBACN,aAAaI,iBAAiB,WAAW;gBACzC,OAAOA,iBAAiB,KAAK;gBAC7B,aAAaH;YACf;QACF,OAAO,IAAIP,kBAAkB;YAC3BS,qBAAqB,MAAMI,gBAAgB;gBACzCR;gBACA,oBAAoBL;gBACpBL;gBACAC;YACF;YACAE,OACEW,mBAAmB,IAAI,EACvB,CAAC,6BAA6B,EAAET,iBAAiB,CAAC,EAChDS,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAEJF,wBAAwBE,mBAAmB,WAAW;YACtDD,kBAAkBC,mBAAmB,KAAK;YAC1CH,aAAaG,mBAAmB,IAAI;QACtC;QAEA,MAAMK,YAAYC,KAAK,GAAG;QAC1B,MAAM,EAAEC,WAAW,EAAEC,IAAI,EAAEC,WAAW,EAAEC,KAAK,EAAEC,iBAAiB,EAAE,GAChE,MAAMC,gBAAgB;YACpBhB;YACA,0BAA0BR;YAC1B,cAAcY;YACdd;YACAC;QACF;QAEF,MAAM0B,WAAWP,KAAK,GAAG,KAAKD;QAC9B,MAAMS,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaV,KAAK,SAAS,CAACM;YAC5B,gBAAgBN,KAAK,SAAS,CAACI;YAC/BG;YACAb;YACAC;YACAC;YACAY;QACF;QAEA,IAAII;QACJ,IAAIR,YAAY,MAAM,EAAE,QACtBQ,WAAW,CAAC,4BAA4B,EAAER,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG3E,MAAMS,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAAS5B;YACX;YACA,gBAAgB,EAAE;YAClB,aAAaoB;YACb,MAAM;YACNM;YACA,YAAY,CAAC,CAACjB;YACd,OAAOkB;QACT;QAEA,MAAME,WAAWV,YAAY,QAAQ,IAAI,EAAE;QAE3C,MAAMW,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACX,gBAAgBC;QAClB;QAEA,IAAIF,UACF,MAAM,IAAIK,aAAaL,UAAUG;QAGnC,IAAID,SAAS,MAAM,GAAG,GACpB,MAAM,IAAIG,aACR,CAAC,0CAA0C,EAAEH,SAAS,MAAM,EAAE,EAC9DC;QAIJ,IAAID,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,SAAS;gBACP,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM;gBAC3B,MAAMA,QAAQ,CAAC,EAAE,CAAE,IAAI;gBACvB,aAAaA,QAAQ,CAAC,EAAE,CAAE,WAAW;YACvC;YACAT;YACAU;QACF;QAGF,OAAO;YACL,SAAS;YACTV;YACAU;QACF;IACF;IAEA,MAAM,QACJG,UAA+B,EAC/BnC,WAAyB,EACzBD,GAA0B,EAC1BqC,eAAwB,EACxBC,gBAAoC,EACpC3B,OAAmB,EACe;QAClCP,OAAOO,SAAS;QAChBP,OACE,AAAsB,YAAtB,OAAOgC,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAGvE,MAAMhB,YAAYC,KAAK,GAAG;QAE1B,IAAIC;QAGJ,IAAIE;QACJ,IAAIC;QACJ,IAAIC;QAEJ,IAAI;YACF,MAAMa,SAAS,MAAMC,qBAAwB;gBAC3C7B;gBACA,WAAWyB;gBACXE;gBACA,eAAetC;gBACfC;gBACAoC;YACF;YACAf,cAAciB,OAAO,WAAW;YAChCf,cAAce,OAAO,WAAW;YAChCd,QAAQc,OAAO,KAAK;YACpBb,oBAAoBa,OAAO,iBAAiB;QAC9C,EAAE,OAAOE,OAAO;YACd,IAAIA,iBAAiBC,sBAAsB;gBAEzC,MAAMd,WAAWP,KAAK,GAAG,KAAKD;gBAC9B,MAAMS,WAA4B;oBAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;oBACtC,YAAYD;oBACZ,aAAaa,MAAM,WAAW;oBAC9B,OAAOA,MAAM,KAAK;gBACpB;gBACA,MAAMR,OAAOC,kBAAkB;oBAC7B,MAAM;oBACN,WAAW;wBAAEE;oBAAW;oBACxB,gBAAgB,EAAE;oBAClB,MAAM;oBACNP;oBACA,OAAOY,MAAM,OAAO;gBACtB;gBACA,MAAM,IAAIN,aAAaM,MAAM,OAAO,EAAER;YACxC;YACA,MAAMQ;QACR;QAEA,MAAMb,WAAWP,KAAK,GAAG,KAAKD;QAC9B,MAAMS,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZJ;YACA,gBAAgBN,KAAK,SAAS,CAACI;YAC/BG;YACAC;QACF;QAEA,IAAII;QACJ,IAAIR,YAAY,MAAM,EAAE,QACtBQ,WAAW,CAAC,qBAAqB,EAAER,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMS,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTK;YACF;YACA,gBAAgB,EAAE;YAClB,MAAM;YACNP;YACA,OAAOC;QACT;QAEA,MAAM,EAAEa,IAAI,EAAEC,OAAO,EAAE,GAAGtB,eAAe,CAAC;QAE1C,MAAMW,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACXY;QACF;QAEA,IAAIb,YAAY,CAACa,MACf,MAAM,IAAIR,aAAaL,UAAUG;QAGnC,OAAO;YACLU;YACAC;YACAnB;YACAC;YACAO;QACF;IACF;IAEA,MAAM,SACJY,MAA+B,EAC/B5C,WAAyB,EACzBD,GAEC,EACwD;QACzDI,OAAOyC,QAAQ;QACf,MAAMlC,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAC7C,MAAM,EAAEmC,QAAQ,EAAE,GAAGnC;QACrB,MAAMoC,mBAAmBpC,QAAQ,UAAU,CAAC,MAAM;QAClDP,OAAO2C,kBAAkB;QAEzB,MAAM,EAAExC,WAAW,EAAE,GAAGN;QACxB,MAAM+C,eAAeC;QAGrB,MAAMC,kBAAkB;QACxB,MAAMC,aAAmBC,MAAM,OAAO,CAACP,UACnC;YACE,MAAMQ,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC/C,KAAKG,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC9C,OAAOA;YACP,QAAQA;QACV,IACAL;QAEJ,IAAIS,eAAe,MAAMC,wBAAwB;YAC/C,gBAAgBR;YAChB,MAAMD;YACN,sBAAsB;gBACpB;oBACE,MAAMK;gBACR;aACD;YACD,iBAAiB;QACnB;QAEA,IAAInD,KAAK,YAAY;YACnB,MAAMY,aAAa4C,iBAAiBL,YAAYL;YAMhDlD,MAAM,oCAAoCgB;YAC1C,MAAM6C,gBAAgB,MAAMC,WAC1BJ,cACA1C,YACAL,AAAgB,iBAAhBA;YAEF+C,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,MAAM,MAAMC,yBAChBF,MACA1D;QAGF,MAAM,EAAE6D,OAAO,EAAE,GAAGF;QACpBxD,OAAO,CAAC0D,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1D1D,OAAO0D,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IAlWA,YACEnD,OAA2D,EAC3DX,GAAoB,CACpB;QAPF;QAEA;QAMEI,OAAOO,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMoD,QAAQ,OAAO,CAACpD;QAGlD,IAAI,AAAyB,WAAlBX,KAAK,UACd,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAqVF"}
+{"version":3,"file":"service/index.mjs","sources":["../../../src/service/index.ts"],"sourcesContent":["import { defaultModelFamilyRequiredForLocateMessage } from '@/ai-model/errors';\nimport {\n  AiExtractElementInfo,\n  AiLocateElement,\n  AiLocateSection,\n  buildSearchAreaConfig,\n} from '@/ai-model/inspect';\nimport type { ModelRuntime } from '@/ai-model/models';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport {\n  AIResponseParseError,\n  callAIWithObjectResponse,\n} from '@/ai-model/service-caller';\nimport type { AIArgs } from '@/ai-model/types';\nimport type { SearchAreaConfig } from '@/ai-model/workflows/inspect/types';\nimport { expandSearchArea } from '@/common';\nimport type {\n  AIDescribeElementResponse,\n  AIUsageInfo,\n  DetailedLocateParam,\n  LocateResultElement,\n  LocateResultWithDump,\n  PartialServiceDumpFromSDK,\n  PlanningLocateParam,\n  Rect,\n  ServiceExtractOption,\n  ServiceExtractParam,\n  ServiceExtractResult,\n  ServiceTaskInfo,\n  UIContext,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TMultimodalPrompt, TUserPrompt } from '../common';\nimport { createServiceDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext;\n  planLocatedElement?: LocateResultElement;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\ninterface ServiceOptions {\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n}\n\ninterface LocateSearchAreaResult {\n  config?: SearchAreaConfig;\n  trace: {\n    sourceRect?: Rect;\n    rawResponse?: string;\n    usage?: AIUsageInfo;\n  };\n}\n\nconst debug = getDebug('ai:service');\nexport default class Service {\n  contextRetrieverFn: () => Promise<UIContext> | UIContext;\n\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n\n  constructor(\n    context: UIContext | (() => Promise<UIContext> | UIContext),\n    opt?: ServiceOptions,\n  ) {\n    assert(context, 'context is required for Service');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    if (typeof opt?.taskInfo !== 'undefined') {\n      this.taskInfo = opt.taskInfo;\n    }\n  }\n\n  async locate(\n    query: PlanningLocateParam,\n    opt: LocateOpts,\n    modelRuntime: ModelRuntime,\n    abortSignal?: AbortSignal,\n  ): Promise<LocateResultWithDump> {\n    const { config: modelConfig } = modelRuntime;\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    if (!modelConfig.modelFamily) {\n      throw new Error(defaultModelFamilyRequiredForLocateMessage);\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn());\n\n    const searchArea = await this.resolveLocateSearchArea({\n      query,\n      queryPrompt,\n      opt,\n      context,\n      modelRuntime,\n      abortSignal,\n    });\n\n    const startTime = Date.now();\n    const { parseResult, rect, rawResponse, usage, reasoning_content } =\n      await AiLocateElement({\n        context,\n        targetElementDescription: queryPrompt,\n        searchConfig: searchArea.config,\n        modelRuntime,\n        abortSignal,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea: searchArea.trace.sourceRect,\n      searchAreaRawResponse: searchArea.trace.rawResponse,\n      searchAreaUsage: searchArea.trace.usage,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `failed to locate element: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedRect: rect,\n      data: null,\n      taskInfo,\n      deepLocate: !!searchArea.trace.sourceRect,\n      error: errorLog,\n    };\n\n    const element = parseResult.element;\n\n    const dump = createServiceDump({\n      ...dumpData,\n      matchedElement: element ? [element] : [],\n    });\n\n    if (errorLog) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    if (element) {\n      return {\n        element: {\n          center: element.center,\n          rect: element.rect,\n          description: element.description,\n        },\n        rect,\n        dump,\n      };\n    }\n\n    return {\n      element: null,\n      rect,\n      dump,\n    };\n  }\n\n  private async resolveLocateSearchArea(options: {\n    query: PlanningLocateParam;\n    queryPrompt: TUserPrompt;\n    opt: LocateOpts;\n    context: UIContext;\n    modelRuntime: ModelRuntime;\n    abortSignal?: AbortSignal;\n  }): Promise<LocateSearchAreaResult> {\n    const { query, queryPrompt, opt, context, modelRuntime, abortSignal } =\n      options;\n    const { adapter } = modelRuntime;\n    const hasPlanLocatedElement = !!opt?.planLocatedElement?.rect;\n\n    if (!query.deepLocate) {\n      return { trace: {} };\n    }\n\n    if (hasPlanLocatedElement) {\n      const config = await buildSearchAreaConfig({\n        context,\n        baseRect: opt.planLocatedElement!.rect,\n      });\n\n      return {\n        config,\n        trace: {\n          sourceRect: config.sourceRect,\n          rawResponse: JSON.stringify({\n            source: 'plan-located-element',\n            rect: opt.planLocatedElement!.rect,\n          }),\n        },\n      };\n    }\n\n    if (adapter.locate.supportsSearchArea) {\n      const searchAreaResponse = await AiLocateSection({\n        context,\n        sectionDescription: queryPrompt,\n        modelRuntime,\n        abortSignal,\n      });\n      const { searchAreaConfig } = searchAreaResponse;\n      assert(\n        searchAreaConfig,\n        `cannot find search area for \"${queryPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n\n      return {\n        config: searchAreaConfig,\n        trace: {\n          sourceRect: searchAreaConfig.sourceRect,\n          rawResponse: searchAreaResponse.rawResponse,\n          usage: searchAreaResponse.usage,\n        },\n      };\n    }\n\n    const firstPassLocateResult = await AiLocateElement({\n      context,\n      targetElementDescription: queryPrompt,\n      modelRuntime,\n      abortSignal,\n    });\n    assert(\n      firstPassLocateResult.rect,\n      `cannot find search area for \"${queryPrompt}\"${\n        firstPassLocateResult.parseResult.errors?.length\n          ? `: ${firstPassLocateResult.parseResult.errors.join('\\n')}`\n          : ''\n      }`,\n    );\n\n    const config = await buildSearchAreaConfig({\n      context,\n      baseRect: firstPassLocateResult.rect,\n    });\n\n    return {\n      config,\n      trace: {\n        sourceRect: config.sourceRect,\n        rawResponse: JSON.stringify({\n          source: 'deep-locate-first-pass',\n          rect: firstPassLocateResult.rect,\n          rawResponse: firstPassLocateResult.rawResponse,\n        }),\n        usage: firstPassLocateResult.usage,\n      },\n    };\n  }\n\n  async extract<T>(\n    dataDemand: ServiceExtractParam,\n    modelRuntime: ModelRuntime,\n    opt?: ServiceExtractOption,\n    pageDescription?: string,\n    multimodalPrompt?: TMultimodalPrompt,\n    context?: UIContext,\n  ): Promise<ServiceExtractResult<T>> {\n    assert(context, 'context is required for extract');\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n\n    const startTime = Date.now();\n\n    let parseResult: Awaited<\n      ReturnType<typeof AiExtractElementInfo<T>>\n    >['parseResult'];\n    let rawResponse: string;\n    let usage: Awaited<ReturnType<typeof AiExtractElementInfo<T>>>['usage'];\n    let reasoning_content: string | undefined;\n\n    try {\n      const result = await AiExtractElementInfo<T>({\n        context,\n        dataQuery: dataDemand,\n        multimodalPrompt,\n        extractOption: opt,\n        modelRuntime,\n        pageDescription,\n      });\n      parseResult = result.parseResult;\n      rawResponse = result.rawResponse;\n      usage = result.usage;\n      reasoning_content = result.reasoning_content;\n    } catch (error) {\n      if (error instanceof AIResponseParseError) {\n        // Create dump with usage and rawResponse from the error\n        const timeCost = Date.now() - startTime;\n        const taskInfo: ServiceTaskInfo = {\n          ...(this.taskInfo ? this.taskInfo : {}),\n          durationMs: timeCost,\n          rawResponse: error.rawResponse,\n          usage: error.usage,\n        };\n        const dump = createServiceDump({\n          type: 'extract',\n          userQuery: { dataDemand },\n          data: null,\n          taskInfo,\n          error: error.message,\n        });\n        throw new ServiceError(error.message, dump);\n      }\n      throw error;\n    }\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse,\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data, thought } = parseResult || {};\n\n    const dump = createServiceDump({\n      ...dumpData,\n      data,\n    });\n\n    if (errorLog && !data) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    return {\n      data,\n      thought,\n      usage,\n      reasoning_content,\n      dump,\n    };\n  }\n\n  async describe(\n    target: Rect | [number, number],\n    modelRuntime: ModelRuntime,\n    opt?: {\n      deepLocate?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for service.describe');\n    const context = await this.contextRetrieverFn();\n    const { shotSize } = context;\n    const screenshotBase64 = context.screenshot.base64;\n    assert(screenshotBase64, 'screenshot is required for service.describe');\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      size: shotSize,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepLocate) {\n      const searchArea = expandSearchArea(targetRect, shotSize);\n      // Always crop in describe mode. Unlike locate's deepLocate (where\n      // cropping too small loses context for finding elements), describe's\n      // deepLocate intentionally zooms in so the model produces a more\n      // precise description from a focused view. expandSearchArea already\n      // guarantees a minimum 400x400 area with surrounding context.\n      // Describe is not a coordinate-parsing flow, so it does not need image\n      // padding for bbox normalization.\n      debug('describe: cropping to searchArea', searchArea);\n      const croppedResult = await cropByRect(imagePayload, searchArea);\n      imagePayload = croppedResult.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      },\n    ];\n\n    const res = await callAIWithObjectResponse<AIDescribeElementResponse>(\n      msgs,\n      modelRuntime,\n    );\n\n    const { content } = res;\n    assert(!content.error, `describe failed: ${content.error}`);\n    assert(content.description, 'failed to describe the element');\n    return content;\n  }\n}\n"],"names":["debug","getDebug","Service","query","opt","modelRuntime","abortSignal","modelConfig","queryPrompt","assert","Error","defaultModelFamilyRequiredForLocateMessage","context","searchArea","startTime","Date","parseResult","rect","rawResponse","usage","reasoning_content","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","element","dump","createServiceDump","ServiceError","options","adapter","hasPlanLocatedElement","config","buildSearchAreaConfig","searchAreaResponse","AiLocateSection","searchAreaConfig","firstPassLocateResult","dataDemand","pageDescription","multimodalPrompt","result","AiExtractElementInfo","error","AIResponseParseError","data","thought","target","shotSize","screenshotBase64","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","croppedResult","cropByRect","msgs","res","callAIWithObjectResponse","content","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;;AA4DA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IAqBnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,YAA0B,EAC1BC,WAAyB,EACM;QAC/B,MAAM,EAAE,QAAQC,WAAW,EAAE,GAAGF;QAChC,MAAMG,cAAc,AAAiB,YAAjB,OAAOL,QAAqBA,QAAQA,MAAM,MAAM;QACpEM,OAAOD,aAAa;QAEpBC,OAAO,AAAiB,YAAjB,OAAON,OAAoB;QAElC,IAAI,CAACI,YAAY,WAAW,EAC1B,MAAM,IAAIG,MAAMC;QAGlB,MAAMC,UAAUR,KAAK,WAAY,MAAM,IAAI,CAAC,kBAAkB;QAE9D,MAAMS,aAAa,MAAM,IAAI,CAAC,uBAAuB,CAAC;YACpDV;YACAK;YACAJ;YACAQ;YACAP;YACAC;QACF;QAEA,MAAMQ,YAAYC,KAAK,GAAG;QAC1B,MAAM,EAAEC,WAAW,EAAEC,IAAI,EAAEC,WAAW,EAAEC,KAAK,EAAEC,iBAAiB,EAAE,GAChE,MAAMC,gBAAgB;YACpBT;YACA,0BAA0BJ;YAC1B,cAAcK,WAAW,MAAM;YAC/BR;YACAC;QACF;QAEF,MAAMgB,WAAWP,KAAK,GAAG,KAAKD;QAC9B,MAAMS,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACN;YAC5B,gBAAgBM,KAAK,SAAS,CAACR;YAC/BG;YACA,YAAYN,WAAW,KAAK,CAAC,UAAU;YACvC,uBAAuBA,WAAW,KAAK,CAAC,WAAW;YACnD,iBAAiBA,WAAW,KAAK,CAAC,KAAK;YACvCO;QACF;QAEA,IAAIK;QACJ,IAAIT,YAAY,MAAM,EAAE,QACtBS,WAAW,CAAC,4BAA4B,EAAET,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG3E,MAAMU,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAASlB;YACX;YACA,aAAaS;YACb,MAAM;YACNM;YACA,YAAY,CAAC,CAACV,WAAW,KAAK,CAAC,UAAU;YACzC,OAAOY;QACT;QAEA,MAAME,UAAUX,YAAY,OAAO;QAEnC,MAAMY,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACX,gBAAgBC,UAAU;gBAACA;aAAQ,GAAG,EAAE;QAC1C;QAEA,IAAIF,UACF,MAAM,IAAIK,aAAaL,UAAUG;QAGnC,IAAID,SACF,OAAO;YACL,SAAS;gBACP,QAAQA,QAAQ,MAAM;gBACtB,MAAMA,QAAQ,IAAI;gBAClB,aAAaA,QAAQ,WAAW;YAClC;YACAV;YACAW;QACF;QAGF,OAAO;YACL,SAAS;YACTX;YACAW;QACF;IACF;IAEA,MAAc,wBAAwBG,OAOrC,EAAmC;QAClC,MAAM,EAAE5B,KAAK,EAAEK,WAAW,EAAEJ,GAAG,EAAEQ,OAAO,EAAEP,YAAY,EAAEC,WAAW,EAAE,GACnEyB;QACF,MAAM,EAAEC,OAAO,EAAE,GAAG3B;QACpB,MAAM4B,wBAAwB,CAAC,CAAC7B,KAAK,oBAAoB;QAEzD,IAAI,CAACD,MAAM,UAAU,EACnB,OAAO;YAAE,OAAO,CAAC;QAAE;QAGrB,IAAI8B,uBAAuB;YACzB,MAAMC,SAAS,MAAMC,sBAAsB;gBACzCvB;gBACA,UAAUR,IAAI,kBAAkB,CAAE,IAAI;YACxC;YAEA,OAAO;gBACL8B;gBACA,OAAO;oBACL,YAAYA,OAAO,UAAU;oBAC7B,aAAaV,KAAK,SAAS,CAAC;wBAC1B,QAAQ;wBACR,MAAMpB,IAAI,kBAAkB,CAAE,IAAI;oBACpC;gBACF;YACF;QACF;QAEA,IAAI4B,QAAQ,MAAM,CAAC,kBAAkB,EAAE;YACrC,MAAMI,qBAAqB,MAAMC,gBAAgB;gBAC/CzB;gBACA,oBAAoBJ;gBACpBH;gBACAC;YACF;YACA,MAAM,EAAEgC,gBAAgB,EAAE,GAAGF;YAC7B3B,OACE6B,kBACA,CAAC,6BAA6B,EAAE9B,YAAY,CAAC,EAC3C4B,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAGJ,OAAO;gBACL,QAAQE;gBACR,OAAO;oBACL,YAAYA,iBAAiB,UAAU;oBACvC,aAAaF,mBAAmB,WAAW;oBAC3C,OAAOA,mBAAmB,KAAK;gBACjC;YACF;QACF;QAEA,MAAMG,wBAAwB,MAAMlB,gBAAgB;YAClDT;YACA,0BAA0BJ;YAC1BH;YACAC;QACF;QACAG,OACE8B,sBAAsB,IAAI,EAC1B,CAAC,6BAA6B,EAAE/B,YAAY,CAAC,EAC3C+B,sBAAsB,WAAW,CAAC,MAAM,EAAE,SACtC,CAAC,EAAE,EAAEA,sBAAsB,WAAW,CAAC,MAAM,CAAC,IAAI,CAAC,OAAO,GAC1D,IACJ;QAGJ,MAAML,SAAS,MAAMC,sBAAsB;YACzCvB;YACA,UAAU2B,sBAAsB,IAAI;QACtC;QAEA,OAAO;YACLL;YACA,OAAO;gBACL,YAAYA,OAAO,UAAU;gBAC7B,aAAaV,KAAK,SAAS,CAAC;oBAC1B,QAAQ;oBACR,MAAMe,sBAAsB,IAAI;oBAChC,aAAaA,sBAAsB,WAAW;gBAChD;gBACA,OAAOA,sBAAsB,KAAK;YACpC;QACF;IACF;IAEA,MAAM,QACJC,UAA+B,EAC/BnC,YAA0B,EAC1BD,GAA0B,EAC1BqC,eAAwB,EACxBC,gBAAoC,EACpC9B,OAAmB,EACe;QAClCH,OAAOG,SAAS;QAChBH,OACE,AAAsB,YAAtB,OAAO+B,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAGvE,MAAM1B,YAAYC,KAAK,GAAG;QAE1B,IAAIC;QAGJ,IAAIE;QACJ,IAAIC;QACJ,IAAIC;QAEJ,IAAI;YACF,MAAMuB,SAAS,MAAMC,qBAAwB;gBAC3ChC;gBACA,WAAW4B;gBACXE;gBACA,eAAetC;gBACfC;gBACAoC;YACF;YACAzB,cAAc2B,OAAO,WAAW;YAChCzB,cAAcyB,OAAO,WAAW;YAChCxB,QAAQwB,OAAO,KAAK;YACpBvB,oBAAoBuB,OAAO,iBAAiB;QAC9C,EAAE,OAAOE,OAAO;YACd,IAAIA,iBAAiBC,sBAAsB;gBAEzC,MAAMxB,WAAWP,KAAK,GAAG,KAAKD;gBAC9B,MAAMS,WAA4B;oBAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;oBACtC,YAAYD;oBACZ,aAAauB,MAAM,WAAW;oBAC9B,OAAOA,MAAM,KAAK;gBACpB;gBACA,MAAMjB,OAAOC,kBAAkB;oBAC7B,MAAM;oBACN,WAAW;wBAAEW;oBAAW;oBACxB,MAAM;oBACNjB;oBACA,OAAOsB,MAAM,OAAO;gBACtB;gBACA,MAAM,IAAIf,aAAae,MAAM,OAAO,EAAEjB;YACxC;YACA,MAAMiB;QACR;QAEA,MAAMvB,WAAWP,KAAK,GAAG,KAAKD;QAC9B,MAAMS,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZJ;YACA,gBAAgBM,KAAK,SAAS,CAACR;YAC/BG;YACAC;QACF;QAEA,IAAIK;QACJ,IAAIT,YAAY,MAAM,EAAE,QACtBS,WAAW,CAAC,qBAAqB,EAAET,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMU,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTc;YACF;YACA,MAAM;YACNjB;YACA,OAAOE;QACT;QAEA,MAAM,EAAEsB,IAAI,EAAEC,OAAO,EAAE,GAAGhC,eAAe,CAAC;QAE1C,MAAMY,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACXqB;QACF;QAEA,IAAItB,YAAY,CAACsB,MACf,MAAM,IAAIjB,aAAaL,UAAUG;QAGnC,OAAO;YACLmB;YACAC;YACA7B;YACAC;YACAQ;QACF;IACF;IAEA,MAAM,SACJqB,MAA+B,EAC/B5C,YAA0B,EAC1BD,GAEC,EACwD;QACzDK,OAAOwC,QAAQ;QACf,MAAMrC,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAC7C,MAAM,EAAEsC,QAAQ,EAAE,GAAGtC;QACrB,MAAMuC,mBAAmBvC,QAAQ,UAAU,CAAC,MAAM;QAClDH,OAAO0C,kBAAkB;QACzB,MAAMC,eAAeC;QAGrB,MAAMC,kBAAkB;QACxB,MAAMC,aAAmBC,MAAM,OAAO,CAACP,UACnC;YACE,MAAMQ,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC/C,KAAKG,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC9C,OAAOA;YACP,QAAQA;QACV,IACAL;QAEJ,IAAIS,eAAe,MAAMC,wBAAwB;YAC/C,gBAAgBR;YAChB,MAAMD;YACN,sBAAsB;gBACpB;oBACE,MAAMK;gBACR;aACD;YACD,iBAAiB;QACnB;QAEA,IAAInD,KAAK,YAAY;YACnB,MAAMS,aAAa+C,iBAAiBL,YAAYL;YAQhDlD,MAAM,oCAAoCa;YAC1C,MAAMgD,gBAAgB,MAAMC,WAAWJ,cAAc7C;YACrD6C,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,MAAM,MAAMC,yBAChBF,MACA1D;QAGF,MAAM,EAAE6D,OAAO,EAAE,GAAGF;QACpBvD,OAAO,CAACyD,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1DzD,OAAOyD,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IAjYA,YACEtD,OAA2D,EAC3DR,GAAoB,CACpB;QAPF;QAEA;QAMEK,OAAOG,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMuD,QAAQ,OAAO,CAACvD;QAGlD,IAAI,AAAyB,WAAlBR,KAAK,UACd,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAoXF"}

package/dist/es/types.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"types.mjs","sources":["../../src/types.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport type { NodeType } from '@midscene/shared/constants';\nimport type { CreateOpenAIClientFn, TModelConfig } from '@midscene/shared/env';\nimport type {\n  BaseElement,\n  LocateResultElement,\n  Rect,\n  Size,\n} from '@midscene/shared/types';\nimport type { z } from 'zod';\nimport type { TUserPrompt } from './common';\nimport type { ScreenshotItem } from './screenshot-item';\nimport type {\n  DetailedLocateParam,\n  MidsceneYamlFlowItem,\n  ServiceExtractOption,\n} from './yaml';\n\nexport type {\n  ElementTreeNode,\n  BaseElement,\n  Rect,\n  Size,\n  Point,\n} from '@midscene/shared/types';\nexport * from './yaml';\n\nexport { ServiceError } from './errors';\nexport {\n  ExecutionDump,\n  ReportActionDump,\n  GroupedActionDump,\n} from './dump/report-action-dump';\n\nexport type AIUsageInfo = Record<string, any> & {\n  prompt_tokens: number | undefined;\n  completion_tokens: number | undefined;\n  total_tokens: number | undefined;\n  cached_input: number | undefined;\n  time_cost: number | undefined;\n  model_name: string | undefined;\n  model_description: string | undefined;\n  /**\n   * Semantic intent of the model call, such as default, planning, or insight.\n   */\n  intent: string | undefined;\n  /**\n   * Config slot where the model config was resolved from. For example, a\n   * planning call may use the default slot when no planning model is configured.\n   */\n  slot: string | undefined;\n  request_id: string | undefined;\n};\n\nexport type { LocateResultElement };\n\nexport type AISingleElementResponseByPosition = {\n  position?: {\n    x: number;\n    y: number;\n  };\n  bbox?: [number, number, number, number];\n  reason: string;\n  text: string;\n};\n\nexport interface AIElementCoordinatesResponse {\n  bbox: [number, number, number, number];\n  errors?: string[];\n}\n\nexport type AIElementResponse = AIElementCoordinatesResponse;\n\nexport interface AIDataExtractionResponse<DataDemand> {\n  data: DataDemand;\n  errors?: string[];\n  thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n  bbox: [number, number, number, number];\n  references_bbox?: [number, number, number, number][];\n  error?: string;\n}\n\nexport interface AIAssertionResponse {\n  pass: boolean;\n  thought: string;\n}\n\nexport interface AIDescribeElementResponse {\n  description: string;\n  error?: string;\n}\n\nexport interface LocatorValidatorOption {\n  centerDistanceThreshold?: number;\n}\n\nexport interface LocateValidatorResult {\n  pass: boolean;\n  rect: Rect;\n  center: [number, number];\n  centerDistance?: number;\n}\n\nexport interface AgentDescribeElementAtPointResult {\n  prompt: string;\n  deepLocate: boolean;\n  verifyResult?: LocateValidatorResult;\n}\n\n/**\n * context\n */\n\nexport abstract class UIContext {\n  /**\n   * screenshot of the current UI state. which size is shotSize(be shrunk by screenshotShrinkFactor),\n   */\n  abstract screenshot: ScreenshotItem;\n\n  /**\n   * screenshot size after shrinking\n   */\n  abstract shotSize: Size;\n\n  /**\n   * The ratio for converting shrunk screenshot coordinates to logical coordinates.\n   *\n   * Example:\n   * - Physical screen width: 3000px, dpr=6\n   * - Logical width: 500px\n   * - User-defined screenshotShrinkFactor: 2\n   * - Actual shrunk screenshot width: 3000 / 2 = 1500px\n   * - shrunkShotToLogicalRatio: dpr / screenshotShrinkFactor = 6 / 2 = 3\n   * - To map back to logical coordinates: 1500 / shrunkShotToLogicalRatio = 500px\n   */\n  abstract shrunkShotToLogicalRatio: number;\n\n  abstract _isFrozen?: boolean;\n\n  // @deprecated - backward compatibility for aiLocate\n  abstract deprecatedDpr?: number;\n}\n\nexport type EnsureObject<T> = { [K in keyof T]: any };\n\nexport type ServiceAction = 'locate' | 'extract' | 'assert' | 'describe';\n\nexport type ServiceExtractParam = string | Record<string, string>;\n\nexport type ElementCacheFeature = Record<string, unknown>;\n\nexport interface LocateResult {\n  element: LocateResultElement | null;\n  rect?: Rect;\n}\n\nexport type ThinkingLevel = 'off' | 'medium' | 'high';\n\nexport type DeepThinkOption = 'unset' | true | false;\n\nexport interface ServiceTaskInfo {\n  durationMs: number;\n  formatResponse?: string;\n  rawResponse?: string;\n  usage?: AIUsageInfo;\n  searchArea?: Rect;\n  searchAreaRawResponse?: string;\n  searchAreaUsage?: AIUsageInfo;\n  reasoning_content?: string;\n}\n\nexport interface DumpMeta {\n  logTime: number;\n}\n\nexport type ReportAttributes = Record<\n  string,\n  string | number | boolean | null | undefined\n>;\n\nexport interface ReportDumpWithAttributes {\n  dumpString: string;\n  attributes?: ReportAttributes;\n}\n\nexport interface ServiceDump extends DumpMeta {\n  type: 'locate' | 'extract' | 'assert';\n  logId: string;\n  userQuery: {\n    element?: TUserPrompt;\n    dataDemand?: ServiceExtractParam;\n    assertion?: TUserPrompt;\n  };\n  matchedElement: LocateResultElement[];\n  matchedRect?: Rect;\n  deepLocate?: boolean;\n  data: any;\n  assertionPass?: boolean;\n  assertionThought?: string;\n  taskInfo: ServiceTaskInfo;\n  error?: string;\n  output?: any;\n}\n\nexport type PartialServiceDumpFromSDK = Omit<\n  ServiceDump,\n  'logTime' | 'logId' | 'model_name'\n>;\n\nexport interface ServiceResultBase {\n  dump: ServiceDump;\n}\n\nexport type LocateResultWithDump = LocateResult & ServiceResultBase;\n\nexport interface ServiceExtractResult<T> extends ServiceResultBase {\n  data: T;\n  thought?: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}\n\n// intermediate variables to optimize the return value by AI\nexport interface LiteUISection {\n  name: string;\n  description: string;\n  sectionCharacteristics: string;\n  textIds: string[];\n}\n\nexport type ElementById = (id: string) => BaseElement | null;\n\nexport type ServiceAssertionResponse = AIAssertionResponse & {\n  usage?: AIUsageInfo;\n};\n\n/**\n * agent\n */\n\nexport type OnTaskStartTip = (tip: string) => Promise<void> | void;\n\nexport interface AgentWaitForOpt extends ServiceExtractOption {\n  checkIntervalMs?: number;\n  timeoutMs?: number;\n}\n\nexport interface AgentAssertOpt {\n  keepRawResponse?: boolean;\n}\n\n/**\n * planning\n *\n */\n\nexport interface PlanningLocateParam extends DetailedLocateParam {\n  bbox?: [number, number, number, number];\n}\n\nexport interface PlanningAction<ParamType = any> {\n  thought?: string;\n  log?: string; // a brief preamble to the user explaining what you’re about to do\n  type: string;\n  param: ParamType;\n}\n\nexport type SubGoalStatus = 'pending' | 'running' | 'finished';\n\nexport interface SubGoal {\n  index: number;\n  status: SubGoalStatus;\n  description: string;\n  logs?: string[];\n}\n\nexport interface RawResponsePlanningAIResponse {\n  action: PlanningAction;\n  thought?: string;\n  log: string;\n  memory?: string;\n  error?: string;\n  finalizeMessage?: string;\n  finalizeSuccess?: boolean;\n  updateSubGoals?: SubGoal[];\n  markFinishedIndexes?: number[];\n}\n\nexport interface PlanningAIResponse\n  extends Omit<RawResponsePlanningAIResponse, 'action'> {\n  actions?: PlanningAction[];\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  yamlString?: string;\n  error?: string;\n  reasoning_content?: string;\n  shouldContinuePlanning: boolean;\n  output?: string; // Output message from <complete> tag (same as finalizeMessage)\n}\n\nexport interface PlanningActionParamSleep {\n  timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n  thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {};\n\nexport interface LongPressParam {\n  duration?: number;\n}\n\nexport interface PullParam {\n  direction: 'up' | 'down';\n  distance?: number;\n  duration?: number;\n}\n/**\n * misc\n */\n\nexport interface Color {\n  name: string;\n  hex: string;\n}\n\nexport interface BaseAgentParserOpt {\n  selector?: string;\n}\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PuppeteerParserOpt extends BaseAgentParserOpt {}\n\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PlaywrightParserOpt extends BaseAgentParserOpt {}\n\n/*\naction\n*/\nexport interface ExecutionTaskProgressOptions {\n  onTaskStart?: (task: ExecutionTask) => Promise<void> | void;\n}\n\nexport interface ExecutionRecorderItem {\n  type: 'screenshot';\n  ts: number;\n  screenshot?: ScreenshotItem;\n  timing?: string;\n}\n\nexport type ExecutionTaskType = 'Planning' | 'Insight' | 'Action Space' | 'Log';\n\nexport interface ExecutorContext {\n  task: ExecutionTask;\n  element?: LocateResultElement | null;\n  uiContext?: UIContext;\n}\n\nexport interface ExecutionTaskApply<\n  Type extends ExecutionTaskType = any,\n  TaskParam = any,\n  TaskOutput = any,\n  TaskLog = any,\n> {\n  type: Type;\n  subType?: string;\n  param?: TaskParam;\n  thought?: string;\n  uiContext?: UIContext;\n  executor: (\n    param: TaskParam,\n    context: ExecutorContext,\n  ) => // biome-ignore lint/suspicious/noConfusingVoidType: void is intentionally allowed as some executors may not return a value\n    | Promise<ExecutionTaskReturn<TaskOutput, TaskLog> | undefined | void>\n    | undefined\n    | void;\n}\n\nexport interface ExecutionTaskHitBy {\n  from: string;\n  context: Record<string, any>;\n}\n\nexport interface ExecutionTaskReturn<TaskOutput = unknown, TaskLog = unknown> {\n  output?: TaskOutput;\n  log?: TaskLog;\n  recorder?: ExecutionRecorderItem[];\n  hitBy?: ExecutionTaskHitBy;\n}\n\nexport type ExecutionTask<\n  E extends ExecutionTaskApply<any, any, any> = ExecutionTaskApply<\n    any,\n    any,\n    any\n  >,\n> = E &\n  ExecutionTaskReturn<\n    E extends ExecutionTaskApply<any, any, infer TaskOutput, any>\n      ? TaskOutput\n      : unknown,\n    E extends ExecutionTaskApply<any, any, any, infer TaskLog>\n      ? TaskLog\n      : unknown\n  > & {\n    taskId: string;\n    status: 'pending' | 'running' | 'finished' | 'failed' | 'cancelled';\n    error?: Error;\n    errorMessage?: string;\n    errorStack?: string;\n    timing?: {\n      start: number;\n      getUiContextStart?: number;\n      getUiContextEnd?: number;\n      callAiStart?: number;\n      callAiEnd?: number;\n      beforeInvokeActionHookStart?: number;\n      beforeInvokeActionHookEnd?: number;\n      callActionStart?: number;\n      callActionEnd?: number;\n      afterInvokeActionHookStart?: number;\n      afterInvokeActionHookEnd?: number;\n      captureAfterCallingSnapshotStart?: number;\n      captureAfterCallingSnapshotEnd?: number;\n      end?: number;\n      cost?: number;\n    };\n    usage?: AIUsageInfo;\n    searchAreaUsage?: AIUsageInfo;\n    reasoning_content?: string;\n  };\n\nexport interface IExecutionDump extends DumpMeta {\n  /** Stable unique identifier for this execution run */\n  id?: string;\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n  aiActContext?: string;\n}\n\n/*\ntask - service-locate\n*/\nexport type ExecutionTaskInsightLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskInsightLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport type ExecutionTaskInsightDump = ServiceDump;\n\nexport type ExecutionTaskInsightLocateApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightLocateParam,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightLocate =\n  ExecutionTask<ExecutionTaskInsightLocateApply>;\n\n/*\ntask - service-query\n*/\nexport interface ExecutionTaskInsightQueryParam {\n  dataDemand: ServiceExtractParam;\n  domIncluded?: boolean | 'visible-only';\n}\n\nexport interface ExecutionTaskInsightQueryOutput {\n  data: any;\n}\n\nexport type ExecutionTaskInsightQueryApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightQueryParam,\n  any,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightQuery =\n  ExecutionTask<ExecutionTaskInsightQueryApply>;\n\n/*\ntask - assertion\n*/\nexport interface ExecutionTaskInsightAssertionParam {\n  assertion: string;\n}\n\nexport type ExecutionTaskInsightAssertionApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightAssertionParam,\n  ServiceAssertionResponse,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightAssertion =\n  ExecutionTask<ExecutionTaskInsightAssertionApply>;\n\n/*\ntask - action (i.e. interact) \n*/\nexport type ExecutionTaskActionApply<ActionParam = any> = ExecutionTaskApply<\n  'Action Space',\n  ActionParam,\n  void,\n  void\n>;\n\nexport type ExecutionTaskAction = ExecutionTask<ExecutionTaskActionApply>;\n\n/*\ntask - Log\n*/\n\nexport type ExecutionTaskLogApply<\n  LogParam = {\n    content: string;\n  },\n> = ExecutionTaskApply<'Log', LogParam, void, void>;\n\nexport type ExecutionTaskLog = ExecutionTask<ExecutionTaskLogApply>;\n\n/*\ntask - planning\n*/\n\nexport interface ExecutionTaskPlanningParam {\n  userInstruction: string;\n  aiActContext?: string;\n  imagesIncludeCount?: number;\n  deepThink?: DeepThinkOption;\n  subGoalStatus?: string;\n  memoriesStatus?: string;\n}\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n  'Planning',\n  ExecutionTaskPlanningParam,\n  PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\ntask - planning-locate\n*/\nexport type ExecutionTaskPlanningLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskPlanningLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport type ExecutionTaskPlanningDump = ServiceDump;\n\nexport type ExecutionTaskPlanningLocateApply = ExecutionTaskApply<\n  'Planning',\n  ExecutionTaskPlanningLocateParam,\n  ExecutionTaskPlanningLocateOutput,\n  ExecutionTaskPlanningDump\n>;\n\nexport type ExecutionTaskPlanningLocate =\n  ExecutionTask<ExecutionTaskPlanningLocateApply>;\n\n/*\nReport metadata - extracted from ReportActionDump for per-execution writes\n*/\nexport interface ReportMeta {\n  groupName: string;\n  groupDescription?: string;\n  sdkVersion: string;\n  modelBriefs: ModelBrief[];\n  deviceType?: string;\n}\n\n// Backward-compatible aliases for existing external consumers.\nexport type GroupMeta = ReportMeta;\n\n/*\nReport dump\n*/\nexport interface IReportActionDump {\n  sdkVersion: string;\n  groupName: string;\n  groupDescription?: string;\n  modelBriefs: ModelBrief[];\n  executions: IExecutionDump[];\n  deviceType?: string;\n}\n\n// Backward-compatible aliases for existing external consumers.\nexport type IGroupedActionDump = IReportActionDump;\n\nexport interface ModelBrief {\n  /**\n   * The intent/category of the model call, for example \"planning\" or \"insight\".\n   */\n  intent?: string;\n\n  /**\n   * The model name returned by usage metadata, for example \"gpt-4o\".\n   */\n  name?: string;\n\n  /**\n   * Optional human-readable model description, for example \"qwen2.5-vl mode\".\n   */\n  modelDescription?: string;\n}\n\nexport type InterfaceType =\n  | 'puppeteer'\n  | 'playwright'\n  | 'static'\n  | 'chrome-extension-proxy'\n  | 'android'\n  | string;\n\nexport interface StreamingCodeGenerationOptions {\n  /** Whether to enable streaming output */\n  stream?: boolean;\n  /** Callback function to handle streaming chunks */\n  onChunk?: StreamingCallback;\n  /** Callback function to handle streaming completion */\n  onComplete?: (finalCode: string) => void;\n  /** Callback function to handle streaming errors */\n  onError?: (error: Error) => void;\n}\n\nexport type StreamingCallback = (chunk: CodeGenerationChunk) => void;\n\nexport interface CodeGenerationChunk {\n  /** The incremental content chunk */\n  content: string;\n  /** The reasoning content */\n  reasoning_content: string;\n  /** The accumulated content so far */\n  accumulated: string;\n  /** Whether this is the final chunk */\n  isComplete: boolean;\n  /** Token usage information if available */\n  usage?: AIUsageInfo;\n}\n\nexport interface StreamingAIResponse {\n  /** The final accumulated content */\n  content: string;\n  /** Token usage information */\n  usage?: AIUsageInfo;\n  /** Whether the response was streamed */\n  isStreamed: boolean;\n}\n\nexport interface DeviceAction<TParam = any, TReturn = any> {\n  name: string;\n  description?: string;\n  interfaceAlias?: string;\n  paramSchema?: z.ZodType<TParam>;\n  call: (param: TParam, context: ExecutorContext) => Promise<TReturn> | TReturn;\n  delayBeforeRunner?: number;\n  delayAfterRunner?: number;\n  /**\n   * An example param object for this action.\n   * Locate fields with { prompt } will automatically get bbox injected when needed.\n   */\n  sample?: { [K in keyof TParam]?: any };\n}\n\n/**\n * Type utilities for extracting types from DeviceAction definitions\n */\n\n/**\n * Extract parameter type from a DeviceAction\n */\nexport type ActionParam<Action extends DeviceAction<any, any>> =\n  Action extends DeviceAction<infer P, any> ? P : never;\n\n/**\n * Extract return type from a DeviceAction\n */\nexport type ActionReturn<Action extends DeviceAction<any, any>> =\n  Action extends DeviceAction<any, infer R> ? R : never;\n\n/**\n * Web-specific types\n */\nexport interface WebElementInfo extends BaseElement {\n  id: string;\n  attributes: {\n    nodeType: NodeType;\n    [key: string]: string;\n  };\n}\n\n/**\n * Agent\n */\n\nexport type CacheConfig = {\n  strategy?: 'read-only' | 'read-write' | 'write-only';\n  id: string;\n  /**\n   * Optional cache directory path.\n   * When set, cache files are written to this directory instead of\n   * `<MIDSCENE_RUN_DIR>/cache`.\n   */\n  cacheDir?: string;\n};\n\nexport type Cache =\n  | false // No read, no write\n  | true // Will throw error at runtime - deprecated\n  | CacheConfig; // Object configuration (requires explicit id)\n\nexport interface AgentOpt {\n  // @deprecated Use `reportFileName` and `cache.id` instead.\n  testId?: string;\n  // @deprecated\n  cacheId?: string; // Keep backward compatibility, but marked as deprecated\n  groupName?: string;\n  groupDescription?: string;\n  /* if auto generate report, default true */\n  generateReport?: boolean;\n  /* if persist per-execution dump files next to the report, default false */\n  persistExecutionDump?: boolean;\n  /* if auto print report msg, default true */\n  autoPrintReportMsg?: boolean;\n\n  /**\n   * Use directory-based report format with separate image files.\n   *\n   * When enabled:\n   * - Screenshots are saved as PNG files in a `screenshots/` subdirectory\n   * - Report is generated as `index.html` with relative image paths\n   * - Reduces memory usage and report file size\n   *\n   * IMPORTANT: 'html-and-external-assets' reports must be served via HTTP server\n   * (e.g., `npx serve ./report-dir`). The file:// protocol will not\n   * work due to browser CORS restrictions.\n   *\n   * @default 'single-html'\n   */\n  outputFormat?: 'single-html' | 'html-and-external-assets';\n\n  onTaskStartTip?: OnTaskStartTip;\n  aiActContext?: string;\n  aiActionContext?: string;\n  /* custom report file name */\n  reportFileName?: string;\n  reportAttributes?: ReportAttributes;\n  modelConfig?: TModelConfig;\n  cache?: Cache;\n  /**\n   * Maximum number of replanning cycles for aiAct.\n   * Defaults to 20 (40 for `vlm-ui-tars`) when not provided.\n   * If omitted, the agent will also read `MIDSCENE_REPLANNING_CYCLE_LIMIT` for backward compatibility.\n   */\n  replanningCycleLimit?: number;\n\n  /**\n   * Wait time in milliseconds after each action execution.\n   * This allows the UI to settle and stabilize before the next action.\n   * Defaults to 300ms when not provided.\n   */\n  waitAfterAction?: number;\n\n  /**\n   * When set to true, Midscene will use the target device's formatted local\n   * time instead of the runtime system time. The target interface must implement\n   * getDeviceLocalTimeString to provide device-local wall-clock time.\n   * Default: false\n   */\n  useDeviceTime?: boolean;\n\n  /**\n   * Custom screenshot shrink factor to reduce AI token usage.\n   * When set, the screenshot will be scaled down by this factor from the physical resolution.\n   *\n   * Example:\n   * - Physical screen width: 3000px, dpr=6\n   * - Logical width: 500px\n   * - screenshotShrinkFactor: 2\n   * - Actual shrunk screenshot width: 3000 / 2 = 1500px\n   * - AI analyzes the 1500px screenshot\n   * - Coordinates are transformed back to logical (500px) before actions execute\n   *\n   * Benefits:\n   * - Reduces token usage for high-resolution screenshots\n   * - Maintains accuracy by scaling coordinates appropriately\n   *\n   * Must be >= 1 (shrinking only, enlarging is not supported).\n   *\n   * @default 1 (no shrinking, uses original physical screenshot)\n   */\n  screenshotShrinkFactor?: number;\n\n  /**\n   * Custom OpenAI client factory function\n   *\n   * If provided, this function will be called to create OpenAI client instances\n   * for each AI call, allowing you to:\n   * - Wrap clients with observability tools (langsmith, langfuse)\n   * - Use custom OpenAI-compatible clients\n   * - Apply different configurations based on intent\n   *\n   * @param config - Resolved model configuration\n   * @returns OpenAI client instance (original or wrapped)\n   *\n   * @example\n   * ```typescript\n   * createOpenAIClient: async (openai, opts) => {\n   *   // Wrap with langsmith for planning tasks\n   *   if (opts.baseURL?.includes('planning')) {\n   *     return wrapOpenAI(openai, { metadata: { task: 'planning' } });\n   *   }\n   *\n   *   return openai;\n   * }\n   * ```\n   */\n  createOpenAIClient?: CreateOpenAIClientFn;\n}\n\nexport type TestStatus =\n  | 'passed'\n  | 'failed'\n  | 'timedOut'\n  | 'skipped'\n  | 'interrupted';\n\nexport interface ReportFileAttributes {\n  testDuration: number;\n  testStatus: TestStatus;\n  testTitle: string;\n  testId: string;\n  testDescription: string;\n}\n\nexport type ReportFileWithAttributes =\n  | {\n      reportFilePath: string;\n      reportAttributes: ReportFileAttributes;\n    }\n  | {\n      reportFilePath?: string;\n      reportAttributes: ReportFileAttributes & { testStatus: 'skipped' };\n    };\n"],"names":["UIContext"],"mappings":";;;AAqHO,MAAeA;AA4BtB"}
+{"version":3,"file":"types.mjs","sources":["../../src/types.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n\nimport type { NodeType } from '@midscene/shared/constants';\nimport type { CreateOpenAIClientFn, TModelConfig } from '@midscene/shared/env';\nimport type {\n  BaseElement,\n  LocateResultElement,\n  Rect,\n  Size,\n} from '@midscene/shared/types';\nimport type { z } from 'zod';\nimport type { TUserPrompt } from './common';\nimport type { ScreenshotItem } from './screenshot-item';\nimport type {\n  DetailedLocateParam,\n  MidsceneYamlFlowItem,\n  ServiceExtractOption,\n} from './yaml';\n\nexport type {\n  ElementTreeNode,\n  BaseElement,\n  Rect,\n  Size,\n  Point,\n} from '@midscene/shared/types';\nexport * from './yaml';\n\nexport { ServiceError } from './errors';\nexport {\n  ExecutionDump,\n  ReportActionDump,\n  GroupedActionDump,\n} from './dump/report-action-dump';\n\nexport type AIUsageInfo = Record<string, any> & {\n  prompt_tokens: number | undefined;\n  completion_tokens: number | undefined;\n  total_tokens: number | undefined;\n  cached_input: number | undefined;\n  time_cost: number | undefined;\n  model_name: string | undefined;\n  model_description: string | undefined;\n  /**\n   * Semantic intent of the model call, such as default, planning, or insight.\n   */\n  intent: string | undefined;\n  /**\n   * Config slot where the model config was resolved from. For example, a\n   * planning call may use the default slot when no planning model is configured.\n   */\n  slot: string | undefined;\n  request_id: string | undefined;\n};\n\nexport type { LocateResultElement };\n\nexport type AISingleElementResponseByPosition = {\n  position?: {\n    x: number;\n    y: number;\n  };\n  bbox?: [number, number, number, number];\n  reason: string;\n  text: string;\n};\n\nexport type LocateResultPoint = [number, number];\nexport type Bbox = [number, number, number, number];\nexport type LocateResultBbox = Bbox;\nexport type PixelBbox = Bbox;\n\nexport interface AIElementLocateResponse {\n  bbox?: LocateResultBbox;\n  point?: LocateResultPoint;\n  errors?: string[];\n}\n\nexport interface AIDataExtractionResponse<DataDemand> {\n  data: DataDemand;\n  errors?: string[];\n  thought?: string;\n}\n\nexport interface AISectionLocatorResponse {\n  bbox?: LocateResultBbox;\n  point?: LocateResultPoint;\n  references_bbox?: LocateResultBbox[];\n  references_point?: LocateResultPoint[];\n  error?: string;\n}\n\nexport interface AIAssertionResponse {\n  pass: boolean;\n  thought: string;\n}\n\nexport interface AIDescribeElementResponse {\n  description: string;\n  error?: string;\n}\n\nexport interface LocatorValidatorOption {\n  centerDistanceThreshold?: number;\n}\n\nexport interface LocateValidatorResult {\n  pass: boolean;\n  rect: Rect;\n  center: [number, number];\n  centerDistance?: number;\n}\n\nexport interface AgentDescribeElementAtPointResult {\n  prompt: string;\n  deepLocate: boolean;\n  verifyResult?: LocateValidatorResult;\n}\n\n/**\n * context\n */\n\nexport abstract class UIContext {\n  /**\n   * screenshot of the current UI state. which size is shotSize(be shrunk by screenshotShrinkFactor),\n   */\n  abstract screenshot: ScreenshotItem;\n\n  /**\n   * screenshot size after shrinking\n   */\n  abstract shotSize: Size;\n\n  /**\n   * The ratio for converting shrunk screenshot coordinates to logical coordinates.\n   *\n   * Example:\n   * - Physical screen width: 3000px, dpr=6\n   * - Logical width: 500px\n   * - User-defined screenshotShrinkFactor: 2\n   * - Actual shrunk screenshot width: 3000 / 2 = 1500px\n   * - shrunkShotToLogicalRatio: dpr / screenshotShrinkFactor = 6 / 2 = 3\n   * - To map back to logical coordinates: 1500 / shrunkShotToLogicalRatio = 500px\n   */\n  abstract shrunkShotToLogicalRatio: number;\n\n  abstract _isFrozen?: boolean;\n\n  // @deprecated - backward compatibility for aiLocate\n  abstract deprecatedDpr?: number;\n}\n\nexport type EnsureObject<T> = { [K in keyof T]: any };\n\nexport type ServiceAction = 'locate' | 'extract' | 'assert' | 'describe';\n\nexport type ServiceExtractParam = string | Record<string, string>;\n\nexport type ElementCacheFeature = Record<string, unknown>;\n\nexport interface LocateResult {\n  element: LocateResultElement | null;\n  rect?: Rect;\n}\n\nexport type ThinkingLevel = 'off' | 'medium' | 'high';\n\nexport type DeepThinkOption = 'unset' | true | false;\n\nexport interface ServiceTaskInfo {\n  durationMs: number;\n  formatResponse?: string;\n  rawResponse?: string;\n  usage?: AIUsageInfo;\n  searchArea?: Rect;\n  searchAreaRawResponse?: string;\n  searchAreaUsage?: AIUsageInfo;\n  reasoning_content?: string;\n}\n\nexport interface DumpMeta {\n  logTime: number;\n}\n\nexport type ReportAttributes = Record<\n  string,\n  string | number | boolean | null | undefined\n>;\n\nexport interface ReportDumpWithAttributes {\n  dumpString: string;\n  attributes?: ReportAttributes;\n}\n\nexport interface ServiceDump extends DumpMeta {\n  type: 'locate' | 'extract' | 'assert';\n  logId: string;\n  userQuery: {\n    element?: TUserPrompt;\n    dataDemand?: ServiceExtractParam;\n    assertion?: TUserPrompt;\n  };\n  matchedElement?: LocateResultElement[];\n  matchedRect?: Rect;\n  deepLocate?: boolean;\n  data: any;\n  assertionPass?: boolean;\n  assertionThought?: string;\n  taskInfo: ServiceTaskInfo;\n  error?: string;\n  output?: any;\n}\n\nexport type PartialServiceDumpFromSDK = Omit<\n  ServiceDump,\n  'logTime' | 'logId' | 'model_name'\n>;\n\nexport interface ServiceResultBase {\n  dump: ServiceDump;\n}\n\nexport type LocateResultWithDump = LocateResult & ServiceResultBase;\n\nexport interface ServiceExtractResult<T> extends ServiceResultBase {\n  data: T;\n  thought?: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}\n\n// intermediate variables to optimize the return value by AI\nexport interface LiteUISection {\n  name: string;\n  description: string;\n  sectionCharacteristics: string;\n  textIds: string[];\n}\n\nexport type ElementById = (id: string) => BaseElement | null;\n\nexport type ServiceAssertionResponse = AIAssertionResponse & {\n  usage?: AIUsageInfo;\n};\n\n/**\n * agent\n */\n\nexport type OnTaskStartTip = (tip: string) => Promise<void> | void;\n\nexport interface AgentWaitForOpt extends ServiceExtractOption {\n  checkIntervalMs?: number;\n  timeoutMs?: number;\n}\n\nexport interface AgentAssertOpt {\n  keepRawResponse?: boolean;\n}\n\n/**\n * planning\n *\n */\n\nexport interface PlanningLocateParam extends DetailedLocateParam {\n  bbox?: LocateResultBbox;\n  point?: LocateResultPoint;\n}\n\nexport type PlanningLocateParamWithLocatedPixelBbox = PlanningLocateParam & {\n  /** Pixel bbox of the located element in screenshot coordinates. */\n  locatedPixelBbox: PixelBbox;\n};\n\nexport interface PlanningAction<ParamType = any> {\n  thought?: string;\n  log?: string; // a brief preamble to the user explaining what you’re about to do\n  type: string;\n  param: ParamType;\n}\n\nexport type SubGoalStatus = 'pending' | 'running' | 'finished';\n\nexport interface SubGoal {\n  index: number;\n  status: SubGoalStatus;\n  description: string;\n  logs?: string[];\n}\n\nexport interface RawResponsePlanningAIResponse {\n  action: PlanningAction;\n  thought?: string;\n  log: string;\n  memory?: string;\n  error?: string;\n  finalizeMessage?: string;\n  finalizeSuccess?: boolean;\n  updateSubGoals?: SubGoal[];\n  markFinishedIndexes?: number[];\n}\n\nexport interface PlanningAIResponse\n  extends Omit<RawResponsePlanningAIResponse, 'action'> {\n  actions?: PlanningAction[];\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  yamlString?: string;\n  error?: string;\n  reasoning_content?: string;\n  shouldContinuePlanning: boolean;\n  output?: string; // Output message from <complete> tag (same as finalizeMessage)\n}\n\nexport interface PlanningActionParamSleep {\n  timeMs: number;\n}\n\nexport interface PlanningActionParamError {\n  thought: string;\n}\n\nexport type PlanningActionParamWaitFor = AgentWaitForOpt & {};\n\nexport interface LongPressParam {\n  duration?: number;\n}\n\nexport interface PullParam {\n  direction: 'up' | 'down';\n  distance?: number;\n  duration?: number;\n}\n/**\n * misc\n */\n\nexport interface Color {\n  name: string;\n  hex: string;\n}\n\nexport interface BaseAgentParserOpt {\n  selector?: string;\n}\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PuppeteerParserOpt extends BaseAgentParserOpt {}\n\n// eslint-disable-next-line @typescript-eslint/no-empty-interface\nexport interface PlaywrightParserOpt extends BaseAgentParserOpt {}\n\n/*\naction\n*/\nexport interface ExecutionTaskProgressOptions {\n  onTaskStart?: (task: ExecutionTask) => Promise<void> | void;\n}\n\nexport interface ExecutionRecorderItem {\n  type: 'screenshot';\n  ts: number;\n  screenshot?: ScreenshotItem;\n  timing?: string;\n}\n\nexport type ExecutionTaskType = 'Planning' | 'Insight' | 'Action Space' | 'Log';\n\nexport interface ExecutorContext {\n  task: ExecutionTask;\n  element?: LocateResultElement | null;\n  uiContext?: UIContext;\n}\n\nexport interface ExecutionTaskApply<\n  Type extends ExecutionTaskType = any,\n  TaskParam = any,\n  TaskOutput = any,\n  TaskLog = any,\n> {\n  type: Type;\n  subType?: string;\n  param?: TaskParam;\n  thought?: string;\n  uiContext?: UIContext;\n  executor: (\n    param: TaskParam,\n    context: ExecutorContext,\n  ) => // biome-ignore lint/suspicious/noConfusingVoidType: void is intentionally allowed as some executors may not return a value\n    | Promise<ExecutionTaskReturn<TaskOutput, TaskLog> | undefined | void>\n    | undefined\n    | void;\n}\n\nexport interface ExecutionTaskHitBy {\n  from: string;\n  context: Record<string, any>;\n}\n\nexport interface ExecutionTaskReturn<TaskOutput = unknown, TaskLog = unknown> {\n  output?: TaskOutput;\n  log?: TaskLog;\n  recorder?: ExecutionRecorderItem[];\n  hitBy?: ExecutionTaskHitBy;\n}\n\nexport type ExecutionTask<\n  E extends ExecutionTaskApply<any, any, any> = ExecutionTaskApply<\n    any,\n    any,\n    any\n  >,\n> = E &\n  ExecutionTaskReturn<\n    E extends ExecutionTaskApply<any, any, infer TaskOutput, any>\n      ? TaskOutput\n      : unknown,\n    E extends ExecutionTaskApply<any, any, any, infer TaskLog>\n      ? TaskLog\n      : unknown\n  > & {\n    taskId: string;\n    status: 'pending' | 'running' | 'finished' | 'failed' | 'cancelled';\n    error?: Error;\n    errorMessage?: string;\n    errorStack?: string;\n    timing?: {\n      start: number;\n      getUiContextStart?: number;\n      getUiContextEnd?: number;\n      callAiStart?: number;\n      callAiEnd?: number;\n      beforeInvokeActionHookStart?: number;\n      beforeInvokeActionHookEnd?: number;\n      callActionStart?: number;\n      callActionEnd?: number;\n      afterInvokeActionHookStart?: number;\n      afterInvokeActionHookEnd?: number;\n      captureAfterCallingSnapshotStart?: number;\n      captureAfterCallingSnapshotEnd?: number;\n      end?: number;\n      cost?: number;\n    };\n    usage?: AIUsageInfo;\n    searchAreaUsage?: AIUsageInfo;\n    reasoning_content?: string;\n  };\n\nexport interface IExecutionDump extends DumpMeta {\n  /** Stable unique identifier for this execution run */\n  id?: string;\n  name: string;\n  description?: string;\n  tasks: ExecutionTask[];\n  aiActContext?: string;\n}\n\n/*\ntask - service-locate\n*/\nexport type ExecutionTaskInsightLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskInsightLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport type ExecutionTaskInsightDump = ServiceDump;\n\nexport type ExecutionTaskInsightLocateApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightLocateParam,\n  ExecutionTaskInsightLocateOutput,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightLocate =\n  ExecutionTask<ExecutionTaskInsightLocateApply>;\n\n/*\ntask - service-query\n*/\nexport interface ExecutionTaskInsightQueryParam {\n  dataDemand: ServiceExtractParam;\n  domIncluded?: boolean | 'visible-only';\n}\n\nexport interface ExecutionTaskInsightQueryOutput {\n  data: any;\n}\n\nexport type ExecutionTaskInsightQueryApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightQueryParam,\n  any,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightQuery =\n  ExecutionTask<ExecutionTaskInsightQueryApply>;\n\n/*\ntask - assertion\n*/\nexport interface ExecutionTaskInsightAssertionParam {\n  assertion: string;\n}\n\nexport type ExecutionTaskInsightAssertionApply = ExecutionTaskApply<\n  'Insight',\n  ExecutionTaskInsightAssertionParam,\n  ServiceAssertionResponse,\n  ExecutionTaskInsightDump\n>;\n\nexport type ExecutionTaskInsightAssertion =\n  ExecutionTask<ExecutionTaskInsightAssertionApply>;\n\n/*\ntask - action (i.e. interact) \n*/\nexport type ExecutionTaskActionApply<ActionParam = any> = ExecutionTaskApply<\n  'Action Space',\n  ActionParam,\n  void,\n  void\n>;\n\nexport type ExecutionTaskAction = ExecutionTask<ExecutionTaskActionApply>;\n\n/*\ntask - Log\n*/\n\nexport type ExecutionTaskLogApply<\n  LogParam = {\n    content: string;\n  },\n> = ExecutionTaskApply<'Log', LogParam, void, void>;\n\nexport type ExecutionTaskLog = ExecutionTask<ExecutionTaskLogApply>;\n\n/*\ntask - planning\n*/\n\nexport interface ExecutionTaskPlanningParam {\n  userInstruction: string;\n  aiActContext?: string;\n  imagesIncludeCount?: number;\n  deepThink?: DeepThinkOption;\n  subGoalStatus?: string;\n  memoriesStatus?: string;\n}\n\nexport type ExecutionTaskPlanningApply = ExecutionTaskApply<\n  'Planning',\n  ExecutionTaskPlanningParam,\n  PlanningAIResponse\n>;\n\nexport type ExecutionTaskPlanning = ExecutionTask<ExecutionTaskPlanningApply>;\n\n/*\ntask - planning-locate\n*/\nexport type ExecutionTaskPlanningLocateParam = PlanningLocateParam;\n\nexport interface ExecutionTaskPlanningLocateOutput {\n  element: LocateResultElement | null;\n}\n\nexport type ExecutionTaskPlanningDump = ServiceDump;\n\nexport type ExecutionTaskPlanningLocateApply = ExecutionTaskApply<\n  'Planning',\n  ExecutionTaskPlanningLocateParam,\n  ExecutionTaskPlanningLocateOutput,\n  ExecutionTaskPlanningDump\n>;\n\nexport type ExecutionTaskPlanningLocate =\n  ExecutionTask<ExecutionTaskPlanningLocateApply>;\n\n/*\nReport metadata - extracted from ReportActionDump for per-execution writes\n*/\nexport interface ReportMeta {\n  groupName: string;\n  groupDescription?: string;\n  sdkVersion: string;\n  modelBriefs: ModelBrief[];\n  deviceType?: string;\n}\n\n// Backward-compatible aliases for existing external consumers.\nexport type GroupMeta = ReportMeta;\n\n/*\nReport dump\n*/\nexport interface IReportActionDump {\n  sdkVersion: string;\n  groupName: string;\n  groupDescription?: string;\n  modelBriefs: ModelBrief[];\n  executions: IExecutionDump[];\n  deviceType?: string;\n}\n\n// Backward-compatible aliases for existing external consumers.\nexport type IGroupedActionDump = IReportActionDump;\n\nexport interface ModelBrief {\n  /**\n   * The intent/category of the model call, for example \"planning\" or \"insight\".\n   */\n  intent?: string;\n\n  /**\n   * The model name returned by usage metadata, for example \"gpt-4o\".\n   */\n  name?: string;\n\n  /**\n   * Optional human-readable model description, for example \"qwen2.5-vl mode\".\n   */\n  modelDescription?: string;\n}\n\nexport type InterfaceType =\n  | 'puppeteer'\n  | 'playwright'\n  | 'static'\n  | 'chrome-extension-proxy'\n  | 'android'\n  | string;\n\nexport interface StreamingCodeGenerationOptions {\n  /** Whether to enable streaming output */\n  stream?: boolean;\n  /** Callback function to handle streaming chunks */\n  onChunk?: StreamingCallback;\n  /** Callback function to handle streaming completion */\n  onComplete?: (finalCode: string) => void;\n  /** Callback function to handle streaming errors */\n  onError?: (error: Error) => void;\n}\n\nexport type StreamingCallback = (chunk: CodeGenerationChunk) => void;\n\nexport interface CodeGenerationChunk {\n  /** The incremental content chunk */\n  content: string;\n  /** The reasoning content */\n  reasoning_content: string;\n  /** The accumulated content so far */\n  accumulated: string;\n  /** Whether this is the final chunk */\n  isComplete: boolean;\n  /** Token usage information if available */\n  usage?: AIUsageInfo;\n}\n\nexport interface StreamingAIResponse {\n  /** The final accumulated content */\n  content: string;\n  /** Token usage information */\n  usage?: AIUsageInfo;\n  /** Whether the response was streamed */\n  isStreamed: boolean;\n}\n\nexport interface DeviceAction<TParam = any, TReturn = any> {\n  name: string;\n  description?: string;\n  interfaceAlias?: string;\n  paramSchema?: z.ZodType<TParam>;\n  call: (param: TParam, context: ExecutorContext) => Promise<TReturn> | TReturn;\n  delayBeforeRunner?: number;\n  delayAfterRunner?: number;\n  /**\n   * An example param object for this action.\n   * Locate fields with { prompt } may be resolved to internal pixel bboxes when needed.\n   */\n  sample?: { [K in keyof TParam]?: any };\n}\n\n/**\n * Type utilities for extracting types from DeviceAction definitions\n */\n\n/**\n * Extract parameter type from a DeviceAction\n */\nexport type ActionParam<Action extends DeviceAction<any, any>> =\n  Action extends DeviceAction<infer P, any> ? P : never;\n\n/**\n * Extract return type from a DeviceAction\n */\nexport type ActionReturn<Action extends DeviceAction<any, any>> =\n  Action extends DeviceAction<any, infer R> ? R : never;\n\n/**\n * Web-specific types\n */\nexport interface WebElementInfo extends BaseElement {\n  id: string;\n  attributes: {\n    nodeType: NodeType;\n    [key: string]: string;\n  };\n}\n\n/**\n * Agent\n */\n\nexport type CacheConfig = {\n  strategy?: 'read-only' | 'read-write' | 'write-only';\n  id: string;\n  /**\n   * Optional cache directory path.\n   * When set, cache files are written to this directory instead of\n   * `<MIDSCENE_RUN_DIR>/cache`.\n   */\n  cacheDir?: string;\n};\n\nexport type Cache =\n  | false // No read, no write\n  | true // Will throw error at runtime - deprecated\n  | CacheConfig; // Object configuration (requires explicit id)\n\nexport interface AgentOpt {\n  // @deprecated Use `reportFileName` and `cache.id` instead.\n  testId?: string;\n  // @deprecated\n  cacheId?: string; // Keep backward compatibility, but marked as deprecated\n  groupName?: string;\n  groupDescription?: string;\n  /* if auto generate report, default true */\n  generateReport?: boolean;\n  /* if persist per-execution dump files next to the report, default false */\n  persistExecutionDump?: boolean;\n  /* if auto print report msg, default true */\n  autoPrintReportMsg?: boolean;\n\n  /**\n   * Use directory-based report format with separate image files.\n   *\n   * When enabled:\n   * - Screenshots are saved as PNG files in a `screenshots/` subdirectory\n   * - Report is generated as `index.html` with relative image paths\n   * - Reduces memory usage and report file size\n   *\n   * IMPORTANT: 'html-and-external-assets' reports must be served via HTTP server\n   * (e.g., `npx serve ./report-dir`). The file:// protocol will not\n   * work due to browser CORS restrictions.\n   *\n   * @default 'single-html'\n   */\n  outputFormat?: 'single-html' | 'html-and-external-assets';\n\n  onTaskStartTip?: OnTaskStartTip;\n  aiActContext?: string;\n  aiActionContext?: string;\n  /* custom report file name */\n  reportFileName?: string;\n  reportAttributes?: ReportAttributes;\n  modelConfig?: TModelConfig;\n  cache?: Cache;\n  /**\n   * Maximum number of replanning cycles for aiAct.\n   * Defaults are resolved by the active model adapter: 20 for standard planning,\n   * 40 for UI-TARS, and 100 for Auto-GLM.\n   * If omitted, the agent will also read `MIDSCENE_REPLANNING_CYCLE_LIMIT` for backward compatibility.\n   */\n  replanningCycleLimit?: number;\n\n  /**\n   * Wait time in milliseconds after each action execution.\n   * This allows the UI to settle and stabilize before the next action.\n   * Defaults to 300ms when not provided.\n   */\n  waitAfterAction?: number;\n\n  /**\n   * When set to true, Midscene will use the target device's formatted local\n   * time instead of the runtime system time. The target interface must implement\n   * getDeviceLocalTimeString to provide device-local wall-clock time.\n   * Default: false\n   */\n  useDeviceTime?: boolean;\n\n  /**\n   * Custom screenshot shrink factor to reduce AI token usage.\n   * When set, the screenshot will be scaled down by this factor from the physical resolution.\n   *\n   * Example:\n   * - Physical screen width: 3000px, dpr=6\n   * - Logical width: 500px\n   * - screenshotShrinkFactor: 2\n   * - Actual shrunk screenshot width: 3000 / 2 = 1500px\n   * - AI analyzes the 1500px screenshot\n   * - Coordinates are transformed back to logical (500px) before actions execute\n   *\n   * Benefits:\n   * - Reduces token usage for high-resolution screenshots\n   * - Maintains accuracy by scaling coordinates appropriately\n   *\n   * Must be >= 1 (shrinking only, enlarging is not supported).\n   *\n   * @default 1 (no shrinking, uses original physical screenshot)\n   */\n  screenshotShrinkFactor?: number;\n\n  /**\n   * Custom OpenAI client factory function\n   *\n   * If provided, this function will be called to create OpenAI client instances\n   * for each AI call, allowing you to:\n   * - Wrap clients with observability tools (langsmith, langfuse)\n   * - Use custom OpenAI-compatible clients\n   * - Apply different configurations based on intent\n   *\n   * @param config - Resolved model configuration\n   * @returns OpenAI client instance (original or wrapped)\n   *\n   * @example\n   * ```typescript\n   * createOpenAIClient: async (openai, opts) => {\n   *   // Wrap with langsmith for planning tasks\n   *   if (opts.baseURL?.includes('planning')) {\n   *     return wrapOpenAI(openai, { metadata: { task: 'planning' } });\n   *   }\n   *\n   *   return openai;\n   * }\n   * ```\n   */\n  createOpenAIClient?: CreateOpenAIClientFn;\n}\n\nexport type TestStatus =\n  | 'passed'\n  | 'failed'\n  | 'timedOut'\n  | 'skipped'\n  | 'interrupted';\n\nexport interface ReportFileAttributes {\n  testDuration: number;\n  testStatus: TestStatus;\n  testTitle: string;\n  testId: string;\n  testDescription: string;\n}\n\nexport type ReportFileWithAttributes =\n  | {\n      reportFilePath: string;\n      reportAttributes: ReportFileAttributes;\n    }\n  | {\n      reportFilePath?: string;\n      reportAttributes: ReportFileAttributes & { testStatus: 'skipped' };\n    };\n"],"names":["UIContext"],"mappings":";;;AA2HO,MAAeA;AA4BtB"}