@midscene/core 0.26.7-beta-20250821134240.0 → 0.27.0

package/dist/lib/ai-model/service-caller/index.js

@@ -75,7 +75,7 @@ var __webpack_exports__ = {};
     const llm_locator_js_namespaceObject = require("../prompt/llm-locator.js");
     const llm_planning_js_namespaceObject = require("../prompt/llm-planning.js");
     async function createChatClient({ AIActionTypeValue, modelPreferences }) {
-        const { socksProxy, httpProxy, modelName, openaiBaseURL, openaiApiKey, openaiExtraConfig, openaiUseAzureDeprecated, useAzureOpenai, azureOpenaiScope, azureOpenaiKey, azureOpenaiEndpoint, azureOpenaiApiVersion, azureOpenaiDeployment, azureExtraConfig, useAnthropicSdk, anthropicApiKey } = (0, env_namespaceObject.decideModelConfig)(modelPreferences, true);
+        const { socksProxy, httpProxy, modelName, openaiBaseURL, openaiApiKey, openaiExtraConfig, openaiUseAzureDeprecated, useAzureOpenai, azureOpenaiScope, azureOpenaiKey, azureOpenaiEndpoint, azureOpenaiApiVersion, azureOpenaiDeployment, azureExtraConfig, useAnthropicSdk, anthropicApiKey, modelDescription } = (0, env_namespaceObject.decideModelConfig)(modelPreferences, true);
         let openai;
         let proxyAgent;
         const debugProxy = (0, logger_namespaceObject.getDebug)('ai:call:proxy');
@@ -136,7 +136,8 @@ var __webpack_exports__ = {};
         if (void 0 !== openai) return {
             completion: openai.chat.completions,
             style: 'openai',
-            modelName
+            modelName,
+            modelDescription
         };
         if (useAnthropicSdk) openai = new sdk_namespaceObject.Anthropic({
             apiKey: anthropicApiKey,
@@ -146,12 +147,13 @@ var __webpack_exports__ = {};
         if (void 0 !== openai && openai.messages) return {
             completion: openai.messages,
             style: 'anthropic',
-            modelName
+            modelName,
+            modelDescription
         };
         throw new Error('Openai SDK or Anthropic SDK is not initialized');
     }
     async function call(messages, AIActionTypeValue, modelPreferences, options) {
-        const { completion, style, modelName } = await createChatClient({
+        const { completion, style, modelName, modelDescription } = await createChatClient({
             AIActionTypeValue,
             modelPreferences
         });
@@ -222,7 +224,9 @@ var __webpack_exports__ = {};
                                     completion_tokens: usage.completion_tokens ?? 0,
                                     total_tokens: usage.total_tokens ?? 0,
                                     time_cost: timeCost ?? 0,
-                                    model_name: modelName
+                                    model_name: modelName,
+                                    model_description: modelDescription,
+                                    intent: modelPreferences.intent
                                 }
                             };
                             options.onChunk(finalChunk);
@@ -303,7 +307,9 @@ var __webpack_exports__ = {};
                                     completion_tokens: anthropicUsage.output_tokens ?? 0,
                                     total_tokens: (anthropicUsage.input_tokens ?? 0) + (anthropicUsage.output_tokens ?? 0),
                                     time_cost: timeCost ?? 0,
-                                    model_name: modelName
+                                    model_name: modelName,
+                                    model_description: modelDescription,
+                                    intent: modelPreferences.intent
                                 } : void 0
                             };
                             options.onChunk(finalChunk);
@@ -343,7 +349,9 @@ var __webpack_exports__ = {};
                     completion_tokens: usage.completion_tokens ?? 0,
                     total_tokens: usage.total_tokens ?? 0,
                     time_cost: timeCost ?? 0,
-                    model_name: modelName
+                    model_name: modelName,
+                    model_description: modelDescription,
+                    intent: modelPreferences.intent
                 } : void 0,
                 isStreamed: !!isStreaming
             };

package/dist/lib/ai-model/service-caller/index.js.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/service-caller/index.js","sources":["webpack://@midscene/core/webpack/runtime/compat_get_default_export","webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/service-caller/index.ts"],"sourcesContent":["// getDefaultExport function for compatibility with non-ESM modules\n__webpack_require__.n = (module) => {\n\tvar getter = module && module.__esModule ?\n\t\t() => (module['default']) :\n\t\t() => (module);\n\t__webpack_require__.d(getter, { a: getter });\n\treturn getter;\n};\n","__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import { AIResponseFormat, type AIUsageInfo } from '@/types';\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\nimport { Anthropic } from '@anthropic-ai/sdk';\nimport {\n  DefaultAzureCredential,\n  getBearerTokenProvider,\n} from '@azure/identity';\nimport {\n  type IModelPreferences,\n  MIDSCENE_API_TYPE,\n  MIDSCENE_LANGSMITH_DEBUG,\n  OPENAI_MAX_TOKENS,\n  decideModelConfig,\n  getAIConfig,\n  getAIConfigInBoolean,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { parseBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { ifInBrowser } from '@midscene/shared/utils';\nimport { HttpsProxyAgent } from 'https-proxy-agent';\nimport { jsonrepair } from 'jsonrepair';\nimport OpenAI, { AzureOpenAI } from 'openai';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { Stream } from 'openai/streaming';\nimport { SocksProxyAgent } from 'socks-proxy-agent';\nimport { AIActionType, type AIArgs } from '../common';\nimport { assertSchema } from '../prompt/assertion';\nimport { locatorSchema } from '../prompt/llm-locator';\nimport { planSchema } from '../prompt/llm-planning';\n\nasync function createChatClient({\n  AIActionTypeValue,\n  modelPreferences,\n}: {\n  AIActionTypeValue: AIActionType;\n  modelPreferences: IModelPreferences;\n}): Promise<{\n  completion: OpenAI.Chat.Completions;\n  style: 'openai' | 'anthropic';\n  modelName: string;\n}> {\n  const {\n    socksProxy,\n    httpProxy,\n    modelName,\n    openaiBaseURL,\n    openaiApiKey,\n    openaiExtraConfig,\n    openaiUseAzureDeprecated,\n    useAzureOpenai,\n    azureOpenaiScope,\n    azureOpenaiKey,\n    azureOpenaiEndpoint,\n    azureOpenaiApiVersion,\n    azureOpenaiDeployment,\n    azureExtraConfig,\n    useAnthropicSdk,\n    anthropicApiKey,\n  } = decideModelConfig(modelPreferences, true);\n\n  let openai: OpenAI | AzureOpenAI | undefined;\n\n  let proxyAgent = undefined;\n  const debugProxy = getDebug('ai:call:proxy');\n  if (httpProxy) {\n    debugProxy('using http proxy', httpProxy);\n    proxyAgent = new HttpsProxyAgent(httpProxy);\n  } else if (socksProxy) {\n    debugProxy('using socks proxy', socksProxy);\n    proxyAgent = new SocksProxyAgent(socksProxy);\n  }\n\n  if (openaiUseAzureDeprecated) {\n    // this is deprecated\n    openai = new AzureOpenAI({\n      baseURL: openaiBaseURL,\n      apiKey: openaiApiKey,\n      httpAgent: proxyAgent,\n      ...openaiExtraConfig,\n      dangerouslyAllowBrowser: true,\n    }) as OpenAI;\n  } else if (useAzureOpenai) {\n    // https://learn.microsoft.com/en-us/azure/ai-services/openai/chatgpt-quickstart?tabs=bash%2Cjavascript-key%2Ctypescript-keyless%2Cpython&pivots=programming-language-javascript#rest-api\n    // keyless authentication\n    let tokenProvider: any = undefined;\n    if (azureOpenaiScope) {\n      assert(\n        !ifInBrowser,\n        'Azure OpenAI is not supported in browser with Midscene.',\n      );\n      const credential = new DefaultAzureCredential();\n\n      tokenProvider = getBearerTokenProvider(credential, azureOpenaiScope);\n\n      openai = new AzureOpenAI({\n        azureADTokenProvider: tokenProvider,\n        endpoint: azureOpenaiEndpoint,\n        apiVersion: azureOpenaiApiVersion,\n        deployment: azureOpenaiDeployment,\n        ...openaiExtraConfig,\n        ...azureExtraConfig,\n      });\n    } else {\n      // endpoint, apiKey, apiVersion, deployment\n      openai = new AzureOpenAI({\n        apiKey: azureOpenaiKey,\n        endpoint: azureOpenaiEndpoint,\n        apiVersion: azureOpenaiApiVersion,\n        deployment: azureOpenaiDeployment,\n        dangerouslyAllowBrowser: true,\n        ...openaiExtraConfig,\n        ...azureExtraConfig,\n      });\n    }\n  } else if (!useAnthropicSdk) {\n    openai = new OpenAI({\n      baseURL: openaiBaseURL,\n      apiKey: openaiApiKey,\n      httpAgent: proxyAgent,\n      ...openaiExtraConfig,\n      defaultHeaders: {\n        ...(openaiExtraConfig?.defaultHeaders || {}),\n        [MIDSCENE_API_TYPE]: AIActionTypeValue.toString(),\n      },\n      dangerouslyAllowBrowser: true,\n    });\n  }\n\n  if (openai && getAIConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)) {\n    if (ifInBrowser) {\n      throw new Error('langsmith is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langsmith wrapper enabled');\n    const { wrapOpenAI } = await import('langsmith/wrappers');\n    openai = wrapOpenAI(openai);\n  }\n\n  if (typeof openai !== 'undefined') {\n    return {\n      completion: openai.chat.completions,\n      style: 'openai',\n      modelName,\n    };\n  }\n\n  // Anthropic\n  if (useAnthropicSdk) {\n    openai = new Anthropic({\n      apiKey: anthropicApiKey,\n      httpAgent: proxyAgent,\n      dangerouslyAllowBrowser: true,\n    }) as any;\n  }\n\n  if (typeof openai !== 'undefined' && (openai as any).messages) {\n    return {\n      completion: (openai as any).messages,\n      style: 'anthropic',\n      modelName,\n    };\n  }\n\n  throw new Error('Openai SDK or Anthropic SDK is not initialized');\n}\n\nexport async function call(\n  messages: ChatCompletionMessageParam[],\n  AIActionTypeValue: AIActionType,\n  modelPreferences: IModelPreferences,\n  options?: {\n    stream?: boolean;\n    onChunk?: StreamingCallback;\n  },\n): Promise<{ content: string; usage?: AIUsageInfo; isStreamed: boolean }> {\n  const { completion, style, modelName } = await createChatClient({\n    AIActionTypeValue,\n    modelPreferences,\n  });\n\n  const responseFormat = getResponseFormat(modelName, AIActionTypeValue);\n\n  const maxTokens = getAIConfig(OPENAI_MAX_TOKENS);\n  const debugCall = getDebug('ai:call');\n  const debugProfileStats = getDebug('ai:profile:stats');\n  const debugProfileDetail = getDebug('ai:profile:detail');\n\n  const startTime = Date.now();\n\n  const isStreaming = options?.stream && options?.onChunk;\n  let content: string | undefined;\n  let accumulated = '';\n  let usage: OpenAI.CompletionUsage | undefined;\n  let timeCost: number | undefined;\n\n  const commonConfig = {\n    temperature: vlLocateMode(modelPreferences) === 'vlm-ui-tars' ? 0.0 : 0.1,\n    stream: !!isStreaming,\n    max_tokens:\n      typeof maxTokens === 'number'\n        ? maxTokens\n        : Number.parseInt(maxTokens || '2048', 10),\n    ...(vlLocateMode(modelPreferences) === 'qwen-vl' // qwen specific config\n      ? {\n          vl_high_resolution_images: true,\n        }\n      : {}),\n  };\n\n  try {\n    if (style === 'openai') {\n      debugCall(\n        `sending ${isStreaming ? 'streaming ' : ''}request to ${modelName}`,\n      );\n\n      if (isStreaming) {\n        const stream = (await completion.create(\n          {\n            model: modelName,\n            messages,\n            response_format: responseFormat,\n            ...commonConfig,\n          },\n          {\n            stream: true,\n          },\n        )) as Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {\n          _request_id?: string | null;\n        };\n\n        for await (const chunk of stream) {\n          const content = chunk.choices?.[0]?.delta?.content || '';\n          const reasoning_content =\n            (chunk.choices?.[0]?.delta as any)?.reasoning_content || '';\n\n          // Check for usage info in any chunk (OpenAI provides usage in separate chunks)\n          if (chunk.usage) {\n            usage = chunk.usage;\n          }\n\n          if (content || reasoning_content) {\n            accumulated += content;\n            const chunkData: CodeGenerationChunk = {\n              content,\n              reasoning_content,\n              accumulated,\n              isComplete: false,\n              usage: undefined,\n            };\n            options.onChunk!(chunkData);\n          }\n\n          // Check if stream is complete\n          if (chunk.choices?.[0]?.finish_reason) {\n            timeCost = Date.now() - startTime;\n\n            // If usage is not available from the stream, provide a basic usage info\n            if (!usage) {\n              // Estimate token counts based on content length (rough approximation)\n              const estimatedTokens = Math.max(\n                1,\n                Math.floor(accumulated.length / 4),\n              );\n              usage = {\n                prompt_tokens: estimatedTokens,\n                completion_tokens: estimatedTokens,\n                total_tokens: estimatedTokens * 2,\n              };\n            }\n\n            // Send final chunk\n            const finalChunk: CodeGenerationChunk = {\n              content: '',\n              accumulated,\n              reasoning_content: '',\n              isComplete: true,\n              usage: {\n                prompt_tokens: usage.prompt_tokens ?? 0,\n                completion_tokens: usage.completion_tokens ?? 0,\n                total_tokens: usage.total_tokens ?? 0,\n                time_cost: timeCost ?? 0,\n                model_name: modelName,\n              },\n            };\n            options.onChunk!(finalChunk);\n            break;\n          }\n        }\n        content = accumulated;\n        debugProfileStats(\n          `streaming model, ${modelName}, mode, ${vlLocateMode(modelPreferences) || 'default'}, cost-ms, ${timeCost}`,\n        );\n      } else {\n        const result = await completion.create({\n          model: modelName,\n          messages,\n          response_format: responseFormat,\n          ...commonConfig,\n        } as any);\n        timeCost = Date.now() - startTime;\n\n        debugProfileStats(\n          `model, ${modelName}, mode, ${vlLocateMode(modelPreferences) || 'default'}, ui-tars-version, ${uiTarsModelVersion(modelPreferences)}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}`,\n        );\n\n        debugProfileDetail(\n          `model usage detail: ${JSON.stringify(result.usage)}`,\n        );\n\n        assert(\n          result.choices,\n          `invalid response from LLM service: ${JSON.stringify(result)}`,\n        );\n        content = result.choices[0].message.content!;\n        usage = result.usage;\n      }\n\n      debugCall(`response: ${content}`);\n      assert(content, 'empty content');\n    } else if (style === 'anthropic') {\n      const convertImageContent = (content: any) => {\n        if (content.type === 'image_url') {\n          const imgBase64 = content.image_url.url;\n          assert(imgBase64, 'image_url is required');\n          const { mimeType, body } = parseBase64(content.image_url.url);\n          return {\n            source: {\n              type: 'base64',\n              media_type: mimeType,\n              data: body,\n            },\n            type: 'image',\n          };\n        }\n        return content;\n      };\n\n      if (isStreaming) {\n        const stream = (await completion.create({\n          model: modelName,\n          system: 'You are a versatile professional in software UI automation',\n          messages: messages.map((m) => ({\n            role: 'user',\n            content: Array.isArray(m.content)\n              ? (m.content as any).map(convertImageContent)\n              : m.content,\n          })),\n          response_format: responseFormat,\n          ...commonConfig,\n        } as any)) as any;\n\n        for await (const chunk of stream) {\n          const content = chunk.delta?.text || '';\n          if (content) {\n            accumulated += content;\n            const chunkData: CodeGenerationChunk = {\n              content,\n              accumulated,\n              reasoning_content: '',\n              isComplete: false,\n              usage: undefined,\n            };\n            options.onChunk!(chunkData);\n          }\n\n          // Check if stream is complete\n          if (chunk.type === 'message_stop') {\n            timeCost = Date.now() - startTime;\n            const anthropicUsage = chunk.usage;\n\n            // Send final chunk\n            const finalChunk: CodeGenerationChunk = {\n              content: '',\n              accumulated,\n              reasoning_content: '',\n              isComplete: true,\n              usage: anthropicUsage\n                ? {\n                    prompt_tokens: anthropicUsage.input_tokens ?? 0,\n                    completion_tokens: anthropicUsage.output_tokens ?? 0,\n                    total_tokens:\n                      (anthropicUsage.input_tokens ?? 0) +\n                      (anthropicUsage.output_tokens ?? 0),\n                    time_cost: timeCost ?? 0,\n                    model_name: modelName,\n                  }\n                : undefined,\n            };\n            options.onChunk!(finalChunk);\n            break;\n          }\n        }\n        content = accumulated;\n      } else {\n        const result = await completion.create({\n          model: modelName,\n          system: 'You are a versatile professional in software UI automation',\n          messages: messages.map((m) => ({\n            role: 'user',\n            content: Array.isArray(m.content)\n              ? (m.content as any).map(convertImageContent)\n              : m.content,\n          })),\n          response_format: responseFormat,\n          ...commonConfig,\n        } as any);\n        timeCost = Date.now() - startTime;\n        content = (result as any).content[0].text as string;\n        usage = result.usage;\n      }\n\n      assert(content, 'empty content');\n    }\n    // Ensure we always have usage info for streaming responses\n    if (isStreaming && !usage) {\n      // Estimate token counts based on content length (rough approximation)\n      const estimatedTokens = Math.max(\n        1,\n        Math.floor((content || '').length / 4),\n      );\n      usage = {\n        prompt_tokens: estimatedTokens,\n        completion_tokens: estimatedTokens,\n        total_tokens: estimatedTokens * 2,\n      };\n    }\n\n    return {\n      content: content || '',\n      usage: usage\n        ? {\n            prompt_tokens: usage.prompt_tokens ?? 0,\n            completion_tokens: usage.completion_tokens ?? 0,\n            total_tokens: usage.total_tokens ?? 0,\n            time_cost: timeCost ?? 0,\n            model_name: modelName,\n          }\n        : undefined,\n      isStreamed: !!isStreaming,\n    };\n  } catch (e: any) {\n    console.error(' call AI error', e);\n    const newError = new Error(\n      `failed to call ${isStreaming ? 'streaming ' : ''}AI model service: ${e.message}. Trouble shooting: https://midscenejs.com/model-provider.html`,\n      {\n        cause: e,\n      },\n    );\n    throw newError;\n  }\n}\n\nexport const getResponseFormat = (\n  modelName: string,\n  AIActionTypeValue: AIActionType,\n):\n  | OpenAI.ChatCompletionCreateParams['response_format']\n  | OpenAI.ResponseFormatJSONObject => {\n  let responseFormat:\n    | OpenAI.ChatCompletionCreateParams['response_format']\n    | OpenAI.ResponseFormatJSONObject\n    | undefined;\n\n  if (modelName.includes('gpt-4')) {\n    switch (AIActionTypeValue) {\n      case AIActionType.ASSERT:\n        responseFormat = assertSchema;\n        break;\n      case AIActionType.INSPECT_ELEMENT:\n        responseFormat = locatorSchema;\n        break;\n      case AIActionType.PLAN:\n        responseFormat = planSchema;\n        break;\n      case AIActionType.EXTRACT_DATA:\n      case AIActionType.DESCRIBE_ELEMENT:\n        responseFormat = { type: AIResponseFormat.JSON };\n        break;\n    }\n  }\n\n  // gpt-4o-2024-05-13 only supports json_object response format\n  if (modelName === 'gpt-4o-2024-05-13') {\n    responseFormat = { type: AIResponseFormat.JSON };\n  }\n\n  return responseFormat;\n};\n\nexport async function callToGetJSONObject<T>(\n  messages: ChatCompletionMessageParam[],\n  AIActionTypeValue: AIActionType,\n  modelPreferences: IModelPreferences,\n): Promise<{ content: T; usage?: AIUsageInfo }> {\n  const response = await call(messages, AIActionTypeValue, modelPreferences);\n  assert(response, 'empty response');\n  const jsonContent = safeParseJson(response.content, modelPreferences);\n  return { content: jsonContent, usage: response.usage };\n}\n\nexport async function callAiFnWithStringResponse<T>(\n  msgs: AIArgs,\n  AIActionTypeValue: AIActionType,\n  modelPreferences: IModelPreferences,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n  const { content, usage } = await call(\n    msgs,\n    AIActionTypeValue,\n    modelPreferences,\n  );\n  return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n  try {\n    // First, try to match a JSON object directly in the response\n    const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\n    if (jsonMatch) {\n      return jsonMatch[1];\n    }\n\n    // If no direct JSON object is found, try to extract JSON from a code block\n    const codeBlockMatch = response.match(\n      /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\n    );\n    if (codeBlockMatch) {\n      return codeBlockMatch[1];\n    }\n\n    // If no code block is found, try to find a JSON-like structure in the text\n    const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\n    if (jsonLikeMatch) {\n      return jsonLikeMatch[0];\n    }\n  } catch {}\n  // If no JSON-like structure is found, return the original response\n  return response;\n}\n\nexport function preprocessDoubaoBboxJson(input: string) {\n  if (input.includes('bbox')) {\n    // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\n    while (/\\d+\\s+\\d+/.test(input)) {\n      input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n    }\n  }\n  return input;\n}\n\nexport function safeParseJson(\n  input: string,\n  modelPreferences: IModelPreferences,\n) {\n  const cleanJsonString = extractJSONFromCodeBlock(input);\n  // match the point\n  if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\n    return cleanJsonString\n      .match(/\\((\\d+),(\\d+)\\)/)\n      ?.slice(1)\n      .map(Number);\n  }\n  try {\n    return JSON.parse(cleanJsonString);\n  } catch {}\n  try {\n    return JSON.parse(jsonrepair(cleanJsonString));\n  } catch (e) {}\n\n  if (\n    vlLocateMode(modelPreferences) === 'doubao-vision' ||\n    vlLocateMode(modelPreferences) === 'vlm-ui-tars'\n  ) {\n    const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\n    return JSON.parse(jsonrepair(jsonString));\n  }\n  throw Error(`failed to parse json response: ${input}`);\n}\n"],"names":["__webpack_require__","module","getter","definition","key","Object","obj","prop","Symbol","createChatClient","AIActionTypeValue","modelPreferences","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","openaiUseAzureDeprecated","useAzureOpenai","azureOpenaiScope","azureOpenaiKey","azureOpenaiEndpoint","azureOpenaiApiVersion","azureOpenaiDeployment","azureExtraConfig","useAnthropicSdk","anthropicApiKey","decideModelConfig","openai","proxyAgent","debugProxy","getDebug","HttpsProxyAgent","SocksProxyAgent","AzureOpenAI","tokenProvider","assert","ifInBrowser","credential","DefaultAzureCredential","getBearerTokenProvider","OpenAI","MIDSCENE_API_TYPE","getAIConfigInBoolean","MIDSCENE_LANGSMITH_DEBUG","Error","console","wrapOpenAI","Anthropic","call","messages","options","completion","style","responseFormat","getResponseFormat","maxTokens","getAIConfig","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","isStreaming","content","accumulated","usage","timeCost","commonConfig","vlLocateMode","Number","stream","chunk","_chunk_choices__delta","_chunk_choices__delta1","_chunk_choices_2","reasoning_content","chunkData","undefined","estimatedTokens","Math","finalChunk","_result_usage","_result_usage1","_result_usage2","result","uiTarsModelVersion","JSON","convertImageContent","imgBase64","mimeType","body","parseBase64","m","Array","_chunk_delta","anthropicUsage","e","newError","AIActionType","assertSchema","locatorSchema","planSchema","AIResponseFormat","callToGetJSONObject","response","jsonContent","safeParseJson","callAiFnWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","cleanJsonString","_cleanJsonString_match","jsonrepair","jsonString"],"mappings":";;;;;;;;;;;;;;;;;;;IACAA,oBAAoB,CAAC,GAAG,CAACC;QACxB,IAAIC,SAASD,UAAUA,OAAO,UAAU,GACvC,IAAOA,MAAM,CAAC,UAAU,GACxB,IAAOA;QACRD,oBAAoB,CAAC,CAACE,QAAQ;YAAE,GAAGA;QAAO;QAC1C,OAAOA;IACR;;;ICPAF,oBAAoB,CAAC,GAAG,CAAC,UAASG;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGH,oBAAoB,CAAC,CAACG,YAAYC,QAAQ,CAACJ,oBAAoB,CAAC,CAAC,UAASI,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAJ,oBAAoB,CAAC,GAAG,CAACM,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFP,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOQ,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;IC2BA,eAAeI,iBAAiB,EAC9BC,iBAAiB,EACjBC,gBAAgB,EAIjB;QAKC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,wBAAwB,EACxBC,cAAc,EACdC,gBAAgB,EAChBC,cAAc,EACdC,mBAAmB,EACnBC,qBAAqB,EACrBC,qBAAqB,EACrBC,gBAAgB,EAChBC,eAAe,EACfC,eAAe,EAChB,GAAGC,AAAAA,IAAAA,oBAAAA,iBAAAA,AAAAA,EAAkBjB,kBAAkB;QAExC,IAAIkB;QAEJ,IAAIC;QACJ,MAAMC,aAAaC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QAC5B,IAAInB,WAAW;YACbkB,WAAW,oBAAoBlB;YAC/BiB,aAAa,IAAIG,2CAAAA,eAAeA,CAACpB;QACnC,OAAO,IAAID,YAAY;YACrBmB,WAAW,qBAAqBnB;YAChCkB,aAAa,IAAII,2CAAAA,eAAeA,CAACtB;QACnC;QAEA,IAAIM,0BAEFW,SAAS,IAAIM,gCAAAA,WAAWA,CAAC;YACvB,SAASpB;YACT,QAAQC;YACR,WAAWc;YACX,GAAGb,iBAAiB;YACpB,yBAAyB;QAC3B;aACK,IAAIE,gBAAgB;YAGzB,IAAIiB;YACJ,IAAIhB,kBAAkB;gBACpBiB,IAAAA,sBAAAA,MAAAA,AAAAA,EACE,CAACC,sBAAAA,WAAWA,EACZ;gBAEF,MAAMC,aAAa,IAAIC,yBAAAA,sBAAsBA;gBAE7CJ,gBAAgBK,AAAAA,IAAAA,yBAAAA,sBAAAA,AAAAA,EAAuBF,YAAYnB;gBAEnDS,SAAS,IAAIM,gCAAAA,WAAWA,CAAC;oBACvB,sBAAsBC;oBACtB,UAAUd;oBACV,YAAYC;oBACZ,YAAYC;oBACZ,GAAGP,iBAAiB;oBACpB,GAAGQ,gBAAgB;gBACrB;YACF,OAEEI,SAAS,IAAIM,gCAAAA,WAAWA,CAAC;gBACvB,QAAQd;gBACR,UAAUC;gBACV,YAAYC;gBACZ,YAAYC;gBACZ,yBAAyB;gBACzB,GAAGP,iBAAiB;gBACpB,GAAGQ,gBAAgB;YACrB;QAEJ,OAAO,IAAI,CAACC,iBACVG,SAAS,IAAIa,CAAAA,yBAAAA,EAAO;YAClB,SAAS3B;YACT,QAAQC;YACR,WAAWc;YACX,GAAGb,iBAAiB;YACpB,gBAAgB;gBACd,GAAIA,AAAAA,CAAAA,QAAAA,oBAAAA,KAAAA,IAAAA,kBAAmB,cAAc,AAAD,KAAK,CAAC,CAAC;gBAC3C,CAAC0B,oBAAAA,iBAAiBA,CAAC,EAAEjC,kBAAkB,QAAQ;YACjD;YACA,yBAAyB;QAC3B;QAGF,IAAImB,UAAUe,AAAAA,IAAAA,oBAAAA,oBAAAA,AAAAA,EAAqBC,oBAAAA,wBAAwBA,GAAG;YAC5D,IAAIP,sBAAAA,WAAWA,EACb,MAAM,IAAIQ,MAAM;YAElBC,QAAQ,GAAG,CAAC;YACZ,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM;YAC7BnB,SAASmB,WAAWnB;QACtB;QAEA,IAAI,AAAkB,WAAXA,QACT,OAAO;YACL,YAAYA,OAAO,IAAI,CAAC,WAAW;YACnC,OAAO;YACPf;QACF;QAIF,IAAIY,iBACFG,SAAS,IAAIoB,oBAAAA,SAASA,CAAC;YACrB,QAAQtB;YACR,WAAWG;YACX,yBAAyB;QAC3B;QAGF,IAAI,AAAkB,WAAXD,UAA2BA,OAAe,QAAQ,EAC3D,OAAO;YACL,YAAaA,OAAe,QAAQ;YACpC,OAAO;YACPf;QACF;QAGF,MAAM,IAAIgC,MAAM;IAClB;IAEO,eAAeI,KACpBC,QAAsC,EACtCzC,iBAA+B,EAC/BC,gBAAmC,EACnCyC,OAGC;QAED,MAAM,EAAEC,UAAU,EAAEC,KAAK,EAAExC,SAAS,EAAE,GAAG,MAAML,iBAAiB;YAC9DC;YACAC;QACF;QAEA,MAAM4C,iBAAiBC,kBAAkB1C,WAAWJ;QAEpD,MAAM+C,YAAYC,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYC,oBAAAA,iBAAiBA;QAC/C,MAAMC,YAAY5B,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QAC3B,MAAM6B,oBAAoB7B,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QACnC,MAAM8B,qBAAqB9B,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QAEpC,MAAM+B,YAAYC,KAAK,GAAG;QAE1B,MAAMC,cAAcb,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM,AAAD,KAAKA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO,AAAD;QACtD,IAAIc;QACJ,IAAIC,cAAc;QAClB,IAAIC;QACJ,IAAIC;QAEJ,MAAMC,eAAe;YACnB,aAAaC,AAAmC,kBAAnCA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa5D,oBAAsC,MAAM;YACtE,QAAQ,CAAC,CAACsD;YACV,YACE,AAAqB,YAArB,OAAOR,YACHA,YACAe,OAAO,QAAQ,CAACf,aAAa,QAAQ;YAC3C,GAAIc,AAAmC,cAAnCA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa5D,oBACb;gBACE,2BAA2B;YAC7B,IACA,CAAC,CAAC;QACR;QAEA,IAAI;YACF,IAAI2C,AAAU,aAAVA,OAAoB;gBACtBM,UACE,CAAC,QAAQ,EAAEK,cAAc,eAAe,GAAG,WAAW,EAAEnD,WAAW;gBAGrE,IAAImD,aAAa;oBACf,MAAMQ,SAAU,MAAMpB,WAAW,MAAM,CACrC;wBACE,OAAOvC;wBACPqC;wBACA,iBAAiBI;wBACjB,GAAGe,YAAY;oBACjB,GACA;wBACE,QAAQ;oBACV;oBAKF,WAAW,MAAMI,SAASD,OAAQ;4BAChBE,uBAAAA,iBAAAA,gBAEbC,wBAAAA,kBAAAA,iBAoBCC,kBAAAA;wBAtBJ,MAAMX,UAAUS,AAAAA,SAAAA,CAAAA,iBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,kBAAAA,cAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,QAAAA,CAAAA,wBAAAA,gBAAoB,KAAK,AAAD,IAAxBA,KAAAA,IAAAA,sBAA2B,OAAO,AAAD,KAAK;wBACtD,MAAMG,oBACJ,AAAC,SAAAF,CAAAA,kBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,mBAAAA,eAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,QAAAA,CAAAA,yBAAAA,iBAAoB,KAAK,AAAD,IAAxBA,KAAAA,IAAAA,uBAAmC,iBAAiB,AAAD,KAAK;wBAG3D,IAAIF,MAAM,KAAK,EACbN,QAAQM,MAAM,KAAK;wBAGrB,IAAIR,WAAWY,mBAAmB;4BAChCX,eAAeD;4BACf,MAAMa,YAAiC;gCACrCb;gCACAY;gCACAX;gCACA,YAAY;gCACZ,OAAOa;4BACT;4BACA5B,QAAQ,OAAO,CAAE2B;wBACnB;wBAGA,IAAI,QAAAF,CAAAA,kBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,mBAAAA,eAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,iBAAoB,aAAa,EAAE;4BACrCR,WAAWL,KAAK,GAAG,KAAKD;4BAGxB,IAAI,CAACK,OAAO;gCAEV,MAAMa,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAACf,YAAY,MAAM,GAAG;gCAElCC,QAAQ;oCACN,eAAea;oCACf,mBAAmBA;oCACnB,cAAcA,AAAkB,IAAlBA;gCAChB;4BACF;4BAGA,MAAME,aAAkC;gCACtC,SAAS;gCACThB;gCACA,mBAAmB;gCACnB,YAAY;gCACZ,OAAO;oCACL,eAAeC,MAAM,aAAa,IAAI;oCACtC,mBAAmBA,MAAM,iBAAiB,IAAI;oCAC9C,cAAcA,MAAM,YAAY,IAAI;oCACpC,WAAWC,YAAY;oCACvB,YAAYvD;gCACd;4BACF;4BACAsC,QAAQ,OAAO,CAAE+B;4BACjB;wBACF;oBACF;oBACAjB,UAAUC;oBACVN,kBACE,CAAC,iBAAiB,EAAE/C,UAAU,QAAQ,EAAEyD,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa5D,qBAAqB,UAAU,WAAW,EAAE0D,UAAU;gBAE/G,OAAO;wBAUoJe,eAAyDC,gBAAwDC;oBAT1Q,MAAMC,SAAS,MAAMlC,WAAW,MAAM,CAAC;wBACrC,OAAOvC;wBACPqC;wBACA,iBAAiBI;wBACjB,GAAGe,YAAY;oBACjB;oBACAD,WAAWL,KAAK,GAAG,KAAKD;oBAExBF,kBACE,CAAC,OAAO,EAAE/C,UAAU,QAAQ,EAAEyD,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa5D,qBAAqB,UAAU,mBAAmB,EAAE6E,AAAAA,IAAAA,oBAAAA,kBAAAA,AAAAA,EAAmB7E,kBAAkB,iBAAiB,EAAEyE,AAAAA,SAAAA,CAAAA,gBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,cAAc,aAAa,AAAD,KAAK,GAAG,qBAAqB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,iBAAiB,AAAD,KAAK,GAAG,gBAAgB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,YAAY,AAAD,KAAK,GAAG,WAAW,EAAEjB,SAAS,aAAa,EAAEkB,OAAO,WAAW,IAAI,IAAI;oBAG1WzB,mBACE,CAAC,oBAAoB,EAAE2B,KAAK,SAAS,CAACF,OAAO,KAAK,GAAG;oBAGvDlD,IAAAA,sBAAAA,MAAAA,AAAAA,EACEkD,OAAO,OAAO,EACd,CAAC,mCAAmC,EAAEE,KAAK,SAAS,CAACF,SAAS;oBAEhErB,UAAUqB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;oBAC3CnB,QAAQmB,OAAO,KAAK;gBACtB;gBAEA3B,UAAU,CAAC,UAAU,EAAEM,SAAS;gBAChC7B,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO6B,SAAS;YAClB,OAAO,IAAIZ,AAAU,gBAAVA,OAAuB;gBAChC,MAAMoC,sBAAsB,CAACxB;oBAC3B,IAAIA,AAAiB,gBAAjBA,QAAQ,IAAI,EAAkB;wBAChC,MAAMyB,YAAYzB,QAAQ,SAAS,CAAC,GAAG;wBACvC7B,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOsD,WAAW;wBAClB,MAAM,EAAEC,QAAQ,EAAEC,IAAI,EAAE,GAAGC,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAY5B,QAAQ,SAAS,CAAC,GAAG;wBAC5D,OAAO;4BACL,QAAQ;gCACN,MAAM;gCACN,YAAY0B;gCACZ,MAAMC;4BACR;4BACA,MAAM;wBACR;oBACF;oBACA,OAAO3B;gBACT;gBAEA,IAAID,aAAa;oBACf,MAAMQ,SAAU,MAAMpB,WAAW,MAAM,CAAC;wBACtC,OAAOvC;wBACP,QAAQ;wBACR,UAAUqC,SAAS,GAAG,CAAC,CAAC4C,IAAO;gCAC7B,MAAM;gCACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;4BACf;wBACA,iBAAiBxC;wBACjB,GAAGe,YAAY;oBACjB;oBAEA,WAAW,MAAMI,SAASD,OAAQ;4BAChBwB;wBAAhB,MAAM/B,UAAU+B,AAAAA,SAAAA,CAAAA,eAAAA,MAAM,KAAK,AAAD,IAAVA,KAAAA,IAAAA,aAAa,IAAI,AAAD,KAAK;wBACrC,IAAI/B,SAAS;4BACXC,eAAeD;4BACf,MAAMa,YAAiC;gCACrCb;gCACAC;gCACA,mBAAmB;gCACnB,YAAY;gCACZ,OAAOa;4BACT;4BACA5B,QAAQ,OAAO,CAAE2B;wBACnB;wBAGA,IAAIL,AAAe,mBAAfA,MAAM,IAAI,EAAqB;4BACjCL,WAAWL,KAAK,GAAG,KAAKD;4BACxB,MAAMmC,iBAAiBxB,MAAM,KAAK;4BAGlC,MAAMS,aAAkC;gCACtC,SAAS;gCACThB;gCACA,mBAAmB;gCACnB,YAAY;gCACZ,OAAO+B,iBACH;oCACE,eAAeA,eAAe,YAAY,IAAI;oCAC9C,mBAAmBA,eAAe,aAAa,IAAI;oCACnD,cACGA,AAAAA,CAAAA,eAAe,YAAY,IAAI,KAC/BA,CAAAA,eAAe,aAAa,IAAI;oCACnC,WAAW7B,YAAY;oCACvB,YAAYvD;gCACd,IACAkE;4BACN;4BACA5B,QAAQ,OAAO,CAAE+B;4BACjB;wBACF;oBACF;oBACAjB,UAAUC;gBACZ,OAAO;oBACL,MAAMoB,SAAS,MAAMlC,WAAW,MAAM,CAAC;wBACrC,OAAOvC;wBACP,QAAQ;wBACR,UAAUqC,SAAS,GAAG,CAAC,CAAC4C,IAAO;gCAC7B,MAAM;gCACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;4BACf;wBACA,iBAAiBxC;wBACjB,GAAGe,YAAY;oBACjB;oBACAD,WAAWL,KAAK,GAAG,KAAKD;oBACxBG,UAAWqB,OAAe,OAAO,CAAC,EAAE,CAAC,IAAI;oBACzCnB,QAAQmB,OAAO,KAAK;gBACtB;gBAEAlD,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO6B,SAAS;YAClB;YAEA,IAAID,eAAe,CAACG,OAAO;gBAEzB,MAAMa,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEhB,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;gBAEtCE,QAAQ;oBACN,eAAea;oBACf,mBAAmBA;oBACnB,cAAcA,AAAkB,IAAlBA;gBAChB;YACF;YAEA,OAAO;gBACL,SAASf,WAAW;gBACpB,OAAOE,QACH;oBACE,eAAeA,MAAM,aAAa,IAAI;oBACtC,mBAAmBA,MAAM,iBAAiB,IAAI;oBAC9C,cAAcA,MAAM,YAAY,IAAI;oBACpC,WAAWC,YAAY;oBACvB,YAAYvD;gBACd,IACAkE;gBACJ,YAAY,CAAC,CAACf;YAChB;QACF,EAAE,OAAOkC,GAAQ;YACfpD,QAAQ,KAAK,CAAC,kBAAkBoD;YAChC,MAAMC,WAAW,IAAItD,MACnB,CAAC,eAAe,EAAEmB,cAAc,eAAe,GAAG,kBAAkB,EAAEkC,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC/I;gBACE,OAAOA;YACT;YAEF,MAAMC;QACR;IACF;IAEO,MAAM5C,oBAAoB,CAC/B1C,WACAJ;QAIA,IAAI6C;QAKJ,IAAIzC,UAAU,QAAQ,CAAC,UACrB,OAAQJ;YACN,KAAK2F,mCAAAA,YAAAA,CAAAA,MAAmB;gBACtB9C,iBAAiB+C,6BAAAA,YAAYA;gBAC7B;YACF,KAAKD,mCAAAA,YAAAA,CAAAA,eAA4B;gBAC/B9C,iBAAiBgD,+BAAAA,aAAaA;gBAC9B;YACF,KAAKF,mCAAAA,YAAAA,CAAAA,IAAiB;gBACpB9C,iBAAiBiD,gCAAAA,UAAUA;gBAC3B;YACF,KAAKH,mCAAAA,YAAAA,CAAAA,YAAyB;YAC9B,KAAKA,mCAAAA,YAAAA,CAAAA,gBAA6B;gBAChC9C,iBAAiB;oBAAE,MAAMkD,kCAAAA,gBAAAA,CAAAA,IAAqB;gBAAC;gBAC/C;QACJ;QAIF,IAAI3F,AAAc,wBAAdA,WACFyC,iBAAiB;YAAE,MAAMkD,kCAAAA,gBAAAA,CAAAA,IAAqB;QAAC;QAGjD,OAAOlD;IACT;IAEO,eAAemD,oBACpBvD,QAAsC,EACtCzC,iBAA+B,EAC/BC,gBAAmC;QAEnC,MAAMgG,WAAW,MAAMzD,KAAKC,UAAUzC,mBAAmBC;QACzD0B,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOsE,UAAU;QACjB,MAAMC,cAAcC,cAAcF,SAAS,OAAO,EAAEhG;QACpD,OAAO;YAAE,SAASiG;YAAa,OAAOD,SAAS,KAAK;QAAC;IACvD;IAEO,eAAeG,2BACpBC,IAAY,EACZrG,iBAA+B,EAC/BC,gBAAmC;QAEnC,MAAM,EAAEuD,OAAO,EAAEE,KAAK,EAAE,GAAG,MAAMlB,KAC/B6D,MACArG,mBACAC;QAEF,OAAO;YAAEuD;YAASE;QAAM;IAC1B;IAEO,SAAS4C,yBAAyBL,QAAgB;QACvD,IAAI;YAEF,MAAMM,YAAYN,SAAS,KAAK,CAAC;YACjC,IAAIM,WACF,OAAOA,SAAS,CAAC,EAAE;YAIrB,MAAMC,iBAAiBP,SAAS,KAAK,CACnC;YAEF,IAAIO,gBACF,OAAOA,cAAc,CAAC,EAAE;YAI1B,MAAMC,gBAAgBR,SAAS,KAAK,CAAC;YACrC,IAAIQ,eACF,OAAOA,aAAa,CAAC,EAAE;QAE3B,EAAE,OAAM,CAAC;QAET,OAAOR;IACT;IAEO,SAASS,yBAAyBC,KAAa;QACpD,IAAIA,MAAM,QAAQ,CAAC,SAEjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;QAG5C,OAAOA;IACT;IAEO,SAASR,cACdQ,KAAa,EACb1G,gBAAmC;QAEnC,MAAM2G,kBAAkBN,yBAAyBK;QAEjD,IAAIC,QAAAA,kBAAAA,KAAAA,IAAAA,gBAAiB,KAAK,CAAC,oBAAoB;gBACtCC;YAAP,OAAO,QAAAA,CAAAA,yBAAAA,gBACJ,KAAK,CAAC,kBAAiB,IADnBA,KAAAA,IAAAA,uBAEH,KAAK,CAAC,GACP,GAAG,CAAC/C;QACT;QACA,IAAI;YACF,OAAOiB,KAAK,KAAK,CAAC6B;QACpB,EAAE,OAAM,CAAC;QACT,IAAI;YACF,OAAO7B,KAAK,KAAK,CAAC+B,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAWF;QAC/B,EAAE,OAAOnB,GAAG,CAAC;QAEb,IACE5B,AAAmC,oBAAnCA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa5D,qBACb4D,AAAmC,kBAAnCA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa5D,mBACb;YACA,MAAM8G,aAAaL,yBAAyBE;YAC5C,OAAO7B,KAAK,KAAK,CAAC+B,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAWC;QAC/B;QACA,MAAM3E,MAAM,CAAC,+BAA+B,EAAEuE,OAAO;IACvD"}
+{"version":3,"file":"ai-model/service-caller/index.js","sources":["webpack://@midscene/core/webpack/runtime/compat_get_default_export","webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/service-caller/index.ts"],"sourcesContent":["// getDefaultExport function for compatibility with non-ESM modules\n__webpack_require__.n = (module) => {\n\tvar getter = module && module.__esModule ?\n\t\t() => (module['default']) :\n\t\t() => (module);\n\t__webpack_require__.d(getter, { a: getter });\n\treturn getter;\n};\n","__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import { AIResponseFormat, type AIUsageInfo } from '@/types';\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\nimport { Anthropic } from '@anthropic-ai/sdk';\nimport {\n  DefaultAzureCredential,\n  getBearerTokenProvider,\n} from '@azure/identity';\nimport {\n  type IModelPreferences,\n  MIDSCENE_API_TYPE,\n  MIDSCENE_LANGSMITH_DEBUG,\n  OPENAI_MAX_TOKENS,\n  decideModelConfig,\n  getAIConfig,\n  getAIConfigInBoolean,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { parseBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { ifInBrowser } from '@midscene/shared/utils';\nimport { HttpsProxyAgent } from 'https-proxy-agent';\nimport { jsonrepair } from 'jsonrepair';\nimport OpenAI, { AzureOpenAI } from 'openai';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { Stream } from 'openai/streaming';\nimport { SocksProxyAgent } from 'socks-proxy-agent';\nimport { AIActionType, type AIArgs } from '../common';\nimport { assertSchema } from '../prompt/assertion';\nimport { locatorSchema } from '../prompt/llm-locator';\nimport { planSchema } from '../prompt/llm-planning';\n\nasync function createChatClient({\n  AIActionTypeValue,\n  modelPreferences,\n}: {\n  AIActionTypeValue: AIActionType;\n  modelPreferences: IModelPreferences;\n}): Promise<{\n  completion: OpenAI.Chat.Completions;\n  style: 'openai' | 'anthropic';\n  modelName: string;\n  modelDescription: string;\n}> {\n  const {\n    socksProxy,\n    httpProxy,\n    modelName,\n    openaiBaseURL,\n    openaiApiKey,\n    openaiExtraConfig,\n    openaiUseAzureDeprecated,\n    useAzureOpenai,\n    azureOpenaiScope,\n    azureOpenaiKey,\n    azureOpenaiEndpoint,\n    azureOpenaiApiVersion,\n    azureOpenaiDeployment,\n    azureExtraConfig,\n    useAnthropicSdk,\n    anthropicApiKey,\n    modelDescription,\n  } = decideModelConfig(modelPreferences, true);\n\n  let openai: OpenAI | AzureOpenAI | undefined;\n\n  let proxyAgent = undefined;\n  const debugProxy = getDebug('ai:call:proxy');\n  if (httpProxy) {\n    debugProxy('using http proxy', httpProxy);\n    proxyAgent = new HttpsProxyAgent(httpProxy);\n  } else if (socksProxy) {\n    debugProxy('using socks proxy', socksProxy);\n    proxyAgent = new SocksProxyAgent(socksProxy);\n  }\n\n  if (openaiUseAzureDeprecated) {\n    // this is deprecated\n    openai = new AzureOpenAI({\n      baseURL: openaiBaseURL,\n      apiKey: openaiApiKey,\n      httpAgent: proxyAgent,\n      ...openaiExtraConfig,\n      dangerouslyAllowBrowser: true,\n    }) as OpenAI;\n  } else if (useAzureOpenai) {\n    // https://learn.microsoft.com/en-us/azure/ai-services/openai/chatgpt-quickstart?tabs=bash%2Cjavascript-key%2Ctypescript-keyless%2Cpython&pivots=programming-language-javascript#rest-api\n    // keyless authentication\n    let tokenProvider: any = undefined;\n    if (azureOpenaiScope) {\n      assert(\n        !ifInBrowser,\n        'Azure OpenAI is not supported in browser with Midscene.',\n      );\n      const credential = new DefaultAzureCredential();\n\n      tokenProvider = getBearerTokenProvider(credential, azureOpenaiScope);\n\n      openai = new AzureOpenAI({\n        azureADTokenProvider: tokenProvider,\n        endpoint: azureOpenaiEndpoint,\n        apiVersion: azureOpenaiApiVersion,\n        deployment: azureOpenaiDeployment,\n        ...openaiExtraConfig,\n        ...azureExtraConfig,\n      });\n    } else {\n      // endpoint, apiKey, apiVersion, deployment\n      openai = new AzureOpenAI({\n        apiKey: azureOpenaiKey,\n        endpoint: azureOpenaiEndpoint,\n        apiVersion: azureOpenaiApiVersion,\n        deployment: azureOpenaiDeployment,\n        dangerouslyAllowBrowser: true,\n        ...openaiExtraConfig,\n        ...azureExtraConfig,\n      });\n    }\n  } else if (!useAnthropicSdk) {\n    openai = new OpenAI({\n      baseURL: openaiBaseURL,\n      apiKey: openaiApiKey,\n      httpAgent: proxyAgent,\n      ...openaiExtraConfig,\n      defaultHeaders: {\n        ...(openaiExtraConfig?.defaultHeaders || {}),\n        [MIDSCENE_API_TYPE]: AIActionTypeValue.toString(),\n      },\n      dangerouslyAllowBrowser: true,\n    });\n  }\n\n  if (openai && getAIConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)) {\n    if (ifInBrowser) {\n      throw new Error('langsmith is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langsmith wrapper enabled');\n    const { wrapOpenAI } = await import('langsmith/wrappers');\n    openai = wrapOpenAI(openai);\n  }\n\n  if (typeof openai !== 'undefined') {\n    return {\n      completion: openai.chat.completions,\n      style: 'openai',\n      modelName,\n      modelDescription,\n    };\n  }\n\n  // Anthropic\n  if (useAnthropicSdk) {\n    openai = new Anthropic({\n      apiKey: anthropicApiKey,\n      httpAgent: proxyAgent,\n      dangerouslyAllowBrowser: true,\n    }) as any;\n  }\n\n  if (typeof openai !== 'undefined' && (openai as any).messages) {\n    return {\n      completion: (openai as any).messages,\n      style: 'anthropic',\n      modelName,\n      modelDescription,\n    };\n  }\n\n  throw new Error('Openai SDK or Anthropic SDK is not initialized');\n}\n\nexport async function call(\n  messages: ChatCompletionMessageParam[],\n  AIActionTypeValue: AIActionType,\n  modelPreferences: IModelPreferences,\n  options?: {\n    stream?: boolean;\n    onChunk?: StreamingCallback;\n  },\n): Promise<{ content: string; usage?: AIUsageInfo; isStreamed: boolean }> {\n  const { completion, style, modelName, modelDescription } =\n    await createChatClient({\n      AIActionTypeValue,\n      modelPreferences,\n    });\n\n  const responseFormat = getResponseFormat(modelName, AIActionTypeValue);\n\n  const maxTokens = getAIConfig(OPENAI_MAX_TOKENS);\n  const debugCall = getDebug('ai:call');\n  const debugProfileStats = getDebug('ai:profile:stats');\n  const debugProfileDetail = getDebug('ai:profile:detail');\n\n  const startTime = Date.now();\n\n  const isStreaming = options?.stream && options?.onChunk;\n  let content: string | undefined;\n  let accumulated = '';\n  let usage: OpenAI.CompletionUsage | undefined;\n  let timeCost: number | undefined;\n\n  const commonConfig = {\n    temperature: vlLocateMode(modelPreferences) === 'vlm-ui-tars' ? 0.0 : 0.1,\n    stream: !!isStreaming,\n    max_tokens:\n      typeof maxTokens === 'number'\n        ? maxTokens\n        : Number.parseInt(maxTokens || '2048', 10),\n    ...(vlLocateMode(modelPreferences) === 'qwen-vl' // qwen specific config\n      ? {\n          vl_high_resolution_images: true,\n        }\n      : {}),\n  };\n\n  try {\n    if (style === 'openai') {\n      debugCall(\n        `sending ${isStreaming ? 'streaming ' : ''}request to ${modelName}`,\n      );\n\n      if (isStreaming) {\n        const stream = (await completion.create(\n          {\n            model: modelName,\n            messages,\n            response_format: responseFormat,\n            ...commonConfig,\n          },\n          {\n            stream: true,\n          },\n        )) as Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {\n          _request_id?: string | null;\n        };\n\n        for await (const chunk of stream) {\n          const content = chunk.choices?.[0]?.delta?.content || '';\n          const reasoning_content =\n            (chunk.choices?.[0]?.delta as any)?.reasoning_content || '';\n\n          // Check for usage info in any chunk (OpenAI provides usage in separate chunks)\n          if (chunk.usage) {\n            usage = chunk.usage;\n          }\n\n          if (content || reasoning_content) {\n            accumulated += content;\n            const chunkData: CodeGenerationChunk = {\n              content,\n              reasoning_content,\n              accumulated,\n              isComplete: false,\n              usage: undefined,\n            };\n            options.onChunk!(chunkData);\n          }\n\n          // Check if stream is complete\n          if (chunk.choices?.[0]?.finish_reason) {\n            timeCost = Date.now() - startTime;\n\n            // If usage is not available from the stream, provide a basic usage info\n            if (!usage) {\n              // Estimate token counts based on content length (rough approximation)\n              const estimatedTokens = Math.max(\n                1,\n                Math.floor(accumulated.length / 4),\n              );\n              usage = {\n                prompt_tokens: estimatedTokens,\n                completion_tokens: estimatedTokens,\n                total_tokens: estimatedTokens * 2,\n              };\n            }\n\n            // Send final chunk\n            const finalChunk: CodeGenerationChunk = {\n              content: '',\n              accumulated,\n              reasoning_content: '',\n              isComplete: true,\n              usage: {\n                prompt_tokens: usage.prompt_tokens ?? 0,\n                completion_tokens: usage.completion_tokens ?? 0,\n                total_tokens: usage.total_tokens ?? 0,\n                time_cost: timeCost ?? 0,\n                model_name: modelName,\n                model_description: modelDescription,\n                intent: modelPreferences.intent,\n              },\n            };\n            options.onChunk!(finalChunk);\n            break;\n          }\n        }\n        content = accumulated;\n        debugProfileStats(\n          `streaming model, ${modelName}, mode, ${vlLocateMode(modelPreferences) || 'default'}, cost-ms, ${timeCost}`,\n        );\n      } else {\n        const result = await completion.create({\n          model: modelName,\n          messages,\n          response_format: responseFormat,\n          ...commonConfig,\n        } as any);\n        timeCost = Date.now() - startTime;\n\n        debugProfileStats(\n          `model, ${modelName}, mode, ${vlLocateMode(modelPreferences) || 'default'}, ui-tars-version, ${uiTarsModelVersion(modelPreferences)}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}`,\n        );\n\n        debugProfileDetail(\n          `model usage detail: ${JSON.stringify(result.usage)}`,\n        );\n\n        assert(\n          result.choices,\n          `invalid response from LLM service: ${JSON.stringify(result)}`,\n        );\n        content = result.choices[0].message.content!;\n        usage = result.usage;\n      }\n\n      debugCall(`response: ${content}`);\n      assert(content, 'empty content');\n    } else if (style === 'anthropic') {\n      const convertImageContent = (content: any) => {\n        if (content.type === 'image_url') {\n          const imgBase64 = content.image_url.url;\n          assert(imgBase64, 'image_url is required');\n          const { mimeType, body } = parseBase64(content.image_url.url);\n          return {\n            source: {\n              type: 'base64',\n              media_type: mimeType,\n              data: body,\n            },\n            type: 'image',\n          };\n        }\n        return content;\n      };\n\n      if (isStreaming) {\n        const stream = (await completion.create({\n          model: modelName,\n          system: 'You are a versatile professional in software UI automation',\n          messages: messages.map((m) => ({\n            role: 'user',\n            content: Array.isArray(m.content)\n              ? (m.content as any).map(convertImageContent)\n              : m.content,\n          })),\n          response_format: responseFormat,\n          ...commonConfig,\n        } as any)) as any;\n\n        for await (const chunk of stream) {\n          const content = chunk.delta?.text || '';\n          if (content) {\n            accumulated += content;\n            const chunkData: CodeGenerationChunk = {\n              content,\n              accumulated,\n              reasoning_content: '',\n              isComplete: false,\n              usage: undefined,\n            };\n            options.onChunk!(chunkData);\n          }\n\n          // Check if stream is complete\n          if (chunk.type === 'message_stop') {\n            timeCost = Date.now() - startTime;\n            const anthropicUsage = chunk.usage;\n\n            // Send final chunk\n            const finalChunk: CodeGenerationChunk = {\n              content: '',\n              accumulated,\n              reasoning_content: '',\n              isComplete: true,\n              usage: anthropicUsage\n                ? {\n                    prompt_tokens: anthropicUsage.input_tokens ?? 0,\n                    completion_tokens: anthropicUsage.output_tokens ?? 0,\n                    total_tokens:\n                      (anthropicUsage.input_tokens ?? 0) +\n                      (anthropicUsage.output_tokens ?? 0),\n                    time_cost: timeCost ?? 0,\n                    model_name: modelName,\n                    model_description: modelDescription,\n                    intent: modelPreferences.intent,\n                  }\n                : undefined,\n            };\n            options.onChunk!(finalChunk);\n            break;\n          }\n        }\n        content = accumulated;\n      } else {\n        const result = await completion.create({\n          model: modelName,\n          system: 'You are a versatile professional in software UI automation',\n          messages: messages.map((m) => ({\n            role: 'user',\n            content: Array.isArray(m.content)\n              ? (m.content as any).map(convertImageContent)\n              : m.content,\n          })),\n          response_format: responseFormat,\n          ...commonConfig,\n        } as any);\n        timeCost = Date.now() - startTime;\n        content = (result as any).content[0].text as string;\n        usage = result.usage;\n      }\n\n      assert(content, 'empty content');\n    }\n    // Ensure we always have usage info for streaming responses\n    if (isStreaming && !usage) {\n      // Estimate token counts based on content length (rough approximation)\n      const estimatedTokens = Math.max(\n        1,\n        Math.floor((content || '').length / 4),\n      );\n      usage = {\n        prompt_tokens: estimatedTokens,\n        completion_tokens: estimatedTokens,\n        total_tokens: estimatedTokens * 2,\n      };\n    }\n\n    return {\n      content: content || '',\n      usage: usage\n        ? {\n            prompt_tokens: usage.prompt_tokens ?? 0,\n            completion_tokens: usage.completion_tokens ?? 0,\n            total_tokens: usage.total_tokens ?? 0,\n            time_cost: timeCost ?? 0,\n            model_name: modelName,\n            model_description: modelDescription,\n            intent: modelPreferences.intent,\n          }\n        : undefined,\n      isStreamed: !!isStreaming,\n    };\n  } catch (e: any) {\n    console.error(' call AI error', e);\n    const newError = new Error(\n      `failed to call ${isStreaming ? 'streaming ' : ''}AI model service: ${e.message}. Trouble shooting: https://midscenejs.com/model-provider.html`,\n      {\n        cause: e,\n      },\n    );\n    throw newError;\n  }\n}\n\nexport const getResponseFormat = (\n  modelName: string,\n  AIActionTypeValue: AIActionType,\n):\n  | OpenAI.ChatCompletionCreateParams['response_format']\n  | OpenAI.ResponseFormatJSONObject => {\n  let responseFormat:\n    | OpenAI.ChatCompletionCreateParams['response_format']\n    | OpenAI.ResponseFormatJSONObject\n    | undefined;\n\n  if (modelName.includes('gpt-4')) {\n    switch (AIActionTypeValue) {\n      case AIActionType.ASSERT:\n        responseFormat = assertSchema;\n        break;\n      case AIActionType.INSPECT_ELEMENT:\n        responseFormat = locatorSchema;\n        break;\n      case AIActionType.PLAN:\n        responseFormat = planSchema;\n        break;\n      case AIActionType.EXTRACT_DATA:\n      case AIActionType.DESCRIBE_ELEMENT:\n        responseFormat = { type: AIResponseFormat.JSON };\n        break;\n    }\n  }\n\n  // gpt-4o-2024-05-13 only supports json_object response format\n  if (modelName === 'gpt-4o-2024-05-13') {\n    responseFormat = { type: AIResponseFormat.JSON };\n  }\n\n  return responseFormat;\n};\n\nexport async function callToGetJSONObject<T>(\n  messages: ChatCompletionMessageParam[],\n  AIActionTypeValue: AIActionType,\n  modelPreferences: IModelPreferences,\n): Promise<{ content: T; usage?: AIUsageInfo }> {\n  const response = await call(messages, AIActionTypeValue, modelPreferences);\n  assert(response, 'empty response');\n  const jsonContent = safeParseJson(response.content, modelPreferences);\n  return { content: jsonContent, usage: response.usage };\n}\n\nexport async function callAiFnWithStringResponse<T>(\n  msgs: AIArgs,\n  AIActionTypeValue: AIActionType,\n  modelPreferences: IModelPreferences,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n  const { content, usage } = await call(\n    msgs,\n    AIActionTypeValue,\n    modelPreferences,\n  );\n  return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n  try {\n    // First, try to match a JSON object directly in the response\n    const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\n    if (jsonMatch) {\n      return jsonMatch[1];\n    }\n\n    // If no direct JSON object is found, try to extract JSON from a code block\n    const codeBlockMatch = response.match(\n      /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\n    );\n    if (codeBlockMatch) {\n      return codeBlockMatch[1];\n    }\n\n    // If no code block is found, try to find a JSON-like structure in the text\n    const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\n    if (jsonLikeMatch) {\n      return jsonLikeMatch[0];\n    }\n  } catch {}\n  // If no JSON-like structure is found, return the original response\n  return response;\n}\n\nexport function preprocessDoubaoBboxJson(input: string) {\n  if (input.includes('bbox')) {\n    // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\n    while (/\\d+\\s+\\d+/.test(input)) {\n      input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n    }\n  }\n  return input;\n}\n\nexport function safeParseJson(\n  input: string,\n  modelPreferences: IModelPreferences,\n) {\n  const cleanJsonString = extractJSONFromCodeBlock(input);\n  // match the point\n  if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\n    return cleanJsonString\n      .match(/\\((\\d+),(\\d+)\\)/)\n      ?.slice(1)\n      .map(Number);\n  }\n  try {\n    return JSON.parse(cleanJsonString);\n  } catch {}\n  try {\n    return JSON.parse(jsonrepair(cleanJsonString));\n  } catch (e) {}\n\n  if (\n    vlLocateMode(modelPreferences) === 'doubao-vision' ||\n    vlLocateMode(modelPreferences) === 'vlm-ui-tars'\n  ) {\n    const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\n    return JSON.parse(jsonrepair(jsonString));\n  }\n  throw Error(`failed to parse json response: ${input}`);\n}\n"],"names":["__webpack_require__","module","getter","definition","key","Object","obj","prop","Symbol","createChatClient","AIActionTypeValue","modelPreferences","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","openaiUseAzureDeprecated","useAzureOpenai","azureOpenaiScope","azureOpenaiKey","azureOpenaiEndpoint","azureOpenaiApiVersion","azureOpenaiDeployment","azureExtraConfig","useAnthropicSdk","anthropicApiKey","modelDescription","decideModelConfig","openai","proxyAgent","debugProxy","getDebug","HttpsProxyAgent","SocksProxyAgent","AzureOpenAI","tokenProvider","assert","ifInBrowser","credential","DefaultAzureCredential","getBearerTokenProvider","OpenAI","MIDSCENE_API_TYPE","getAIConfigInBoolean","MIDSCENE_LANGSMITH_DEBUG","Error","console","wrapOpenAI","Anthropic","call","messages","options","completion","style","responseFormat","getResponseFormat","maxTokens","getAIConfig","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","isStreaming","content","accumulated","usage","timeCost","commonConfig","vlLocateMode","Number","stream","chunk","_chunk_choices__delta","_chunk_choices__delta1","_chunk_choices_2","reasoning_content","chunkData","undefined","estimatedTokens","Math","finalChunk","_result_usage","_result_usage1","_result_usage2","result","uiTarsModelVersion","JSON","convertImageContent","imgBase64","mimeType","body","parseBase64","m","Array","_chunk_delta","anthropicUsage","e","newError","AIActionType","assertSchema","locatorSchema","planSchema","AIResponseFormat","callToGetJSONObject","response","jsonContent","safeParseJson","callAiFnWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","cleanJsonString","_cleanJsonString_match","jsonrepair","jsonString"],"mappings":";;;;;;;;;;;;;;;;;;;IACAA,oBAAoB,CAAC,GAAG,CAACC;QACxB,IAAIC,SAASD,UAAUA,OAAO,UAAU,GACvC,IAAOA,MAAM,CAAC,UAAU,GACxB,IAAOA;QACRD,oBAAoB,CAAC,CAACE,QAAQ;YAAE,GAAGA;QAAO;QAC1C,OAAOA;IACR;;;ICPAF,oBAAoB,CAAC,GAAG,CAAC,UAASG;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGH,oBAAoB,CAAC,CAACG,YAAYC,QAAQ,CAACJ,oBAAoB,CAAC,CAAC,UAASI,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAJ,oBAAoB,CAAC,GAAG,CAACM,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFP,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOQ,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;IC2BA,eAAeI,iBAAiB,EAC9BC,iBAAiB,EACjBC,gBAAgB,EAIjB;QAMC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,wBAAwB,EACxBC,cAAc,EACdC,gBAAgB,EAChBC,cAAc,EACdC,mBAAmB,EACnBC,qBAAqB,EACrBC,qBAAqB,EACrBC,gBAAgB,EAChBC,eAAe,EACfC,eAAe,EACfC,gBAAgB,EACjB,GAAGC,AAAAA,IAAAA,oBAAAA,iBAAAA,AAAAA,EAAkBlB,kBAAkB;QAExC,IAAImB;QAEJ,IAAIC;QACJ,MAAMC,aAAaC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QAC5B,IAAIpB,WAAW;YACbmB,WAAW,oBAAoBnB;YAC/BkB,aAAa,IAAIG,2CAAAA,eAAeA,CAACrB;QACnC,OAAO,IAAID,YAAY;YACrBoB,WAAW,qBAAqBpB;YAChCmB,aAAa,IAAII,2CAAAA,eAAeA,CAACvB;QACnC;QAEA,IAAIM,0BAEFY,SAAS,IAAIM,gCAAAA,WAAWA,CAAC;YACvB,SAASrB;YACT,QAAQC;YACR,WAAWe;YACX,GAAGd,iBAAiB;YACpB,yBAAyB;QAC3B;aACK,IAAIE,gBAAgB;YAGzB,IAAIkB;YACJ,IAAIjB,kBAAkB;gBACpBkB,IAAAA,sBAAAA,MAAAA,AAAAA,EACE,CAACC,sBAAAA,WAAWA,EACZ;gBAEF,MAAMC,aAAa,IAAIC,yBAAAA,sBAAsBA;gBAE7CJ,gBAAgBK,AAAAA,IAAAA,yBAAAA,sBAAAA,AAAAA,EAAuBF,YAAYpB;gBAEnDU,SAAS,IAAIM,gCAAAA,WAAWA,CAAC;oBACvB,sBAAsBC;oBACtB,UAAUf;oBACV,YAAYC;oBACZ,YAAYC;oBACZ,GAAGP,iBAAiB;oBACpB,GAAGQ,gBAAgB;gBACrB;YACF,OAEEK,SAAS,IAAIM,gCAAAA,WAAWA,CAAC;gBACvB,QAAQf;gBACR,UAAUC;gBACV,YAAYC;gBACZ,YAAYC;gBACZ,yBAAyB;gBACzB,GAAGP,iBAAiB;gBACpB,GAAGQ,gBAAgB;YACrB;QAEJ,OAAO,IAAI,CAACC,iBACVI,SAAS,IAAIa,CAAAA,yBAAAA,EAAO;YAClB,SAAS5B;YACT,QAAQC;YACR,WAAWe;YACX,GAAGd,iBAAiB;YACpB,gBAAgB;gBACd,GAAIA,AAAAA,CAAAA,QAAAA,oBAAAA,KAAAA,IAAAA,kBAAmB,cAAc,AAAD,KAAK,CAAC,CAAC;gBAC3C,CAAC2B,oBAAAA,iBAAiBA,CAAC,EAAElC,kBAAkB,QAAQ;YACjD;YACA,yBAAyB;QAC3B;QAGF,IAAIoB,UAAUe,AAAAA,IAAAA,oBAAAA,oBAAAA,AAAAA,EAAqBC,oBAAAA,wBAAwBA,GAAG;YAC5D,IAAIP,sBAAAA,WAAWA,EACb,MAAM,IAAIQ,MAAM;YAElBC,QAAQ,GAAG,CAAC;YACZ,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM;YAC7BnB,SAASmB,WAAWnB;QACtB;QAEA,IAAI,AAAkB,WAAXA,QACT,OAAO;YACL,YAAYA,OAAO,IAAI,CAAC,WAAW;YACnC,OAAO;YACPhB;YACAc;QACF;QAIF,IAAIF,iBACFI,SAAS,IAAIoB,oBAAAA,SAASA,CAAC;YACrB,QAAQvB;YACR,WAAWI;YACX,yBAAyB;QAC3B;QAGF,IAAI,AAAkB,WAAXD,UAA2BA,OAAe,QAAQ,EAC3D,OAAO;YACL,YAAaA,OAAe,QAAQ;YACpC,OAAO;YACPhB;YACAc;QACF;QAGF,MAAM,IAAImB,MAAM;IAClB;IAEO,eAAeI,KACpBC,QAAsC,EACtC1C,iBAA+B,EAC/BC,gBAAmC,EACnC0C,OAGC;QAED,MAAM,EAAEC,UAAU,EAAEC,KAAK,EAAEzC,SAAS,EAAEc,gBAAgB,EAAE,GACtD,MAAMnB,iBAAiB;YACrBC;YACAC;QACF;QAEF,MAAM6C,iBAAiBC,kBAAkB3C,WAAWJ;QAEpD,MAAMgD,YAAYC,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYC,oBAAAA,iBAAiBA;QAC/C,MAAMC,YAAY5B,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QAC3B,MAAM6B,oBAAoB7B,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QACnC,MAAM8B,qBAAqB9B,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QAEpC,MAAM+B,YAAYC,KAAK,GAAG;QAE1B,MAAMC,cAAcb,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM,AAAD,KAAKA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO,AAAD;QACtD,IAAIc;QACJ,IAAIC,cAAc;QAClB,IAAIC;QACJ,IAAIC;QAEJ,MAAMC,eAAe;YACnB,aAAaC,AAAmC,kBAAnCA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa7D,oBAAsC,MAAM;YACtE,QAAQ,CAAC,CAACuD;YACV,YACE,AAAqB,YAArB,OAAOR,YACHA,YACAe,OAAO,QAAQ,CAACf,aAAa,QAAQ;YAC3C,GAAIc,AAAmC,cAAnCA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa7D,oBACb;gBACE,2BAA2B;YAC7B,IACA,CAAC,CAAC;QACR;QAEA,IAAI;YACF,IAAI4C,AAAU,aAAVA,OAAoB;gBACtBM,UACE,CAAC,QAAQ,EAAEK,cAAc,eAAe,GAAG,WAAW,EAAEpD,WAAW;gBAGrE,IAAIoD,aAAa;oBACf,MAAMQ,SAAU,MAAMpB,WAAW,MAAM,CACrC;wBACE,OAAOxC;wBACPsC;wBACA,iBAAiBI;wBACjB,GAAGe,YAAY;oBACjB,GACA;wBACE,QAAQ;oBACV;oBAKF,WAAW,MAAMI,SAASD,OAAQ;4BAChBE,uBAAAA,iBAAAA,gBAEbC,wBAAAA,kBAAAA,iBAoBCC,kBAAAA;wBAtBJ,MAAMX,UAAUS,AAAAA,SAAAA,CAAAA,iBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,kBAAAA,cAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,QAAAA,CAAAA,wBAAAA,gBAAoB,KAAK,AAAD,IAAxBA,KAAAA,IAAAA,sBAA2B,OAAO,AAAD,KAAK;wBACtD,MAAMG,oBACJ,AAAC,SAAAF,CAAAA,kBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,mBAAAA,eAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,QAAAA,CAAAA,yBAAAA,iBAAoB,KAAK,AAAD,IAAxBA,KAAAA,IAAAA,uBAAmC,iBAAiB,AAAD,KAAK;wBAG3D,IAAIF,MAAM,KAAK,EACbN,QAAQM,MAAM,KAAK;wBAGrB,IAAIR,WAAWY,mBAAmB;4BAChCX,eAAeD;4BACf,MAAMa,YAAiC;gCACrCb;gCACAY;gCACAX;gCACA,YAAY;gCACZ,OAAOa;4BACT;4BACA5B,QAAQ,OAAO,CAAE2B;wBACnB;wBAGA,IAAI,QAAAF,CAAAA,kBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,mBAAAA,eAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,iBAAoB,aAAa,EAAE;4BACrCR,WAAWL,KAAK,GAAG,KAAKD;4BAGxB,IAAI,CAACK,OAAO;gCAEV,MAAMa,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAACf,YAAY,MAAM,GAAG;gCAElCC,QAAQ;oCACN,eAAea;oCACf,mBAAmBA;oCACnB,cAAcA,AAAkB,IAAlBA;gCAChB;4BACF;4BAGA,MAAME,aAAkC;gCACtC,SAAS;gCACThB;gCACA,mBAAmB;gCACnB,YAAY;gCACZ,OAAO;oCACL,eAAeC,MAAM,aAAa,IAAI;oCACtC,mBAAmBA,MAAM,iBAAiB,IAAI;oCAC9C,cAAcA,MAAM,YAAY,IAAI;oCACpC,WAAWC,YAAY;oCACvB,YAAYxD;oCACZ,mBAAmBc;oCACnB,QAAQjB,iBAAiB,MAAM;gCACjC;4BACF;4BACA0C,QAAQ,OAAO,CAAE+B;4BACjB;wBACF;oBACF;oBACAjB,UAAUC;oBACVN,kBACE,CAAC,iBAAiB,EAAEhD,UAAU,QAAQ,EAAE0D,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa7D,qBAAqB,UAAU,WAAW,EAAE2D,UAAU;gBAE/G,OAAO;wBAUoJe,eAAyDC,gBAAwDC;oBAT1Q,MAAMC,SAAS,MAAMlC,WAAW,MAAM,CAAC;wBACrC,OAAOxC;wBACPsC;wBACA,iBAAiBI;wBACjB,GAAGe,YAAY;oBACjB;oBACAD,WAAWL,KAAK,GAAG,KAAKD;oBAExBF,kBACE,CAAC,OAAO,EAAEhD,UAAU,QAAQ,EAAE0D,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa7D,qBAAqB,UAAU,mBAAmB,EAAE8E,AAAAA,IAAAA,oBAAAA,kBAAAA,AAAAA,EAAmB9E,kBAAkB,iBAAiB,EAAE0E,AAAAA,SAAAA,CAAAA,gBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,cAAc,aAAa,AAAD,KAAK,GAAG,qBAAqB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,iBAAiB,AAAD,KAAK,GAAG,gBAAgB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,YAAY,AAAD,KAAK,GAAG,WAAW,EAAEjB,SAAS,aAAa,EAAEkB,OAAO,WAAW,IAAI,IAAI;oBAG1WzB,mBACE,CAAC,oBAAoB,EAAE2B,KAAK,SAAS,CAACF,OAAO,KAAK,GAAG;oBAGvDlD,IAAAA,sBAAAA,MAAAA,AAAAA,EACEkD,OAAO,OAAO,EACd,CAAC,mCAAmC,EAAEE,KAAK,SAAS,CAACF,SAAS;oBAEhErB,UAAUqB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;oBAC3CnB,QAAQmB,OAAO,KAAK;gBACtB;gBAEA3B,UAAU,CAAC,UAAU,EAAEM,SAAS;gBAChC7B,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO6B,SAAS;YAClB,OAAO,IAAIZ,AAAU,gBAAVA,OAAuB;gBAChC,MAAMoC,sBAAsB,CAACxB;oBAC3B,IAAIA,AAAiB,gBAAjBA,QAAQ,IAAI,EAAkB;wBAChC,MAAMyB,YAAYzB,QAAQ,SAAS,CAAC,GAAG;wBACvC7B,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOsD,WAAW;wBAClB,MAAM,EAAEC,QAAQ,EAAEC,IAAI,EAAE,GAAGC,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAY5B,QAAQ,SAAS,CAAC,GAAG;wBAC5D,OAAO;4BACL,QAAQ;gCACN,MAAM;gCACN,YAAY0B;gCACZ,MAAMC;4BACR;4BACA,MAAM;wBACR;oBACF;oBACA,OAAO3B;gBACT;gBAEA,IAAID,aAAa;oBACf,MAAMQ,SAAU,MAAMpB,WAAW,MAAM,CAAC;wBACtC,OAAOxC;wBACP,QAAQ;wBACR,UAAUsC,SAAS,GAAG,CAAC,CAAC4C,IAAO;gCAC7B,MAAM;gCACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;4BACf;wBACA,iBAAiBxC;wBACjB,GAAGe,YAAY;oBACjB;oBAEA,WAAW,MAAMI,SAASD,OAAQ;4BAChBwB;wBAAhB,MAAM/B,UAAU+B,AAAAA,SAAAA,CAAAA,eAAAA,MAAM,KAAK,AAAD,IAAVA,KAAAA,IAAAA,aAAa,IAAI,AAAD,KAAK;wBACrC,IAAI/B,SAAS;4BACXC,eAAeD;4BACf,MAAMa,YAAiC;gCACrCb;gCACAC;gCACA,mBAAmB;gCACnB,YAAY;gCACZ,OAAOa;4BACT;4BACA5B,QAAQ,OAAO,CAAE2B;wBACnB;wBAGA,IAAIL,AAAe,mBAAfA,MAAM,IAAI,EAAqB;4BACjCL,WAAWL,KAAK,GAAG,KAAKD;4BACxB,MAAMmC,iBAAiBxB,MAAM,KAAK;4BAGlC,MAAMS,aAAkC;gCACtC,SAAS;gCACThB;gCACA,mBAAmB;gCACnB,YAAY;gCACZ,OAAO+B,iBACH;oCACE,eAAeA,eAAe,YAAY,IAAI;oCAC9C,mBAAmBA,eAAe,aAAa,IAAI;oCACnD,cACGA,AAAAA,CAAAA,eAAe,YAAY,IAAI,KAC/BA,CAAAA,eAAe,aAAa,IAAI;oCACnC,WAAW7B,YAAY;oCACvB,YAAYxD;oCACZ,mBAAmBc;oCACnB,QAAQjB,iBAAiB,MAAM;gCACjC,IACAsE;4BACN;4BACA5B,QAAQ,OAAO,CAAE+B;4BACjB;wBACF;oBACF;oBACAjB,UAAUC;gBACZ,OAAO;oBACL,MAAMoB,SAAS,MAAMlC,WAAW,MAAM,CAAC;wBACrC,OAAOxC;wBACP,QAAQ;wBACR,UAAUsC,SAAS,GAAG,CAAC,CAAC4C,IAAO;gCAC7B,MAAM;gCACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;4BACf;wBACA,iBAAiBxC;wBACjB,GAAGe,YAAY;oBACjB;oBACAD,WAAWL,KAAK,GAAG,KAAKD;oBACxBG,UAAWqB,OAAe,OAAO,CAAC,EAAE,CAAC,IAAI;oBACzCnB,QAAQmB,OAAO,KAAK;gBACtB;gBAEAlD,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO6B,SAAS;YAClB;YAEA,IAAID,eAAe,CAACG,OAAO;gBAEzB,MAAMa,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEhB,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;gBAEtCE,QAAQ;oBACN,eAAea;oBACf,mBAAmBA;oBACnB,cAAcA,AAAkB,IAAlBA;gBAChB;YACF;YAEA,OAAO;gBACL,SAASf,WAAW;gBACpB,OAAOE,QACH;oBACE,eAAeA,MAAM,aAAa,IAAI;oBACtC,mBAAmBA,MAAM,iBAAiB,IAAI;oBAC9C,cAAcA,MAAM,YAAY,IAAI;oBACpC,WAAWC,YAAY;oBACvB,YAAYxD;oBACZ,mBAAmBc;oBACnB,QAAQjB,iBAAiB,MAAM;gBACjC,IACAsE;gBACJ,YAAY,CAAC,CAACf;YAChB;QACF,EAAE,OAAOkC,GAAQ;YACfpD,QAAQ,KAAK,CAAC,kBAAkBoD;YAChC,MAAMC,WAAW,IAAItD,MACnB,CAAC,eAAe,EAAEmB,cAAc,eAAe,GAAG,kBAAkB,EAAEkC,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC/I;gBACE,OAAOA;YACT;YAEF,MAAMC;QACR;IACF;IAEO,MAAM5C,oBAAoB,CAC/B3C,WACAJ;QAIA,IAAI8C;QAKJ,IAAI1C,UAAU,QAAQ,CAAC,UACrB,OAAQJ;YACN,KAAK4F,mCAAAA,YAAAA,CAAAA,MAAmB;gBACtB9C,iBAAiB+C,6BAAAA,YAAYA;gBAC7B;YACF,KAAKD,mCAAAA,YAAAA,CAAAA,eAA4B;gBAC/B9C,iBAAiBgD,+BAAAA,aAAaA;gBAC9B;YACF,KAAKF,mCAAAA,YAAAA,CAAAA,IAAiB;gBACpB9C,iBAAiBiD,gCAAAA,UAAUA;gBAC3B;YACF,KAAKH,mCAAAA,YAAAA,CAAAA,YAAyB;YAC9B,KAAKA,mCAAAA,YAAAA,CAAAA,gBAA6B;gBAChC9C,iBAAiB;oBAAE,MAAMkD,kCAAAA,gBAAAA,CAAAA,IAAqB;gBAAC;gBAC/C;QACJ;QAIF,IAAI5F,AAAc,wBAAdA,WACF0C,iBAAiB;YAAE,MAAMkD,kCAAAA,gBAAAA,CAAAA,IAAqB;QAAC;QAGjD,OAAOlD;IACT;IAEO,eAAemD,oBACpBvD,QAAsC,EACtC1C,iBAA+B,EAC/BC,gBAAmC;QAEnC,MAAMiG,WAAW,MAAMzD,KAAKC,UAAU1C,mBAAmBC;QACzD2B,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOsE,UAAU;QACjB,MAAMC,cAAcC,cAAcF,SAAS,OAAO,EAAEjG;QACpD,OAAO;YAAE,SAASkG;YAAa,OAAOD,SAAS,KAAK;QAAC;IACvD;IAEO,eAAeG,2BACpBC,IAAY,EACZtG,iBAA+B,EAC/BC,gBAAmC;QAEnC,MAAM,EAAEwD,OAAO,EAAEE,KAAK,EAAE,GAAG,MAAMlB,KAC/B6D,MACAtG,mBACAC;QAEF,OAAO;YAAEwD;YAASE;QAAM;IAC1B;IAEO,SAAS4C,yBAAyBL,QAAgB;QACvD,IAAI;YAEF,MAAMM,YAAYN,SAAS,KAAK,CAAC;YACjC,IAAIM,WACF,OAAOA,SAAS,CAAC,EAAE;YAIrB,MAAMC,iBAAiBP,SAAS,KAAK,CACnC;YAEF,IAAIO,gBACF,OAAOA,cAAc,CAAC,EAAE;YAI1B,MAAMC,gBAAgBR,SAAS,KAAK,CAAC;YACrC,IAAIQ,eACF,OAAOA,aAAa,CAAC,EAAE;QAE3B,EAAE,OAAM,CAAC;QAET,OAAOR;IACT;IAEO,SAASS,yBAAyBC,KAAa;QACpD,IAAIA,MAAM,QAAQ,CAAC,SAEjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;QAG5C,OAAOA;IACT;IAEO,SAASR,cACdQ,KAAa,EACb3G,gBAAmC;QAEnC,MAAM4G,kBAAkBN,yBAAyBK;QAEjD,IAAIC,QAAAA,kBAAAA,KAAAA,IAAAA,gBAAiB,KAAK,CAAC,oBAAoB;gBACtCC;YAAP,OAAO,QAAAA,CAAAA,yBAAAA,gBACJ,KAAK,CAAC,kBAAiB,IADnBA,KAAAA,IAAAA,uBAEH,KAAK,CAAC,GACP,GAAG,CAAC/C;QACT;QACA,IAAI;YACF,OAAOiB,KAAK,KAAK,CAAC6B;QACpB,EAAE,OAAM,CAAC;QACT,IAAI;YACF,OAAO7B,KAAK,KAAK,CAAC+B,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAWF;QAC/B,EAAE,OAAOnB,GAAG,CAAC;QAEb,IACE5B,AAAmC,oBAAnCA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa7D,qBACb6D,AAAmC,kBAAnCA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa7D,mBACb;YACA,MAAM+G,aAAaL,yBAAyBE;YAC5C,OAAO7B,KAAK,KAAK,CAAC+B,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAWC;QAC/B;QACA,MAAM3E,MAAM,CAAC,+BAA+B,EAAEuE,OAAO;IACvD"}

package/dist/lib/ai-model/ui-tars-planning.js

@@ -68,33 +68,23 @@ async function vlmPlanning(options) {
         },
         modelVer: modelVer || void 0
     });
-    debug('modelVer', modelVer, 'parsed', JSON.stringify(parsed));
+    debug('ui-tars modelVer', modelVer, ', parsed', JSON.stringify(parsed));
     const transformActions = [];
     parsed.forEach((action)=>{
         if ('click' === action.action_type) {
             (0, utils_namespaceObject.assert)(action.action_inputs.start_box, 'start_box is required');
             const point = getPoint(action.action_inputs.start_box, size);
-            transformActions.push({
-                type: 'Locate',
-                param: {},
-                locate: {
-                    prompt: action.thought || '',
-                    bbox: pointToBbox({
-                        x: point[0],
-                        y: point[1]
-                    }, size.width, size.height)
-                }
-            });
             transformActions.push({
                 type: 'Tap',
-                locate: {
-                    prompt: action.thought || '',
-                    bbox: pointToBbox({
-                        x: point[0],
-                        y: point[1]
-                    }, size.width, size.height)
-                },
-                param: action.thought || ''
+                param: {
+                    locate: {
+                        prompt: action.thought || '',
+                        bbox: pointToBbox({
+                            x: point[0],
+                            y: point[1]
+                        }, size.width, size.height)
+                    }
+                }
             });
         } else if ('drag' === action.action_type) {
             (0, utils_namespaceObject.assert)(action.action_inputs.start_box, 'start_box is required');
@@ -102,18 +92,23 @@ async function vlmPlanning(options) {
             const startPoint = getPoint(action.action_inputs.start_box, size);
             const endPoint = getPoint(action.action_inputs.end_box, size);
             transformActions.push({
-                type: 'Drag',
+                type: 'DragAndDrop',
                 param: {
-                    start_box: {
-                        x: startPoint[0],
-                        y: startPoint[1]
+                    from: {
+                        prompt: action.thought || '',
+                        bbox: pointToBbox({
+                            x: startPoint[0],
+                            y: startPoint[1]
+                        }, size.width, size.height)
                     },
-                    end_box: {
-                        x: endPoint[0],
-                        y: endPoint[1]
+                    to: {
+                        prompt: action.thought || '',
+                        bbox: pointToBbox({
+                            x: endPoint[0],
+                            y: endPoint[1]
+                        }, size.width, size.height)
                     }
                 },
-                locate: null,
                 thought: action.thought || ''
             });
         } else if ('type' === action.action_type) transformActions.push({
@@ -121,7 +116,6 @@ async function vlmPlanning(options) {
             param: {
                 value: action.action_inputs.content
             },
-            locate: null,
             thought: action.thought || ''
         });
         else if ('scroll' === action.action_type) transformActions.push({
@@ -129,13 +123,11 @@ async function vlmPlanning(options) {
             param: {
                 direction: action.action_inputs.direction
             },
-            locate: null,
             thought: action.thought || ''
         });
         else if ('finished' === action.action_type) transformActions.push({
             type: 'Finished',
             param: {},
-            locate: null,
             thought: action.thought || ''
         });
         else if ('hotkey' === action.action_type) if (action.action_inputs.key) {
@@ -143,9 +135,8 @@ async function vlmPlanning(options) {
             transformActions.push({
                 type: 'KeyboardPress',
                 param: {
-                    value: keys
+                    keyName: keys
                 },
-                locate: null,
                 thought: action.thought || ''
             });
         } else console.warn('No key found in action: hotkey. Will not perform action.');
@@ -154,56 +145,8 @@ async function vlmPlanning(options) {
             param: {
                 timeMs: 1000
             },
-            locate: null,
-            thought: action.thought || ''
-        });
-        else if ('androidBackButton' === action.action_type) transformActions.push({
-            type: 'AndroidBackButton',
-            param: {},
-            locate: null,
-            thought: action.thought || ''
-        });
-        else if ('androidHomeButton' === action.action_type) transformActions.push({
-            type: 'AndroidHomeButton',
-            param: {},
-            locate: null,
             thought: action.thought || ''
         });
-        else if ('androidRecentAppsButton' === action.action_type) transformActions.push({
-            type: 'AndroidRecentAppsButton',
-            param: {}
-        });
-        else if ('androidLongPress' === action.action_type) {
-            (0, utils_namespaceObject.assert)(action.action_inputs.start_coords, 'start_coords is required for androidLongPress');
-            const point = action.action_inputs.start_coords;
-            transformActions.push({
-                type: 'AndroidLongPress',
-                param: {
-                    x: point[0],
-                    y: point[1],
-                    duration: 1000
-                },
-                locate: null,
-                thought: action.thought || ''
-            });
-        } else if ('androidPull' === action.action_type) {
-            const pullDirection = action.action_inputs.direction || 'down';
-            const startPoint = action.action_inputs.start_coords ? {
-                x: action.action_inputs.start_coords[0],
-                y: action.action_inputs.start_coords[1]
-            } : void 0;
-            transformActions.push({
-                type: 'AndroidPull',
-                param: {
-                    direction: pullDirection,
-                    startPoint,
-                    distance: action.action_inputs.distance,
-                    duration: action.action_inputs.duration || 500
-                },
-                locate: null,
-                thought: action.thought || ''
-            });
-        }
     });
     if (0 === transformActions.length) throw new Error(`No actions found, response: ${res.content}`, {
         cause: {
@@ -211,6 +154,7 @@ async function vlmPlanning(options) {
             parsed
         }
     });
+    debug('transformActions', JSON.stringify(transformActions, null, 2));
     return {
         actions: transformActions,
         actionsFromModel: parsed,

package/dist/lib/ai-model/ui-tars-planning.js.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/ui-tars-planning.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/ui-tars-planning.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n  AIUsageInfo,\n  MidsceneYamlFlowItem,\n  PlanningAction,\n  Size,\n} from '@/types';\nimport {\n  type IModelPreferences,\n  UITarsModelVersion,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { resizeImgBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { transformHotkeyInput } from '@midscene/shared/us-keyboard-layout';\nimport { assert } from '@midscene/shared/utils';\nimport { actionParser } from '@ui-tars/action-parser';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport { AIActionType } from './common';\nimport { getSummary, getUiTarsPlanningPrompt } from './prompt/ui-tars-planning';\nimport { call } from './service-caller/index';\ntype ActionType =\n  | 'click'\n  | 'drag'\n  | 'type'\n  | 'hotkey'\n  | 'finished'\n  | 'scroll'\n  | 'wait'\n  | 'androidBackButton'\n  | 'androidHomeButton'\n  | 'androidRecentAppsButton'\n  | 'androidLongPress'\n  | 'androidPull';\n\nconst debug = getDebug('ui-tars-planning');\nconst bboxSize = 10;\nconst pointToBbox = (\n  point: { x: number; y: number },\n  width: number,\n  height: number,\n): [number, number, number, number] => {\n  return [\n    Math.round(Math.max(point.x - bboxSize / 2, 0)),\n    Math.round(Math.max(point.y - bboxSize / 2, 0)),\n    Math.round(Math.min(point.x + bboxSize / 2, width)),\n    Math.round(Math.min(point.y + bboxSize / 2, height)),\n  ];\n};\n\nexport async function vlmPlanning(options: {\n  userInstruction: string;\n  conversationHistory: ChatCompletionMessageParam[];\n  size: { width: number; height: number };\n  modelPreferences: IModelPreferences;\n}): Promise<{\n  actions: PlanningAction<any>[];\n  actionsFromModel: ReturnType<typeof actionParser>['parsed'];\n  action_summary: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n}> {\n  const { conversationHistory, userInstruction, size, modelPreferences } =\n    options;\n  const systemPrompt = getUiTarsPlanningPrompt() + userInstruction;\n\n  const res = await call(\n    [\n      {\n        role: 'user',\n        content: systemPrompt,\n      },\n      ...conversationHistory,\n    ],\n    AIActionType.INSPECT_ELEMENT,\n    modelPreferences,\n  );\n  const convertedText = convertBboxToCoordinates(res.content);\n\n  const modelVer = uiTarsModelVersion(modelPreferences);\n\n  const { parsed } = actionParser({\n    prediction: convertedText,\n    factor: [1000, 1000],\n    screenContext: {\n      width: size.width,\n      height: size.height,\n    },\n    modelVer: modelVer || undefined,\n  });\n\n  debug('modelVer', modelVer, 'parsed', JSON.stringify(parsed));\n\n  const transformActions: PlanningAction[] = [];\n  parsed.forEach((action) => {\n    if (action.action_type === 'click') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n      transformActions.push({\n        type: 'Locate',\n        param: {},\n        locate: {\n          prompt: action.thought || '',\n          bbox: pointToBbox(\n            { x: point[0], y: point[1] },\n            size.width,\n            size.height,\n          ),\n        },\n      });\n      transformActions.push({\n        type: 'Tap',\n        locate: {\n          prompt: action.thought || '',\n          bbox: pointToBbox(\n            { x: point[0], y: point[1] },\n            size.width,\n            size.height,\n          ),\n        },\n        param: action.thought || '',\n      });\n    } else if (action.action_type === 'drag') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      assert(action.action_inputs.end_box, 'end_box is required');\n      const startPoint = getPoint(action.action_inputs.start_box, size);\n      const endPoint = getPoint(action.action_inputs.end_box, size);\n      transformActions.push({\n        type: 'Drag',\n        param: {\n          start_box: { x: startPoint[0], y: startPoint[1] },\n          end_box: { x: endPoint[0], y: endPoint[1] },\n        },\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'type') {\n      transformActions.push({\n        type: 'Input',\n        param: {\n          value: action.action_inputs.content,\n        },\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'scroll') {\n      transformActions.push({\n        type: 'Scroll',\n        param: {\n          direction: action.action_inputs.direction,\n        },\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'finished') {\n      transformActions.push({\n        type: 'Finished',\n        param: {},\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'hotkey') {\n      if (!action.action_inputs.key) {\n        console.warn(\n          'No key found in action: hotkey. Will not perform action.',\n        );\n      } else {\n        const keys = transformHotkeyInput(action.action_inputs.key);\n\n        transformActions.push({\n          type: 'KeyboardPress',\n          param: {\n            value: keys,\n          },\n          locate: null,\n          thought: action.thought || '',\n        });\n      }\n    } else if (action.action_type === 'wait') {\n      transformActions.push({\n        type: 'Sleep',\n        param: {\n          timeMs: 1000,\n        },\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'androidBackButton') {\n      transformActions.push({\n        type: 'AndroidBackButton',\n        param: {},\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'androidHomeButton') {\n      transformActions.push({\n        type: 'AndroidHomeButton',\n        param: {},\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'androidRecentAppsButton') {\n      transformActions.push({\n        type: 'AndroidRecentAppsButton',\n        param: {},\n      });\n    } else if (action.action_type === 'androidLongPress') {\n      assert(\n        action.action_inputs.start_coords,\n        'start_coords is required for androidLongPress',\n      );\n      const point = action.action_inputs.start_coords;\n      transformActions.push({\n        type: 'AndroidLongPress',\n        param: {\n          x: point[0],\n          y: point[1],\n          duration: 1000,\n        },\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'androidPull') {\n      const pullDirection = action.action_inputs.direction || 'down';\n      const startPoint = action.action_inputs.start_coords\n        ? {\n            x: action.action_inputs.start_coords[0],\n            y: action.action_inputs.start_coords[1],\n          }\n        : undefined;\n\n      transformActions.push({\n        type: 'AndroidPull',\n        param: {\n          direction: pullDirection as 'up' | 'down',\n          startPoint,\n          distance: (action.action_inputs as any).distance,\n          duration: (action.action_inputs as any).duration || 500,\n        },\n        locate: null,\n        thought: action.thought || '',\n      });\n    }\n  });\n\n  if (transformActions.length === 0) {\n    throw new Error(`No actions found, response: ${res.content}`, {\n      cause: {\n        prediction: res.content,\n        parsed,\n      },\n    });\n  }\n\n  return {\n    actions: transformActions,\n    actionsFromModel: parsed,\n    action_summary: getSummary(res.content),\n    usage: res.usage,\n    rawResponse: JSON.stringify(res.content, undefined, 2),\n  };\n}\n\n/**\n * Converts bounding box notation to coordinate points\n * @param text - The text containing bbox tags to be converted\n * @returns The text with bbox tags replaced by coordinate points\n */\nfunction convertBboxToCoordinates(text: string): string {\n  // Match the four numbers after <bbox>\n  const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n  function replaceMatch(\n    match: string,\n    x1: string,\n    y1: string,\n    x2: string,\n    y2: string,\n  ): string {\n    // Convert strings to numbers and calculate center point\n    const x1Num = Number.parseInt(x1, 10);\n    const y1Num = Number.parseInt(y1, 10);\n    const x2Num = Number.parseInt(x2, 10);\n    const y2Num = Number.parseInt(y2, 10);\n\n    // Use Math.floor to truncate and calculate center point\n    const x = Math.floor((x1Num + x2Num) / 2);\n    const y = Math.floor((y1Num + y2Num) / 2);\n\n    // Return formatted coordinate string\n    return `(${x},${y})`;\n  }\n\n  // Remove [EOS] and replace <bbox> coordinates\n  const cleanedText = text.replace(/\\[EOS\\]/g, '');\n  return cleanedText.replace(pattern, replaceMatch).trim();\n}\n\nfunction getPoint(startBox: string, size: { width: number; height: number }) {\n  const [x, y] = JSON.parse(startBox);\n  return [x * size.width, y * size.height];\n}\n\ninterface BaseAction {\n  action_type: ActionType;\n  action_inputs: Record<string, any>;\n  reflection: string | null;\n  thought: string | null;\n}\n\ninterface ClickAction extends BaseAction {\n  action_type: 'click';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface DragAction extends BaseAction {\n  action_type: 'drag';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n    end_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface WaitAction extends BaseAction {\n  action_type: 'wait';\n  action_inputs: {\n    time: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface TypeAction extends BaseAction {\n  action_type: 'type';\n  action_inputs: {\n    content: string;\n  };\n}\n\ninterface HotkeyAction extends BaseAction {\n  action_type: 'hotkey';\n  action_inputs: {\n    key: string;\n  };\n}\n\ninterface ScrollAction extends BaseAction {\n  action_type: 'scroll';\n  action_inputs: {\n    direction: 'up' | 'down';\n  };\n}\n\ninterface FinishedAction extends BaseAction {\n  action_type: 'finished';\n  action_inputs: Record<string, never>;\n}\n\ninterface AndroidLongPressAction extends BaseAction {\n  action_type: 'androidLongPress';\n  action_inputs: {\n    start_coords: [number, number]; // Coordinates for long press\n    duration?: number; // Duration in milliseconds\n  };\n}\n\nexport type Action =\n  | ClickAction\n  | DragAction\n  | TypeAction\n  | HotkeyAction\n  | ScrollAction\n  | FinishedAction\n  | WaitAction\n  | AndroidLongPressAction;\n\nexport async function resizeImageForUiTars(\n  imageBase64: string,\n  size: Size,\n  modelPreferences: IModelPreferences,\n) {\n  if (\n    vlLocateMode(modelPreferences) === 'vlm-ui-tars' &&\n    uiTarsModelVersion(modelPreferences) === UITarsModelVersion.V1_5\n  ) {\n    debug('ui-tars-v1.5, will check image size', size);\n    const currentPixels = size.width * size.height;\n    const maxPixels = 16384 * 28 * 28; //\n    if (currentPixels > maxPixels) {\n      const resizeFactor = Math.sqrt(maxPixels / currentPixels);\n      const newWidth = Math.floor(size.width * resizeFactor);\n      const newHeight = Math.floor(size.height * resizeFactor);\n      debug(\n        'resize image for ui-tars, new width: %s, new height: %s',\n        newWidth,\n        newHeight,\n      );\n      const resizedImage = await resizeImgBase64(imageBase64, {\n        width: newWidth,\n        height: newHeight,\n      });\n      return resizedImage;\n    }\n  }\n  return imageBase64;\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","bboxSize","pointToBbox","point","width","height","Math","vlmPlanning","options","conversationHistory","userInstruction","size","modelPreferences","systemPrompt","getUiTarsPlanningPrompt","res","call","AIActionType","convertedText","convertBboxToCoordinates","modelVer","uiTarsModelVersion","parsed","actionParser","undefined","JSON","transformActions","action","assert","getPoint","startPoint","endPoint","keys","transformHotkeyInput","console","pullDirection","Error","getSummary","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","y","cleanedText","startBox","resizeImageForUiTars","imageBase64","vlLocateMode","UITarsModelVersion","currentPixels","maxPixels","resizeFactor","newWidth","newHeight","resizedImage","resizeImgBase64"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;AC6BA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AACvB,MAAMC,WAAW;AACjB,MAAMC,cAAc,CAClBC,OACAC,OACAC,SAEO;QACLC,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGG;QAC5CE,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGI;KAC7C;AAGI,eAAeE,YAAYC,OAKjC;IAQC,MAAM,EAAEC,mBAAmB,EAAEC,eAAe,EAAEC,IAAI,EAAEC,gBAAgB,EAAE,GACpEJ;IACF,MAAMK,eAAeC,AAAAA,IAAAA,oCAAAA,uBAAAA,AAAAA,MAA4BJ;IAEjD,MAAMK,MAAM,MAAMC,AAAAA,IAAAA,yBAAAA,IAAAA,AAAAA,EAChB;QACE;YACE,MAAM;YACN,SAASH;QACX;WACGJ;KACJ,EACDQ,mCAAAA,YAAAA,CAAAA,eAA4B,EAC5BL;IAEF,MAAMM,gBAAgBC,yBAAyBJ,IAAI,OAAO;IAE1D,MAAMK,WAAWC,AAAAA,IAAAA,oBAAAA,kBAAAA,AAAAA,EAAmBT;IAEpC,MAAM,EAAEU,MAAM,EAAE,GAAGC,AAAAA,IAAAA,8BAAAA,YAAAA,AAAAA,EAAa;QAC9B,YAAYL;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOP,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUS,YAAYI;IACxB;IAEAzB,MAAM,YAAYqB,UAAU,UAAUK,KAAK,SAAS,CAACH;IAErD,MAAMI,mBAAqC,EAAE;IAC7CJ,OAAO,OAAO,CAAC,CAACK;QACd,IAAIA,AAAuB,YAAvBA,OAAO,WAAW,EAAc;YAClCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAMxB,QAAQ0B,SAASF,OAAO,aAAa,CAAC,SAAS,EAAEhB;YACvDe,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO,CAAC;gBACR,QAAQ;oBACN,QAAQC,OAAO,OAAO,IAAI;oBAC1B,MAAMzB,YACJ;wBAAE,GAAGC,KAAK,CAAC,EAAE;wBAAE,GAAGA,KAAK,CAAC,EAAE;oBAAC,GAC3BQ,KAAK,KAAK,EACVA,KAAK,MAAM;gBAEf;YACF;YACAe,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,QAAQ;oBACN,QAAQC,OAAO,OAAO,IAAI;oBAC1B,MAAMzB,YACJ;wBAAE,GAAGC,KAAK,CAAC,EAAE;wBAAE,GAAGA,KAAK,CAAC,EAAE;oBAAC,GAC3BQ,KAAK,KAAK,EACVA,KAAK,MAAM;gBAEf;gBACA,OAAOgB,OAAO,OAAO,IAAI;YAC3B;QACF,OAAO,IAAIA,AAAuB,WAAvBA,OAAO,WAAW,EAAa;YACxCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMG,aAAaD,SAASF,OAAO,aAAa,CAAC,SAAS,EAAEhB;YAC5D,MAAMoB,WAAWF,SAASF,OAAO,aAAa,CAAC,OAAO,EAAEhB;YACxDe,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,WAAW;wBAAE,GAAGI,UAAU,CAAC,EAAE;wBAAE,GAAGA,UAAU,CAAC,EAAE;oBAAC;oBAChD,SAAS;wBAAE,GAAGC,QAAQ,CAAC,EAAE;wBAAE,GAAGA,QAAQ,CAAC,EAAE;oBAAC;gBAC5C;gBACA,QAAQ;gBACR,SAASJ,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIA,AAAuB,WAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,OAAOC,OAAO,aAAa,CAAC,OAAO;YACrC;YACA,QAAQ;YACR,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,aAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWC,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,QAAQ;YACR,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,eAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO,CAAC;YACR,QAAQ;YACR,SAASC,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,aAAvBA,OAAO,WAAW,EAC3B,IAAKA,OAAO,aAAa,CAAC,GAAG,EAItB;YACL,MAAMK,OAAOC,AAAAA,IAAAA,mCAAAA,oBAAAA,AAAAA,EAAqBN,OAAO,aAAa,CAAC,GAAG;YAE1DD,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,OAAOM;gBACT;gBACA,QAAQ;gBACR,SAASL,OAAO,OAAO,IAAI;YAC7B;QACF,OAdEO,QAAQ,IAAI,CACV;aAcC,IAAIP,AAAuB,WAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,QAAQ;YACR,SAASC,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,wBAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO,CAAC;YACR,QAAQ;YACR,SAASC,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,wBAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO,CAAC;YACR,QAAQ;YACR,SAASC,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,8BAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO,CAAC;QACV;aACK,IAAIC,AAAuB,uBAAvBA,OAAO,WAAW,EAAyB;YACpDC,IAAAA,sBAAAA,MAAAA,AAAAA,EACED,OAAO,aAAa,CAAC,YAAY,EACjC;YAEF,MAAMxB,QAAQwB,OAAO,aAAa,CAAC,YAAY;YAC/CD,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,GAAGvB,KAAK,CAAC,EAAE;oBACX,GAAGA,KAAK,CAAC,EAAE;oBACX,UAAU;gBACZ;gBACA,QAAQ;gBACR,SAASwB,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIA,AAAuB,kBAAvBA,OAAO,WAAW,EAAoB;YAC/C,MAAMQ,gBAAgBR,OAAO,aAAa,CAAC,SAAS,IAAI;YACxD,MAAMG,aAAaH,OAAO,aAAa,CAAC,YAAY,GAChD;gBACE,GAAGA,OAAO,aAAa,CAAC,YAAY,CAAC,EAAE;gBACvC,GAAGA,OAAO,aAAa,CAAC,YAAY,CAAC,EAAE;YACzC,IACAH;YAEJE,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,WAAWS;oBACXL;oBACA,UAAWH,OAAO,aAAa,CAAS,QAAQ;oBAChD,UAAWA,OAAO,aAAa,CAAS,QAAQ,IAAI;gBACtD;gBACA,QAAQ;gBACR,SAASA,OAAO,OAAO,IAAI;YAC7B;QACF;IACF;IAEA,IAAID,AAA4B,MAA5BA,iBAAiB,MAAM,EACzB,MAAM,IAAIU,MAAM,CAAC,4BAA4B,EAAErB,IAAI,OAAO,EAAE,EAAE;QAC5D,OAAO;YACL,YAAYA,IAAI,OAAO;YACvBO;QACF;IACF;IAGF,OAAO;QACL,SAASI;QACT,kBAAkBJ;QAClB,gBAAgBe,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAWtB,IAAI,OAAO;QACtC,OAAOA,IAAI,KAAK;QAChB,aAAaU,KAAK,SAAS,CAACV,IAAI,OAAO,EAAES,QAAW;IACtD;AACF;AAOA,SAASL,yBAAyBmB,IAAY;IAE5C,MAAMC,UAAU;IAEhB,SAASC,aACPC,KAAa,EACbC,EAAU,EACVC,EAAU,EACVC,EAAU,EACVC,EAAU;QAGV,MAAMC,QAAQC,OAAO,QAAQ,CAACL,IAAI;QAClC,MAAMM,QAAQD,OAAO,QAAQ,CAACJ,IAAI;QAClC,MAAMM,QAAQF,OAAO,QAAQ,CAACH,IAAI;QAClC,MAAMM,QAAQH,OAAO,QAAQ,CAACF,IAAI;QAGlC,MAAMM,IAAI7C,KAAK,KAAK,CAAEwC,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMG,IAAI9C,KAAK,KAAK,CAAE0C,AAAAA,CAAAA,QAAQE,KAAI,IAAK;QAGvC,OAAO,CAAC,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC,CAAC;IACtB;IAGA,MAAMC,cAAcf,KAAK,OAAO,CAAC,YAAY;IAC7C,OAAOe,YAAY,OAAO,CAACd,SAASC,cAAc,IAAI;AACxD;AAEA,SAASX,SAASyB,QAAgB,EAAE3C,IAAuC;IACzE,MAAM,CAACwC,GAAGC,EAAE,GAAG3B,KAAK,KAAK,CAAC6B;IAC1B,OAAO;QAACH,IAAIxC,KAAK,KAAK;QAAEyC,IAAIzC,KAAK,MAAM;KAAC;AAC1C;AA2EO,eAAe4C,qBACpBC,WAAmB,EACnB7C,IAAU,EACVC,gBAAmC;IAEnC,IACE6C,AAAmC,kBAAnCA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa7C,qBACbS,AAAAA,IAAAA,oBAAAA,kBAAAA,AAAAA,EAAmBT,sBAAsB8C,oBAAAA,kBAAAA,CAAAA,IAAuB,EAChE;QACA3D,MAAM,uCAAuCY;QAC7C,MAAMgD,gBAAgBhD,KAAK,KAAK,GAAGA,KAAK,MAAM;QAC9C,MAAMiD,YAAY;QAClB,IAAID,gBAAgBC,WAAW;YAC7B,MAAMC,eAAevD,KAAK,IAAI,CAACsD,YAAYD;YAC3C,MAAMG,WAAWxD,KAAK,KAAK,CAACK,KAAK,KAAK,GAAGkD;YACzC,MAAME,YAAYzD,KAAK,KAAK,CAACK,KAAK,MAAM,GAAGkD;YAC3C9D,MACE,2DACA+D,UACAC;YAEF,MAAMC,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,eAAAA,AAAAA,EAAgBT,aAAa;gBACtD,OAAOM;gBACP,QAAQC;YACV;YACA,OAAOC;QACT;IACF;IACA,OAAOR;AACT"}
+{"version":3,"file":"ai-model/ui-tars-planning.js","sources":["webpack://@midscene/core/webpack/runtime/define_property_getters","webpack://@midscene/core/webpack/runtime/has_own_property","webpack://@midscene/core/webpack/runtime/make_namespace_object","webpack://@midscene/core/./src/ai-model/ui-tars-planning.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import type {\n  AIUsageInfo,\n  MidsceneYamlFlowItem,\n  PlanningAction,\n  Size,\n} from '@/types';\nimport {\n  type IModelPreferences,\n  UITarsModelVersion,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { resizeImgBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { transformHotkeyInput } from '@midscene/shared/us-keyboard-layout';\nimport { assert } from '@midscene/shared/utils';\nimport { actionParser } from '@ui-tars/action-parser';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport { AIActionType } from './common';\nimport { getSummary, getUiTarsPlanningPrompt } from './prompt/ui-tars-planning';\nimport { call } from './service-caller/index';\ntype ActionType =\n  | 'click'\n  | 'drag'\n  | 'type'\n  | 'hotkey'\n  | 'finished'\n  | 'scroll'\n  | 'wait';\n\nconst debug = getDebug('ui-tars-planning');\nconst bboxSize = 10;\nconst pointToBbox = (\n  point: { x: number; y: number },\n  width: number,\n  height: number,\n): [number, number, number, number] => {\n  return [\n    Math.round(Math.max(point.x - bboxSize / 2, 0)),\n    Math.round(Math.max(point.y - bboxSize / 2, 0)),\n    Math.round(Math.min(point.x + bboxSize / 2, width)),\n    Math.round(Math.min(point.y + bboxSize / 2, height)),\n  ];\n};\n\nexport async function vlmPlanning(options: {\n  userInstruction: string;\n  conversationHistory: ChatCompletionMessageParam[];\n  size: { width: number; height: number };\n  modelPreferences: IModelPreferences;\n}): Promise<{\n  actions: PlanningAction<any>[];\n  actionsFromModel: ReturnType<typeof actionParser>['parsed'];\n  action_summary: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n}> {\n  const { conversationHistory, userInstruction, size, modelPreferences } =\n    options;\n  const systemPrompt = getUiTarsPlanningPrompt() + userInstruction;\n\n  const res = await call(\n    [\n      {\n        role: 'user',\n        content: systemPrompt,\n      },\n      ...conversationHistory,\n    ],\n    AIActionType.INSPECT_ELEMENT,\n    modelPreferences,\n  );\n  const convertedText = convertBboxToCoordinates(res.content);\n\n  const modelVer = uiTarsModelVersion(modelPreferences);\n\n  const { parsed } = actionParser({\n    prediction: convertedText,\n    factor: [1000, 1000],\n    screenContext: {\n      width: size.width,\n      height: size.height,\n    },\n    modelVer: modelVer || undefined,\n  });\n\n  debug('ui-tars modelVer', modelVer, ', parsed', JSON.stringify(parsed));\n\n  const transformActions: PlanningAction[] = [];\n  parsed.forEach((action) => {\n    if (action.action_type === 'click') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n      transformActions.push({\n        type: 'Tap',\n        param: {\n          locate: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: point[0], y: point[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n      });\n    } else if (action.action_type === 'drag') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      assert(action.action_inputs.end_box, 'end_box is required');\n      const startPoint = getPoint(action.action_inputs.start_box, size);\n      const endPoint = getPoint(action.action_inputs.end_box, size);\n      transformActions.push({\n        type: 'DragAndDrop',\n        param: {\n          from: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: startPoint[0], y: startPoint[1] },\n              size.width,\n              size.height,\n            ),\n          },\n          to: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: endPoint[0], y: endPoint[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'type') {\n      transformActions.push({\n        type: 'Input',\n        param: {\n          value: action.action_inputs.content,\n        },\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'scroll') {\n      transformActions.push({\n        type: 'Scroll',\n        param: {\n          direction: action.action_inputs.direction,\n        },\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'finished') {\n      transformActions.push({\n        type: 'Finished',\n        param: {},\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'hotkey') {\n      if (!action.action_inputs.key) {\n        console.warn(\n          'No key found in action: hotkey. Will not perform action.',\n        );\n      } else {\n        const keys = transformHotkeyInput(action.action_inputs.key);\n\n        transformActions.push({\n          type: 'KeyboardPress',\n          param: {\n            keyName: keys,\n          },\n          thought: action.thought || '',\n        });\n      }\n    } else if (action.action_type === 'wait') {\n      transformActions.push({\n        type: 'Sleep',\n        param: {\n          timeMs: 1000,\n        },\n        thought: action.thought || '',\n      });\n    }\n  });\n\n  if (transformActions.length === 0) {\n    throw new Error(`No actions found, response: ${res.content}`, {\n      cause: {\n        prediction: res.content,\n        parsed,\n      },\n    });\n  }\n\n  debug('transformActions', JSON.stringify(transformActions, null, 2));\n\n  return {\n    actions: transformActions,\n    actionsFromModel: parsed,\n    action_summary: getSummary(res.content),\n    usage: res.usage,\n    rawResponse: JSON.stringify(res.content, undefined, 2),\n  };\n}\n\n/**\n * Converts bounding box notation to coordinate points\n * @param text - The text containing bbox tags to be converted\n * @returns The text with bbox tags replaced by coordinate points\n */\nfunction convertBboxToCoordinates(text: string): string {\n  // Match the four numbers after <bbox>\n  const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n  function replaceMatch(\n    match: string,\n    x1: string,\n    y1: string,\n    x2: string,\n    y2: string,\n  ): string {\n    // Convert strings to numbers and calculate center point\n    const x1Num = Number.parseInt(x1, 10);\n    const y1Num = Number.parseInt(y1, 10);\n    const x2Num = Number.parseInt(x2, 10);\n    const y2Num = Number.parseInt(y2, 10);\n\n    // Use Math.floor to truncate and calculate center point\n    const x = Math.floor((x1Num + x2Num) / 2);\n    const y = Math.floor((y1Num + y2Num) / 2);\n\n    // Return formatted coordinate string\n    return `(${x},${y})`;\n  }\n\n  // Remove [EOS] and replace <bbox> coordinates\n  const cleanedText = text.replace(/\\[EOS\\]/g, '');\n  return cleanedText.replace(pattern, replaceMatch).trim();\n}\n\nfunction getPoint(startBox: string, size: { width: number; height: number }) {\n  const [x, y] = JSON.parse(startBox);\n  return [x * size.width, y * size.height];\n}\n\ninterface BaseAction {\n  action_type: ActionType;\n  action_inputs: Record<string, any>;\n  reflection: string | null;\n  thought: string | null;\n}\n\ninterface ClickAction extends BaseAction {\n  action_type: 'click';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface DragAction extends BaseAction {\n  action_type: 'drag';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n    end_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface WaitAction extends BaseAction {\n  action_type: 'wait';\n  action_inputs: {\n    time: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface TypeAction extends BaseAction {\n  action_type: 'type';\n  action_inputs: {\n    content: string;\n  };\n}\n\ninterface HotkeyAction extends BaseAction {\n  action_type: 'hotkey';\n  action_inputs: {\n    key: string;\n  };\n}\n\ninterface ScrollAction extends BaseAction {\n  action_type: 'scroll';\n  action_inputs: {\n    direction: 'up' | 'down';\n  };\n}\n\ninterface FinishedAction extends BaseAction {\n  action_type: 'finished';\n  action_inputs: Record<string, never>;\n}\n\nexport type Action =\n  | ClickAction\n  | DragAction\n  | TypeAction\n  | HotkeyAction\n  | ScrollAction\n  | FinishedAction\n  | WaitAction;\n\nexport async function resizeImageForUiTars(\n  imageBase64: string,\n  size: Size,\n  modelPreferences: IModelPreferences,\n) {\n  if (\n    vlLocateMode(modelPreferences) === 'vlm-ui-tars' &&\n    uiTarsModelVersion(modelPreferences) === UITarsModelVersion.V1_5\n  ) {\n    debug('ui-tars-v1.5, will check image size', size);\n    const currentPixels = size.width * size.height;\n    const maxPixels = 16384 * 28 * 28; //\n    if (currentPixels > maxPixels) {\n      const resizeFactor = Math.sqrt(maxPixels / currentPixels);\n      const newWidth = Math.floor(size.width * resizeFactor);\n      const newHeight = Math.floor(size.height * resizeFactor);\n      debug(\n        'resize image for ui-tars, new width: %s, new height: %s',\n        newWidth,\n        newHeight,\n      );\n      const resizedImage = await resizeImgBase64(imageBase64, {\n        width: newWidth,\n        height: newHeight,\n      });\n      return resizedImage;\n    }\n  }\n  return imageBase64;\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","bboxSize","pointToBbox","point","width","height","Math","vlmPlanning","options","conversationHistory","userInstruction","size","modelPreferences","systemPrompt","getUiTarsPlanningPrompt","res","call","AIActionType","convertedText","convertBboxToCoordinates","modelVer","uiTarsModelVersion","parsed","actionParser","undefined","JSON","transformActions","action","assert","getPoint","startPoint","endPoint","keys","transformHotkeyInput","console","Error","getSummary","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","y","cleanedText","startBox","resizeImageForUiTars","imageBase64","vlLocateMode","UITarsModelVersion","currentPixels","maxPixels","resizeFactor","newWidth","newHeight","resizedImage","resizeImgBase64"],"mappings":";;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;ACwBA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AACvB,MAAMC,WAAW;AACjB,MAAMC,cAAc,CAClBC,OACAC,OACAC,SAEO;QACLC,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGG;QAC5CE,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGI;KAC7C;AAGI,eAAeE,YAAYC,OAKjC;IAQC,MAAM,EAAEC,mBAAmB,EAAEC,eAAe,EAAEC,IAAI,EAAEC,gBAAgB,EAAE,GACpEJ;IACF,MAAMK,eAAeC,AAAAA,IAAAA,oCAAAA,uBAAAA,AAAAA,MAA4BJ;IAEjD,MAAMK,MAAM,MAAMC,AAAAA,IAAAA,yBAAAA,IAAAA,AAAAA,EAChB;QACE;YACE,MAAM;YACN,SAASH;QACX;WACGJ;KACJ,EACDQ,mCAAAA,YAAAA,CAAAA,eAA4B,EAC5BL;IAEF,MAAMM,gBAAgBC,yBAAyBJ,IAAI,OAAO;IAE1D,MAAMK,WAAWC,AAAAA,IAAAA,oBAAAA,kBAAAA,AAAAA,EAAmBT;IAEpC,MAAM,EAAEU,MAAM,EAAE,GAAGC,AAAAA,IAAAA,8BAAAA,YAAAA,AAAAA,EAAa;QAC9B,YAAYL;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOP,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUS,YAAYI;IACxB;IAEAzB,MAAM,oBAAoBqB,UAAU,YAAYK,KAAK,SAAS,CAACH;IAE/D,MAAMI,mBAAqC,EAAE;IAC7CJ,OAAO,OAAO,CAAC,CAACK;QACd,IAAIA,AAAuB,YAAvBA,OAAO,WAAW,EAAc;YAClCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAMxB,QAAQ0B,SAASF,OAAO,aAAa,CAAC,SAAS,EAAEhB;YACvDe,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQC,OAAO,OAAO,IAAI;wBAC1B,MAAMzB,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BQ,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;YACF;QACF,OAAO,IAAIgB,AAAuB,WAAvBA,OAAO,WAAW,EAAa;YACxCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMG,aAAaD,SAASF,OAAO,aAAa,CAAC,SAAS,EAAEhB;YAC5D,MAAMoB,WAAWF,SAASF,OAAO,aAAa,CAAC,OAAO,EAAEhB;YACxDe,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,MAAM;wBACJ,QAAQC,OAAO,OAAO,IAAI;wBAC1B,MAAMzB,YACJ;4BAAE,GAAG4B,UAAU,CAAC,EAAE;4BAAE,GAAGA,UAAU,CAAC,EAAE;wBAAC,GACrCnB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;oBACA,IAAI;wBACF,QAAQgB,OAAO,OAAO,IAAI;wBAC1B,MAAMzB,YACJ;4BAAE,GAAG6B,QAAQ,CAAC,EAAE;4BAAE,GAAGA,QAAQ,CAAC,EAAE;wBAAC,GACjCpB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASgB,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIA,AAAuB,WAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,OAAOC,OAAO,aAAa,CAAC,OAAO;YACrC;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,aAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWC,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,eAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO,CAAC;YACR,SAASC,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,aAAvBA,OAAO,WAAW,EAC3B,IAAKA,OAAO,aAAa,CAAC,GAAG,EAItB;YACL,MAAMK,OAAOC,AAAAA,IAAAA,mCAAAA,oBAAAA,AAAAA,EAAqBN,OAAO,aAAa,CAAC,GAAG;YAE1DD,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,SAASM;gBACX;gBACA,SAASL,OAAO,OAAO,IAAI;YAC7B;QACF,OAbEO,QAAQ,IAAI,CACV;aAaC,IAAIP,AAAuB,WAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,SAASC,OAAO,OAAO,IAAI;QAC7B;IAEJ;IAEA,IAAID,AAA4B,MAA5BA,iBAAiB,MAAM,EACzB,MAAM,IAAIS,MAAM,CAAC,4BAA4B,EAAEpB,IAAI,OAAO,EAAE,EAAE;QAC5D,OAAO;YACL,YAAYA,IAAI,OAAO;YACvBO;QACF;IACF;IAGFvB,MAAM,oBAAoB0B,KAAK,SAAS,CAACC,kBAAkB,MAAM;IAEjE,OAAO;QACL,SAASA;QACT,kBAAkBJ;QAClB,gBAAgBc,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAWrB,IAAI,OAAO;QACtC,OAAOA,IAAI,KAAK;QAChB,aAAaU,KAAK,SAAS,CAACV,IAAI,OAAO,EAAES,QAAW;IACtD;AACF;AAOA,SAASL,yBAAyBkB,IAAY;IAE5C,MAAMC,UAAU;IAEhB,SAASC,aACPC,KAAa,EACbC,EAAU,EACVC,EAAU,EACVC,EAAU,EACVC,EAAU;QAGV,MAAMC,QAAQC,OAAO,QAAQ,CAACL,IAAI;QAClC,MAAMM,QAAQD,OAAO,QAAQ,CAACJ,IAAI;QAClC,MAAMM,QAAQF,OAAO,QAAQ,CAACH,IAAI;QAClC,MAAMM,QAAQH,OAAO,QAAQ,CAACF,IAAI;QAGlC,MAAMM,IAAI5C,KAAK,KAAK,CAAEuC,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMG,IAAI7C,KAAK,KAAK,CAAEyC,AAAAA,CAAAA,QAAQE,KAAI,IAAK;QAGvC,OAAO,CAAC,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC,CAAC;IACtB;IAGA,MAAMC,cAAcf,KAAK,OAAO,CAAC,YAAY;IAC7C,OAAOe,YAAY,OAAO,CAACd,SAASC,cAAc,IAAI;AACxD;AAEA,SAASV,SAASwB,QAAgB,EAAE1C,IAAuC;IACzE,MAAM,CAACuC,GAAGC,EAAE,GAAG1B,KAAK,KAAK,CAAC4B;IAC1B,OAAO;QAACH,IAAIvC,KAAK,KAAK;QAAEwC,IAAIxC,KAAK,MAAM;KAAC;AAC1C;AAkEO,eAAe2C,qBACpBC,WAAmB,EACnB5C,IAAU,EACVC,gBAAmC;IAEnC,IACE4C,AAAmC,kBAAnCA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAa5C,qBACbS,AAAAA,IAAAA,oBAAAA,kBAAAA,AAAAA,EAAmBT,sBAAsB6C,oBAAAA,kBAAAA,CAAAA,IAAuB,EAChE;QACA1D,MAAM,uCAAuCY;QAC7C,MAAM+C,gBAAgB/C,KAAK,KAAK,GAAGA,KAAK,MAAM;QAC9C,MAAMgD,YAAY;QAClB,IAAID,gBAAgBC,WAAW;YAC7B,MAAMC,eAAetD,KAAK,IAAI,CAACqD,YAAYD;YAC3C,MAAMG,WAAWvD,KAAK,KAAK,CAACK,KAAK,KAAK,GAAGiD;YACzC,MAAME,YAAYxD,KAAK,KAAK,CAACK,KAAK,MAAM,GAAGiD;YAC3C7D,MACE,2DACA8D,UACAC;YAEF,MAAMC,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,eAAAA,AAAAA,EAAgBT,aAAa;gBACtD,OAAOM;gBACP,QAAQC;YACV;YACA,OAAOC;QACT;IACF;IACA,OAAOR;AACT"}

package/dist/lib/insight/index.js

@@ -196,6 +196,9 @@ class Insight {
         const context = await this.contextRetrieverFn('describe');
         const { screenshotBase64, size } = context;
         (0, utils_namespaceObject.assert)(screenshotBase64, 'screenshot is required for insight.describe');
+        const modelPreferences = {
+            intent: 'grounding'
+        };
         const systemPrompt = (0, describe_js_namespaceObject.elementDescriberInstruction)();
         const defaultRectSize = 30;
         const targetRect = Array.isArray(target) ? {
@@ -215,9 +218,6 @@ class Insight {
             borderThickness: 3
         });
         if (null == opt ? void 0 : opt.deepThink) {
-            const modelPreferences = {
-                intent: 'grounding'
-            };
             const searchArea = (0, common_js_namespaceObject.expandSearchArea)(targetRect, context.size, modelPreferences);
             debug('describe: set searchArea', searchArea);
             imagePayload = await (0, img_namespaceObject.cropByRect)(imagePayload, searchArea, (0, env_namespaceObject.getIsUseQwenVl)(modelPreferences));
@@ -241,7 +241,7 @@ class Insight {
             }
         ];
         const callAIFn = this.aiVendorFn || index_js_namespaceObject.callToGetJSONObject;
-        const res = await callAIFn(msgs, common_js_namespaceObject.AIActionType.DESCRIBE_ELEMENT);
+        const res = await callAIFn(msgs, common_js_namespaceObject.AIActionType.DESCRIBE_ELEMENT, modelPreferences);
         const { content } = res;
         (0, utils_namespaceObject.assert)(!content.error, `describe failed: ${content.error}`);
         (0, utils_namespaceObject.assert)(content.description, 'failed to describe the element');