@rpascene/core 0.30.8

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

@@ -0,0 +1 @@
+{"version":3,"file":"ai-model\\service-caller\\index.js","sources":["webpack://@rpascene/core/webpack/runtime/compat_get_default_export","webpack://@rpascene/core/webpack/runtime/define_property_getters","webpack://@rpascene/core/webpack/runtime/has_own_property","webpack://@rpascene/core/webpack/runtime/make_namespace_object","webpack://@rpascene/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 IModelConfig,\n  RPASCENE_API_TYPE,\n  RPASCENE_LANGSMITH_DEBUG,\n  OPENAI_MAX_TOKENS,\n  type TVlModeTypes,\n  type UITarsModelVersion,\n  globalConfigManager,\n} from '@rpascene/shared/env';\n\nimport { parseBase64 } from '@rpascene/shared/img';\nimport { getDebug } from '@rpascene/shared/logger';\nimport { assert } from '@rpascene/shared/utils';\nimport { ifInBrowser } from '@rpascene/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\n// 用友AI服务配置接口\nexport interface AIServiceConfig {\n  domain: string; // 领域客户端的领域编码\n  yht_access_token: string; // 用友登录获取鉴权\n  chatType: number; // 大模型厂商类别，如14: '火山引擎'\n  model?: string; // 具体模型值，如Doubao-vision-pro\n  modelCategory: number; // 模型类别：1-文本内容、2-图片理解\n  stream?: number; // 是否以流式接口的形式返回数据；默认0，0-关闭 1-开启\n  temperature?: number; // 温度参数，默认0.95，范围 (0, 1.0]\n  top_p?: number; // 核取样参数，默认值为（智普：0.7，百度 0.8）\n  extraParams?: Record<string, any>; // 额外大模型参数\n  topic?: string; // 默认可为空，用于聊天历史记录获取显示\n  baseURL: string; // API基础URL/chat\n}\n\n// 默认用友AI服务配置\nexport const defaultYhtConfig: AIServiceConfig = {\n  domain: 'APILink',\n  yht_access_token: '',\n  // 用友\n  // chatType: 5,\n  // model: 'qwenvl-7b',\n  // 通义\n  // chatType: 12,\n  // model: 'qwen3-vl-plus',\n  // 火山\n  chatType: 14,\n  // model: 'Doubao-vision-pro',\n  model: 'doubao-seed-1-6-vision-250815',\n  modelCategory: 2,\n  stream: 0,\n  temperature: 0.95,\n  top_p: 0.7,\n  baseURL: 'https://c2.yonyoucloud.com/iuap-aip-service/report/rest/api/aiService/chat'\n};\n\nasync function createChatClient({\n  AIActionTypeValue,\n  modelConfig,\n}: {\n  AIActionTypeValue: AIActionType;\n  modelConfig: IModelConfig;\n}): Promise<{\n  completion: OpenAI.Chat.Completions;\n  style: 'openai' | 'anthropic';\n  modelName: string;\n  modelDescription: string;\n  uiTarsVersion?: UITarsModelVersion;\n  vlMode: TVlModeTypes | undefined;\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    uiTarsModelVersion: uiTarsVersion,\n    vlMode,\n  } = modelConfig;\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 Rpascene.',\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        [RPASCENE_API_TYPE]: AIActionTypeValue.toString(),\n      },\n      dangerouslyAllowBrowser: true,\n    });\n  }\n\n  if (\n    openai &&\n    globalConfigManager.getEnvConfigInBoolean(RPASCENE_LANGSMITH_DEBUG)\n  ) {\n    if (ifInBrowser) {\n      throw new Error('langsmith is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langsmith wrapper enabled');\n    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      uiTarsVersion,\n      vlMode,\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      uiTarsVersion,\n      vlMode,\n    };\n  }\n\n  throw new Error('Openai SDK or Anthropic SDK is not initialized');\n}\n\nexport async function callAIOld(\n  messages: ChatCompletionMessageParam[],\n  AIActionTypeValue: AIActionType,\n  modelConfig: IModelConfig,\n  options?: {\n    stream?: boolean;\n    onChunk?: StreamingCallback;\n  },\n): Promise<{ content: string; usage?: AIUsageInfo; isStreamed: boolean }> {\n  const {\n    completion,\n    style,\n    modelName,\n    modelDescription,\n    uiTarsVersion,\n    vlMode,\n  } = await createChatClient({\n    AIActionTypeValue,\n    modelConfig,\n  });\n\n  const responseFormat = getResponseFormat(modelName, AIActionTypeValue);\n\n  const maxTokens = globalConfigManager.getEnvConfigValue(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: vlMode === '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    ...(vlMode === 'qwen-vl' || vlMode === 'qwen3-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: modelConfig.intent,\n              },\n            };\n            options.onChunk!(finalChunk);\n            break;\n          }\n        }\n        content = accumulated;\n        debugProfileStats(\n          `streaming model, ${modelName}, mode, ${vlMode || '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, ${vlMode || 'default'}, ui-tars-version, ${uiTarsVersion}, 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: modelConfig.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: modelConfig.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://rpascenejs.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      case AIActionType.TEXT:\n        // No response format for plain text - return as-is\n        responseFormat = undefined;\n        break;\n    }\n  }\n\n  // gpt-4o-2024-05-13 only supports json_object response format\n  // Skip for plain text to allow string output\n  if (\n    modelName === 'gpt-4o-2024-05-13' &&\n    AIActionTypeValue !== AIActionType.TEXT\n  ) {\n    responseFormat = { type: AIResponseFormat.JSON };\n  }\n\n  return responseFormat;\n};\n\nexport async function callAIWithObjectResponse<T>(\n  messages: ChatCompletionMessageParam[],\n  AIActionTypeValue: AIActionType,\n  modelConfig: IModelConfig,\n): Promise<{ content: T; usage?: AIUsageInfo }> {\n  const response = await callAI(messages, AIActionTypeValue, modelConfig);\n  assert(response, 'empty response');\n  const vlMode = modelConfig.vlMode;\n  const jsonContent = safeParseJson(response.content, vlMode);\n  return { content: jsonContent, usage: response.usage };\n}\n\nexport async function callAIWithStringResponse(\n  msgs: AIArgs,\n  AIActionTypeValue: AIActionType,\n  modelConfig: IModelConfig,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n  const { content, usage } = await callAI(msgs, AIActionTypeValue, modelConfig);\n  return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n  try {\n    // First, try to match a JSON object directly in the response\n    const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\n    if (jsonMatch) {\n      return jsonMatch[1];\n    }\n\n    // If no direct JSON object is found, try to extract JSON from a code block\n    const codeBlockMatch = response.match(\n      /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\n    );\n    if (codeBlockMatch) {\n      return codeBlockMatch[1];\n    }\n\n    // If no code block is found, try to find a JSON-like structure in the text\n    const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\n    if (jsonLikeMatch) {\n      return jsonLikeMatch[0];\n    }\n  } catch { }\n  // If no JSON-like structure is found, return the original response\n  return response;\n}\n\nexport function preprocessDoubaoBboxJson(input: string) {\n  if (input.includes('bbox')) {\n    // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\n    while (/\\d+\\s+\\d+/.test(input)) {\n      input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n    }\n  }\n  return input;\n}\n\nexport function safeParseJson(input: string, vlMode: TVlModeTypes | undefined) {\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 (vlMode === 'doubao-vision' || vlMode === 'vlm-ui-tars') {\n    const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\n    return JSON.parse(jsonrepair(jsonString));\n  }\n  throw Error(`failed to parse json response: ${input}`);\n}\n\n// 用友AI服务调用方法\nexport async function callAI(\n  messages: ChatCompletionMessageParam[],\n  AIActionTypeValue: AIActionType,\n  modelConfig: IModelConfig,\n  options?: {\n    stream?: boolean;\n    onChunk?: StreamingCallback;\n  },\n): Promise<{ content: string; usage?: AIUsageInfo; isStreamed: boolean }> {\n\n  const allEnvConfig = globalConfigManager.getAllEnvConfig();\n  const YhtConfig: any = allEnvConfig || defaultYhtConfig\n  // console.log('YhtConfig:', YhtConfig)\n  // console.log('调用大模型的参数:', messages, AIActionTypeValue, modelConfig, options);\n  const debugCall = getDebug('ai:call:yht');\n  const debugProfileStats = getDebug('ai:profile:stats:yht');\n  const debugProfileDetail = getDebug('ai:profile:detail:yht');\n  const startTime = Date.now();\n\n  // 使用默认配置或自定义配置\n  const config: AIServiceConfig = YhtConfig;\n  const isStreaming = options?.stream && options?.onChunk;\n\n  try {\n    debugCall(`准备调用AI服务，模型: ${config.model}, 动作类型: ${AIActionTypeValue}`);\n\n    // 转换消息格式\n    const yhtMessages = convertToYhtMessageFormat(messages);\n    debugProfileDetail('转换后的消息格式:', JSON.stringify(yhtMessages));\n\n    // 构建请求体\n    const requestBody = {\n      domain: config.domain,\n      messages: yhtMessages,\n      chatType: config.chatType,\n      model: config.model,\n      modelCategory: config.modelCategory,\n      stream: isStreaming ? 1 : config.stream,\n      temperature: config.temperature,\n      top_p: config.top_p,\n      extraParams: config.extraParams,\n      topic: config.topic,\n    };\n\n    // console.log('请求体:', requestBody);\n    debugCall('发送请求到AI服务');\n    const response = await fetch(config.baseURL || defaultYhtConfig.baseURL, {\n      method: 'POST',\n      headers: {\n        'Content-Type': 'application/json',\n        'yht_access_token': config.yht_access_token,\n      },\n      body: JSON.stringify(requestBody),\n    });\n\n    if (!response.ok) {\n      const errorText = await response.text();\n      debugCall(`用友AI服务返回错误: ${response.status} ${errorText}`);\n\n\n      throw new Error(`用友AI服务返回错误: ${response.status} ${errorText}`);\n    }\n\n    const responseText = await response.text();\n    debugProfileDetail('用友AI服务原始响应:', responseText);\n\n    // 处理响应数据\n    let result: any;\n    try {\n      result = JSON.parse(responseText);\n    } catch (parseError) {\n      debugCall('JSON解析失败，尝试检查是否为XML格式');\n      // 特殊处理用友AI服务可能返回的XML格式\n      const processedContent = extractContentFromResponse(responseText);\n      result = {\n        result: {\n          content: processedContent\n        },\n        usage: undefined\n      };\n    }\n    // console.log('调用大模型的结果:', result);\n\n    const timeCost = Date.now() - startTime;\n\n    debugProfileStats(\n      `用友模型调用完成，耗时: ${timeCost}ms, token使用: ${result.usage?.total_tokens || 'unknown'}`\n    );\n\n    debugProfileDetail(\n      `用友模型使用详情: ${JSON.stringify(result.usage)}`\n    );\n\n    // 提取内容并进行后处理\n    let content = result.result?.content || '';\n\n    // 处理可能包含在XML标签中的JSON\n    content = extractContentFromResponse(content);\n\n    debugCall(`用友AI服务响应内容: ${content}`);\n    assert(content, '用友AI服务返回空内容');\n\n    // 构建使用信息\n    const usage: AIUsageInfo = {\n      prompt_tokens: result.usage?.prompt_tokens || 0,\n      completion_tokens: result.usage?.completion_tokens || 0,\n      total_tokens: result.usage?.total_tokens || 0,\n      time_cost: timeCost,\n      model_name: config.model || '',\n      model_description: '用友AI服务',\n      intent: modelConfig.intent,\n    };\n\n    return {\n      content,\n      usage,\n      isStreamed: false, // 目前只实现非流式调用\n    };\n  } catch (e: any) {\n    const errorMsg = e.message || '未知错误';\n    debugCall(`用友AI调用错误: ${errorMsg}`);\n    console.error('用友AI调用错误:', e);\n\n    // 根据错误类型提供更具体的错误信息\n    let enhancedErrorMsg = `用友AI服务调用失败: ${errorMsg}`;\n    if (errorMsg.includes('fetch')) {\n      enhancedErrorMsg += '。请检查网络连接或API地址是否正确。';\n    } else if (errorMsg.includes('token')) {\n      enhancedErrorMsg += '。请检查用友访问令牌是否有效。';\n    }\n\n    const newError = new Error(enhancedErrorMsg, { cause: e });\n    throw newError;\n  }\n}\n\n// 从响应中提取内容，处理XML格式\nfunction extractContentFromResponse(response: string): string {\n  try {\n    // 特殊处理用友AI服务可能返回的XML格式\n    if (response.includes('<json>') && response.includes('</json>')) {\n      const xmlMatch = response.match(/<json>([\\s\\S]*?)<\\/json>/);\n      if (xmlMatch && xmlMatch[1]) {\n        return xmlMatch[1].trim();\n      }\n    }\n\n    // 尝试直接提取JSON格式内容\n    return extractJSONFromCodeBlock(response);\n  } catch (error) {\n    console.warn('从响应中提取内容时出错:', error);\n    return response;\n  }\n}\n\n// 将OpenAI格式的消息转换为用友格式\nfunction convertToYhtMessageFormat(messages: ChatCompletionMessageParam[]): any[] {\n  return messages.map(msg => {\n    const yhtMessage: any = {\n      role: msg.role === 'system' ? 'system' : 'user',\n    };\n\n    // 处理内容\n    if (typeof msg.content === 'string') {\n      yhtMessage.vlContent = [\n        {\n          type: 'text',\n          text: msg.content\n        }\n      ];\n    } else if (Array.isArray(msg.content)) {\n      // 处理多模态内容（图片+文本）\n      const textParts = msg.content.filter(part => part.type === 'text');\n      const imageParts = msg.content.filter(part => part.type === 'image_url');\n\n      if (imageParts.length > 0) {\n        // 处理包含图片的消息\n        yhtMessage.vlContent = [\n          {\n            type: 'text',\n            text: textParts.map(part => part.text).join(' ') || '请分析图片内容'\n          },\n          ...imageParts.map(part => ({\n            type: 'image_url',\n            image_url: {\n              url: part.image_url.url,\n              detail: 'high'\n            }\n          }))\n        ];\n      } else if (textParts.length > 0) {\n        // 处理纯文本消息（数组形式）\n        yhtMessage.vlContent = [\n          {\n            type: 'text',\n            text: textParts.map(part => part.text).join(' ')\n          }\n        ];\n      }\n    }\n\n    return yhtMessage;\n  });\n}\n"],"names":["__webpack_require__","module","getter","definition","key","Object","obj","prop","Symbol","defaultYhtConfig","createChatClient","AIActionTypeValue","modelConfig","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","openaiUseAzureDeprecated","useAzureOpenai","azureOpenaiScope","azureOpenaiKey","azureOpenaiEndpoint","azureOpenaiApiVersion","azureOpenaiDeployment","azureExtraConfig","useAnthropicSdk","anthropicApiKey","modelDescription","uiTarsVersion","vlMode","openai","proxyAgent","debugProxy","getDebug","HttpsProxyAgent","SocksProxyAgent","AzureOpenAI","tokenProvider","assert","ifInBrowser","credential","DefaultAzureCredential","getBearerTokenProvider","OpenAI","RPASCENE_API_TYPE","globalConfigManager","RPASCENE_LANGSMITH_DEBUG","Error","console","wrapOpenAI","Anthropic","callAIOld","messages","options","completion","style","responseFormat","getResponseFormat","maxTokens","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","isStreaming","content","accumulated","usage","timeCost","commonConfig","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","JSON","convertImageContent","imgBase64","mimeType","body","parseBase64","m","Array","_chunk_delta","anthropicUsage","e","newError","AIActionType","assertSchema","locatorSchema","planSchema","AIResponseFormat","callAIWithObjectResponse","response","callAI","jsonContent","safeParseJson","callAIWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","cleanJsonString","_cleanJsonString_match","jsonrepair","jsonString","allEnvConfig","YhtConfig","config","_result_result","_result_usage3","yhtMessages","convertToYhtMessageFormat","requestBody","fetch","errorText","responseText","parseError","processedContent","extractContentFromResponse","errorMsg","enhancedErrorMsg","xmlMatch","error","msg","yhtMessage","textParts","part","imageParts"],"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;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;IC0CO,MAAMI,mBAAoC;QAC/C,QAAQ;QACR,kBAAkB;QAQlB,UAAU;QAEV,OAAO;QACP,eAAe;QACf,QAAQ;QACR,aAAa;QACb,OAAO;QACP,SAAS;IACX;IAEA,eAAeC,iBAAiB,EAC9BC,iBAAiB,EACjBC,WAAW,EAIZ;QAQC,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,EAChB,oBAAoBC,aAAa,EACjCC,MAAM,EACP,GAAGnB;QAEJ,IAAIoB;QAEJ,IAAIC;QACJ,MAAMC,aAAaC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QAC5B,IAAIrB,WAAW;YACboB,WAAW,oBAAoBpB;YAC/BmB,aAAa,IAAIG,2CAAAA,eAAeA,CAACtB;QACnC,OAAO,IAAID,YAAY;YACrBqB,WAAW,qBAAqBrB;YAChCoB,aAAa,IAAII,2CAAAA,eAAeA,CAACxB;QACnC;QAEA,IAAIM,0BAEFa,SAAS,IAAIM,gCAAAA,WAAWA,CAAC;YACvB,SAAStB;YACT,QAAQC;YACR,WAAWgB;YACX,GAAGf,iBAAiB;YACpB,yBAAyB;QAC3B;aACK,IAAIE,gBAAgB;YAGzB,IAAImB;YACJ,IAAIlB,kBAAkB;gBACpBmB,IAAAA,sBAAAA,MAAAA,AAAAA,EACE,CAACC,sBAAAA,WAAWA,EACZ;gBAEF,MAAMC,aAAa,IAAIC,yBAAAA,sBAAsBA;gBAE7CJ,gBAAgBK,AAAAA,IAAAA,yBAAAA,sBAAAA,AAAAA,EAAuBF,YAAYrB;gBAEnDW,SAAS,IAAIM,gCAAAA,WAAWA,CAAC;oBACvB,sBAAsBC;oBACtB,UAAUhB;oBACV,YAAYC;oBACZ,YAAYC;oBACZ,GAAGP,iBAAiB;oBACpB,GAAGQ,gBAAgB;gBACrB;YACF,OAEEM,SAAS,IAAIM,gCAAAA,WAAWA,CAAC;gBACvB,QAAQhB;gBACR,UAAUC;gBACV,YAAYC;gBACZ,YAAYC;gBACZ,yBAAyB;gBACzB,GAAGP,iBAAiB;gBACpB,GAAGQ,gBAAgB;YACrB;QAEJ,OAAO,IAAI,CAACC,iBACVK,SAAS,IAAIa,CAAAA,yBAAAA,EAAO;YAClB,SAAS7B;YACT,QAAQC;YACR,WAAWgB;YACX,GAAGf,iBAAiB;YACpB,gBAAgB;gBACd,GAAIA,AAAAA,CAAAA,QAAAA,oBAAAA,KAAAA,IAAAA,kBAAmB,cAAc,AAAD,KAAK,CAAC,CAAC;gBAC3C,CAAC4B,oBAAAA,iBAAiBA,CAAC,EAAEnC,kBAAkB,QAAQ;YACjD;YACA,yBAAyB;QAC3B;QAGF,IACEqB,UACAe,oBAAAA,mBAAAA,CAAAA,qBAAyC,CAACC,oBAAAA,wBAAwBA,GAClE;YACA,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;YACPjB;YACAc;YACAC;YACAC;QACF;QAIF,IAAIJ,iBACFK,SAAS,IAAIoB,oBAAAA,SAASA,CAAC;YACrB,QAAQxB;YACR,WAAWK;YACX,yBAAyB;QAC3B;QAGF,IAAI,AAAkB,WAAXD,UAA2BA,OAAe,QAAQ,EAC3D,OAAO;YACL,YAAaA,OAAe,QAAQ;YACpC,OAAO;YACPjB;YACAc;YACAC;YACAC;QACF;QAGF,MAAM,IAAIkB,MAAM;IAClB;IAEO,eAAeI,UACpBC,QAAsC,EACtC3C,iBAA+B,EAC/BC,WAAyB,EACzB2C,OAGC;QAED,MAAM,EACJC,UAAU,EACVC,KAAK,EACL1C,SAAS,EACTc,gBAAgB,EAChBC,aAAa,EACbC,MAAM,EACP,GAAG,MAAMrB,iBAAiB;YACzBC;YACAC;QACF;QAEA,MAAM8C,iBAAiBC,kBAAkB5C,WAAWJ;QAEpD,MAAMiD,YAAYb,oBAAAA,mBAAAA,CAAAA,iBAAqC,CAACc,oBAAAA,iBAAiBA;QACzE,MAAMC,YAAY3B,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QAC3B,MAAM4B,oBAAoB5B,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QACnC,MAAM6B,qBAAqB7B,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QAEpC,MAAM8B,YAAYC,KAAK,GAAG;QAE1B,MAAMC,cAAcZ,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM,AAAD,KAAKA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO,AAAD;QACtD,IAAIa;QACJ,IAAIC,cAAc;QAClB,IAAIC;QACJ,IAAIC;QAEJ,MAAMC,eAAe;YACnB,aAAazC,AAAW,kBAAXA,SAA2B,MAAM;YAC9C,QAAQ,CAAC,CAACoC;YACV,YACE,AAAqB,YAArB,OAAOP,YACHA,YACAa,OAAO,QAAQ,CAACb,aAAa,QAAQ;YAC3C,GAAI7B,AAAW,cAAXA,UAAwBA,AAAW,eAAXA,SACxB;gBACA,2BAA2B;YAC7B,IACE,CAAC,CAAC;QACR;QAEA,IAAI;YACF,IAAI0B,AAAU,aAAVA,OAAoB;gBACtBK,UACE,CAAC,QAAQ,EAAEK,cAAc,eAAe,GAAG,WAAW,EAAEpD,WAAW;gBAGrE,IAAIoD,aAAa;oBACf,MAAMO,SAAU,MAAMlB,WAAW,MAAM,CACrC;wBACE,OAAOzC;wBACPuC;wBACA,iBAAiBI;wBACjB,GAAGc,YAAY;oBACjB,GACA;wBACE,QAAQ;oBACV;oBAKF,WAAW,MAAMG,SAASD,OAAQ;4BAChBE,uBAAAA,iBAAAA,gBAEbC,wBAAAA,kBAAAA,iBAoBCC,kBAAAA;wBAtBJ,MAAMV,UAAUQ,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,EACbL,QAAQK,MAAM,KAAK;wBAGrB,IAAIP,WAAWW,mBAAmB;4BAChCV,eAAeD;4BACf,MAAMY,YAAiC;gCACrCZ;gCACAW;gCACAV;gCACA,YAAY;gCACZ,OAAOY;4BACT;4BACA1B,QAAQ,OAAO,CAAEyB;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;4BACrCP,WAAWL,KAAK,GAAG,KAAKD;4BAGxB,IAAI,CAACK,OAAO;gCAEV,MAAMY,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAACd,YAAY,MAAM,GAAG;gCAElCC,QAAQ;oCACN,eAAeY;oCACf,mBAAmBA;oCACnB,cAAcA,AAAkB,IAAlBA;gCAChB;4BACF;4BAGA,MAAME,aAAkC;gCACtC,SAAS;gCACTf;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,YAAY,MAAM;gCAC5B;4BACF;4BACA2C,QAAQ,OAAO,CAAE6B;4BACjB;wBACF;oBACF;oBACAhB,UAAUC;oBACVN,kBACE,CAAC,iBAAiB,EAAEhD,UAAU,QAAQ,EAAEgB,UAAU,UAAU,WAAW,EAAEwC,UAAU;gBAEvF,OAAO;wBAUqGc,eAAyDC,gBAAwDC;oBAT3N,MAAMC,SAAS,MAAMhC,WAAW,MAAM,CAAC;wBACrC,OAAOzC;wBACPuC;wBACA,iBAAiBI;wBACjB,GAAGc,YAAY;oBACjB;oBACAD,WAAWL,KAAK,GAAG,KAAKD;oBAExBF,kBACE,CAAC,OAAO,EAAEhD,UAAU,QAAQ,EAAEgB,UAAU,UAAU,mBAAmB,EAAED,cAAc,iBAAiB,EAAEuD,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,EAAEhB,SAAS,aAAa,EAAEiB,OAAO,WAAW,IAAI,IAAI;oBAG3TxB,mBACE,CAAC,oBAAoB,EAAEyB,KAAK,SAAS,CAACD,OAAO,KAAK,GAAG;oBAGvDhD,IAAAA,sBAAAA,MAAAA,AAAAA,EACEgD,OAAO,OAAO,EACd,CAAC,mCAAmC,EAAEC,KAAK,SAAS,CAACD,SAAS;oBAEhEpB,UAAUoB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;oBAC3ClB,QAAQkB,OAAO,KAAK;gBACtB;gBAEA1B,UAAU,CAAC,UAAU,EAAEM,SAAS;gBAChC5B,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO4B,SAAS;YAClB,OAAO,IAAIX,AAAU,gBAAVA,OAAuB;gBAChC,MAAMiC,sBAAsB,CAACtB;oBAC3B,IAAIA,AAAiB,gBAAjBA,QAAQ,IAAI,EAAkB;wBAChC,MAAMuB,YAAYvB,QAAQ,SAAS,CAAC,GAAG;wBACvC5B,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOmD,WAAW;wBAClB,MAAM,EAAEC,QAAQ,EAAEC,IAAI,EAAE,GAAGC,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAY1B,QAAQ,SAAS,CAAC,GAAG;wBAC5D,OAAO;4BACL,QAAQ;gCACN,MAAM;gCACN,YAAYwB;gCACZ,MAAMC;4BACR;4BACA,MAAM;wBACR;oBACF;oBACA,OAAOzB;gBACT;gBAEA,IAAID,aAAa;oBACf,MAAMO,SAAU,MAAMlB,WAAW,MAAM,CAAC;wBACtC,OAAOzC;wBACP,QAAQ;wBACR,UAAUuC,SAAS,GAAG,CAAC,CAACyC,IAAO;gCAC7B,MAAM;gCACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;4BACf;wBACA,iBAAiBrC;wBACjB,GAAGc,YAAY;oBACjB;oBAEA,WAAW,MAAMG,SAASD,OAAQ;4BAChBuB;wBAAhB,MAAM7B,UAAU6B,AAAAA,SAAAA,CAAAA,eAAAA,MAAM,KAAK,AAAD,IAAVA,KAAAA,IAAAA,aAAa,IAAI,AAAD,KAAK;wBACrC,IAAI7B,SAAS;4BACXC,eAAeD;4BACf,MAAMY,YAAiC;gCACrCZ;gCACAC;gCACA,mBAAmB;gCACnB,YAAY;gCACZ,OAAOY;4BACT;4BACA1B,QAAQ,OAAO,CAAEyB;wBACnB;wBAGA,IAAIL,AAAe,mBAAfA,MAAM,IAAI,EAAqB;4BACjCJ,WAAWL,KAAK,GAAG,KAAKD;4BACxB,MAAMiC,iBAAiBvB,MAAM,KAAK;4BAGlC,MAAMS,aAAkC;gCACtC,SAAS;gCACTf;gCACA,mBAAmB;gCACnB,YAAY;gCACZ,OAAO6B,iBACH;oCACA,eAAeA,eAAe,YAAY,IAAI;oCAC9C,mBAAmBA,eAAe,aAAa,IAAI;oCACnD,cACGA,AAAAA,CAAAA,eAAe,YAAY,IAAI,KAC/BA,CAAAA,eAAe,aAAa,IAAI;oCACnC,WAAW3B,YAAY;oCACvB,YAAYxD;oCACZ,mBAAmBc;oCACnB,QAAQjB,YAAY,MAAM;gCAC5B,IACEqE;4BACN;4BACA1B,QAAQ,OAAO,CAAE6B;4BACjB;wBACF;oBACF;oBACAhB,UAAUC;gBACZ,OAAO;oBACL,MAAMmB,SAAS,MAAMhC,WAAW,MAAM,CAAC;wBACrC,OAAOzC;wBACP,QAAQ;wBACR,UAAUuC,SAAS,GAAG,CAAC,CAACyC,IAAO;gCAC7B,MAAM;gCACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;4BACf;wBACA,iBAAiBrC;wBACjB,GAAGc,YAAY;oBACjB;oBACAD,WAAWL,KAAK,GAAG,KAAKD;oBACxBG,UAAWoB,OAAe,OAAO,CAAC,EAAE,CAAC,IAAI;oBACzClB,QAAQkB,OAAO,KAAK;gBACtB;gBAEAhD,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO4B,SAAS;YAClB;YAEA,IAAID,eAAe,CAACG,OAAO;gBAEzB,MAAMY,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEf,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;gBAEtCE,QAAQ;oBACN,eAAeY;oBACf,mBAAmBA;oBACnB,cAAcA,AAAkB,IAAlBA;gBAChB;YACF;YAEA,OAAO;gBACL,SAASd,WAAW;gBACpB,OAAOE,QACH;oBACA,eAAeA,MAAM,aAAa,IAAI;oBACtC,mBAAmBA,MAAM,iBAAiB,IAAI;oBAC9C,cAAcA,MAAM,YAAY,IAAI;oBACpC,WAAWC,YAAY;oBACvB,YAAYxD;oBACZ,mBAAmBc;oBACnB,QAAQjB,YAAY,MAAM;gBAC5B,IACEqE;gBACJ,YAAY,CAAC,CAACd;YAChB;QACF,EAAE,OAAOgC,GAAQ;YACfjD,QAAQ,KAAK,CAAC,kBAAkBiD;YAChC,MAAMC,WAAW,IAAInD,MACnB,CAAC,eAAe,EAAEkB,cAAc,eAAe,GAAG,kBAAkB,EAAEgC,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC/I;gBACE,OAAOA;YACT;YAEF,MAAMC;QACR;IACF;IAEO,MAAMzC,oBAAoB,CAC/B5C,WACAJ;QAIA,IAAI+C;QAKJ,IAAI3C,UAAU,QAAQ,CAAC,UACrB,OAAQJ;YACN,KAAK0F,mCAAAA,YAAAA,CAAAA,MAAmB;gBACtB3C,iBAAiB4C,6BAAAA,YAAYA;gBAC7B;YACF,KAAKD,mCAAAA,YAAAA,CAAAA,eAA4B;gBAC/B3C,iBAAiB6C,+BAAAA,aAAaA;gBAC9B;YACF,KAAKF,mCAAAA,YAAAA,CAAAA,IAAiB;gBACpB3C,iBAAiB8C,gCAAAA,UAAUA;gBAC3B;YACF,KAAKH,mCAAAA,YAAAA,CAAAA,YAAyB;YAC9B,KAAKA,mCAAAA,YAAAA,CAAAA,gBAA6B;gBAChC3C,iBAAiB;oBAAE,MAAM+C,kCAAAA,gBAAAA,CAAAA,IAAqB;gBAAC;gBAC/C;YACF,KAAKJ,mCAAAA,YAAAA,CAAAA,IAAiB;gBAEpB3C,iBAAiBuB;gBACjB;QACJ;QAKF,IACElE,AAAc,wBAAdA,aACAJ,sBAAsB0F,mCAAAA,YAAAA,CAAAA,IAAiB,EAEvC3C,iBAAiB;YAAE,MAAM+C,kCAAAA,gBAAAA,CAAAA,IAAqB;QAAC;QAGjD,OAAO/C;IACT;IAEO,eAAegD,yBACpBpD,QAAsC,EACtC3C,iBAA+B,EAC/BC,WAAyB;QAEzB,MAAM+F,WAAW,MAAMC,OAAOtD,UAAU3C,mBAAmBC;QAC3D4B,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOmE,UAAU;QACjB,MAAM5E,SAASnB,YAAY,MAAM;QACjC,MAAMiG,cAAcC,cAAcH,SAAS,OAAO,EAAE5E;QACpD,OAAO;YAAE,SAAS8E;YAAa,OAAOF,SAAS,KAAK;QAAC;IACvD;IAEO,eAAeI,yBACpBC,IAAY,EACZrG,iBAA+B,EAC/BC,WAAyB;QAEzB,MAAM,EAAEwD,OAAO,EAAEE,KAAK,EAAE,GAAG,MAAMsC,OAAOI,MAAMrG,mBAAmBC;QACjE,OAAO;YAAEwD;YAASE;QAAM;IAC1B;IAEO,SAAS2C,yBAAyBN,QAAgB;QACvD,IAAI;YAEF,MAAMO,YAAYP,SAAS,KAAK,CAAC;YACjC,IAAIO,WACF,OAAOA,SAAS,CAAC,EAAE;YAIrB,MAAMC,iBAAiBR,SAAS,KAAK,CACnC;YAEF,IAAIQ,gBACF,OAAOA,cAAc,CAAC,EAAE;YAI1B,MAAMC,gBAAgBT,SAAS,KAAK,CAAC;YACrC,IAAIS,eACF,OAAOA,aAAa,CAAC,EAAE;QAE3B,EAAE,OAAM,CAAE;QAEV,OAAOT;IACT;IAEO,SAASU,yBAAyBC,KAAa;QACpD,IAAIA,MAAM,QAAQ,CAAC,SAEjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;QAG5C,OAAOA;IACT;IAEO,SAASR,cAAcQ,KAAa,EAAEvF,MAAgC;QAC3E,MAAMwF,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,OAAOgB,KAAK,KAAK,CAAC8B;QACpB,EAAE,OAAM,CAAE;QACV,IAAI;YACF,OAAO9B,KAAK,KAAK,CAACgC,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAWF;QAC/B,EAAE,OAAOpB,GAAG,CAAE;QAEd,IAAIpE,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,QAA0B;YAC1D,MAAM2F,aAAaL,yBAAyBE;YAC5C,OAAO9B,KAAK,KAAK,CAACgC,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAWC;QAC/B;QACA,MAAMzE,MAAM,CAAC,+BAA+B,EAAEqE,OAAO;IACvD;IAGO,eAAeV,OACpBtD,QAAsC,EACtC3C,iBAA+B,EAC/BC,WAAyB,EACzB2C,OAGC;QAGD,MAAMoE,eAAe5E,oBAAAA,mBAAAA,CAAAA,eAAmC;QACxD,MAAM6E,YAAiBD,gBAAgBlH;QAGvC,MAAMqD,YAAY3B,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QAC3B,MAAM4B,oBAAoB5B,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QACnC,MAAM6B,qBAAqB7B,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QACpC,MAAM8B,YAAYC,KAAK,GAAG;QAG1B,MAAM2D,SAA0BD;QAChC,MAAMzD,cAAcZ,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM,AAAD,KAAKA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO,AAAD;QAEtD,IAAI;gBA+DwC8B,eAQ5ByC,gBAUGxC,gBACIC,gBACLwC;YAlFhBjE,UAAU,CAAC,4EAAa,EAAE+D,OAAO,KAAK,CAAC,oCAAQ,EAAElH,mBAAmB;YAGpE,MAAMqH,cAAcC,0BAA0B3E;YAC9CU,mBAAmB,qDAAayB,KAAK,SAAS,CAACuC;YAG/C,MAAME,cAAc;gBAClB,QAAQL,OAAO,MAAM;gBACrB,UAAUG;gBACV,UAAUH,OAAO,QAAQ;gBACzB,OAAOA,OAAO,KAAK;gBACnB,eAAeA,OAAO,aAAa;gBACnC,QAAQ1D,cAAc,IAAI0D,OAAO,MAAM;gBACvC,aAAaA,OAAO,WAAW;gBAC/B,OAAOA,OAAO,KAAK;gBACnB,aAAaA,OAAO,WAAW;gBAC/B,OAAOA,OAAO,KAAK;YACrB;YAGA/D,UAAU;YACV,MAAM6C,WAAW,MAAMwB,MAAMN,OAAO,OAAO,IAAIpH,iBAAiB,OAAO,EAAE;gBACvE,QAAQ;gBACR,SAAS;oBACP,gBAAgB;oBAChB,kBAAoBoH,OAAO,gBAAgB;gBAC7C;gBACA,MAAMpC,KAAK,SAAS,CAACyC;YACvB;YAEA,IAAI,CAACvB,SAAS,EAAE,EAAE;gBAChB,MAAMyB,YAAY,MAAMzB,SAAS,IAAI;gBACrC7C,UAAU,CAAC,oEAAY,EAAE6C,SAAS,MAAM,CAAC,CAAC,EAAEyB,WAAW;gBAGvD,MAAM,IAAInF,MAAM,CAAC,oEAAY,EAAE0D,SAAS,MAAM,CAAC,CAAC,EAAEyB,WAAW;YAC/D;YAEA,MAAMC,eAAe,MAAM1B,SAAS,IAAI;YACxC3C,mBAAmB,uDAAeqE;YAGlC,IAAI7C;YACJ,IAAI;gBACFA,SAASC,KAAK,KAAK,CAAC4C;YACtB,EAAE,OAAOC,YAAY;gBACnBxE,UAAU;gBAEV,MAAMyE,mBAAmBC,2BAA2BH;gBACpD7C,SAAS;oBACP,QAAQ;wBACN,SAAS+C;oBACX;oBACA,OAAOtD;gBACT;YACF;YAGA,MAAMV,WAAWL,KAAK,GAAG,KAAKD;YAE9BF,kBACE,CAAC,0FAAa,EAAEQ,SAAS,2BAAa,EAAEc,AAAAA,SAAAA,CAAAA,gBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,cAAc,YAAY,AAAD,KAAK,WAAW;YAGnFrB,mBACE,CAAC,kEAAU,EAAEyB,KAAK,SAAS,CAACD,OAAO,KAAK,GAAG;YAI7C,IAAIpB,UAAU0D,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,MAAM,AAAD,IAAZA,KAAAA,IAAAA,eAAe,OAAO,AAAD,KAAK;YAGxC1D,UAAUoE,2BAA2BpE;YAErCN,UAAU,CAAC,oEAAY,EAAEM,SAAS;YAClC5B,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO4B,SAAS;YAGhB,MAAME,QAAqB;gBACzB,eAAegB,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,aAAa,AAAD,KAAK;gBAC9C,mBAAmBC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,iBAAiB,AAAD,KAAK;gBACtD,cAAcwC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,YAAY,AAAD,KAAK;gBAC5C,WAAWxD;gBACX,YAAYsD,OAAO,KAAK,IAAI;gBAC5B,mBAAmB;gBACnB,QAAQjH,YAAY,MAAM;YAC5B;YAEA,OAAO;gBACLwD;gBACAE;gBACA,YAAY;YACd;QACF,EAAE,OAAO6B,GAAQ;YACf,MAAMsC,WAAWtC,EAAE,OAAO,IAAI;YAC9BrC,UAAU,CAAC,oDAAU,EAAE2E,UAAU;YACjCvF,QAAQ,KAAK,CAAC,2CAAaiD;YAG3B,IAAIuC,mBAAmB,CAAC,oEAAY,EAAED,UAAU;YAChD,IAAIA,SAAS,QAAQ,CAAC,UACpBC,oBAAoB;iBACf,IAAID,SAAS,QAAQ,CAAC,UAC3BC,oBAAoB;YAGtB,MAAMtC,WAAW,IAAInD,MAAMyF,kBAAkB;gBAAE,OAAOvC;YAAE;YACxD,MAAMC;QACR;IACF;IAGA,SAASoC,2BAA2B7B,QAAgB;QAClD,IAAI;YAEF,IAAIA,SAAS,QAAQ,CAAC,aAAaA,SAAS,QAAQ,CAAC,YAAY;gBAC/D,MAAMgC,WAAWhC,SAAS,KAAK,CAAC;gBAChC,IAAIgC,YAAYA,QAAQ,CAAC,EAAE,EACzB,OAAOA,QAAQ,CAAC,EAAE,CAAC,IAAI;YAE3B;YAGA,OAAO1B,yBAAyBN;QAClC,EAAE,OAAOiC,OAAO;YACd1F,QAAQ,IAAI,CAAC,uEAAgB0F;YAC7B,OAAOjC;QACT;IACF;IAGA,SAASsB,0BAA0B3E,QAAsC;QACvE,OAAOA,SAAS,GAAG,CAACuF,CAAAA;YAClB,MAAMC,aAAkB;gBACtB,MAAMD,AAAa,aAAbA,IAAI,IAAI,GAAgB,WAAW;YAC3C;YAGA,IAAI,AAAuB,YAAvB,OAAOA,IAAI,OAAO,EACpBC,WAAW,SAAS,GAAG;gBACrB;oBACE,MAAM;oBACN,MAAMD,IAAI,OAAO;gBACnB;aACD;iBACI,IAAI7C,MAAM,OAAO,CAAC6C,IAAI,OAAO,GAAG;gBAErC,MAAME,YAAYF,IAAI,OAAO,CAAC,MAAM,CAACG,CAAAA,OAAQA,AAAc,WAAdA,KAAK,IAAI;gBACtD,MAAMC,aAAaJ,IAAI,OAAO,CAAC,MAAM,CAACG,CAAAA,OAAQA,AAAc,gBAAdA,KAAK,IAAI;gBAEvD,IAAIC,WAAW,MAAM,GAAG,GAEtBH,WAAW,SAAS,GAAG;oBACrB;wBACE,MAAM;wBACN,MAAMC,UAAU,GAAG,CAACC,CAAAA,OAAQA,KAAK,IAAI,EAAE,IAAI,CAAC,QAAQ;oBACtD;uBACGC,WAAW,GAAG,CAACD,CAAAA,OAAS;4BACzB,MAAM;4BACN,WAAW;gCACT,KAAKA,KAAK,SAAS,CAAC,GAAG;gCACvB,QAAQ;4BACV;wBACF;iBACD;qBACI,IAAID,UAAU,MAAM,GAAG,GAE5BD,WAAW,SAAS,GAAG;oBACrB;wBACE,MAAM;wBACN,MAAMC,UAAU,GAAG,CAACC,CAAAA,OAAQA,KAAK,IAAI,EAAE,IAAI,CAAC;oBAC9C;iBACD;YAEL;YAEA,OAAOF;QACT;IACF"}

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

@@ -0,0 +1,238 @@
+"use strict";
+var __webpack_require__ = {};
+(()=>{
+    __webpack_require__.d = (exports1, definition)=>{
+        for(var key in definition)if (__webpack_require__.o(definition, key) && !__webpack_require__.o(exports1, key)) Object.defineProperty(exports1, key, {
+            enumerable: true,
+            get: definition[key]
+        });
+    };
+})();
+(()=>{
+    __webpack_require__.o = (obj, prop)=>Object.prototype.hasOwnProperty.call(obj, prop);
+})();
+(()=>{
+    __webpack_require__.r = (exports1)=>{
+        if ('undefined' != typeof Symbol && Symbol.toStringTag) Object.defineProperty(exports1, Symbol.toStringTag, {
+            value: 'Module'
+        });
+        Object.defineProperty(exports1, '__esModule', {
+            value: true
+        });
+    };
+})();
+var __webpack_exports__ = {};
+__webpack_require__.r(__webpack_exports__);
+__webpack_require__.d(__webpack_exports__, {
+    resizeImageForUiTars: ()=>resizeImageForUiTars,
+    uiTarsPlanning: ()=>uiTarsPlanning
+});
+const env_namespaceObject = require("@rpascene/shared/env");
+const img_namespaceObject = require("@rpascene/shared/img");
+const logger_namespaceObject = require("@rpascene/shared/logger");
+const us_keyboard_layout_namespaceObject = require("@rpascene/shared/us-keyboard-layout");
+const utils_namespaceObject = require("@rpascene/shared/utils");
+const action_parser_namespaceObject = require("@ui-tars/action-parser");
+const external_common_js_namespaceObject = require("./common.js");
+const ui_tars_planning_js_namespaceObject = require("./prompt/ui-tars-planning.js");
+const index_js_namespaceObject = require("./service-caller/index.js");
+const debug = (0, logger_namespaceObject.getDebug)('ui-tars-planning');
+const bboxSize = 10;
+const pointToBbox = (point, width, height)=>[
+        Math.round(Math.max(point.x - bboxSize / 2, 0)),
+        Math.round(Math.max(point.y - bboxSize / 2, 0)),
+        Math.round(Math.min(point.x + bboxSize / 2, width)),
+        Math.round(Math.min(point.y + bboxSize / 2, height))
+    ];
+async function uiTarsPlanning(userInstruction, options) {
+    const { conversationHistory, context, modelConfig } = options;
+    const { uiTarsModelVersion } = modelConfig;
+    const systemPrompt = (0, ui_tars_planning_js_namespaceObject.getUiTarsPlanningPrompt)() + userInstruction;
+    const imagePayload = await resizeImageForUiTars(context.screenshotBase64, context.size, uiTarsModelVersion);
+    conversationHistory.append({
+        role: 'user',
+        content: [
+            {
+                type: 'image_url',
+                image_url: {
+                    url: imagePayload
+                }
+            }
+        ]
+    });
+    const res = await (0, index_js_namespaceObject.callAIWithStringResponse)([
+        {
+            role: 'user',
+            content: systemPrompt
+        },
+        ...conversationHistory.snapshot()
+    ], external_common_js_namespaceObject.AIActionType.INSPECT_ELEMENT, modelConfig);
+    const convertedText = convertBboxToCoordinates(res.content);
+    const { size } = context;
+    const { parsed } = (0, action_parser_namespaceObject.actionParser)({
+        prediction: convertedText,
+        factor: [
+            1000,
+            1000
+        ],
+        screenContext: {
+            width: size.width,
+            height: size.height
+        },
+        modelVer: uiTarsModelVersion
+    });
+    debug('ui-tars modelVer', uiTarsModelVersion, ', parsed', JSON.stringify(parsed));
+    const transformActions = [];
+    let shouldContinue = true;
+    parsed.forEach((action)=>{
+        const actionType = (action.action_type || '').toLowerCase();
+        if ('click' === actionType) {
+            (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: 'Tap',
+                param: {
+                    locate: {
+                        prompt: action.thought || '',
+                        bbox: pointToBbox({
+                            x: point[0],
+                            y: point[1]
+                        }, size.width, size.height)
+                    }
+                }
+            });
+        } else if ('drag' === actionType) {
+            (0, utils_namespaceObject.assert)(action.action_inputs.start_box, 'start_box is required');
+            (0, utils_namespaceObject.assert)(action.action_inputs.end_box, 'end_box is required');
+            const startPoint = getPoint(action.action_inputs.start_box, size);
+            const endPoint = getPoint(action.action_inputs.end_box, size);
+            transformActions.push({
+                type: 'DragAndDrop',
+                param: {
+                    from: {
+                        prompt: action.thought || '',
+                        bbox: pointToBbox({
+                            x: startPoint[0],
+                            y: startPoint[1]
+                        }, size.width, size.height)
+                    },
+                    to: {
+                        prompt: action.thought || '',
+                        bbox: pointToBbox({
+                            x: endPoint[0],
+                            y: endPoint[1]
+                        }, size.width, size.height)
+                    }
+                },
+                thought: action.thought || ''
+            });
+        } else if ('type' === actionType) transformActions.push({
+            type: 'Input',
+            param: {
+                value: action.action_inputs.content
+            },
+            thought: action.thought || ''
+        });
+        else if ('scroll' === actionType) transformActions.push({
+            type: 'Scroll',
+            param: {
+                direction: action.action_inputs.direction
+            },
+            thought: action.thought || ''
+        });
+        else if ('finished' === actionType) {
+            shouldContinue = false;
+            transformActions.push({
+                type: 'Finished',
+                param: {},
+                thought: action.thought || ''
+            });
+        } else if ('hotkey' === actionType) if (action.action_inputs.key) {
+            const keys = (0, us_keyboard_layout_namespaceObject.transformHotkeyInput)(action.action_inputs.key);
+            transformActions.push({
+                type: 'KeyboardPress',
+                param: {
+                    keyName: keys
+                },
+                thought: action.thought || ''
+            });
+        } else console.warn('No key found in action: hotkey. Will not perform action.');
+        else if ('wait' === actionType) transformActions.push({
+            type: 'Sleep',
+            param: {
+                timeMs: 1000
+            },
+            thought: action.thought || ''
+        });
+    });
+    if (0 === transformActions.length) throw new Error(`No actions found, response: ${res.content}`, {
+        cause: {
+            prediction: res.content,
+            parsed
+        }
+    });
+    debug('transformActions', JSON.stringify(transformActions, null, 2));
+    const log = (0, ui_tars_planning_js_namespaceObject.getSummary)(res.content);
+    conversationHistory.append({
+        role: 'assistant',
+        content: log
+    });
+    return {
+        actions: transformActions,
+        log,
+        usage: res.usage,
+        rawResponse: JSON.stringify(res.content, void 0, 2),
+        more_actions_needed_by_instruction: shouldContinue
+    };
+}
+function convertBboxToCoordinates(text) {
+    const pattern = /<bbox>(\d+)\s+(\d+)\s+(\d+)\s+(\d+)<\/bbox>/g;
+    function replaceMatch(match, x1, y1, x2, y2) {
+        const x1Num = Number.parseInt(x1, 10);
+        const y1Num = Number.parseInt(y1, 10);
+        const x2Num = Number.parseInt(x2, 10);
+        const y2Num = Number.parseInt(y2, 10);
+        const x = Math.floor((x1Num + x2Num) / 2);
+        const y = Math.floor((y1Num + y2Num) / 2);
+        return `(${x},${y})`;
+    }
+    const cleanedText = text.replace(/\[EOS\]/g, '');
+    return cleanedText.replace(pattern, replaceMatch).trim();
+}
+function getPoint(startBox, size) {
+    const [x, y] = JSON.parse(startBox);
+    return [
+        x * size.width,
+        y * size.height
+    ];
+}
+async function resizeImageForUiTars(imageBase64, size, uiTarsVersion) {
+    if (uiTarsVersion === env_namespaceObject.UITarsModelVersion.V1_5) {
+        debug('ui-tars-v1.5, will check image size', size);
+        const currentPixels = size.width * size.height;
+        const maxPixels = 12845056;
+        if (currentPixels > maxPixels) {
+            const resizeFactor = Math.sqrt(maxPixels / currentPixels);
+            const newWidth = Math.floor(size.width * resizeFactor);
+            const newHeight = Math.floor(size.height * resizeFactor);
+            debug('resize image for ui-tars, new width: %s, new height: %s', newWidth, newHeight);
+            const resizedImage = await (0, img_namespaceObject.resizeImgBase64)(imageBase64, {
+                width: newWidth,
+                height: newHeight
+            });
+            return resizedImage;
+        }
+    }
+    return imageBase64;
+}
+exports.resizeImageForUiTars = __webpack_exports__.resizeImageForUiTars;
+exports.uiTarsPlanning = __webpack_exports__.uiTarsPlanning;
+for(var __webpack_i__ in __webpack_exports__)if (-1 === [
+    "resizeImageForUiTars",
+    "uiTarsPlanning"
+].indexOf(__webpack_i__)) exports[__webpack_i__] = __webpack_exports__[__webpack_i__];
+Object.defineProperty(exports, '__esModule', {
+    value: true
+});
+
+//# sourceMappingURL=ui-tars-planning.js.map

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

@@ -0,0 +1 @@
+{"version":3,"file":"ai-model\\ui-tars-planning.js","sources":["webpack://@rpascene/core/webpack/runtime/define_property_getters","webpack://@rpascene/core/webpack/runtime/has_own_property","webpack://@rpascene/core/webpack/runtime/make_namespace_object","webpack://@rpascene/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  PlanningAIResponse,\n  PlanningAction,\n  Size,\n  UIContext,\n} from '@/types';\nimport { type IModelConfig, UITarsModelVersion } from '@rpascene/shared/env';\nimport { resizeImgBase64 } from '@rpascene/shared/img';\nimport { getDebug } from '@rpascene/shared/logger';\nimport { transformHotkeyInput } from '@rpascene/shared/us-keyboard-layout';\nimport { assert } from '@rpascene/shared/utils';\nimport { actionParser } from '@ui-tars/action-parser';\nimport { AIActionType } from './common';\nimport type { ConversationHistory } from './conversation-history';\nimport { getSummary, getUiTarsPlanningPrompt } from './prompt/ui-tars-planning';\nimport { callAIWithStringResponse } 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 uiTarsPlanning(\n  userInstruction: string,\n  options: {\n    conversationHistory: ConversationHistory;\n    context: UIContext;\n    modelConfig: IModelConfig;\n  },\n): Promise<PlanningAIResponse> {\n  const { conversationHistory, context, modelConfig } = options;\n  const { uiTarsModelVersion } = modelConfig;\n  const systemPrompt = getUiTarsPlanningPrompt() + userInstruction;\n\n  const imagePayload = await resizeImageForUiTars(\n    context.screenshotBase64,\n    context.size,\n    uiTarsModelVersion,\n  );\n\n  conversationHistory.append({\n    role: 'user',\n    content: [\n      {\n        type: 'image_url',\n        image_url: {\n          url: imagePayload,\n        },\n      },\n    ],\n  });\n\n  const res = await callAIWithStringResponse(\n    [\n      {\n        role: 'user',\n        content: systemPrompt,\n      },\n      ...conversationHistory.snapshot(),\n    ],\n    AIActionType.INSPECT_ELEMENT,\n    modelConfig,\n  );\n  const convertedText = convertBboxToCoordinates(res.content);\n\n  const { size } = context;\n  const { parsed } = actionParser({\n    prediction: convertedText,\n    factor: [1000, 1000],\n    screenContext: {\n      width: size.width,\n      height: size.height,\n    },\n    modelVer: uiTarsModelVersion,\n  });\n\n  debug(\n    'ui-tars modelVer',\n    uiTarsModelVersion,\n    ', parsed',\n    JSON.stringify(parsed),\n  );\n\n  const transformActions: PlanningAction[] = [];\n  let shouldContinue = true;\n  parsed.forEach((action) => {\n    const actionType = (action.action_type || '').toLowerCase();\n    if (actionType === 'click') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, 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 (actionType === 'drag') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      assert(action.action_inputs.end_box, 'end_box is required');\n      const startPoint = getPoint(action.action_inputs.start_box, 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 (actionType === 'type') {\n      transformActions.push({\n        type: 'Input',\n        param: {\n          value: action.action_inputs.content,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'scroll') {\n      transformActions.push({\n        type: 'Scroll',\n        param: {\n          direction: action.action_inputs.direction,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'finished') {\n      shouldContinue = false;\n      transformActions.push({\n        type: 'Finished',\n        param: {},\n        thought: action.thought || '',\n      });\n    } else if (actionType === '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 (actionType === '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: ${res.content}`, {\n      cause: {\n        prediction: res.content,\n        parsed,\n      },\n    });\n  }\n\n  debug('transformActions', JSON.stringify(transformActions, null, 2));\n  const log = getSummary(res.content);\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: log,\n  });\n\n  return {\n    actions: transformActions,\n    log,\n    usage: res.usage,\n    rawResponse: JSON.stringify(res.content, undefined, 2),\n    more_actions_needed_by_instruction: shouldContinue,\n  };\n}\n\n/**\n * Converts bounding box notation to coordinate points\n * @param text - The text containing bbox tags to be converted\n * @returns The text with bbox tags replaced by coordinate points\n */\nfunction convertBboxToCoordinates(text: string): string {\n  // Match the four numbers after <bbox>\n  const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n  function replaceMatch(\n    match: string,\n    x1: string,\n    y1: string,\n    x2: string,\n    y2: string,\n  ): string {\n    // Convert strings to numbers and calculate center point\n    const x1Num = Number.parseInt(x1, 10);\n    const y1Num = Number.parseInt(y1, 10);\n    const x2Num = Number.parseInt(x2, 10);\n    const y2Num = Number.parseInt(y2, 10);\n\n    // Use Math.floor to truncate and calculate center point\n    const x = Math.floor((x1Num + x2Num) / 2);\n    const y = Math.floor((y1Num + y2Num) / 2);\n\n    // Return formatted coordinate string\n    return `(${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  uiTarsVersion: UITarsModelVersion | undefined,\n) {\n  if (uiTarsVersion === UITarsModelVersion.V1_5) {\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","uiTarsPlanning","userInstruction","options","conversationHistory","context","modelConfig","uiTarsModelVersion","systemPrompt","getUiTarsPlanningPrompt","imagePayload","resizeImageForUiTars","res","callAIWithStringResponse","AIActionType","convertedText","convertBboxToCoordinates","size","parsed","actionParser","JSON","transformActions","shouldContinue","action","actionType","assert","getPoint","startPoint","endPoint","keys","transformHotkeyInput","console","Error","log","getSummary","undefined","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","y","cleanedText","startBox","imageBase64","uiTarsVersion","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;;;;;;;;;;;;;;;;;ACmBA,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,eACpBC,eAAuB,EACvBC,OAIC;IAED,MAAM,EAAEC,mBAAmB,EAAEC,OAAO,EAAEC,WAAW,EAAE,GAAGH;IACtD,MAAM,EAAEI,kBAAkB,EAAE,GAAGD;IAC/B,MAAME,eAAeC,AAAAA,IAAAA,oCAAAA,uBAAAA,AAAAA,MAA4BP;IAEjD,MAAMQ,eAAe,MAAMC,qBACzBN,QAAQ,gBAAgB,EACxBA,QAAQ,IAAI,EACZE;IAGFH,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKM;gBACP;YACF;SACD;IACH;IAEA,MAAME,MAAM,MAAMC,AAAAA,IAAAA,yBAAAA,wBAAAA,AAAAA,EAChB;QACE;YACE,MAAM;YACN,SAASL;QACX;WACGJ,oBAAoB,QAAQ;KAChC,EACDU,mCAAAA,YAAAA,CAAAA,eAA4B,EAC5BR;IAEF,MAAMS,gBAAgBC,yBAAyBJ,IAAI,OAAO;IAE1D,MAAM,EAAEK,IAAI,EAAE,GAAGZ;IACjB,MAAM,EAAEa,MAAM,EAAE,GAAGC,AAAAA,IAAAA,8BAAAA,YAAAA,AAAAA,EAAa;QAC9B,YAAYJ;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOE,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUV;IACZ;IAEAd,MACE,oBACAc,oBACA,YACAa,KAAK,SAAS,CAACF;IAGjB,MAAMG,mBAAqC,EAAE;IAC7C,IAAIC,iBAAiB;IACrBJ,OAAO,OAAO,CAAC,CAACK;QACd,MAAMC,aAAcD,AAAAA,CAAAA,OAAO,WAAW,IAAI,EAAC,EAAG,WAAW;QACzD,IAAIC,AAAe,YAAfA,YAAwB;YAC1BC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM1B,QAAQ6B,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEN;YACvDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQE,OAAO,OAAO,IAAI;wBAC1B,MAAM3B,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BoB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;YACF;QACF,OAAO,IAAIO,AAAe,WAAfA,YAAuB;YAChCC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCE,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOF,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMI,aAAaD,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEN;YAC5D,MAAMW,WAAWF,SAASH,OAAO,aAAa,CAAC,OAAO,EAAEN;YACxDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,MAAM;wBACJ,QAAQE,OAAO,OAAO,IAAI;wBAC1B,MAAM3B,YACJ;4BAAE,GAAG+B,UAAU,CAAC,EAAE;4BAAE,GAAGA,UAAU,CAAC,EAAE;wBAAC,GACrCV,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;oBACA,IAAI;wBACF,QAAQM,OAAO,OAAO,IAAI;wBAC1B,MAAM3B,YACJ;4BAAE,GAAGgC,QAAQ,CAAC,EAAE;4BAAE,GAAGA,QAAQ,CAAC,EAAE;wBAAC,GACjCX,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASM,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YACTH,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,OAAOE,OAAO,aAAa,CAAC,OAAO;YACrC;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,aAAfA,YACTH,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWE,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,eAAfA,YAA2B;YACpCF,iBAAiB;YACjBD,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO,CAAC;gBACR,SAASE,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,aAAfA,YACT,IAAKD,OAAO,aAAa,CAAC,GAAG,EAItB;YACL,MAAMM,OAAOC,AAAAA,IAAAA,mCAAAA,oBAAAA,AAAAA,EAAqBP,OAAO,aAAa,CAAC,GAAG;YAE1DF,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,SAASQ;gBACX;gBACA,SAASN,OAAO,OAAO,IAAI;YAC7B;QACF,OAbEQ,QAAQ,IAAI,CACV;aAaC,IAAIP,AAAe,WAAfA,YACTH,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,SAASE,OAAO,OAAO,IAAI;QAC7B;IAEJ;IAEA,IAAIF,AAA4B,MAA5BA,iBAAiB,MAAM,EACzB,MAAM,IAAIW,MAAM,CAAC,4BAA4B,EAAEpB,IAAI,OAAO,EAAE,EAAE;QAC5D,OAAO;YACL,YAAYA,IAAI,OAAO;YACvBM;QACF;IACF;IAGFzB,MAAM,oBAAoB2B,KAAK,SAAS,CAACC,kBAAkB,MAAM;IACjE,MAAMY,MAAMC,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAWtB,IAAI,OAAO;IAElCR,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS6B;IACX;IAEA,OAAO;QACL,SAASZ;QACTY;QACA,OAAOrB,IAAI,KAAK;QAChB,aAAaQ,KAAK,SAAS,CAACR,IAAI,OAAO,EAAEuB,QAAW;QACpD,oCAAoCb;IACtC;AACF;AAOA,SAASN,yBAAyBoB,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,IAAIjD,KAAK,KAAK,CAAE4C,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMG,IAAIlD,KAAK,KAAK,CAAE8C,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,SAASZ,SAAS0B,QAAgB,EAAEnC,IAAuC;IACzE,MAAM,CAACgC,GAAGC,EAAE,GAAG9B,KAAK,KAAK,CAACgC;IAC1B,OAAO;QAACH,IAAIhC,KAAK,KAAK;QAAEiC,IAAIjC,KAAK,MAAM;KAAC;AAC1C;AAkEO,eAAeN,qBACpB0C,WAAmB,EACnBpC,IAAU,EACVqC,aAA6C;IAE7C,IAAIA,kBAAkBC,oBAAAA,kBAAAA,CAAAA,IAAuB,EAAE;QAC7C9D,MAAM,uCAAuCwB;QAC7C,MAAMuC,gBAAgBvC,KAAK,KAAK,GAAGA,KAAK,MAAM;QAC9C,MAAMwC,YAAY;QAClB,IAAID,gBAAgBC,WAAW;YAC7B,MAAMC,eAAe1D,KAAK,IAAI,CAACyD,YAAYD;YAC3C,MAAMG,WAAW3D,KAAK,KAAK,CAACiB,KAAK,KAAK,GAAGyC;YACzC,MAAME,YAAY5D,KAAK,KAAK,CAACiB,KAAK,MAAM,GAAGyC;YAC3CjE,MACE,2DACAkE,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/device/index.js

@@ -0,0 +1,255 @@
+"use strict";
+var __webpack_require__ = {};
+(()=>{
+    __webpack_require__.d = (exports1, definition)=>{
+        for(var key in definition)if (__webpack_require__.o(definition, key) && !__webpack_require__.o(exports1, key)) Object.defineProperty(exports1, key, {
+            enumerable: true,
+            get: definition[key]
+        });
+    };
+})();
+(()=>{
+    __webpack_require__.o = (obj, prop)=>Object.prototype.hasOwnProperty.call(obj, prop);
+})();
+(()=>{
+    __webpack_require__.r = (exports1)=>{
+        if ('undefined' != typeof Symbol && Symbol.toStringTag) Object.defineProperty(exports1, Symbol.toStringTag, {
+            value: 'Module'
+        });
+        Object.defineProperty(exports1, '__esModule', {
+            value: true
+        });
+    };
+})();
+var __webpack_exports__ = {};
+__webpack_require__.r(__webpack_exports__);
+__webpack_require__.d(__webpack_exports__, {
+    defineActionSwipe: ()=>defineActionSwipe,
+    actionRightClickParamSchema: ()=>actionRightClickParamSchema,
+    ActionSwipeParamSchema: ()=>ActionSwipeParamSchema,
+    actionDoubleClickParamSchema: ()=>actionDoubleClickParamSchema,
+    defineAction: ()=>defineAction,
+    AbstractInterface: ()=>AbstractInterface,
+    defineActionLongPress: ()=>defineActionLongPress,
+    defineActionTap: ()=>defineActionTap,
+    defineActionDragAndDrop: ()=>defineActionDragAndDrop,
+    actionTapParamSchema: ()=>actionTapParamSchema,
+    actionKeyboardPressParamSchema: ()=>actionKeyboardPressParamSchema,
+    defineActionHover: ()=>defineActionHover,
+    actionScrollParamSchema: ()=>actionScrollParamSchema,
+    actionDragAndDropParamSchema: ()=>actionDragAndDropParamSchema,
+    defineActionClearInput: ()=>defineActionClearInput,
+    defineActionRightClick: ()=>defineActionRightClick,
+    ActionLongPressParamSchema: ()=>ActionLongPressParamSchema,
+    actionClearInputParamSchema: ()=>actionClearInputParamSchema,
+    actionInputParamSchema: ()=>actionInputParamSchema,
+    defineActionDoubleClick: ()=>defineActionDoubleClick,
+    defineActionKeyboardPress: ()=>defineActionKeyboardPress,
+    defineActionScroll: ()=>defineActionScroll,
+    actionHoverParamSchema: ()=>actionHoverParamSchema,
+    defineActionInput: ()=>defineActionInput
+});
+const index_js_namespaceObject = require("../ai-model/index.js");
+const external_zod_namespaceObject = require("zod");
+class AbstractInterface {
+}
+const defineAction = (config)=>config;
+const actionTapParamSchema = external_zod_namespaceObject.z.object({
+    locate: (0, index_js_namespaceObject.getRpasceneLocationSchema)().describe('The element to be tapped')
+});
+const defineActionTap = (call)=>defineAction({
+        name: 'Tap',
+        description: 'Tap the element',
+        interfaceAlias: 'aiTap',
+        paramSchema: actionTapParamSchema,
+        call
+    });
+const actionRightClickParamSchema = external_zod_namespaceObject.z.object({
+    locate: (0, index_js_namespaceObject.getRpasceneLocationSchema)().describe('The element to be right clicked')
+});
+const defineActionRightClick = (call)=>defineAction({
+        name: 'RightClick',
+        description: 'Right click the element',
+        interfaceAlias: 'aiRightClick',
+        paramSchema: actionRightClickParamSchema,
+        call
+    });
+const actionDoubleClickParamSchema = external_zod_namespaceObject.z.object({
+    locate: (0, index_js_namespaceObject.getRpasceneLocationSchema)().describe('The element to be double clicked')
+});
+const defineActionDoubleClick = (call)=>defineAction({
+        name: 'DoubleClick',
+        description: 'Double click the element',
+        interfaceAlias: 'aiDoubleClick',
+        paramSchema: actionDoubleClickParamSchema,
+        call
+    });
+const actionHoverParamSchema = external_zod_namespaceObject.z.object({
+    locate: (0, index_js_namespaceObject.getRpasceneLocationSchema)().describe('The element to be hovered')
+});
+const defineActionHover = (call)=>defineAction({
+        name: 'Hover',
+        description: 'Move the mouse to the element',
+        interfaceAlias: 'aiHover',
+        paramSchema: actionHoverParamSchema,
+        call
+    });
+const actionInputParamSchema = external_zod_namespaceObject.z.object({
+    value: external_zod_namespaceObject.z.union([
+        external_zod_namespaceObject.z.string(),
+        external_zod_namespaceObject.z.number()
+    ]).transform((val)=>String(val)).describe('The text to input. Provide the final content for replace/append modes, or an empty string when using clear mode to remove existing text.'),
+    locate: (0, index_js_namespaceObject.getRpasceneLocationSchema)().describe('The element to be input').optional(),
+    mode: external_zod_namespaceObject.z["enum"]([
+        'replace',
+        'clear',
+        'append'
+    ]).default('replace').optional().describe('Input mode: "replace" (default) - clear the field and input the value; "append" - append the value to existing content; "clear" - clear the field without inputting new text.')
+});
+const defineActionInput = (call)=>defineAction({
+        name: 'Input',
+        description: 'Input the value into the element',
+        interfaceAlias: 'aiInput',
+        paramSchema: actionInputParamSchema,
+        call
+    });
+const actionKeyboardPressParamSchema = external_zod_namespaceObject.z.object({
+    locate: (0, index_js_namespaceObject.getRpasceneLocationSchema)().describe('The element to be clicked before pressing the key').optional(),
+    keyName: external_zod_namespaceObject.z.string().describe("The key to be pressed. Use '+' for key combinations, e.g., 'Control+A', 'Shift+Enter'")
+});
+const defineActionKeyboardPress = (call)=>defineAction({
+        name: 'KeyboardPress',
+        description: 'Press a key or key combination, like "Enter", "Tab", "Escape", or "Control+A", "Shift+Enter". Do not use this to type text.',
+        interfaceAlias: 'aiKeyboardPress',
+        paramSchema: actionKeyboardPressParamSchema,
+        call
+    });
+const actionScrollParamSchema = external_zod_namespaceObject.z.object({
+    direction: external_zod_namespaceObject.z["enum"]([
+        'down',
+        'up',
+        'right',
+        'left'
+    ]).default('down').describe('The direction to scroll'),
+    scrollType: external_zod_namespaceObject.z["enum"]([
+        'once',
+        'untilBottom',
+        'untilTop',
+        'untilRight',
+        'untilLeft'
+    ]).default('once').describe('The scroll type'),
+    distance: external_zod_namespaceObject.z.number().nullable().optional().describe('The distance in pixels to scroll'),
+    locate: (0, index_js_namespaceObject.getRpasceneLocationSchema)().optional().describe('The element to be scrolled')
+});
+const defineActionScroll = (call)=>defineAction({
+        name: 'Scroll',
+        description: 'Scroll the page or an element. The direction to scroll, the scroll type, and the distance to scroll. The distance is the number of pixels to scroll. If not specified, use `down` direction, `once` scroll type, and `null` distance.',
+        interfaceAlias: 'aiScroll',
+        paramSchema: actionScrollParamSchema,
+        call
+    });
+const actionDragAndDropParamSchema = external_zod_namespaceObject.z.object({
+    from: (0, index_js_namespaceObject.getRpasceneLocationSchema)().describe('The position to be dragged'),
+    to: (0, index_js_namespaceObject.getRpasceneLocationSchema)().describe('The position to be dropped')
+});
+const defineActionDragAndDrop = (call)=>defineAction({
+        name: 'DragAndDrop',
+        description: 'Drag and drop the element',
+        interfaceAlias: 'aiDragAndDrop',
+        paramSchema: actionDragAndDropParamSchema,
+        call
+    });
+const ActionLongPressParamSchema = external_zod_namespaceObject.z.object({
+    locate: (0, index_js_namespaceObject.getRpasceneLocationSchema)().describe('The element to be long pressed'),
+    duration: external_zod_namespaceObject.z.number().default(500).optional().describe('Long press duration in milliseconds')
+});
+const defineActionLongPress = (call)=>defineAction({
+        name: 'LongPress',
+        description: 'Long press the element',
+        paramSchema: ActionLongPressParamSchema,
+        call
+    });
+const ActionSwipeParamSchema = external_zod_namespaceObject.z.object({
+    start: (0, index_js_namespaceObject.getRpasceneLocationSchema)().optional().describe('Starting point of the swipe gesture, if not specified, the center of the page will be used'),
+    direction: external_zod_namespaceObject.z["enum"]([
+        'up',
+        'down',
+        'left',
+        'right'
+    ]).optional().describe('The direction to swipe (required when using distance). The direction means the direction of the finger swipe.'),
+    distance: external_zod_namespaceObject.z.number().optional().describe('The distance in pixels to swipe (mutually exclusive with end)'),
+    end: (0, index_js_namespaceObject.getRpasceneLocationSchema)().optional().describe('Ending point of the swipe gesture (mutually exclusive with distance)'),
+    duration: external_zod_namespaceObject.z.number().default(300).describe('Duration of the swipe gesture in milliseconds'),
+    repeat: external_zod_namespaceObject.z.number().optional().describe('The number of times to repeat the swipe gesture. 1 for default, 0 for infinite (e.g. endless swipe until the end of the page)')
+});
+const defineActionSwipe = (call)=>defineAction({
+        name: 'Swipe',
+        description: 'Perform a swipe gesture. You must specify either "end" (target location) or "distance" + "direction" - they are mutually exclusive. Use "end" for precise location-based swipes, or "distance" + "direction" for relative movement.',
+        paramSchema: ActionSwipeParamSchema,
+        call
+    });
+const actionClearInputParamSchema = external_zod_namespaceObject.z.object({
+    locate: (0, index_js_namespaceObject.getRpasceneLocationSchema)().describe('The input field to be cleared')
+});
+const defineActionClearInput = (call)=>defineAction({
+        name: 'ClearInput',
+        description: 'Clear the text content of an input field',
+        interfaceAlias: 'aiClearInput',
+        paramSchema: actionClearInputParamSchema,
+        call
+    });
+exports.AbstractInterface = __webpack_exports__.AbstractInterface;
+exports.ActionLongPressParamSchema = __webpack_exports__.ActionLongPressParamSchema;
+exports.ActionSwipeParamSchema = __webpack_exports__.ActionSwipeParamSchema;
+exports.actionClearInputParamSchema = __webpack_exports__.actionClearInputParamSchema;
+exports.actionDoubleClickParamSchema = __webpack_exports__.actionDoubleClickParamSchema;
+exports.actionDragAndDropParamSchema = __webpack_exports__.actionDragAndDropParamSchema;
+exports.actionHoverParamSchema = __webpack_exports__.actionHoverParamSchema;
+exports.actionInputParamSchema = __webpack_exports__.actionInputParamSchema;
+exports.actionKeyboardPressParamSchema = __webpack_exports__.actionKeyboardPressParamSchema;
+exports.actionRightClickParamSchema = __webpack_exports__.actionRightClickParamSchema;
+exports.actionScrollParamSchema = __webpack_exports__.actionScrollParamSchema;
+exports.actionTapParamSchema = __webpack_exports__.actionTapParamSchema;
+exports.defineAction = __webpack_exports__.defineAction;
+exports.defineActionClearInput = __webpack_exports__.defineActionClearInput;
+exports.defineActionDoubleClick = __webpack_exports__.defineActionDoubleClick;
+exports.defineActionDragAndDrop = __webpack_exports__.defineActionDragAndDrop;
+exports.defineActionHover = __webpack_exports__.defineActionHover;
+exports.defineActionInput = __webpack_exports__.defineActionInput;
+exports.defineActionKeyboardPress = __webpack_exports__.defineActionKeyboardPress;
+exports.defineActionLongPress = __webpack_exports__.defineActionLongPress;
+exports.defineActionRightClick = __webpack_exports__.defineActionRightClick;
+exports.defineActionScroll = __webpack_exports__.defineActionScroll;
+exports.defineActionSwipe = __webpack_exports__.defineActionSwipe;
+exports.defineActionTap = __webpack_exports__.defineActionTap;
+for(var __webpack_i__ in __webpack_exports__)if (-1 === [
+    "AbstractInterface",
+    "ActionLongPressParamSchema",
+    "ActionSwipeParamSchema",
+    "actionClearInputParamSchema",
+    "actionDoubleClickParamSchema",
+    "actionDragAndDropParamSchema",
+    "actionHoverParamSchema",
+    "actionInputParamSchema",
+    "actionKeyboardPressParamSchema",
+    "actionRightClickParamSchema",
+    "actionScrollParamSchema",
+    "actionTapParamSchema",
+    "defineAction",
+    "defineActionClearInput",
+    "defineActionDoubleClick",
+    "defineActionDragAndDrop",
+    "defineActionHover",
+    "defineActionInput",
+    "defineActionKeyboardPress",
+    "defineActionLongPress",
+    "defineActionRightClick",
+    "defineActionScroll",
+    "defineActionSwipe",
+    "defineActionTap"
+].indexOf(__webpack_i__)) exports[__webpack_i__] = __webpack_exports__[__webpack_i__];
+Object.defineProperty(exports, '__esModule', {
+    value: true
+});
+
+//# sourceMappingURL=index.js.map