@midscene/core 0.26.7-beta-20250818081955.0 → 0.26.7-beta-20250820105545.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/es/ai-model/common.mjs +58 -16
- package/dist/es/ai-model/common.mjs.map +1 -1
- package/dist/es/ai-model/index.mjs +3 -3
- package/dist/es/ai-model/inspect.mjs +28 -16
- package/dist/es/ai-model/inspect.mjs.map +1 -1
- package/dist/es/ai-model/llm-planning.mjs +26 -23
- package/dist/es/ai-model/llm-planning.mjs.map +1 -1
- package/dist/es/ai-model/prompt/llm-planning.mjs +50 -23
- package/dist/es/ai-model/prompt/llm-planning.mjs.map +1 -1
- package/dist/es/ai-model/prompt/playwright-generator.mjs +9 -3
- package/dist/es/ai-model/prompt/playwright-generator.mjs.map +1 -1
- package/dist/es/ai-model/prompt/util.mjs +2 -2
- package/dist/es/ai-model/prompt/util.mjs.map +1 -1
- package/dist/es/ai-model/prompt/yaml-generator.mjs +9 -3
- package/dist/es/ai-model/prompt/yaml-generator.mjs.map +1 -1
- package/dist/es/ai-model/service-caller/index.mjs +72 -118
- package/dist/es/ai-model/service-caller/index.mjs.map +1 -1
- package/dist/es/ai-model/ui-tars-planning.mjs +5 -5
- package/dist/es/ai-model/ui-tars-planning.mjs.map +1 -1
- package/dist/es/index.mjs +3 -2
- package/dist/es/index.mjs.map +1 -1
- package/dist/es/insight/index.mjs +13 -61
- package/dist/es/insight/index.mjs.map +1 -1
- package/dist/es/types.mjs.map +1 -1
- package/dist/es/utils.mjs +5 -6
- package/dist/es/utils.mjs.map +1 -1
- package/dist/lib/ai-model/common.js +80 -20
- package/dist/lib/ai-model/common.js.map +1 -1
- package/dist/lib/ai-model/index.js +14 -5
- package/dist/lib/ai-model/inspect.js +27 -15
- package/dist/lib/ai-model/inspect.js.map +1 -1
- package/dist/lib/ai-model/llm-planning.js +25 -22
- package/dist/lib/ai-model/llm-planning.js.map +1 -1
- package/dist/lib/ai-model/prompt/llm-planning.js +52 -25
- package/dist/lib/ai-model/prompt/llm-planning.js.map +1 -1
- package/dist/lib/ai-model/prompt/playwright-generator.js +9 -3
- package/dist/lib/ai-model/prompt/playwright-generator.js.map +1 -1
- package/dist/lib/ai-model/prompt/util.js +2 -2
- package/dist/lib/ai-model/prompt/util.js.map +1 -1
- package/dist/lib/ai-model/prompt/yaml-generator.js +9 -3
- package/dist/lib/ai-model/prompt/yaml-generator.js.map +1 -1
- package/dist/lib/ai-model/service-caller/index.js +75 -124
- package/dist/lib/ai-model/service-caller/index.js.map +1 -1
- package/dist/lib/ai-model/ui-tars-planning.js +5 -5
- package/dist/lib/ai-model/ui-tars-planning.js.map +1 -1
- package/dist/lib/index.js +20 -4
- package/dist/lib/index.js.map +1 -1
- package/dist/lib/insight/index.js +10 -58
- package/dist/lib/insight/index.js.map +1 -1
- package/dist/lib/types.js.map +1 -1
- package/dist/lib/utils.js +4 -5
- package/dist/lib/utils.js.map +1 -1
- package/dist/types/ai-model/common.d.ts +160 -7
- package/dist/types/ai-model/index.d.ts +2 -2
- package/dist/types/ai-model/inspect.d.ts +2 -0
- package/dist/types/ai-model/llm-planning.d.ts +1 -1
- package/dist/types/ai-model/prompt/llm-planning.d.ts +2 -2
- package/dist/types/ai-model/prompt/util.d.ts +2 -1
- package/dist/types/ai-model/service-caller/index.d.ts +6 -6
- package/dist/types/ai-model/ui-tars-planning.d.ts +3 -1
- package/dist/types/index.d.ts +3 -1
- package/dist/types/insight/index.d.ts +1 -4
- package/dist/types/types.d.ts +8 -11
- package/dist/types/yaml.d.ts +1 -0
- package/package.json +4 -3
|
@@ -1 +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 ANTHROPIC_API_KEY,\n AZURE_OPENAI_API_VERSION,\n AZURE_OPENAI_DEPLOYMENT,\n AZURE_OPENAI_ENDPOINT,\n AZURE_OPENAI_KEY,\n MIDSCENE_API_TYPE,\n MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON,\n MIDSCENE_AZURE_OPENAI_SCOPE,\n MIDSCENE_DEBUG_AI_PROFILE,\n MIDSCENE_DEBUG_AI_RESPONSE,\n MIDSCENE_LANGSMITH_DEBUG,\n MIDSCENE_MODEL_NAME,\n MIDSCENE_OPENAI_HTTP_PROXY,\n MIDSCENE_OPENAI_INIT_CONFIG_JSON,\n MIDSCENE_OPENAI_SOCKS_PROXY,\n MIDSCENE_USE_ANTHROPIC_SDK,\n MIDSCENE_USE_AZURE_OPENAI,\n OPENAI_API_KEY,\n OPENAI_BASE_URL,\n OPENAI_MAX_TOKENS,\n OPENAI_USE_AZURE,\n getAIConfig,\n getAIConfigInBoolean,\n getAIConfigInJson,\n uiTarsModelVersion,\n vlLocateMode,\n} from '@midscene/shared/env';\nimport { parseBase64 } from '@midscene/shared/img';\nimport { enableDebug, 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\nexport function checkAIConfig() {\n const openaiKey = getAIConfig(OPENAI_API_KEY);\n const azureConfig = getAIConfig(MIDSCENE_USE_AZURE_OPENAI);\n const anthropicKey = getAIConfig(ANTHROPIC_API_KEY);\n const initConfigJson = getAIConfig(MIDSCENE_OPENAI_INIT_CONFIG_JSON);\n\n if (openaiKey) return true;\n if (azureConfig) return true;\n if (anthropicKey) return true;\n\n return Boolean(initConfigJson);\n}\n\n// if debug config is initialized\nlet debugConfigInitialized = false;\n\nfunction initDebugConfig() {\n // if debug config is initialized, return\n if (debugConfigInitialized) return;\n\n const shouldPrintTiming = getAIConfigInBoolean(MIDSCENE_DEBUG_AI_PROFILE);\n let debugConfig = '';\n if (shouldPrintTiming) {\n console.warn(\n 'MIDSCENE_DEBUG_AI_PROFILE is deprecated, use DEBUG=midscene:ai:profile instead',\n );\n debugConfig = 'ai:profile';\n }\n const shouldPrintAIResponse = getAIConfigInBoolean(\n MIDSCENE_DEBUG_AI_RESPONSE,\n );\n if (shouldPrintAIResponse) {\n console.warn(\n 'MIDSCENE_DEBUG_AI_RESPONSE is deprecated, use DEBUG=midscene:ai:response instead',\n );\n if (debugConfig) {\n debugConfig = 'ai:*';\n } else {\n debugConfig = 'ai:call';\n }\n }\n if (debugConfig) {\n enableDebug(debugConfig);\n }\n\n // mark as initialized\n debugConfigInitialized = true;\n}\n\n// default model\nconst defaultModel = 'gpt-4o';\nexport function getModelName() {\n let modelName = defaultModel;\n const nameInConfig = getAIConfig(MIDSCENE_MODEL_NAME);\n if (nameInConfig) {\n modelName = nameInConfig;\n }\n return modelName;\n}\n\nasync function createChatClient({\n AIActionTypeValue,\n}: {\n AIActionTypeValue: AIActionType;\n}): Promise<{\n completion: OpenAI.Chat.Completions;\n style: 'openai' | 'anthropic';\n}> {\n initDebugConfig();\n let openai: OpenAI | AzureOpenAI | undefined;\n const extraConfig = getAIConfigInJson(MIDSCENE_OPENAI_INIT_CONFIG_JSON);\n\n const socksProxy = getAIConfig(MIDSCENE_OPENAI_SOCKS_PROXY);\n const httpProxy = getAIConfig(MIDSCENE_OPENAI_HTTP_PROXY);\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 (getAIConfig(OPENAI_USE_AZURE)) {\n // this is deprecated\n openai = new AzureOpenAI({\n baseURL: getAIConfig(OPENAI_BASE_URL),\n apiKey: getAIConfig(OPENAI_API_KEY),\n httpAgent: proxyAgent,\n ...extraConfig,\n dangerouslyAllowBrowser: true,\n }) as OpenAI;\n } else if (getAIConfig(MIDSCENE_USE_AZURE_OPENAI)) {\n const extraAzureConfig = getAIConfigInJson(\n MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON,\n );\n\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 const scope = getAIConfig(MIDSCENE_AZURE_OPENAI_SCOPE);\n let tokenProvider: any = undefined;\n if (scope) {\n assert(\n !ifInBrowser,\n 'Azure OpenAI is not supported in browser with Midscene.',\n );\n const credential = new DefaultAzureCredential();\n\n assert(scope, 'MIDSCENE_AZURE_OPENAI_SCOPE is required');\n tokenProvider = getBearerTokenProvider(credential, scope);\n\n openai = new AzureOpenAI({\n azureADTokenProvider: tokenProvider,\n endpoint: getAIConfig(AZURE_OPENAI_ENDPOINT),\n apiVersion: getAIConfig(AZURE_OPENAI_API_VERSION),\n deployment: getAIConfig(AZURE_OPENAI_DEPLOYMENT),\n ...extraConfig,\n ...extraAzureConfig,\n });\n } else {\n // endpoint, apiKey, apiVersion, deployment\n openai = new AzureOpenAI({\n apiKey: getAIConfig(AZURE_OPENAI_KEY),\n endpoint: getAIConfig(AZURE_OPENAI_ENDPOINT),\n apiVersion: getAIConfig(AZURE_OPENAI_API_VERSION),\n deployment: getAIConfig(AZURE_OPENAI_DEPLOYMENT),\n dangerouslyAllowBrowser: true,\n ...extraConfig,\n ...extraAzureConfig,\n });\n }\n } else if (!getAIConfig(MIDSCENE_USE_ANTHROPIC_SDK)) {\n const baseURL = getAIConfig(OPENAI_BASE_URL);\n if (typeof baseURL === 'string') {\n if (!/^https?:\\/\\//.test(baseURL)) {\n throw new Error(\n `OPENAI_BASE_URL must be a valid URL starting with http:// or https://, but got: ${baseURL}\\nPlease check your config.`,\n );\n }\n }\n\n openai = new OpenAI({\n baseURL: getAIConfig(OPENAI_BASE_URL),\n apiKey: getAIConfig(OPENAI_API_KEY),\n httpAgent: proxyAgent,\n ...extraConfig,\n defaultHeaders: {\n ...(extraConfig?.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 };\n }\n\n // Anthropic\n if (getAIConfig(MIDSCENE_USE_ANTHROPIC_SDK)) {\n const apiKey = getAIConfig(ANTHROPIC_API_KEY);\n assert(apiKey, 'ANTHROPIC_API_KEY is required');\n openai = new Anthropic({\n apiKey,\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 };\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 responseFormat?:\n | OpenAI.ChatCompletionCreateParams['response_format']\n | OpenAI.ResponseFormatJSONObject,\n options?: {\n stream?: boolean;\n onChunk?: StreamingCallback;\n },\n): Promise<{ content: string; usage?: AIUsageInfo; isStreamed: boolean }> {\n assert(\n checkAIConfig(),\n 'Cannot find config for AI model service. If you are using a self-hosted model without validating the API key, please set `OPENAI_API_KEY` to any non-null value. https://midscenejs.com/model-provider.html',\n );\n\n const { completion, style } = await createChatClient({\n AIActionTypeValue,\n });\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 const model = getModelName();\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() === '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() === '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 ${model}`,\n );\n\n if (isStreaming) {\n const stream = (await completion.create(\n {\n model,\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: model,\n },\n };\n options.onChunk!(finalChunk);\n break;\n }\n }\n content = accumulated;\n debugProfileStats(\n `streaming model, ${model}, mode, ${vlLocateMode() || 'default'}, cost-ms, ${timeCost}`,\n );\n } else {\n const result = await completion.create({\n model,\n messages,\n response_format: responseFormat,\n ...commonConfig,\n } as any);\n timeCost = Date.now() - startTime;\n\n debugProfileStats(\n `model, ${model}, mode, ${vlLocateMode() || 'default'}, ui-tars-version, ${uiTarsModelVersion()}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}`,\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,\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: model,\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,\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: model,\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: ${e.message}. Trouble shooting: https://midscenejs.com/model-provider.html`,\n {\n cause: e,\n },\n );\n throw newError;\n }\n}\n\nexport async function callToGetJSONObject<T>(\n messages: ChatCompletionMessageParam[],\n AIActionTypeValue: AIActionType,\n): Promise<{ content: T; usage?: AIUsageInfo }> {\n let responseFormat:\n | OpenAI.ChatCompletionCreateParams['response_format']\n | OpenAI.ResponseFormatJSONObject\n | undefined;\n\n const model = getModelName();\n\n if (model.includes('gpt-4')) {\n switch (AIActionTypeValue) {\n case AIActionType.ASSERT:\n responseFormat = assertSchema;\n break;\n case AIActionType.INSPECT_ELEMENT:\n responseFormat = locatorSchema;\n break;\n case AIActionType.PLAN:\n responseFormat = planSchema;\n break;\n case AIActionType.EXTRACT_DATA:\n case AIActionType.DESCRIBE_ELEMENT:\n responseFormat = { type: AIResponseFormat.JSON };\n break;\n }\n }\n\n // gpt-4o-2024-05-13 only supports json_object response format\n if (model === 'gpt-4o-2024-05-13') {\n responseFormat = { type: AIResponseFormat.JSON };\n }\n\n const response = await call(messages, AIActionTypeValue, responseFormat);\n assert(response, 'empty response');\n const jsonContent = safeParseJson(response.content);\n return { content: jsonContent, usage: response.usage };\n}\n\nexport async function callAiFnWithStringResponse<T>(\n msgs: AIArgs,\n AIActionTypeValue: AIActionType,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n const { content, usage } = await call(msgs, AIActionTypeValue);\n return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n try {\n // First, try to match a JSON object directly in the response\n const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\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) {\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 (vlLocateMode() === 'doubao-vision' || vlLocateMode() === '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"],"names":["__webpack_require__","module","getter","definition","key","Object","obj","prop","Symbol","checkAIConfig","openaiKey","getAIConfig","OPENAI_API_KEY","azureConfig","MIDSCENE_USE_AZURE_OPENAI","anthropicKey","ANTHROPIC_API_KEY","initConfigJson","MIDSCENE_OPENAI_INIT_CONFIG_JSON","Boolean","debugConfigInitialized","initDebugConfig","shouldPrintTiming","getAIConfigInBoolean","MIDSCENE_DEBUG_AI_PROFILE","debugConfig","console","shouldPrintAIResponse","MIDSCENE_DEBUG_AI_RESPONSE","enableDebug","defaultModel","getModelName","modelName","nameInConfig","MIDSCENE_MODEL_NAME","createChatClient","AIActionTypeValue","openai","extraConfig","getAIConfigInJson","socksProxy","MIDSCENE_OPENAI_SOCKS_PROXY","httpProxy","MIDSCENE_OPENAI_HTTP_PROXY","proxyAgent","debugProxy","getDebug","HttpsProxyAgent","SocksProxyAgent","OPENAI_USE_AZURE","AzureOpenAI","OPENAI_BASE_URL","extraAzureConfig","MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON","scope","MIDSCENE_AZURE_OPENAI_SCOPE","tokenProvider","assert","ifInBrowser","credential","DefaultAzureCredential","getBearerTokenProvider","AZURE_OPENAI_ENDPOINT","AZURE_OPENAI_API_VERSION","AZURE_OPENAI_DEPLOYMENT","AZURE_OPENAI_KEY","MIDSCENE_USE_ANTHROPIC_SDK","baseURL","Error","OpenAI","MIDSCENE_API_TYPE","MIDSCENE_LANGSMITH_DEBUG","wrapOpenAI","apiKey","Anthropic","call","messages","responseFormat","options","completion","style","maxTokens","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","model","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","callToGetJSONObject","AIActionType","assertSchema","locatorSchema","planSchema","AIResponseFormat","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;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;IC4CO,SAASI;QACd,MAAMC,YAAYC,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYC,oBAAAA,cAAcA;QAC5C,MAAMC,cAAcF,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYG,oBAAAA,yBAAyBA;QACzD,MAAMC,eAAeJ,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYK,oBAAAA,iBAAiBA;QAClD,MAAMC,iBAAiBN,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYO,oBAAAA,gCAAgCA;QAEnE,IAAIR,WAAW,OAAO;QACtB,IAAIG,aAAa,OAAO;QACxB,IAAIE,cAAc,OAAO;QAEzB,OAAOI,QAAQF;IACjB;IAGA,IAAIG,yBAAyB;IAE7B,SAASC;QAEP,IAAID,wBAAwB;QAE5B,MAAME,oBAAoBC,AAAAA,IAAAA,oBAAAA,oBAAAA,AAAAA,EAAqBC,oBAAAA,yBAAyBA;QACxE,IAAIC,cAAc;QAClB,IAAIH,mBAAmB;YACrBI,QAAQ,IAAI,CACV;YAEFD,cAAc;QAChB;QACA,MAAME,wBAAwBJ,AAAAA,IAAAA,oBAAAA,oBAAAA,AAAAA,EAC5BK,oBAAAA,0BAA0BA;QAE5B,IAAID,uBAAuB;YACzBD,QAAQ,IAAI,CACV;YAGAD,cADEA,cACY,SAEA;QAElB;QACA,IAAIA,aACFI,AAAAA,IAAAA,uBAAAA,WAAAA,AAAAA,EAAYJ;QAIdL,yBAAyB;IAC3B;IAGA,MAAMU,eAAe;IACd,SAASC;QACd,IAAIC,YAAYF;QAChB,MAAMG,eAAetB,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYuB,oBAAAA,mBAAmBA;QACpD,IAAID,cACFD,YAAYC;QAEd,OAAOD;IACT;IAEA,eAAeG,iBAAiB,EAC9BC,iBAAiB,EAGlB;QAICf;QACA,IAAIgB;QACJ,MAAMC,cAAcC,AAAAA,IAAAA,oBAAAA,iBAAAA,AAAAA,EAAkBrB,oBAAAA,gCAAgCA;QAEtE,MAAMsB,aAAa7B,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAY8B,oBAAAA,2BAA2BA;QAC1D,MAAMC,YAAY/B,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYgC,oBAAAA,0BAA0BA;QAExD,IAAIC;QACJ,MAAMC,aAAaC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QAC5B,IAAIJ,WAAW;YACbG,WAAW,oBAAoBH;YAC/BE,aAAa,IAAIG,2CAAAA,eAAeA,CAACL;QACnC,OAAO,IAAIF,YAAY;YACrBK,WAAW,qBAAqBL;YAChCI,aAAa,IAAII,2CAAAA,eAAeA,CAACR;QACnC;QAEA,IAAI7B,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYsC,oBAAAA,gBAAgBA,GAE9BZ,SAAS,IAAIa,gCAAAA,WAAWA,CAAC;YACvB,SAASvC,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYwC,oBAAAA,eAAeA;YACpC,QAAQxC,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYC,oBAAAA,cAAcA;YAClC,WAAWgC;YACX,GAAGN,WAAW;YACd,yBAAyB;QAC3B;aACK,IAAI3B,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYG,oBAAAA,yBAAyBA,GAAG;YACjD,MAAMsC,mBAAmBb,AAAAA,IAAAA,oBAAAA,iBAAAA,AAAAA,EACvBc,oBAAAA,sCAAsCA;YAKxC,MAAMC,QAAQ3C,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAY4C,oBAAAA,2BAA2BA;YACrD,IAAIC;YACJ,IAAIF,OAAO;gBACTG,IAAAA,sBAAAA,MAAAA,AAAAA,EACE,CAACC,sBAAAA,WAAWA,EACZ;gBAEF,MAAMC,aAAa,IAAIC,yBAAAA,sBAAsBA;gBAE7CH,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOH,OAAO;gBACdE,gBAAgBK,AAAAA,IAAAA,yBAAAA,sBAAAA,AAAAA,EAAuBF,YAAYL;gBAEnDjB,SAAS,IAAIa,gCAAAA,WAAWA,CAAC;oBACvB,sBAAsBM;oBACtB,UAAU7C,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYmD,oBAAAA,qBAAqBA;oBAC3C,YAAYnD,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYoD,oBAAAA,wBAAwBA;oBAChD,YAAYpD,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYqD,oBAAAA,uBAAuBA;oBAC/C,GAAG1B,WAAW;oBACd,GAAGc,gBAAgB;gBACrB;YACF,OAEEf,SAAS,IAAIa,gCAAAA,WAAWA,CAAC;gBACvB,QAAQvC,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYsD,oBAAAA,gBAAgBA;gBACpC,UAAUtD,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYmD,oBAAAA,qBAAqBA;gBAC3C,YAAYnD,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYoD,oBAAAA,wBAAwBA;gBAChD,YAAYpD,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYqD,oBAAAA,uBAAuBA;gBAC/C,yBAAyB;gBACzB,GAAG1B,WAAW;gBACd,GAAGc,gBAAgB;YACrB;QAEJ,OAAO,IAAI,CAACzC,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYuD,oBAAAA,0BAA0BA,GAAG;YACnD,MAAMC,UAAUxD,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYwC,oBAAAA,eAAeA;YAC3C,IAAI,AAAmB,YAAnB,OAAOgB,SACT;gBAAA,IAAI,CAAC,eAAe,IAAI,CAACA,UACvB,MAAM,IAAIC,MACR,CAAC,gFAAgF,EAAED,QAAQ,2BAA2B,CAAC;YAE3H;YAGF9B,SAAS,IAAIgC,CAAAA,yBAAAA,EAAO;gBAClB,SAAS1D,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYwC,oBAAAA,eAAeA;gBACpC,QAAQxC,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYC,oBAAAA,cAAcA;gBAClC,WAAWgC;gBACX,GAAGN,WAAW;gBACd,gBAAgB;oBACd,GAAIA,AAAAA,CAAAA,QAAAA,cAAAA,KAAAA,IAAAA,YAAa,cAAc,AAAD,KAAK,CAAC,CAAC;oBACrC,CAACgC,oBAAAA,iBAAiBA,CAAC,EAAElC,kBAAkB,QAAQ;gBACjD;gBACA,yBAAyB;YAC3B;QACF;QAEA,IAAIC,UAAUd,AAAAA,IAAAA,oBAAAA,oBAAAA,AAAAA,EAAqBgD,oBAAAA,wBAAwBA,GAAG;YAC5D,IAAIb,sBAAAA,WAAWA,EACb,MAAM,IAAIU,MAAM;YAElB1C,QAAQ,GAAG,CAAC;YACZ,MAAM,EAAE8C,UAAU,EAAE,GAAG,MAAM;YAC7BnC,SAASmC,WAAWnC;QACtB;QAEA,IAAI,AAAkB,WAAXA,QACT,OAAO;YACL,YAAYA,OAAO,IAAI,CAAC,WAAW;YACnC,OAAO;QACT;QAIF,IAAI1B,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYuD,oBAAAA,0BAA0BA,GAAG;YAC3C,MAAMO,SAAS9D,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYK,oBAAAA,iBAAiBA;YAC5CyC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOgB,QAAQ;YACfpC,SAAS,IAAIqC,oBAAAA,SAASA,CAAC;gBACrBD;gBACA,WAAW7B;gBACX,yBAAyB;YAC3B;QACF;QAEA,IAAI,AAAkB,WAAXP,UAA2BA,OAAe,QAAQ,EAC3D,OAAO;YACL,YAAaA,OAAe,QAAQ;YACpC,OAAO;QACT;QAGF,MAAM,IAAI+B,MAAM;IAClB;IAEO,eAAeO,KACpBC,QAAsC,EACtCxC,iBAA+B,EAC/ByC,cAEmC,EACnCC,OAGC;QAEDrB,IAAAA,sBAAAA,MAAAA,AAAAA,EACEhD,iBACA;QAGF,MAAM,EAAEsE,UAAU,EAAEC,KAAK,EAAE,GAAG,MAAM7C,iBAAiB;YACnDC;QACF;QAEA,MAAM6C,YAAYtE,AAAAA,IAAAA,oBAAAA,WAAAA,AAAAA,EAAYuE,oBAAAA,iBAAiBA;QAC/C,MAAMC,YAAYrC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QAC3B,MAAMsC,oBAAoBtC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QACnC,MAAMuC,qBAAqBvC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;QAEpC,MAAMwC,YAAYC,KAAK,GAAG;QAC1B,MAAMC,QAAQzD;QACd,MAAM0D,cAAcX,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM,AAAD,KAAKA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO,AAAD;QACtD,IAAIY;QACJ,IAAIC,cAAc;QAClB,IAAIC;QACJ,IAAIC;QAEJ,MAAMC,eAAe;YACnB,aAAaC,AAAmB,kBAAnBA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,MAAmC,MAAM;YACtD,QAAQ,CAAC,CAACN;YACV,YACE,AAAqB,YAArB,OAAOR,YACHA,YACAe,OAAO,QAAQ,CAACf,aAAa,QAAQ;YAC3C,GAAIc,AAAmB,cAAnBA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,MACA;gBACE,2BAA2B;YAC7B,IACA,CAAC,CAAC;QACR;QAEA,IAAI;YACF,IAAIf,AAAU,aAAVA,OAAoB;gBACtBG,UACE,CAAC,QAAQ,EAAEM,cAAc,eAAe,GAAG,WAAW,EAAED,OAAO;gBAGjE,IAAIC,aAAa;oBACf,MAAMQ,SAAU,MAAMlB,WAAW,MAAM,CACrC;wBACES;wBACAZ;wBACA,iBAAiBC;wBACjB,GAAGiB,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;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;4BACrCR,WAAWN,KAAK,GAAG,KAAKD;4BAGxB,IAAI,CAACM,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,YAAYL;gCACd;4BACF;4BACAV,QAAQ,OAAO,CAAE6B;4BACjB;wBACF;oBACF;oBACAjB,UAAUC;oBACVP,kBACE,CAAC,iBAAiB,EAAEI,MAAM,QAAQ,EAAEO,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,OAAkB,UAAU,WAAW,EAAEF,UAAU;gBAE3F,OAAO;wBAUgHe,eAAyDC,gBAAwDC;oBATtO,MAAMC,SAAS,MAAMhC,WAAW,MAAM,CAAC;wBACrCS;wBACAZ;wBACA,iBAAiBC;wBACjB,GAAGiB,YAAY;oBACjB;oBACAD,WAAWN,KAAK,GAAG,KAAKD;oBAExBF,kBACE,CAAC,OAAO,EAAEI,MAAM,QAAQ,EAAEO,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,OAAkB,UAAU,mBAAmB,EAAEiB,AAAAA,IAAAA,oBAAAA,kBAAAA,AAAAA,IAAqB,iBAAiB,EAAEJ,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;oBAGtU1B,mBACE,CAAC,oBAAoB,EAAE4B,KAAK,SAAS,CAACF,OAAO,KAAK,GAAG;oBAGvDtD,IAAAA,sBAAAA,MAAAA,AAAAA,EACEsD,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;gBAEA5B,UAAU,CAAC,UAAU,EAAEO,SAAS;gBAChCjC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOiC,SAAS;YAClB,OAAO,IAAIV,AAAU,gBAAVA,OAAuB;gBAChC,MAAMkC,sBAAsB,CAACxB;oBAC3B,IAAIA,AAAiB,gBAAjBA,QAAQ,IAAI,EAAkB;wBAChC,MAAMyB,YAAYzB,QAAQ,SAAS,CAAC,GAAG;wBACvCjC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO0D,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,MAAMlB,WAAW,MAAM,CAAC;wBACtCS;wBACA,QAAQ;wBACR,UAAUZ,SAAS,GAAG,CAAC,CAAC2C,IAAO;gCAC7B,MAAM;gCACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;4BACf;wBACA,iBAAiB1C;wBACjB,GAAGiB,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;4BACA1B,QAAQ,OAAO,CAAEyB;wBACnB;wBAGA,IAAIL,AAAe,mBAAfA,MAAM,IAAI,EAAqB;4BACjCL,WAAWN,KAAK,GAAG,KAAKD;4BACxB,MAAMoC,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,YAAYL;gCACd,IACAgB;4BACN;4BACA1B,QAAQ,OAAO,CAAE6B;4BACjB;wBACF;oBACF;oBACAjB,UAAUC;gBACZ,OAAO;oBACL,MAAMoB,SAAS,MAAMhC,WAAW,MAAM,CAAC;wBACrCS;wBACA,QAAQ;wBACR,UAAUZ,SAAS,GAAG,CAAC,CAAC2C,IAAO;gCAC7B,MAAM;gCACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;4BACf;wBACA,iBAAiB1C;wBACjB,GAAGiB,YAAY;oBACjB;oBACAD,WAAWN,KAAK,GAAG,KAAKD;oBACxBI,UAAWqB,OAAe,OAAO,CAAC,EAAE,CAAC,IAAI;oBACzCnB,QAAQmB,OAAO,KAAK;gBACtB;gBAEAtD,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOiC,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,YAAYL;gBACd,IACAgB;gBACJ,YAAY,CAAC,CAACf;YAChB;QACF,EAAE,OAAOkC,GAAQ;YACfjG,QAAQ,KAAK,CAAC,kBAAkBiG;YAChC,MAAMC,WAAW,IAAIxD,MACnB,CAAC,eAAe,EAAEqB,cAAc,eAAe,GAAG,kBAAkB,EAAEkC,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC/I;gBACE,OAAOA;YACT;YAEF,MAAMC;QACR;IACF;IAEO,eAAeC,oBACpBjD,QAAsC,EACtCxC,iBAA+B;QAE/B,IAAIyC;QAKJ,MAAMW,QAAQzD;QAEd,IAAIyD,MAAM,QAAQ,CAAC,UACjB,OAAQpD;YACN,KAAK0F,mCAAAA,YAAAA,CAAAA,MAAmB;gBACtBjD,iBAAiBkD,6BAAAA,YAAYA;gBAC7B;YACF,KAAKD,mCAAAA,YAAAA,CAAAA,eAA4B;gBAC/BjD,iBAAiBmD,+BAAAA,aAAaA;gBAC9B;YACF,KAAKF,mCAAAA,YAAAA,CAAAA,IAAiB;gBACpBjD,iBAAiBoD,gCAAAA,UAAUA;gBAC3B;YACF,KAAKH,mCAAAA,YAAAA,CAAAA,YAAyB;YAC9B,KAAKA,mCAAAA,YAAAA,CAAAA,gBAA6B;gBAChCjD,iBAAiB;oBAAE,MAAMqD,kCAAAA,gBAAAA,CAAAA,IAAqB;gBAAC;gBAC/C;QACJ;QAIF,IAAI1C,AAAU,wBAAVA,OACFX,iBAAiB;YAAE,MAAMqD,kCAAAA,gBAAAA,CAAAA,IAAqB;QAAC;QAGjD,MAAMC,WAAW,MAAMxD,KAAKC,UAAUxC,mBAAmByC;QACzDpB,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO0E,UAAU;QACjB,MAAMC,cAAcC,cAAcF,SAAS,OAAO;QAClD,OAAO;YAAE,SAASC;YAAa,OAAOD,SAAS,KAAK;QAAC;IACvD;IAEO,eAAeG,2BACpBC,IAAY,EACZnG,iBAA+B;QAE/B,MAAM,EAAEsD,OAAO,EAAEE,KAAK,EAAE,GAAG,MAAMjB,KAAK4D,MAAMnG;QAC5C,OAAO;YAAEsD;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,cAAcQ,KAAa;QACzC,MAAMC,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,IAAI5B,AAAmB,oBAAnBA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,OAAsCA,AAAmB,kBAAnBA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,KAAkC;YAC1E,MAAMkD,aAAaL,yBAAyBE;YAC5C,OAAO7B,KAAK,KAAK,CAAC+B,AAAAA,IAAAA,oCAAAA,UAAAA,AAAAA,EAAWC;QAC/B;QACA,MAAM7E,MAAM,CAAC,+BAA+B,EAAEyE,OAAO;IACvD"}
|
|
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"}
|
|
@@ -45,7 +45,7 @@ const pointToBbox = (point, width, height)=>[
|
|
|
45
45
|
Math.round(Math.min(point.y + bboxSize / 2, height))
|
|
46
46
|
];
|
|
47
47
|
async function vlmPlanning(options) {
|
|
48
|
-
const { conversationHistory, userInstruction, size } = options;
|
|
48
|
+
const { conversationHistory, userInstruction, size, modelPreferences } = options;
|
|
49
49
|
const systemPrompt = (0, ui_tars_planning_js_namespaceObject.getUiTarsPlanningPrompt)() + userInstruction;
|
|
50
50
|
const res = await (0, index_js_namespaceObject.call)([
|
|
51
51
|
{
|
|
@@ -53,9 +53,9 @@ async function vlmPlanning(options) {
|
|
|
53
53
|
content: systemPrompt
|
|
54
54
|
},
|
|
55
55
|
...conversationHistory
|
|
56
|
-
], external_common_js_namespaceObject.AIActionType.INSPECT_ELEMENT);
|
|
56
|
+
], external_common_js_namespaceObject.AIActionType.INSPECT_ELEMENT, modelPreferences);
|
|
57
57
|
const convertedText = convertBboxToCoordinates(res.content);
|
|
58
|
-
const modelVer = (0, env_namespaceObject.uiTarsModelVersion)();
|
|
58
|
+
const modelVer = (0, env_namespaceObject.uiTarsModelVersion)(modelPreferences);
|
|
59
59
|
const { parsed } = (0, action_parser_namespaceObject.actionParser)({
|
|
60
60
|
prediction: convertedText,
|
|
61
61
|
factor: [
|
|
@@ -240,8 +240,8 @@ function getPoint(startBox, size) {
|
|
|
240
240
|
y * size.height
|
|
241
241
|
];
|
|
242
242
|
}
|
|
243
|
-
async function resizeImageForUiTars(imageBase64, size) {
|
|
244
|
-
if ('vlm-ui-tars' === (0, env_namespaceObject.vlLocateMode)() && (0, env_namespaceObject.uiTarsModelVersion)() === env_namespaceObject.UITarsModelVersion.V1_5) {
|
|
243
|
+
async function resizeImageForUiTars(imageBase64, size, modelPreferences) {
|
|
244
|
+
if ('vlm-ui-tars' === (0, env_namespaceObject.vlLocateMode)(modelPreferences) && (0, env_namespaceObject.uiTarsModelVersion)(modelPreferences) === env_namespaceObject.UITarsModelVersion.V1_5) {
|
|
245
245
|
debug('ui-tars-v1.5, will check image size', size);
|
|
246
246
|
const currentPixels = size.width * size.height;
|
|
247
247
|
const maxPixels = 12845056;
|
|
@@ -1 +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 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}): 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 } = 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 );\n const convertedText = convertBboxToCoordinates(res.content);\n\n const modelVer = uiTarsModelVersion();\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(imageBase64: string, size: Size) {\n if (\n vlLocateMode() === 'vlm-ui-tars' &&\n uiTarsModelVersion() === 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","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;;;;;;;;;;;;;;;;;AC4BA,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,OAIjC;IAQC,MAAM,EAAEC,mBAAmB,EAAEC,eAAe,EAAEC,IAAI,EAAE,GAAGH;IACvD,MAAMI,eAAeC,AAAAA,IAAAA,oCAAAA,uBAAAA,AAAAA,MAA4BH;IAEjD,MAAMI,MAAM,MAAMC,AAAAA,IAAAA,yBAAAA,IAAAA,AAAAA,EAChB;QACE;YACE,MAAM;YACN,SAASH;QACX;WACGH;KACJ,EACDO,mCAAAA,YAAAA,CAAAA,eAA4B;IAE9B,MAAMC,gBAAgBC,yBAAyBJ,IAAI,OAAO;IAE1D,MAAMK,WAAWC,AAAAA,IAAAA,oBAAAA,kBAAAA,AAAAA;IAEjB,MAAM,EAAEC,MAAM,EAAE,GAAGC,AAAAA,IAAAA,8BAAAA,YAAAA,AAAAA,EAAa;QAC9B,YAAYL;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAON,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUQ,YAAYI;IACxB;IAEAxB,MAAM,YAAYoB,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,MAAMvB,QAAQyB,SAASF,OAAO,aAAa,CAAC,SAAS,EAAEf;YACvDc,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO,CAAC;gBACR,QAAQ;oBACN,QAAQC,OAAO,OAAO,IAAI;oBAC1B,MAAMxB,YACJ;wBAAE,GAAGC,KAAK,CAAC,EAAE;wBAAE,GAAGA,KAAK,CAAC,EAAE;oBAAC,GAC3BQ,KAAK,KAAK,EACVA,KAAK,MAAM;gBAEf;YACF;YACAc,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,QAAQ;oBACN,QAAQC,OAAO,OAAO,IAAI;oBAC1B,MAAMxB,YACJ;wBAAE,GAAGC,KAAK,CAAC,EAAE;wBAAE,GAAGA,KAAK,CAAC,EAAE;oBAAC,GAC3BQ,KAAK,KAAK,EACVA,KAAK,MAAM;gBAEf;gBACA,OAAOe,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,EAAEf;YAC5D,MAAMmB,WAAWF,SAASF,OAAO,aAAa,CAAC,OAAO,EAAEf;YACxDc,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,MAAMvB,QAAQuB,OAAO,aAAa,CAAC,YAAY;YAC/CD,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,GAAGtB,KAAK,CAAC,EAAE;oBACX,GAAGA,KAAK,CAAC,EAAE;oBACX,UAAU;gBACZ;gBACA,QAAQ;gBACR,SAASuB,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,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,SAASX,SAASyB,QAAgB,EAAE1C,IAAuC;IACzE,MAAM,CAACuC,GAAGC,EAAE,GAAG3B,KAAK,KAAK,CAAC6B;IAC1B,OAAO;QAACH,IAAIvC,KAAK,KAAK;QAAEwC,IAAIxC,KAAK,MAAM;KAAC;AAC1C;AA2EO,eAAe2C,qBAAqBC,WAAmB,EAAE5C,IAAU;IACxE,IACE6C,AAAmB,kBAAnBA,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,OACApC,AAAAA,IAAAA,oBAAAA,kBAAAA,AAAAA,QAAyBqC,oBAAAA,kBAAAA,CAAAA,IAAuB,EAChD;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"}
|
|
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"}
|
package/dist/lib/index.js
CHANGED
|
@@ -33,16 +33,22 @@ var __webpack_require__ = {};
|
|
|
33
33
|
var __webpack_exports__ = {};
|
|
34
34
|
__webpack_require__.r(__webpack_exports__);
|
|
35
35
|
__webpack_require__.d(__webpack_exports__, {
|
|
36
|
+
MidsceneLocation: ()=>external_ai_model_index_js_namespaceObject.MidsceneLocation,
|
|
36
37
|
MIDSCENE_MODEL_NAME: ()=>env_namespaceObject.MIDSCENE_MODEL_NAME,
|
|
38
|
+
RectSchema: ()=>external_ai_model_index_js_namespaceObject.RectSchema,
|
|
39
|
+
PointSchema: ()=>external_ai_model_index_js_namespaceObject.PointSchema,
|
|
40
|
+
SizeSchema: ()=>external_ai_model_index_js_namespaceObject.SizeSchema,
|
|
37
41
|
default: ()=>src,
|
|
38
|
-
getVersion: ()=>external_utils_js_namespaceObject.getVersion,
|
|
39
|
-
plan: ()=>external_ai_model_index_js_namespaceObject.plan,
|
|
40
42
|
getAIConfig: ()=>env_namespaceObject.getAIConfig,
|
|
41
43
|
AiLocateElement: ()=>external_ai_model_index_js_namespaceObject.AiLocateElement,
|
|
42
44
|
Insight: ()=>index_js_default(),
|
|
43
45
|
describeUserPage: ()=>external_ai_model_index_js_namespaceObject.describeUserPage,
|
|
44
|
-
Executor: ()=>action_executor_js_namespaceObject.Executor
|
|
46
|
+
Executor: ()=>action_executor_js_namespaceObject.Executor,
|
|
47
|
+
getVersion: ()=>external_utils_js_namespaceObject.getVersion,
|
|
48
|
+
plan: ()=>external_ai_model_index_js_namespaceObject.plan,
|
|
49
|
+
z: ()=>external_zod_namespaceObject.z
|
|
45
50
|
});
|
|
51
|
+
const external_zod_namespaceObject = require("zod");
|
|
46
52
|
const action_executor_js_namespaceObject = require("./ai-model/action-executor.js");
|
|
47
53
|
const index_js_namespaceObject = require("./insight/index.js");
|
|
48
54
|
var index_js_default = /*#__PURE__*/ __webpack_require__.n(index_js_namespaceObject);
|
|
@@ -54,21 +60,31 @@ exports.AiLocateElement = __webpack_exports__.AiLocateElement;
|
|
|
54
60
|
exports.Executor = __webpack_exports__.Executor;
|
|
55
61
|
exports.Insight = __webpack_exports__.Insight;
|
|
56
62
|
exports.MIDSCENE_MODEL_NAME = __webpack_exports__.MIDSCENE_MODEL_NAME;
|
|
63
|
+
exports.MidsceneLocation = __webpack_exports__.MidsceneLocation;
|
|
64
|
+
exports.PointSchema = __webpack_exports__.PointSchema;
|
|
65
|
+
exports.RectSchema = __webpack_exports__.RectSchema;
|
|
66
|
+
exports.SizeSchema = __webpack_exports__.SizeSchema;
|
|
57
67
|
exports["default"] = __webpack_exports__["default"];
|
|
58
68
|
exports.describeUserPage = __webpack_exports__.describeUserPage;
|
|
59
69
|
exports.getAIConfig = __webpack_exports__.getAIConfig;
|
|
60
70
|
exports.getVersion = __webpack_exports__.getVersion;
|
|
61
71
|
exports.plan = __webpack_exports__.plan;
|
|
72
|
+
exports.z = __webpack_exports__.z;
|
|
62
73
|
for(var __webpack_i__ in __webpack_exports__)if (-1 === [
|
|
63
74
|
"AiLocateElement",
|
|
64
75
|
"Executor",
|
|
65
76
|
"Insight",
|
|
66
77
|
"MIDSCENE_MODEL_NAME",
|
|
78
|
+
"MidsceneLocation",
|
|
79
|
+
"PointSchema",
|
|
80
|
+
"RectSchema",
|
|
81
|
+
"SizeSchema",
|
|
67
82
|
"default",
|
|
68
83
|
"describeUserPage",
|
|
69
84
|
"getAIConfig",
|
|
70
85
|
"getVersion",
|
|
71
|
-
"plan"
|
|
86
|
+
"plan",
|
|
87
|
+
"z"
|
|
72
88
|
].indexOf(__webpack_i__)) exports[__webpack_i__] = __webpack_exports__[__webpack_i__];
|
|
73
89
|
Object.defineProperty(exports, '__esModule', {
|
|
74
90
|
value: true
|
package/dist/lib/index.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"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/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 { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n plan,\n describeUserPage,\n AiLocateElement,\n} from './ai-model/index';\n\nexport { getAIConfig, MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n MidsceneYamlScript,\n MidsceneYamlTask,\n MidsceneYamlFlowItem,\n MidsceneYamlFlowItemAIRightClick,\n MidsceneYamlConfigResult,\n LocateOption,\n DetailedLocateParam,\n} from './yaml';\n"],"names":["__webpack_require__","module","getter","definition","key","Object","obj","prop","Symbol","Insight"],"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
|
|
1
|
+
{"version":3,"file":"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/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 { z } from 'zod';\nimport { Executor } from './ai-model/action-executor';\nimport Insight from './insight/index';\nimport { getVersion } from './utils';\n\nexport {\n plan,\n describeUserPage,\n AiLocateElement,\n MidsceneLocation,\n type MidsceneLocationType,\n PointSchema,\n SizeSchema,\n RectSchema,\n} from './ai-model/index';\n\nexport { getAIConfig, MIDSCENE_MODEL_NAME } from '@midscene/shared/env';\n\nexport type * from './types';\n\nexport { z };\n\nexport default Insight;\nexport { Executor, Insight, getVersion };\n\nexport type {\n MidsceneYamlScript,\n MidsceneYamlTask,\n MidsceneYamlFlowItem,\n MidsceneYamlFlowItemAIRightClick,\n MidsceneYamlConfigResult,\n LocateOption,\n DetailedLocateParam,\n} from './yaml';\n"],"names":["__webpack_require__","module","getter","definition","key","Object","obj","prop","Symbol","Insight"],"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;;;;;;;;;;;;;;;;;;;;;;;;;;;ACgBA,YAAeI"}
|
|
@@ -29,9 +29,7 @@ __webpack_require__.d(__webpack_exports__, {
|
|
|
29
29
|
const common_js_namespaceObject = require("../ai-model/common.js");
|
|
30
30
|
const index_js_namespaceObject = require("../ai-model/index.js");
|
|
31
31
|
const inspect_js_namespaceObject = require("../ai-model/inspect.js");
|
|
32
|
-
const describe_js_namespaceObject = require("../ai-model/prompt/describe.js");
|
|
33
32
|
const env_namespaceObject = require("@midscene/shared/env");
|
|
34
|
-
const img_namespaceObject = require("@midscene/shared/img");
|
|
35
33
|
const logger_namespaceObject = require("@midscene/shared/logger");
|
|
36
34
|
const utils_namespaceObject = require("@midscene/shared/utils");
|
|
37
35
|
const external_utils_js_namespaceObject = require("./utils.js");
|
|
@@ -59,7 +57,10 @@ class Insight {
|
|
|
59
57
|
if (globalDeepThinkSwitch) debug('globalDeepThinkSwitch', globalDeepThinkSwitch);
|
|
60
58
|
let searchAreaPrompt;
|
|
61
59
|
if (query.deepThink || globalDeepThinkSwitch) searchAreaPrompt = query.prompt;
|
|
62
|
-
|
|
60
|
+
const modelPreferences = {
|
|
61
|
+
intent: 'grounding'
|
|
62
|
+
};
|
|
63
|
+
if (searchAreaPrompt && !(0, env_namespaceObject.vlLocateMode)(modelPreferences)) {
|
|
63
64
|
console.warn('The "deepThink" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model');
|
|
64
65
|
searchAreaPrompt = void 0;
|
|
65
66
|
}
|
|
@@ -146,13 +147,17 @@ class Insight {
|
|
|
146
147
|
(0, utils_namespaceObject.assert)('object' == typeof dataDemand || 'string' == typeof dataDemand, `dataDemand should be object or string, but get ${typeof dataDemand}`);
|
|
147
148
|
const dumpSubscriber = this.onceDumpUpdatedFn;
|
|
148
149
|
this.onceDumpUpdatedFn = void 0;
|
|
150
|
+
const modelPreferences = {
|
|
151
|
+
intent: 'VQA'
|
|
152
|
+
};
|
|
149
153
|
const context = await this.contextRetrieverFn('extract');
|
|
150
154
|
const startTime = Date.now();
|
|
151
155
|
const { parseResult, usage } = await (0, index_js_namespaceObject.AiExtractElementInfo)({
|
|
152
156
|
context,
|
|
153
157
|
dataQuery: dataDemand,
|
|
154
158
|
multimodalPrompt,
|
|
155
|
-
extractOption: opt
|
|
159
|
+
extractOption: opt,
|
|
160
|
+
modelPreferences
|
|
156
161
|
});
|
|
157
162
|
const timeCost = Date.now() - startTime;
|
|
158
163
|
const taskInfo = {
|
|
@@ -177,66 +182,13 @@ class Insight {
|
|
|
177
182
|
...dumpData,
|
|
178
183
|
data
|
|
179
184
|
}, dumpSubscriber);
|
|
180
|
-
if (errorLog && !data) throw new Error(errorLog);
|
|
185
|
+
if (errorLog && !data && !(null == opt ? void 0 : opt.doNotThrowError)) throw new Error(errorLog);
|
|
181
186
|
return {
|
|
182
187
|
data,
|
|
183
188
|
thought,
|
|
184
189
|
usage
|
|
185
190
|
};
|
|
186
191
|
}
|
|
187
|
-
async describe(target, opt) {
|
|
188
|
-
(0, utils_namespaceObject.assert)(target, 'target is required for insight.describe');
|
|
189
|
-
const context = await this.contextRetrieverFn('describe');
|
|
190
|
-
const { screenshotBase64, size } = context;
|
|
191
|
-
(0, utils_namespaceObject.assert)(screenshotBase64, 'screenshot is required for insight.describe');
|
|
192
|
-
const systemPrompt = (0, describe_js_namespaceObject.elementDescriberInstruction)();
|
|
193
|
-
const defaultRectSize = 30;
|
|
194
|
-
const targetRect = Array.isArray(target) ? {
|
|
195
|
-
left: Math.floor(target[0] - defaultRectSize / 2),
|
|
196
|
-
top: Math.floor(target[1] - defaultRectSize / 2),
|
|
197
|
-
width: defaultRectSize,
|
|
198
|
-
height: defaultRectSize
|
|
199
|
-
} : target;
|
|
200
|
-
let imagePayload = await (0, img_namespaceObject.compositeElementInfoImg)({
|
|
201
|
-
inputImgBase64: screenshotBase64,
|
|
202
|
-
size,
|
|
203
|
-
elementsPositionInfo: [
|
|
204
|
-
{
|
|
205
|
-
rect: targetRect
|
|
206
|
-
}
|
|
207
|
-
],
|
|
208
|
-
borderThickness: 3
|
|
209
|
-
});
|
|
210
|
-
if (null == opt ? void 0 : opt.deepThink) {
|
|
211
|
-
const searchArea = (0, common_js_namespaceObject.expandSearchArea)(targetRect, context.size);
|
|
212
|
-
debug('describe: set searchArea', searchArea);
|
|
213
|
-
imagePayload = await (0, img_namespaceObject.cropByRect)(imagePayload, searchArea, (0, env_namespaceObject.getAIConfigInBoolean)(env_namespaceObject.MIDSCENE_USE_QWEN_VL));
|
|
214
|
-
}
|
|
215
|
-
const msgs = [
|
|
216
|
-
{
|
|
217
|
-
role: 'system',
|
|
218
|
-
content: systemPrompt
|
|
219
|
-
},
|
|
220
|
-
{
|
|
221
|
-
role: 'user',
|
|
222
|
-
content: [
|
|
223
|
-
{
|
|
224
|
-
type: 'image_url',
|
|
225
|
-
image_url: {
|
|
226
|
-
url: imagePayload,
|
|
227
|
-
detail: 'high'
|
|
228
|
-
}
|
|
229
|
-
}
|
|
230
|
-
]
|
|
231
|
-
}
|
|
232
|
-
];
|
|
233
|
-
const callAIFn = this.aiVendorFn || index_js_namespaceObject.callToGetJSONObject;
|
|
234
|
-
const res = await callAIFn(msgs, common_js_namespaceObject.AIActionType.DESCRIBE_ELEMENT);
|
|
235
|
-
const { content } = res;
|
|
236
|
-
(0, utils_namespaceObject.assert)(!content.error, `describe failed: ${content.error}`);
|
|
237
|
-
(0, utils_namespaceObject.assert)(content.description, 'failed to describe the element');
|
|
238
|
-
return content;
|
|
239
|
-
}
|
|
240
192
|
constructor(context, opt){
|
|
241
193
|
_define_property(this, "contextRetrieverFn", void 0);
|
|
242
194
|
_define_property(this, "aiVendorFn", common_js_namespaceObject.callAiFn);
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"insight/index.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/insight/index.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 {\n AIActionType,\n type AIArgs,\n callAiFn,\n expandSearchArea,\n} from '@/ai-model/common';\nimport {\n AiExtractElementInfo,\n AiLocateElement,\n callToGetJSONObject,\n} from '@/ai-model/index';\nimport { AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n AIDescribeElementResponse,\n AIElementResponse,\n AIUsageInfo,\n BaseElement,\n DetailedLocateParam,\n DumpSubscriber,\n InsightAction,\n InsightAssertionResponse,\n InsightExtractOption,\n InsightExtractParam,\n InsightOptions,\n InsightTaskInfo,\n LocateResult,\n PartialInsightDumpFromSDK,\n Rect,\n TMultimodalPrompt,\n TUserPrompt,\n UIContext,\n} from '@/types';\nimport {\n MIDSCENE_FORCE_DEEP_THINK,\n MIDSCENE_USE_QWEN_VL,\n getAIConfigInBoolean,\n vlLocateMode,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n context?: UIContext<BaseElement>;\n callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n ElementType extends BaseElement = BaseElement,\n ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n contextRetrieverFn: (\n action: InsightAction,\n ) => Promise<ContextType> | ContextType;\n\n aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n onceDumpUpdatedFn?: DumpSubscriber;\n\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n constructor(\n context:\n | ContextType\n | ((action: InsightAction) => Promise<ContextType> | ContextType),\n opt?: InsightOptions,\n ) {\n assert(context, 'context is required for Insight');\n if (typeof context === 'function') {\n this.contextRetrieverFn = context;\n } else {\n this.contextRetrieverFn = () => Promise.resolve(context);\n }\n\n if (typeof opt?.aiVendorFn !== 'undefined') {\n this.aiVendorFn = opt.aiVendorFn;\n }\n if (typeof opt?.taskInfo !== 'undefined') {\n this.taskInfo = opt.taskInfo;\n }\n }\n\n async locate(\n query: DetailedLocateParam,\n opt?: LocateOpts,\n ): Promise<LocateResult> {\n const { callAI } = opt || {};\n const queryPrompt = typeof query === 'string' ? query : query.prompt;\n assert(queryPrompt, 'query is required for locate');\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n assert(typeof query === 'object', 'query should be an object for locate');\n\n const globalDeepThinkSwitch = getAIConfigInBoolean(\n MIDSCENE_FORCE_DEEP_THINK,\n );\n if (globalDeepThinkSwitch) {\n debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n }\n let searchAreaPrompt;\n if (query.deepThink || globalDeepThinkSwitch) {\n searchAreaPrompt = query.prompt;\n }\n\n if (searchAreaPrompt && !vlLocateMode()) {\n console.warn(\n 'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n );\n searchAreaPrompt = undefined;\n }\n\n const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n let searchArea: Rect | undefined = undefined;\n let searchAreaRawResponse: string | undefined = undefined;\n let searchAreaUsage: AIUsageInfo | undefined = undefined;\n let searchAreaResponse:\n | Awaited<ReturnType<typeof AiLocateSection>>\n | undefined = undefined;\n if (searchAreaPrompt) {\n searchAreaResponse = await AiLocateSection({\n context,\n sectionDescription: searchAreaPrompt,\n });\n assert(\n searchAreaResponse.rect,\n `cannot find search area for \"${searchAreaPrompt}\"${\n searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n }`,\n );\n searchAreaRawResponse = searchAreaResponse.rawResponse;\n searchAreaUsage = searchAreaResponse.usage;\n searchArea = searchAreaResponse.rect;\n }\n\n const startTime = Date.now();\n const {\n parseResult,\n rect,\n elementById,\n rawResponse,\n usage,\n isOrderSensitive,\n } = await AiLocateElement({\n callAI: callAI || this.aiVendorFn,\n context,\n targetElementDescription: queryPrompt,\n searchConfig: searchAreaResponse,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(rawResponse),\n formatResponse: JSON.stringify(parseResult),\n usage,\n searchArea,\n searchAreaRawResponse,\n searchAreaUsage,\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'locate',\n userQuery: {\n element: queryPrompt,\n },\n matchedElement: [],\n matchedRect: rect,\n data: null,\n taskInfo,\n deepThink: !!searchArea,\n error: errorLog,\n };\n\n const elements: BaseElement[] = [];\n (parseResult.elements || []).forEach((item) => {\n if ('id' in item) {\n const element = elementById(item?.id);\n\n if (!element) {\n console.warn(\n `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n );\n return;\n }\n elements.push(element);\n }\n });\n\n emitInsightDump(\n {\n ...dumpData,\n matchedElement: elements,\n },\n dumpSubscriber,\n );\n\n if (errorLog) {\n throw new Error(errorLog);\n }\n\n assert(\n elements.length <= 1,\n `locate: multiple elements found, length = ${elements.length}`,\n );\n\n if (elements.length === 1) {\n return {\n element: {\n id: elements[0]!.id,\n indexId: elements[0]!.indexId,\n center: elements[0]!.center,\n rect: elements[0]!.rect,\n xpaths: elements[0]!.xpaths || [],\n attributes: elements[0]!.attributes,\n isOrderSensitive,\n },\n rect,\n };\n }\n return {\n element: null,\n rect,\n };\n }\n\n async extract<T>(\n dataDemand: InsightExtractParam,\n opt?: InsightExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ): Promise<{\n data: T;\n thought?: string;\n usage?: AIUsageInfo;\n }> {\n assert(\n typeof dataDemand === 'object' || typeof dataDemand === 'string',\n `dataDemand should be object or string, but get ${typeof dataDemand}`,\n );\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n const context = await this.contextRetrieverFn('extract');\n\n const startTime = Date.now();\n const { parseResult, usage } = await AiExtractElementInfo<T>({\n context,\n dataQuery: dataDemand,\n multimodalPrompt,\n extractOption: opt,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(parseResult),\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'extract',\n userQuery: {\n dataDemand,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n error: errorLog,\n };\n\n const { data, thought } = parseResult || {};\n\n // 4\n emitInsightDump(\n {\n ...dumpData,\n data,\n },\n dumpSubscriber,\n );\n\n if (errorLog && !data) {\n throw new Error(errorLog);\n }\n\n return {\n data,\n thought,\n usage,\n };\n }\n\n async describe(\n target: Rect | [number, number],\n opt?: {\n deepThink?: boolean;\n },\n ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n assert(target, 'target is required for insight.describe');\n const context = await this.contextRetrieverFn('describe');\n const { screenshotBase64, size } = context;\n assert(screenshotBase64, 'screenshot is required for insight.describe');\n\n const systemPrompt = elementDescriberInstruction();\n\n // Convert [x,y] center point to Rect if needed\n const defaultRectSize = 30;\n const targetRect: Rect = Array.isArray(target)\n ? {\n left: Math.floor(target[0] - defaultRectSize / 2),\n top: Math.floor(target[1] - defaultRectSize / 2),\n width: defaultRectSize,\n height: defaultRectSize,\n }\n : target;\n\n let imagePayload = await compositeElementInfoImg({\n inputImgBase64: screenshotBase64,\n size,\n elementsPositionInfo: [\n {\n rect: targetRect,\n },\n ],\n borderThickness: 3,\n });\n\n if (opt?.deepThink) {\n const searchArea = expandSearchArea(targetRect, context.size);\n debug('describe: set searchArea', searchArea);\n imagePayload = await cropByRect(\n imagePayload,\n searchArea,\n getAIConfigInBoolean(MIDSCENE_USE_QWEN_VL),\n );\n }\n\n const msgs: AIArgs = [\n { role: 'system', content: systemPrompt },\n {\n role: 'user',\n content: [\n {\n type: 'image_url',\n image_url: {\n url: imagePayload,\n detail: 'high',\n },\n },\n ],\n },\n ];\n\n const callAIFn =\n this.aiVendorFn || callToGetJSONObject<AIDescribeElementResponse>;\n\n const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT);\n\n const { content } = res;\n assert(!content.error, `describe failed: ${content.error}`);\n assert(content.description, 'failed to describe the element');\n return content;\n }\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","Insight","query","opt","_parseResult_errors","callAI","queryPrompt","assert","dumpSubscriber","undefined","globalDeepThinkSwitch","getAIConfigInBoolean","MIDSCENE_FORCE_DEEP_THINK","searchAreaPrompt","vlLocateMode","console","context","searchArea","searchAreaRawResponse","searchAreaUsage","searchAreaResponse","AiLocateSection","startTime","Date","parseResult","rect","elementById","rawResponse","usage","isOrderSensitive","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","elements","item","element","emitInsightDump","Error","dataDemand","multimodalPrompt","AiExtractElementInfo","data","thought","target","screenshotBase64","size","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","cropByRect","MIDSCENE_USE_QWEN_VL","msgs","callAIFn","callToGetJSONObject","res","AIActionType","content","callAiFn","Promise"],"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;;;;;;;;;;;;;;;;;;;;;;;;;;AC+CA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AACR,MAAMC;IAmCnB,MAAM,OACJC,KAA0B,EAC1BC,GAAgB,EACO;YA+EnBC;QA9EJ,MAAM,EAAEC,MAAM,EAAE,GAAGF,OAAO,CAAC;QAC3B,MAAMG,cAAc,AAAiB,YAAjB,OAAOJ,QAAqBA,QAAQA,MAAM,MAAM;QACpEK,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,aAAa;QACpB,MAAME,iBAAiB,IAAI,CAAC,iBAAiB;QAC7C,IAAI,CAAC,iBAAiB,GAAGC;QAEzBF,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,AAAiB,YAAjB,OAAOL,OAAoB;QAElC,MAAMQ,wBAAwBC,AAAAA,IAAAA,oBAAAA,oBAAAA,AAAAA,EAC5BC,oBAAAA,yBAAyBA;QAE3B,IAAIF,uBACFX,MAAM,yBAAyBW;QAEjC,IAAIG;QACJ,IAAIX,MAAM,SAAS,IAAIQ,uBACrBG,mBAAmBX,MAAM,MAAM;QAGjC,IAAIW,oBAAoB,CAACC,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,KAAgB;YACvCC,QAAQ,IAAI,CACV;YAEFF,mBAAmBJ;QACrB;QAEA,MAAMO,UAAUb,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAM,MAAM,IAAI,CAAC,kBAAkB,CAAC;QAE/D,IAAIc;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAGJ,IAAIP,kBAAkB;YACpBO,qBAAqB,MAAMC,AAAAA,IAAAA,2BAAAA,eAAAA,AAAAA,EAAgB;gBACzCL;gBACA,oBAAoBH;YACtB;YACAN,IAAAA,sBAAAA,MAAAA,AAAAA,EACEa,mBAAmB,IAAI,EACvB,CAAC,6BAA6B,EAAEP,iBAAiB,CAAC,EAChDO,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAEJF,wBAAwBE,mBAAmB,WAAW;YACtDD,kBAAkBC,mBAAmB,KAAK;YAC1CH,aAAaG,mBAAmB,IAAI;QACtC;QAEA,MAAME,YAAYC,KAAK,GAAG;QAC1B,MAAM,EACJC,WAAW,EACXC,IAAI,EACJC,WAAW,EACXC,WAAW,EACXC,KAAK,EACLC,gBAAgB,EACjB,GAAG,MAAMC,AAAAA,IAAAA,yBAAAA,eAAAA,AAAAA,EAAgB;YACxB,QAAQzB,UAAU,IAAI,CAAC,UAAU;YACjCW;YACA,0BAA0BV;YAC1B,cAAcc;QAChB;QAEA,MAAMW,WAAWR,KAAK,GAAG,KAAKD;QAC9B,MAAMU,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACN;YAC5B,gBAAgBM,KAAK,SAAS,CAACT;YAC/BI;YACAX;YACAC;YACAC;QACF;QAEA,IAAIe;QACJ,IAAI,QAAA9B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B8B,WAAW,CAAC,6BAA6B,EAAEV,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG5E,MAAMW,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAAS7B;YACX;YACA,gBAAgB,EAAE;YAClB,aAAamB;YACb,MAAM;YACNO;YACA,WAAW,CAAC,CAACf;YACb,OAAOiB;QACT;QAEA,MAAME,WAA0B,EAAE;QACjCZ,CAAAA,YAAY,QAAQ,IAAI,EAAC,EAAG,OAAO,CAAC,CAACa;YACpC,IAAI,QAAQA,MAAM;gBAChB,MAAMC,UAAUZ,YAAYW,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,EAAE;gBAEpC,IAAI,CAACC,SAAS,YACZvB,QAAQ,IAAI,CACV,CAAC,+BAA+B,EAAEsB,KAAK,EAAE,CAAC,0CAA0C,CAAC;gBAIzFD,SAAS,IAAI,CAACE;YAChB;QACF;QAEAC,IAAAA,kCAAAA,eAAAA,AAAAA,EACE;YACE,GAAGJ,QAAQ;YACX,gBAAgBC;QAClB,GACA5B;QAGF,IAAI0B,UACF,MAAM,IAAIM,MAAMN;QAGlB3B,IAAAA,sBAAAA,MAAAA,AAAAA,EACE6B,SAAS,MAAM,IAAI,GACnB,CAAC,0CAA0C,EAAEA,SAAS,MAAM,EAAE;QAGhE,IAAIA,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,SAAS;gBACP,IAAIA,QAAQ,CAAC,EAAE,CAAE,EAAE;gBACnB,SAASA,QAAQ,CAAC,EAAE,CAAE,OAAO;gBAC7B,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM;gBAC3B,MAAMA,QAAQ,CAAC,EAAE,CAAE,IAAI;gBACvB,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM,IAAI,EAAE;gBACjC,YAAYA,QAAQ,CAAC,EAAE,CAAE,UAAU;gBACnCP;YACF;YACAJ;QACF;QAEF,OAAO;YACL,SAAS;YACTA;QACF;IACF;IAEA,MAAM,QACJgB,UAA+B,EAC/BtC,GAA0B,EAC1BuC,gBAAoC,EAKnC;YA0BGtC;QAzBJG,IAAAA,sBAAAA,MAAAA,AAAAA,EACE,AAAsB,YAAtB,OAAOkC,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAEvE,MAAMjC,iBAAiB,IAAI,CAAC,iBAAiB;QAC7C,IAAI,CAAC,iBAAiB,GAAGC;QAEzB,MAAMO,UAAU,MAAM,IAAI,CAAC,kBAAkB,CAAC;QAE9C,MAAMM,YAAYC,KAAK,GAAG;QAC1B,MAAM,EAAEC,WAAW,EAAEI,KAAK,EAAE,GAAG,MAAMe,AAAAA,IAAAA,yBAAAA,oBAAAA,AAAAA,EAAwB;YAC3D3B;YACA,WAAWyB;YACXC;YACA,eAAevC;QACjB;QAEA,MAAM4B,WAAWR,KAAK,GAAG,KAAKD;QAC9B,MAAMU,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACT;QAC9B;QAEA,IAAIU;QACJ,IAAI,QAAA9B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B8B,WAAW,CAAC,qBAAqB,EAAEV,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMW,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTM;YACF;YACA,gBAAgB,EAAE;YAClB,MAAM;YACNT;YACA,OAAOE;QACT;QAEA,MAAM,EAAEU,IAAI,EAAEC,OAAO,EAAE,GAAGrB,eAAe,CAAC;QAG1Ce,IAAAA,kCAAAA,eAAAA,AAAAA,EACE;YACE,GAAGJ,QAAQ;YACXS;QACF,GACApC;QAGF,IAAI0B,YAAY,CAACU,MACf,MAAM,IAAIJ,MAAMN;QAGlB,OAAO;YACLU;YACAC;YACAjB;QACF;IACF;IAEA,MAAM,SACJkB,MAA+B,EAC/B3C,GAEC,EACwD;QACzDI,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOuC,QAAQ;QACf,MAAM9B,UAAU,MAAM,IAAI,CAAC,kBAAkB,CAAC;QAC9C,MAAM,EAAE+B,gBAAgB,EAAEC,IAAI,EAAE,GAAGhC;QACnCT,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOwC,kBAAkB;QAEzB,MAAME,eAAeC,AAAAA,IAAAA,4BAAAA,2BAAAA,AAAAA;QAGrB,MAAMC,kBAAkB;QACxB,MAAMC,aAAmBC,MAAM,OAAO,CAACP,UACnC;YACE,MAAMQ,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC/C,KAAKG,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC9C,OAAOA;YACP,QAAQA;QACV,IACAL;QAEJ,IAAIS,eAAe,MAAMC,AAAAA,IAAAA,oBAAAA,uBAAAA,AAAAA,EAAwB;YAC/C,gBAAgBT;YAChBC;YACA,sBAAsB;gBACpB;oBACE,MAAMI;gBACR;aACD;YACD,iBAAiB;QACnB;QAEA,IAAIjD,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,EAAE;YAClB,MAAMc,aAAawC,AAAAA,IAAAA,0BAAAA,gBAAAA,AAAAA,EAAiBL,YAAYpC,QAAQ,IAAI;YAC5DjB,MAAM,4BAA4BkB;YAClCsC,eAAe,MAAMG,AAAAA,IAAAA,oBAAAA,UAAAA,AAAAA,EACnBH,cACAtC,YACAN,AAAAA,IAAAA,oBAAAA,oBAAAA,AAAAA,EAAqBgD,oBAAAA,oBAAoBA;QAE7C;QAEA,MAAMC,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,WACJ,IAAI,CAAC,UAAU,IAAIC,yBAAAA,mBAAmBA;QAExC,MAAMC,MAAM,MAAMF,SAASD,MAAMI,0BAAAA,YAAAA,CAAAA,gBAA6B;QAE9D,MAAM,EAAEC,OAAO,EAAE,GAAGF;QACpBxD,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,CAAC0D,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1D1D,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO0D,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IAzTA,YACEjD,OAEmE,EACnEb,GAAoB,CACpB;QAfF;QAIA,qCAAoD+D,0BAAAA,QAAQA;QAE5D;QAEA;QAQE3D,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOS,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMmD,QAAQ,OAAO,CAACnD;QAGlD,IAAI,AAA2B,WAApBb,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,UAAU,AAAD,GACvB,IAAI,CAAC,UAAU,GAAGA,IAAI,UAAU;QAElC,IAAI,AAAyB,WAAlBA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,QAAQ,AAAD,GACrB,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAuSF"}
|
|
1
|
+
{"version":3,"file":"insight/index.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/insight/index.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 { callAiFn } from '@/ai-model/common';\nimport { AiExtractElementInfo, AiLocateElement } from '@/ai-model/index';\nimport { AiLocateSection } from '@/ai-model/inspect';\nimport type {\n AIElementResponse,\n AIUsageInfo,\n BaseElement,\n DetailedLocateParam,\n DumpSubscriber,\n InsightAction,\n InsightExtractOption,\n InsightExtractParam,\n InsightOptions,\n InsightTaskInfo,\n LocateResult,\n PartialInsightDumpFromSDK,\n Rect,\n TMultimodalPrompt,\n UIContext,\n} from '@/types';\nimport {\n type IModelPreferences,\n MIDSCENE_FORCE_DEEP_THINK,\n getAIConfigInBoolean,\n vlLocateMode,\n} from '@midscene/shared/env';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n context?: UIContext<BaseElement>;\n callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n ElementType extends BaseElement = BaseElement,\n ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n contextRetrieverFn: (\n action: InsightAction,\n ) => Promise<ContextType> | ContextType;\n\n aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n onceDumpUpdatedFn?: DumpSubscriber;\n\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n constructor(\n context:\n | ContextType\n | ((action: InsightAction) => Promise<ContextType> | ContextType),\n opt?: InsightOptions,\n ) {\n assert(context, 'context is required for Insight');\n if (typeof context === 'function') {\n this.contextRetrieverFn = context;\n } else {\n this.contextRetrieverFn = () => Promise.resolve(context);\n }\n\n if (typeof opt?.aiVendorFn !== 'undefined') {\n this.aiVendorFn = opt.aiVendorFn;\n }\n if (typeof opt?.taskInfo !== 'undefined') {\n this.taskInfo = opt.taskInfo;\n }\n }\n\n async locate(\n query: DetailedLocateParam,\n opt?: LocateOpts,\n ): Promise<LocateResult> {\n const { callAI } = opt || {};\n const queryPrompt = typeof query === 'string' ? query : query.prompt;\n assert(queryPrompt, 'query is required for locate');\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n assert(typeof query === 'object', 'query should be an object for locate');\n\n const globalDeepThinkSwitch = getAIConfigInBoolean(\n MIDSCENE_FORCE_DEEP_THINK,\n );\n if (globalDeepThinkSwitch) {\n debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n }\n let searchAreaPrompt;\n if (query.deepThink || globalDeepThinkSwitch) {\n searchAreaPrompt = query.prompt;\n }\n const modelPreferences: IModelPreferences = {\n intent: 'grounding',\n };\n\n if (searchAreaPrompt && !vlLocateMode(modelPreferences)) {\n console.warn(\n 'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n );\n searchAreaPrompt = undefined;\n }\n\n const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n let searchArea: Rect | undefined = undefined;\n let searchAreaRawResponse: string | undefined = undefined;\n let searchAreaUsage: AIUsageInfo | undefined = undefined;\n let searchAreaResponse:\n | Awaited<ReturnType<typeof AiLocateSection>>\n | undefined = undefined;\n if (searchAreaPrompt) {\n searchAreaResponse = await AiLocateSection({\n context,\n sectionDescription: searchAreaPrompt,\n });\n assert(\n searchAreaResponse.rect,\n `cannot find search area for \"${searchAreaPrompt}\"${\n searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n }`,\n );\n searchAreaRawResponse = searchAreaResponse.rawResponse;\n searchAreaUsage = searchAreaResponse.usage;\n searchArea = searchAreaResponse.rect;\n }\n\n const startTime = Date.now();\n const {\n parseResult,\n rect,\n elementById,\n rawResponse,\n usage,\n isOrderSensitive,\n } = await AiLocateElement({\n callAI: callAI || this.aiVendorFn,\n context,\n targetElementDescription: queryPrompt,\n searchConfig: searchAreaResponse,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(rawResponse),\n formatResponse: JSON.stringify(parseResult),\n usage,\n searchArea,\n searchAreaRawResponse,\n searchAreaUsage,\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'locate',\n userQuery: {\n element: queryPrompt,\n },\n matchedElement: [],\n matchedRect: rect,\n data: null,\n taskInfo,\n deepThink: !!searchArea,\n error: errorLog,\n };\n\n const elements: BaseElement[] = [];\n (parseResult.elements || []).forEach((item) => {\n if ('id' in item) {\n const element = elementById(item?.id);\n\n if (!element) {\n console.warn(\n `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n );\n return;\n }\n elements.push(element);\n }\n });\n\n emitInsightDump(\n {\n ...dumpData,\n matchedElement: elements,\n },\n dumpSubscriber,\n );\n\n if (errorLog) {\n throw new Error(errorLog);\n }\n\n assert(\n elements.length <= 1,\n `locate: multiple elements found, length = ${elements.length}`,\n );\n\n if (elements.length === 1) {\n return {\n element: {\n id: elements[0]!.id,\n indexId: elements[0]!.indexId,\n center: elements[0]!.center,\n rect: elements[0]!.rect,\n xpaths: elements[0]!.xpaths || [],\n attributes: elements[0]!.attributes,\n isOrderSensitive,\n },\n rect,\n };\n }\n return {\n element: null,\n rect,\n };\n }\n\n async extract<T>(\n dataDemand: InsightExtractParam,\n opt?: InsightExtractOption,\n multimodalPrompt?: TMultimodalPrompt,\n ): Promise<{\n data: T;\n thought?: string;\n usage?: AIUsageInfo;\n }> {\n assert(\n typeof dataDemand === 'object' || typeof dataDemand === 'string',\n `dataDemand should be object or string, but get ${typeof dataDemand}`,\n );\n const dumpSubscriber = this.onceDumpUpdatedFn;\n this.onceDumpUpdatedFn = undefined;\n\n const modelPreferences: IModelPreferences = {\n intent: 'VQA',\n };\n\n const context = await this.contextRetrieverFn('extract');\n\n const startTime = Date.now();\n const { parseResult, usage } = await AiExtractElementInfo<T>({\n context,\n dataQuery: dataDemand,\n multimodalPrompt,\n extractOption: opt,\n modelPreferences,\n });\n\n const timeCost = Date.now() - startTime;\n const taskInfo: InsightTaskInfo = {\n ...(this.taskInfo ? this.taskInfo : {}),\n durationMs: timeCost,\n rawResponse: JSON.stringify(parseResult),\n };\n\n let errorLog: string | undefined;\n if (parseResult.errors?.length) {\n errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n }\n\n const dumpData: PartialInsightDumpFromSDK = {\n type: 'extract',\n userQuery: {\n dataDemand,\n },\n matchedElement: [],\n data: null,\n taskInfo,\n error: errorLog,\n };\n\n const { data, thought } = parseResult || {};\n\n // 4\n emitInsightDump(\n {\n ...dumpData,\n data,\n },\n dumpSubscriber,\n );\n\n if (errorLog && !data && !opt?.doNotThrowError) {\n throw new Error(errorLog);\n }\n\n return {\n data,\n thought,\n usage,\n };\n }\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","debug","getDebug","Insight","query","opt","_parseResult_errors","callAI","queryPrompt","assert","dumpSubscriber","undefined","globalDeepThinkSwitch","getAIConfigInBoolean","MIDSCENE_FORCE_DEEP_THINK","searchAreaPrompt","modelPreferences","vlLocateMode","console","context","searchArea","searchAreaRawResponse","searchAreaUsage","searchAreaResponse","AiLocateSection","startTime","Date","parseResult","rect","elementById","rawResponse","usage","isOrderSensitive","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","elements","item","element","emitInsightDump","Error","dataDemand","multimodalPrompt","AiExtractElementInfo","data","thought","callAiFn","Promise"],"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;;;;;;;;;;;;;;;;;;;;;;;;ACiCA,MAAMI,QAAQC,AAAAA,IAAAA,uBAAAA,QAAAA,AAAAA,EAAS;AACR,MAAMC;IAmCnB,MAAM,OACJC,KAA0B,EAC1BC,GAAgB,EACO;YAkFnBC;QAjFJ,MAAM,EAAEC,MAAM,EAAE,GAAGF,OAAO,CAAC;QAC3B,MAAMG,cAAc,AAAiB,YAAjB,OAAOJ,QAAqBA,QAAQA,MAAM,MAAM;QACpEK,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOD,aAAa;QACpB,MAAME,iBAAiB,IAAI,CAAC,iBAAiB;QAC7C,IAAI,CAAC,iBAAiB,GAAGC;QAEzBF,IAAAA,sBAAAA,MAAAA,AAAAA,EAAO,AAAiB,YAAjB,OAAOL,OAAoB;QAElC,MAAMQ,wBAAwBC,AAAAA,IAAAA,oBAAAA,oBAAAA,AAAAA,EAC5BC,oBAAAA,yBAAyBA;QAE3B,IAAIF,uBACFX,MAAM,yBAAyBW;QAEjC,IAAIG;QACJ,IAAIX,MAAM,SAAS,IAAIQ,uBACrBG,mBAAmBX,MAAM,MAAM;QAEjC,MAAMY,mBAAsC;YAC1C,QAAQ;QACV;QAEA,IAAID,oBAAoB,CAACE,AAAAA,IAAAA,oBAAAA,YAAAA,AAAAA,EAAaD,mBAAmB;YACvDE,QAAQ,IAAI,CACV;YAEFH,mBAAmBJ;QACrB;QAEA,MAAMQ,UAAUd,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAM,MAAM,IAAI,CAAC,kBAAkB,CAAC;QAE/D,IAAIe;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAGJ,IAAIR,kBAAkB;YACpBQ,qBAAqB,MAAMC,AAAAA,IAAAA,2BAAAA,eAAAA,AAAAA,EAAgB;gBACzCL;gBACA,oBAAoBJ;YACtB;YACAN,IAAAA,sBAAAA,MAAAA,AAAAA,EACEc,mBAAmB,IAAI,EACvB,CAAC,6BAA6B,EAAER,iBAAiB,CAAC,EAChDQ,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAEJF,wBAAwBE,mBAAmB,WAAW;YACtDD,kBAAkBC,mBAAmB,KAAK;YAC1CH,aAAaG,mBAAmB,IAAI;QACtC;QAEA,MAAME,YAAYC,KAAK,GAAG;QAC1B,MAAM,EACJC,WAAW,EACXC,IAAI,EACJC,WAAW,EACXC,WAAW,EACXC,KAAK,EACLC,gBAAgB,EACjB,GAAG,MAAMC,AAAAA,IAAAA,yBAAAA,eAAAA,AAAAA,EAAgB;YACxB,QAAQ1B,UAAU,IAAI,CAAC,UAAU;YACjCY;YACA,0BAA0BX;YAC1B,cAAce;QAChB;QAEA,MAAMW,WAAWR,KAAK,GAAG,KAAKD;QAC9B,MAAMU,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACN;YAC5B,gBAAgBM,KAAK,SAAS,CAACT;YAC/BI;YACAX;YACAC;YACAC;QACF;QAEA,IAAIe;QACJ,IAAI,QAAA/B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B+B,WAAW,CAAC,6BAA6B,EAAEV,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG5E,MAAMW,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAAS9B;YACX;YACA,gBAAgB,EAAE;YAClB,aAAaoB;YACb,MAAM;YACNO;YACA,WAAW,CAAC,CAACf;YACb,OAAOiB;QACT;QAEA,MAAME,WAA0B,EAAE;QACjCZ,CAAAA,YAAY,QAAQ,IAAI,EAAC,EAAG,OAAO,CAAC,CAACa;YACpC,IAAI,QAAQA,MAAM;gBAChB,MAAMC,UAAUZ,YAAYW,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,EAAE;gBAEpC,IAAI,CAACC,SAAS,YACZvB,QAAQ,IAAI,CACV,CAAC,+BAA+B,EAAEsB,KAAK,EAAE,CAAC,0CAA0C,CAAC;gBAIzFD,SAAS,IAAI,CAACE;YAChB;QACF;QAEAC,IAAAA,kCAAAA,eAAAA,AAAAA,EACE;YACE,GAAGJ,QAAQ;YACX,gBAAgBC;QAClB,GACA7B;QAGF,IAAI2B,UACF,MAAM,IAAIM,MAAMN;QAGlB5B,IAAAA,sBAAAA,MAAAA,AAAAA,EACE8B,SAAS,MAAM,IAAI,GACnB,CAAC,0CAA0C,EAAEA,SAAS,MAAM,EAAE;QAGhE,IAAIA,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,SAAS;gBACP,IAAIA,QAAQ,CAAC,EAAE,CAAE,EAAE;gBACnB,SAASA,QAAQ,CAAC,EAAE,CAAE,OAAO;gBAC7B,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM;gBAC3B,MAAMA,QAAQ,CAAC,EAAE,CAAE,IAAI;gBACvB,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM,IAAI,EAAE;gBACjC,YAAYA,QAAQ,CAAC,EAAE,CAAE,UAAU;gBACnCP;YACF;YACAJ;QACF;QAEF,OAAO;YACL,SAAS;YACTA;QACF;IACF;IAEA,MAAM,QACJgB,UAA+B,EAC/BvC,GAA0B,EAC1BwC,gBAAoC,EAKnC;YA+BGvC;QA9BJG,IAAAA,sBAAAA,MAAAA,AAAAA,EACE,AAAsB,YAAtB,OAAOmC,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAEvE,MAAMlC,iBAAiB,IAAI,CAAC,iBAAiB;QAC7C,IAAI,CAAC,iBAAiB,GAAGC;QAEzB,MAAMK,mBAAsC;YAC1C,QAAQ;QACV;QAEA,MAAMG,UAAU,MAAM,IAAI,CAAC,kBAAkB,CAAC;QAE9C,MAAMM,YAAYC,KAAK,GAAG;QAC1B,MAAM,EAAEC,WAAW,EAAEI,KAAK,EAAE,GAAG,MAAMe,AAAAA,IAAAA,yBAAAA,oBAAAA,AAAAA,EAAwB;YAC3D3B;YACA,WAAWyB;YACXC;YACA,eAAexC;YACfW;QACF;QAEA,MAAMkB,WAAWR,KAAK,GAAG,KAAKD;QAC9B,MAAMU,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACT;QAC9B;QAEA,IAAIU;QACJ,IAAI,QAAA/B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B+B,WAAW,CAAC,qBAAqB,EAAEV,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMW,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTM;YACF;YACA,gBAAgB,EAAE;YAClB,MAAM;YACNT;YACA,OAAOE;QACT;QAEA,MAAM,EAAEU,IAAI,EAAEC,OAAO,EAAE,GAAGrB,eAAe,CAAC;QAG1Ce,IAAAA,kCAAAA,eAAAA,AAAAA,EACE;YACE,GAAGJ,QAAQ;YACXS;QACF,GACArC;QAGF,IAAI2B,YAAY,CAACU,QAAQ,CAAC1C,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,AAAD,GAC3C,MAAM,IAAIsC,MAAMN;QAGlB,OAAO;YACLU;YACAC;YACAjB;QACF;IACF;IAzPA,YACEZ,OAEmE,EACnEd,GAAoB,CACpB;QAfF;QAIA,qCAAoD4C,0BAAAA,QAAQA;QAE5D;QAEA;QAQExC,IAAAA,sBAAAA,MAAAA,AAAAA,EAAOU,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAM+B,QAAQ,OAAO,CAAC/B;QAGlD,IAAI,AAA2B,WAApBd,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,UAAU,AAAD,GACvB,IAAI,CAAC,UAAU,GAAGA,IAAI,UAAU;QAElC,IAAI,AAAyB,WAAlBA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,QAAQ,AAAD,GACrB,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAuOF"}
|