@sqaitech/core 0.30.13 → 0.30.15
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/agent/utils.mjs +1 -1
- package/dist/es/ai-model/service-caller/index.mjs +1 -1
- package/dist/es/ai-model/service-caller/index.mjs.map +1 -1
- package/dist/es/insight/index.mjs +1 -1
- package/dist/es/insight/index.mjs.map +1 -1
- package/dist/es/utils.mjs +3 -3
- package/dist/es/utils.mjs.map +1 -1
- package/dist/lib/agent/utils.js +1 -1
- package/dist/lib/ai-model/service-caller/index.js +1 -1
- package/dist/lib/ai-model/service-caller/index.js.map +1 -1
- package/dist/lib/insight/index.js +1 -1
- package/dist/lib/insight/index.js.map +1 -1
- package/dist/lib/utils.js +3 -3
- package/dist/lib/utils.js.map +1 -1
- package/package.json +3 -3
package/dist/es/agent/utils.mjs
CHANGED
|
@@ -143,7 +143,7 @@ function trimContextByViewport(execution) {
|
|
|
143
143
|
}) : execution.tasks
|
|
144
144
|
};
|
|
145
145
|
}
|
|
146
|
-
const getMidsceneVersion = ()=>"0.30.
|
|
146
|
+
const getMidsceneVersion = ()=>"0.30.14";
|
|
147
147
|
const parsePrompt = (prompt)=>{
|
|
148
148
|
if ('string' == typeof prompt) return {
|
|
149
149
|
textPrompt: prompt,
|
|
@@ -300,7 +300,7 @@ async function callAI(messages, AIActionTypeValue, modelConfig, options) {
|
|
|
300
300
|
};
|
|
301
301
|
} catch (e) {
|
|
302
302
|
console.error(' call AI error', e);
|
|
303
|
-
const newError = new Error(`failed to call ${isStreaming ? 'streaming ' : ''}AI model service: ${e.message}. Trouble shooting: https://
|
|
303
|
+
const newError = new Error(`failed to call ${isStreaming ? 'streaming ' : ''}AI model service: ${e.message}. Trouble shooting: https://sqai.tech/model-provider.html`, {
|
|
304
304
|
cause: e
|
|
305
305
|
});
|
|
306
306
|
throw newError;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"ai-model\\service-caller\\index.mjs","sources":["webpack://@sqaitech/core/./src/ai-model/service-caller/index.ts"],"sourcesContent":["import { AIResponseFormat, type AIUsageInfo } from '@/types';\r\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\r\nimport { Anthropic } from '@anthropic-ai/sdk';\r\nimport {\r\n DefaultAzureCredential,\r\n getBearerTokenProvider,\r\n} from '@azure/identity';\r\nimport {\r\n type IModelConfig,\r\n SQAI_API_TYPE,\r\n SQAI_LANGSMITH_DEBUG,\r\n OPENAI_MAX_TOKENS,\r\n type TVlModeTypes,\r\n type UITarsModelVersion,\r\n globalConfigManager,\r\n} from '@sqaitech/shared/env';\r\n\r\nimport { parseBase64 } from '@sqaitech/shared/img';\r\nimport { getDebug } from '@sqaitech/shared/logger';\r\nimport { assert } from '@sqaitech/shared/utils';\r\nimport { ifInBrowser } from '@sqaitech/shared/utils';\r\nimport { HttpsProxyAgent } from 'https-proxy-agent';\r\nimport { jsonrepair } from 'jsonrepair';\r\nimport OpenAI, { AzureOpenAI } from 'openai';\r\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\r\nimport type { Stream } from 'openai/streaming';\r\nimport { SocksProxyAgent } from 'socks-proxy-agent';\r\nimport { AIActionType, type AIArgs } from '../common';\r\nimport { assertSchema } from '../prompt/assertion';\r\nimport { locatorSchema } from '../prompt/llm-locator';\r\nimport { planSchema } from '../prompt/llm-planning';\r\n\r\nasync function createChatClient({\r\n AIActionTypeValue,\r\n modelConfig,\r\n}: {\r\n AIActionTypeValue: AIActionType;\r\n modelConfig: IModelConfig;\r\n}): Promise<{\r\n completion: OpenAI.Chat.Completions;\r\n style: 'openai' | 'anthropic';\r\n modelName: string;\r\n modelDescription: string;\r\n uiTarsVersion?: UITarsModelVersion;\r\n vlMode: TVlModeTypes | undefined;\r\n}> {\r\n const {\r\n socksProxy,\r\n httpProxy,\r\n modelName,\r\n openaiBaseURL,\r\n openaiApiKey,\r\n openaiExtraConfig,\r\n openaiUseAzureDeprecated,\r\n useAzureOpenai,\r\n azureOpenaiScope,\r\n azureOpenaiKey,\r\n azureOpenaiEndpoint,\r\n azureOpenaiApiVersion,\r\n azureOpenaiDeployment,\r\n azureExtraConfig,\r\n useAnthropicSdk,\r\n anthropicApiKey,\r\n modelDescription,\r\n uiTarsModelVersion: uiTarsVersion,\r\n vlMode,\r\n } = modelConfig;\r\n\r\n let openai: OpenAI | AzureOpenAI | undefined;\r\n\r\n let proxyAgent = undefined;\r\n const debugProxy = getDebug('ai:call:proxy');\r\n if (httpProxy) {\r\n debugProxy('using http proxy', httpProxy);\r\n proxyAgent = new HttpsProxyAgent(httpProxy);\r\n } else if (socksProxy) {\r\n debugProxy('using socks proxy', socksProxy);\r\n proxyAgent = new SocksProxyAgent(socksProxy);\r\n }\r\n\r\n if (openaiUseAzureDeprecated) {\r\n // this is deprecated\r\n openai = new AzureOpenAI({\r\n baseURL: openaiBaseURL,\r\n apiKey: openaiApiKey,\r\n httpAgent: proxyAgent,\r\n ...openaiExtraConfig,\r\n dangerouslyAllowBrowser: true,\r\n }) as OpenAI;\r\n } else if (useAzureOpenai) {\r\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\r\n // keyless authentication\r\n let tokenProvider: any = undefined;\r\n if (azureOpenaiScope) {\r\n assert(\r\n !ifInBrowser,\r\n 'Azure OpenAI is not supported in browser with Midscene.',\r\n );\r\n const credential = new DefaultAzureCredential();\r\n\r\n tokenProvider = getBearerTokenProvider(credential, azureOpenaiScope);\r\n\r\n openai = new AzureOpenAI({\r\n azureADTokenProvider: tokenProvider,\r\n endpoint: azureOpenaiEndpoint,\r\n apiVersion: azureOpenaiApiVersion,\r\n deployment: azureOpenaiDeployment,\r\n ...openaiExtraConfig,\r\n ...azureExtraConfig,\r\n });\r\n } else {\r\n // endpoint, apiKey, apiVersion, deployment\r\n openai = new AzureOpenAI({\r\n apiKey: azureOpenaiKey,\r\n endpoint: azureOpenaiEndpoint,\r\n apiVersion: azureOpenaiApiVersion,\r\n deployment: azureOpenaiDeployment,\r\n dangerouslyAllowBrowser: true,\r\n ...openaiExtraConfig,\r\n ...azureExtraConfig,\r\n });\r\n }\r\n } else if (!useAnthropicSdk) {\r\n openai = new OpenAI({\r\n baseURL: openaiBaseURL,\r\n apiKey: openaiApiKey,\r\n httpAgent: proxyAgent,\r\n ...openaiExtraConfig,\r\n defaultHeaders: {\r\n ...(openaiExtraConfig?.defaultHeaders || {}),\r\n [SQAI_API_TYPE]: AIActionTypeValue.toString(),\r\n },\r\n dangerouslyAllowBrowser: true,\r\n });\r\n }\r\n\r\n if (\r\n openai &&\r\n globalConfigManager.getEnvConfigInBoolean(SQAI_LANGSMITH_DEBUG)\r\n ) {\r\n if (ifInBrowser) {\r\n throw new Error('langsmith is not supported in browser');\r\n }\r\n console.log('DEBUGGING MODE: langsmith wrapper enabled');\r\n const { wrapOpenAI } = await import('langsmith/wrappers');\r\n openai = wrapOpenAI(openai);\r\n }\r\n\r\n if (typeof openai !== 'undefined') {\r\n return {\r\n completion: openai.chat.completions,\r\n style: 'openai',\r\n modelName,\r\n modelDescription,\r\n uiTarsVersion,\r\n vlMode,\r\n };\r\n }\r\n\r\n // Anthropic\r\n if (useAnthropicSdk) {\r\n openai = new Anthropic({\r\n apiKey: anthropicApiKey,\r\n httpAgent: proxyAgent,\r\n dangerouslyAllowBrowser: true,\r\n }) as any;\r\n }\r\n\r\n if (typeof openai !== 'undefined' && (openai as any).messages) {\r\n return {\r\n completion: (openai as any).messages,\r\n style: 'anthropic',\r\n modelName,\r\n modelDescription,\r\n uiTarsVersion,\r\n vlMode,\r\n };\r\n }\r\n\r\n throw new Error('Openai SDK or Anthropic SDK is not initialized');\r\n}\r\n\r\nexport async function callAI(\r\n messages: ChatCompletionMessageParam[],\r\n AIActionTypeValue: AIActionType,\r\n modelConfig: IModelConfig,\r\n options?: {\r\n stream?: boolean;\r\n onChunk?: StreamingCallback;\r\n },\r\n): Promise<{ content: string; usage?: AIUsageInfo; isStreamed: boolean }> {\r\n const {\r\n completion,\r\n style,\r\n modelName,\r\n modelDescription,\r\n uiTarsVersion,\r\n vlMode,\r\n } = await createChatClient({\r\n AIActionTypeValue,\r\n modelConfig,\r\n });\r\n\r\n const responseFormat = getResponseFormat(modelName, AIActionTypeValue);\r\n\r\n const maxTokens = globalConfigManager.getEnvConfigValue(OPENAI_MAX_TOKENS);\r\n const debugCall = getDebug('ai:call');\r\n const debugProfileStats = getDebug('ai:profile:stats');\r\n const debugProfileDetail = getDebug('ai:profile:detail');\r\n\r\n const startTime = Date.now();\r\n\r\n const isStreaming = options?.stream && options?.onChunk;\r\n let content: string | undefined;\r\n let accumulated = '';\r\n let usage: OpenAI.CompletionUsage | undefined;\r\n let timeCost: number | undefined;\r\n\r\n const commonConfig = {\r\n temperature: vlMode === 'vlm-ui-tars' ? 0.0 : 0.1,\r\n stream: !!isStreaming,\r\n max_tokens:\r\n typeof maxTokens === 'number'\r\n ? maxTokens\r\n : Number.parseInt(maxTokens || '2048', 10),\r\n ...(vlMode === 'qwen-vl' || vlMode === 'qwen3-vl' // qwen specific config\r\n ? {\r\n vl_high_resolution_images: true,\r\n }\r\n : {}),\r\n };\r\n\r\n try {\r\n if (style === 'openai') {\r\n debugCall(\r\n `sending ${isStreaming ? 'streaming ' : ''}request to ${modelName}`,\r\n );\r\n\r\n if (isStreaming) {\r\n const stream = (await completion.create(\r\n {\r\n model: modelName,\r\n messages,\r\n response_format: responseFormat,\r\n ...commonConfig,\r\n },\r\n {\r\n stream: true,\r\n },\r\n )) as Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {\r\n _request_id?: string | null;\r\n };\r\n\r\n for await (const chunk of stream) {\r\n const content = chunk.choices?.[0]?.delta?.content || '';\r\n const reasoning_content =\r\n (chunk.choices?.[0]?.delta as any)?.reasoning_content || '';\r\n\r\n // Check for usage info in any chunk (OpenAI provides usage in separate chunks)\r\n if (chunk.usage) {\r\n usage = chunk.usage;\r\n }\r\n\r\n if (content || reasoning_content) {\r\n accumulated += content;\r\n const chunkData: CodeGenerationChunk = {\r\n content,\r\n reasoning_content,\r\n accumulated,\r\n isComplete: false,\r\n usage: undefined,\r\n };\r\n options.onChunk!(chunkData);\r\n }\r\n\r\n // Check if stream is complete\r\n if (chunk.choices?.[0]?.finish_reason) {\r\n timeCost = Date.now() - startTime;\r\n\r\n // If usage is not available from the stream, provide a basic usage info\r\n if (!usage) {\r\n // Estimate token counts based on content length (rough approximation)\r\n const estimatedTokens = Math.max(\r\n 1,\r\n Math.floor(accumulated.length / 4),\r\n );\r\n usage = {\r\n prompt_tokens: estimatedTokens,\r\n completion_tokens: estimatedTokens,\r\n total_tokens: estimatedTokens * 2,\r\n };\r\n }\r\n\r\n // Send final chunk\r\n const finalChunk: CodeGenerationChunk = {\r\n content: '',\r\n accumulated,\r\n reasoning_content: '',\r\n isComplete: true,\r\n usage: {\r\n prompt_tokens: usage.prompt_tokens ?? 0,\r\n completion_tokens: usage.completion_tokens ?? 0,\r\n total_tokens: usage.total_tokens ?? 0,\r\n time_cost: timeCost ?? 0,\r\n model_name: modelName,\r\n model_description: modelDescription,\r\n intent: modelConfig.intent,\r\n },\r\n };\r\n options.onChunk!(finalChunk);\r\n break;\r\n }\r\n }\r\n content = accumulated;\r\n debugProfileStats(\r\n `streaming model, ${modelName}, mode, ${vlMode || 'default'}, cost-ms, ${timeCost}`,\r\n );\r\n } else {\r\n const result = await completion.create({\r\n model: modelName,\r\n messages,\r\n response_format: responseFormat,\r\n ...commonConfig,\r\n } as any);\r\n timeCost = Date.now() - startTime;\r\n\r\n debugProfileStats(\r\n `model, ${modelName}, mode, ${vlMode || 'default'}, ui-tars-version, ${uiTarsVersion}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}`,\r\n );\r\n\r\n debugProfileDetail(\r\n `model usage detail: ${JSON.stringify(result.usage)}`,\r\n );\r\n\r\n assert(\r\n result.choices,\r\n `invalid response from LLM service: ${JSON.stringify(result)}`,\r\n );\r\n content = result.choices[0].message.content!;\r\n usage = result.usage;\r\n }\r\n\r\n debugCall(`response: ${content}`);\r\n assert(content, 'empty content');\r\n } else if (style === 'anthropic') {\r\n const convertImageContent = (content: any) => {\r\n if (content.type === 'image_url') {\r\n const imgBase64 = content.image_url.url;\r\n assert(imgBase64, 'image_url is required');\r\n const { mimeType, body } = parseBase64(content.image_url.url);\r\n return {\r\n source: {\r\n type: 'base64',\r\n media_type: mimeType,\r\n data: body,\r\n },\r\n type: 'image',\r\n };\r\n }\r\n return content;\r\n };\r\n\r\n if (isStreaming) {\r\n const stream = (await completion.create({\r\n model: modelName,\r\n system: 'You are a versatile professional in software UI automation',\r\n messages: messages.map((m) => ({\r\n role: 'user',\r\n content: Array.isArray(m.content)\r\n ? (m.content as any).map(convertImageContent)\r\n : m.content,\r\n })),\r\n response_format: responseFormat,\r\n ...commonConfig,\r\n } as any)) as any;\r\n\r\n for await (const chunk of stream) {\r\n const content = chunk.delta?.text || '';\r\n if (content) {\r\n accumulated += content;\r\n const chunkData: CodeGenerationChunk = {\r\n content,\r\n accumulated,\r\n reasoning_content: '',\r\n isComplete: false,\r\n usage: undefined,\r\n };\r\n options.onChunk!(chunkData);\r\n }\r\n\r\n // Check if stream is complete\r\n if (chunk.type === 'message_stop') {\r\n timeCost = Date.now() - startTime;\r\n const anthropicUsage = chunk.usage;\r\n\r\n // Send final chunk\r\n const finalChunk: CodeGenerationChunk = {\r\n content: '',\r\n accumulated,\r\n reasoning_content: '',\r\n isComplete: true,\r\n usage: anthropicUsage\r\n ? {\r\n prompt_tokens: anthropicUsage.input_tokens ?? 0,\r\n completion_tokens: anthropicUsage.output_tokens ?? 0,\r\n total_tokens:\r\n (anthropicUsage.input_tokens ?? 0) +\r\n (anthropicUsage.output_tokens ?? 0),\r\n time_cost: timeCost ?? 0,\r\n model_name: modelName,\r\n model_description: modelDescription,\r\n intent: modelConfig.intent,\r\n }\r\n : undefined,\r\n };\r\n options.onChunk!(finalChunk);\r\n break;\r\n }\r\n }\r\n content = accumulated;\r\n } else {\r\n const result = await completion.create({\r\n model: modelName,\r\n system: 'You are a versatile professional in software UI automation',\r\n messages: messages.map((m) => ({\r\n role: 'user',\r\n content: Array.isArray(m.content)\r\n ? (m.content as any).map(convertImageContent)\r\n : m.content,\r\n })),\r\n response_format: responseFormat,\r\n ...commonConfig,\r\n } as any);\r\n timeCost = Date.now() - startTime;\r\n content = (result as any).content[0].text as string;\r\n usage = result.usage;\r\n }\r\n\r\n assert(content, 'empty content');\r\n }\r\n // Ensure we always have usage info for streaming responses\r\n if (isStreaming && !usage) {\r\n // Estimate token counts based on content length (rough approximation)\r\n const estimatedTokens = Math.max(\r\n 1,\r\n Math.floor((content || '').length / 4),\r\n );\r\n usage = {\r\n prompt_tokens: estimatedTokens,\r\n completion_tokens: estimatedTokens,\r\n total_tokens: estimatedTokens * 2,\r\n };\r\n }\r\n\r\n return {\r\n content: content || '',\r\n usage: usage\r\n ? {\r\n prompt_tokens: usage.prompt_tokens ?? 0,\r\n completion_tokens: usage.completion_tokens ?? 0,\r\n total_tokens: usage.total_tokens ?? 0,\r\n time_cost: timeCost ?? 0,\r\n model_name: modelName,\r\n model_description: modelDescription,\r\n intent: modelConfig.intent,\r\n }\r\n : undefined,\r\n isStreamed: !!isStreaming,\r\n };\r\n } catch (e: any) {\r\n console.error(' call AI error', e);\r\n const newError = new Error(\r\n `failed to call ${isStreaming ? 'streaming ' : ''}AI model service: ${e.message}. Trouble shooting: https://midscenejs.com/model-provider.html`,\r\n {\r\n cause: e,\r\n },\r\n );\r\n throw newError;\r\n }\r\n}\r\n\r\nexport const getResponseFormat = (\r\n modelName: string,\r\n AIActionTypeValue: AIActionType,\r\n):\r\n | OpenAI.ChatCompletionCreateParams['response_format']\r\n | OpenAI.ResponseFormatJSONObject => {\r\n let responseFormat:\r\n | OpenAI.ChatCompletionCreateParams['response_format']\r\n | OpenAI.ResponseFormatJSONObject\r\n | undefined;\r\n\r\n if (modelName.includes('gpt-4')) {\r\n switch (AIActionTypeValue) {\r\n case AIActionType.ASSERT:\r\n responseFormat = assertSchema;\r\n break;\r\n case AIActionType.INSPECT_ELEMENT:\r\n responseFormat = locatorSchema;\r\n break;\r\n case AIActionType.PLAN:\r\n responseFormat = planSchema;\r\n break;\r\n case AIActionType.EXTRACT_DATA:\r\n case AIActionType.DESCRIBE_ELEMENT:\r\n responseFormat = { type: AIResponseFormat.JSON };\r\n break;\r\n case AIActionType.TEXT:\r\n // No response format for plain text - return as-is\r\n responseFormat = undefined;\r\n break;\r\n }\r\n }\r\n\r\n // gpt-4o-2024-05-13 only supports json_object response format\r\n // Skip for plain text to allow string output\r\n if (\r\n modelName === 'gpt-4o-2024-05-13' &&\r\n AIActionTypeValue !== AIActionType.TEXT\r\n ) {\r\n responseFormat = { type: AIResponseFormat.JSON };\r\n }\r\n\r\n return responseFormat;\r\n};\r\n\r\nexport async function callAIWithObjectResponse<T>(\r\n messages: ChatCompletionMessageParam[],\r\n AIActionTypeValue: AIActionType,\r\n modelConfig: IModelConfig,\r\n): Promise<{ content: T; usage?: AIUsageInfo }> {\r\n const response = await callAI(messages, AIActionTypeValue, modelConfig);\r\n assert(response, 'empty response');\r\n const vlMode = modelConfig.vlMode;\r\n const jsonContent = safeParseJson(response.content, vlMode);\r\n return { content: jsonContent, usage: response.usage };\r\n}\r\n\r\nexport async function callAIWithStringResponse(\r\n msgs: AIArgs,\r\n AIActionTypeValue: AIActionType,\r\n modelConfig: IModelConfig,\r\n): Promise<{ content: string; usage?: AIUsageInfo }> {\r\n const { content, usage } = await callAI(msgs, AIActionTypeValue, modelConfig);\r\n return { content, usage };\r\n}\r\n\r\nexport function extractJSONFromCodeBlock(response: string) {\r\n try {\r\n // First, try to match a JSON object directly in the response\r\n const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\r\n if (jsonMatch) {\r\n return jsonMatch[1];\r\n }\r\n\r\n // If no direct JSON object is found, try to extract JSON from a code block\r\n const codeBlockMatch = response.match(\r\n /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\r\n );\r\n if (codeBlockMatch) {\r\n return codeBlockMatch[1];\r\n }\r\n\r\n // If no code block is found, try to find a JSON-like structure in the text\r\n const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\r\n if (jsonLikeMatch) {\r\n return jsonLikeMatch[0];\r\n }\r\n } catch {}\r\n // If no JSON-like structure is found, return the original response\r\n return response;\r\n}\r\n\r\nexport function preprocessDoubaoBboxJson(input: string) {\r\n if (input.includes('bbox')) {\r\n // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\r\n while (/\\d+\\s+\\d+/.test(input)) {\r\n input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\r\n }\r\n }\r\n return input;\r\n}\r\n\r\nexport function safeParseJson(input: string, vlMode: TVlModeTypes | undefined) {\r\n const cleanJsonString = extractJSONFromCodeBlock(input);\r\n // match the point\r\n if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\r\n return cleanJsonString\r\n .match(/\\((\\d+),(\\d+)\\)/)\r\n ?.slice(1)\r\n .map(Number);\r\n }\r\n try {\r\n return JSON.parse(cleanJsonString);\r\n } catch {}\r\n try {\r\n return JSON.parse(jsonrepair(cleanJsonString));\r\n } catch (e) {}\r\n\r\n if (vlMode === 'doubao-vision' || vlMode === 'vlm-ui-tars') {\r\n const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\r\n return JSON.parse(jsonrepair(jsonString));\r\n }\r\n throw Error(`failed to parse json response: ${input}`);\r\n}\r\n"],"names":["createChatClient","AIActionTypeValue","modelConfig","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","openaiUseAzureDeprecated","useAzureOpenai","azureOpenaiScope","azureOpenaiKey","azureOpenaiEndpoint","azureOpenaiApiVersion","azureOpenaiDeployment","azureExtraConfig","useAnthropicSdk","anthropicApiKey","modelDescription","uiTarsVersion","vlMode","openai","proxyAgent","debugProxy","getDebug","HttpsProxyAgent","SocksProxyAgent","AzureOpenAI","tokenProvider","assert","ifInBrowser","credential","DefaultAzureCredential","getBearerTokenProvider","OpenAI","SQAI_API_TYPE","globalConfigManager","SQAI_LANGSMITH_DEBUG","Error","console","wrapOpenAI","Anthropic","callAI","messages","options","completion","style","responseFormat","getResponseFormat","maxTokens","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","isStreaming","content","accumulated","usage","timeCost","commonConfig","Number","stream","chunk","_chunk_choices__delta","_chunk_choices__delta1","_chunk_choices_2","reasoning_content","chunkData","undefined","estimatedTokens","Math","finalChunk","_result_usage","_result_usage1","_result_usage2","result","JSON","convertImageContent","imgBase64","mimeType","body","parseBase64","m","Array","_chunk_delta","anthropicUsage","e","newError","AIActionType","assertSchema","locatorSchema","planSchema","AIResponseFormat","callAIWithObjectResponse","response","jsonContent","safeParseJson","callAIWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","cleanJsonString","_cleanJsonString_match","jsonrepair","jsonString"],"mappings":";;;;;;;;;;;;;;;AAgCA,eAAeA,iBAAiB,EAC9BC,iBAAiB,EACjBC,WAAW,EAIZ;IAQC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,wBAAwB,EACxBC,cAAc,EACdC,gBAAgB,EAChBC,cAAc,EACdC,mBAAmB,EACnBC,qBAAqB,EACrBC,qBAAqB,EACrBC,gBAAgB,EAChBC,eAAe,EACfC,eAAe,EACfC,gBAAgB,EAChB,oBAAoBC,aAAa,EACjCC,MAAM,EACP,GAAGnB;IAEJ,IAAIoB;IAEJ,IAAIC;IACJ,MAAMC,aAAaC,SAAS;IAC5B,IAAIrB,WAAW;QACboB,WAAW,oBAAoBpB;QAC/BmB,aAAa,IAAIG,gBAAgBtB;IACnC,OAAO,IAAID,YAAY;QACrBqB,WAAW,qBAAqBrB;QAChCoB,aAAa,IAAII,gBAAgBxB;IACnC;IAEA,IAAIM,0BAEFa,SAAS,IAAIM,YAAY;QACvB,SAAStB;QACT,QAAQC;QACR,WAAWgB;QACX,GAAGf,iBAAiB;QACpB,yBAAyB;IAC3B;SACK,IAAIE,gBAAgB;QAGzB,IAAImB;QACJ,IAAIlB,kBAAkB;YACpBmB,OACE,CAACC,aACD;YAEF,MAAMC,aAAa,IAAIC;YAEvBJ,gBAAgBK,uBAAuBF,YAAYrB;YAEnDW,SAAS,IAAIM,YAAY;gBACvB,sBAAsBC;gBACtB,UAAUhB;gBACV,YAAYC;gBACZ,YAAYC;gBACZ,GAAGP,iBAAiB;gBACpB,GAAGQ,gBAAgB;YACrB;QACF,OAEEM,SAAS,IAAIM,YAAY;YACvB,QAAQhB;YACR,UAAUC;YACV,YAAYC;YACZ,YAAYC;YACZ,yBAAyB;YACzB,GAAGP,iBAAiB;YACpB,GAAGQ,gBAAgB;QACrB;IAEJ,OAAO,IAAI,CAACC,iBACVK,SAAS,IAAIa,SAAO;QAClB,SAAS7B;QACT,QAAQC;QACR,WAAWgB;QACX,GAAGf,iBAAiB;QACpB,gBAAgB;YACd,GAAIA,AAAAA,CAAAA,QAAAA,oBAAAA,KAAAA,IAAAA,kBAAmB,cAAc,AAAD,KAAK,CAAC,CAAC;YAC3C,CAAC4B,cAAc,EAAEnC,kBAAkB,QAAQ;QAC7C;QACA,yBAAyB;IAC3B;IAGF,IACEqB,UACAe,oBAAoB,qBAAqB,CAACC,uBAC1C;QACA,IAAIP,aACF,MAAM,IAAIQ,MAAM;QAElBC,QAAQ,GAAG,CAAC;QACZ,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAAC;QACpCnB,SAASmB,WAAWnB;IACtB;IAEA,IAAI,AAAkB,WAAXA,QACT,OAAO;QACL,YAAYA,OAAO,IAAI,CAAC,WAAW;QACnC,OAAO;QACPjB;QACAc;QACAC;QACAC;IACF;IAIF,IAAIJ,iBACFK,SAAS,IAAIoB,UAAU;QACrB,QAAQxB;QACR,WAAWK;QACX,yBAAyB;IAC3B;IAGF,IAAI,AAAkB,WAAXD,UAA2BA,OAAe,QAAQ,EAC3D,OAAO;QACL,YAAaA,OAAe,QAAQ;QACpC,OAAO;QACPjB;QACAc;QACAC;QACAC;IACF;IAGF,MAAM,IAAIkB,MAAM;AAClB;AAEO,eAAeI,OACpBC,QAAsC,EACtC3C,iBAA+B,EAC/BC,WAAyB,EACzB2C,OAGC;IAED,MAAM,EACJC,UAAU,EACVC,KAAK,EACL1C,SAAS,EACTc,gBAAgB,EAChBC,aAAa,EACbC,MAAM,EACP,GAAG,MAAMrB,iBAAiB;QACzBC;QACAC;IACF;IAEA,MAAM8C,iBAAiBC,kBAAkB5C,WAAWJ;IAEpD,MAAMiD,YAAYb,oBAAoB,iBAAiB,CAACc;IACxD,MAAMC,YAAY3B,SAAS;IAC3B,MAAM4B,oBAAoB5B,SAAS;IACnC,MAAM6B,qBAAqB7B,SAAS;IAEpC,MAAM8B,YAAYC,KAAK,GAAG;IAE1B,MAAMC,cAAcZ,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM,AAAD,KAAKA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO,AAAD;IACtD,IAAIa;IACJ,IAAIC,cAAc;IAClB,IAAIC;IACJ,IAAIC;IAEJ,MAAMC,eAAe;QACnB,aAAazC,AAAW,kBAAXA,SAA2B,MAAM;QAC9C,QAAQ,CAAC,CAACoC;QACV,YACE,AAAqB,YAArB,OAAOP,YACHA,YACAa,OAAO,QAAQ,CAACb,aAAa,QAAQ;QAC3C,GAAI7B,AAAW,cAAXA,UAAwBA,AAAW,eAAXA,SACxB;YACE,2BAA2B;QAC7B,IACA,CAAC,CAAC;IACR;IAEA,IAAI;QACF,IAAI0B,AAAU,aAAVA,OAAoB;YACtBK,UACE,CAAC,QAAQ,EAAEK,cAAc,eAAe,GAAG,WAAW,EAAEpD,WAAW;YAGrE,IAAIoD,aAAa;gBACf,MAAMO,SAAU,MAAMlB,WAAW,MAAM,CACrC;oBACE,OAAOzC;oBACPuC;oBACA,iBAAiBI;oBACjB,GAAGc,YAAY;gBACjB,GACA;oBACE,QAAQ;gBACV;gBAKF,WAAW,MAAMG,SAASD,OAAQ;wBAChBE,uBAAAA,iBAAAA,gBAEbC,wBAAAA,kBAAAA,iBAoBCC,kBAAAA;oBAtBJ,MAAMV,UAAUQ,AAAAA,SAAAA,CAAAA,iBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,kBAAAA,cAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,QAAAA,CAAAA,wBAAAA,gBAAoB,KAAK,AAAD,IAAxBA,KAAAA,IAAAA,sBAA2B,OAAO,AAAD,KAAK;oBACtD,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;oBAG3D,IAAIF,MAAM,KAAK,EACbL,QAAQK,MAAM,KAAK;oBAGrB,IAAIP,WAAWW,mBAAmB;wBAChCV,eAAeD;wBACf,MAAMY,YAAiC;4BACrCZ;4BACAW;4BACAV;4BACA,YAAY;4BACZ,OAAOY;wBACT;wBACA1B,QAAQ,OAAO,CAAEyB;oBACnB;oBAGA,IAAI,QAAAF,CAAAA,kBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,mBAAAA,eAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,iBAAoB,aAAa,EAAE;wBACrCP,WAAWL,KAAK,GAAG,KAAKD;wBAGxB,IAAI,CAACK,OAAO;4BAEV,MAAMY,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAACd,YAAY,MAAM,GAAG;4BAElCC,QAAQ;gCACN,eAAeY;gCACf,mBAAmBA;gCACnB,cAAcA,AAAkB,IAAlBA;4BAChB;wBACF;wBAGA,MAAME,aAAkC;4BACtC,SAAS;4BACTf;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAO;gCACL,eAAeC,MAAM,aAAa,IAAI;gCACtC,mBAAmBA,MAAM,iBAAiB,IAAI;gCAC9C,cAAcA,MAAM,YAAY,IAAI;gCACpC,WAAWC,YAAY;gCACvB,YAAYxD;gCACZ,mBAAmBc;gCACnB,QAAQjB,YAAY,MAAM;4BAC5B;wBACF;wBACA2C,QAAQ,OAAO,CAAE6B;wBACjB;oBACF;gBACF;gBACAhB,UAAUC;gBACVN,kBACE,CAAC,iBAAiB,EAAEhD,UAAU,QAAQ,EAAEgB,UAAU,UAAU,WAAW,EAAEwC,UAAU;YAEvF,OAAO;oBAUqGc,eAAyDC,gBAAwDC;gBAT3N,MAAMC,SAAS,MAAMhC,WAAW,MAAM,CAAC;oBACrC,OAAOzC;oBACPuC;oBACA,iBAAiBI;oBACjB,GAAGc,YAAY;gBACjB;gBACAD,WAAWL,KAAK,GAAG,KAAKD;gBAExBF,kBACE,CAAC,OAAO,EAAEhD,UAAU,QAAQ,EAAEgB,UAAU,UAAU,mBAAmB,EAAED,cAAc,iBAAiB,EAAEuD,AAAAA,SAAAA,CAAAA,gBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,cAAc,aAAa,AAAD,KAAK,GAAG,qBAAqB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,iBAAiB,AAAD,KAAK,GAAG,gBAAgB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,YAAY,AAAD,KAAK,GAAG,WAAW,EAAEhB,SAAS,aAAa,EAAEiB,OAAO,WAAW,IAAI,IAAI;gBAG3TxB,mBACE,CAAC,oBAAoB,EAAEyB,KAAK,SAAS,CAACD,OAAO,KAAK,GAAG;gBAGvDhD,OACEgD,OAAO,OAAO,EACd,CAAC,mCAAmC,EAAEC,KAAK,SAAS,CAACD,SAAS;gBAEhEpB,UAAUoB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;gBAC3ClB,QAAQkB,OAAO,KAAK;YACtB;YAEA1B,UAAU,CAAC,UAAU,EAAEM,SAAS;YAChC5B,OAAO4B,SAAS;QAClB,OAAO,IAAIX,AAAU,gBAAVA,OAAuB;YAChC,MAAMiC,sBAAsB,CAACtB;gBAC3B,IAAIA,AAAiB,gBAAjBA,QAAQ,IAAI,EAAkB;oBAChC,MAAMuB,YAAYvB,QAAQ,SAAS,CAAC,GAAG;oBACvC5B,OAAOmD,WAAW;oBAClB,MAAM,EAAEC,QAAQ,EAAEC,IAAI,EAAE,GAAGC,YAAY1B,QAAQ,SAAS,CAAC,GAAG;oBAC5D,OAAO;wBACL,QAAQ;4BACN,MAAM;4BACN,YAAYwB;4BACZ,MAAMC;wBACR;wBACA,MAAM;oBACR;gBACF;gBACA,OAAOzB;YACT;YAEA,IAAID,aAAa;gBACf,MAAMO,SAAU,MAAMlB,WAAW,MAAM,CAAC;oBACtC,OAAOzC;oBACP,QAAQ;oBACR,UAAUuC,SAAS,GAAG,CAAC,CAACyC,IAAO;4BAC7B,MAAM;4BACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;wBACf;oBACA,iBAAiBrC;oBACjB,GAAGc,YAAY;gBACjB;gBAEA,WAAW,MAAMG,SAASD,OAAQ;wBAChBuB;oBAAhB,MAAM7B,UAAU6B,AAAAA,SAAAA,CAAAA,eAAAA,MAAM,KAAK,AAAD,IAAVA,KAAAA,IAAAA,aAAa,IAAI,AAAD,KAAK;oBACrC,IAAI7B,SAAS;wBACXC,eAAeD;wBACf,MAAMY,YAAiC;4BACrCZ;4BACAC;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAOY;wBACT;wBACA1B,QAAQ,OAAO,CAAEyB;oBACnB;oBAGA,IAAIL,AAAe,mBAAfA,MAAM,IAAI,EAAqB;wBACjCJ,WAAWL,KAAK,GAAG,KAAKD;wBACxB,MAAMiC,iBAAiBvB,MAAM,KAAK;wBAGlC,MAAMS,aAAkC;4BACtC,SAAS;4BACTf;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAO6B,iBACH;gCACE,eAAeA,eAAe,YAAY,IAAI;gCAC9C,mBAAmBA,eAAe,aAAa,IAAI;gCACnD,cACGA,AAAAA,CAAAA,eAAe,YAAY,IAAI,KAC/BA,CAAAA,eAAe,aAAa,IAAI;gCACnC,WAAW3B,YAAY;gCACvB,YAAYxD;gCACZ,mBAAmBc;gCACnB,QAAQjB,YAAY,MAAM;4BAC5B,IACAqE;wBACN;wBACA1B,QAAQ,OAAO,CAAE6B;wBACjB;oBACF;gBACF;gBACAhB,UAAUC;YACZ,OAAO;gBACL,MAAMmB,SAAS,MAAMhC,WAAW,MAAM,CAAC;oBACrC,OAAOzC;oBACP,QAAQ;oBACR,UAAUuC,SAAS,GAAG,CAAC,CAACyC,IAAO;4BAC7B,MAAM;4BACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;wBACf;oBACA,iBAAiBrC;oBACjB,GAAGc,YAAY;gBACjB;gBACAD,WAAWL,KAAK,GAAG,KAAKD;gBACxBG,UAAWoB,OAAe,OAAO,CAAC,EAAE,CAAC,IAAI;gBACzClB,QAAQkB,OAAO,KAAK;YACtB;YAEAhD,OAAO4B,SAAS;QAClB;QAEA,IAAID,eAAe,CAACG,OAAO;YAEzB,MAAMY,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEf,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;YAEtCE,QAAQ;gBACN,eAAeY;gBACf,mBAAmBA;gBACnB,cAAcA,AAAkB,IAAlBA;YAChB;QACF;QAEA,OAAO;YACL,SAASd,WAAW;YACpB,OAAOE,QACH;gBACE,eAAeA,MAAM,aAAa,IAAI;gBACtC,mBAAmBA,MAAM,iBAAiB,IAAI;gBAC9C,cAAcA,MAAM,YAAY,IAAI;gBACpC,WAAWC,YAAY;gBACvB,YAAYxD;gBACZ,mBAAmBc;gBACnB,QAAQjB,YAAY,MAAM;YAC5B,IACAqE;YACJ,YAAY,CAAC,CAACd;QAChB;IACF,EAAE,OAAOgC,GAAQ;QACfjD,QAAQ,KAAK,CAAC,kBAAkBiD;QAChC,MAAMC,WAAW,IAAInD,MACnB,CAAC,eAAe,EAAEkB,cAAc,eAAe,GAAG,kBAAkB,EAAEgC,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC/I;YACE,OAAOA;QACT;QAEF,MAAMC;IACR;AACF;AAEO,MAAMzC,oBAAoB,CAC/B5C,WACAJ;IAIA,IAAI+C;IAKJ,IAAI3C,UAAU,QAAQ,CAAC,UACrB,OAAQJ;QACN,KAAK0F,aAAa,MAAM;YACtB3C,iBAAiB4C;YACjB;QACF,KAAKD,aAAa,eAAe;YAC/B3C,iBAAiB6C;YACjB;QACF,KAAKF,aAAa,IAAI;YACpB3C,iBAAiB8C;YACjB;QACF,KAAKH,aAAa,YAAY;QAC9B,KAAKA,aAAa,gBAAgB;YAChC3C,iBAAiB;gBAAE,MAAM+C,iBAAiB,IAAI;YAAC;YAC/C;QACF,KAAKJ,aAAa,IAAI;YAEpB3C,iBAAiBuB;YACjB;IACJ;IAKF,IACElE,AAAc,wBAAdA,aACAJ,sBAAsB0F,aAAa,IAAI,EAEvC3C,iBAAiB;QAAE,MAAM+C,iBAAiB,IAAI;IAAC;IAGjD,OAAO/C;AACT;AAEO,eAAegD,yBACpBpD,QAAsC,EACtC3C,iBAA+B,EAC/BC,WAAyB;IAEzB,MAAM+F,WAAW,MAAMtD,OAAOC,UAAU3C,mBAAmBC;IAC3D4B,OAAOmE,UAAU;IACjB,MAAM5E,SAASnB,YAAY,MAAM;IACjC,MAAMgG,cAAcC,cAAcF,SAAS,OAAO,EAAE5E;IACpD,OAAO;QAAE,SAAS6E;QAAa,OAAOD,SAAS,KAAK;IAAC;AACvD;AAEO,eAAeG,yBACpBC,IAAY,EACZpG,iBAA+B,EAC/BC,WAAyB;IAEzB,MAAM,EAAEwD,OAAO,EAAEE,KAAK,EAAE,GAAG,MAAMjB,OAAO0D,MAAMpG,mBAAmBC;IACjE,OAAO;QAAEwD;QAASE;IAAM;AAC1B;AAEO,SAAS0C,yBAAyBL,QAAgB;IACvD,IAAI;QAEF,MAAMM,YAAYN,SAAS,KAAK,CAAC;QACjC,IAAIM,WACF,OAAOA,SAAS,CAAC,EAAE;QAIrB,MAAMC,iBAAiBP,SAAS,KAAK,CACnC;QAEF,IAAIO,gBACF,OAAOA,cAAc,CAAC,EAAE;QAI1B,MAAMC,gBAAgBR,SAAS,KAAK,CAAC;QACrC,IAAIQ,eACF,OAAOA,aAAa,CAAC,EAAE;IAE3B,EAAE,OAAM,CAAC;IAET,OAAOR;AACT;AAEO,SAASS,yBAAyBC,KAAa;IACpD,IAAIA,MAAM,QAAQ,CAAC,SAEjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;IAG5C,OAAOA;AACT;AAEO,SAASR,cAAcQ,KAAa,EAAEtF,MAAgC;IAC3E,MAAMuF,kBAAkBN,yBAAyBK;IAEjD,IAAIC,QAAAA,kBAAAA,KAAAA,IAAAA,gBAAiB,KAAK,CAAC,oBAAoB;YACtCC;QAAP,OAAO,QAAAA,CAAAA,yBAAAA,gBACJ,KAAK,CAAC,kBAAiB,IADnBA,KAAAA,IAAAA,uBAEH,KAAK,CAAC,GACP,GAAG,CAAC9C;IACT;IACA,IAAI;QACF,OAAOgB,KAAK,KAAK,CAAC6B;IACpB,EAAE,OAAM,CAAC;IACT,IAAI;QACF,OAAO7B,KAAK,KAAK,CAAC+B,WAAWF;IAC/B,EAAE,OAAOnB,GAAG,CAAC;IAEb,IAAIpE,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,QAA0B;QAC1D,MAAM0F,aAAaL,yBAAyBE;QAC5C,OAAO7B,KAAK,KAAK,CAAC+B,WAAWC;IAC/B;IACA,MAAMxE,MAAM,CAAC,+BAA+B,EAAEoE,OAAO;AACvD"}
|
|
1
|
+
{"version":3,"file":"ai-model\\service-caller\\index.mjs","sources":["webpack://@sqaitech/core/./src/ai-model/service-caller/index.ts"],"sourcesContent":["import { AIResponseFormat, type AIUsageInfo } from '@/types';\r\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\r\nimport { Anthropic } from '@anthropic-ai/sdk';\r\nimport {\r\n DefaultAzureCredential,\r\n getBearerTokenProvider,\r\n} from '@azure/identity';\r\nimport {\r\n type IModelConfig,\r\n SQAI_API_TYPE,\r\n SQAI_LANGSMITH_DEBUG,\r\n OPENAI_MAX_TOKENS,\r\n type TVlModeTypes,\r\n type UITarsModelVersion,\r\n globalConfigManager,\r\n} from '@sqaitech/shared/env';\r\n\r\nimport { parseBase64 } from '@sqaitech/shared/img';\r\nimport { getDebug } from '@sqaitech/shared/logger';\r\nimport { assert } from '@sqaitech/shared/utils';\r\nimport { ifInBrowser } from '@sqaitech/shared/utils';\r\nimport { HttpsProxyAgent } from 'https-proxy-agent';\r\nimport { jsonrepair } from 'jsonrepair';\r\nimport OpenAI, { AzureOpenAI } from 'openai';\r\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\r\nimport type { Stream } from 'openai/streaming';\r\nimport { SocksProxyAgent } from 'socks-proxy-agent';\r\nimport { AIActionType, type AIArgs } from '../common';\r\nimport { assertSchema } from '../prompt/assertion';\r\nimport { locatorSchema } from '../prompt/llm-locator';\r\nimport { planSchema } from '../prompt/llm-planning';\r\n\r\nasync function createChatClient({\r\n AIActionTypeValue,\r\n modelConfig,\r\n}: {\r\n AIActionTypeValue: AIActionType;\r\n modelConfig: IModelConfig;\r\n}): Promise<{\r\n completion: OpenAI.Chat.Completions;\r\n style: 'openai' | 'anthropic';\r\n modelName: string;\r\n modelDescription: string;\r\n uiTarsVersion?: UITarsModelVersion;\r\n vlMode: TVlModeTypes | undefined;\r\n}> {\r\n const {\r\n socksProxy,\r\n httpProxy,\r\n modelName,\r\n openaiBaseURL,\r\n openaiApiKey,\r\n openaiExtraConfig,\r\n openaiUseAzureDeprecated,\r\n useAzureOpenai,\r\n azureOpenaiScope,\r\n azureOpenaiKey,\r\n azureOpenaiEndpoint,\r\n azureOpenaiApiVersion,\r\n azureOpenaiDeployment,\r\n azureExtraConfig,\r\n useAnthropicSdk,\r\n anthropicApiKey,\r\n modelDescription,\r\n uiTarsModelVersion: uiTarsVersion,\r\n vlMode,\r\n } = modelConfig;\r\n\r\n let openai: OpenAI | AzureOpenAI | undefined;\r\n\r\n let proxyAgent = undefined;\r\n const debugProxy = getDebug('ai:call:proxy');\r\n if (httpProxy) {\r\n debugProxy('using http proxy', httpProxy);\r\n proxyAgent = new HttpsProxyAgent(httpProxy);\r\n } else if (socksProxy) {\r\n debugProxy('using socks proxy', socksProxy);\r\n proxyAgent = new SocksProxyAgent(socksProxy);\r\n }\r\n\r\n if (openaiUseAzureDeprecated) {\r\n // this is deprecated\r\n openai = new AzureOpenAI({\r\n baseURL: openaiBaseURL,\r\n apiKey: openaiApiKey,\r\n httpAgent: proxyAgent,\r\n ...openaiExtraConfig,\r\n dangerouslyAllowBrowser: true,\r\n }) as OpenAI;\r\n } else if (useAzureOpenai) {\r\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\r\n // keyless authentication\r\n let tokenProvider: any = undefined;\r\n if (azureOpenaiScope) {\r\n assert(\r\n !ifInBrowser,\r\n 'Azure OpenAI is not supported in browser with Midscene.',\r\n );\r\n const credential = new DefaultAzureCredential();\r\n\r\n tokenProvider = getBearerTokenProvider(credential, azureOpenaiScope);\r\n\r\n openai = new AzureOpenAI({\r\n azureADTokenProvider: tokenProvider,\r\n endpoint: azureOpenaiEndpoint,\r\n apiVersion: azureOpenaiApiVersion,\r\n deployment: azureOpenaiDeployment,\r\n ...openaiExtraConfig,\r\n ...azureExtraConfig,\r\n });\r\n } else {\r\n // endpoint, apiKey, apiVersion, deployment\r\n openai = new AzureOpenAI({\r\n apiKey: azureOpenaiKey,\r\n endpoint: azureOpenaiEndpoint,\r\n apiVersion: azureOpenaiApiVersion,\r\n deployment: azureOpenaiDeployment,\r\n dangerouslyAllowBrowser: true,\r\n ...openaiExtraConfig,\r\n ...azureExtraConfig,\r\n });\r\n }\r\n } else if (!useAnthropicSdk) {\r\n openai = new OpenAI({\r\n baseURL: openaiBaseURL,\r\n apiKey: openaiApiKey,\r\n httpAgent: proxyAgent,\r\n ...openaiExtraConfig,\r\n defaultHeaders: {\r\n ...(openaiExtraConfig?.defaultHeaders || {}),\r\n [SQAI_API_TYPE]: AIActionTypeValue.toString(),\r\n },\r\n dangerouslyAllowBrowser: true,\r\n });\r\n }\r\n\r\n if (\r\n openai &&\r\n globalConfigManager.getEnvConfigInBoolean(SQAI_LANGSMITH_DEBUG)\r\n ) {\r\n if (ifInBrowser) {\r\n throw new Error('langsmith is not supported in browser');\r\n }\r\n console.log('DEBUGGING MODE: langsmith wrapper enabled');\r\n const { wrapOpenAI } = await import('langsmith/wrappers');\r\n openai = wrapOpenAI(openai);\r\n }\r\n\r\n if (typeof openai !== 'undefined') {\r\n return {\r\n completion: openai.chat.completions,\r\n style: 'openai',\r\n modelName,\r\n modelDescription,\r\n uiTarsVersion,\r\n vlMode,\r\n };\r\n }\r\n\r\n // Anthropic\r\n if (useAnthropicSdk) {\r\n openai = new Anthropic({\r\n apiKey: anthropicApiKey,\r\n httpAgent: proxyAgent,\r\n dangerouslyAllowBrowser: true,\r\n }) as any;\r\n }\r\n\r\n if (typeof openai !== 'undefined' && (openai as any).messages) {\r\n return {\r\n completion: (openai as any).messages,\r\n style: 'anthropic',\r\n modelName,\r\n modelDescription,\r\n uiTarsVersion,\r\n vlMode,\r\n };\r\n }\r\n\r\n throw new Error('Openai SDK or Anthropic SDK is not initialized');\r\n}\r\n\r\nexport async function callAI(\r\n messages: ChatCompletionMessageParam[],\r\n AIActionTypeValue: AIActionType,\r\n modelConfig: IModelConfig,\r\n options?: {\r\n stream?: boolean;\r\n onChunk?: StreamingCallback;\r\n },\r\n): Promise<{ content: string; usage?: AIUsageInfo; isStreamed: boolean }> {\r\n const {\r\n completion,\r\n style,\r\n modelName,\r\n modelDescription,\r\n uiTarsVersion,\r\n vlMode,\r\n } = await createChatClient({\r\n AIActionTypeValue,\r\n modelConfig,\r\n });\r\n\r\n const responseFormat = getResponseFormat(modelName, AIActionTypeValue);\r\n\r\n const maxTokens = globalConfigManager.getEnvConfigValue(OPENAI_MAX_TOKENS);\r\n const debugCall = getDebug('ai:call');\r\n const debugProfileStats = getDebug('ai:profile:stats');\r\n const debugProfileDetail = getDebug('ai:profile:detail');\r\n\r\n const startTime = Date.now();\r\n\r\n const isStreaming = options?.stream && options?.onChunk;\r\n let content: string | undefined;\r\n let accumulated = '';\r\n let usage: OpenAI.CompletionUsage | undefined;\r\n let timeCost: number | undefined;\r\n\r\n const commonConfig = {\r\n temperature: vlMode === 'vlm-ui-tars' ? 0.0 : 0.1,\r\n stream: !!isStreaming,\r\n max_tokens:\r\n typeof maxTokens === 'number'\r\n ? maxTokens\r\n : Number.parseInt(maxTokens || '2048', 10),\r\n ...(vlMode === 'qwen-vl' || vlMode === 'qwen3-vl' // qwen specific config\r\n ? {\r\n vl_high_resolution_images: true,\r\n }\r\n : {}),\r\n };\r\n\r\n try {\r\n if (style === 'openai') {\r\n debugCall(\r\n `sending ${isStreaming ? 'streaming ' : ''}request to ${modelName}`,\r\n );\r\n\r\n if (isStreaming) {\r\n const stream = (await completion.create(\r\n {\r\n model: modelName,\r\n messages,\r\n response_format: responseFormat,\r\n ...commonConfig,\r\n },\r\n {\r\n stream: true,\r\n },\r\n )) as Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {\r\n _request_id?: string | null;\r\n };\r\n\r\n for await (const chunk of stream) {\r\n const content = chunk.choices?.[0]?.delta?.content || '';\r\n const reasoning_content =\r\n (chunk.choices?.[0]?.delta as any)?.reasoning_content || '';\r\n\r\n // Check for usage info in any chunk (OpenAI provides usage in separate chunks)\r\n if (chunk.usage) {\r\n usage = chunk.usage;\r\n }\r\n\r\n if (content || reasoning_content) {\r\n accumulated += content;\r\n const chunkData: CodeGenerationChunk = {\r\n content,\r\n reasoning_content,\r\n accumulated,\r\n isComplete: false,\r\n usage: undefined,\r\n };\r\n options.onChunk!(chunkData);\r\n }\r\n\r\n // Check if stream is complete\r\n if (chunk.choices?.[0]?.finish_reason) {\r\n timeCost = Date.now() - startTime;\r\n\r\n // If usage is not available from the stream, provide a basic usage info\r\n if (!usage) {\r\n // Estimate token counts based on content length (rough approximation)\r\n const estimatedTokens = Math.max(\r\n 1,\r\n Math.floor(accumulated.length / 4),\r\n );\r\n usage = {\r\n prompt_tokens: estimatedTokens,\r\n completion_tokens: estimatedTokens,\r\n total_tokens: estimatedTokens * 2,\r\n };\r\n }\r\n\r\n // Send final chunk\r\n const finalChunk: CodeGenerationChunk = {\r\n content: '',\r\n accumulated,\r\n reasoning_content: '',\r\n isComplete: true,\r\n usage: {\r\n prompt_tokens: usage.prompt_tokens ?? 0,\r\n completion_tokens: usage.completion_tokens ?? 0,\r\n total_tokens: usage.total_tokens ?? 0,\r\n time_cost: timeCost ?? 0,\r\n model_name: modelName,\r\n model_description: modelDescription,\r\n intent: modelConfig.intent,\r\n },\r\n };\r\n options.onChunk!(finalChunk);\r\n break;\r\n }\r\n }\r\n content = accumulated;\r\n debugProfileStats(\r\n `streaming model, ${modelName}, mode, ${vlMode || 'default'}, cost-ms, ${timeCost}`,\r\n );\r\n } else {\r\n const result = await completion.create({\r\n model: modelName,\r\n messages,\r\n response_format: responseFormat,\r\n ...commonConfig,\r\n } as any);\r\n timeCost = Date.now() - startTime;\r\n\r\n debugProfileStats(\r\n `model, ${modelName}, mode, ${vlMode || 'default'}, ui-tars-version, ${uiTarsVersion}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}`,\r\n );\r\n\r\n debugProfileDetail(\r\n `model usage detail: ${JSON.stringify(result.usage)}`,\r\n );\r\n\r\n assert(\r\n result.choices,\r\n `invalid response from LLM service: ${JSON.stringify(result)}`,\r\n );\r\n content = result.choices[0].message.content!;\r\n usage = result.usage;\r\n }\r\n\r\n debugCall(`response: ${content}`);\r\n assert(content, 'empty content');\r\n } else if (style === 'anthropic') {\r\n const convertImageContent = (content: any) => {\r\n if (content.type === 'image_url') {\r\n const imgBase64 = content.image_url.url;\r\n assert(imgBase64, 'image_url is required');\r\n const { mimeType, body } = parseBase64(content.image_url.url);\r\n return {\r\n source: {\r\n type: 'base64',\r\n media_type: mimeType,\r\n data: body,\r\n },\r\n type: 'image',\r\n };\r\n }\r\n return content;\r\n };\r\n\r\n if (isStreaming) {\r\n const stream = (await completion.create({\r\n model: modelName,\r\n system: 'You are a versatile professional in software UI automation',\r\n messages: messages.map((m) => ({\r\n role: 'user',\r\n content: Array.isArray(m.content)\r\n ? (m.content as any).map(convertImageContent)\r\n : m.content,\r\n })),\r\n response_format: responseFormat,\r\n ...commonConfig,\r\n } as any)) as any;\r\n\r\n for await (const chunk of stream) {\r\n const content = chunk.delta?.text || '';\r\n if (content) {\r\n accumulated += content;\r\n const chunkData: CodeGenerationChunk = {\r\n content,\r\n accumulated,\r\n reasoning_content: '',\r\n isComplete: false,\r\n usage: undefined,\r\n };\r\n options.onChunk!(chunkData);\r\n }\r\n\r\n // Check if stream is complete\r\n if (chunk.type === 'message_stop') {\r\n timeCost = Date.now() - startTime;\r\n const anthropicUsage = chunk.usage;\r\n\r\n // Send final chunk\r\n const finalChunk: CodeGenerationChunk = {\r\n content: '',\r\n accumulated,\r\n reasoning_content: '',\r\n isComplete: true,\r\n usage: anthropicUsage\r\n ? {\r\n prompt_tokens: anthropicUsage.input_tokens ?? 0,\r\n completion_tokens: anthropicUsage.output_tokens ?? 0,\r\n total_tokens:\r\n (anthropicUsage.input_tokens ?? 0) +\r\n (anthropicUsage.output_tokens ?? 0),\r\n time_cost: timeCost ?? 0,\r\n model_name: modelName,\r\n model_description: modelDescription,\r\n intent: modelConfig.intent,\r\n }\r\n : undefined,\r\n };\r\n options.onChunk!(finalChunk);\r\n break;\r\n }\r\n }\r\n content = accumulated;\r\n } else {\r\n const result = await completion.create({\r\n model: modelName,\r\n system: 'You are a versatile professional in software UI automation',\r\n messages: messages.map((m) => ({\r\n role: 'user',\r\n content: Array.isArray(m.content)\r\n ? (m.content as any).map(convertImageContent)\r\n : m.content,\r\n })),\r\n response_format: responseFormat,\r\n ...commonConfig,\r\n } as any);\r\n timeCost = Date.now() - startTime;\r\n content = (result as any).content[0].text as string;\r\n usage = result.usage;\r\n }\r\n\r\n assert(content, 'empty content');\r\n }\r\n // Ensure we always have usage info for streaming responses\r\n if (isStreaming && !usage) {\r\n // Estimate token counts based on content length (rough approximation)\r\n const estimatedTokens = Math.max(\r\n 1,\r\n Math.floor((content || '').length / 4),\r\n );\r\n usage = {\r\n prompt_tokens: estimatedTokens,\r\n completion_tokens: estimatedTokens,\r\n total_tokens: estimatedTokens * 2,\r\n };\r\n }\r\n\r\n return {\r\n content: content || '',\r\n usage: usage\r\n ? {\r\n prompt_tokens: usage.prompt_tokens ?? 0,\r\n completion_tokens: usage.completion_tokens ?? 0,\r\n total_tokens: usage.total_tokens ?? 0,\r\n time_cost: timeCost ?? 0,\r\n model_name: modelName,\r\n model_description: modelDescription,\r\n intent: modelConfig.intent,\r\n }\r\n : undefined,\r\n isStreamed: !!isStreaming,\r\n };\r\n } catch (e: any) {\r\n console.error(' call AI error', e);\r\n const newError = new Error(\r\n `failed to call ${isStreaming ? 'streaming ' : ''}AI model service: ${e.message}. Trouble shooting: https://sqai.tech/model-provider.html`,\r\n {\r\n cause: e,\r\n },\r\n );\r\n throw newError;\r\n }\r\n}\r\n\r\nexport const getResponseFormat = (\r\n modelName: string,\r\n AIActionTypeValue: AIActionType,\r\n):\r\n | OpenAI.ChatCompletionCreateParams['response_format']\r\n | OpenAI.ResponseFormatJSONObject => {\r\n let responseFormat:\r\n | OpenAI.ChatCompletionCreateParams['response_format']\r\n | OpenAI.ResponseFormatJSONObject\r\n | undefined;\r\n\r\n if (modelName.includes('gpt-4')) {\r\n switch (AIActionTypeValue) {\r\n case AIActionType.ASSERT:\r\n responseFormat = assertSchema;\r\n break;\r\n case AIActionType.INSPECT_ELEMENT:\r\n responseFormat = locatorSchema;\r\n break;\r\n case AIActionType.PLAN:\r\n responseFormat = planSchema;\r\n break;\r\n case AIActionType.EXTRACT_DATA:\r\n case AIActionType.DESCRIBE_ELEMENT:\r\n responseFormat = { type: AIResponseFormat.JSON };\r\n break;\r\n case AIActionType.TEXT:\r\n // No response format for plain text - return as-is\r\n responseFormat = undefined;\r\n break;\r\n }\r\n }\r\n\r\n // gpt-4o-2024-05-13 only supports json_object response format\r\n // Skip for plain text to allow string output\r\n if (\r\n modelName === 'gpt-4o-2024-05-13' &&\r\n AIActionTypeValue !== AIActionType.TEXT\r\n ) {\r\n responseFormat = { type: AIResponseFormat.JSON };\r\n }\r\n\r\n return responseFormat;\r\n};\r\n\r\nexport async function callAIWithObjectResponse<T>(\r\n messages: ChatCompletionMessageParam[],\r\n AIActionTypeValue: AIActionType,\r\n modelConfig: IModelConfig,\r\n): Promise<{ content: T; usage?: AIUsageInfo }> {\r\n const response = await callAI(messages, AIActionTypeValue, modelConfig);\r\n assert(response, 'empty response');\r\n const vlMode = modelConfig.vlMode;\r\n const jsonContent = safeParseJson(response.content, vlMode);\r\n return { content: jsonContent, usage: response.usage };\r\n}\r\n\r\nexport async function callAIWithStringResponse(\r\n msgs: AIArgs,\r\n AIActionTypeValue: AIActionType,\r\n modelConfig: IModelConfig,\r\n): Promise<{ content: string; usage?: AIUsageInfo }> {\r\n const { content, usage } = await callAI(msgs, AIActionTypeValue, modelConfig);\r\n return { content, usage };\r\n}\r\n\r\nexport function extractJSONFromCodeBlock(response: string) {\r\n try {\r\n // First, try to match a JSON object directly in the response\r\n const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\r\n if (jsonMatch) {\r\n return jsonMatch[1];\r\n }\r\n\r\n // If no direct JSON object is found, try to extract JSON from a code block\r\n const codeBlockMatch = response.match(\r\n /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\r\n );\r\n if (codeBlockMatch) {\r\n return codeBlockMatch[1];\r\n }\r\n\r\n // If no code block is found, try to find a JSON-like structure in the text\r\n const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\r\n if (jsonLikeMatch) {\r\n return jsonLikeMatch[0];\r\n }\r\n } catch {}\r\n // If no JSON-like structure is found, return the original response\r\n return response;\r\n}\r\n\r\nexport function preprocessDoubaoBboxJson(input: string) {\r\n if (input.includes('bbox')) {\r\n // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\r\n while (/\\d+\\s+\\d+/.test(input)) {\r\n input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\r\n }\r\n }\r\n return input;\r\n}\r\n\r\nexport function safeParseJson(input: string, vlMode: TVlModeTypes | undefined) {\r\n const cleanJsonString = extractJSONFromCodeBlock(input);\r\n // match the point\r\n if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\r\n return cleanJsonString\r\n .match(/\\((\\d+),(\\d+)\\)/)\r\n ?.slice(1)\r\n .map(Number);\r\n }\r\n try {\r\n return JSON.parse(cleanJsonString);\r\n } catch {}\r\n try {\r\n return JSON.parse(jsonrepair(cleanJsonString));\r\n } catch (e) {}\r\n\r\n if (vlMode === 'doubao-vision' || vlMode === 'vlm-ui-tars') {\r\n const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\r\n return JSON.parse(jsonrepair(jsonString));\r\n }\r\n throw Error(`failed to parse json response: ${input}`);\r\n}\r\n"],"names":["createChatClient","AIActionTypeValue","modelConfig","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","openaiUseAzureDeprecated","useAzureOpenai","azureOpenaiScope","azureOpenaiKey","azureOpenaiEndpoint","azureOpenaiApiVersion","azureOpenaiDeployment","azureExtraConfig","useAnthropicSdk","anthropicApiKey","modelDescription","uiTarsVersion","vlMode","openai","proxyAgent","debugProxy","getDebug","HttpsProxyAgent","SocksProxyAgent","AzureOpenAI","tokenProvider","assert","ifInBrowser","credential","DefaultAzureCredential","getBearerTokenProvider","OpenAI","SQAI_API_TYPE","globalConfigManager","SQAI_LANGSMITH_DEBUG","Error","console","wrapOpenAI","Anthropic","callAI","messages","options","completion","style","responseFormat","getResponseFormat","maxTokens","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","isStreaming","content","accumulated","usage","timeCost","commonConfig","Number","stream","chunk","_chunk_choices__delta","_chunk_choices__delta1","_chunk_choices_2","reasoning_content","chunkData","undefined","estimatedTokens","Math","finalChunk","_result_usage","_result_usage1","_result_usage2","result","JSON","convertImageContent","imgBase64","mimeType","body","parseBase64","m","Array","_chunk_delta","anthropicUsage","e","newError","AIActionType","assertSchema","locatorSchema","planSchema","AIResponseFormat","callAIWithObjectResponse","response","jsonContent","safeParseJson","callAIWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","cleanJsonString","_cleanJsonString_match","jsonrepair","jsonString"],"mappings":";;;;;;;;;;;;;;;AAgCA,eAAeA,iBAAiB,EAC9BC,iBAAiB,EACjBC,WAAW,EAIZ;IAQC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,wBAAwB,EACxBC,cAAc,EACdC,gBAAgB,EAChBC,cAAc,EACdC,mBAAmB,EACnBC,qBAAqB,EACrBC,qBAAqB,EACrBC,gBAAgB,EAChBC,eAAe,EACfC,eAAe,EACfC,gBAAgB,EAChB,oBAAoBC,aAAa,EACjCC,MAAM,EACP,GAAGnB;IAEJ,IAAIoB;IAEJ,IAAIC;IACJ,MAAMC,aAAaC,SAAS;IAC5B,IAAIrB,WAAW;QACboB,WAAW,oBAAoBpB;QAC/BmB,aAAa,IAAIG,gBAAgBtB;IACnC,OAAO,IAAID,YAAY;QACrBqB,WAAW,qBAAqBrB;QAChCoB,aAAa,IAAII,gBAAgBxB;IACnC;IAEA,IAAIM,0BAEFa,SAAS,IAAIM,YAAY;QACvB,SAAStB;QACT,QAAQC;QACR,WAAWgB;QACX,GAAGf,iBAAiB;QACpB,yBAAyB;IAC3B;SACK,IAAIE,gBAAgB;QAGzB,IAAImB;QACJ,IAAIlB,kBAAkB;YACpBmB,OACE,CAACC,aACD;YAEF,MAAMC,aAAa,IAAIC;YAEvBJ,gBAAgBK,uBAAuBF,YAAYrB;YAEnDW,SAAS,IAAIM,YAAY;gBACvB,sBAAsBC;gBACtB,UAAUhB;gBACV,YAAYC;gBACZ,YAAYC;gBACZ,GAAGP,iBAAiB;gBACpB,GAAGQ,gBAAgB;YACrB;QACF,OAEEM,SAAS,IAAIM,YAAY;YACvB,QAAQhB;YACR,UAAUC;YACV,YAAYC;YACZ,YAAYC;YACZ,yBAAyB;YACzB,GAAGP,iBAAiB;YACpB,GAAGQ,gBAAgB;QACrB;IAEJ,OAAO,IAAI,CAACC,iBACVK,SAAS,IAAIa,SAAO;QAClB,SAAS7B;QACT,QAAQC;QACR,WAAWgB;QACX,GAAGf,iBAAiB;QACpB,gBAAgB;YACd,GAAIA,AAAAA,CAAAA,QAAAA,oBAAAA,KAAAA,IAAAA,kBAAmB,cAAc,AAAD,KAAK,CAAC,CAAC;YAC3C,CAAC4B,cAAc,EAAEnC,kBAAkB,QAAQ;QAC7C;QACA,yBAAyB;IAC3B;IAGF,IACEqB,UACAe,oBAAoB,qBAAqB,CAACC,uBAC1C;QACA,IAAIP,aACF,MAAM,IAAIQ,MAAM;QAElBC,QAAQ,GAAG,CAAC;QACZ,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAAC;QACpCnB,SAASmB,WAAWnB;IACtB;IAEA,IAAI,AAAkB,WAAXA,QACT,OAAO;QACL,YAAYA,OAAO,IAAI,CAAC,WAAW;QACnC,OAAO;QACPjB;QACAc;QACAC;QACAC;IACF;IAIF,IAAIJ,iBACFK,SAAS,IAAIoB,UAAU;QACrB,QAAQxB;QACR,WAAWK;QACX,yBAAyB;IAC3B;IAGF,IAAI,AAAkB,WAAXD,UAA2BA,OAAe,QAAQ,EAC3D,OAAO;QACL,YAAaA,OAAe,QAAQ;QACpC,OAAO;QACPjB;QACAc;QACAC;QACAC;IACF;IAGF,MAAM,IAAIkB,MAAM;AAClB;AAEO,eAAeI,OACpBC,QAAsC,EACtC3C,iBAA+B,EAC/BC,WAAyB,EACzB2C,OAGC;IAED,MAAM,EACJC,UAAU,EACVC,KAAK,EACL1C,SAAS,EACTc,gBAAgB,EAChBC,aAAa,EACbC,MAAM,EACP,GAAG,MAAMrB,iBAAiB;QACzBC;QACAC;IACF;IAEA,MAAM8C,iBAAiBC,kBAAkB5C,WAAWJ;IAEpD,MAAMiD,YAAYb,oBAAoB,iBAAiB,CAACc;IACxD,MAAMC,YAAY3B,SAAS;IAC3B,MAAM4B,oBAAoB5B,SAAS;IACnC,MAAM6B,qBAAqB7B,SAAS;IAEpC,MAAM8B,YAAYC,KAAK,GAAG;IAE1B,MAAMC,cAAcZ,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM,AAAD,KAAKA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO,AAAD;IACtD,IAAIa;IACJ,IAAIC,cAAc;IAClB,IAAIC;IACJ,IAAIC;IAEJ,MAAMC,eAAe;QACnB,aAAazC,AAAW,kBAAXA,SAA2B,MAAM;QAC9C,QAAQ,CAAC,CAACoC;QACV,YACE,AAAqB,YAArB,OAAOP,YACHA,YACAa,OAAO,QAAQ,CAACb,aAAa,QAAQ;QAC3C,GAAI7B,AAAW,cAAXA,UAAwBA,AAAW,eAAXA,SACxB;YACE,2BAA2B;QAC7B,IACA,CAAC,CAAC;IACR;IAEA,IAAI;QACF,IAAI0B,AAAU,aAAVA,OAAoB;YACtBK,UACE,CAAC,QAAQ,EAAEK,cAAc,eAAe,GAAG,WAAW,EAAEpD,WAAW;YAGrE,IAAIoD,aAAa;gBACf,MAAMO,SAAU,MAAMlB,WAAW,MAAM,CACrC;oBACE,OAAOzC;oBACPuC;oBACA,iBAAiBI;oBACjB,GAAGc,YAAY;gBACjB,GACA;oBACE,QAAQ;gBACV;gBAKF,WAAW,MAAMG,SAASD,OAAQ;wBAChBE,uBAAAA,iBAAAA,gBAEbC,wBAAAA,kBAAAA,iBAoBCC,kBAAAA;oBAtBJ,MAAMV,UAAUQ,AAAAA,SAAAA,CAAAA,iBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,kBAAAA,cAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,QAAAA,CAAAA,wBAAAA,gBAAoB,KAAK,AAAD,IAAxBA,KAAAA,IAAAA,sBAA2B,OAAO,AAAD,KAAK;oBACtD,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;oBAG3D,IAAIF,MAAM,KAAK,EACbL,QAAQK,MAAM,KAAK;oBAGrB,IAAIP,WAAWW,mBAAmB;wBAChCV,eAAeD;wBACf,MAAMY,YAAiC;4BACrCZ;4BACAW;4BACAV;4BACA,YAAY;4BACZ,OAAOY;wBACT;wBACA1B,QAAQ,OAAO,CAAEyB;oBACnB;oBAGA,IAAI,QAAAF,CAAAA,kBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,mBAAAA,eAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,iBAAoB,aAAa,EAAE;wBACrCP,WAAWL,KAAK,GAAG,KAAKD;wBAGxB,IAAI,CAACK,OAAO;4BAEV,MAAMY,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAACd,YAAY,MAAM,GAAG;4BAElCC,QAAQ;gCACN,eAAeY;gCACf,mBAAmBA;gCACnB,cAAcA,AAAkB,IAAlBA;4BAChB;wBACF;wBAGA,MAAME,aAAkC;4BACtC,SAAS;4BACTf;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAO;gCACL,eAAeC,MAAM,aAAa,IAAI;gCACtC,mBAAmBA,MAAM,iBAAiB,IAAI;gCAC9C,cAAcA,MAAM,YAAY,IAAI;gCACpC,WAAWC,YAAY;gCACvB,YAAYxD;gCACZ,mBAAmBc;gCACnB,QAAQjB,YAAY,MAAM;4BAC5B;wBACF;wBACA2C,QAAQ,OAAO,CAAE6B;wBACjB;oBACF;gBACF;gBACAhB,UAAUC;gBACVN,kBACE,CAAC,iBAAiB,EAAEhD,UAAU,QAAQ,EAAEgB,UAAU,UAAU,WAAW,EAAEwC,UAAU;YAEvF,OAAO;oBAUqGc,eAAyDC,gBAAwDC;gBAT3N,MAAMC,SAAS,MAAMhC,WAAW,MAAM,CAAC;oBACrC,OAAOzC;oBACPuC;oBACA,iBAAiBI;oBACjB,GAAGc,YAAY;gBACjB;gBACAD,WAAWL,KAAK,GAAG,KAAKD;gBAExBF,kBACE,CAAC,OAAO,EAAEhD,UAAU,QAAQ,EAAEgB,UAAU,UAAU,mBAAmB,EAAED,cAAc,iBAAiB,EAAEuD,AAAAA,SAAAA,CAAAA,gBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,cAAc,aAAa,AAAD,KAAK,GAAG,qBAAqB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,iBAAiB,AAAD,KAAK,GAAG,gBAAgB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,YAAY,AAAD,KAAK,GAAG,WAAW,EAAEhB,SAAS,aAAa,EAAEiB,OAAO,WAAW,IAAI,IAAI;gBAG3TxB,mBACE,CAAC,oBAAoB,EAAEyB,KAAK,SAAS,CAACD,OAAO,KAAK,GAAG;gBAGvDhD,OACEgD,OAAO,OAAO,EACd,CAAC,mCAAmC,EAAEC,KAAK,SAAS,CAACD,SAAS;gBAEhEpB,UAAUoB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;gBAC3ClB,QAAQkB,OAAO,KAAK;YACtB;YAEA1B,UAAU,CAAC,UAAU,EAAEM,SAAS;YAChC5B,OAAO4B,SAAS;QAClB,OAAO,IAAIX,AAAU,gBAAVA,OAAuB;YAChC,MAAMiC,sBAAsB,CAACtB;gBAC3B,IAAIA,AAAiB,gBAAjBA,QAAQ,IAAI,EAAkB;oBAChC,MAAMuB,YAAYvB,QAAQ,SAAS,CAAC,GAAG;oBACvC5B,OAAOmD,WAAW;oBAClB,MAAM,EAAEC,QAAQ,EAAEC,IAAI,EAAE,GAAGC,YAAY1B,QAAQ,SAAS,CAAC,GAAG;oBAC5D,OAAO;wBACL,QAAQ;4BACN,MAAM;4BACN,YAAYwB;4BACZ,MAAMC;wBACR;wBACA,MAAM;oBACR;gBACF;gBACA,OAAOzB;YACT;YAEA,IAAID,aAAa;gBACf,MAAMO,SAAU,MAAMlB,WAAW,MAAM,CAAC;oBACtC,OAAOzC;oBACP,QAAQ;oBACR,UAAUuC,SAAS,GAAG,CAAC,CAACyC,IAAO;4BAC7B,MAAM;4BACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;wBACf;oBACA,iBAAiBrC;oBACjB,GAAGc,YAAY;gBACjB;gBAEA,WAAW,MAAMG,SAASD,OAAQ;wBAChBuB;oBAAhB,MAAM7B,UAAU6B,AAAAA,SAAAA,CAAAA,eAAAA,MAAM,KAAK,AAAD,IAAVA,KAAAA,IAAAA,aAAa,IAAI,AAAD,KAAK;oBACrC,IAAI7B,SAAS;wBACXC,eAAeD;wBACf,MAAMY,YAAiC;4BACrCZ;4BACAC;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAOY;wBACT;wBACA1B,QAAQ,OAAO,CAAEyB;oBACnB;oBAGA,IAAIL,AAAe,mBAAfA,MAAM,IAAI,EAAqB;wBACjCJ,WAAWL,KAAK,GAAG,KAAKD;wBACxB,MAAMiC,iBAAiBvB,MAAM,KAAK;wBAGlC,MAAMS,aAAkC;4BACtC,SAAS;4BACTf;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAO6B,iBACH;gCACE,eAAeA,eAAe,YAAY,IAAI;gCAC9C,mBAAmBA,eAAe,aAAa,IAAI;gCACnD,cACGA,AAAAA,CAAAA,eAAe,YAAY,IAAI,KAC/BA,CAAAA,eAAe,aAAa,IAAI;gCACnC,WAAW3B,YAAY;gCACvB,YAAYxD;gCACZ,mBAAmBc;gCACnB,QAAQjB,YAAY,MAAM;4BAC5B,IACAqE;wBACN;wBACA1B,QAAQ,OAAO,CAAE6B;wBACjB;oBACF;gBACF;gBACAhB,UAAUC;YACZ,OAAO;gBACL,MAAMmB,SAAS,MAAMhC,WAAW,MAAM,CAAC;oBACrC,OAAOzC;oBACP,QAAQ;oBACR,UAAUuC,SAAS,GAAG,CAAC,CAACyC,IAAO;4BAC7B,MAAM;4BACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;wBACf;oBACA,iBAAiBrC;oBACjB,GAAGc,YAAY;gBACjB;gBACAD,WAAWL,KAAK,GAAG,KAAKD;gBACxBG,UAAWoB,OAAe,OAAO,CAAC,EAAE,CAAC,IAAI;gBACzClB,QAAQkB,OAAO,KAAK;YACtB;YAEAhD,OAAO4B,SAAS;QAClB;QAEA,IAAID,eAAe,CAACG,OAAO;YAEzB,MAAMY,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEf,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;YAEtCE,QAAQ;gBACN,eAAeY;gBACf,mBAAmBA;gBACnB,cAAcA,AAAkB,IAAlBA;YAChB;QACF;QAEA,OAAO;YACL,SAASd,WAAW;YACpB,OAAOE,QACH;gBACE,eAAeA,MAAM,aAAa,IAAI;gBACtC,mBAAmBA,MAAM,iBAAiB,IAAI;gBAC9C,cAAcA,MAAM,YAAY,IAAI;gBACpC,WAAWC,YAAY;gBACvB,YAAYxD;gBACZ,mBAAmBc;gBACnB,QAAQjB,YAAY,MAAM;YAC5B,IACAqE;YACJ,YAAY,CAAC,CAACd;QAChB;IACF,EAAE,OAAOgC,GAAQ;QACfjD,QAAQ,KAAK,CAAC,kBAAkBiD;QAChC,MAAMC,WAAW,IAAInD,MACnB,CAAC,eAAe,EAAEkB,cAAc,eAAe,GAAG,kBAAkB,EAAEgC,EAAE,OAAO,CAAC,yDAAyD,CAAC,EAC1I;YACE,OAAOA;QACT;QAEF,MAAMC;IACR;AACF;AAEO,MAAMzC,oBAAoB,CAC/B5C,WACAJ;IAIA,IAAI+C;IAKJ,IAAI3C,UAAU,QAAQ,CAAC,UACrB,OAAQJ;QACN,KAAK0F,aAAa,MAAM;YACtB3C,iBAAiB4C;YACjB;QACF,KAAKD,aAAa,eAAe;YAC/B3C,iBAAiB6C;YACjB;QACF,KAAKF,aAAa,IAAI;YACpB3C,iBAAiB8C;YACjB;QACF,KAAKH,aAAa,YAAY;QAC9B,KAAKA,aAAa,gBAAgB;YAChC3C,iBAAiB;gBAAE,MAAM+C,iBAAiB,IAAI;YAAC;YAC/C;QACF,KAAKJ,aAAa,IAAI;YAEpB3C,iBAAiBuB;YACjB;IACJ;IAKF,IACElE,AAAc,wBAAdA,aACAJ,sBAAsB0F,aAAa,IAAI,EAEvC3C,iBAAiB;QAAE,MAAM+C,iBAAiB,IAAI;IAAC;IAGjD,OAAO/C;AACT;AAEO,eAAegD,yBACpBpD,QAAsC,EACtC3C,iBAA+B,EAC/BC,WAAyB;IAEzB,MAAM+F,WAAW,MAAMtD,OAAOC,UAAU3C,mBAAmBC;IAC3D4B,OAAOmE,UAAU;IACjB,MAAM5E,SAASnB,YAAY,MAAM;IACjC,MAAMgG,cAAcC,cAAcF,SAAS,OAAO,EAAE5E;IACpD,OAAO;QAAE,SAAS6E;QAAa,OAAOD,SAAS,KAAK;IAAC;AACvD;AAEO,eAAeG,yBACpBC,IAAY,EACZpG,iBAA+B,EAC/BC,WAAyB;IAEzB,MAAM,EAAEwD,OAAO,EAAEE,KAAK,EAAE,GAAG,MAAMjB,OAAO0D,MAAMpG,mBAAmBC;IACjE,OAAO;QAAEwD;QAASE;IAAM;AAC1B;AAEO,SAAS0C,yBAAyBL,QAAgB;IACvD,IAAI;QAEF,MAAMM,YAAYN,SAAS,KAAK,CAAC;QACjC,IAAIM,WACF,OAAOA,SAAS,CAAC,EAAE;QAIrB,MAAMC,iBAAiBP,SAAS,KAAK,CACnC;QAEF,IAAIO,gBACF,OAAOA,cAAc,CAAC,EAAE;QAI1B,MAAMC,gBAAgBR,SAAS,KAAK,CAAC;QACrC,IAAIQ,eACF,OAAOA,aAAa,CAAC,EAAE;IAE3B,EAAE,OAAM,CAAC;IAET,OAAOR;AACT;AAEO,SAASS,yBAAyBC,KAAa;IACpD,IAAIA,MAAM,QAAQ,CAAC,SAEjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;IAG5C,OAAOA;AACT;AAEO,SAASR,cAAcQ,KAAa,EAAEtF,MAAgC;IAC3E,MAAMuF,kBAAkBN,yBAAyBK;IAEjD,IAAIC,QAAAA,kBAAAA,KAAAA,IAAAA,gBAAiB,KAAK,CAAC,oBAAoB;YACtCC;QAAP,OAAO,QAAAA,CAAAA,yBAAAA,gBACJ,KAAK,CAAC,kBAAiB,IADnBA,KAAAA,IAAAA,uBAEH,KAAK,CAAC,GACP,GAAG,CAAC9C;IACT;IACA,IAAI;QACF,OAAOgB,KAAK,KAAK,CAAC6B;IACpB,EAAE,OAAM,CAAC;IACT,IAAI;QACF,OAAO7B,KAAK,KAAK,CAAC+B,WAAWF;IAC/B,EAAE,OAAOnB,GAAG,CAAC;IAEb,IAAIpE,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,QAA0B;QAC1D,MAAM0F,aAAaL,yBAAyBE;QAC5C,OAAO7B,KAAK,KAAK,CAAC+B,WAAWC;IAC/B;IACA,MAAMxE,MAAM,CAAC,+BAA+B,EAAEoE,OAAO;AACvD"}
|
|
@@ -32,7 +32,7 @@ class Insight {
|
|
|
32
32
|
if (query.deepThink || globalDeepThinkSwitch) searchAreaPrompt = query.prompt;
|
|
33
33
|
const { vlMode } = modelConfig;
|
|
34
34
|
if (searchAreaPrompt && !vlMode) {
|
|
35
|
-
console.warn('The "deepThink" feature is not supported with multimodal LLM. Please config VL model for
|
|
35
|
+
console.warn('The "deepThink" feature is not supported with multimodal LLM. Please config VL model for SQAI. https://sqai.tech/choose-a-model');
|
|
36
36
|
searchAreaPrompt = void 0;
|
|
37
37
|
}
|
|
38
38
|
const context = (null == opt ? void 0 : opt.context) || await this.contextRetrieverFn('locate');
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"insight\\index.mjs","sources":["webpack://@sqaitech/core/./src/insight/index.ts"],"sourcesContent":["import { AIActionType, type AIArgs, expandSearchArea } from '@/ai-model/common';\r\nimport {\r\n AiExtractElementInfo,\r\n AiLocateElement,\r\n callAIWithObjectResponse,\r\n} from '@/ai-model/index';\r\nimport { AiLocateSection } from '@/ai-model/inspect';\r\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\r\nimport type {\r\n AIDescribeElementResponse,\r\n AIUsageInfo,\r\n BaseElement,\r\n DetailedLocateParam,\r\n DumpSubscriber,\r\n InsightAction,\r\n InsightExtractOption,\r\n InsightExtractParam,\r\n InsightTaskInfo,\r\n LocateResult,\r\n PartialInsightDumpFromSDK,\r\n Rect,\r\n UIContext,\r\n} from '@/types';\r\nimport {\r\n type IModelConfig,\r\n SQAI_FORCE_DEEP_THINK,\r\n globalConfigManager,\r\n} from '@sqaitech/shared/env';\r\nimport { compositeElementInfoImg, cropByRect } from '@sqaitech/shared/img';\r\nimport { getDebug } from '@sqaitech/shared/logger';\r\nimport { assert } from '@sqaitech/shared/utils';\r\nimport type { TMultimodalPrompt } from '../ai-model/common';\r\nimport { emitInsightDump } from './utils';\r\n\r\nexport interface LocateOpts {\r\n context?: UIContext<BaseElement>;\r\n}\r\n\r\nexport type AnyValue<T> = {\r\n [K in keyof T]: unknown extends T[K] ? any : T[K];\r\n};\r\n\r\ninterface InsightOptions {\r\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\r\n aiVendorFn?: typeof callAIWithObjectResponse;\r\n}\r\n\r\nconst debug = getDebug('ai:insight');\r\nexport default class Insight<\r\n ElementType extends BaseElement = BaseElement,\r\n ContextType extends UIContext<ElementType> = UIContext<ElementType>,\r\n> {\r\n contextRetrieverFn: (\r\n action: InsightAction,\r\n ) => Promise<ContextType> | ContextType;\r\n\r\n aiVendorFn: Exclude<InsightOptions['aiVendorFn'], undefined> =\r\n callAIWithObjectResponse;\r\n\r\n onceDumpUpdatedFn?: DumpSubscriber;\r\n\r\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\r\n\r\n constructor(\r\n context:\r\n | ContextType\r\n | ((action: InsightAction) => Promise<ContextType> | ContextType),\r\n opt?: InsightOptions,\r\n ) {\r\n assert(context, 'context is required for Insight');\r\n if (typeof context === 'function') {\r\n this.contextRetrieverFn = context;\r\n } else {\r\n this.contextRetrieverFn = () => Promise.resolve(context);\r\n }\r\n\r\n // just for unit test, aiVendorFn is callAIWithObjectResponse by default\r\n if (typeof opt?.aiVendorFn !== 'undefined') {\r\n this.aiVendorFn = opt.aiVendorFn;\r\n }\r\n if (typeof opt?.taskInfo !== 'undefined') {\r\n this.taskInfo = opt.taskInfo;\r\n }\r\n }\r\n\r\n async locate(\r\n query: DetailedLocateParam,\r\n opt: LocateOpts,\r\n modelConfig: IModelConfig,\r\n ): Promise<LocateResult> {\r\n const queryPrompt = typeof query === 'string' ? query : query.prompt;\r\n assert(queryPrompt, 'query is required for locate');\r\n const dumpSubscriber = this.onceDumpUpdatedFn;\r\n this.onceDumpUpdatedFn = undefined;\r\n\r\n assert(typeof query === 'object', 'query should be an object for locate');\r\n\r\n const globalDeepThinkSwitch = globalConfigManager.getEnvConfigInBoolean(\r\n SQAI_FORCE_DEEP_THINK,\r\n );\r\n if (globalDeepThinkSwitch) {\r\n debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\r\n }\r\n let searchAreaPrompt;\r\n if (query.deepThink || globalDeepThinkSwitch) {\r\n searchAreaPrompt = query.prompt;\r\n }\r\n\r\n const { vlMode } = modelConfig;\r\n\r\n if (searchAreaPrompt && !vlMode) {\r\n console.warn(\r\n 'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\r\n );\r\n searchAreaPrompt = undefined;\r\n }\r\n\r\n const context = opt?.context || (await this.contextRetrieverFn('locate'));\r\n\r\n let searchArea: Rect | undefined = undefined;\r\n let searchAreaRawResponse: string | undefined = undefined;\r\n let searchAreaUsage: AIUsageInfo | undefined = undefined;\r\n let searchAreaResponse:\r\n | Awaited<ReturnType<typeof AiLocateSection>>\r\n | undefined = undefined;\r\n if (searchAreaPrompt) {\r\n searchAreaResponse = await AiLocateSection({\r\n context,\r\n sectionDescription: searchAreaPrompt,\r\n modelConfig,\r\n });\r\n assert(\r\n searchAreaResponse.rect,\r\n `cannot find search area for \"${searchAreaPrompt}\"${\r\n searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\r\n }`,\r\n );\r\n searchAreaRawResponse = searchAreaResponse.rawResponse;\r\n searchAreaUsage = searchAreaResponse.usage;\r\n searchArea = searchAreaResponse.rect;\r\n }\r\n\r\n const startTime = Date.now();\r\n const {\r\n parseResult,\r\n rect,\r\n elementById,\r\n rawResponse,\r\n usage,\r\n isOrderSensitive,\r\n } = await AiLocateElement({\r\n callAIFn: this.aiVendorFn,\r\n context,\r\n targetElementDescription: queryPrompt,\r\n searchConfig: searchAreaResponse,\r\n modelConfig,\r\n });\r\n\r\n const timeCost = Date.now() - startTime;\r\n const taskInfo: InsightTaskInfo = {\r\n ...(this.taskInfo ? this.taskInfo : {}),\r\n durationMs: timeCost,\r\n rawResponse: JSON.stringify(rawResponse),\r\n formatResponse: JSON.stringify(parseResult),\r\n usage,\r\n searchArea,\r\n searchAreaRawResponse,\r\n searchAreaUsage,\r\n };\r\n\r\n let errorLog: string | undefined;\r\n if (parseResult.errors?.length) {\r\n errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\r\n }\r\n\r\n const dumpData: PartialInsightDumpFromSDK = {\r\n type: 'locate',\r\n userQuery: {\r\n element: queryPrompt,\r\n },\r\n matchedElement: [],\r\n matchedRect: rect,\r\n data: null,\r\n taskInfo,\r\n deepThink: !!searchArea,\r\n error: errorLog,\r\n };\r\n\r\n const elements: BaseElement[] = [];\r\n (parseResult.elements || []).forEach((item) => {\r\n if ('id' in item) {\r\n const element = elementById(item?.id);\r\n\r\n if (!element) {\r\n console.warn(\r\n `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\r\n );\r\n return;\r\n }\r\n elements.push(element);\r\n }\r\n });\r\n\r\n emitInsightDump(\r\n {\r\n ...dumpData,\r\n matchedElement: elements,\r\n },\r\n dumpSubscriber,\r\n );\r\n\r\n if (errorLog) {\r\n throw new Error(errorLog);\r\n }\r\n\r\n assert(\r\n elements.length <= 1,\r\n `locate: multiple elements found, length = ${elements.length}`,\r\n );\r\n\r\n if (elements.length === 1) {\r\n return {\r\n element: {\r\n id: elements[0]!.id,\r\n indexId: elements[0]!.indexId,\r\n center: elements[0]!.center,\r\n rect: elements[0]!.rect,\r\n xpaths: elements[0]!.xpaths || [],\r\n attributes: elements[0]!.attributes,\r\n isOrderSensitive,\r\n },\r\n rect,\r\n };\r\n }\r\n return {\r\n element: null,\r\n rect,\r\n };\r\n }\r\n\r\n async extract<T>(\r\n dataDemand: InsightExtractParam,\r\n modelConfig: IModelConfig,\r\n opt?: InsightExtractOption,\r\n multimodalPrompt?: TMultimodalPrompt,\r\n ): Promise<{\r\n data: T;\r\n thought?: string;\r\n usage?: AIUsageInfo;\r\n }> {\r\n assert(\r\n typeof dataDemand === 'object' || typeof dataDemand === 'string',\r\n `dataDemand should be object or string, but get ${typeof dataDemand}`,\r\n );\r\n const dumpSubscriber = this.onceDumpUpdatedFn;\r\n this.onceDumpUpdatedFn = undefined;\r\n\r\n const context = await this.contextRetrieverFn('extract');\r\n\r\n const startTime = Date.now();\r\n\r\n const { parseResult, usage } = await AiExtractElementInfo<T>({\r\n context,\r\n dataQuery: dataDemand,\r\n multimodalPrompt,\r\n extractOption: opt,\r\n modelConfig,\r\n });\r\n\r\n const timeCost = Date.now() - startTime;\r\n const taskInfo: InsightTaskInfo = {\r\n ...(this.taskInfo ? this.taskInfo : {}),\r\n durationMs: timeCost,\r\n rawResponse: JSON.stringify(parseResult),\r\n };\r\n\r\n let errorLog: string | undefined;\r\n if (parseResult.errors?.length) {\r\n errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\r\n }\r\n\r\n const dumpData: PartialInsightDumpFromSDK = {\r\n type: 'extract',\r\n userQuery: {\r\n dataDemand,\r\n },\r\n matchedElement: [],\r\n data: null,\r\n taskInfo,\r\n error: errorLog,\r\n };\r\n\r\n const { data, thought } = parseResult || {};\r\n\r\n // 4\r\n emitInsightDump(\r\n {\r\n ...dumpData,\r\n data,\r\n },\r\n dumpSubscriber,\r\n );\r\n\r\n if (errorLog && !data && !opt?.doNotThrowError) {\r\n throw new Error(errorLog);\r\n }\r\n\r\n return {\r\n data,\r\n thought,\r\n usage,\r\n };\r\n }\r\n\r\n async describe(\r\n target: Rect | [number, number],\r\n modelConfig: IModelConfig,\r\n opt?: {\r\n deepThink?: boolean;\r\n },\r\n ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\r\n assert(target, 'target is required for insight.describe');\r\n const context = await this.contextRetrieverFn('describe');\r\n const { screenshotBase64, size } = context;\r\n assert(screenshotBase64, 'screenshot is required for insight.describe');\r\n // The result of the \"describe\" function will be used for positioning, so essentially it is a form of grounding.\r\n const { vlMode } = modelConfig;\r\n const systemPrompt = elementDescriberInstruction();\r\n\r\n // Convert [x,y] center point to Rect if needed\r\n const defaultRectSize = 30;\r\n const targetRect: Rect = Array.isArray(target)\r\n ? {\r\n left: Math.floor(target[0] - defaultRectSize / 2),\r\n top: Math.floor(target[1] - defaultRectSize / 2),\r\n width: defaultRectSize,\r\n height: defaultRectSize,\r\n }\r\n : target;\r\n\r\n let imagePayload = await compositeElementInfoImg({\r\n inputImgBase64: screenshotBase64,\r\n size,\r\n elementsPositionInfo: [\r\n {\r\n rect: targetRect,\r\n },\r\n ],\r\n borderThickness: 3,\r\n });\r\n\r\n if (opt?.deepThink) {\r\n const searchArea = expandSearchArea(targetRect, context.size, vlMode);\r\n debug('describe: set searchArea', searchArea);\r\n const croppedResult = await cropByRect(\r\n imagePayload,\r\n searchArea,\r\n vlMode === 'qwen-vl',\r\n );\r\n imagePayload = croppedResult.imageBase64;\r\n }\r\n\r\n const msgs: AIArgs = [\r\n { role: 'system', content: systemPrompt },\r\n {\r\n role: 'user',\r\n content: [\r\n {\r\n type: 'image_url',\r\n image_url: {\r\n url: imagePayload,\r\n detail: 'high',\r\n },\r\n },\r\n ],\r\n },\r\n ];\r\n\r\n const callAIFn = this\r\n .aiVendorFn as typeof callAIWithObjectResponse<AIDescribeElementResponse>;\r\n\r\n const res = await callAIFn(\r\n msgs,\r\n AIActionType.DESCRIBE_ELEMENT,\r\n modelConfig,\r\n );\r\n\r\n const { content } = res;\r\n assert(!content.error, `describe failed: ${content.error}`);\r\n assert(content.description, 'failed to describe the element');\r\n return content;\r\n }\r\n}\r\n"],"names":["debug","getDebug","Insight","query","opt","modelConfig","_parseResult_errors","queryPrompt","assert","dumpSubscriber","undefined","globalDeepThinkSwitch","globalConfigManager","SQAI_FORCE_DEEP_THINK","searchAreaPrompt","vlMode","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","croppedResult","cropByRect","msgs","callAIFn","res","AIActionType","content","callAIWithObjectResponse","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;AA+CA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IAqCnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,WAAyB,EACF;YAkFnBC;QAjFJ,MAAMC,cAAc,AAAiB,YAAjB,OAAOJ,QAAqBA,QAAQA,MAAM,MAAM;QACpEK,OAAOD,aAAa;QACpB,MAAME,iBAAiB,IAAI,CAAC,iBAAiB;QAC7C,IAAI,CAAC,iBAAiB,GAAGC;QAEzBF,OAAO,AAAiB,YAAjB,OAAOL,OAAoB;QAElC,MAAMQ,wBAAwBC,oBAAoB,qBAAqB,CACrEC;QAEF,IAAIF,uBACFX,MAAM,yBAAyBW;QAEjC,IAAIG;QACJ,IAAIX,MAAM,SAAS,IAAIQ,uBACrBG,mBAAmBX,MAAM,MAAM;QAGjC,MAAM,EAAEY,MAAM,EAAE,GAAGV;QAEnB,IAAIS,oBAAoB,CAACC,QAAQ;YAC/BC,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,gBAAgB;gBACzCL;gBACA,oBAAoBH;gBACpBT;YACF;YACAG,OACEa,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,gBAAgB;YACxB,UAAU,IAAI,CAAC,UAAU;YACzBd;YACA,0BAA0BV;YAC1B,cAAcc;YACdhB;QACF;QAEA,MAAM2B,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,QAAA7B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B6B,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,gBACE;YACE,GAAGJ,QAAQ;YACX,gBAAgBC;QAClB,GACA5B;QAGF,IAAI0B,UACF,MAAM,IAAIM,MAAMN;QAGlB3B,OACE6B,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/BrC,WAAyB,EACzBD,GAA0B,EAC1BuC,gBAAoC,EAKnC;YA4BGrC;QA3BJE,OACE,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;QAE1B,MAAM,EAAEC,WAAW,EAAEI,KAAK,EAAE,GAAG,MAAMe,qBAAwB;YAC3D3B;YACA,WAAWyB;YACXC;YACA,eAAevC;YACfC;QACF;QAEA,MAAM2B,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,QAAA7B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B6B,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,gBACE;YACE,GAAGJ,QAAQ;YACXS;QACF,GACApC;QAGF,IAAI0B,YAAY,CAACU,QAAQ,CAACzC,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,AAAD,GAC3C,MAAM,IAAIqC,MAAMN;QAGlB,OAAO;YACLU;YACAC;YACAjB;QACF;IACF;IAEA,MAAM,SACJkB,MAA+B,EAC/B1C,WAAyB,EACzBD,GAEC,EACwD;QACzDI,OAAOuC,QAAQ;QACf,MAAM9B,UAAU,MAAM,IAAI,CAAC,kBAAkB,CAAC;QAC9C,MAAM,EAAE+B,gBAAgB,EAAEC,IAAI,EAAE,GAAGhC;QACnCT,OAAOwC,kBAAkB;QAEzB,MAAM,EAAEjC,MAAM,EAAE,GAAGV;QACnB,MAAM6C,eAAeC;QAGrB,MAAMC,kBAAkB;QACxB,MAAMC,aAAmBC,MAAM,OAAO,CAACP,UACnC;YACE,MAAMQ,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC/C,KAAKG,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC9C,OAAOA;YACP,QAAQA;QACV,IACAL;QAEJ,IAAIS,eAAe,MAAMC,wBAAwB;YAC/C,gBAAgBT;YAChBC;YACA,sBAAsB;gBACpB;oBACE,MAAMI;gBACR;aACD;YACD,iBAAiB;QACnB;QAEA,IAAIjD,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,EAAE;YAClB,MAAMc,aAAawC,iBAAiBL,YAAYpC,QAAQ,IAAI,EAAEF;YAC9Df,MAAM,4BAA4BkB;YAClC,MAAMyC,gBAAgB,MAAMC,WAC1BJ,cACAtC,YACAH,AAAW,cAAXA;YAEFyC,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,WAAW,IAAI,CAClB,UAAU;QAEb,MAAMC,MAAM,MAAMD,SAChBD,MACAG,aAAa,gBAAgB,EAC7B3D;QAGF,MAAM,EAAE4D,OAAO,EAAE,GAAGF;QACpBvD,OAAO,CAACyD,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1DzD,OAAOyD,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IAxUA,YACEhD,OAEmE,EACnEb,GAAoB,CACpB;QAhBF;QAIA,qCACE8D;QAEF;QAEA;QAQE1D,OAAOS,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMkD,QAAQ,OAAO,CAAClD;QAIlD,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;AAqTF"}
|
|
1
|
+
{"version":3,"file":"insight\\index.mjs","sources":["webpack://@sqaitech/core/./src/insight/index.ts"],"sourcesContent":["import { AIActionType, type AIArgs, expandSearchArea } from '@/ai-model/common';\r\nimport {\r\n AiExtractElementInfo,\r\n AiLocateElement,\r\n callAIWithObjectResponse,\r\n} from '@/ai-model/index';\r\nimport { AiLocateSection } from '@/ai-model/inspect';\r\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\r\nimport type {\r\n AIDescribeElementResponse,\r\n AIUsageInfo,\r\n BaseElement,\r\n DetailedLocateParam,\r\n DumpSubscriber,\r\n InsightAction,\r\n InsightExtractOption,\r\n InsightExtractParam,\r\n InsightTaskInfo,\r\n LocateResult,\r\n PartialInsightDumpFromSDK,\r\n Rect,\r\n UIContext,\r\n} from '@/types';\r\nimport {\r\n type IModelConfig,\r\n SQAI_FORCE_DEEP_THINK,\r\n globalConfigManager,\r\n} from '@sqaitech/shared/env';\r\nimport { compositeElementInfoImg, cropByRect } from '@sqaitech/shared/img';\r\nimport { getDebug } from '@sqaitech/shared/logger';\r\nimport { assert } from '@sqaitech/shared/utils';\r\nimport type { TMultimodalPrompt } from '../ai-model/common';\r\nimport { emitInsightDump } from './utils';\r\n\r\nexport interface LocateOpts {\r\n context?: UIContext<BaseElement>;\r\n}\r\n\r\nexport type AnyValue<T> = {\r\n [K in keyof T]: unknown extends T[K] ? any : T[K];\r\n};\r\n\r\ninterface InsightOptions {\r\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\r\n aiVendorFn?: typeof callAIWithObjectResponse;\r\n}\r\n\r\nconst debug = getDebug('ai:insight');\r\nexport default class Insight<\r\n ElementType extends BaseElement = BaseElement,\r\n ContextType extends UIContext<ElementType> = UIContext<ElementType>,\r\n> {\r\n contextRetrieverFn: (\r\n action: InsightAction,\r\n ) => Promise<ContextType> | ContextType;\r\n\r\n aiVendorFn: Exclude<InsightOptions['aiVendorFn'], undefined> =\r\n callAIWithObjectResponse;\r\n\r\n onceDumpUpdatedFn?: DumpSubscriber;\r\n\r\n taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\r\n\r\n constructor(\r\n context:\r\n | ContextType\r\n | ((action: InsightAction) => Promise<ContextType> | ContextType),\r\n opt?: InsightOptions,\r\n ) {\r\n assert(context, 'context is required for Insight');\r\n if (typeof context === 'function') {\r\n this.contextRetrieverFn = context;\r\n } else {\r\n this.contextRetrieverFn = () => Promise.resolve(context);\r\n }\r\n\r\n // just for unit test, aiVendorFn is callAIWithObjectResponse by default\r\n if (typeof opt?.aiVendorFn !== 'undefined') {\r\n this.aiVendorFn = opt.aiVendorFn;\r\n }\r\n if (typeof opt?.taskInfo !== 'undefined') {\r\n this.taskInfo = opt.taskInfo;\r\n }\r\n }\r\n\r\n async locate(\r\n query: DetailedLocateParam,\r\n opt: LocateOpts,\r\n modelConfig: IModelConfig,\r\n ): Promise<LocateResult> {\r\n const queryPrompt = typeof query === 'string' ? query : query.prompt;\r\n assert(queryPrompt, 'query is required for locate');\r\n const dumpSubscriber = this.onceDumpUpdatedFn;\r\n this.onceDumpUpdatedFn = undefined;\r\n\r\n assert(typeof query === 'object', 'query should be an object for locate');\r\n\r\n const globalDeepThinkSwitch = globalConfigManager.getEnvConfigInBoolean(\r\n SQAI_FORCE_DEEP_THINK,\r\n );\r\n if (globalDeepThinkSwitch) {\r\n debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\r\n }\r\n let searchAreaPrompt;\r\n if (query.deepThink || globalDeepThinkSwitch) {\r\n searchAreaPrompt = query.prompt;\r\n }\r\n\r\n const { vlMode } = modelConfig;\r\n\r\n if (searchAreaPrompt && !vlMode) {\r\n console.warn(\r\n 'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for SQAI. https://sqai.tech/choose-a-model',\r\n );\r\n searchAreaPrompt = undefined;\r\n }\r\n\r\n const context = opt?.context || (await this.contextRetrieverFn('locate'));\r\n\r\n let searchArea: Rect | undefined = undefined;\r\n let searchAreaRawResponse: string | undefined = undefined;\r\n let searchAreaUsage: AIUsageInfo | undefined = undefined;\r\n let searchAreaResponse:\r\n | Awaited<ReturnType<typeof AiLocateSection>>\r\n | undefined = undefined;\r\n if (searchAreaPrompt) {\r\n searchAreaResponse = await AiLocateSection({\r\n context,\r\n sectionDescription: searchAreaPrompt,\r\n modelConfig,\r\n });\r\n assert(\r\n searchAreaResponse.rect,\r\n `cannot find search area for \"${searchAreaPrompt}\"${\r\n searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\r\n }`,\r\n );\r\n searchAreaRawResponse = searchAreaResponse.rawResponse;\r\n searchAreaUsage = searchAreaResponse.usage;\r\n searchArea = searchAreaResponse.rect;\r\n }\r\n\r\n const startTime = Date.now();\r\n const {\r\n parseResult,\r\n rect,\r\n elementById,\r\n rawResponse,\r\n usage,\r\n isOrderSensitive,\r\n } = await AiLocateElement({\r\n callAIFn: this.aiVendorFn,\r\n context,\r\n targetElementDescription: queryPrompt,\r\n searchConfig: searchAreaResponse,\r\n modelConfig,\r\n });\r\n\r\n const timeCost = Date.now() - startTime;\r\n const taskInfo: InsightTaskInfo = {\r\n ...(this.taskInfo ? this.taskInfo : {}),\r\n durationMs: timeCost,\r\n rawResponse: JSON.stringify(rawResponse),\r\n formatResponse: JSON.stringify(parseResult),\r\n usage,\r\n searchArea,\r\n searchAreaRawResponse,\r\n searchAreaUsage,\r\n };\r\n\r\n let errorLog: string | undefined;\r\n if (parseResult.errors?.length) {\r\n errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\r\n }\r\n\r\n const dumpData: PartialInsightDumpFromSDK = {\r\n type: 'locate',\r\n userQuery: {\r\n element: queryPrompt,\r\n },\r\n matchedElement: [],\r\n matchedRect: rect,\r\n data: null,\r\n taskInfo,\r\n deepThink: !!searchArea,\r\n error: errorLog,\r\n };\r\n\r\n const elements: BaseElement[] = [];\r\n (parseResult.elements || []).forEach((item) => {\r\n if ('id' in item) {\r\n const element = elementById(item?.id);\r\n\r\n if (!element) {\r\n console.warn(\r\n `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\r\n );\r\n return;\r\n }\r\n elements.push(element);\r\n }\r\n });\r\n\r\n emitInsightDump(\r\n {\r\n ...dumpData,\r\n matchedElement: elements,\r\n },\r\n dumpSubscriber,\r\n );\r\n\r\n if (errorLog) {\r\n throw new Error(errorLog);\r\n }\r\n\r\n assert(\r\n elements.length <= 1,\r\n `locate: multiple elements found, length = ${elements.length}`,\r\n );\r\n\r\n if (elements.length === 1) {\r\n return {\r\n element: {\r\n id: elements[0]!.id,\r\n indexId: elements[0]!.indexId,\r\n center: elements[0]!.center,\r\n rect: elements[0]!.rect,\r\n xpaths: elements[0]!.xpaths || [],\r\n attributes: elements[0]!.attributes,\r\n isOrderSensitive,\r\n },\r\n rect,\r\n };\r\n }\r\n return {\r\n element: null,\r\n rect,\r\n };\r\n }\r\n\r\n async extract<T>(\r\n dataDemand: InsightExtractParam,\r\n modelConfig: IModelConfig,\r\n opt?: InsightExtractOption,\r\n multimodalPrompt?: TMultimodalPrompt,\r\n ): Promise<{\r\n data: T;\r\n thought?: string;\r\n usage?: AIUsageInfo;\r\n }> {\r\n assert(\r\n typeof dataDemand === 'object' || typeof dataDemand === 'string',\r\n `dataDemand should be object or string, but get ${typeof dataDemand}`,\r\n );\r\n const dumpSubscriber = this.onceDumpUpdatedFn;\r\n this.onceDumpUpdatedFn = undefined;\r\n\r\n const context = await this.contextRetrieverFn('extract');\r\n\r\n const startTime = Date.now();\r\n\r\n const { parseResult, usage } = await AiExtractElementInfo<T>({\r\n context,\r\n dataQuery: dataDemand,\r\n multimodalPrompt,\r\n extractOption: opt,\r\n modelConfig,\r\n });\r\n\r\n const timeCost = Date.now() - startTime;\r\n const taskInfo: InsightTaskInfo = {\r\n ...(this.taskInfo ? this.taskInfo : {}),\r\n durationMs: timeCost,\r\n rawResponse: JSON.stringify(parseResult),\r\n };\r\n\r\n let errorLog: string | undefined;\r\n if (parseResult.errors?.length) {\r\n errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\r\n }\r\n\r\n const dumpData: PartialInsightDumpFromSDK = {\r\n type: 'extract',\r\n userQuery: {\r\n dataDemand,\r\n },\r\n matchedElement: [],\r\n data: null,\r\n taskInfo,\r\n error: errorLog,\r\n };\r\n\r\n const { data, thought } = parseResult || {};\r\n\r\n // 4\r\n emitInsightDump(\r\n {\r\n ...dumpData,\r\n data,\r\n },\r\n dumpSubscriber,\r\n );\r\n\r\n if (errorLog && !data && !opt?.doNotThrowError) {\r\n throw new Error(errorLog);\r\n }\r\n\r\n return {\r\n data,\r\n thought,\r\n usage,\r\n };\r\n }\r\n\r\n async describe(\r\n target: Rect | [number, number],\r\n modelConfig: IModelConfig,\r\n opt?: {\r\n deepThink?: boolean;\r\n },\r\n ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\r\n assert(target, 'target is required for insight.describe');\r\n const context = await this.contextRetrieverFn('describe');\r\n const { screenshotBase64, size } = context;\r\n assert(screenshotBase64, 'screenshot is required for insight.describe');\r\n // The result of the \"describe\" function will be used for positioning, so essentially it is a form of grounding.\r\n const { vlMode } = modelConfig;\r\n const systemPrompt = elementDescriberInstruction();\r\n\r\n // Convert [x,y] center point to Rect if needed\r\n const defaultRectSize = 30;\r\n const targetRect: Rect = Array.isArray(target)\r\n ? {\r\n left: Math.floor(target[0] - defaultRectSize / 2),\r\n top: Math.floor(target[1] - defaultRectSize / 2),\r\n width: defaultRectSize,\r\n height: defaultRectSize,\r\n }\r\n : target;\r\n\r\n let imagePayload = await compositeElementInfoImg({\r\n inputImgBase64: screenshotBase64,\r\n size,\r\n elementsPositionInfo: [\r\n {\r\n rect: targetRect,\r\n },\r\n ],\r\n borderThickness: 3,\r\n });\r\n\r\n if (opt?.deepThink) {\r\n const searchArea = expandSearchArea(targetRect, context.size, vlMode);\r\n debug('describe: set searchArea', searchArea);\r\n const croppedResult = await cropByRect(\r\n imagePayload,\r\n searchArea,\r\n vlMode === 'qwen-vl',\r\n );\r\n imagePayload = croppedResult.imageBase64;\r\n }\r\n\r\n const msgs: AIArgs = [\r\n { role: 'system', content: systemPrompt },\r\n {\r\n role: 'user',\r\n content: [\r\n {\r\n type: 'image_url',\r\n image_url: {\r\n url: imagePayload,\r\n detail: 'high',\r\n },\r\n },\r\n ],\r\n },\r\n ];\r\n\r\n const callAIFn = this\r\n .aiVendorFn as typeof callAIWithObjectResponse<AIDescribeElementResponse>;\r\n\r\n const res = await callAIFn(\r\n msgs,\r\n AIActionType.DESCRIBE_ELEMENT,\r\n modelConfig,\r\n );\r\n\r\n const { content } = res;\r\n assert(!content.error, `describe failed: ${content.error}`);\r\n assert(content.description, 'failed to describe the element');\r\n return content;\r\n }\r\n}\r\n"],"names":["debug","getDebug","Insight","query","opt","modelConfig","_parseResult_errors","queryPrompt","assert","dumpSubscriber","undefined","globalDeepThinkSwitch","globalConfigManager","SQAI_FORCE_DEEP_THINK","searchAreaPrompt","vlMode","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","croppedResult","cropByRect","msgs","callAIFn","res","AIActionType","content","callAIWithObjectResponse","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;AA+CA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IAqCnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,WAAyB,EACF;YAkFnBC;QAjFJ,MAAMC,cAAc,AAAiB,YAAjB,OAAOJ,QAAqBA,QAAQA,MAAM,MAAM;QACpEK,OAAOD,aAAa;QACpB,MAAME,iBAAiB,IAAI,CAAC,iBAAiB;QAC7C,IAAI,CAAC,iBAAiB,GAAGC;QAEzBF,OAAO,AAAiB,YAAjB,OAAOL,OAAoB;QAElC,MAAMQ,wBAAwBC,oBAAoB,qBAAqB,CACrEC;QAEF,IAAIF,uBACFX,MAAM,yBAAyBW;QAEjC,IAAIG;QACJ,IAAIX,MAAM,SAAS,IAAIQ,uBACrBG,mBAAmBX,MAAM,MAAM;QAGjC,MAAM,EAAEY,MAAM,EAAE,GAAGV;QAEnB,IAAIS,oBAAoB,CAACC,QAAQ;YAC/BC,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,gBAAgB;gBACzCL;gBACA,oBAAoBH;gBACpBT;YACF;YACAG,OACEa,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,gBAAgB;YACxB,UAAU,IAAI,CAAC,UAAU;YACzBd;YACA,0BAA0BV;YAC1B,cAAcc;YACdhB;QACF;QAEA,MAAM2B,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,QAAA7B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B6B,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,gBACE;YACE,GAAGJ,QAAQ;YACX,gBAAgBC;QAClB,GACA5B;QAGF,IAAI0B,UACF,MAAM,IAAIM,MAAMN;QAGlB3B,OACE6B,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/BrC,WAAyB,EACzBD,GAA0B,EAC1BuC,gBAAoC,EAKnC;YA4BGrC;QA3BJE,OACE,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;QAE1B,MAAM,EAAEC,WAAW,EAAEI,KAAK,EAAE,GAAG,MAAMe,qBAAwB;YAC3D3B;YACA,WAAWyB;YACXC;YACA,eAAevC;YACfC;QACF;QAEA,MAAM2B,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,QAAA7B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B6B,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,gBACE;YACE,GAAGJ,QAAQ;YACXS;QACF,GACApC;QAGF,IAAI0B,YAAY,CAACU,QAAQ,CAACzC,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,AAAD,GAC3C,MAAM,IAAIqC,MAAMN;QAGlB,OAAO;YACLU;YACAC;YACAjB;QACF;IACF;IAEA,MAAM,SACJkB,MAA+B,EAC/B1C,WAAyB,EACzBD,GAEC,EACwD;QACzDI,OAAOuC,QAAQ;QACf,MAAM9B,UAAU,MAAM,IAAI,CAAC,kBAAkB,CAAC;QAC9C,MAAM,EAAE+B,gBAAgB,EAAEC,IAAI,EAAE,GAAGhC;QACnCT,OAAOwC,kBAAkB;QAEzB,MAAM,EAAEjC,MAAM,EAAE,GAAGV;QACnB,MAAM6C,eAAeC;QAGrB,MAAMC,kBAAkB;QACxB,MAAMC,aAAmBC,MAAM,OAAO,CAACP,UACnC;YACE,MAAMQ,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC/C,KAAKG,KAAK,KAAK,CAACR,MAAM,CAAC,EAAE,GAAGK,kBAAkB;YAC9C,OAAOA;YACP,QAAQA;QACV,IACAL;QAEJ,IAAIS,eAAe,MAAMC,wBAAwB;YAC/C,gBAAgBT;YAChBC;YACA,sBAAsB;gBACpB;oBACE,MAAMI;gBACR;aACD;YACD,iBAAiB;QACnB;QAEA,IAAIjD,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,EAAE;YAClB,MAAMc,aAAawC,iBAAiBL,YAAYpC,QAAQ,IAAI,EAAEF;YAC9Df,MAAM,4BAA4BkB;YAClC,MAAMyC,gBAAgB,MAAMC,WAC1BJ,cACAtC,YACAH,AAAW,cAAXA;YAEFyC,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,WAAW,IAAI,CAClB,UAAU;QAEb,MAAMC,MAAM,MAAMD,SAChBD,MACAG,aAAa,gBAAgB,EAC7B3D;QAGF,MAAM,EAAE4D,OAAO,EAAE,GAAGF;QACpBvD,OAAO,CAACyD,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1DzD,OAAOyD,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IAxUA,YACEhD,OAEmE,EACnEb,GAAoB,CACpB;QAhBF;QAIA,qCACE8D;QAEF;QAEA;QAQE1D,OAAOS,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMkD,QAAQ,OAAO,CAAClD;QAIlD,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;AAqTF"}
|