@midscene/core 0.26.7-beta-20250821134240.0 → 0.27.0

package/dist/es/ai-model/service-caller/index.mjs

@@ -14,7 +14,7 @@ import { assertSchema } from "../prompt/assertion.mjs";
 import { locatorSchema } from "../prompt/llm-locator.mjs";
 import { planSchema } from "../prompt/llm-planning.mjs";
 async function createChatClient({ AIActionTypeValue, modelPreferences }) {
-    const { socksProxy, httpProxy, modelName, openaiBaseURL, openaiApiKey, openaiExtraConfig, openaiUseAzureDeprecated, useAzureOpenai, azureOpenaiScope, azureOpenaiKey, azureOpenaiEndpoint, azureOpenaiApiVersion, azureOpenaiDeployment, azureExtraConfig, useAnthropicSdk, anthropicApiKey } = decideModelConfig(modelPreferences, true);
+    const { socksProxy, httpProxy, modelName, openaiBaseURL, openaiApiKey, openaiExtraConfig, openaiUseAzureDeprecated, useAzureOpenai, azureOpenaiScope, azureOpenaiKey, azureOpenaiEndpoint, azureOpenaiApiVersion, azureOpenaiDeployment, azureExtraConfig, useAnthropicSdk, anthropicApiKey, modelDescription } = decideModelConfig(modelPreferences, true);
     let openai;
     let proxyAgent;
     const debugProxy = getDebug('ai:call:proxy');
@@ -75,7 +75,8 @@ async function createChatClient({ AIActionTypeValue, modelPreferences }) {
     if (void 0 !== openai) return {
         completion: openai.chat.completions,
         style: 'openai',
-        modelName
+        modelName,
+        modelDescription
     };
     if (useAnthropicSdk) openai = new Anthropic({
         apiKey: anthropicApiKey,
@@ -85,12 +86,13 @@ async function createChatClient({ AIActionTypeValue, modelPreferences }) {
     if (void 0 !== openai && openai.messages) return {
         completion: openai.messages,
         style: 'anthropic',
-        modelName
+        modelName,
+        modelDescription
     };
     throw new Error('Openai SDK or Anthropic SDK is not initialized');
 }
 async function call(messages, AIActionTypeValue, modelPreferences, options) {
-    const { completion, style, modelName } = await createChatClient({
+    const { completion, style, modelName, modelDescription } = await createChatClient({
         AIActionTypeValue,
         modelPreferences
     });
@@ -161,7 +163,9 @@ async function call(messages, AIActionTypeValue, modelPreferences, options) {
                                 completion_tokens: usage.completion_tokens ?? 0,
                                 total_tokens: usage.total_tokens ?? 0,
                                 time_cost: timeCost ?? 0,
-                                model_name: modelName
+                                model_name: modelName,
+                                model_description: modelDescription,
+                                intent: modelPreferences.intent
                             }
                         };
                         options.onChunk(finalChunk);
@@ -242,7 +246,9 @@ async function call(messages, AIActionTypeValue, modelPreferences, options) {
                                 completion_tokens: anthropicUsage.output_tokens ?? 0,
                                 total_tokens: (anthropicUsage.input_tokens ?? 0) + (anthropicUsage.output_tokens ?? 0),
                                 time_cost: timeCost ?? 0,
-                                model_name: modelName
+                                model_name: modelName,
+                                model_description: modelDescription,
+                                intent: modelPreferences.intent
                             } : void 0
                         };
                         options.onChunk(finalChunk);
@@ -282,7 +288,9 @@ async function call(messages, AIActionTypeValue, modelPreferences, options) {
                 completion_tokens: usage.completion_tokens ?? 0,
                 total_tokens: usage.total_tokens ?? 0,
                 time_cost: timeCost ?? 0,
-                model_name: modelName
+                model_name: modelName,
+                model_description: modelDescription,
+                intent: modelPreferences.intent
             } : void 0,
             isStreamed: !!isStreaming
         };

package/dist/es/ai-model/service-caller/index.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/service-caller/index.mjs","sources":["webpack://@midscene/core/./src/ai-model/service-caller/index.ts"],"sourcesContent":["import { AIResponseFormat, type AIUsageInfo } from '@/types';\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\nimport { Anthropic } from '@anthropic-ai/sdk';\nimport {\n  DefaultAzureCredential,\n  getBearerTokenProvider,\n} from '@azure/identity';\nimport {\n  type IModelPreferences,\n  MIDSCENE_API_TYPE,\n  MIDSCENE_LANGSMITH_DEBUG,\n  OPENAI_MAX_TOKENS,\n  decideModelConfig,\n  getAIConfig,\n  getAIConfigInBoolean,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { parseBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { ifInBrowser } from '@midscene/shared/utils';\nimport { HttpsProxyAgent } from 'https-proxy-agent';\nimport { jsonrepair } from 'jsonrepair';\nimport OpenAI, { AzureOpenAI } from 'openai';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { Stream } from 'openai/streaming';\nimport { SocksProxyAgent } from 'socks-proxy-agent';\nimport { AIActionType, type AIArgs } from '../common';\nimport { assertSchema } from '../prompt/assertion';\nimport { locatorSchema } from '../prompt/llm-locator';\nimport { planSchema } from '../prompt/llm-planning';\n\nasync function createChatClient({\n  AIActionTypeValue,\n  modelPreferences,\n}: {\n  AIActionTypeValue: AIActionType;\n  modelPreferences: IModelPreferences;\n}): Promise<{\n  completion: OpenAI.Chat.Completions;\n  style: 'openai' | 'anthropic';\n  modelName: string;\n}> {\n  const {\n    socksProxy,\n    httpProxy,\n    modelName,\n    openaiBaseURL,\n    openaiApiKey,\n    openaiExtraConfig,\n    openaiUseAzureDeprecated,\n    useAzureOpenai,\n    azureOpenaiScope,\n    azureOpenaiKey,\n    azureOpenaiEndpoint,\n    azureOpenaiApiVersion,\n    azureOpenaiDeployment,\n    azureExtraConfig,\n    useAnthropicSdk,\n    anthropicApiKey,\n  } = decideModelConfig(modelPreferences, true);\n\n  let openai: OpenAI | AzureOpenAI | undefined;\n\n  let proxyAgent = undefined;\n  const debugProxy = getDebug('ai:call:proxy');\n  if (httpProxy) {\n    debugProxy('using http proxy', httpProxy);\n    proxyAgent = new HttpsProxyAgent(httpProxy);\n  } else if (socksProxy) {\n    debugProxy('using socks proxy', socksProxy);\n    proxyAgent = new SocksProxyAgent(socksProxy);\n  }\n\n  if (openaiUseAzureDeprecated) {\n    // this is deprecated\n    openai = new AzureOpenAI({\n      baseURL: openaiBaseURL,\n      apiKey: openaiApiKey,\n      httpAgent: proxyAgent,\n      ...openaiExtraConfig,\n      dangerouslyAllowBrowser: true,\n    }) as OpenAI;\n  } else if (useAzureOpenai) {\n    // https://learn.microsoft.com/en-us/azure/ai-services/openai/chatgpt-quickstart?tabs=bash%2Cjavascript-key%2Ctypescript-keyless%2Cpython&pivots=programming-language-javascript#rest-api\n    // keyless authentication\n    let tokenProvider: any = undefined;\n    if (azureOpenaiScope) {\n      assert(\n        !ifInBrowser,\n        'Azure OpenAI is not supported in browser with Midscene.',\n      );\n      const credential = new DefaultAzureCredential();\n\n      tokenProvider = getBearerTokenProvider(credential, azureOpenaiScope);\n\n      openai = new AzureOpenAI({\n        azureADTokenProvider: tokenProvider,\n        endpoint: azureOpenaiEndpoint,\n        apiVersion: azureOpenaiApiVersion,\n        deployment: azureOpenaiDeployment,\n        ...openaiExtraConfig,\n        ...azureExtraConfig,\n      });\n    } else {\n      // endpoint, apiKey, apiVersion, deployment\n      openai = new AzureOpenAI({\n        apiKey: azureOpenaiKey,\n        endpoint: azureOpenaiEndpoint,\n        apiVersion: azureOpenaiApiVersion,\n        deployment: azureOpenaiDeployment,\n        dangerouslyAllowBrowser: true,\n        ...openaiExtraConfig,\n        ...azureExtraConfig,\n      });\n    }\n  } else if (!useAnthropicSdk) {\n    openai = new OpenAI({\n      baseURL: openaiBaseURL,\n      apiKey: openaiApiKey,\n      httpAgent: proxyAgent,\n      ...openaiExtraConfig,\n      defaultHeaders: {\n        ...(openaiExtraConfig?.defaultHeaders || {}),\n        [MIDSCENE_API_TYPE]: AIActionTypeValue.toString(),\n      },\n      dangerouslyAllowBrowser: true,\n    });\n  }\n\n  if (openai && getAIConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)) {\n    if (ifInBrowser) {\n      throw new Error('langsmith is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langsmith wrapper enabled');\n    const { wrapOpenAI } = await import('langsmith/wrappers');\n    openai = wrapOpenAI(openai);\n  }\n\n  if (typeof openai !== 'undefined') {\n    return {\n      completion: openai.chat.completions,\n      style: 'openai',\n      modelName,\n    };\n  }\n\n  // Anthropic\n  if (useAnthropicSdk) {\n    openai = new Anthropic({\n      apiKey: anthropicApiKey,\n      httpAgent: proxyAgent,\n      dangerouslyAllowBrowser: true,\n    }) as any;\n  }\n\n  if (typeof openai !== 'undefined' && (openai as any).messages) {\n    return {\n      completion: (openai as any).messages,\n      style: 'anthropic',\n      modelName,\n    };\n  }\n\n  throw new Error('Openai SDK or Anthropic SDK is not initialized');\n}\n\nexport async function call(\n  messages: ChatCompletionMessageParam[],\n  AIActionTypeValue: AIActionType,\n  modelPreferences: IModelPreferences,\n  options?: {\n    stream?: boolean;\n    onChunk?: StreamingCallback;\n  },\n): Promise<{ content: string; usage?: AIUsageInfo; isStreamed: boolean }> {\n  const { completion, style, modelName } = await createChatClient({\n    AIActionTypeValue,\n    modelPreferences,\n  });\n\n  const responseFormat = getResponseFormat(modelName, AIActionTypeValue);\n\n  const maxTokens = getAIConfig(OPENAI_MAX_TOKENS);\n  const debugCall = getDebug('ai:call');\n  const debugProfileStats = getDebug('ai:profile:stats');\n  const debugProfileDetail = getDebug('ai:profile:detail');\n\n  const startTime = Date.now();\n\n  const isStreaming = options?.stream && options?.onChunk;\n  let content: string | undefined;\n  let accumulated = '';\n  let usage: OpenAI.CompletionUsage | undefined;\n  let timeCost: number | undefined;\n\n  const commonConfig = {\n    temperature: vlLocateMode(modelPreferences) === 'vlm-ui-tars' ? 0.0 : 0.1,\n    stream: !!isStreaming,\n    max_tokens:\n      typeof maxTokens === 'number'\n        ? maxTokens\n        : Number.parseInt(maxTokens || '2048', 10),\n    ...(vlLocateMode(modelPreferences) === 'qwen-vl' // qwen specific config\n      ? {\n          vl_high_resolution_images: true,\n        }\n      : {}),\n  };\n\n  try {\n    if (style === 'openai') {\n      debugCall(\n        `sending ${isStreaming ? 'streaming ' : ''}request to ${modelName}`,\n      );\n\n      if (isStreaming) {\n        const stream = (await completion.create(\n          {\n            model: modelName,\n            messages,\n            response_format: responseFormat,\n            ...commonConfig,\n          },\n          {\n            stream: true,\n          },\n        )) as Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {\n          _request_id?: string | null;\n        };\n\n        for await (const chunk of stream) {\n          const content = chunk.choices?.[0]?.delta?.content || '';\n          const reasoning_content =\n            (chunk.choices?.[0]?.delta as any)?.reasoning_content || '';\n\n          // Check for usage info in any chunk (OpenAI provides usage in separate chunks)\n          if (chunk.usage) {\n            usage = chunk.usage;\n          }\n\n          if (content || reasoning_content) {\n            accumulated += content;\n            const chunkData: CodeGenerationChunk = {\n              content,\n              reasoning_content,\n              accumulated,\n              isComplete: false,\n              usage: undefined,\n            };\n            options.onChunk!(chunkData);\n          }\n\n          // Check if stream is complete\n          if (chunk.choices?.[0]?.finish_reason) {\n            timeCost = Date.now() - startTime;\n\n            // If usage is not available from the stream, provide a basic usage info\n            if (!usage) {\n              // Estimate token counts based on content length (rough approximation)\n              const estimatedTokens = Math.max(\n                1,\n                Math.floor(accumulated.length / 4),\n              );\n              usage = {\n                prompt_tokens: estimatedTokens,\n                completion_tokens: estimatedTokens,\n                total_tokens: estimatedTokens * 2,\n              };\n            }\n\n            // Send final chunk\n            const finalChunk: CodeGenerationChunk = {\n              content: '',\n              accumulated,\n              reasoning_content: '',\n              isComplete: true,\n              usage: {\n                prompt_tokens: usage.prompt_tokens ?? 0,\n                completion_tokens: usage.completion_tokens ?? 0,\n                total_tokens: usage.total_tokens ?? 0,\n                time_cost: timeCost ?? 0,\n                model_name: modelName,\n              },\n            };\n            options.onChunk!(finalChunk);\n            break;\n          }\n        }\n        content = accumulated;\n        debugProfileStats(\n          `streaming model, ${modelName}, mode, ${vlLocateMode(modelPreferences) || 'default'}, cost-ms, ${timeCost}`,\n        );\n      } else {\n        const result = await completion.create({\n          model: modelName,\n          messages,\n          response_format: responseFormat,\n          ...commonConfig,\n        } as any);\n        timeCost = Date.now() - startTime;\n\n        debugProfileStats(\n          `model, ${modelName}, mode, ${vlLocateMode(modelPreferences) || 'default'}, ui-tars-version, ${uiTarsModelVersion(modelPreferences)}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}`,\n        );\n\n        debugProfileDetail(\n          `model usage detail: ${JSON.stringify(result.usage)}`,\n        );\n\n        assert(\n          result.choices,\n          `invalid response from LLM service: ${JSON.stringify(result)}`,\n        );\n        content = result.choices[0].message.content!;\n        usage = result.usage;\n      }\n\n      debugCall(`response: ${content}`);\n      assert(content, 'empty content');\n    } else if (style === 'anthropic') {\n      const convertImageContent = (content: any) => {\n        if (content.type === 'image_url') {\n          const imgBase64 = content.image_url.url;\n          assert(imgBase64, 'image_url is required');\n          const { mimeType, body } = parseBase64(content.image_url.url);\n          return {\n            source: {\n              type: 'base64',\n              media_type: mimeType,\n              data: body,\n            },\n            type: 'image',\n          };\n        }\n        return content;\n      };\n\n      if (isStreaming) {\n        const stream = (await completion.create({\n          model: modelName,\n          system: 'You are a versatile professional in software UI automation',\n          messages: messages.map((m) => ({\n            role: 'user',\n            content: Array.isArray(m.content)\n              ? (m.content as any).map(convertImageContent)\n              : m.content,\n          })),\n          response_format: responseFormat,\n          ...commonConfig,\n        } as any)) as any;\n\n        for await (const chunk of stream) {\n          const content = chunk.delta?.text || '';\n          if (content) {\n            accumulated += content;\n            const chunkData: CodeGenerationChunk = {\n              content,\n              accumulated,\n              reasoning_content: '',\n              isComplete: false,\n              usage: undefined,\n            };\n            options.onChunk!(chunkData);\n          }\n\n          // Check if stream is complete\n          if (chunk.type === 'message_stop') {\n            timeCost = Date.now() - startTime;\n            const anthropicUsage = chunk.usage;\n\n            // Send final chunk\n            const finalChunk: CodeGenerationChunk = {\n              content: '',\n              accumulated,\n              reasoning_content: '',\n              isComplete: true,\n              usage: anthropicUsage\n                ? {\n                    prompt_tokens: anthropicUsage.input_tokens ?? 0,\n                    completion_tokens: anthropicUsage.output_tokens ?? 0,\n                    total_tokens:\n                      (anthropicUsage.input_tokens ?? 0) +\n                      (anthropicUsage.output_tokens ?? 0),\n                    time_cost: timeCost ?? 0,\n                    model_name: modelName,\n                  }\n                : undefined,\n            };\n            options.onChunk!(finalChunk);\n            break;\n          }\n        }\n        content = accumulated;\n      } else {\n        const result = await completion.create({\n          model: modelName,\n          system: 'You are a versatile professional in software UI automation',\n          messages: messages.map((m) => ({\n            role: 'user',\n            content: Array.isArray(m.content)\n              ? (m.content as any).map(convertImageContent)\n              : m.content,\n          })),\n          response_format: responseFormat,\n          ...commonConfig,\n        } as any);\n        timeCost = Date.now() - startTime;\n        content = (result as any).content[0].text as string;\n        usage = result.usage;\n      }\n\n      assert(content, 'empty content');\n    }\n    // Ensure we always have usage info for streaming responses\n    if (isStreaming && !usage) {\n      // Estimate token counts based on content length (rough approximation)\n      const estimatedTokens = Math.max(\n        1,\n        Math.floor((content || '').length / 4),\n      );\n      usage = {\n        prompt_tokens: estimatedTokens,\n        completion_tokens: estimatedTokens,\n        total_tokens: estimatedTokens * 2,\n      };\n    }\n\n    return {\n      content: content || '',\n      usage: usage\n        ? {\n            prompt_tokens: usage.prompt_tokens ?? 0,\n            completion_tokens: usage.completion_tokens ?? 0,\n            total_tokens: usage.total_tokens ?? 0,\n            time_cost: timeCost ?? 0,\n            model_name: modelName,\n          }\n        : undefined,\n      isStreamed: !!isStreaming,\n    };\n  } catch (e: any) {\n    console.error(' call AI error', e);\n    const newError = new Error(\n      `failed to call ${isStreaming ? 'streaming ' : ''}AI model service: ${e.message}. Trouble shooting: https://midscenejs.com/model-provider.html`,\n      {\n        cause: e,\n      },\n    );\n    throw newError;\n  }\n}\n\nexport const getResponseFormat = (\n  modelName: string,\n  AIActionTypeValue: AIActionType,\n):\n  | OpenAI.ChatCompletionCreateParams['response_format']\n  | OpenAI.ResponseFormatJSONObject => {\n  let responseFormat:\n    | OpenAI.ChatCompletionCreateParams['response_format']\n    | OpenAI.ResponseFormatJSONObject\n    | undefined;\n\n  if (modelName.includes('gpt-4')) {\n    switch (AIActionTypeValue) {\n      case AIActionType.ASSERT:\n        responseFormat = assertSchema;\n        break;\n      case AIActionType.INSPECT_ELEMENT:\n        responseFormat = locatorSchema;\n        break;\n      case AIActionType.PLAN:\n        responseFormat = planSchema;\n        break;\n      case AIActionType.EXTRACT_DATA:\n      case AIActionType.DESCRIBE_ELEMENT:\n        responseFormat = { type: AIResponseFormat.JSON };\n        break;\n    }\n  }\n\n  // gpt-4o-2024-05-13 only supports json_object response format\n  if (modelName === 'gpt-4o-2024-05-13') {\n    responseFormat = { type: AIResponseFormat.JSON };\n  }\n\n  return responseFormat;\n};\n\nexport async function callToGetJSONObject<T>(\n  messages: ChatCompletionMessageParam[],\n  AIActionTypeValue: AIActionType,\n  modelPreferences: IModelPreferences,\n): Promise<{ content: T; usage?: AIUsageInfo }> {\n  const response = await call(messages, AIActionTypeValue, modelPreferences);\n  assert(response, 'empty response');\n  const jsonContent = safeParseJson(response.content, modelPreferences);\n  return { content: jsonContent, usage: response.usage };\n}\n\nexport async function callAiFnWithStringResponse<T>(\n  msgs: AIArgs,\n  AIActionTypeValue: AIActionType,\n  modelPreferences: IModelPreferences,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n  const { content, usage } = await call(\n    msgs,\n    AIActionTypeValue,\n    modelPreferences,\n  );\n  return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n  try {\n    // First, try to match a JSON object directly in the response\n    const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\n    if (jsonMatch) {\n      return jsonMatch[1];\n    }\n\n    // If no direct JSON object is found, try to extract JSON from a code block\n    const codeBlockMatch = response.match(\n      /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\n    );\n    if (codeBlockMatch) {\n      return codeBlockMatch[1];\n    }\n\n    // If no code block is found, try to find a JSON-like structure in the text\n    const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\n    if (jsonLikeMatch) {\n      return jsonLikeMatch[0];\n    }\n  } catch {}\n  // If no JSON-like structure is found, return the original response\n  return response;\n}\n\nexport function preprocessDoubaoBboxJson(input: string) {\n  if (input.includes('bbox')) {\n    // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\n    while (/\\d+\\s+\\d+/.test(input)) {\n      input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n    }\n  }\n  return input;\n}\n\nexport function safeParseJson(\n  input: string,\n  modelPreferences: IModelPreferences,\n) {\n  const cleanJsonString = extractJSONFromCodeBlock(input);\n  // match the point\n  if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\n    return cleanJsonString\n      .match(/\\((\\d+),(\\d+)\\)/)\n      ?.slice(1)\n      .map(Number);\n  }\n  try {\n    return JSON.parse(cleanJsonString);\n  } catch {}\n  try {\n    return JSON.parse(jsonrepair(cleanJsonString));\n  } catch (e) {}\n\n  if (\n    vlLocateMode(modelPreferences) === 'doubao-vision' ||\n    vlLocateMode(modelPreferences) === 'vlm-ui-tars'\n  ) {\n    const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\n    return JSON.parse(jsonrepair(jsonString));\n  }\n  throw Error(`failed to parse json response: ${input}`);\n}\n"],"names":["createChatClient","AIActionTypeValue","modelPreferences","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","openaiUseAzureDeprecated","useAzureOpenai","azureOpenaiScope","azureOpenaiKey","azureOpenaiEndpoint","azureOpenaiApiVersion","azureOpenaiDeployment","azureExtraConfig","useAnthropicSdk","anthropicApiKey","decideModelConfig","openai","proxyAgent","debugProxy","getDebug","HttpsProxyAgent","SocksProxyAgent","AzureOpenAI","tokenProvider","assert","ifInBrowser","credential","DefaultAzureCredential","getBearerTokenProvider","OpenAI","MIDSCENE_API_TYPE","getAIConfigInBoolean","MIDSCENE_LANGSMITH_DEBUG","Error","console","wrapOpenAI","Anthropic","call","messages","options","completion","style","responseFormat","getResponseFormat","maxTokens","getAIConfig","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","isStreaming","content","accumulated","usage","timeCost","commonConfig","vlLocateMode","Number","stream","chunk","_chunk_choices__delta","_chunk_choices__delta1","_chunk_choices_2","reasoning_content","chunkData","undefined","estimatedTokens","Math","finalChunk","_result_usage","_result_usage1","_result_usage2","result","uiTarsModelVersion","JSON","convertImageContent","imgBase64","mimeType","body","parseBase64","m","Array","_chunk_delta","anthropicUsage","e","newError","AIActionType","assertSchema","locatorSchema","planSchema","AIResponseFormat","callToGetJSONObject","response","jsonContent","safeParseJson","callAiFnWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","cleanJsonString","_cleanJsonString_match","jsonrepair","jsonString"],"mappings":";;;;;;;;;;;;;;;AAiCA,eAAeA,iBAAiB,EAC9BC,iBAAiB,EACjBC,gBAAgB,EAIjB;IAKC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,wBAAwB,EACxBC,cAAc,EACdC,gBAAgB,EAChBC,cAAc,EACdC,mBAAmB,EACnBC,qBAAqB,EACrBC,qBAAqB,EACrBC,gBAAgB,EAChBC,eAAe,EACfC,eAAe,EAChB,GAAGC,kBAAkBjB,kBAAkB;IAExC,IAAIkB;IAEJ,IAAIC;IACJ,MAAMC,aAAaC,SAAS;IAC5B,IAAInB,WAAW;QACbkB,WAAW,oBAAoBlB;QAC/BiB,aAAa,IAAIG,gBAAgBpB;IACnC,OAAO,IAAID,YAAY;QACrBmB,WAAW,qBAAqBnB;QAChCkB,aAAa,IAAII,gBAAgBtB;IACnC;IAEA,IAAIM,0BAEFW,SAAS,IAAIM,YAAY;QACvB,SAASpB;QACT,QAAQC;QACR,WAAWc;QACX,GAAGb,iBAAiB;QACpB,yBAAyB;IAC3B;SACK,IAAIE,gBAAgB;QAGzB,IAAIiB;QACJ,IAAIhB,kBAAkB;YACpBiB,OACE,CAACC,aACD;YAEF,MAAMC,aAAa,IAAIC;YAEvBJ,gBAAgBK,uBAAuBF,YAAYnB;YAEnDS,SAAS,IAAIM,YAAY;gBACvB,sBAAsBC;gBACtB,UAAUd;gBACV,YAAYC;gBACZ,YAAYC;gBACZ,GAAGP,iBAAiB;gBACpB,GAAGQ,gBAAgB;YACrB;QACF,OAEEI,SAAS,IAAIM,YAAY;YACvB,QAAQd;YACR,UAAUC;YACV,YAAYC;YACZ,YAAYC;YACZ,yBAAyB;YACzB,GAAGP,iBAAiB;YACpB,GAAGQ,gBAAgB;QACrB;IAEJ,OAAO,IAAI,CAACC,iBACVG,SAAS,IAAIa,SAAO;QAClB,SAAS3B;QACT,QAAQC;QACR,WAAWc;QACX,GAAGb,iBAAiB;QACpB,gBAAgB;YACd,GAAIA,AAAAA,CAAAA,QAAAA,oBAAAA,KAAAA,IAAAA,kBAAmB,cAAc,AAAD,KAAK,CAAC,CAAC;YAC3C,CAAC0B,kBAAkB,EAAEjC,kBAAkB,QAAQ;QACjD;QACA,yBAAyB;IAC3B;IAGF,IAAImB,UAAUe,qBAAqBC,2BAA2B;QAC5D,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;QACPf;IACF;IAIF,IAAIY,iBACFG,SAAS,IAAIoB,UAAU;QACrB,QAAQtB;QACR,WAAWG;QACX,yBAAyB;IAC3B;IAGF,IAAI,AAAkB,WAAXD,UAA2BA,OAAe,QAAQ,EAC3D,OAAO;QACL,YAAaA,OAAe,QAAQ;QACpC,OAAO;QACPf;IACF;IAGF,MAAM,IAAIgC,MAAM;AAClB;AAEO,eAAeI,KACpBC,QAAsC,EACtCzC,iBAA+B,EAC/BC,gBAAmC,EACnCyC,OAGC;IAED,MAAM,EAAEC,UAAU,EAAEC,KAAK,EAAExC,SAAS,EAAE,GAAG,MAAML,iBAAiB;QAC9DC;QACAC;IACF;IAEA,MAAM4C,iBAAiBC,kBAAkB1C,WAAWJ;IAEpD,MAAM+C,YAAYC,YAAYC;IAC9B,MAAMC,YAAY5B,SAAS;IAC3B,MAAM6B,oBAAoB7B,SAAS;IACnC,MAAM8B,qBAAqB9B,SAAS;IAEpC,MAAM+B,YAAYC,KAAK,GAAG;IAE1B,MAAMC,cAAcb,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM,AAAD,KAAKA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO,AAAD;IACtD,IAAIc;IACJ,IAAIC,cAAc;IAClB,IAAIC;IACJ,IAAIC;IAEJ,MAAMC,eAAe;QACnB,aAAaC,AAAmC,kBAAnCA,aAAa5D,oBAAsC,MAAM;QACtE,QAAQ,CAAC,CAACsD;QACV,YACE,AAAqB,YAArB,OAAOR,YACHA,YACAe,OAAO,QAAQ,CAACf,aAAa,QAAQ;QAC3C,GAAIc,AAAmC,cAAnCA,aAAa5D,oBACb;YACE,2BAA2B;QAC7B,IACA,CAAC,CAAC;IACR;IAEA,IAAI;QACF,IAAI2C,AAAU,aAAVA,OAAoB;YACtBM,UACE,CAAC,QAAQ,EAAEK,cAAc,eAAe,GAAG,WAAW,EAAEnD,WAAW;YAGrE,IAAImD,aAAa;gBACf,MAAMQ,SAAU,MAAMpB,WAAW,MAAM,CACrC;oBACE,OAAOvC;oBACPqC;oBACA,iBAAiBI;oBACjB,GAAGe,YAAY;gBACjB,GACA;oBACE,QAAQ;gBACV;gBAKF,WAAW,MAAMI,SAASD,OAAQ;wBAChBE,uBAAAA,iBAAAA,gBAEbC,wBAAAA,kBAAAA,iBAoBCC,kBAAAA;oBAtBJ,MAAMX,UAAUS,AAAAA,SAAAA,CAAAA,iBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,kBAAAA,cAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,QAAAA,CAAAA,wBAAAA,gBAAoB,KAAK,AAAD,IAAxBA,KAAAA,IAAAA,sBAA2B,OAAO,AAAD,KAAK;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,EACbN,QAAQM,MAAM,KAAK;oBAGrB,IAAIR,WAAWY,mBAAmB;wBAChCX,eAAeD;wBACf,MAAMa,YAAiC;4BACrCb;4BACAY;4BACAX;4BACA,YAAY;4BACZ,OAAOa;wBACT;wBACA5B,QAAQ,OAAO,CAAE2B;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;wBACrCR,WAAWL,KAAK,GAAG,KAAKD;wBAGxB,IAAI,CAACK,OAAO;4BAEV,MAAMa,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAACf,YAAY,MAAM,GAAG;4BAElCC,QAAQ;gCACN,eAAea;gCACf,mBAAmBA;gCACnB,cAAcA,AAAkB,IAAlBA;4BAChB;wBACF;wBAGA,MAAME,aAAkC;4BACtC,SAAS;4BACThB;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,YAAYvD;4BACd;wBACF;wBACAsC,QAAQ,OAAO,CAAE+B;wBACjB;oBACF;gBACF;gBACAjB,UAAUC;gBACVN,kBACE,CAAC,iBAAiB,EAAE/C,UAAU,QAAQ,EAAEyD,aAAa5D,qBAAqB,UAAU,WAAW,EAAE0D,UAAU;YAE/G,OAAO;oBAUoJe,eAAyDC,gBAAwDC;gBAT1Q,MAAMC,SAAS,MAAMlC,WAAW,MAAM,CAAC;oBACrC,OAAOvC;oBACPqC;oBACA,iBAAiBI;oBACjB,GAAGe,YAAY;gBACjB;gBACAD,WAAWL,KAAK,GAAG,KAAKD;gBAExBF,kBACE,CAAC,OAAO,EAAE/C,UAAU,QAAQ,EAAEyD,aAAa5D,qBAAqB,UAAU,mBAAmB,EAAE6E,mBAAmB7E,kBAAkB,iBAAiB,EAAEyE,AAAAA,SAAAA,CAAAA,gBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,cAAc,aAAa,AAAD,KAAK,GAAG,qBAAqB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,iBAAiB,AAAD,KAAK,GAAG,gBAAgB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,YAAY,AAAD,KAAK,GAAG,WAAW,EAAEjB,SAAS,aAAa,EAAEkB,OAAO,WAAW,IAAI,IAAI;gBAG1WzB,mBACE,CAAC,oBAAoB,EAAE2B,KAAK,SAAS,CAACF,OAAO,KAAK,GAAG;gBAGvDlD,OACEkD,OAAO,OAAO,EACd,CAAC,mCAAmC,EAAEE,KAAK,SAAS,CAACF,SAAS;gBAEhErB,UAAUqB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;gBAC3CnB,QAAQmB,OAAO,KAAK;YACtB;YAEA3B,UAAU,CAAC,UAAU,EAAEM,SAAS;YAChC7B,OAAO6B,SAAS;QAClB,OAAO,IAAIZ,AAAU,gBAAVA,OAAuB;YAChC,MAAMoC,sBAAsB,CAACxB;gBAC3B,IAAIA,AAAiB,gBAAjBA,QAAQ,IAAI,EAAkB;oBAChC,MAAMyB,YAAYzB,QAAQ,SAAS,CAAC,GAAG;oBACvC7B,OAAOsD,WAAW;oBAClB,MAAM,EAAEC,QAAQ,EAAEC,IAAI,EAAE,GAAGC,YAAY5B,QAAQ,SAAS,CAAC,GAAG;oBAC5D,OAAO;wBACL,QAAQ;4BACN,MAAM;4BACN,YAAY0B;4BACZ,MAAMC;wBACR;wBACA,MAAM;oBACR;gBACF;gBACA,OAAO3B;YACT;YAEA,IAAID,aAAa;gBACf,MAAMQ,SAAU,MAAMpB,WAAW,MAAM,CAAC;oBACtC,OAAOvC;oBACP,QAAQ;oBACR,UAAUqC,SAAS,GAAG,CAAC,CAAC4C,IAAO;4BAC7B,MAAM;4BACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;wBACf;oBACA,iBAAiBxC;oBACjB,GAAGe,YAAY;gBACjB;gBAEA,WAAW,MAAMI,SAASD,OAAQ;wBAChBwB;oBAAhB,MAAM/B,UAAU+B,AAAAA,SAAAA,CAAAA,eAAAA,MAAM,KAAK,AAAD,IAAVA,KAAAA,IAAAA,aAAa,IAAI,AAAD,KAAK;oBACrC,IAAI/B,SAAS;wBACXC,eAAeD;wBACf,MAAMa,YAAiC;4BACrCb;4BACAC;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAOa;wBACT;wBACA5B,QAAQ,OAAO,CAAE2B;oBACnB;oBAGA,IAAIL,AAAe,mBAAfA,MAAM,IAAI,EAAqB;wBACjCL,WAAWL,KAAK,GAAG,KAAKD;wBACxB,MAAMmC,iBAAiBxB,MAAM,KAAK;wBAGlC,MAAMS,aAAkC;4BACtC,SAAS;4BACThB;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAO+B,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,WAAW7B,YAAY;gCACvB,YAAYvD;4BACd,IACAkE;wBACN;wBACA5B,QAAQ,OAAO,CAAE+B;wBACjB;oBACF;gBACF;gBACAjB,UAAUC;YACZ,OAAO;gBACL,MAAMoB,SAAS,MAAMlC,WAAW,MAAM,CAAC;oBACrC,OAAOvC;oBACP,QAAQ;oBACR,UAAUqC,SAAS,GAAG,CAAC,CAAC4C,IAAO;4BAC7B,MAAM;4BACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;wBACf;oBACA,iBAAiBxC;oBACjB,GAAGe,YAAY;gBACjB;gBACAD,WAAWL,KAAK,GAAG,KAAKD;gBACxBG,UAAWqB,OAAe,OAAO,CAAC,EAAE,CAAC,IAAI;gBACzCnB,QAAQmB,OAAO,KAAK;YACtB;YAEAlD,OAAO6B,SAAS;QAClB;QAEA,IAAID,eAAe,CAACG,OAAO;YAEzB,MAAMa,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEhB,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;YAEtCE,QAAQ;gBACN,eAAea;gBACf,mBAAmBA;gBACnB,cAAcA,AAAkB,IAAlBA;YAChB;QACF;QAEA,OAAO;YACL,SAASf,WAAW;YACpB,OAAOE,QACH;gBACE,eAAeA,MAAM,aAAa,IAAI;gBACtC,mBAAmBA,MAAM,iBAAiB,IAAI;gBAC9C,cAAcA,MAAM,YAAY,IAAI;gBACpC,WAAWC,YAAY;gBACvB,YAAYvD;YACd,IACAkE;YACJ,YAAY,CAAC,CAACf;QAChB;IACF,EAAE,OAAOkC,GAAQ;QACfpD,QAAQ,KAAK,CAAC,kBAAkBoD;QAChC,MAAMC,WAAW,IAAItD,MACnB,CAAC,eAAe,EAAEmB,cAAc,eAAe,GAAG,kBAAkB,EAAEkC,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC/I;YACE,OAAOA;QACT;QAEF,MAAMC;IACR;AACF;AAEO,MAAM5C,oBAAoB,CAC/B1C,WACAJ;IAIA,IAAI6C;IAKJ,IAAIzC,UAAU,QAAQ,CAAC,UACrB,OAAQJ;QACN,KAAK2F,aAAa,MAAM;YACtB9C,iBAAiB+C;YACjB;QACF,KAAKD,aAAa,eAAe;YAC/B9C,iBAAiBgD;YACjB;QACF,KAAKF,aAAa,IAAI;YACpB9C,iBAAiBiD;YACjB;QACF,KAAKH,aAAa,YAAY;QAC9B,KAAKA,aAAa,gBAAgB;YAChC9C,iBAAiB;gBAAE,MAAMkD,iBAAiB,IAAI;YAAC;YAC/C;IACJ;IAIF,IAAI3F,AAAc,wBAAdA,WACFyC,iBAAiB;QAAE,MAAMkD,iBAAiB,IAAI;IAAC;IAGjD,OAAOlD;AACT;AAEO,eAAemD,oBACpBvD,QAAsC,EACtCzC,iBAA+B,EAC/BC,gBAAmC;IAEnC,MAAMgG,WAAW,MAAMzD,KAAKC,UAAUzC,mBAAmBC;IACzD0B,OAAOsE,UAAU;IACjB,MAAMC,cAAcC,cAAcF,SAAS,OAAO,EAAEhG;IACpD,OAAO;QAAE,SAASiG;QAAa,OAAOD,SAAS,KAAK;IAAC;AACvD;AAEO,eAAeG,2BACpBC,IAAY,EACZrG,iBAA+B,EAC/BC,gBAAmC;IAEnC,MAAM,EAAEuD,OAAO,EAAEE,KAAK,EAAE,GAAG,MAAMlB,KAC/B6D,MACArG,mBACAC;IAEF,OAAO;QAAEuD;QAASE;IAAM;AAC1B;AAEO,SAAS4C,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,cACdQ,KAAa,EACb1G,gBAAmC;IAEnC,MAAM2G,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,CAAC/C;IACT;IACA,IAAI;QACF,OAAOiB,KAAK,KAAK,CAAC6B;IACpB,EAAE,OAAM,CAAC;IACT,IAAI;QACF,OAAO7B,KAAK,KAAK,CAAC+B,WAAWF;IAC/B,EAAE,OAAOnB,GAAG,CAAC;IAEb,IACE5B,AAAmC,oBAAnCA,aAAa5D,qBACb4D,AAAmC,kBAAnCA,aAAa5D,mBACb;QACA,MAAM8G,aAAaL,yBAAyBE;QAC5C,OAAO7B,KAAK,KAAK,CAAC+B,WAAWC;IAC/B;IACA,MAAM3E,MAAM,CAAC,+BAA+B,EAAEuE,OAAO;AACvD"}
+{"version":3,"file":"ai-model/service-caller/index.mjs","sources":["webpack://@midscene/core/./src/ai-model/service-caller/index.ts"],"sourcesContent":["import { AIResponseFormat, type AIUsageInfo } from '@/types';\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\nimport { Anthropic } from '@anthropic-ai/sdk';\nimport {\n  DefaultAzureCredential,\n  getBearerTokenProvider,\n} from '@azure/identity';\nimport {\n  type IModelPreferences,\n  MIDSCENE_API_TYPE,\n  MIDSCENE_LANGSMITH_DEBUG,\n  OPENAI_MAX_TOKENS,\n  decideModelConfig,\n  getAIConfig,\n  getAIConfigInBoolean,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { parseBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport { ifInBrowser } from '@midscene/shared/utils';\nimport { HttpsProxyAgent } from 'https-proxy-agent';\nimport { jsonrepair } from 'jsonrepair';\nimport OpenAI, { AzureOpenAI } from 'openai';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { Stream } from 'openai/streaming';\nimport { SocksProxyAgent } from 'socks-proxy-agent';\nimport { AIActionType, type AIArgs } from '../common';\nimport { assertSchema } from '../prompt/assertion';\nimport { locatorSchema } from '../prompt/llm-locator';\nimport { planSchema } from '../prompt/llm-planning';\n\nasync function createChatClient({\n  AIActionTypeValue,\n  modelPreferences,\n}: {\n  AIActionTypeValue: AIActionType;\n  modelPreferences: IModelPreferences;\n}): Promise<{\n  completion: OpenAI.Chat.Completions;\n  style: 'openai' | 'anthropic';\n  modelName: string;\n  modelDescription: string;\n}> {\n  const {\n    socksProxy,\n    httpProxy,\n    modelName,\n    openaiBaseURL,\n    openaiApiKey,\n    openaiExtraConfig,\n    openaiUseAzureDeprecated,\n    useAzureOpenai,\n    azureOpenaiScope,\n    azureOpenaiKey,\n    azureOpenaiEndpoint,\n    azureOpenaiApiVersion,\n    azureOpenaiDeployment,\n    azureExtraConfig,\n    useAnthropicSdk,\n    anthropicApiKey,\n    modelDescription,\n  } = decideModelConfig(modelPreferences, true);\n\n  let openai: OpenAI | AzureOpenAI | undefined;\n\n  let proxyAgent = undefined;\n  const debugProxy = getDebug('ai:call:proxy');\n  if (httpProxy) {\n    debugProxy('using http proxy', httpProxy);\n    proxyAgent = new HttpsProxyAgent(httpProxy);\n  } else if (socksProxy) {\n    debugProxy('using socks proxy', socksProxy);\n    proxyAgent = new SocksProxyAgent(socksProxy);\n  }\n\n  if (openaiUseAzureDeprecated) {\n    // this is deprecated\n    openai = new AzureOpenAI({\n      baseURL: openaiBaseURL,\n      apiKey: openaiApiKey,\n      httpAgent: proxyAgent,\n      ...openaiExtraConfig,\n      dangerouslyAllowBrowser: true,\n    }) as OpenAI;\n  } else if (useAzureOpenai) {\n    // https://learn.microsoft.com/en-us/azure/ai-services/openai/chatgpt-quickstart?tabs=bash%2Cjavascript-key%2Ctypescript-keyless%2Cpython&pivots=programming-language-javascript#rest-api\n    // keyless authentication\n    let tokenProvider: any = undefined;\n    if (azureOpenaiScope) {\n      assert(\n        !ifInBrowser,\n        'Azure OpenAI is not supported in browser with Midscene.',\n      );\n      const credential = new DefaultAzureCredential();\n\n      tokenProvider = getBearerTokenProvider(credential, azureOpenaiScope);\n\n      openai = new AzureOpenAI({\n        azureADTokenProvider: tokenProvider,\n        endpoint: azureOpenaiEndpoint,\n        apiVersion: azureOpenaiApiVersion,\n        deployment: azureOpenaiDeployment,\n        ...openaiExtraConfig,\n        ...azureExtraConfig,\n      });\n    } else {\n      // endpoint, apiKey, apiVersion, deployment\n      openai = new AzureOpenAI({\n        apiKey: azureOpenaiKey,\n        endpoint: azureOpenaiEndpoint,\n        apiVersion: azureOpenaiApiVersion,\n        deployment: azureOpenaiDeployment,\n        dangerouslyAllowBrowser: true,\n        ...openaiExtraConfig,\n        ...azureExtraConfig,\n      });\n    }\n  } else if (!useAnthropicSdk) {\n    openai = new OpenAI({\n      baseURL: openaiBaseURL,\n      apiKey: openaiApiKey,\n      httpAgent: proxyAgent,\n      ...openaiExtraConfig,\n      defaultHeaders: {\n        ...(openaiExtraConfig?.defaultHeaders || {}),\n        [MIDSCENE_API_TYPE]: AIActionTypeValue.toString(),\n      },\n      dangerouslyAllowBrowser: true,\n    });\n  }\n\n  if (openai && getAIConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)) {\n    if (ifInBrowser) {\n      throw new Error('langsmith is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langsmith wrapper enabled');\n    const { wrapOpenAI } = await import('langsmith/wrappers');\n    openai = wrapOpenAI(openai);\n  }\n\n  if (typeof openai !== 'undefined') {\n    return {\n      completion: openai.chat.completions,\n      style: 'openai',\n      modelName,\n      modelDescription,\n    };\n  }\n\n  // Anthropic\n  if (useAnthropicSdk) {\n    openai = new Anthropic({\n      apiKey: anthropicApiKey,\n      httpAgent: proxyAgent,\n      dangerouslyAllowBrowser: true,\n    }) as any;\n  }\n\n  if (typeof openai !== 'undefined' && (openai as any).messages) {\n    return {\n      completion: (openai as any).messages,\n      style: 'anthropic',\n      modelName,\n      modelDescription,\n    };\n  }\n\n  throw new Error('Openai SDK or Anthropic SDK is not initialized');\n}\n\nexport async function call(\n  messages: ChatCompletionMessageParam[],\n  AIActionTypeValue: AIActionType,\n  modelPreferences: IModelPreferences,\n  options?: {\n    stream?: boolean;\n    onChunk?: StreamingCallback;\n  },\n): Promise<{ content: string; usage?: AIUsageInfo; isStreamed: boolean }> {\n  const { completion, style, modelName, modelDescription } =\n    await createChatClient({\n      AIActionTypeValue,\n      modelPreferences,\n    });\n\n  const responseFormat = getResponseFormat(modelName, AIActionTypeValue);\n\n  const maxTokens = getAIConfig(OPENAI_MAX_TOKENS);\n  const debugCall = getDebug('ai:call');\n  const debugProfileStats = getDebug('ai:profile:stats');\n  const debugProfileDetail = getDebug('ai:profile:detail');\n\n  const startTime = Date.now();\n\n  const isStreaming = options?.stream && options?.onChunk;\n  let content: string | undefined;\n  let accumulated = '';\n  let usage: OpenAI.CompletionUsage | undefined;\n  let timeCost: number | undefined;\n\n  const commonConfig = {\n    temperature: vlLocateMode(modelPreferences) === 'vlm-ui-tars' ? 0.0 : 0.1,\n    stream: !!isStreaming,\n    max_tokens:\n      typeof maxTokens === 'number'\n        ? maxTokens\n        : Number.parseInt(maxTokens || '2048', 10),\n    ...(vlLocateMode(modelPreferences) === 'qwen-vl' // qwen specific config\n      ? {\n          vl_high_resolution_images: true,\n        }\n      : {}),\n  };\n\n  try {\n    if (style === 'openai') {\n      debugCall(\n        `sending ${isStreaming ? 'streaming ' : ''}request to ${modelName}`,\n      );\n\n      if (isStreaming) {\n        const stream = (await completion.create(\n          {\n            model: modelName,\n            messages,\n            response_format: responseFormat,\n            ...commonConfig,\n          },\n          {\n            stream: true,\n          },\n        )) as Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {\n          _request_id?: string | null;\n        };\n\n        for await (const chunk of stream) {\n          const content = chunk.choices?.[0]?.delta?.content || '';\n          const reasoning_content =\n            (chunk.choices?.[0]?.delta as any)?.reasoning_content || '';\n\n          // Check for usage info in any chunk (OpenAI provides usage in separate chunks)\n          if (chunk.usage) {\n            usage = chunk.usage;\n          }\n\n          if (content || reasoning_content) {\n            accumulated += content;\n            const chunkData: CodeGenerationChunk = {\n              content,\n              reasoning_content,\n              accumulated,\n              isComplete: false,\n              usage: undefined,\n            };\n            options.onChunk!(chunkData);\n          }\n\n          // Check if stream is complete\n          if (chunk.choices?.[0]?.finish_reason) {\n            timeCost = Date.now() - startTime;\n\n            // If usage is not available from the stream, provide a basic usage info\n            if (!usage) {\n              // Estimate token counts based on content length (rough approximation)\n              const estimatedTokens = Math.max(\n                1,\n                Math.floor(accumulated.length / 4),\n              );\n              usage = {\n                prompt_tokens: estimatedTokens,\n                completion_tokens: estimatedTokens,\n                total_tokens: estimatedTokens * 2,\n              };\n            }\n\n            // Send final chunk\n            const finalChunk: CodeGenerationChunk = {\n              content: '',\n              accumulated,\n              reasoning_content: '',\n              isComplete: true,\n              usage: {\n                prompt_tokens: usage.prompt_tokens ?? 0,\n                completion_tokens: usage.completion_tokens ?? 0,\n                total_tokens: usage.total_tokens ?? 0,\n                time_cost: timeCost ?? 0,\n                model_name: modelName,\n                model_description: modelDescription,\n                intent: modelPreferences.intent,\n              },\n            };\n            options.onChunk!(finalChunk);\n            break;\n          }\n        }\n        content = accumulated;\n        debugProfileStats(\n          `streaming model, ${modelName}, mode, ${vlLocateMode(modelPreferences) || 'default'}, cost-ms, ${timeCost}`,\n        );\n      } else {\n        const result = await completion.create({\n          model: modelName,\n          messages,\n          response_format: responseFormat,\n          ...commonConfig,\n        } as any);\n        timeCost = Date.now() - startTime;\n\n        debugProfileStats(\n          `model, ${modelName}, mode, ${vlLocateMode(modelPreferences) || 'default'}, ui-tars-version, ${uiTarsModelVersion(modelPreferences)}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}`,\n        );\n\n        debugProfileDetail(\n          `model usage detail: ${JSON.stringify(result.usage)}`,\n        );\n\n        assert(\n          result.choices,\n          `invalid response from LLM service: ${JSON.stringify(result)}`,\n        );\n        content = result.choices[0].message.content!;\n        usage = result.usage;\n      }\n\n      debugCall(`response: ${content}`);\n      assert(content, 'empty content');\n    } else if (style === 'anthropic') {\n      const convertImageContent = (content: any) => {\n        if (content.type === 'image_url') {\n          const imgBase64 = content.image_url.url;\n          assert(imgBase64, 'image_url is required');\n          const { mimeType, body } = parseBase64(content.image_url.url);\n          return {\n            source: {\n              type: 'base64',\n              media_type: mimeType,\n              data: body,\n            },\n            type: 'image',\n          };\n        }\n        return content;\n      };\n\n      if (isStreaming) {\n        const stream = (await completion.create({\n          model: modelName,\n          system: 'You are a versatile professional in software UI automation',\n          messages: messages.map((m) => ({\n            role: 'user',\n            content: Array.isArray(m.content)\n              ? (m.content as any).map(convertImageContent)\n              : m.content,\n          })),\n          response_format: responseFormat,\n          ...commonConfig,\n        } as any)) as any;\n\n        for await (const chunk of stream) {\n          const content = chunk.delta?.text || '';\n          if (content) {\n            accumulated += content;\n            const chunkData: CodeGenerationChunk = {\n              content,\n              accumulated,\n              reasoning_content: '',\n              isComplete: false,\n              usage: undefined,\n            };\n            options.onChunk!(chunkData);\n          }\n\n          // Check if stream is complete\n          if (chunk.type === 'message_stop') {\n            timeCost = Date.now() - startTime;\n            const anthropicUsage = chunk.usage;\n\n            // Send final chunk\n            const finalChunk: CodeGenerationChunk = {\n              content: '',\n              accumulated,\n              reasoning_content: '',\n              isComplete: true,\n              usage: anthropicUsage\n                ? {\n                    prompt_tokens: anthropicUsage.input_tokens ?? 0,\n                    completion_tokens: anthropicUsage.output_tokens ?? 0,\n                    total_tokens:\n                      (anthropicUsage.input_tokens ?? 0) +\n                      (anthropicUsage.output_tokens ?? 0),\n                    time_cost: timeCost ?? 0,\n                    model_name: modelName,\n                    model_description: modelDescription,\n                    intent: modelPreferences.intent,\n                  }\n                : undefined,\n            };\n            options.onChunk!(finalChunk);\n            break;\n          }\n        }\n        content = accumulated;\n      } else {\n        const result = await completion.create({\n          model: modelName,\n          system: 'You are a versatile professional in software UI automation',\n          messages: messages.map((m) => ({\n            role: 'user',\n            content: Array.isArray(m.content)\n              ? (m.content as any).map(convertImageContent)\n              : m.content,\n          })),\n          response_format: responseFormat,\n          ...commonConfig,\n        } as any);\n        timeCost = Date.now() - startTime;\n        content = (result as any).content[0].text as string;\n        usage = result.usage;\n      }\n\n      assert(content, 'empty content');\n    }\n    // Ensure we always have usage info for streaming responses\n    if (isStreaming && !usage) {\n      // Estimate token counts based on content length (rough approximation)\n      const estimatedTokens = Math.max(\n        1,\n        Math.floor((content || '').length / 4),\n      );\n      usage = {\n        prompt_tokens: estimatedTokens,\n        completion_tokens: estimatedTokens,\n        total_tokens: estimatedTokens * 2,\n      };\n    }\n\n    return {\n      content: content || '',\n      usage: usage\n        ? {\n            prompt_tokens: usage.prompt_tokens ?? 0,\n            completion_tokens: usage.completion_tokens ?? 0,\n            total_tokens: usage.total_tokens ?? 0,\n            time_cost: timeCost ?? 0,\n            model_name: modelName,\n            model_description: modelDescription,\n            intent: modelPreferences.intent,\n          }\n        : undefined,\n      isStreamed: !!isStreaming,\n    };\n  } catch (e: any) {\n    console.error(' call AI error', e);\n    const newError = new Error(\n      `failed to call ${isStreaming ? 'streaming ' : ''}AI model service: ${e.message}. Trouble shooting: https://midscenejs.com/model-provider.html`,\n      {\n        cause: e,\n      },\n    );\n    throw newError;\n  }\n}\n\nexport const getResponseFormat = (\n  modelName: string,\n  AIActionTypeValue: AIActionType,\n):\n  | OpenAI.ChatCompletionCreateParams['response_format']\n  | OpenAI.ResponseFormatJSONObject => {\n  let responseFormat:\n    | OpenAI.ChatCompletionCreateParams['response_format']\n    | OpenAI.ResponseFormatJSONObject\n    | undefined;\n\n  if (modelName.includes('gpt-4')) {\n    switch (AIActionTypeValue) {\n      case AIActionType.ASSERT:\n        responseFormat = assertSchema;\n        break;\n      case AIActionType.INSPECT_ELEMENT:\n        responseFormat = locatorSchema;\n        break;\n      case AIActionType.PLAN:\n        responseFormat = planSchema;\n        break;\n      case AIActionType.EXTRACT_DATA:\n      case AIActionType.DESCRIBE_ELEMENT:\n        responseFormat = { type: AIResponseFormat.JSON };\n        break;\n    }\n  }\n\n  // gpt-4o-2024-05-13 only supports json_object response format\n  if (modelName === 'gpt-4o-2024-05-13') {\n    responseFormat = { type: AIResponseFormat.JSON };\n  }\n\n  return responseFormat;\n};\n\nexport async function callToGetJSONObject<T>(\n  messages: ChatCompletionMessageParam[],\n  AIActionTypeValue: AIActionType,\n  modelPreferences: IModelPreferences,\n): Promise<{ content: T; usage?: AIUsageInfo }> {\n  const response = await call(messages, AIActionTypeValue, modelPreferences);\n  assert(response, 'empty response');\n  const jsonContent = safeParseJson(response.content, modelPreferences);\n  return { content: jsonContent, usage: response.usage };\n}\n\nexport async function callAiFnWithStringResponse<T>(\n  msgs: AIArgs,\n  AIActionTypeValue: AIActionType,\n  modelPreferences: IModelPreferences,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n  const { content, usage } = await call(\n    msgs,\n    AIActionTypeValue,\n    modelPreferences,\n  );\n  return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n  try {\n    // First, try to match a JSON object directly in the response\n    const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\n    if (jsonMatch) {\n      return jsonMatch[1];\n    }\n\n    // If no direct JSON object is found, try to extract JSON from a code block\n    const codeBlockMatch = response.match(\n      /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\n    );\n    if (codeBlockMatch) {\n      return codeBlockMatch[1];\n    }\n\n    // If no code block is found, try to find a JSON-like structure in the text\n    const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\n    if (jsonLikeMatch) {\n      return jsonLikeMatch[0];\n    }\n  } catch {}\n  // If no JSON-like structure is found, return the original response\n  return response;\n}\n\nexport function preprocessDoubaoBboxJson(input: string) {\n  if (input.includes('bbox')) {\n    // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\n    while (/\\d+\\s+\\d+/.test(input)) {\n      input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n    }\n  }\n  return input;\n}\n\nexport function safeParseJson(\n  input: string,\n  modelPreferences: IModelPreferences,\n) {\n  const cleanJsonString = extractJSONFromCodeBlock(input);\n  // match the point\n  if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\n    return cleanJsonString\n      .match(/\\((\\d+),(\\d+)\\)/)\n      ?.slice(1)\n      .map(Number);\n  }\n  try {\n    return JSON.parse(cleanJsonString);\n  } catch {}\n  try {\n    return JSON.parse(jsonrepair(cleanJsonString));\n  } catch (e) {}\n\n  if (\n    vlLocateMode(modelPreferences) === 'doubao-vision' ||\n    vlLocateMode(modelPreferences) === 'vlm-ui-tars'\n  ) {\n    const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\n    return JSON.parse(jsonrepair(jsonString));\n  }\n  throw Error(`failed to parse json response: ${input}`);\n}\n"],"names":["createChatClient","AIActionTypeValue","modelPreferences","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","openaiUseAzureDeprecated","useAzureOpenai","azureOpenaiScope","azureOpenaiKey","azureOpenaiEndpoint","azureOpenaiApiVersion","azureOpenaiDeployment","azureExtraConfig","useAnthropicSdk","anthropicApiKey","modelDescription","decideModelConfig","openai","proxyAgent","debugProxy","getDebug","HttpsProxyAgent","SocksProxyAgent","AzureOpenAI","tokenProvider","assert","ifInBrowser","credential","DefaultAzureCredential","getBearerTokenProvider","OpenAI","MIDSCENE_API_TYPE","getAIConfigInBoolean","MIDSCENE_LANGSMITH_DEBUG","Error","console","wrapOpenAI","Anthropic","call","messages","options","completion","style","responseFormat","getResponseFormat","maxTokens","getAIConfig","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","isStreaming","content","accumulated","usage","timeCost","commonConfig","vlLocateMode","Number","stream","chunk","_chunk_choices__delta","_chunk_choices__delta1","_chunk_choices_2","reasoning_content","chunkData","undefined","estimatedTokens","Math","finalChunk","_result_usage","_result_usage1","_result_usage2","result","uiTarsModelVersion","JSON","convertImageContent","imgBase64","mimeType","body","parseBase64","m","Array","_chunk_delta","anthropicUsage","e","newError","AIActionType","assertSchema","locatorSchema","planSchema","AIResponseFormat","callToGetJSONObject","response","jsonContent","safeParseJson","callAiFnWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","cleanJsonString","_cleanJsonString_match","jsonrepair","jsonString"],"mappings":";;;;;;;;;;;;;;;AAiCA,eAAeA,iBAAiB,EAC9BC,iBAAiB,EACjBC,gBAAgB,EAIjB;IAMC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,wBAAwB,EACxBC,cAAc,EACdC,gBAAgB,EAChBC,cAAc,EACdC,mBAAmB,EACnBC,qBAAqB,EACrBC,qBAAqB,EACrBC,gBAAgB,EAChBC,eAAe,EACfC,eAAe,EACfC,gBAAgB,EACjB,GAAGC,kBAAkBlB,kBAAkB;IAExC,IAAImB;IAEJ,IAAIC;IACJ,MAAMC,aAAaC,SAAS;IAC5B,IAAIpB,WAAW;QACbmB,WAAW,oBAAoBnB;QAC/BkB,aAAa,IAAIG,gBAAgBrB;IACnC,OAAO,IAAID,YAAY;QACrBoB,WAAW,qBAAqBpB;QAChCmB,aAAa,IAAII,gBAAgBvB;IACnC;IAEA,IAAIM,0BAEFY,SAAS,IAAIM,YAAY;QACvB,SAASrB;QACT,QAAQC;QACR,WAAWe;QACX,GAAGd,iBAAiB;QACpB,yBAAyB;IAC3B;SACK,IAAIE,gBAAgB;QAGzB,IAAIkB;QACJ,IAAIjB,kBAAkB;YACpBkB,OACE,CAACC,aACD;YAEF,MAAMC,aAAa,IAAIC;YAEvBJ,gBAAgBK,uBAAuBF,YAAYpB;YAEnDU,SAAS,IAAIM,YAAY;gBACvB,sBAAsBC;gBACtB,UAAUf;gBACV,YAAYC;gBACZ,YAAYC;gBACZ,GAAGP,iBAAiB;gBACpB,GAAGQ,gBAAgB;YACrB;QACF,OAEEK,SAAS,IAAIM,YAAY;YACvB,QAAQf;YACR,UAAUC;YACV,YAAYC;YACZ,YAAYC;YACZ,yBAAyB;YACzB,GAAGP,iBAAiB;YACpB,GAAGQ,gBAAgB;QACrB;IAEJ,OAAO,IAAI,CAACC,iBACVI,SAAS,IAAIa,SAAO;QAClB,SAAS5B;QACT,QAAQC;QACR,WAAWe;QACX,GAAGd,iBAAiB;QACpB,gBAAgB;YACd,GAAIA,AAAAA,CAAAA,QAAAA,oBAAAA,KAAAA,IAAAA,kBAAmB,cAAc,AAAD,KAAK,CAAC,CAAC;YAC3C,CAAC2B,kBAAkB,EAAElC,kBAAkB,QAAQ;QACjD;QACA,yBAAyB;IAC3B;IAGF,IAAIoB,UAAUe,qBAAqBC,2BAA2B;QAC5D,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;QACPhB;QACAc;IACF;IAIF,IAAIF,iBACFI,SAAS,IAAIoB,UAAU;QACrB,QAAQvB;QACR,WAAWI;QACX,yBAAyB;IAC3B;IAGF,IAAI,AAAkB,WAAXD,UAA2BA,OAAe,QAAQ,EAC3D,OAAO;QACL,YAAaA,OAAe,QAAQ;QACpC,OAAO;QACPhB;QACAc;IACF;IAGF,MAAM,IAAImB,MAAM;AAClB;AAEO,eAAeI,KACpBC,QAAsC,EACtC1C,iBAA+B,EAC/BC,gBAAmC,EACnC0C,OAGC;IAED,MAAM,EAAEC,UAAU,EAAEC,KAAK,EAAEzC,SAAS,EAAEc,gBAAgB,EAAE,GACtD,MAAMnB,iBAAiB;QACrBC;QACAC;IACF;IAEF,MAAM6C,iBAAiBC,kBAAkB3C,WAAWJ;IAEpD,MAAMgD,YAAYC,YAAYC;IAC9B,MAAMC,YAAY5B,SAAS;IAC3B,MAAM6B,oBAAoB7B,SAAS;IACnC,MAAM8B,qBAAqB9B,SAAS;IAEpC,MAAM+B,YAAYC,KAAK,GAAG;IAE1B,MAAMC,cAAcb,AAAAA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,MAAM,AAAD,KAAKA,CAAAA,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO,AAAD;IACtD,IAAIc;IACJ,IAAIC,cAAc;IAClB,IAAIC;IACJ,IAAIC;IAEJ,MAAMC,eAAe;QACnB,aAAaC,AAAmC,kBAAnCA,aAAa7D,oBAAsC,MAAM;QACtE,QAAQ,CAAC,CAACuD;QACV,YACE,AAAqB,YAArB,OAAOR,YACHA,YACAe,OAAO,QAAQ,CAACf,aAAa,QAAQ;QAC3C,GAAIc,AAAmC,cAAnCA,aAAa7D,oBACb;YACE,2BAA2B;QAC7B,IACA,CAAC,CAAC;IACR;IAEA,IAAI;QACF,IAAI4C,AAAU,aAAVA,OAAoB;YACtBM,UACE,CAAC,QAAQ,EAAEK,cAAc,eAAe,GAAG,WAAW,EAAEpD,WAAW;YAGrE,IAAIoD,aAAa;gBACf,MAAMQ,SAAU,MAAMpB,WAAW,MAAM,CACrC;oBACE,OAAOxC;oBACPsC;oBACA,iBAAiBI;oBACjB,GAAGe,YAAY;gBACjB,GACA;oBACE,QAAQ;gBACV;gBAKF,WAAW,MAAMI,SAASD,OAAQ;wBAChBE,uBAAAA,iBAAAA,gBAEbC,wBAAAA,kBAAAA,iBAoBCC,kBAAAA;oBAtBJ,MAAMX,UAAUS,AAAAA,SAAAA,CAAAA,iBAAAA,MAAM,OAAO,AAAD,IAAZA,KAAAA,IAAAA,QAAAA,CAAAA,kBAAAA,cAAe,CAAC,EAAE,AAAD,IAAjBA,KAAAA,IAAAA,QAAAA,CAAAA,wBAAAA,gBAAoB,KAAK,AAAD,IAAxBA,KAAAA,IAAAA,sBAA2B,OAAO,AAAD,KAAK;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,EACbN,QAAQM,MAAM,KAAK;oBAGrB,IAAIR,WAAWY,mBAAmB;wBAChCX,eAAeD;wBACf,MAAMa,YAAiC;4BACrCb;4BACAY;4BACAX;4BACA,YAAY;4BACZ,OAAOa;wBACT;wBACA5B,QAAQ,OAAO,CAAE2B;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;wBACrCR,WAAWL,KAAK,GAAG,KAAKD;wBAGxB,IAAI,CAACK,OAAO;4BAEV,MAAMa,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAACf,YAAY,MAAM,GAAG;4BAElCC,QAAQ;gCACN,eAAea;gCACf,mBAAmBA;gCACnB,cAAcA,AAAkB,IAAlBA;4BAChB;wBACF;wBAGA,MAAME,aAAkC;4BACtC,SAAS;4BACThB;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,iBAAiB,MAAM;4BACjC;wBACF;wBACA0C,QAAQ,OAAO,CAAE+B;wBACjB;oBACF;gBACF;gBACAjB,UAAUC;gBACVN,kBACE,CAAC,iBAAiB,EAAEhD,UAAU,QAAQ,EAAE0D,aAAa7D,qBAAqB,UAAU,WAAW,EAAE2D,UAAU;YAE/G,OAAO;oBAUoJe,eAAyDC,gBAAwDC;gBAT1Q,MAAMC,SAAS,MAAMlC,WAAW,MAAM,CAAC;oBACrC,OAAOxC;oBACPsC;oBACA,iBAAiBI;oBACjB,GAAGe,YAAY;gBACjB;gBACAD,WAAWL,KAAK,GAAG,KAAKD;gBAExBF,kBACE,CAAC,OAAO,EAAEhD,UAAU,QAAQ,EAAE0D,aAAa7D,qBAAqB,UAAU,mBAAmB,EAAE8E,mBAAmB9E,kBAAkB,iBAAiB,EAAE0E,AAAAA,SAAAA,CAAAA,gBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,cAAc,aAAa,AAAD,KAAK,GAAG,qBAAqB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,iBAAiB,AAAD,KAAK,GAAG,gBAAgB,EAAEC,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,KAAK,AAAD,IAAXA,KAAAA,IAAAA,eAAc,YAAY,AAAD,KAAK,GAAG,WAAW,EAAEjB,SAAS,aAAa,EAAEkB,OAAO,WAAW,IAAI,IAAI;gBAG1WzB,mBACE,CAAC,oBAAoB,EAAE2B,KAAK,SAAS,CAACF,OAAO,KAAK,GAAG;gBAGvDlD,OACEkD,OAAO,OAAO,EACd,CAAC,mCAAmC,EAAEE,KAAK,SAAS,CAACF,SAAS;gBAEhErB,UAAUqB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;gBAC3CnB,QAAQmB,OAAO,KAAK;YACtB;YAEA3B,UAAU,CAAC,UAAU,EAAEM,SAAS;YAChC7B,OAAO6B,SAAS;QAClB,OAAO,IAAIZ,AAAU,gBAAVA,OAAuB;YAChC,MAAMoC,sBAAsB,CAACxB;gBAC3B,IAAIA,AAAiB,gBAAjBA,QAAQ,IAAI,EAAkB;oBAChC,MAAMyB,YAAYzB,QAAQ,SAAS,CAAC,GAAG;oBACvC7B,OAAOsD,WAAW;oBAClB,MAAM,EAAEC,QAAQ,EAAEC,IAAI,EAAE,GAAGC,YAAY5B,QAAQ,SAAS,CAAC,GAAG;oBAC5D,OAAO;wBACL,QAAQ;4BACN,MAAM;4BACN,YAAY0B;4BACZ,MAAMC;wBACR;wBACA,MAAM;oBACR;gBACF;gBACA,OAAO3B;YACT;YAEA,IAAID,aAAa;gBACf,MAAMQ,SAAU,MAAMpB,WAAW,MAAM,CAAC;oBACtC,OAAOxC;oBACP,QAAQ;oBACR,UAAUsC,SAAS,GAAG,CAAC,CAAC4C,IAAO;4BAC7B,MAAM;4BACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;wBACf;oBACA,iBAAiBxC;oBACjB,GAAGe,YAAY;gBACjB;gBAEA,WAAW,MAAMI,SAASD,OAAQ;wBAChBwB;oBAAhB,MAAM/B,UAAU+B,AAAAA,SAAAA,CAAAA,eAAAA,MAAM,KAAK,AAAD,IAAVA,KAAAA,IAAAA,aAAa,IAAI,AAAD,KAAK;oBACrC,IAAI/B,SAAS;wBACXC,eAAeD;wBACf,MAAMa,YAAiC;4BACrCb;4BACAC;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAOa;wBACT;wBACA5B,QAAQ,OAAO,CAAE2B;oBACnB;oBAGA,IAAIL,AAAe,mBAAfA,MAAM,IAAI,EAAqB;wBACjCL,WAAWL,KAAK,GAAG,KAAKD;wBACxB,MAAMmC,iBAAiBxB,MAAM,KAAK;wBAGlC,MAAMS,aAAkC;4BACtC,SAAS;4BACThB;4BACA,mBAAmB;4BACnB,YAAY;4BACZ,OAAO+B,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,WAAW7B,YAAY;gCACvB,YAAYxD;gCACZ,mBAAmBc;gCACnB,QAAQjB,iBAAiB,MAAM;4BACjC,IACAsE;wBACN;wBACA5B,QAAQ,OAAO,CAAE+B;wBACjB;oBACF;gBACF;gBACAjB,UAAUC;YACZ,OAAO;gBACL,MAAMoB,SAAS,MAAMlC,WAAW,MAAM,CAAC;oBACrC,OAAOxC;oBACP,QAAQ;oBACR,UAAUsC,SAAS,GAAG,CAAC,CAAC4C,IAAO;4BAC7B,MAAM;4BACN,SAASC,MAAM,OAAO,CAACD,EAAE,OAAO,IAC3BA,EAAE,OAAO,CAAS,GAAG,CAACL,uBACvBK,EAAE,OAAO;wBACf;oBACA,iBAAiBxC;oBACjB,GAAGe,YAAY;gBACjB;gBACAD,WAAWL,KAAK,GAAG,KAAKD;gBACxBG,UAAWqB,OAAe,OAAO,CAAC,EAAE,CAAC,IAAI;gBACzCnB,QAAQmB,OAAO,KAAK;YACtB;YAEAlD,OAAO6B,SAAS;QAClB;QAEA,IAAID,eAAe,CAACG,OAAO;YAEzB,MAAMa,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEhB,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;YAEtCE,QAAQ;gBACN,eAAea;gBACf,mBAAmBA;gBACnB,cAAcA,AAAkB,IAAlBA;YAChB;QACF;QAEA,OAAO;YACL,SAASf,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,iBAAiB,MAAM;YACjC,IACAsE;YACJ,YAAY,CAAC,CAACf;QAChB;IACF,EAAE,OAAOkC,GAAQ;QACfpD,QAAQ,KAAK,CAAC,kBAAkBoD;QAChC,MAAMC,WAAW,IAAItD,MACnB,CAAC,eAAe,EAAEmB,cAAc,eAAe,GAAG,kBAAkB,EAAEkC,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC/I;YACE,OAAOA;QACT;QAEF,MAAMC;IACR;AACF;AAEO,MAAM5C,oBAAoB,CAC/B3C,WACAJ;IAIA,IAAI8C;IAKJ,IAAI1C,UAAU,QAAQ,CAAC,UACrB,OAAQJ;QACN,KAAK4F,aAAa,MAAM;YACtB9C,iBAAiB+C;YACjB;QACF,KAAKD,aAAa,eAAe;YAC/B9C,iBAAiBgD;YACjB;QACF,KAAKF,aAAa,IAAI;YACpB9C,iBAAiBiD;YACjB;QACF,KAAKH,aAAa,YAAY;QAC9B,KAAKA,aAAa,gBAAgB;YAChC9C,iBAAiB;gBAAE,MAAMkD,iBAAiB,IAAI;YAAC;YAC/C;IACJ;IAIF,IAAI5F,AAAc,wBAAdA,WACF0C,iBAAiB;QAAE,MAAMkD,iBAAiB,IAAI;IAAC;IAGjD,OAAOlD;AACT;AAEO,eAAemD,oBACpBvD,QAAsC,EACtC1C,iBAA+B,EAC/BC,gBAAmC;IAEnC,MAAMiG,WAAW,MAAMzD,KAAKC,UAAU1C,mBAAmBC;IACzD2B,OAAOsE,UAAU;IACjB,MAAMC,cAAcC,cAAcF,SAAS,OAAO,EAAEjG;IACpD,OAAO;QAAE,SAASkG;QAAa,OAAOD,SAAS,KAAK;IAAC;AACvD;AAEO,eAAeG,2BACpBC,IAAY,EACZtG,iBAA+B,EAC/BC,gBAAmC;IAEnC,MAAM,EAAEwD,OAAO,EAAEE,KAAK,EAAE,GAAG,MAAMlB,KAC/B6D,MACAtG,mBACAC;IAEF,OAAO;QAAEwD;QAASE;IAAM;AAC1B;AAEO,SAAS4C,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,cACdQ,KAAa,EACb3G,gBAAmC;IAEnC,MAAM4G,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,CAAC/C;IACT;IACA,IAAI;QACF,OAAOiB,KAAK,KAAK,CAAC6B;IACpB,EAAE,OAAM,CAAC;IACT,IAAI;QACF,OAAO7B,KAAK,KAAK,CAAC+B,WAAWF;IAC/B,EAAE,OAAOnB,GAAG,CAAC;IAEb,IACE5B,AAAmC,oBAAnCA,aAAa7D,qBACb6D,AAAmC,kBAAnCA,aAAa7D,mBACb;QACA,MAAM+G,aAAaL,yBAAyBE;QAC5C,OAAO7B,KAAK,KAAK,CAAC+B,WAAWC;IAC/B;IACA,MAAM3E,MAAM,CAAC,+BAA+B,EAAEuE,OAAO;AACvD"}

package/dist/es/ai-model/ui-tars-planning.mjs

@@ -39,33 +39,23 @@ async function vlmPlanning(options) {
         },
         modelVer: modelVer || void 0
     });
-    debug('modelVer', modelVer, 'parsed', JSON.stringify(parsed));
+    debug('ui-tars modelVer', modelVer, ', parsed', JSON.stringify(parsed));
     const transformActions = [];
     parsed.forEach((action)=>{
         if ('click' === action.action_type) {
             assert(action.action_inputs.start_box, 'start_box is required');
             const point = getPoint(action.action_inputs.start_box, size);
-            transformActions.push({
-                type: 'Locate',
-                param: {},
-                locate: {
-                    prompt: action.thought || '',
-                    bbox: pointToBbox({
-                        x: point[0],
-                        y: point[1]
-                    }, size.width, size.height)
-                }
-            });
             transformActions.push({
                 type: 'Tap',
-                locate: {
-                    prompt: action.thought || '',
-                    bbox: pointToBbox({
-                        x: point[0],
-                        y: point[1]
-                    }, size.width, size.height)
-                },
-                param: action.thought || ''
+                param: {
+                    locate: {
+                        prompt: action.thought || '',
+                        bbox: pointToBbox({
+                            x: point[0],
+                            y: point[1]
+                        }, size.width, size.height)
+                    }
+                }
             });
         } else if ('drag' === action.action_type) {
             assert(action.action_inputs.start_box, 'start_box is required');
@@ -73,18 +63,23 @@ async function vlmPlanning(options) {
             const startPoint = getPoint(action.action_inputs.start_box, size);
             const endPoint = getPoint(action.action_inputs.end_box, size);
             transformActions.push({
-                type: 'Drag',
+                type: 'DragAndDrop',
                 param: {
-                    start_box: {
-                        x: startPoint[0],
-                        y: startPoint[1]
+                    from: {
+                        prompt: action.thought || '',
+                        bbox: pointToBbox({
+                            x: startPoint[0],
+                            y: startPoint[1]
+                        }, size.width, size.height)
                     },
-                    end_box: {
-                        x: endPoint[0],
-                        y: endPoint[1]
+                    to: {
+                        prompt: action.thought || '',
+                        bbox: pointToBbox({
+                            x: endPoint[0],
+                            y: endPoint[1]
+                        }, size.width, size.height)
                     }
                 },
-                locate: null,
                 thought: action.thought || ''
             });
         } else if ('type' === action.action_type) transformActions.push({
@@ -92,7 +87,6 @@ async function vlmPlanning(options) {
             param: {
                 value: action.action_inputs.content
             },
-            locate: null,
             thought: action.thought || ''
         });
         else if ('scroll' === action.action_type) transformActions.push({
@@ -100,13 +94,11 @@ async function vlmPlanning(options) {
             param: {
                 direction: action.action_inputs.direction
             },
-            locate: null,
             thought: action.thought || ''
         });
         else if ('finished' === action.action_type) transformActions.push({
             type: 'Finished',
             param: {},
-            locate: null,
             thought: action.thought || ''
         });
         else if ('hotkey' === action.action_type) if (action.action_inputs.key) {
@@ -114,9 +106,8 @@ async function vlmPlanning(options) {
             transformActions.push({
                 type: 'KeyboardPress',
                 param: {
-                    value: keys
+                    keyName: keys
                 },
-                locate: null,
                 thought: action.thought || ''
             });
         } else console.warn('No key found in action: hotkey. Will not perform action.');
@@ -125,56 +116,8 @@ async function vlmPlanning(options) {
             param: {
                 timeMs: 1000
             },
-            locate: null,
-            thought: action.thought || ''
-        });
-        else if ('androidBackButton' === action.action_type) transformActions.push({
-            type: 'AndroidBackButton',
-            param: {},
-            locate: null,
-            thought: action.thought || ''
-        });
-        else if ('androidHomeButton' === action.action_type) transformActions.push({
-            type: 'AndroidHomeButton',
-            param: {},
-            locate: null,
             thought: action.thought || ''
         });
-        else if ('androidRecentAppsButton' === action.action_type) transformActions.push({
-            type: 'AndroidRecentAppsButton',
-            param: {}
-        });
-        else if ('androidLongPress' === action.action_type) {
-            assert(action.action_inputs.start_coords, 'start_coords is required for androidLongPress');
-            const point = action.action_inputs.start_coords;
-            transformActions.push({
-                type: 'AndroidLongPress',
-                param: {
-                    x: point[0],
-                    y: point[1],
-                    duration: 1000
-                },
-                locate: null,
-                thought: action.thought || ''
-            });
-        } else if ('androidPull' === action.action_type) {
-            const pullDirection = action.action_inputs.direction || 'down';
-            const startPoint = action.action_inputs.start_coords ? {
-                x: action.action_inputs.start_coords[0],
-                y: action.action_inputs.start_coords[1]
-            } : void 0;
-            transformActions.push({
-                type: 'AndroidPull',
-                param: {
-                    direction: pullDirection,
-                    startPoint,
-                    distance: action.action_inputs.distance,
-                    duration: action.action_inputs.duration || 500
-                },
-                locate: null,
-                thought: action.thought || ''
-            });
-        }
     });
     if (0 === transformActions.length) throw new Error(`No actions found, response: ${res.content}`, {
         cause: {
@@ -182,6 +125,7 @@ async function vlmPlanning(options) {
             parsed
         }
     });
+    debug('transformActions', JSON.stringify(transformActions, null, 2));
     return {
         actions: transformActions,
         actionsFromModel: parsed,

package/dist/es/ai-model/ui-tars-planning.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/ui-tars-planning.mjs","sources":["webpack://@midscene/core/./src/ai-model/ui-tars-planning.ts"],"sourcesContent":["import type {\n  AIUsageInfo,\n  MidsceneYamlFlowItem,\n  PlanningAction,\n  Size,\n} from '@/types';\nimport {\n  type IModelPreferences,\n  UITarsModelVersion,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { resizeImgBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { transformHotkeyInput } from '@midscene/shared/us-keyboard-layout';\nimport { assert } from '@midscene/shared/utils';\nimport { actionParser } from '@ui-tars/action-parser';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport { AIActionType } from './common';\nimport { getSummary, getUiTarsPlanningPrompt } from './prompt/ui-tars-planning';\nimport { call } from './service-caller/index';\ntype ActionType =\n  | 'click'\n  | 'drag'\n  | 'type'\n  | 'hotkey'\n  | 'finished'\n  | 'scroll'\n  | 'wait'\n  | 'androidBackButton'\n  | 'androidHomeButton'\n  | 'androidRecentAppsButton'\n  | 'androidLongPress'\n  | 'androidPull';\n\nconst debug = getDebug('ui-tars-planning');\nconst bboxSize = 10;\nconst pointToBbox = (\n  point: { x: number; y: number },\n  width: number,\n  height: number,\n): [number, number, number, number] => {\n  return [\n    Math.round(Math.max(point.x - bboxSize / 2, 0)),\n    Math.round(Math.max(point.y - bboxSize / 2, 0)),\n    Math.round(Math.min(point.x + bboxSize / 2, width)),\n    Math.round(Math.min(point.y + bboxSize / 2, height)),\n  ];\n};\n\nexport async function vlmPlanning(options: {\n  userInstruction: string;\n  conversationHistory: ChatCompletionMessageParam[];\n  size: { width: number; height: number };\n  modelPreferences: IModelPreferences;\n}): Promise<{\n  actions: PlanningAction<any>[];\n  actionsFromModel: ReturnType<typeof actionParser>['parsed'];\n  action_summary: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n}> {\n  const { conversationHistory, userInstruction, size, modelPreferences } =\n    options;\n  const systemPrompt = getUiTarsPlanningPrompt() + userInstruction;\n\n  const res = await call(\n    [\n      {\n        role: 'user',\n        content: systemPrompt,\n      },\n      ...conversationHistory,\n    ],\n    AIActionType.INSPECT_ELEMENT,\n    modelPreferences,\n  );\n  const convertedText = convertBboxToCoordinates(res.content);\n\n  const modelVer = uiTarsModelVersion(modelPreferences);\n\n  const { parsed } = actionParser({\n    prediction: convertedText,\n    factor: [1000, 1000],\n    screenContext: {\n      width: size.width,\n      height: size.height,\n    },\n    modelVer: modelVer || undefined,\n  });\n\n  debug('modelVer', modelVer, 'parsed', JSON.stringify(parsed));\n\n  const transformActions: PlanningAction[] = [];\n  parsed.forEach((action) => {\n    if (action.action_type === 'click') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n      transformActions.push({\n        type: 'Locate',\n        param: {},\n        locate: {\n          prompt: action.thought || '',\n          bbox: pointToBbox(\n            { x: point[0], y: point[1] },\n            size.width,\n            size.height,\n          ),\n        },\n      });\n      transformActions.push({\n        type: 'Tap',\n        locate: {\n          prompt: action.thought || '',\n          bbox: pointToBbox(\n            { x: point[0], y: point[1] },\n            size.width,\n            size.height,\n          ),\n        },\n        param: action.thought || '',\n      });\n    } else if (action.action_type === 'drag') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      assert(action.action_inputs.end_box, 'end_box is required');\n      const startPoint = getPoint(action.action_inputs.start_box, size);\n      const endPoint = getPoint(action.action_inputs.end_box, size);\n      transformActions.push({\n        type: 'Drag',\n        param: {\n          start_box: { x: startPoint[0], y: startPoint[1] },\n          end_box: { x: endPoint[0], y: endPoint[1] },\n        },\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'type') {\n      transformActions.push({\n        type: 'Input',\n        param: {\n          value: action.action_inputs.content,\n        },\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'scroll') {\n      transformActions.push({\n        type: 'Scroll',\n        param: {\n          direction: action.action_inputs.direction,\n        },\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'finished') {\n      transformActions.push({\n        type: 'Finished',\n        param: {},\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'hotkey') {\n      if (!action.action_inputs.key) {\n        console.warn(\n          'No key found in action: hotkey. Will not perform action.',\n        );\n      } else {\n        const keys = transformHotkeyInput(action.action_inputs.key);\n\n        transformActions.push({\n          type: 'KeyboardPress',\n          param: {\n            value: keys,\n          },\n          locate: null,\n          thought: action.thought || '',\n        });\n      }\n    } else if (action.action_type === 'wait') {\n      transformActions.push({\n        type: 'Sleep',\n        param: {\n          timeMs: 1000,\n        },\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'androidBackButton') {\n      transformActions.push({\n        type: 'AndroidBackButton',\n        param: {},\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'androidHomeButton') {\n      transformActions.push({\n        type: 'AndroidHomeButton',\n        param: {},\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'androidRecentAppsButton') {\n      transformActions.push({\n        type: 'AndroidRecentAppsButton',\n        param: {},\n      });\n    } else if (action.action_type === 'androidLongPress') {\n      assert(\n        action.action_inputs.start_coords,\n        'start_coords is required for androidLongPress',\n      );\n      const point = action.action_inputs.start_coords;\n      transformActions.push({\n        type: 'AndroidLongPress',\n        param: {\n          x: point[0],\n          y: point[1],\n          duration: 1000,\n        },\n        locate: null,\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'androidPull') {\n      const pullDirection = action.action_inputs.direction || 'down';\n      const startPoint = action.action_inputs.start_coords\n        ? {\n            x: action.action_inputs.start_coords[0],\n            y: action.action_inputs.start_coords[1],\n          }\n        : undefined;\n\n      transformActions.push({\n        type: 'AndroidPull',\n        param: {\n          direction: pullDirection as 'up' | 'down',\n          startPoint,\n          distance: (action.action_inputs as any).distance,\n          duration: (action.action_inputs as any).duration || 500,\n        },\n        locate: null,\n        thought: action.thought || '',\n      });\n    }\n  });\n\n  if (transformActions.length === 0) {\n    throw new Error(`No actions found, response: ${res.content}`, {\n      cause: {\n        prediction: res.content,\n        parsed,\n      },\n    });\n  }\n\n  return {\n    actions: transformActions,\n    actionsFromModel: parsed,\n    action_summary: getSummary(res.content),\n    usage: res.usage,\n    rawResponse: JSON.stringify(res.content, undefined, 2),\n  };\n}\n\n/**\n * Converts bounding box notation to coordinate points\n * @param text - The text containing bbox tags to be converted\n * @returns The text with bbox tags replaced by coordinate points\n */\nfunction convertBboxToCoordinates(text: string): string {\n  // Match the four numbers after <bbox>\n  const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n  function replaceMatch(\n    match: string,\n    x1: string,\n    y1: string,\n    x2: string,\n    y2: string,\n  ): string {\n    // Convert strings to numbers and calculate center point\n    const x1Num = Number.parseInt(x1, 10);\n    const y1Num = Number.parseInt(y1, 10);\n    const x2Num = Number.parseInt(x2, 10);\n    const y2Num = Number.parseInt(y2, 10);\n\n    // Use Math.floor to truncate and calculate center point\n    const x = Math.floor((x1Num + x2Num) / 2);\n    const y = Math.floor((y1Num + y2Num) / 2);\n\n    // Return formatted coordinate string\n    return `(${x},${y})`;\n  }\n\n  // Remove [EOS] and replace <bbox> coordinates\n  const cleanedText = text.replace(/\\[EOS\\]/g, '');\n  return cleanedText.replace(pattern, replaceMatch).trim();\n}\n\nfunction getPoint(startBox: string, size: { width: number; height: number }) {\n  const [x, y] = JSON.parse(startBox);\n  return [x * size.width, y * size.height];\n}\n\ninterface BaseAction {\n  action_type: ActionType;\n  action_inputs: Record<string, any>;\n  reflection: string | null;\n  thought: string | null;\n}\n\ninterface ClickAction extends BaseAction {\n  action_type: 'click';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface DragAction extends BaseAction {\n  action_type: 'drag';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n    end_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface WaitAction extends BaseAction {\n  action_type: 'wait';\n  action_inputs: {\n    time: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface TypeAction extends BaseAction {\n  action_type: 'type';\n  action_inputs: {\n    content: string;\n  };\n}\n\ninterface HotkeyAction extends BaseAction {\n  action_type: 'hotkey';\n  action_inputs: {\n    key: string;\n  };\n}\n\ninterface ScrollAction extends BaseAction {\n  action_type: 'scroll';\n  action_inputs: {\n    direction: 'up' | 'down';\n  };\n}\n\ninterface FinishedAction extends BaseAction {\n  action_type: 'finished';\n  action_inputs: Record<string, never>;\n}\n\ninterface AndroidLongPressAction extends BaseAction {\n  action_type: 'androidLongPress';\n  action_inputs: {\n    start_coords: [number, number]; // Coordinates for long press\n    duration?: number; // Duration in milliseconds\n  };\n}\n\nexport type Action =\n  | ClickAction\n  | DragAction\n  | TypeAction\n  | HotkeyAction\n  | ScrollAction\n  | FinishedAction\n  | WaitAction\n  | AndroidLongPressAction;\n\nexport async function resizeImageForUiTars(\n  imageBase64: string,\n  size: Size,\n  modelPreferences: IModelPreferences,\n) {\n  if (\n    vlLocateMode(modelPreferences) === 'vlm-ui-tars' &&\n    uiTarsModelVersion(modelPreferences) === UITarsModelVersion.V1_5\n  ) {\n    debug('ui-tars-v1.5, will check image size', size);\n    const currentPixels = size.width * size.height;\n    const maxPixels = 16384 * 28 * 28; //\n    if (currentPixels > maxPixels) {\n      const resizeFactor = Math.sqrt(maxPixels / currentPixels);\n      const newWidth = Math.floor(size.width * resizeFactor);\n      const newHeight = Math.floor(size.height * resizeFactor);\n      debug(\n        'resize image for ui-tars, new width: %s, new height: %s',\n        newWidth,\n        newHeight,\n      );\n      const resizedImage = await resizeImgBase64(imageBase64, {\n        width: newWidth,\n        height: newHeight,\n      });\n      return resizedImage;\n    }\n  }\n  return imageBase64;\n}\n"],"names":["debug","getDebug","bboxSize","pointToBbox","point","width","height","Math","vlmPlanning","options","conversationHistory","userInstruction","size","modelPreferences","systemPrompt","getUiTarsPlanningPrompt","res","call","AIActionType","convertedText","convertBboxToCoordinates","modelVer","uiTarsModelVersion","parsed","actionParser","undefined","JSON","transformActions","action","assert","getPoint","startPoint","endPoint","keys","transformHotkeyInput","console","pullDirection","Error","getSummary","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","y","cleanedText","startBox","resizeImageForUiTars","imageBase64","vlLocateMode","UITarsModelVersion","currentPixels","maxPixels","resizeFactor","newWidth","newHeight","resizedImage","resizeImgBase64"],"mappings":";;;;;;;;;AAmCA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,WAAW;AACjB,MAAMC,cAAc,CAClBC,OACAC,OACAC,SAEO;QACLC,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGG;QAC5CE,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGI;KAC7C;AAGI,eAAeE,YAAYC,OAKjC;IAQC,MAAM,EAAEC,mBAAmB,EAAEC,eAAe,EAAEC,IAAI,EAAEC,gBAAgB,EAAE,GACpEJ;IACF,MAAMK,eAAeC,4BAA4BJ;IAEjD,MAAMK,MAAM,MAAMC,KAChB;QACE;YACE,MAAM;YACN,SAASH;QACX;WACGJ;KACJ,EACDQ,aAAa,eAAe,EAC5BL;IAEF,MAAMM,gBAAgBC,yBAAyBJ,IAAI,OAAO;IAE1D,MAAMK,WAAWC,mBAAmBT;IAEpC,MAAM,EAAEU,MAAM,EAAE,GAAGC,aAAa;QAC9B,YAAYL;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOP,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUS,YAAYI;IACxB;IAEAzB,MAAM,YAAYqB,UAAU,UAAUK,KAAK,SAAS,CAACH;IAErD,MAAMI,mBAAqC,EAAE;IAC7CJ,OAAO,OAAO,CAAC,CAACK;QACd,IAAIA,AAAuB,YAAvBA,OAAO,WAAW,EAAc;YAClCC,OAAOD,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAMxB,QAAQ0B,SAASF,OAAO,aAAa,CAAC,SAAS,EAAEhB;YACvDe,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO,CAAC;gBACR,QAAQ;oBACN,QAAQC,OAAO,OAAO,IAAI;oBAC1B,MAAMzB,YACJ;wBAAE,GAAGC,KAAK,CAAC,EAAE;wBAAE,GAAGA,KAAK,CAAC,EAAE;oBAAC,GAC3BQ,KAAK,KAAK,EACVA,KAAK,MAAM;gBAEf;YACF;YACAe,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,QAAQ;oBACN,QAAQC,OAAO,OAAO,IAAI;oBAC1B,MAAMzB,YACJ;wBAAE,GAAGC,KAAK,CAAC,EAAE;wBAAE,GAAGA,KAAK,CAAC,EAAE;oBAAC,GAC3BQ,KAAK,KAAK,EACVA,KAAK,MAAM;gBAEf;gBACA,OAAOgB,OAAO,OAAO,IAAI;YAC3B;QACF,OAAO,IAAIA,AAAuB,WAAvBA,OAAO,WAAW,EAAa;YACxCC,OAAOD,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCC,OAAOD,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMG,aAAaD,SAASF,OAAO,aAAa,CAAC,SAAS,EAAEhB;YAC5D,MAAMoB,WAAWF,SAASF,OAAO,aAAa,CAAC,OAAO,EAAEhB;YACxDe,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,WAAW;wBAAE,GAAGI,UAAU,CAAC,EAAE;wBAAE,GAAGA,UAAU,CAAC,EAAE;oBAAC;oBAChD,SAAS;wBAAE,GAAGC,QAAQ,CAAC,EAAE;wBAAE,GAAGA,QAAQ,CAAC,EAAE;oBAAC;gBAC5C;gBACA,QAAQ;gBACR,SAASJ,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIA,AAAuB,WAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,OAAOC,OAAO,aAAa,CAAC,OAAO;YACrC;YACA,QAAQ;YACR,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,aAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWC,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,QAAQ;YACR,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,eAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO,CAAC;YACR,QAAQ;YACR,SAASC,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,aAAvBA,OAAO,WAAW,EAC3B,IAAKA,OAAO,aAAa,CAAC,GAAG,EAItB;YACL,MAAMK,OAAOC,qBAAqBN,OAAO,aAAa,CAAC,GAAG;YAE1DD,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,OAAOM;gBACT;gBACA,QAAQ;gBACR,SAASL,OAAO,OAAO,IAAI;YAC7B;QACF,OAdEO,QAAQ,IAAI,CACV;aAcC,IAAIP,AAAuB,WAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,QAAQ;YACR,SAASC,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,wBAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO,CAAC;YACR,QAAQ;YACR,SAASC,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,wBAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO,CAAC;YACR,QAAQ;YACR,SAASC,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,8BAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO,CAAC;QACV;aACK,IAAIC,AAAuB,uBAAvBA,OAAO,WAAW,EAAyB;YACpDC,OACED,OAAO,aAAa,CAAC,YAAY,EACjC;YAEF,MAAMxB,QAAQwB,OAAO,aAAa,CAAC,YAAY;YAC/CD,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,GAAGvB,KAAK,CAAC,EAAE;oBACX,GAAGA,KAAK,CAAC,EAAE;oBACX,UAAU;gBACZ;gBACA,QAAQ;gBACR,SAASwB,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIA,AAAuB,kBAAvBA,OAAO,WAAW,EAAoB;YAC/C,MAAMQ,gBAAgBR,OAAO,aAAa,CAAC,SAAS,IAAI;YACxD,MAAMG,aAAaH,OAAO,aAAa,CAAC,YAAY,GAChD;gBACE,GAAGA,OAAO,aAAa,CAAC,YAAY,CAAC,EAAE;gBACvC,GAAGA,OAAO,aAAa,CAAC,YAAY,CAAC,EAAE;YACzC,IACAH;YAEJE,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,WAAWS;oBACXL;oBACA,UAAWH,OAAO,aAAa,CAAS,QAAQ;oBAChD,UAAWA,OAAO,aAAa,CAAS,QAAQ,IAAI;gBACtD;gBACA,QAAQ;gBACR,SAASA,OAAO,OAAO,IAAI;YAC7B;QACF;IACF;IAEA,IAAID,AAA4B,MAA5BA,iBAAiB,MAAM,EACzB,MAAM,IAAIU,MAAM,CAAC,4BAA4B,EAAErB,IAAI,OAAO,EAAE,EAAE;QAC5D,OAAO;YACL,YAAYA,IAAI,OAAO;YACvBO;QACF;IACF;IAGF,OAAO;QACL,SAASI;QACT,kBAAkBJ;QAClB,gBAAgBe,WAAWtB,IAAI,OAAO;QACtC,OAAOA,IAAI,KAAK;QAChB,aAAaU,KAAK,SAAS,CAACV,IAAI,OAAO,EAAES,QAAW;IACtD;AACF;AAOA,SAASL,yBAAyBmB,IAAY;IAE5C,MAAMC,UAAU;IAEhB,SAASC,aACPC,KAAa,EACbC,EAAU,EACVC,EAAU,EACVC,EAAU,EACVC,EAAU;QAGV,MAAMC,QAAQC,OAAO,QAAQ,CAACL,IAAI;QAClC,MAAMM,QAAQD,OAAO,QAAQ,CAACJ,IAAI;QAClC,MAAMM,QAAQF,OAAO,QAAQ,CAACH,IAAI;QAClC,MAAMM,QAAQH,OAAO,QAAQ,CAACF,IAAI;QAGlC,MAAMM,IAAI7C,KAAK,KAAK,CAAEwC,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMG,IAAI9C,KAAK,KAAK,CAAE0C,AAAAA,CAAAA,QAAQE,KAAI,IAAK;QAGvC,OAAO,CAAC,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC,CAAC;IACtB;IAGA,MAAMC,cAAcf,KAAK,OAAO,CAAC,YAAY;IAC7C,OAAOe,YAAY,OAAO,CAACd,SAASC,cAAc,IAAI;AACxD;AAEA,SAASX,SAASyB,QAAgB,EAAE3C,IAAuC;IACzE,MAAM,CAACwC,GAAGC,EAAE,GAAG3B,KAAK,KAAK,CAAC6B;IAC1B,OAAO;QAACH,IAAIxC,KAAK,KAAK;QAAEyC,IAAIzC,KAAK,MAAM;KAAC;AAC1C;AA2EO,eAAe4C,qBACpBC,WAAmB,EACnB7C,IAAU,EACVC,gBAAmC;IAEnC,IACE6C,AAAmC,kBAAnCA,aAAa7C,qBACbS,mBAAmBT,sBAAsB8C,mBAAmB,IAAI,EAChE;QACA3D,MAAM,uCAAuCY;QAC7C,MAAMgD,gBAAgBhD,KAAK,KAAK,GAAGA,KAAK,MAAM;QAC9C,MAAMiD,YAAY;QAClB,IAAID,gBAAgBC,WAAW;YAC7B,MAAMC,eAAevD,KAAK,IAAI,CAACsD,YAAYD;YAC3C,MAAMG,WAAWxD,KAAK,KAAK,CAACK,KAAK,KAAK,GAAGkD;YACzC,MAAME,YAAYzD,KAAK,KAAK,CAACK,KAAK,MAAM,GAAGkD;YAC3C9D,MACE,2DACA+D,UACAC;YAEF,MAAMC,eAAe,MAAMC,gBAAgBT,aAAa;gBACtD,OAAOM;gBACP,QAAQC;YACV;YACA,OAAOC;QACT;IACF;IACA,OAAOR;AACT"}
+{"version":3,"file":"ai-model/ui-tars-planning.mjs","sources":["webpack://@midscene/core/./src/ai-model/ui-tars-planning.ts"],"sourcesContent":["import type {\n  AIUsageInfo,\n  MidsceneYamlFlowItem,\n  PlanningAction,\n  Size,\n} from '@/types';\nimport {\n  type IModelPreferences,\n  UITarsModelVersion,\n  uiTarsModelVersion,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { resizeImgBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { transformHotkeyInput } from '@midscene/shared/us-keyboard-layout';\nimport { assert } from '@midscene/shared/utils';\nimport { actionParser } from '@ui-tars/action-parser';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport { AIActionType } from './common';\nimport { getSummary, getUiTarsPlanningPrompt } from './prompt/ui-tars-planning';\nimport { call } from './service-caller/index';\ntype ActionType =\n  | 'click'\n  | 'drag'\n  | 'type'\n  | 'hotkey'\n  | 'finished'\n  | 'scroll'\n  | 'wait';\n\nconst debug = getDebug('ui-tars-planning');\nconst bboxSize = 10;\nconst pointToBbox = (\n  point: { x: number; y: number },\n  width: number,\n  height: number,\n): [number, number, number, number] => {\n  return [\n    Math.round(Math.max(point.x - bboxSize / 2, 0)),\n    Math.round(Math.max(point.y - bboxSize / 2, 0)),\n    Math.round(Math.min(point.x + bboxSize / 2, width)),\n    Math.round(Math.min(point.y + bboxSize / 2, height)),\n  ];\n};\n\nexport async function vlmPlanning(options: {\n  userInstruction: string;\n  conversationHistory: ChatCompletionMessageParam[];\n  size: { width: number; height: number };\n  modelPreferences: IModelPreferences;\n}): Promise<{\n  actions: PlanningAction<any>[];\n  actionsFromModel: ReturnType<typeof actionParser>['parsed'];\n  action_summary: string;\n  yamlFlow?: MidsceneYamlFlowItem[];\n  usage?: AIUsageInfo;\n  rawResponse?: string;\n}> {\n  const { conversationHistory, userInstruction, size, modelPreferences } =\n    options;\n  const systemPrompt = getUiTarsPlanningPrompt() + userInstruction;\n\n  const res = await call(\n    [\n      {\n        role: 'user',\n        content: systemPrompt,\n      },\n      ...conversationHistory,\n    ],\n    AIActionType.INSPECT_ELEMENT,\n    modelPreferences,\n  );\n  const convertedText = convertBboxToCoordinates(res.content);\n\n  const modelVer = uiTarsModelVersion(modelPreferences);\n\n  const { parsed } = actionParser({\n    prediction: convertedText,\n    factor: [1000, 1000],\n    screenContext: {\n      width: size.width,\n      height: size.height,\n    },\n    modelVer: modelVer || undefined,\n  });\n\n  debug('ui-tars modelVer', modelVer, ', parsed', JSON.stringify(parsed));\n\n  const transformActions: PlanningAction[] = [];\n  parsed.forEach((action) => {\n    if (action.action_type === 'click') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n      transformActions.push({\n        type: 'Tap',\n        param: {\n          locate: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: point[0], y: point[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n      });\n    } else if (action.action_type === 'drag') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      assert(action.action_inputs.end_box, 'end_box is required');\n      const startPoint = getPoint(action.action_inputs.start_box, size);\n      const endPoint = getPoint(action.action_inputs.end_box, size);\n      transformActions.push({\n        type: 'DragAndDrop',\n        param: {\n          from: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: startPoint[0], y: startPoint[1] },\n              size.width,\n              size.height,\n            ),\n          },\n          to: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: endPoint[0], y: endPoint[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'type') {\n      transformActions.push({\n        type: 'Input',\n        param: {\n          value: action.action_inputs.content,\n        },\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'scroll') {\n      transformActions.push({\n        type: 'Scroll',\n        param: {\n          direction: action.action_inputs.direction,\n        },\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'finished') {\n      transformActions.push({\n        type: 'Finished',\n        param: {},\n        thought: action.thought || '',\n      });\n    } else if (action.action_type === 'hotkey') {\n      if (!action.action_inputs.key) {\n        console.warn(\n          'No key found in action: hotkey. Will not perform action.',\n        );\n      } else {\n        const keys = transformHotkeyInput(action.action_inputs.key);\n\n        transformActions.push({\n          type: 'KeyboardPress',\n          param: {\n            keyName: keys,\n          },\n          thought: action.thought || '',\n        });\n      }\n    } else if (action.action_type === 'wait') {\n      transformActions.push({\n        type: 'Sleep',\n        param: {\n          timeMs: 1000,\n        },\n        thought: action.thought || '',\n      });\n    }\n  });\n\n  if (transformActions.length === 0) {\n    throw new Error(`No actions found, response: ${res.content}`, {\n      cause: {\n        prediction: res.content,\n        parsed,\n      },\n    });\n  }\n\n  debug('transformActions', JSON.stringify(transformActions, null, 2));\n\n  return {\n    actions: transformActions,\n    actionsFromModel: parsed,\n    action_summary: getSummary(res.content),\n    usage: res.usage,\n    rawResponse: JSON.stringify(res.content, undefined, 2),\n  };\n}\n\n/**\n * Converts bounding box notation to coordinate points\n * @param text - The text containing bbox tags to be converted\n * @returns The text with bbox tags replaced by coordinate points\n */\nfunction convertBboxToCoordinates(text: string): string {\n  // Match the four numbers after <bbox>\n  const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n  function replaceMatch(\n    match: string,\n    x1: string,\n    y1: string,\n    x2: string,\n    y2: string,\n  ): string {\n    // Convert strings to numbers and calculate center point\n    const x1Num = Number.parseInt(x1, 10);\n    const y1Num = Number.parseInt(y1, 10);\n    const x2Num = Number.parseInt(x2, 10);\n    const y2Num = Number.parseInt(y2, 10);\n\n    // Use Math.floor to truncate and calculate center point\n    const x = Math.floor((x1Num + x2Num) / 2);\n    const y = Math.floor((y1Num + y2Num) / 2);\n\n    // Return formatted coordinate string\n    return `(${x},${y})`;\n  }\n\n  // Remove [EOS] and replace <bbox> coordinates\n  const cleanedText = text.replace(/\\[EOS\\]/g, '');\n  return cleanedText.replace(pattern, replaceMatch).trim();\n}\n\nfunction getPoint(startBox: string, size: { width: number; height: number }) {\n  const [x, y] = JSON.parse(startBox);\n  return [x * size.width, y * size.height];\n}\n\ninterface BaseAction {\n  action_type: ActionType;\n  action_inputs: Record<string, any>;\n  reflection: string | null;\n  thought: string | null;\n}\n\ninterface ClickAction extends BaseAction {\n  action_type: 'click';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface DragAction extends BaseAction {\n  action_type: 'drag';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n    end_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface WaitAction extends BaseAction {\n  action_type: 'wait';\n  action_inputs: {\n    time: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface TypeAction extends BaseAction {\n  action_type: 'type';\n  action_inputs: {\n    content: string;\n  };\n}\n\ninterface HotkeyAction extends BaseAction {\n  action_type: 'hotkey';\n  action_inputs: {\n    key: string;\n  };\n}\n\ninterface ScrollAction extends BaseAction {\n  action_type: 'scroll';\n  action_inputs: {\n    direction: 'up' | 'down';\n  };\n}\n\ninterface FinishedAction extends BaseAction {\n  action_type: 'finished';\n  action_inputs: Record<string, never>;\n}\n\nexport type Action =\n  | ClickAction\n  | DragAction\n  | TypeAction\n  | HotkeyAction\n  | ScrollAction\n  | FinishedAction\n  | WaitAction;\n\nexport async function resizeImageForUiTars(\n  imageBase64: string,\n  size: Size,\n  modelPreferences: IModelPreferences,\n) {\n  if (\n    vlLocateMode(modelPreferences) === 'vlm-ui-tars' &&\n    uiTarsModelVersion(modelPreferences) === UITarsModelVersion.V1_5\n  ) {\n    debug('ui-tars-v1.5, will check image size', size);\n    const currentPixels = size.width * size.height;\n    const maxPixels = 16384 * 28 * 28; //\n    if (currentPixels > maxPixels) {\n      const resizeFactor = Math.sqrt(maxPixels / currentPixels);\n      const newWidth = Math.floor(size.width * resizeFactor);\n      const newHeight = Math.floor(size.height * resizeFactor);\n      debug(\n        'resize image for ui-tars, new width: %s, new height: %s',\n        newWidth,\n        newHeight,\n      );\n      const resizedImage = await resizeImgBase64(imageBase64, {\n        width: newWidth,\n        height: newHeight,\n      });\n      return resizedImage;\n    }\n  }\n  return imageBase64;\n}\n"],"names":["debug","getDebug","bboxSize","pointToBbox","point","width","height","Math","vlmPlanning","options","conversationHistory","userInstruction","size","modelPreferences","systemPrompt","getUiTarsPlanningPrompt","res","call","AIActionType","convertedText","convertBboxToCoordinates","modelVer","uiTarsModelVersion","parsed","actionParser","undefined","JSON","transformActions","action","assert","getPoint","startPoint","endPoint","keys","transformHotkeyInput","console","Error","getSummary","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","y","cleanedText","startBox","resizeImageForUiTars","imageBase64","vlLocateMode","UITarsModelVersion","currentPixels","maxPixels","resizeFactor","newWidth","newHeight","resizedImage","resizeImgBase64"],"mappings":";;;;;;;;;AA8BA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,WAAW;AACjB,MAAMC,cAAc,CAClBC,OACAC,OACAC,SAEO;QACLC,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGG;QAC5CE,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGI;KAC7C;AAGI,eAAeE,YAAYC,OAKjC;IAQC,MAAM,EAAEC,mBAAmB,EAAEC,eAAe,EAAEC,IAAI,EAAEC,gBAAgB,EAAE,GACpEJ;IACF,MAAMK,eAAeC,4BAA4BJ;IAEjD,MAAMK,MAAM,MAAMC,KAChB;QACE;YACE,MAAM;YACN,SAASH;QACX;WACGJ;KACJ,EACDQ,aAAa,eAAe,EAC5BL;IAEF,MAAMM,gBAAgBC,yBAAyBJ,IAAI,OAAO;IAE1D,MAAMK,WAAWC,mBAAmBT;IAEpC,MAAM,EAAEU,MAAM,EAAE,GAAGC,aAAa;QAC9B,YAAYL;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOP,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUS,YAAYI;IACxB;IAEAzB,MAAM,oBAAoBqB,UAAU,YAAYK,KAAK,SAAS,CAACH;IAE/D,MAAMI,mBAAqC,EAAE;IAC7CJ,OAAO,OAAO,CAAC,CAACK;QACd,IAAIA,AAAuB,YAAvBA,OAAO,WAAW,EAAc;YAClCC,OAAOD,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAMxB,QAAQ0B,SAASF,OAAO,aAAa,CAAC,SAAS,EAAEhB;YACvDe,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQC,OAAO,OAAO,IAAI;wBAC1B,MAAMzB,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BQ,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;YACF;QACF,OAAO,IAAIgB,AAAuB,WAAvBA,OAAO,WAAW,EAAa;YACxCC,OAAOD,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCC,OAAOD,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMG,aAAaD,SAASF,OAAO,aAAa,CAAC,SAAS,EAAEhB;YAC5D,MAAMoB,WAAWF,SAASF,OAAO,aAAa,CAAC,OAAO,EAAEhB;YACxDe,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,MAAM;wBACJ,QAAQC,OAAO,OAAO,IAAI;wBAC1B,MAAMzB,YACJ;4BAAE,GAAG4B,UAAU,CAAC,EAAE;4BAAE,GAAGA,UAAU,CAAC,EAAE;wBAAC,GACrCnB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;oBACA,IAAI;wBACF,QAAQgB,OAAO,OAAO,IAAI;wBAC1B,MAAMzB,YACJ;4BAAE,GAAG6B,QAAQ,CAAC,EAAE;4BAAE,GAAGA,QAAQ,CAAC,EAAE;wBAAC,GACjCpB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASgB,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIA,AAAuB,WAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,OAAOC,OAAO,aAAa,CAAC,OAAO;YACrC;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,aAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWC,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,eAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO,CAAC;YACR,SAASC,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIA,AAAuB,aAAvBA,OAAO,WAAW,EAC3B,IAAKA,OAAO,aAAa,CAAC,GAAG,EAItB;YACL,MAAMK,OAAOC,qBAAqBN,OAAO,aAAa,CAAC,GAAG;YAE1DD,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,SAASM;gBACX;gBACA,SAASL,OAAO,OAAO,IAAI;YAC7B;QACF,OAbEO,QAAQ,IAAI,CACV;aAaC,IAAIP,AAAuB,WAAvBA,OAAO,WAAW,EAC3BD,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,SAASC,OAAO,OAAO,IAAI;QAC7B;IAEJ;IAEA,IAAID,AAA4B,MAA5BA,iBAAiB,MAAM,EACzB,MAAM,IAAIS,MAAM,CAAC,4BAA4B,EAAEpB,IAAI,OAAO,EAAE,EAAE;QAC5D,OAAO;YACL,YAAYA,IAAI,OAAO;YACvBO;QACF;IACF;IAGFvB,MAAM,oBAAoB0B,KAAK,SAAS,CAACC,kBAAkB,MAAM;IAEjE,OAAO;QACL,SAASA;QACT,kBAAkBJ;QAClB,gBAAgBc,WAAWrB,IAAI,OAAO;QACtC,OAAOA,IAAI,KAAK;QAChB,aAAaU,KAAK,SAAS,CAACV,IAAI,OAAO,EAAES,QAAW;IACtD;AACF;AAOA,SAASL,yBAAyBkB,IAAY;IAE5C,MAAMC,UAAU;IAEhB,SAASC,aACPC,KAAa,EACbC,EAAU,EACVC,EAAU,EACVC,EAAU,EACVC,EAAU;QAGV,MAAMC,QAAQC,OAAO,QAAQ,CAACL,IAAI;QAClC,MAAMM,QAAQD,OAAO,QAAQ,CAACJ,IAAI;QAClC,MAAMM,QAAQF,OAAO,QAAQ,CAACH,IAAI;QAClC,MAAMM,QAAQH,OAAO,QAAQ,CAACF,IAAI;QAGlC,MAAMM,IAAI5C,KAAK,KAAK,CAAEuC,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMG,IAAI7C,KAAK,KAAK,CAAEyC,AAAAA,CAAAA,QAAQE,KAAI,IAAK;QAGvC,OAAO,CAAC,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC,CAAC;IACtB;IAGA,MAAMC,cAAcf,KAAK,OAAO,CAAC,YAAY;IAC7C,OAAOe,YAAY,OAAO,CAACd,SAASC,cAAc,IAAI;AACxD;AAEA,SAASV,SAASwB,QAAgB,EAAE1C,IAAuC;IACzE,MAAM,CAACuC,GAAGC,EAAE,GAAG1B,KAAK,KAAK,CAAC4B;IAC1B,OAAO;QAACH,IAAIvC,KAAK,KAAK;QAAEwC,IAAIxC,KAAK,MAAM;KAAC;AAC1C;AAkEO,eAAe2C,qBACpBC,WAAmB,EACnB5C,IAAU,EACVC,gBAAmC;IAEnC,IACE4C,AAAmC,kBAAnCA,aAAa5C,qBACbS,mBAAmBT,sBAAsB6C,mBAAmB,IAAI,EAChE;QACA1D,MAAM,uCAAuCY;QAC7C,MAAM+C,gBAAgB/C,KAAK,KAAK,GAAGA,KAAK,MAAM;QAC9C,MAAMgD,YAAY;QAClB,IAAID,gBAAgBC,WAAW;YAC7B,MAAMC,eAAetD,KAAK,IAAI,CAACqD,YAAYD;YAC3C,MAAMG,WAAWvD,KAAK,KAAK,CAACK,KAAK,KAAK,GAAGiD;YACzC,MAAME,YAAYxD,KAAK,KAAK,CAACK,KAAK,MAAM,GAAGiD;YAC3C7D,MACE,2DACA8D,UACAC;YAEF,MAAMC,eAAe,MAAMC,gBAAgBT,aAAa;gBACtD,OAAOM;gBACP,QAAQC;YACV;YACA,OAAOC;QACT;IACF;IACA,OAAOR;AACT"}

package/dist/es/insight/index.mjs

@@ -168,6 +168,9 @@ class Insight {
         const context = await this.contextRetrieverFn('describe');
         const { screenshotBase64, size } = context;
         assert(screenshotBase64, 'screenshot is required for insight.describe');
+        const modelPreferences = {
+            intent: 'grounding'
+        };
         const systemPrompt = elementDescriberInstruction();
         const defaultRectSize = 30;
         const targetRect = Array.isArray(target) ? {
@@ -187,9 +190,6 @@ class Insight {
             borderThickness: 3
         });
         if (null == opt ? void 0 : opt.deepThink) {
-            const modelPreferences = {
-                intent: 'grounding'
-            };
             const searchArea = expandSearchArea(targetRect, context.size, modelPreferences);
             debug('describe: set searchArea', searchArea);
             imagePayload = await cropByRect(imagePayload, searchArea, getIsUseQwenVl(modelPreferences));
@@ -213,7 +213,7 @@ class Insight {
             }
         ];
         const callAIFn = this.aiVendorFn || callToGetJSONObject;
-        const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT);
+        const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT, modelPreferences);
         const { content } = res;
         assert(!content.error, `describe failed: ${content.error}`);
         assert(content.description, 'failed to describe the element');

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

@@ -1 +1 @@
-{"version":3,"file":"insight/index.mjs","sources":["webpack://@midscene/core/./src/insight/index.ts"],"sourcesContent":["import {\n  AIActionType,\n  type AIArgs,\n  callAiFn,\n  expandSearchArea,\n} from '@/ai-model/common';\nimport {\n  AiExtractElementInfo,\n  AiLocateElement,\n  callToGetJSONObject,\n} from '@/ai-model/index';\nimport { AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n  AIDescribeElementResponse,\n  AIElementResponse,\n  AIUsageInfo,\n  BaseElement,\n  DetailedLocateParam,\n  DumpSubscriber,\n  InsightAction,\n  InsightExtractOption,\n  InsightExtractParam,\n  InsightOptions,\n  InsightTaskInfo,\n  LocateResult,\n  PartialInsightDumpFromSDK,\n  Rect,\n  UIContext,\n} from '@/types';\nimport {\n  type IModelPreferences,\n  MIDSCENE_FORCE_DEEP_THINK,\n  getAIConfigInBoolean,\n  getIsUseQwenVl,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TMultimodalPrompt } from '../ai-model/common';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext<BaseElement>;\n  callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n  ElementType extends BaseElement = BaseElement,\n  ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n  contextRetrieverFn: (\n    action: InsightAction,\n  ) => Promise<ContextType> | ContextType;\n\n  aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n  onceDumpUpdatedFn?: DumpSubscriber;\n\n  taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n  constructor(\n    context:\n      | ContextType\n      | ((action: InsightAction) => Promise<ContextType> | ContextType),\n    opt?: InsightOptions,\n  ) {\n    assert(context, 'context is required for Insight');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    if (typeof opt?.aiVendorFn !== 'undefined') {\n      this.aiVendorFn = opt.aiVendorFn;\n    }\n    if (typeof opt?.taskInfo !== 'undefined') {\n      this.taskInfo = opt.taskInfo;\n    }\n  }\n\n  async locate(\n    query: DetailedLocateParam,\n    opt?: LocateOpts,\n  ): Promise<LocateResult> {\n    const { callAI } = opt || {};\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    const globalDeepThinkSwitch = getAIConfigInBoolean(\n      MIDSCENE_FORCE_DEEP_THINK,\n    );\n    if (globalDeepThinkSwitch) {\n      debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n    }\n    let searchAreaPrompt;\n    if (query.deepThink || globalDeepThinkSwitch) {\n      searchAreaPrompt = query.prompt;\n    }\n    const modelPreferences: IModelPreferences = {\n      intent: 'grounding',\n    };\n\n    if (searchAreaPrompt && !vlLocateMode(modelPreferences)) {\n      console.warn(\n        'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n      );\n      searchAreaPrompt = undefined;\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n    let searchArea: Rect | undefined = undefined;\n    let searchAreaRawResponse: string | undefined = undefined;\n    let searchAreaUsage: AIUsageInfo | undefined = undefined;\n    let searchAreaResponse:\n      | Awaited<ReturnType<typeof AiLocateSection>>\n      | undefined = undefined;\n    if (searchAreaPrompt) {\n      searchAreaResponse = await AiLocateSection({\n        context,\n        sectionDescription: searchAreaPrompt,\n      });\n      assert(\n        searchAreaResponse.rect,\n        `cannot find search area for \"${searchAreaPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const {\n      parseResult,\n      rect,\n      elementById,\n      rawResponse,\n      usage,\n      isOrderSensitive,\n    } = await AiLocateElement({\n      callAI: callAI || this.aiVendorFn,\n      context,\n      targetElementDescription: queryPrompt,\n      searchConfig: searchAreaResponse,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedElement: [],\n      matchedRect: rect,\n      data: null,\n      taskInfo,\n      deepThink: !!searchArea,\n      error: errorLog,\n    };\n\n    const elements: BaseElement[] = [];\n    (parseResult.elements || []).forEach((item) => {\n      if ('id' in item) {\n        const element = elementById(item?.id);\n\n        if (!element) {\n          console.warn(\n            `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n          );\n          return;\n        }\n        elements.push(element);\n      }\n    });\n\n    emitInsightDump(\n      {\n        ...dumpData,\n        matchedElement: elements,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog) {\n      throw new Error(errorLog);\n    }\n\n    assert(\n      elements.length <= 1,\n      `locate: multiple elements found, length = ${elements.length}`,\n    );\n\n    if (elements.length === 1) {\n      return {\n        element: {\n          id: elements[0]!.id,\n          indexId: elements[0]!.indexId,\n          center: elements[0]!.center,\n          rect: elements[0]!.rect,\n          xpaths: elements[0]!.xpaths || [],\n          attributes: elements[0]!.attributes,\n          isOrderSensitive,\n        },\n        rect,\n      };\n    }\n    return {\n      element: null,\n      rect,\n    };\n  }\n\n  async extract<T>(\n    dataDemand: InsightExtractParam,\n    opt?: InsightExtractOption,\n    multimodalPrompt?: TMultimodalPrompt,\n  ): Promise<{\n    data: T;\n    thought?: string;\n    usage?: AIUsageInfo;\n  }> {\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const modelPreferences: IModelPreferences = {\n      intent: 'VQA',\n    };\n\n    const context = await this.contextRetrieverFn('extract');\n\n    const startTime = Date.now();\n    const { parseResult, usage } = await AiExtractElementInfo<T>({\n      context,\n      dataQuery: dataDemand,\n      multimodalPrompt,\n      extractOption: opt,\n      modelPreferences,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(parseResult),\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data, thought } = parseResult || {};\n\n    // 4\n    emitInsightDump(\n      {\n        ...dumpData,\n        data,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog && !data && !opt?.doNotThrowError) {\n      throw new Error(errorLog);\n    }\n\n    return {\n      data,\n      thought,\n      usage,\n    };\n  }\n\n  async describe(\n    target: Rect | [number, number],\n    opt?: {\n      deepThink?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for insight.describe');\n    const context = await this.contextRetrieverFn('describe');\n    const { screenshotBase64, size } = context;\n    assert(screenshotBase64, 'screenshot is required for insight.describe');\n\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      size,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepThink) {\n      // The result of the \"describe\" function will be used for positioning, so essentially it is a form of grounding.\n      const modelPreferences: IModelPreferences = { intent: 'grounding' };\n      const searchArea = expandSearchArea(\n        targetRect,\n        context.size,\n        modelPreferences,\n      );\n      debug('describe: set searchArea', searchArea);\n      imagePayload = await cropByRect(\n        imagePayload,\n        searchArea,\n        getIsUseQwenVl(modelPreferences),\n      );\n    }\n\n    const msgs: AIArgs = [\n      { role: 'system', content: systemPrompt },\n      {\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: imagePayload,\n              detail: 'high',\n            },\n          },\n        ],\n      },\n    ];\n\n    const callAIFn =\n      this.aiVendorFn || callToGetJSONObject<AIDescribeElementResponse>;\n\n    const res = await callAIFn(msgs, AIActionType.DESCRIBE_ELEMENT);\n\n    const { content } = res;\n    assert(!content.error, `describe failed: ${content.error}`);\n    assert(content.description, 'failed to describe the element');\n    return content;\n  }\n}\n"],"names":["debug","getDebug","Insight","query","opt","_parseResult_errors","callAI","queryPrompt","assert","dumpSubscriber","undefined","globalDeepThinkSwitch","getAIConfigInBoolean","MIDSCENE_FORCE_DEEP_THINK","searchAreaPrompt","modelPreferences","vlLocateMode","console","context","searchArea","searchAreaRawResponse","searchAreaUsage","searchAreaResponse","AiLocateSection","startTime","Date","parseResult","rect","elementById","rawResponse","usage","isOrderSensitive","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","elements","item","element","emitInsightDump","Error","dataDemand","multimodalPrompt","AiExtractElementInfo","data","thought","target","screenshotBase64","size","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","cropByRect","getIsUseQwenVl","msgs","callAIFn","callToGetJSONObject","res","AIActionType","content","callAiFn","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;AAoDA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IAmCnB,MAAM,OACJC,KAA0B,EAC1BC,GAAgB,EACO;YAkFnBC;QAjFJ,MAAM,EAAEC,MAAM,EAAE,GAAGF,OAAO,CAAC;QAC3B,MAAMG,cAAc,AAAiB,YAAjB,OAAOJ,QAAqBA,QAAQA,MAAM,MAAM;QACpEK,OAAOD,aAAa;QACpB,MAAME,iBAAiB,IAAI,CAAC,iBAAiB;QAC7C,IAAI,CAAC,iBAAiB,GAAGC;QAEzBF,OAAO,AAAiB,YAAjB,OAAOL,OAAoB;QAElC,MAAMQ,wBAAwBC,qBAC5BC;QAEF,IAAIF,uBACFX,MAAM,yBAAyBW;QAEjC,IAAIG;QACJ,IAAIX,MAAM,SAAS,IAAIQ,uBACrBG,mBAAmBX,MAAM,MAAM;QAEjC,MAAMY,mBAAsC;YAC1C,QAAQ;QACV;QAEA,IAAID,oBAAoB,CAACE,aAAaD,mBAAmB;YACvDE,QAAQ,IAAI,CACV;YAEFH,mBAAmBJ;QACrB;QAEA,MAAMQ,UAAUd,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAM,MAAM,IAAI,CAAC,kBAAkB,CAAC;QAE/D,IAAIe;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAGJ,IAAIR,kBAAkB;YACpBQ,qBAAqB,MAAMC,gBAAgB;gBACzCL;gBACA,oBAAoBJ;YACtB;YACAN,OACEc,mBAAmB,IAAI,EACvB,CAAC,6BAA6B,EAAER,iBAAiB,CAAC,EAChDQ,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAEJF,wBAAwBE,mBAAmB,WAAW;YACtDD,kBAAkBC,mBAAmB,KAAK;YAC1CH,aAAaG,mBAAmB,IAAI;QACtC;QAEA,MAAME,YAAYC,KAAK,GAAG;QAC1B,MAAM,EACJC,WAAW,EACXC,IAAI,EACJC,WAAW,EACXC,WAAW,EACXC,KAAK,EACLC,gBAAgB,EACjB,GAAG,MAAMC,gBAAgB;YACxB,QAAQ1B,UAAU,IAAI,CAAC,UAAU;YACjCY;YACA,0BAA0BX;YAC1B,cAAce;QAChB;QAEA,MAAMW,WAAWR,KAAK,GAAG,KAAKD;QAC9B,MAAMU,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACN;YAC5B,gBAAgBM,KAAK,SAAS,CAACT;YAC/BI;YACAX;YACAC;YACAC;QACF;QAEA,IAAIe;QACJ,IAAI,QAAA/B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B+B,WAAW,CAAC,6BAA6B,EAAEV,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG5E,MAAMW,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAAS9B;YACX;YACA,gBAAgB,EAAE;YAClB,aAAaoB;YACb,MAAM;YACNO;YACA,WAAW,CAAC,CAACf;YACb,OAAOiB;QACT;QAEA,MAAME,WAA0B,EAAE;QACjCZ,CAAAA,YAAY,QAAQ,IAAI,EAAC,EAAG,OAAO,CAAC,CAACa;YACpC,IAAI,QAAQA,MAAM;gBAChB,MAAMC,UAAUZ,YAAYW,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,EAAE;gBAEpC,IAAI,CAACC,SAAS,YACZvB,QAAQ,IAAI,CACV,CAAC,+BAA+B,EAAEsB,KAAK,EAAE,CAAC,0CAA0C,CAAC;gBAIzFD,SAAS,IAAI,CAACE;YAChB;QACF;QAEAC,gBACE;YACE,GAAGJ,QAAQ;YACX,gBAAgBC;QAClB,GACA7B;QAGF,IAAI2B,UACF,MAAM,IAAIM,MAAMN;QAGlB5B,OACE8B,SAAS,MAAM,IAAI,GACnB,CAAC,0CAA0C,EAAEA,SAAS,MAAM,EAAE;QAGhE,IAAIA,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,SAAS;gBACP,IAAIA,QAAQ,CAAC,EAAE,CAAE,EAAE;gBACnB,SAASA,QAAQ,CAAC,EAAE,CAAE,OAAO;gBAC7B,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM;gBAC3B,MAAMA,QAAQ,CAAC,EAAE,CAAE,IAAI;gBACvB,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM,IAAI,EAAE;gBACjC,YAAYA,QAAQ,CAAC,EAAE,CAAE,UAAU;gBACnCP;YACF;YACAJ;QACF;QAEF,OAAO;YACL,SAAS;YACTA;QACF;IACF;IAEA,MAAM,QACJgB,UAA+B,EAC/BvC,GAA0B,EAC1BwC,gBAAoC,EAKnC;YA+BGvC;QA9BJG,OACE,AAAsB,YAAtB,OAAOmC,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAEvE,MAAMlC,iBAAiB,IAAI,CAAC,iBAAiB;QAC7C,IAAI,CAAC,iBAAiB,GAAGC;QAEzB,MAAMK,mBAAsC;YAC1C,QAAQ;QACV;QAEA,MAAMG,UAAU,MAAM,IAAI,CAAC,kBAAkB,CAAC;QAE9C,MAAMM,YAAYC,KAAK,GAAG;QAC1B,MAAM,EAAEC,WAAW,EAAEI,KAAK,EAAE,GAAG,MAAMe,qBAAwB;YAC3D3B;YACA,WAAWyB;YACXC;YACA,eAAexC;YACfW;QACF;QAEA,MAAMkB,WAAWR,KAAK,GAAG,KAAKD;QAC9B,MAAMU,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACT;QAC9B;QAEA,IAAIU;QACJ,IAAI,QAAA/B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B+B,WAAW,CAAC,qBAAqB,EAAEV,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMW,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTM;YACF;YACA,gBAAgB,EAAE;YAClB,MAAM;YACNT;YACA,OAAOE;QACT;QAEA,MAAM,EAAEU,IAAI,EAAEC,OAAO,EAAE,GAAGrB,eAAe,CAAC;QAG1Ce,gBACE;YACE,GAAGJ,QAAQ;YACXS;QACF,GACArC;QAGF,IAAI2B,YAAY,CAACU,QAAQ,CAAC1C,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,AAAD,GAC3C,MAAM,IAAIsC,MAAMN;QAGlB,OAAO;YACLU;YACAC;YACAjB;QACF;IACF;IAEA,MAAM,SACJkB,MAA+B,EAC/B5C,GAEC,EACwD;QACzDI,OAAOwC,QAAQ;QACf,MAAM9B,UAAU,MAAM,IAAI,CAAC,kBAAkB,CAAC;QAC9C,MAAM,EAAE+B,gBAAgB,EAAEC,IAAI,EAAE,GAAGhC;QACnCV,OAAOyC,kBAAkB;QAEzB,MAAME,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,IAAIlD,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,EAAE;YAElB,MAAMW,mBAAsC;gBAAE,QAAQ;YAAY;YAClE,MAAMI,aAAawC,iBACjBL,YACApC,QAAQ,IAAI,EACZH;YAEFf,MAAM,4BAA4BmB;YAClCsC,eAAe,MAAMG,WACnBH,cACAtC,YACA0C,eAAe9C;QAEnB;QAEA,MAAM+C,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,WACJ,IAAI,CAAC,UAAU,IAAIC;QAErB,MAAMC,MAAM,MAAMF,SAASD,MAAMI,aAAa,gBAAgB;QAE9D,MAAM,EAAEC,OAAO,EAAE,GAAGF;QACpBzD,OAAO,CAAC2D,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1D3D,OAAO2D,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IAvUA,YACEjD,OAEmE,EACnEd,GAAoB,CACpB;QAfF;QAIA,qCAAoDgE;QAEpD;QAEA;QAQE5D,OAAOU,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMmD,QAAQ,OAAO,CAACnD;QAGlD,IAAI,AAA2B,WAApBd,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,UAAU,AAAD,GACvB,IAAI,CAAC,UAAU,GAAGA,IAAI,UAAU;QAElC,IAAI,AAAyB,WAAlBA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,QAAQ,AAAD,GACrB,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAqTF"}
+{"version":3,"file":"insight/index.mjs","sources":["webpack://@midscene/core/./src/insight/index.ts"],"sourcesContent":["import {\n  AIActionType,\n  type AIArgs,\n  callAiFn,\n  expandSearchArea,\n} from '@/ai-model/common';\nimport {\n  AiExtractElementInfo,\n  AiLocateElement,\n  callToGetJSONObject,\n} from '@/ai-model/index';\nimport { AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport type {\n  AIDescribeElementResponse,\n  AIElementResponse,\n  AIUsageInfo,\n  BaseElement,\n  DetailedLocateParam,\n  DumpSubscriber,\n  InsightAction,\n  InsightExtractOption,\n  InsightExtractParam,\n  InsightOptions,\n  InsightTaskInfo,\n  LocateResult,\n  PartialInsightDumpFromSDK,\n  Rect,\n  UIContext,\n} from '@/types';\nimport {\n  type IModelPreferences,\n  MIDSCENE_FORCE_DEEP_THINK,\n  getAIConfigInBoolean,\n  getIsUseQwenVl,\n  vlLocateMode,\n} from '@midscene/shared/env';\nimport { compositeElementInfoImg, cropByRect } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert } from '@midscene/shared/utils';\nimport type { TMultimodalPrompt } from '../ai-model/common';\nimport { emitInsightDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext<BaseElement>;\n  callAI?: typeof callAiFn<AIElementResponse>;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\nconst debug = getDebug('ai:insight');\nexport default class Insight<\n  ElementType extends BaseElement = BaseElement,\n  ContextType extends UIContext<ElementType> = UIContext<ElementType>,\n> {\n  contextRetrieverFn: (\n    action: InsightAction,\n  ) => Promise<ContextType> | ContextType;\n\n  aiVendorFn: (...args: Array<any>) => Promise<any> = callAiFn;\n\n  onceDumpUpdatedFn?: DumpSubscriber;\n\n  taskInfo?: Omit<InsightTaskInfo, 'durationMs'>;\n\n  constructor(\n    context:\n      | ContextType\n      | ((action: InsightAction) => Promise<ContextType> | ContextType),\n    opt?: InsightOptions,\n  ) {\n    assert(context, 'context is required for Insight');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    if (typeof opt?.aiVendorFn !== 'undefined') {\n      this.aiVendorFn = opt.aiVendorFn;\n    }\n    if (typeof opt?.taskInfo !== 'undefined') {\n      this.taskInfo = opt.taskInfo;\n    }\n  }\n\n  async locate(\n    query: DetailedLocateParam,\n    opt?: LocateOpts,\n  ): Promise<LocateResult> {\n    const { callAI } = opt || {};\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    const globalDeepThinkSwitch = getAIConfigInBoolean(\n      MIDSCENE_FORCE_DEEP_THINK,\n    );\n    if (globalDeepThinkSwitch) {\n      debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n    }\n    let searchAreaPrompt;\n    if (query.deepThink || globalDeepThinkSwitch) {\n      searchAreaPrompt = query.prompt;\n    }\n    const modelPreferences: IModelPreferences = {\n      intent: 'grounding',\n    };\n\n    if (searchAreaPrompt && !vlLocateMode(modelPreferences)) {\n      console.warn(\n        'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/choose-a-model',\n      );\n      searchAreaPrompt = undefined;\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn('locate'));\n\n    let searchArea: Rect | undefined = undefined;\n    let searchAreaRawResponse: string | undefined = undefined;\n    let searchAreaUsage: AIUsageInfo | undefined = undefined;\n    let searchAreaResponse:\n      | Awaited<ReturnType<typeof AiLocateSection>>\n      | undefined = undefined;\n    if (searchAreaPrompt) {\n      searchAreaResponse = await AiLocateSection({\n        context,\n        sectionDescription: searchAreaPrompt,\n      });\n      assert(\n        searchAreaResponse.rect,\n        `cannot find search area for \"${searchAreaPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const {\n      parseResult,\n      rect,\n      elementById,\n      rawResponse,\n      usage,\n      isOrderSensitive,\n    } = await AiLocateElement({\n      callAI: callAI || this.aiVendorFn,\n      context,\n      targetElementDescription: queryPrompt,\n      searchConfig: searchAreaResponse,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI model failed to locate: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'locate',\n      userQuery: {\n        element: queryPrompt,\n      },\n      matchedElement: [],\n      matchedRect: rect,\n      data: null,\n      taskInfo,\n      deepThink: !!searchArea,\n      error: errorLog,\n    };\n\n    const elements: BaseElement[] = [];\n    (parseResult.elements || []).forEach((item) => {\n      if ('id' in item) {\n        const element = elementById(item?.id);\n\n        if (!element) {\n          console.warn(\n            `locate: cannot find element id=${item.id}. Maybe an unstable response from AI model`,\n          );\n          return;\n        }\n        elements.push(element);\n      }\n    });\n\n    emitInsightDump(\n      {\n        ...dumpData,\n        matchedElement: elements,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog) {\n      throw new Error(errorLog);\n    }\n\n    assert(\n      elements.length <= 1,\n      `locate: multiple elements found, length = ${elements.length}`,\n    );\n\n    if (elements.length === 1) {\n      return {\n        element: {\n          id: elements[0]!.id,\n          indexId: elements[0]!.indexId,\n          center: elements[0]!.center,\n          rect: elements[0]!.rect,\n          xpaths: elements[0]!.xpaths || [],\n          attributes: elements[0]!.attributes,\n          isOrderSensitive,\n        },\n        rect,\n      };\n    }\n    return {\n      element: null,\n      rect,\n    };\n  }\n\n  async extract<T>(\n    dataDemand: InsightExtractParam,\n    opt?: InsightExtractOption,\n    multimodalPrompt?: TMultimodalPrompt,\n  ): Promise<{\n    data: T;\n    thought?: string;\n    usage?: AIUsageInfo;\n  }> {\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n    const dumpSubscriber = this.onceDumpUpdatedFn;\n    this.onceDumpUpdatedFn = undefined;\n\n    const modelPreferences: IModelPreferences = {\n      intent: 'VQA',\n    };\n\n    const context = await this.contextRetrieverFn('extract');\n\n    const startTime = Date.now();\n    const { parseResult, usage } = await AiExtractElementInfo<T>({\n      context,\n      dataQuery: dataDemand,\n      multimodalPrompt,\n      extractOption: opt,\n      modelPreferences,\n    });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: InsightTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(parseResult),\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialInsightDumpFromSDK = {\n      type: 'extract',\n      userQuery: {\n        dataDemand,\n      },\n      matchedElement: [],\n      data: null,\n      taskInfo,\n      error: errorLog,\n    };\n\n    const { data, thought } = parseResult || {};\n\n    // 4\n    emitInsightDump(\n      {\n        ...dumpData,\n        data,\n      },\n      dumpSubscriber,\n    );\n\n    if (errorLog && !data && !opt?.doNotThrowError) {\n      throw new Error(errorLog);\n    }\n\n    return {\n      data,\n      thought,\n      usage,\n    };\n  }\n\n  async describe(\n    target: Rect | [number, number],\n    opt?: {\n      deepThink?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for insight.describe');\n    const context = await this.contextRetrieverFn('describe');\n    const { screenshotBase64, size } = context;\n    assert(screenshotBase64, 'screenshot is required for insight.describe');\n    // The result of the \"describe\" function will be used for positioning, so essentially it is a form of grounding.\n    const modelPreferences: IModelPreferences = { intent: 'grounding' };\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      size,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepThink) {\n      const searchArea = expandSearchArea(\n        targetRect,\n        context.size,\n        modelPreferences,\n      );\n      debug('describe: set searchArea', searchArea);\n      imagePayload = await cropByRect(\n        imagePayload,\n        searchArea,\n        getIsUseQwenVl(modelPreferences),\n      );\n    }\n\n    const msgs: AIArgs = [\n      { role: 'system', content: systemPrompt },\n      {\n        role: 'user',\n        content: [\n          {\n            type: 'image_url',\n            image_url: {\n              url: imagePayload,\n              detail: 'high',\n            },\n          },\n        ],\n      },\n    ];\n\n    const callAIFn =\n      this.aiVendorFn || callToGetJSONObject<AIDescribeElementResponse>;\n\n    const res = await callAIFn(\n      msgs,\n      AIActionType.DESCRIBE_ELEMENT,\n      modelPreferences,\n    );\n\n    const { content } = res;\n    assert(!content.error, `describe failed: ${content.error}`);\n    assert(content.description, 'failed to describe the element');\n    return content;\n  }\n}\n"],"names":["debug","getDebug","Insight","query","opt","_parseResult_errors","callAI","queryPrompt","assert","dumpSubscriber","undefined","globalDeepThinkSwitch","getAIConfigInBoolean","MIDSCENE_FORCE_DEEP_THINK","searchAreaPrompt","modelPreferences","vlLocateMode","console","context","searchArea","searchAreaRawResponse","searchAreaUsage","searchAreaResponse","AiLocateSection","startTime","Date","parseResult","rect","elementById","rawResponse","usage","isOrderSensitive","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","elements","item","element","emitInsightDump","Error","dataDemand","multimodalPrompt","AiExtractElementInfo","data","thought","target","screenshotBase64","size","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","cropByRect","getIsUseQwenVl","msgs","callAIFn","callToGetJSONObject","res","AIActionType","content","callAiFn","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;AAoDA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IAmCnB,MAAM,OACJC,KAA0B,EAC1BC,GAAgB,EACO;YAkFnBC;QAjFJ,MAAM,EAAEC,MAAM,EAAE,GAAGF,OAAO,CAAC;QAC3B,MAAMG,cAAc,AAAiB,YAAjB,OAAOJ,QAAqBA,QAAQA,MAAM,MAAM;QACpEK,OAAOD,aAAa;QACpB,MAAME,iBAAiB,IAAI,CAAC,iBAAiB;QAC7C,IAAI,CAAC,iBAAiB,GAAGC;QAEzBF,OAAO,AAAiB,YAAjB,OAAOL,OAAoB;QAElC,MAAMQ,wBAAwBC,qBAC5BC;QAEF,IAAIF,uBACFX,MAAM,yBAAyBW;QAEjC,IAAIG;QACJ,IAAIX,MAAM,SAAS,IAAIQ,uBACrBG,mBAAmBX,MAAM,MAAM;QAEjC,MAAMY,mBAAsC;YAC1C,QAAQ;QACV;QAEA,IAAID,oBAAoB,CAACE,aAAaD,mBAAmB;YACvDE,QAAQ,IAAI,CACV;YAEFH,mBAAmBJ;QACrB;QAEA,MAAMQ,UAAUd,AAAAA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,OAAO,AAAD,KAAM,MAAM,IAAI,CAAC,kBAAkB,CAAC;QAE/D,IAAIe;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAGJ,IAAIR,kBAAkB;YACpBQ,qBAAqB,MAAMC,gBAAgB;gBACzCL;gBACA,oBAAoBJ;YACtB;YACAN,OACEc,mBAAmB,IAAI,EACvB,CAAC,6BAA6B,EAAER,iBAAiB,CAAC,EAChDQ,mBAAmB,KAAK,GAAG,CAAC,EAAE,EAAEA,mBAAmB,KAAK,EAAE,GAAG,IAC7D;YAEJF,wBAAwBE,mBAAmB,WAAW;YACtDD,kBAAkBC,mBAAmB,KAAK;YAC1CH,aAAaG,mBAAmB,IAAI;QACtC;QAEA,MAAME,YAAYC,KAAK,GAAG;QAC1B,MAAM,EACJC,WAAW,EACXC,IAAI,EACJC,WAAW,EACXC,WAAW,EACXC,KAAK,EACLC,gBAAgB,EACjB,GAAG,MAAMC,gBAAgB;YACxB,QAAQ1B,UAAU,IAAI,CAAC,UAAU;YACjCY;YACA,0BAA0BX;YAC1B,cAAce;QAChB;QAEA,MAAMW,WAAWR,KAAK,GAAG,KAAKD;QAC9B,MAAMU,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACN;YAC5B,gBAAgBM,KAAK,SAAS,CAACT;YAC/BI;YACAX;YACAC;YACAC;QACF;QAEA,IAAIe;QACJ,IAAI,QAAA/B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B+B,WAAW,CAAC,6BAA6B,EAAEV,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG5E,MAAMW,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAAS9B;YACX;YACA,gBAAgB,EAAE;YAClB,aAAaoB;YACb,MAAM;YACNO;YACA,WAAW,CAAC,CAACf;YACb,OAAOiB;QACT;QAEA,MAAME,WAA0B,EAAE;QACjCZ,CAAAA,YAAY,QAAQ,IAAI,EAAC,EAAG,OAAO,CAAC,CAACa;YACpC,IAAI,QAAQA,MAAM;gBAChB,MAAMC,UAAUZ,YAAYW,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,EAAE;gBAEpC,IAAI,CAACC,SAAS,YACZvB,QAAQ,IAAI,CACV,CAAC,+BAA+B,EAAEsB,KAAK,EAAE,CAAC,0CAA0C,CAAC;gBAIzFD,SAAS,IAAI,CAACE;YAChB;QACF;QAEAC,gBACE;YACE,GAAGJ,QAAQ;YACX,gBAAgBC;QAClB,GACA7B;QAGF,IAAI2B,UACF,MAAM,IAAIM,MAAMN;QAGlB5B,OACE8B,SAAS,MAAM,IAAI,GACnB,CAAC,0CAA0C,EAAEA,SAAS,MAAM,EAAE;QAGhE,IAAIA,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,SAAS;gBACP,IAAIA,QAAQ,CAAC,EAAE,CAAE,EAAE;gBACnB,SAASA,QAAQ,CAAC,EAAE,CAAE,OAAO;gBAC7B,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM;gBAC3B,MAAMA,QAAQ,CAAC,EAAE,CAAE,IAAI;gBACvB,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM,IAAI,EAAE;gBACjC,YAAYA,QAAQ,CAAC,EAAE,CAAE,UAAU;gBACnCP;YACF;YACAJ;QACF;QAEF,OAAO;YACL,SAAS;YACTA;QACF;IACF;IAEA,MAAM,QACJgB,UAA+B,EAC/BvC,GAA0B,EAC1BwC,gBAAoC,EAKnC;YA+BGvC;QA9BJG,OACE,AAAsB,YAAtB,OAAOmC,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAEvE,MAAMlC,iBAAiB,IAAI,CAAC,iBAAiB;QAC7C,IAAI,CAAC,iBAAiB,GAAGC;QAEzB,MAAMK,mBAAsC;YAC1C,QAAQ;QACV;QAEA,MAAMG,UAAU,MAAM,IAAI,CAAC,kBAAkB,CAAC;QAE9C,MAAMM,YAAYC,KAAK,GAAG;QAC1B,MAAM,EAAEC,WAAW,EAAEI,KAAK,EAAE,GAAG,MAAMe,qBAAwB;YAC3D3B;YACA,WAAWyB;YACXC;YACA,eAAexC;YACfW;QACF;QAEA,MAAMkB,WAAWR,KAAK,GAAG,KAAKD;QAC9B,MAAMU,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACT;QAC9B;QAEA,IAAIU;QACJ,IAAI,QAAA/B,CAAAA,sBAAAA,YAAY,MAAM,AAAD,IAAjBA,KAAAA,IAAAA,oBAAoB,MAAM,EAC5B+B,WAAW,CAAC,qBAAqB,EAAEV,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMW,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTM;YACF;YACA,gBAAgB,EAAE;YAClB,MAAM;YACNT;YACA,OAAOE;QACT;QAEA,MAAM,EAAEU,IAAI,EAAEC,OAAO,EAAE,GAAGrB,eAAe,CAAC;QAG1Ce,gBACE;YACE,GAAGJ,QAAQ;YACXS;QACF,GACArC;QAGF,IAAI2B,YAAY,CAACU,QAAQ,CAAC1C,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,eAAe,AAAD,GAC3C,MAAM,IAAIsC,MAAMN;QAGlB,OAAO;YACLU;YACAC;YACAjB;QACF;IACF;IAEA,MAAM,SACJkB,MAA+B,EAC/B5C,GAEC,EACwD;QACzDI,OAAOwC,QAAQ;QACf,MAAM9B,UAAU,MAAM,IAAI,CAAC,kBAAkB,CAAC;QAC9C,MAAM,EAAE+B,gBAAgB,EAAEC,IAAI,EAAE,GAAGhC;QACnCV,OAAOyC,kBAAkB;QAEzB,MAAMlC,mBAAsC;YAAE,QAAQ;QAAY;QAClE,MAAMoC,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,IAAIlD,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,SAAS,EAAE;YAClB,MAAMe,aAAawC,iBACjBL,YACApC,QAAQ,IAAI,EACZH;YAEFf,MAAM,4BAA4BmB;YAClCsC,eAAe,MAAMG,WACnBH,cACAtC,YACA0C,eAAe9C;QAEnB;QAEA,MAAM+C,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,WACJ,IAAI,CAAC,UAAU,IAAIC;QAErB,MAAMC,MAAM,MAAMF,SAChBD,MACAI,aAAa,gBAAgB,EAC7BnD;QAGF,MAAM,EAAEoD,OAAO,EAAE,GAAGF;QACpBzD,OAAO,CAAC2D,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1D3D,OAAO2D,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IA1UA,YACEjD,OAEmE,EACnEd,GAAoB,CACpB;QAfF;QAIA,qCAAoDgE;QAEpD;QAEA;QAQE5D,OAAOU,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMmD,QAAQ,OAAO,CAACnD;QAGlD,IAAI,AAA2B,WAApBd,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,UAAU,AAAD,GACvB,IAAI,CAAC,UAAU,GAAGA,IAAI,UAAU;QAElC,IAAI,AAAyB,WAAlBA,CAAAA,QAAAA,MAAAA,KAAAA,IAAAA,IAAK,QAAQ,AAAD,GACrB,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAwTF"}