@midscene/core 1.2.2-beta-20260115034338.0 → 1.2.2-beta-20260115090041.0

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

@@ -1 +1 @@
-{"version":3,"file":"ai-model/service-caller/index.mjs","sources":["../../../../src/ai-model/service-caller/index.ts"],"sourcesContent":["import {\n  AIResponseFormat,\n  type AIUsageInfo,\n  type DeepThinkOption,\n} from '@/types';\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\nimport {\n  type IModelConfig,\n  MIDSCENE_LANGFUSE_DEBUG,\n  MIDSCENE_LANGSMITH_DEBUG,\n  MIDSCENE_MODEL_MAX_TOKENS,\n  OPENAI_MAX_TOKENS,\n  type TVlModeTypes,\n  type UITarsModelVersion,\n  globalConfigManager,\n} from '@midscene/shared/env';\n\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert, ifInBrowser } from '@midscene/shared/utils';\nimport { jsonrepair } from 'jsonrepair';\nimport OpenAI from 'openai';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { Stream } from 'openai/streaming';\nimport type { AIArgs } from '../../common';\n\nasync function createChatClient({\n  modelConfig,\n}: {\n  modelConfig: IModelConfig;\n}): Promise<{\n  completion: OpenAI.Chat.Completions;\n  modelName: string;\n  modelDescription: string;\n  uiTarsVersion?: UITarsModelVersion;\n  vlMode: TVlModeTypes | undefined;\n}> {\n  const {\n    socksProxy,\n    httpProxy,\n    modelName,\n    openaiBaseURL,\n    openaiApiKey,\n    openaiExtraConfig,\n    modelDescription,\n    uiTarsModelVersion: uiTarsVersion,\n    vlMode,\n    createOpenAIClient,\n    timeout,\n  } = modelConfig;\n\n  let proxyAgent: any = undefined;\n  const debugProxy = getDebug('ai:call:proxy');\n\n  // Helper function to sanitize proxy URL for logging (remove credentials)\n  // Uses URL API instead of regex to avoid ReDoS vulnerabilities\n  const sanitizeProxyUrl = (url: string): string => {\n    try {\n      const parsed = new URL(url);\n      if (parsed.username) {\n        // Keep username for debugging, hide password for security\n        parsed.password = '****';\n        return parsed.href;\n      }\n      return url;\n    } catch {\n      // If URL parsing fails, return original URL (will be caught later)\n      return url;\n    }\n  };\n\n  if (httpProxy) {\n    debugProxy('using http proxy', sanitizeProxyUrl(httpProxy));\n    if (ifInBrowser) {\n      console.warn(\n        'HTTP proxy is configured but not supported in browser environment',\n      );\n    } else {\n      // Dynamic import with variable to avoid bundler static analysis\n      const moduleName = 'undici';\n      const { ProxyAgent } = await import(moduleName);\n      proxyAgent = new ProxyAgent({\n        uri: httpProxy,\n        // Note: authentication is handled via the URI (e.g., http://user:pass@proxy.com:8080)\n      });\n    }\n  } else if (socksProxy) {\n    debugProxy('using socks proxy', sanitizeProxyUrl(socksProxy));\n    if (ifInBrowser) {\n      console.warn(\n        'SOCKS proxy is configured but not supported in browser environment',\n      );\n    } else {\n      try {\n        // Dynamic import with variable to avoid bundler static analysis\n        const moduleName = 'fetch-socks';\n        const { socksDispatcher } = await import(moduleName);\n        // Parse SOCKS proxy URL (e.g., socks5://127.0.0.1:1080)\n        const proxyUrl = new URL(socksProxy);\n\n        // Validate hostname\n        if (!proxyUrl.hostname) {\n          throw new Error('SOCKS proxy URL must include a valid hostname');\n        }\n\n        // Validate and parse port\n        const port = Number.parseInt(proxyUrl.port, 10);\n        if (!proxyUrl.port || Number.isNaN(port)) {\n          throw new Error('SOCKS proxy URL must include a valid port');\n        }\n\n        // Parse SOCKS version from protocol\n        const protocol = proxyUrl.protocol.replace(':', '');\n        const socksType =\n          protocol === 'socks4' ? 4 : protocol === 'socks5' ? 5 : 5;\n\n        proxyAgent = socksDispatcher({\n          type: socksType,\n          host: proxyUrl.hostname,\n          port,\n          ...(proxyUrl.username\n            ? {\n                userId: decodeURIComponent(proxyUrl.username),\n                password: decodeURIComponent(proxyUrl.password || ''),\n              }\n            : {}),\n        });\n        debugProxy('socks proxy configured successfully', {\n          type: socksType,\n          host: proxyUrl.hostname,\n          port: port,\n        });\n      } catch (error) {\n        console.error('Failed to configure SOCKS proxy:', error);\n        throw new Error(\n          `Invalid SOCKS proxy URL: ${socksProxy}. Expected format: socks4://host:port, socks5://host:port, or with authentication: socks5://user:pass@host:port`,\n        );\n      }\n    }\n  }\n\n  const openAIOptions = {\n    baseURL: openaiBaseURL,\n    apiKey: openaiApiKey,\n    // Use fetchOptions.dispatcher for fetch-based SDK instead of httpAgent\n    // Note: Type assertion needed due to undici version mismatch between dependencies\n    ...(proxyAgent ? { fetchOptions: { dispatcher: proxyAgent as any } } : {}),\n    ...openaiExtraConfig,\n    ...(typeof timeout === 'number' ? { timeout } : {}),\n    dangerouslyAllowBrowser: true,\n  };\n\n  const baseOpenAI = new OpenAI(openAIOptions);\n\n  let openai: OpenAI = baseOpenAI;\n\n  // LangSmith wrapper\n  if (\n    openai &&\n    globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)\n  ) {\n    if (ifInBrowser) {\n      throw new Error('langsmith is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langsmith wrapper enabled');\n    // Use variable to prevent static analysis by bundlers\n    const langsmithModule = 'langsmith/wrappers';\n    const { wrapOpenAI } = await import(langsmithModule);\n    openai = wrapOpenAI(openai);\n  }\n\n  // Langfuse wrapper\n  if (\n    openai &&\n    globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGFUSE_DEBUG)\n  ) {\n    if (ifInBrowser) {\n      throw new Error('langfuse is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langfuse wrapper enabled');\n    // Use variable to prevent static analysis by bundlers\n    const langfuseModule = 'langfuse';\n    const { observeOpenAI } = await import(langfuseModule);\n    openai = observeOpenAI(openai);\n  }\n\n  if (createOpenAIClient) {\n    const wrappedClient = await createOpenAIClient(baseOpenAI, openAIOptions);\n\n    if (wrappedClient) {\n      openai = wrappedClient as OpenAI;\n    }\n  }\n\n  return {\n    completion: openai.chat.completions,\n    modelName,\n    modelDescription,\n    uiTarsVersion,\n    vlMode,\n  };\n}\n\nexport async function callAI(\n  messages: ChatCompletionMessageParam[],\n  modelConfig: IModelConfig,\n  options?: {\n    stream?: boolean;\n    onChunk?: StreamingCallback;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<{\n  content: string;\n  reasoning_content?: string;\n  usage?: AIUsageInfo;\n  isStreamed: boolean;\n}> {\n  const { completion, modelName, modelDescription, uiTarsVersion, vlMode } =\n    await createChatClient({\n      modelConfig,\n    });\n\n  const maxTokens =\n    globalConfigManager.getEnvConfigValueAsNumber(MIDSCENE_MODEL_MAX_TOKENS) ??\n    globalConfigManager.getEnvConfigValueAsNumber(OPENAI_MAX_TOKENS);\n  const debugCall = getDebug('ai:call');\n  const debugProfileStats = getDebug('ai:profile:stats');\n  const debugProfileDetail = getDebug('ai:profile:detail');\n\n  const startTime = Date.now();\n  const temperature = modelConfig.temperature ?? 0;\n\n  const isStreaming = options?.stream && options?.onChunk;\n  let content: string | undefined;\n  let accumulated = '';\n  let accumulatedReasoning = '';\n  let usage: OpenAI.CompletionUsage | undefined;\n  let timeCost: number | undefined;\n\n  const buildUsageInfo = (usageData?: OpenAI.CompletionUsage) => {\n    if (!usageData) return undefined;\n\n    const cachedInputTokens = (\n      usageData as { prompt_tokens_details?: { cached_tokens?: number } }\n    )?.prompt_tokens_details?.cached_tokens;\n\n    return {\n      prompt_tokens: usageData.prompt_tokens ?? 0,\n      completion_tokens: usageData.completion_tokens ?? 0,\n      total_tokens: usageData.total_tokens ?? 0,\n      cached_input: cachedInputTokens ?? 0,\n      time_cost: timeCost ?? 0,\n      model_name: modelName,\n      model_description: modelDescription,\n      intent: modelConfig.intent,\n    } satisfies AIUsageInfo;\n  };\n\n  const commonConfig = {\n    temperature,\n    stream: !!isStreaming,\n    max_tokens: maxTokens,\n    ...(vlMode === 'qwen2.5-vl' // qwen vl v2 specific config\n      ? {\n          vl_high_resolution_images: true,\n        }\n      : {}),\n  };\n  const {\n    config: deepThinkConfig,\n    debugMessage,\n    warningMessage,\n  } = resolveDeepThinkConfig({\n    deepThink: options?.deepThink,\n    vlMode,\n  });\n  if (debugMessage) {\n    debugCall(debugMessage);\n  }\n  if (warningMessage) {\n    debugCall(warningMessage);\n    console.warn(warningMessage);\n  }\n\n  try {\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          ...commonConfig,\n          ...deepThinkConfig,\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          accumulatedReasoning += reasoning_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: buildUsageInfo(usage),\n          };\n          options.onChunk!(finalChunk);\n          break;\n        }\n      }\n      content = accumulated;\n      debugProfileStats(\n        `streaming model, ${modelName}, mode, ${vlMode || 'default'}, cost-ms, ${timeCost}, temperature, ${temperature ?? ''}`,\n      );\n    } else {\n      const result = await completion.create({\n        model: modelName,\n        messages,\n        ...commonConfig,\n        ...deepThinkConfig,\n      } as any);\n      timeCost = Date.now() - startTime;\n\n      debugProfileStats(\n        `model, ${modelName}, mode, ${vlMode || 'default'}, ui-tars-version, ${uiTarsVersion}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}, temperature, ${temperature ?? ''}`,\n      );\n\n      debugProfileDetail(`model usage detail: ${JSON.stringify(result.usage)}`);\n\n      assert(\n        result.choices,\n        `invalid response from LLM service: ${JSON.stringify(result)}`,\n      );\n      content = result.choices[0].message.content!;\n      accumulatedReasoning =\n        (result.choices[0].message as any)?.reasoning_content || '';\n      usage = result.usage;\n    }\n\n    debugCall(`response reasoning content: ${accumulatedReasoning}`);\n    debugCall(`response content: ${content}`);\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      } as OpenAI.CompletionUsage;\n    }\n\n    return {\n      content: content || '',\n      reasoning_content: accumulatedReasoning || undefined,\n      usage: buildUsageInfo(usage),\n      isStreamed: !!isStreaming,\n    };\n  } catch (e: any) {\n    console.error(' call AI error', e);\n    const newError = new Error(\n      `failed to call ${isStreaming ? 'streaming ' : ''}AI model service (${modelName}): ${e.message}\\nTrouble shooting: https://midscenejs.com/model-provider.html`,\n      {\n        cause: e,\n      },\n    );\n    throw newError;\n  }\n}\n\nexport async function callAIWithObjectResponse<T>(\n  messages: ChatCompletionMessageParam[],\n  modelConfig: IModelConfig,\n  options?: {\n    deepThink?: DeepThinkOption;\n  },\n): Promise<{\n  content: T;\n  contentString: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}> {\n  const response = await callAI(messages, modelConfig, {\n    deepThink: options?.deepThink,\n  });\n  assert(response, 'empty response');\n  const vlMode = modelConfig.vlMode;\n  const jsonContent = safeParseJson(response.content, vlMode);\n  assert(\n    typeof jsonContent === 'object',\n    `failed to parse json response from model (${modelConfig.modelName}): ${response.content}`,\n  );\n  return {\n    content: jsonContent,\n    contentString: response.content,\n    usage: response.usage,\n    reasoning_content: response.reasoning_content,\n  };\n}\n\nexport async function callAIWithStringResponse(\n  msgs: AIArgs,\n  modelConfig: IModelConfig,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n  const { content, usage } = await callAI(msgs, modelConfig);\n  return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n  try {\n    // First, try to match a JSON object directly in the response\n    const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\n    if (jsonMatch) {\n      return jsonMatch[1];\n    }\n\n    // If no direct JSON object is found, try to extract JSON from a code block\n    const codeBlockMatch = response.match(\n      /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\n    );\n    if (codeBlockMatch) {\n      return codeBlockMatch[1];\n    }\n\n    // If no code block is found, try to find a JSON-like structure in the text\n    const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\n    if (jsonLikeMatch) {\n      return jsonLikeMatch[0];\n    }\n  } catch {}\n  // If no JSON-like structure is found, return the original response\n  return response;\n}\n\nexport function preprocessDoubaoBboxJson(input: string) {\n  if (input.includes('bbox')) {\n    // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\n    while (/\\d+\\s+\\d+/.test(input)) {\n      input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n    }\n  }\n  return input;\n}\n\nexport function resolveDeepThinkConfig({\n  deepThink,\n  vlMode,\n}: {\n  deepThink?: DeepThinkOption;\n  vlMode?: TVlModeTypes;\n}): {\n  config: Record<string, unknown>;\n  debugMessage?: string;\n  warningMessage?: string;\n} {\n  const normalizedDeepThink = deepThink === 'unset' ? undefined : deepThink;\n\n  if (normalizedDeepThink === undefined) {\n    return { config: {}, debugMessage: undefined };\n  }\n\n  if (vlMode === 'qwen3-vl') {\n    return {\n      config: { enable_thinking: normalizedDeepThink },\n      debugMessage: `deepThink mapped to enable_thinking=${normalizedDeepThink} for qwen3-vl`,\n    };\n  }\n\n  if (vlMode === 'doubao-vision') {\n    return {\n      config: {\n        thinking: { type: normalizedDeepThink ? 'enabled' : 'disabled' },\n      },\n      debugMessage: `deepThink mapped to thinking.type=${normalizedDeepThink ? 'enabled' : 'disabled'} for doubao-vision`,\n    };\n  }\n\n  if (vlMode === 'glm-v') {\n    return {\n      config: {\n        thinking: { type: normalizedDeepThink ? 'enabled' : 'disabled' },\n      },\n      debugMessage: `deepThink mapped to thinking.type=${normalizedDeepThink ? 'enabled' : 'disabled'} for glm-v`,\n    };\n  }\n\n  return {\n    config: {},\n    debugMessage: `deepThink ignored: unsupported model_family \"${vlMode ?? 'default'}\"`,\n    warningMessage: `The \"deepThink\" option is not supported for model_family \"${vlMode ?? 'default'}\".`,\n  };\n}\n\n/**\n * Normalize a parsed JSON object by trimming whitespace from:\n * 1. All object keys (e.g., \" prompt \" -> \"prompt\")\n * 2. All string values (e.g., \" Tap \" -> \"Tap\")\n * This handles LLM output that may include leading/trailing spaces.\n */\nfunction normalizeJsonObject(obj: any): any {\n  // Handle null and undefined\n  if (obj === null || obj === undefined) {\n    return obj;\n  }\n\n  // Handle arrays - recursively normalize each element\n  if (Array.isArray(obj)) {\n    return obj.map((item) => normalizeJsonObject(item));\n  }\n\n  // Handle objects\n  if (typeof obj === 'object') {\n    const normalized: any = {};\n\n    for (const [key, value] of Object.entries(obj)) {\n      // Trim the key to remove leading/trailing spaces\n      const trimmedKey = key.trim();\n\n      // Recursively normalize the value\n      let normalizedValue = normalizeJsonObject(value);\n\n      // Trim all string values\n      if (typeof normalizedValue === 'string') {\n        normalizedValue = normalizedValue.trim();\n      }\n\n      normalized[trimmedKey] = normalizedValue;\n    }\n\n    return normalized;\n  }\n\n  // Handle primitive strings\n  if (typeof obj === 'string') {\n    return obj.trim();\n  }\n\n  // Return other primitives as-is\n  return obj;\n}\n\nexport function safeParseJson(input: string, vlMode: TVlModeTypes | undefined) {\n  const cleanJsonString = extractJSONFromCodeBlock(input);\n  // match the point\n  if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\n    return cleanJsonString\n      .match(/\\((\\d+),(\\d+)\\)/)\n      ?.slice(1)\n      .map(Number);\n  }\n\n  let parsed: any;\n  let lastError: unknown;\n  try {\n    parsed = JSON.parse(cleanJsonString);\n    return normalizeJsonObject(parsed);\n  } catch (error) {\n    lastError = error;\n  }\n  try {\n    parsed = JSON.parse(jsonrepair(cleanJsonString));\n    return normalizeJsonObject(parsed);\n  } catch (error) {\n    lastError = error;\n  }\n\n  if (vlMode === 'doubao-vision' || vlMode === 'vlm-ui-tars') {\n    const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\n    try {\n      parsed = JSON.parse(jsonrepair(jsonString));\n      return normalizeJsonObject(parsed);\n    } catch (error) {\n      lastError = error;\n    }\n  }\n  throw Error(\n    `failed to parse LLM response into JSON. Error - ${String(\n      lastError ?? 'unknown error',\n    )}. Response - \\n ${input}`,\n  );\n}\n"],"names":["createChatClient","modelConfig","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","modelDescription","uiTarsVersion","vlMode","createOpenAIClient","timeout","proxyAgent","debugProxy","getDebug","sanitizeProxyUrl","url","parsed","URL","ifInBrowser","console","moduleName","ProxyAgent","socksDispatcher","proxyUrl","Error","port","Number","protocol","socksType","decodeURIComponent","error","openAIOptions","baseOpenAI","OpenAI","openai","globalConfigManager","MIDSCENE_LANGSMITH_DEBUG","langsmithModule","wrapOpenAI","MIDSCENE_LANGFUSE_DEBUG","langfuseModule","observeOpenAI","wrappedClient","callAI","messages","options","completion","maxTokens","MIDSCENE_MODEL_MAX_TOKENS","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","temperature","isStreaming","content","accumulated","accumulatedReasoning","usage","timeCost","buildUsageInfo","usageData","cachedInputTokens","commonConfig","deepThinkConfig","debugMessage","warningMessage","resolveDeepThinkConfig","stream","chunk","reasoning_content","chunkData","undefined","estimatedTokens","Math","finalChunk","result","JSON","assert","e","newError","callAIWithObjectResponse","response","jsonContent","safeParseJson","callAIWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","deepThink","normalizedDeepThink","normalizeJsonObject","obj","Array","item","normalized","key","value","Object","trimmedKey","normalizedValue","cleanJsonString","lastError","jsonrepair","jsonString","String"],"mappings":";;;;;AAyBA,eAAeA,iBAAiB,EAC9BC,WAAW,EAGZ;IAOC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,gBAAgB,EAChB,oBAAoBC,aAAa,EACjCC,MAAM,EACNC,kBAAkB,EAClBC,OAAO,EACR,GAAGX;IAEJ,IAAIY;IACJ,MAAMC,aAAaC,SAAS;IAI5B,MAAMC,mBAAmB,CAACC;QACxB,IAAI;YACF,MAAMC,SAAS,IAAIC,IAAIF;YACvB,IAAIC,OAAO,QAAQ,EAAE;gBAEnBA,OAAO,QAAQ,GAAG;gBAClB,OAAOA,OAAO,IAAI;YACpB;YACA,OAAOD;QACT,EAAE,OAAM;YAEN,OAAOA;QACT;IACF;IAEA,IAAId,WAAW;QACbW,WAAW,oBAAoBE,iBAAiBb;QAChD,IAAIiB,aACFC,QAAQ,IAAI,CACV;aAEG;YAEL,MAAMC,aAAa;YACnB,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAACD;YACpCT,aAAa,IAAIU,WAAW;gBAC1B,KAAKpB;YAEP;QACF;IACF,OAAO,IAAID,YAAY;QACrBY,WAAW,qBAAqBE,iBAAiBd;QACjD,IAAIkB,aACFC,QAAQ,IAAI,CACV;aAGF,IAAI;YAEF,MAAMC,aAAa;YACnB,MAAM,EAAEE,eAAe,EAAE,GAAG,MAAM,MAAM,CAACF;YAEzC,MAAMG,WAAW,IAAIN,IAAIjB;YAGzB,IAAI,CAACuB,SAAS,QAAQ,EACpB,MAAM,IAAIC,MAAM;YAIlB,MAAMC,OAAOC,OAAO,QAAQ,CAACH,SAAS,IAAI,EAAE;YAC5C,IAAI,CAACA,SAAS,IAAI,IAAIG,OAAO,KAAK,CAACD,OACjC,MAAM,IAAID,MAAM;YAIlB,MAAMG,WAAWJ,SAAS,QAAQ,CAAC,OAAO,CAAC,KAAK;YAChD,MAAMK,YACJD,AAAa,aAAbA,WAAwB,IAAIA,AAAa,aAAbA,WAAwB,IAAI;YAE1DhB,aAAaW,gBAAgB;gBAC3B,MAAMM;gBACN,MAAML,SAAS,QAAQ;gBACvBE;gBACA,GAAIF,SAAS,QAAQ,GACjB;oBACE,QAAQM,mBAAmBN,SAAS,QAAQ;oBAC5C,UAAUM,mBAAmBN,SAAS,QAAQ,IAAI;gBACpD,IACA,CAAC,CAAC;YACR;YACAX,WAAW,uCAAuC;gBAChD,MAAMgB;gBACN,MAAML,SAAS,QAAQ;gBACvB,MAAME;YACR;QACF,EAAE,OAAOK,OAAO;YACdX,QAAQ,KAAK,CAAC,oCAAoCW;YAClD,MAAM,IAAIN,MACR,CAAC,yBAAyB,EAAExB,WAAW,+GAA+G,CAAC;QAE3J;IAEJ;IAEA,MAAM+B,gBAAgB;QACpB,SAAS5B;QACT,QAAQC;QAGR,GAAIO,aAAa;YAAE,cAAc;gBAAE,YAAYA;YAAkB;QAAE,IAAI,CAAC,CAAC;QACzE,GAAGN,iBAAiB;QACpB,GAAI,AAAmB,YAAnB,OAAOK,UAAuB;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAClD,yBAAyB;IAC3B;IAEA,MAAMsB,aAAa,IAAIC,SAAOF;IAE9B,IAAIG,SAAiBF;IAGrB,IACEE,UACAC,oBAAoB,qBAAqB,CAACC,2BAC1C;QACA,IAAIlB,aACF,MAAM,IAAIM,MAAM;QAElBL,QAAQ,GAAG,CAAC;QAEZ,MAAMkB,kBAAkB;QACxB,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAACD;QACpCH,SAASI,WAAWJ;IACtB;IAGA,IACEA,UACAC,oBAAoB,qBAAqB,CAACI,0BAC1C;QACA,IAAIrB,aACF,MAAM,IAAIM,MAAM;QAElBL,QAAQ,GAAG,CAAC;QAEZ,MAAMqB,iBAAiB;QACvB,MAAM,EAAEC,aAAa,EAAE,GAAG,MAAM,MAAM,CAACD;QACvCN,SAASO,cAAcP;IACzB;IAEA,IAAIzB,oBAAoB;QACtB,MAAMiC,gBAAgB,MAAMjC,mBAAmBuB,YAAYD;QAE3D,IAAIW,eACFR,SAASQ;IAEb;IAEA,OAAO;QACL,YAAYR,OAAO,IAAI,CAAC,WAAW;QACnChC;QACAI;QACAC;QACAC;IACF;AACF;AAEO,eAAemC,OACpBC,QAAsC,EACtC7C,WAAyB,EACzB8C,OAIC;IAOD,MAAM,EAAEC,UAAU,EAAE5C,SAAS,EAAEI,gBAAgB,EAAEC,aAAa,EAAEC,MAAM,EAAE,GACtE,MAAMV,iBAAiB;QACrBC;IACF;IAEF,MAAMgD,YACJZ,oBAAoB,yBAAyB,CAACa,8BAC9Cb,oBAAoB,yBAAyB,CAACc;IAChD,MAAMC,YAAYrC,SAAS;IAC3B,MAAMsC,oBAAoBtC,SAAS;IACnC,MAAMuC,qBAAqBvC,SAAS;IAEpC,MAAMwC,YAAYC,KAAK,GAAG;IAC1B,MAAMC,cAAcxD,YAAY,WAAW,IAAI;IAE/C,MAAMyD,cAAcX,SAAS,UAAUA,SAAS;IAChD,IAAIY;IACJ,IAAIC,cAAc;IAClB,IAAIC,uBAAuB;IAC3B,IAAIC;IACJ,IAAIC;IAEJ,MAAMC,iBAAiB,CAACC;QACtB,IAAI,CAACA,WAAW;QAEhB,MAAMC,oBACJD,WACC,uBAAuB;QAE1B,OAAO;YACL,eAAeA,UAAU,aAAa,IAAI;YAC1C,mBAAmBA,UAAU,iBAAiB,IAAI;YAClD,cAAcA,UAAU,YAAY,IAAI;YACxC,cAAcC,qBAAqB;YACnC,WAAWH,YAAY;YACvB,YAAY3D;YACZ,mBAAmBI;YACnB,QAAQP,YAAY,MAAM;QAC5B;IACF;IAEA,MAAMkE,eAAe;QACnBV;QACA,QAAQ,CAAC,CAACC;QACV,YAAYT;QACZ,GAAIvC,AAAW,iBAAXA,SACA;YACE,2BAA2B;QAC7B,IACA,CAAC,CAAC;IACR;IACA,MAAM,EACJ,QAAQ0D,eAAe,EACvBC,YAAY,EACZC,cAAc,EACf,GAAGC,uBAAuB;QACzB,WAAWxB,SAAS;QACpBrC;IACF;IACA,IAAI2D,cACFjB,UAAUiB;IAEZ,IAAIC,gBAAgB;QAClBlB,UAAUkB;QACVjD,QAAQ,IAAI,CAACiD;IACf;IAEA,IAAI;QACFlB,UACE,CAAC,QAAQ,EAAEM,cAAc,eAAe,GAAG,WAAW,EAAEtD,WAAW;QAGrE,IAAIsD,aAAa;YACf,MAAMc,SAAU,MAAMxB,WAAW,MAAM,CACrC;gBACE,OAAO5C;gBACP0C;gBACA,GAAGqB,YAAY;gBACf,GAAGC,eAAe;YACpB,GACA;gBACE,QAAQ;YACV;YAKF,WAAW,MAAMK,SAASD,OAAQ;gBAChC,MAAMb,UAAUc,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,OAAO,WAAW;gBACtD,MAAMC,oBACHD,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,OAAe,qBAAqB;gBAG3D,IAAIA,MAAM,KAAK,EACbX,QAAQW,MAAM,KAAK;gBAGrB,IAAId,WAAWe,mBAAmB;oBAChCd,eAAeD;oBACfE,wBAAwBa;oBACxB,MAAMC,YAAiC;wBACrChB;wBACAe;wBACAd;wBACA,YAAY;wBACZ,OAAOgB;oBACT;oBACA7B,QAAQ,OAAO,CAAE4B;gBACnB;gBAGA,IAAIF,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,eAAe;oBACrCV,WAAWP,KAAK,GAAG,KAAKD;oBAGxB,IAAI,CAACO,OAAO;wBAEV,MAAMe,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAClB,YAAY,MAAM,GAAG;wBAElCE,QAAQ;4BACN,eAAee;4BACf,mBAAmBA;4BACnB,cAAcA,AAAkB,IAAlBA;wBAChB;oBACF;oBAGA,MAAME,aAAkC;wBACtC,SAAS;wBACTnB;wBACA,mBAAmB;wBACnB,YAAY;wBACZ,OAAOI,eAAeF;oBACxB;oBACAf,QAAQ,OAAO,CAAEgC;oBACjB;gBACF;YACF;YACApB,UAAUC;YACVP,kBACE,CAAC,iBAAiB,EAAEjD,UAAU,QAAQ,EAAEM,UAAU,UAAU,WAAW,EAAEqD,SAAS,eAAe,EAAEN,eAAe,IAAI;QAE1H,OAAO;YACL,MAAMuB,SAAS,MAAMhC,WAAW,MAAM,CAAC;gBACrC,OAAO5C;gBACP0C;gBACA,GAAGqB,YAAY;gBACf,GAAGC,eAAe;YACpB;YACAL,WAAWP,KAAK,GAAG,KAAKD;YAExBF,kBACE,CAAC,OAAO,EAAEjD,UAAU,QAAQ,EAAEM,UAAU,UAAU,mBAAmB,EAAED,cAAc,iBAAiB,EAAEuE,OAAO,KAAK,EAAE,iBAAiB,GAAG,qBAAqB,EAAEA,OAAO,KAAK,EAAE,qBAAqB,GAAG,gBAAgB,EAAEA,OAAO,KAAK,EAAE,gBAAgB,GAAG,WAAW,EAAEjB,SAAS,aAAa,EAAEiB,OAAO,WAAW,IAAI,GAAG,eAAe,EAAEvB,eAAe,IAAI;YAG9VH,mBAAmB,CAAC,oBAAoB,EAAE2B,KAAK,SAAS,CAACD,OAAO,KAAK,GAAG;YAExEE,OACEF,OAAO,OAAO,EACd,CAAC,mCAAmC,EAAEC,KAAK,SAAS,CAACD,SAAS;YAEhErB,UAAUqB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;YAC3CnB,uBACGmB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,EAAU,qBAAqB;YAC3DlB,QAAQkB,OAAO,KAAK;QACtB;QAEA5B,UAAU,CAAC,4BAA4B,EAAES,sBAAsB;QAC/DT,UAAU,CAAC,kBAAkB,EAAEO,SAAS;QACxCuB,OAAOvB,SAAS;QAGhB,IAAID,eAAe,CAACI,OAAO;YAEzB,MAAMe,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEnB,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;YAEtCG,QAAQ;gBACN,eAAee;gBACf,mBAAmBA;gBACnB,cAAcA,AAAkB,IAAlBA;YAChB;QACF;QAEA,OAAO;YACL,SAASlB,WAAW;YACpB,mBAAmBE,wBAAwBe;YAC3C,OAAOZ,eAAeF;YACtB,YAAY,CAAC,CAACJ;QAChB;IACF,EAAE,OAAOyB,GAAQ;QACf9D,QAAQ,KAAK,CAAC,kBAAkB8D;QAChC,MAAMC,WAAW,IAAI1D,MACnB,CAAC,eAAe,EAAEgC,cAAc,eAAe,GAAG,kBAAkB,EAAEtD,UAAU,GAAG,EAAE+E,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC9J;YACE,OAAOA;QACT;QAEF,MAAMC;IACR;AACF;AAEO,eAAeC,yBACpBvC,QAAsC,EACtC7C,WAAyB,EACzB8C,OAEC;IAOD,MAAMuC,WAAW,MAAMzC,OAAOC,UAAU7C,aAAa;QACnD,WAAW8C,SAAS;IACtB;IACAmC,OAAOI,UAAU;IACjB,MAAM5E,SAAST,YAAY,MAAM;IACjC,MAAMsF,cAAcC,cAAcF,SAAS,OAAO,EAAE5E;IACpDwE,OACE,AAAuB,YAAvB,OAAOK,aACP,CAAC,0CAA0C,EAAEtF,YAAY,SAAS,CAAC,GAAG,EAAEqF,SAAS,OAAO,EAAE;IAE5F,OAAO;QACL,SAASC;QACT,eAAeD,SAAS,OAAO;QAC/B,OAAOA,SAAS,KAAK;QACrB,mBAAmBA,SAAS,iBAAiB;IAC/C;AACF;AAEO,eAAeG,yBACpBC,IAAY,EACZzF,WAAyB;IAEzB,MAAM,EAAE0D,OAAO,EAAEG,KAAK,EAAE,GAAG,MAAMjB,OAAO6C,MAAMzF;IAC9C,OAAO;QAAE0D;QAASG;IAAM;AAC1B;AAEO,SAAS6B,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,SAASzB,uBAAuB,EACrC0B,SAAS,EACTvF,MAAM,EAIP;IAKC,MAAMwF,sBAAsBD,AAAc,YAAdA,YAAwBrB,SAAYqB;IAEhE,IAAIC,AAAwBtB,WAAxBsB,qBACF,OAAO;QAAE,QAAQ,CAAC;QAAG,cAActB;IAAU;IAG/C,IAAIlE,AAAW,eAAXA,QACF,OAAO;QACL,QAAQ;YAAE,iBAAiBwF;QAAoB;QAC/C,cAAc,CAAC,oCAAoC,EAAEA,oBAAoB,aAAa,CAAC;IACzF;IAGF,IAAIxF,AAAW,oBAAXA,QACF,OAAO;QACL,QAAQ;YACN,UAAU;gBAAE,MAAMwF,sBAAsB,YAAY;YAAW;QACjE;QACA,cAAc,CAAC,kCAAkC,EAAEA,sBAAsB,YAAY,WAAW,kBAAkB,CAAC;IACrH;IAGF,IAAIxF,AAAW,YAAXA,QACF,OAAO;QACL,QAAQ;YACN,UAAU;gBAAE,MAAMwF,sBAAsB,YAAY;YAAW;QACjE;QACA,cAAc,CAAC,kCAAkC,EAAEA,sBAAsB,YAAY,WAAW,UAAU,CAAC;IAC7G;IAGF,OAAO;QACL,QAAQ,CAAC;QACT,cAAc,CAAC,6CAA6C,EAAExF,UAAU,UAAU,CAAC,CAAC;QACpF,gBAAgB,CAAC,0DAA0D,EAAEA,UAAU,UAAU,EAAE,CAAC;IACtG;AACF;AAQA,SAASyF,oBAAoBC,GAAQ;IAEnC,IAAIA,QAAAA,KACF,OAAOA;IAIT,IAAIC,MAAM,OAAO,CAACD,MAChB,OAAOA,IAAI,GAAG,CAAC,CAACE,OAASH,oBAAoBG;IAI/C,IAAI,AAAe,YAAf,OAAOF,KAAkB;QAC3B,MAAMG,aAAkB,CAAC;QAEzB,KAAK,MAAM,CAACC,KAAKC,MAAM,IAAIC,OAAO,OAAO,CAACN,KAAM;YAE9C,MAAMO,aAAaH,IAAI,IAAI;YAG3B,IAAII,kBAAkBT,oBAAoBM;YAG1C,IAAI,AAA2B,YAA3B,OAAOG,iBACTA,kBAAkBA,gBAAgB,IAAI;YAGxCL,UAAU,CAACI,WAAW,GAAGC;QAC3B;QAEA,OAAOL;IACT;IAGA,IAAI,AAAe,YAAf,OAAOH,KACT,OAAOA,IAAI,IAAI;IAIjB,OAAOA;AACT;AAEO,SAASZ,cAAcQ,KAAa,EAAEtF,MAAgC;IAC3E,MAAMmG,kBAAkBlB,yBAAyBK;IAEjD,IAAIa,iBAAiB,MAAM,oBACzB,OAAOA,gBACJ,KAAK,CAAC,oBACL,MAAM,GACP,IAAIjF;IAGT,IAAIV;IACJ,IAAI4F;IACJ,IAAI;QACF5F,SAAS+D,KAAK,KAAK,CAAC4B;QACpB,OAAOV,oBAAoBjF;IAC7B,EAAE,OAAOc,OAAO;QACd8E,YAAY9E;IACd;IACA,IAAI;QACFd,SAAS+D,KAAK,KAAK,CAAC8B,WAAWF;QAC/B,OAAOV,oBAAoBjF;IAC7B,EAAE,OAAOc,OAAO;QACd8E,YAAY9E;IACd;IAEA,IAAItB,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,QAA0B;QAC1D,MAAMsG,aAAajB,yBAAyBc;QAC5C,IAAI;YACF3F,SAAS+D,KAAK,KAAK,CAAC8B,WAAWC;YAC/B,OAAOb,oBAAoBjF;QAC7B,EAAE,OAAOc,OAAO;YACd8E,YAAY9E;QACd;IACF;IACA,MAAMN,MACJ,CAAC,gDAAgD,EAAEuF,OACjDH,aAAa,iBACb,gBAAgB,EAAEd,OAAO;AAE/B"}
+{"version":3,"file":"ai-model/service-caller/index.mjs","sources":["../../../../src/ai-model/service-caller/index.ts"],"sourcesContent":["import {\n  AIResponseFormat,\n  type AIUsageInfo,\n  type DeepThinkOption,\n} from '@/types';\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\nimport {\n  type IModelConfig,\n  MIDSCENE_LANGFUSE_DEBUG,\n  MIDSCENE_LANGSMITH_DEBUG,\n  MIDSCENE_MODEL_MAX_TOKENS,\n  OPENAI_MAX_TOKENS,\n  type TVlModeTypes,\n  type UITarsModelVersion,\n  globalConfigManager,\n} from '@midscene/shared/env';\n\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert, ifInBrowser } from '@midscene/shared/utils';\nimport { jsonrepair } from 'jsonrepair';\nimport OpenAI from 'openai';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { Stream } from 'openai/streaming';\nimport type { AIArgs } from '../../common';\nimport { isAutoGLM } from '../auto-glm/util';\n\nasync function createChatClient({\n  modelConfig,\n}: {\n  modelConfig: IModelConfig;\n}): Promise<{\n  completion: OpenAI.Chat.Completions;\n  modelName: string;\n  modelDescription: string;\n  uiTarsVersion?: UITarsModelVersion;\n  vlMode: TVlModeTypes | undefined;\n}> {\n  const {\n    socksProxy,\n    httpProxy,\n    modelName,\n    openaiBaseURL,\n    openaiApiKey,\n    openaiExtraConfig,\n    modelDescription,\n    uiTarsModelVersion: uiTarsVersion,\n    vlMode,\n    createOpenAIClient,\n    timeout,\n  } = modelConfig;\n\n  let proxyAgent: any = undefined;\n  const debugProxy = getDebug('ai:call:proxy');\n\n  // Helper function to sanitize proxy URL for logging (remove credentials)\n  // Uses URL API instead of regex to avoid ReDoS vulnerabilities\n  const sanitizeProxyUrl = (url: string): string => {\n    try {\n      const parsed = new URL(url);\n      if (parsed.username) {\n        // Keep username for debugging, hide password for security\n        parsed.password = '****';\n        return parsed.href;\n      }\n      return url;\n    } catch {\n      // If URL parsing fails, return original URL (will be caught later)\n      return url;\n    }\n  };\n\n  if (httpProxy) {\n    debugProxy('using http proxy', sanitizeProxyUrl(httpProxy));\n    if (ifInBrowser) {\n      console.warn(\n        'HTTP proxy is configured but not supported in browser environment',\n      );\n    } else {\n      // Dynamic import with variable to avoid bundler static analysis\n      const moduleName = 'undici';\n      const { ProxyAgent } = await import(moduleName);\n      proxyAgent = new ProxyAgent({\n        uri: httpProxy,\n        // Note: authentication is handled via the URI (e.g., http://user:pass@proxy.com:8080)\n      });\n    }\n  } else if (socksProxy) {\n    debugProxy('using socks proxy', sanitizeProxyUrl(socksProxy));\n    if (ifInBrowser) {\n      console.warn(\n        'SOCKS proxy is configured but not supported in browser environment',\n      );\n    } else {\n      try {\n        // Dynamic import with variable to avoid bundler static analysis\n        const moduleName = 'fetch-socks';\n        const { socksDispatcher } = await import(moduleName);\n        // Parse SOCKS proxy URL (e.g., socks5://127.0.0.1:1080)\n        const proxyUrl = new URL(socksProxy);\n\n        // Validate hostname\n        if (!proxyUrl.hostname) {\n          throw new Error('SOCKS proxy URL must include a valid hostname');\n        }\n\n        // Validate and parse port\n        const port = Number.parseInt(proxyUrl.port, 10);\n        if (!proxyUrl.port || Number.isNaN(port)) {\n          throw new Error('SOCKS proxy URL must include a valid port');\n        }\n\n        // Parse SOCKS version from protocol\n        const protocol = proxyUrl.protocol.replace(':', '');\n        const socksType =\n          protocol === 'socks4' ? 4 : protocol === 'socks5' ? 5 : 5;\n\n        proxyAgent = socksDispatcher({\n          type: socksType,\n          host: proxyUrl.hostname,\n          port,\n          ...(proxyUrl.username\n            ? {\n                userId: decodeURIComponent(proxyUrl.username),\n                password: decodeURIComponent(proxyUrl.password || ''),\n              }\n            : {}),\n        });\n        debugProxy('socks proxy configured successfully', {\n          type: socksType,\n          host: proxyUrl.hostname,\n          port: port,\n        });\n      } catch (error) {\n        console.error('Failed to configure SOCKS proxy:', error);\n        throw new Error(\n          `Invalid SOCKS proxy URL: ${socksProxy}. Expected format: socks4://host:port, socks5://host:port, or with authentication: socks5://user:pass@host:port`,\n        );\n      }\n    }\n  }\n\n  const openAIOptions = {\n    baseURL: openaiBaseURL,\n    apiKey: openaiApiKey,\n    // Use fetchOptions.dispatcher for fetch-based SDK instead of httpAgent\n    // Note: Type assertion needed due to undici version mismatch between dependencies\n    ...(proxyAgent ? { fetchOptions: { dispatcher: proxyAgent as any } } : {}),\n    ...openaiExtraConfig,\n    ...(typeof timeout === 'number' ? { timeout } : {}),\n    dangerouslyAllowBrowser: true,\n  };\n\n  const baseOpenAI = new OpenAI(openAIOptions);\n\n  let openai: OpenAI = baseOpenAI;\n\n  // LangSmith wrapper\n  if (\n    openai &&\n    globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)\n  ) {\n    if (ifInBrowser) {\n      throw new Error('langsmith is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langsmith wrapper enabled');\n    // Use variable to prevent static analysis by bundlers\n    const langsmithModule = 'langsmith/wrappers';\n    const { wrapOpenAI } = await import(langsmithModule);\n    openai = wrapOpenAI(openai);\n  }\n\n  // Langfuse wrapper\n  if (\n    openai &&\n    globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGFUSE_DEBUG)\n  ) {\n    if (ifInBrowser) {\n      throw new Error('langfuse is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langfuse wrapper enabled');\n    // Use variable to prevent static analysis by bundlers\n    const langfuseModule = 'langfuse';\n    const { observeOpenAI } = await import(langfuseModule);\n    openai = observeOpenAI(openai);\n  }\n\n  if (createOpenAIClient) {\n    const wrappedClient = await createOpenAIClient(baseOpenAI, openAIOptions);\n\n    if (wrappedClient) {\n      openai = wrappedClient as OpenAI;\n    }\n  }\n\n  return {\n    completion: openai.chat.completions,\n    modelName,\n    modelDescription,\n    uiTarsVersion,\n    vlMode,\n  };\n}\n\nexport async function callAI(\n  messages: ChatCompletionMessageParam[],\n  modelConfig: IModelConfig,\n  options?: {\n    stream?: boolean;\n    onChunk?: StreamingCallback;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<{\n  content: string;\n  reasoning_content?: string;\n  usage?: AIUsageInfo;\n  isStreamed: boolean;\n}> {\n  const { completion, modelName, modelDescription, uiTarsVersion, vlMode } =\n    await createChatClient({\n      modelConfig,\n    });\n\n  const maxTokens =\n    globalConfigManager.getEnvConfigValueAsNumber(MIDSCENE_MODEL_MAX_TOKENS) ??\n    globalConfigManager.getEnvConfigValueAsNumber(OPENAI_MAX_TOKENS);\n  const debugCall = getDebug('ai:call');\n  const debugProfileStats = getDebug('ai:profile:stats');\n  const debugProfileDetail = getDebug('ai:profile:detail');\n\n  const startTime = Date.now();\n  const temperature = modelConfig.temperature ?? 0;\n\n  const isStreaming = options?.stream && options?.onChunk;\n  let content: string | undefined;\n  let accumulated = '';\n  let accumulatedReasoning = '';\n  let usage: OpenAI.CompletionUsage | undefined;\n  let timeCost: number | undefined;\n\n  const buildUsageInfo = (usageData?: OpenAI.CompletionUsage) => {\n    if (!usageData) return undefined;\n\n    const cachedInputTokens = (\n      usageData as { prompt_tokens_details?: { cached_tokens?: number } }\n    )?.prompt_tokens_details?.cached_tokens;\n\n    return {\n      prompt_tokens: usageData.prompt_tokens ?? 0,\n      completion_tokens: usageData.completion_tokens ?? 0,\n      total_tokens: usageData.total_tokens ?? 0,\n      cached_input: cachedInputTokens ?? 0,\n      time_cost: timeCost ?? 0,\n      model_name: modelName,\n      model_description: modelDescription,\n      intent: modelConfig.intent,\n    } satisfies AIUsageInfo;\n  };\n\n  const commonConfig = {\n    temperature,\n    stream: !!isStreaming,\n    max_tokens: maxTokens,\n    ...(vlMode === 'qwen2.5-vl' // qwen vl v2 specific config\n      ? {\n          vl_high_resolution_images: true,\n        }\n      : {}),\n  };\n\n  if (isAutoGLM(vlMode)) {\n    (commonConfig as unknown as Record<string, number>).top_p = 0.85;\n    (commonConfig as unknown as Record<string, number>).frequency_penalty = 0.2;\n  }\n\n  const {\n    config: deepThinkConfig,\n    debugMessage,\n    warningMessage,\n  } = resolveDeepThinkConfig({\n    deepThink: options?.deepThink,\n    vlMode,\n  });\n  if (debugMessage) {\n    debugCall(debugMessage);\n  }\n  if (warningMessage) {\n    debugCall(warningMessage);\n    console.warn(warningMessage);\n  }\n\n  try {\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          ...commonConfig,\n          ...deepThinkConfig,\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          accumulatedReasoning += reasoning_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: buildUsageInfo(usage),\n          };\n          options.onChunk!(finalChunk);\n          break;\n        }\n      }\n      content = accumulated;\n      debugProfileStats(\n        `streaming model, ${modelName}, mode, ${vlMode || 'default'}, cost-ms, ${timeCost}, temperature, ${temperature ?? ''}`,\n      );\n    } else {\n      const result = await completion.create({\n        model: modelName,\n        messages,\n        ...commonConfig,\n        ...deepThinkConfig,\n      } as any);\n      timeCost = Date.now() - startTime;\n\n      debugProfileStats(\n        `model, ${modelName}, mode, ${vlMode || 'default'}, ui-tars-version, ${uiTarsVersion}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}, temperature, ${temperature ?? ''}`,\n      );\n\n      debugProfileDetail(`model usage detail: ${JSON.stringify(result.usage)}`);\n\n      assert(\n        result.choices,\n        `invalid response from LLM service: ${JSON.stringify(result)}`,\n      );\n      content = result.choices[0].message.content!;\n      accumulatedReasoning =\n        (result.choices[0].message as any)?.reasoning_content || '';\n      usage = result.usage;\n    }\n\n    debugCall(`response reasoning content: ${accumulatedReasoning}`);\n    debugCall(`response content: ${content}`);\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      } as OpenAI.CompletionUsage;\n    }\n\n    return {\n      content: content || '',\n      reasoning_content: accumulatedReasoning || undefined,\n      usage: buildUsageInfo(usage),\n      isStreamed: !!isStreaming,\n    };\n  } catch (e: any) {\n    console.error(' call AI error', e);\n    const newError = new Error(\n      `failed to call ${isStreaming ? 'streaming ' : ''}AI model service (${modelName}): ${e.message}\\nTrouble shooting: https://midscenejs.com/model-provider.html`,\n      {\n        cause: e,\n      },\n    );\n    throw newError;\n  }\n}\n\nexport async function callAIWithObjectResponse<T>(\n  messages: ChatCompletionMessageParam[],\n  modelConfig: IModelConfig,\n  options?: {\n    deepThink?: DeepThinkOption;\n  },\n): Promise<{\n  content: T;\n  contentString: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}> {\n  const response = await callAI(messages, modelConfig, {\n    deepThink: options?.deepThink,\n  });\n  assert(response, 'empty response');\n  const vlMode = modelConfig.vlMode;\n  const jsonContent = safeParseJson(response.content, vlMode);\n  assert(\n    typeof jsonContent === 'object',\n    `failed to parse json response from model (${modelConfig.modelName}): ${response.content}`,\n  );\n  return {\n    content: jsonContent,\n    contentString: response.content,\n    usage: response.usage,\n    reasoning_content: response.reasoning_content,\n  };\n}\n\nexport async function callAIWithStringResponse(\n  msgs: AIArgs,\n  modelConfig: IModelConfig,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n  const { content, usage } = await callAI(msgs, modelConfig);\n  return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n  try {\n    // First, try to match a JSON object directly in the response\n    const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\n    if (jsonMatch) {\n      return jsonMatch[1];\n    }\n\n    // If no direct JSON object is found, try to extract JSON from a code block\n    const codeBlockMatch = response.match(\n      /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\n    );\n    if (codeBlockMatch) {\n      return codeBlockMatch[1];\n    }\n\n    // If no code block is found, try to find a JSON-like structure in the text\n    const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\n    if (jsonLikeMatch) {\n      return jsonLikeMatch[0];\n    }\n  } catch {}\n  // If no JSON-like structure is found, return the original response\n  return response;\n}\n\nexport function preprocessDoubaoBboxJson(input: string) {\n  if (input.includes('bbox')) {\n    // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\n    while (/\\d+\\s+\\d+/.test(input)) {\n      input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n    }\n  }\n  return input;\n}\n\nexport function resolveDeepThinkConfig({\n  deepThink,\n  vlMode,\n}: {\n  deepThink?: DeepThinkOption;\n  vlMode?: TVlModeTypes;\n}): {\n  config: Record<string, unknown>;\n  debugMessage?: string;\n  warningMessage?: string;\n} {\n  const normalizedDeepThink = deepThink === 'unset' ? undefined : deepThink;\n\n  if (normalizedDeepThink === undefined) {\n    return { config: {}, debugMessage: undefined };\n  }\n\n  if (vlMode === 'qwen3-vl') {\n    return {\n      config: { enable_thinking: normalizedDeepThink },\n      debugMessage: `deepThink mapped to enable_thinking=${normalizedDeepThink} for qwen3-vl`,\n    };\n  }\n\n  if (vlMode === 'doubao-vision') {\n    return {\n      config: {\n        thinking: { type: normalizedDeepThink ? 'enabled' : 'disabled' },\n      },\n      debugMessage: `deepThink mapped to thinking.type=${normalizedDeepThink ? 'enabled' : 'disabled'} for doubao-vision`,\n    };\n  }\n\n  if (vlMode === 'glm-v') {\n    return {\n      config: {\n        thinking: { type: normalizedDeepThink ? 'enabled' : 'disabled' },\n      },\n      debugMessage: `deepThink mapped to thinking.type=${normalizedDeepThink ? 'enabled' : 'disabled'} for glm-v`,\n    };\n  }\n\n  return {\n    config: {},\n    debugMessage: `deepThink ignored: unsupported model_family \"${vlMode ?? 'default'}\"`,\n    warningMessage: `The \"deepThink\" option is not supported for model_family \"${vlMode ?? 'default'}\".`,\n  };\n}\n\n/**\n * Normalize a parsed JSON object by trimming whitespace from:\n * 1. All object keys (e.g., \" prompt \" -> \"prompt\")\n * 2. All string values (e.g., \" Tap \" -> \"Tap\")\n * This handles LLM output that may include leading/trailing spaces.\n */\nfunction normalizeJsonObject(obj: any): any {\n  // Handle null and undefined\n  if (obj === null || obj === undefined) {\n    return obj;\n  }\n\n  // Handle arrays - recursively normalize each element\n  if (Array.isArray(obj)) {\n    return obj.map((item) => normalizeJsonObject(item));\n  }\n\n  // Handle objects\n  if (typeof obj === 'object') {\n    const normalized: any = {};\n\n    for (const [key, value] of Object.entries(obj)) {\n      // Trim the key to remove leading/trailing spaces\n      const trimmedKey = key.trim();\n\n      // Recursively normalize the value\n      let normalizedValue = normalizeJsonObject(value);\n\n      // Trim all string values\n      if (typeof normalizedValue === 'string') {\n        normalizedValue = normalizedValue.trim();\n      }\n\n      normalized[trimmedKey] = normalizedValue;\n    }\n\n    return normalized;\n  }\n\n  // Handle primitive strings\n  if (typeof obj === 'string') {\n    return obj.trim();\n  }\n\n  // Return other primitives as-is\n  return obj;\n}\n\nexport function safeParseJson(input: string, vlMode: TVlModeTypes | undefined) {\n  const cleanJsonString = extractJSONFromCodeBlock(input);\n  // match the point\n  if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\n    return cleanJsonString\n      .match(/\\((\\d+),(\\d+)\\)/)\n      ?.slice(1)\n      .map(Number);\n  }\n\n  let parsed: any;\n  let lastError: unknown;\n  try {\n    parsed = JSON.parse(cleanJsonString);\n    return normalizeJsonObject(parsed);\n  } catch (error) {\n    lastError = error;\n  }\n  try {\n    parsed = JSON.parse(jsonrepair(cleanJsonString));\n    return normalizeJsonObject(parsed);\n  } catch (error) {\n    lastError = error;\n  }\n\n  if (vlMode === 'doubao-vision' || vlMode === 'vlm-ui-tars') {\n    const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\n    try {\n      parsed = JSON.parse(jsonrepair(jsonString));\n      return normalizeJsonObject(parsed);\n    } catch (error) {\n      lastError = error;\n    }\n  }\n  throw Error(\n    `failed to parse LLM response into JSON. Error - ${String(\n      lastError ?? 'unknown error',\n    )}. Response - \\n ${input}`,\n  );\n}\n"],"names":["createChatClient","modelConfig","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","modelDescription","uiTarsVersion","vlMode","createOpenAIClient","timeout","proxyAgent","debugProxy","getDebug","sanitizeProxyUrl","url","parsed","URL","ifInBrowser","console","moduleName","ProxyAgent","socksDispatcher","proxyUrl","Error","port","Number","protocol","socksType","decodeURIComponent","error","openAIOptions","baseOpenAI","OpenAI","openai","globalConfigManager","MIDSCENE_LANGSMITH_DEBUG","langsmithModule","wrapOpenAI","MIDSCENE_LANGFUSE_DEBUG","langfuseModule","observeOpenAI","wrappedClient","callAI","messages","options","completion","maxTokens","MIDSCENE_MODEL_MAX_TOKENS","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","temperature","isStreaming","content","accumulated","accumulatedReasoning","usage","timeCost","buildUsageInfo","usageData","cachedInputTokens","commonConfig","isAutoGLM","deepThinkConfig","debugMessage","warningMessage","resolveDeepThinkConfig","stream","chunk","reasoning_content","chunkData","undefined","estimatedTokens","Math","finalChunk","result","JSON","assert","e","newError","callAIWithObjectResponse","response","jsonContent","safeParseJson","callAIWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","deepThink","normalizedDeepThink","normalizeJsonObject","obj","Array","item","normalized","key","value","Object","trimmedKey","normalizedValue","cleanJsonString","lastError","jsonrepair","jsonString","String"],"mappings":";;;;;;AA0BA,eAAeA,iBAAiB,EAC9BC,WAAW,EAGZ;IAOC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,gBAAgB,EAChB,oBAAoBC,aAAa,EACjCC,MAAM,EACNC,kBAAkB,EAClBC,OAAO,EACR,GAAGX;IAEJ,IAAIY;IACJ,MAAMC,aAAaC,SAAS;IAI5B,MAAMC,mBAAmB,CAACC;QACxB,IAAI;YACF,MAAMC,SAAS,IAAIC,IAAIF;YACvB,IAAIC,OAAO,QAAQ,EAAE;gBAEnBA,OAAO,QAAQ,GAAG;gBAClB,OAAOA,OAAO,IAAI;YACpB;YACA,OAAOD;QACT,EAAE,OAAM;YAEN,OAAOA;QACT;IACF;IAEA,IAAId,WAAW;QACbW,WAAW,oBAAoBE,iBAAiBb;QAChD,IAAIiB,aACFC,QAAQ,IAAI,CACV;aAEG;YAEL,MAAMC,aAAa;YACnB,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAACD;YACpCT,aAAa,IAAIU,WAAW;gBAC1B,KAAKpB;YAEP;QACF;IACF,OAAO,IAAID,YAAY;QACrBY,WAAW,qBAAqBE,iBAAiBd;QACjD,IAAIkB,aACFC,QAAQ,IAAI,CACV;aAGF,IAAI;YAEF,MAAMC,aAAa;YACnB,MAAM,EAAEE,eAAe,EAAE,GAAG,MAAM,MAAM,CAACF;YAEzC,MAAMG,WAAW,IAAIN,IAAIjB;YAGzB,IAAI,CAACuB,SAAS,QAAQ,EACpB,MAAM,IAAIC,MAAM;YAIlB,MAAMC,OAAOC,OAAO,QAAQ,CAACH,SAAS,IAAI,EAAE;YAC5C,IAAI,CAACA,SAAS,IAAI,IAAIG,OAAO,KAAK,CAACD,OACjC,MAAM,IAAID,MAAM;YAIlB,MAAMG,WAAWJ,SAAS,QAAQ,CAAC,OAAO,CAAC,KAAK;YAChD,MAAMK,YACJD,AAAa,aAAbA,WAAwB,IAAIA,AAAa,aAAbA,WAAwB,IAAI;YAE1DhB,aAAaW,gBAAgB;gBAC3B,MAAMM;gBACN,MAAML,SAAS,QAAQ;gBACvBE;gBACA,GAAIF,SAAS,QAAQ,GACjB;oBACE,QAAQM,mBAAmBN,SAAS,QAAQ;oBAC5C,UAAUM,mBAAmBN,SAAS,QAAQ,IAAI;gBACpD,IACA,CAAC,CAAC;YACR;YACAX,WAAW,uCAAuC;gBAChD,MAAMgB;gBACN,MAAML,SAAS,QAAQ;gBACvB,MAAME;YACR;QACF,EAAE,OAAOK,OAAO;YACdX,QAAQ,KAAK,CAAC,oCAAoCW;YAClD,MAAM,IAAIN,MACR,CAAC,yBAAyB,EAAExB,WAAW,+GAA+G,CAAC;QAE3J;IAEJ;IAEA,MAAM+B,gBAAgB;QACpB,SAAS5B;QACT,QAAQC;QAGR,GAAIO,aAAa;YAAE,cAAc;gBAAE,YAAYA;YAAkB;QAAE,IAAI,CAAC,CAAC;QACzE,GAAGN,iBAAiB;QACpB,GAAI,AAAmB,YAAnB,OAAOK,UAAuB;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAClD,yBAAyB;IAC3B;IAEA,MAAMsB,aAAa,IAAIC,SAAOF;IAE9B,IAAIG,SAAiBF;IAGrB,IACEE,UACAC,oBAAoB,qBAAqB,CAACC,2BAC1C;QACA,IAAIlB,aACF,MAAM,IAAIM,MAAM;QAElBL,QAAQ,GAAG,CAAC;QAEZ,MAAMkB,kBAAkB;QACxB,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAACD;QACpCH,SAASI,WAAWJ;IACtB;IAGA,IACEA,UACAC,oBAAoB,qBAAqB,CAACI,0BAC1C;QACA,IAAIrB,aACF,MAAM,IAAIM,MAAM;QAElBL,QAAQ,GAAG,CAAC;QAEZ,MAAMqB,iBAAiB;QACvB,MAAM,EAAEC,aAAa,EAAE,GAAG,MAAM,MAAM,CAACD;QACvCN,SAASO,cAAcP;IACzB;IAEA,IAAIzB,oBAAoB;QACtB,MAAMiC,gBAAgB,MAAMjC,mBAAmBuB,YAAYD;QAE3D,IAAIW,eACFR,SAASQ;IAEb;IAEA,OAAO;QACL,YAAYR,OAAO,IAAI,CAAC,WAAW;QACnChC;QACAI;QACAC;QACAC;IACF;AACF;AAEO,eAAemC,OACpBC,QAAsC,EACtC7C,WAAyB,EACzB8C,OAIC;IAOD,MAAM,EAAEC,UAAU,EAAE5C,SAAS,EAAEI,gBAAgB,EAAEC,aAAa,EAAEC,MAAM,EAAE,GACtE,MAAMV,iBAAiB;QACrBC;IACF;IAEF,MAAMgD,YACJZ,oBAAoB,yBAAyB,CAACa,8BAC9Cb,oBAAoB,yBAAyB,CAACc;IAChD,MAAMC,YAAYrC,SAAS;IAC3B,MAAMsC,oBAAoBtC,SAAS;IACnC,MAAMuC,qBAAqBvC,SAAS;IAEpC,MAAMwC,YAAYC,KAAK,GAAG;IAC1B,MAAMC,cAAcxD,YAAY,WAAW,IAAI;IAE/C,MAAMyD,cAAcX,SAAS,UAAUA,SAAS;IAChD,IAAIY;IACJ,IAAIC,cAAc;IAClB,IAAIC,uBAAuB;IAC3B,IAAIC;IACJ,IAAIC;IAEJ,MAAMC,iBAAiB,CAACC;QACtB,IAAI,CAACA,WAAW;QAEhB,MAAMC,oBACJD,WACC,uBAAuB;QAE1B,OAAO;YACL,eAAeA,UAAU,aAAa,IAAI;YAC1C,mBAAmBA,UAAU,iBAAiB,IAAI;YAClD,cAAcA,UAAU,YAAY,IAAI;YACxC,cAAcC,qBAAqB;YACnC,WAAWH,YAAY;YACvB,YAAY3D;YACZ,mBAAmBI;YACnB,QAAQP,YAAY,MAAM;QAC5B;IACF;IAEA,MAAMkE,eAAe;QACnBV;QACA,QAAQ,CAAC,CAACC;QACV,YAAYT;QACZ,GAAIvC,AAAW,iBAAXA,SACA;YACE,2BAA2B;QAC7B,IACA,CAAC,CAAC;IACR;IAEA,IAAI0D,UAAU1D,SAAS;QACpByD,aAAmD,KAAK,GAAG;QAC3DA,aAAmD,iBAAiB,GAAG;IAC1E;IAEA,MAAM,EACJ,QAAQE,eAAe,EACvBC,YAAY,EACZC,cAAc,EACf,GAAGC,uBAAuB;QACzB,WAAWzB,SAAS;QACpBrC;IACF;IACA,IAAI4D,cACFlB,UAAUkB;IAEZ,IAAIC,gBAAgB;QAClBnB,UAAUmB;QACVlD,QAAQ,IAAI,CAACkD;IACf;IAEA,IAAI;QACFnB,UACE,CAAC,QAAQ,EAAEM,cAAc,eAAe,GAAG,WAAW,EAAEtD,WAAW;QAGrE,IAAIsD,aAAa;YACf,MAAMe,SAAU,MAAMzB,WAAW,MAAM,CACrC;gBACE,OAAO5C;gBACP0C;gBACA,GAAGqB,YAAY;gBACf,GAAGE,eAAe;YACpB,GACA;gBACE,QAAQ;YACV;YAKF,WAAW,MAAMK,SAASD,OAAQ;gBAChC,MAAMd,UAAUe,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,OAAO,WAAW;gBACtD,MAAMC,oBACHD,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,OAAe,qBAAqB;gBAG3D,IAAIA,MAAM,KAAK,EACbZ,QAAQY,MAAM,KAAK;gBAGrB,IAAIf,WAAWgB,mBAAmB;oBAChCf,eAAeD;oBACfE,wBAAwBc;oBACxB,MAAMC,YAAiC;wBACrCjB;wBACAgB;wBACAf;wBACA,YAAY;wBACZ,OAAOiB;oBACT;oBACA9B,QAAQ,OAAO,CAAE6B;gBACnB;gBAGA,IAAIF,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,eAAe;oBACrCX,WAAWP,KAAK,GAAG,KAAKD;oBAGxB,IAAI,CAACO,OAAO;wBAEV,MAAMgB,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAACnB,YAAY,MAAM,GAAG;wBAElCE,QAAQ;4BACN,eAAegB;4BACf,mBAAmBA;4BACnB,cAAcA,AAAkB,IAAlBA;wBAChB;oBACF;oBAGA,MAAME,aAAkC;wBACtC,SAAS;wBACTpB;wBACA,mBAAmB;wBACnB,YAAY;wBACZ,OAAOI,eAAeF;oBACxB;oBACAf,QAAQ,OAAO,CAAEiC;oBACjB;gBACF;YACF;YACArB,UAAUC;YACVP,kBACE,CAAC,iBAAiB,EAAEjD,UAAU,QAAQ,EAAEM,UAAU,UAAU,WAAW,EAAEqD,SAAS,eAAe,EAAEN,eAAe,IAAI;QAE1H,OAAO;YACL,MAAMwB,SAAS,MAAMjC,WAAW,MAAM,CAAC;gBACrC,OAAO5C;gBACP0C;gBACA,GAAGqB,YAAY;gBACf,GAAGE,eAAe;YACpB;YACAN,WAAWP,KAAK,GAAG,KAAKD;YAExBF,kBACE,CAAC,OAAO,EAAEjD,UAAU,QAAQ,EAAEM,UAAU,UAAU,mBAAmB,EAAED,cAAc,iBAAiB,EAAEwE,OAAO,KAAK,EAAE,iBAAiB,GAAG,qBAAqB,EAAEA,OAAO,KAAK,EAAE,qBAAqB,GAAG,gBAAgB,EAAEA,OAAO,KAAK,EAAE,gBAAgB,GAAG,WAAW,EAAElB,SAAS,aAAa,EAAEkB,OAAO,WAAW,IAAI,GAAG,eAAe,EAAExB,eAAe,IAAI;YAG9VH,mBAAmB,CAAC,oBAAoB,EAAE4B,KAAK,SAAS,CAACD,OAAO,KAAK,GAAG;YAExEE,OACEF,OAAO,OAAO,EACd,CAAC,mCAAmC,EAAEC,KAAK,SAAS,CAACD,SAAS;YAEhEtB,UAAUsB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;YAC3CpB,uBACGoB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,EAAU,qBAAqB;YAC3DnB,QAAQmB,OAAO,KAAK;QACtB;QAEA7B,UAAU,CAAC,4BAA4B,EAAES,sBAAsB;QAC/DT,UAAU,CAAC,kBAAkB,EAAEO,SAAS;QACxCwB,OAAOxB,SAAS;QAGhB,IAAID,eAAe,CAACI,OAAO;YAEzB,MAAMgB,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEpB,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;YAEtCG,QAAQ;gBACN,eAAegB;gBACf,mBAAmBA;gBACnB,cAAcA,AAAkB,IAAlBA;YAChB;QACF;QAEA,OAAO;YACL,SAASnB,WAAW;YACpB,mBAAmBE,wBAAwBgB;YAC3C,OAAOb,eAAeF;YACtB,YAAY,CAAC,CAACJ;QAChB;IACF,EAAE,OAAO0B,GAAQ;QACf/D,QAAQ,KAAK,CAAC,kBAAkB+D;QAChC,MAAMC,WAAW,IAAI3D,MACnB,CAAC,eAAe,EAAEgC,cAAc,eAAe,GAAG,kBAAkB,EAAEtD,UAAU,GAAG,EAAEgF,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC9J;YACE,OAAOA;QACT;QAEF,MAAMC;IACR;AACF;AAEO,eAAeC,yBACpBxC,QAAsC,EACtC7C,WAAyB,EACzB8C,OAEC;IAOD,MAAMwC,WAAW,MAAM1C,OAAOC,UAAU7C,aAAa;QACnD,WAAW8C,SAAS;IACtB;IACAoC,OAAOI,UAAU;IACjB,MAAM7E,SAAST,YAAY,MAAM;IACjC,MAAMuF,cAAcC,cAAcF,SAAS,OAAO,EAAE7E;IACpDyE,OACE,AAAuB,YAAvB,OAAOK,aACP,CAAC,0CAA0C,EAAEvF,YAAY,SAAS,CAAC,GAAG,EAAEsF,SAAS,OAAO,EAAE;IAE5F,OAAO;QACL,SAASC;QACT,eAAeD,SAAS,OAAO;QAC/B,OAAOA,SAAS,KAAK;QACrB,mBAAmBA,SAAS,iBAAiB;IAC/C;AACF;AAEO,eAAeG,yBACpBC,IAAY,EACZ1F,WAAyB;IAEzB,MAAM,EAAE0D,OAAO,EAAEG,KAAK,EAAE,GAAG,MAAMjB,OAAO8C,MAAM1F;IAC9C,OAAO;QAAE0D;QAASG;IAAM;AAC1B;AAEO,SAAS8B,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,SAASzB,uBAAuB,EACrC0B,SAAS,EACTxF,MAAM,EAIP;IAKC,MAAMyF,sBAAsBD,AAAc,YAAdA,YAAwBrB,SAAYqB;IAEhE,IAAIC,AAAwBtB,WAAxBsB,qBACF,OAAO;QAAE,QAAQ,CAAC;QAAG,cAActB;IAAU;IAG/C,IAAInE,AAAW,eAAXA,QACF,OAAO;QACL,QAAQ;YAAE,iBAAiByF;QAAoB;QAC/C,cAAc,CAAC,oCAAoC,EAAEA,oBAAoB,aAAa,CAAC;IACzF;IAGF,IAAIzF,AAAW,oBAAXA,QACF,OAAO;QACL,QAAQ;YACN,UAAU;gBAAE,MAAMyF,sBAAsB,YAAY;YAAW;QACjE;QACA,cAAc,CAAC,kCAAkC,EAAEA,sBAAsB,YAAY,WAAW,kBAAkB,CAAC;IACrH;IAGF,IAAIzF,AAAW,YAAXA,QACF,OAAO;QACL,QAAQ;YACN,UAAU;gBAAE,MAAMyF,sBAAsB,YAAY;YAAW;QACjE;QACA,cAAc,CAAC,kCAAkC,EAAEA,sBAAsB,YAAY,WAAW,UAAU,CAAC;IAC7G;IAGF,OAAO;QACL,QAAQ,CAAC;QACT,cAAc,CAAC,6CAA6C,EAAEzF,UAAU,UAAU,CAAC,CAAC;QACpF,gBAAgB,CAAC,0DAA0D,EAAEA,UAAU,UAAU,EAAE,CAAC;IACtG;AACF;AAQA,SAAS0F,oBAAoBC,GAAQ;IAEnC,IAAIA,QAAAA,KACF,OAAOA;IAIT,IAAIC,MAAM,OAAO,CAACD,MAChB,OAAOA,IAAI,GAAG,CAAC,CAACE,OAASH,oBAAoBG;IAI/C,IAAI,AAAe,YAAf,OAAOF,KAAkB;QAC3B,MAAMG,aAAkB,CAAC;QAEzB,KAAK,MAAM,CAACC,KAAKC,MAAM,IAAIC,OAAO,OAAO,CAACN,KAAM;YAE9C,MAAMO,aAAaH,IAAI,IAAI;YAG3B,IAAII,kBAAkBT,oBAAoBM;YAG1C,IAAI,AAA2B,YAA3B,OAAOG,iBACTA,kBAAkBA,gBAAgB,IAAI;YAGxCL,UAAU,CAACI,WAAW,GAAGC;QAC3B;QAEA,OAAOL;IACT;IAGA,IAAI,AAAe,YAAf,OAAOH,KACT,OAAOA,IAAI,IAAI;IAIjB,OAAOA;AACT;AAEO,SAASZ,cAAcQ,KAAa,EAAEvF,MAAgC;IAC3E,MAAMoG,kBAAkBlB,yBAAyBK;IAEjD,IAAIa,iBAAiB,MAAM,oBACzB,OAAOA,gBACJ,KAAK,CAAC,oBACL,MAAM,GACP,IAAIlF;IAGT,IAAIV;IACJ,IAAI6F;IACJ,IAAI;QACF7F,SAASgE,KAAK,KAAK,CAAC4B;QACpB,OAAOV,oBAAoBlF;IAC7B,EAAE,OAAOc,OAAO;QACd+E,YAAY/E;IACd;IACA,IAAI;QACFd,SAASgE,KAAK,KAAK,CAAC8B,WAAWF;QAC/B,OAAOV,oBAAoBlF;IAC7B,EAAE,OAAOc,OAAO;QACd+E,YAAY/E;IACd;IAEA,IAAItB,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,QAA0B;QAC1D,MAAMuG,aAAajB,yBAAyBc;QAC5C,IAAI;YACF5F,SAASgE,KAAK,KAAK,CAAC8B,WAAWC;YAC/B,OAAOb,oBAAoBlF;QAC7B,EAAE,OAAOc,OAAO;YACd+E,YAAY/E;QACd;IACF;IACA,MAAMN,MACJ,CAAC,gDAAgD,EAAEwF,OACjDH,aAAa,iBACb,gBAAgB,EAAEd,OAAO;AAE/B"}

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

@@ -4,6 +4,7 @@ import { getDebug } from "@midscene/shared/logger";
 import { transformHotkeyInput } from "@midscene/shared/us-keyboard-layout";
 import { assert } from "@midscene/shared/utils";
 import { actionParser } from "@ui-tars/action-parser";
+import { LatestLocateRecorder } from "./latest-locate-recorder.mjs";
 import { getSummary, getUiTarsPlanningPrompt } from "./prompt/ui-tars-planning.mjs";
 import { callAIWithStringResponse } from "./service-caller/index.mjs";
 const debug = getDebug('ui-tars-planning');
@@ -14,6 +15,7 @@ const pointToBbox = (point, width, height)=>[
         Math.round(Math.min(point.x + bboxSize / 2, width)),
         Math.round(Math.min(point.y + bboxSize / 2, height))
     ];
+const lastLocateRecorder = new LatestLocateRecorder();
 async function uiTarsPlanning(userInstruction, options) {
     const { conversationHistory, context, modelConfig, actionContext } = options;
     const { uiTarsModelVersion } = modelConfig;
@@ -62,47 +64,53 @@ async function uiTarsPlanning(userInstruction, options) {
         if ('click' === actionType) {
             assert(action.action_inputs.start_box, 'start_box is required');
             const point = getPoint(action.action_inputs.start_box, size);
+            const locate = {
+                prompt: action.thought || '',
+                bbox: pointToBbox({
+                    x: point[0],
+                    y: point[1]
+                }, size.width, size.height)
+            };
+            lastLocateRecorder.recordLocate(locate, 'click');
             transformActions.push({
                 type: 'Tap',
                 param: {
-                    locate: {
-                        prompt: action.thought || '',
-                        bbox: pointToBbox({
-                            x: point[0],
-                            y: point[1]
-                        }, size.width, size.height)
-                    }
+                    locate: locate
                 }
             });
         } else if ('left_double' === actionType) {
             assert(action.action_inputs.start_box, 'start_box is required');
             const point = getPoint(action.action_inputs.start_box, size);
+            const locate = {
+                prompt: action.thought || '',
+                bbox: pointToBbox({
+                    x: point[0],
+                    y: point[1]
+                }, size.width, size.height)
+            };
+            lastLocateRecorder.recordLocate(locate, 'left_double');
             transformActions.push({
                 type: 'DoubleClick',
                 param: {
-                    locate: {
-                        prompt: action.thought || '',
-                        bbox: pointToBbox({
-                            x: point[0],
-                            y: point[1]
-                        }, size.width, size.height)
-                    }
+                    locate: locate
                 },
                 thought: action.thought || ''
             });
         } else if ('right_single' === actionType) {
             assert(action.action_inputs.start_box, 'start_box is required');
             const point = getPoint(action.action_inputs.start_box, size);
+            const locate = {
+                prompt: action.thought || '',
+                bbox: pointToBbox({
+                    x: point[0],
+                    y: point[1]
+                }, size.width, size.height)
+            };
+            lastLocateRecorder.recordLocate(locate, 'right_single');
             transformActions.push({
                 type: 'RightClick',
                 param: {
-                    locate: {
-                        prompt: action.thought || '',
-                        bbox: pointToBbox({
-                            x: point[0],
-                            y: point[1]
-                        }, size.width, size.height)
-                    }
+                    locate: locate
                 },
                 thought: action.thought || ''
             });
@@ -131,14 +139,18 @@ async function uiTarsPlanning(userInstruction, options) {
                 },
                 thought: action.thought || ''
             });
-        } else if ('type' === actionType) transformActions.push({
-            type: 'Input',
-            param: {
-                value: action.action_inputs.content
-            },
-            thought: action.thought || ''
-        });
-        else if ('scroll' === actionType) transformActions.push({
+        } else if ('type' === actionType) {
+            const { locate: latestLocate, source } = lastLocateRecorder.getLatestLocate();
+            debug(`use latestLocate from ${source} as locate when Input`, latestLocate);
+            transformActions.push({
+                type: 'Input',
+                param: {
+                    value: action.action_inputs.content,
+                    locate: latestLocate
+                },
+                thought: action.thought || ''
+            });
+        } else if ('scroll' === actionType) transformActions.push({
             type: 'Scroll',
             param: {
                 direction: action.action_inputs.direction

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

@@ -1 +1 @@
-{"version":3,"file":"ai-model/ui-tars-planning.mjs","sources":["../../../src/ai-model/ui-tars-planning.ts"],"sourcesContent":["import type {\n  PlanningAIResponse,\n  PlanningAction,\n  Size,\n  UIContext,\n} from '@/types';\nimport { type IModelConfig, UITarsModelVersion } 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 { ConversationHistory } from './conversation-history';\nimport { getSummary, getUiTarsPlanningPrompt } from './prompt/ui-tars-planning';\nimport { callAIWithStringResponse } from './service-caller/index';\ntype ActionType =\n  | 'click'\n  | 'left_double'\n  | 'right_single'\n  | 'drag'\n  | 'type'\n  | 'hotkey'\n  | 'finished'\n  | 'scroll'\n  | 'wait';\n\nconst debug = getDebug('ui-tars-planning');\nconst bboxSize = 10;\nconst pointToBbox = (\n  point: { x: number; y: number },\n  width: number,\n  height: number,\n): [number, number, number, number] => {\n  return [\n    Math.round(Math.max(point.x - bboxSize / 2, 0)),\n    Math.round(Math.max(point.y - bboxSize / 2, 0)),\n    Math.round(Math.min(point.x + bboxSize / 2, width)),\n    Math.round(Math.min(point.y + bboxSize / 2, height)),\n  ];\n};\n\nexport async function uiTarsPlanning(\n  userInstruction: string,\n  options: {\n    conversationHistory: ConversationHistory;\n    context: UIContext;\n    modelConfig: IModelConfig;\n    actionContext?: string;\n  },\n): Promise<PlanningAIResponse> {\n  const { conversationHistory, context, modelConfig, actionContext } = options;\n  const { uiTarsModelVersion } = modelConfig;\n\n  let instruction = userInstruction;\n  if (actionContext) {\n    instruction = `<high_priority_knowledge>${actionContext}</high_priority_knowledge>\\n<user_instruction>${userInstruction}</user_instruction>`;\n  }\n\n  const systemPrompt = getUiTarsPlanningPrompt() + instruction;\n\n  const imagePayload = await resizeImageForUiTars(\n    context.screenshotBase64,\n    context.size,\n    uiTarsModelVersion,\n  );\n\n  conversationHistory.append({\n    role: 'user',\n    content: [\n      {\n        type: 'image_url',\n        image_url: {\n          url: imagePayload,\n        },\n      },\n    ],\n  });\n\n  const res = await callAIWithStringResponse(\n    [\n      {\n        role: 'user',\n        content: systemPrompt,\n      },\n      ...conversationHistory.snapshot(),\n    ],\n    modelConfig,\n  );\n  const convertedText = convertBboxToCoordinates(res.content);\n\n  const { size } = context;\n  const { parsed } = actionParser({\n    prediction: convertedText,\n    factor: [1000, 1000],\n    screenContext: {\n      width: size.width,\n      height: size.height,\n    },\n    modelVer: uiTarsModelVersion,\n  });\n\n  debug(\n    'ui-tars modelVer',\n    uiTarsModelVersion,\n    ', parsed',\n    JSON.stringify(parsed),\n  );\n\n  const transformActions: PlanningAction[] = [];\n  const unhandledActions: Array<{ type: string; thought: string }> = [];\n  let shouldContinue = true;\n  parsed.forEach((action) => {\n    const actionType = (action.action_type || '').toLowerCase();\n    if (actionType === 'click') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n      transformActions.push({\n        type: 'Tap',\n        param: {\n          locate: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: point[0], y: point[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n      });\n    } else if (actionType === 'left_double') {\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: 'DoubleClick',\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        thought: action.thought || '',\n      });\n    } else if (actionType === 'right_single') {\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: 'RightClick',\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        thought: action.thought || '',\n      });\n    } else if (actionType === 'drag') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      assert(action.action_inputs.end_box, 'end_box is required');\n      const startPoint = getPoint(action.action_inputs.start_box, size);\n      const endPoint = getPoint(action.action_inputs.end_box, size);\n      transformActions.push({\n        type: 'DragAndDrop',\n        param: {\n          from: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: startPoint[0], y: startPoint[1] },\n              size.width,\n              size.height,\n            ),\n          },\n          to: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: endPoint[0], y: endPoint[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'type') {\n      transformActions.push({\n        type: 'Input',\n        param: {\n          value: action.action_inputs.content,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'scroll') {\n      transformActions.push({\n        type: 'Scroll',\n        param: {\n          direction: action.action_inputs.direction,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'finished') {\n      shouldContinue = false;\n      transformActions.push({\n        type: 'Finished',\n        param: {},\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'hotkey') {\n      if (!action.action_inputs.key) {\n        console.warn(\n          'No key found in action: hotkey. Will not perform action.',\n        );\n      } else {\n        const keys = transformHotkeyInput(action.action_inputs.key);\n\n        transformActions.push({\n          type: 'KeyboardPress',\n          param: {\n            keyName: keys.join('+'),\n          },\n          thought: action.thought || '',\n        });\n      }\n    } else if (actionType === 'wait') {\n      transformActions.push({\n        type: 'Sleep',\n        param: {\n          timeMs: 1000,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType) {\n      // Track unhandled action types\n      unhandledActions.push({\n        type: actionType,\n        thought: action.thought || '',\n      });\n      debug('Unhandled action type:', actionType, 'thought:', action.thought);\n    }\n  });\n\n  if (transformActions.length === 0) {\n    const errorDetails: string[] = [];\n\n    // Check if parsing failed\n    if (parsed.length === 0) {\n      errorDetails.push('Action parser returned no actions');\n\n      // Check if response has Thought but no Action\n      if (\n        res.content.includes('Thought:') &&\n        !res.content.includes('Action:')\n      ) {\n        errorDetails.push(\n          'Response contains \"Thought:\" but missing \"Action:\" line',\n        );\n      } else {\n        errorDetails.push('Response may be malformed or empty');\n      }\n    }\n\n    // Check if we have unhandled action types\n    if (unhandledActions.length > 0) {\n      const types = unhandledActions.map((a) => a.type).join(', ');\n      errorDetails.push(`Unhandled action types: ${types}`);\n    }\n\n    const errorMessage = [\n      'No actions found in UI-TARS response.',\n      ...errorDetails,\n      `\\nRaw response: ${res.content}`,\n    ].join('\\n');\n\n    throw new Error(errorMessage, {\n      cause: {\n        prediction: res.content,\n        parsed,\n        unhandledActions,\n        convertedText,\n      },\n    });\n  }\n\n  debug('transformActions', JSON.stringify(transformActions, null, 2));\n  const log = getSummary(res.content);\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: log,\n  });\n\n  return {\n    actions: transformActions,\n    log,\n    usage: res.usage,\n    rawResponse: JSON.stringify(res.content, undefined, 2),\n    more_actions_needed_by_instruction: shouldContinue,\n  };\n}\n\n/**\n * Converts bounding box notation to coordinate points\n * @param text - The text containing bbox tags to be converted\n * @returns The text with bbox tags replaced by coordinate points\n */\nfunction convertBboxToCoordinates(text: string): string {\n  // Match the four numbers after <bbox>\n  const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n  function replaceMatch(\n    match: string,\n    x1: string,\n    y1: string,\n    x2: string,\n    y2: string,\n  ): string {\n    // Convert strings to numbers and calculate center point\n    const x1Num = Number.parseInt(x1, 10);\n    const y1Num = Number.parseInt(y1, 10);\n    const x2Num = Number.parseInt(x2, 10);\n    const y2Num = Number.parseInt(y2, 10);\n\n    // Use Math.floor to truncate and calculate center point\n    const x = Math.floor((x1Num + x2Num) / 2);\n    const y = Math.floor((y1Num + y2Num) / 2);\n\n    // Return formatted coordinate string\n    return `(${x},${y})`;\n  }\n\n  // Remove [EOS] and replace <bbox> coordinates\n  const cleanedText = text.replace(/\\[EOS\\]/g, '');\n  return cleanedText.replace(pattern, replaceMatch).trim();\n}\n\nfunction getPoint(startBox: string, size: { width: number; height: number }) {\n  const [x, y] = JSON.parse(startBox);\n  return [x * size.width, y * size.height];\n}\n\ninterface BaseAction {\n  action_type: ActionType;\n  action_inputs: Record<string, any>;\n  reflection: string | null;\n  thought: string | null;\n}\n\ninterface ClickAction extends BaseAction {\n  action_type: 'click';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface DragAction extends BaseAction {\n  action_type: 'drag';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n    end_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface WaitAction extends BaseAction {\n  action_type: 'wait';\n  action_inputs: {\n    time: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface LeftDoubleAction extends BaseAction {\n  action_type: 'left_double';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface RightSingleAction extends BaseAction {\n  action_type: 'right_single';\n  action_inputs: {\n    start_box: 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  | LeftDoubleAction\n  | RightSingleAction\n  | DragAction\n  | TypeAction\n  | HotkeyAction\n  | ScrollAction\n  | FinishedAction\n  | WaitAction;\n\nexport async function resizeImageForUiTars(\n  imageBase64: string,\n  size: Size,\n  uiTarsVersion: UITarsModelVersion | undefined,\n) {\n  if (uiTarsVersion === UITarsModelVersion.V1_5) {\n    debug('ui-tars-v1.5, will check image size', size);\n    const currentPixels = size.width * size.height;\n    const maxPixels = 16384 * 28 * 28; //\n    if (currentPixels > maxPixels) {\n      const resizeFactor = Math.sqrt(maxPixels / currentPixels);\n      const newWidth = Math.floor(size.width * resizeFactor);\n      const newHeight = Math.floor(size.height * resizeFactor);\n      debug(\n        'resize image for ui-tars, new width: %s, new height: %s',\n        newWidth,\n        newHeight,\n      );\n      const resizedImage = await resizeImgBase64(imageBase64, {\n        width: newWidth,\n        height: newHeight,\n      });\n      return resizedImage;\n    }\n  }\n  return imageBase64;\n}\n"],"names":["debug","getDebug","bboxSize","pointToBbox","point","width","height","Math","uiTarsPlanning","userInstruction","options","conversationHistory","context","modelConfig","actionContext","uiTarsModelVersion","instruction","systemPrompt","getUiTarsPlanningPrompt","imagePayload","resizeImageForUiTars","res","callAIWithStringResponse","convertedText","convertBboxToCoordinates","size","parsed","actionParser","JSON","transformActions","unhandledActions","shouldContinue","action","actionType","assert","getPoint","startPoint","endPoint","keys","transformHotkeyInput","console","errorDetails","types","a","errorMessage","Error","log","getSummary","undefined","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","y","cleanedText","startBox","imageBase64","uiTarsVersion","UITarsModelVersion","currentPixels","maxPixels","resizeFactor","newWidth","newHeight","resizedImage","resizeImgBase64"],"mappings":";;;;;;;;AA0BA,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,eACpBC,eAAuB,EACvBC,OAKC;IAED,MAAM,EAAEC,mBAAmB,EAAEC,OAAO,EAAEC,WAAW,EAAEC,aAAa,EAAE,GAAGJ;IACrE,MAAM,EAAEK,kBAAkB,EAAE,GAAGF;IAE/B,IAAIG,cAAcP;IAClB,IAAIK,eACFE,cAAc,CAAC,yBAAyB,EAAEF,cAAc,8CAA8C,EAAEL,gBAAgB,mBAAmB,CAAC;IAG9I,MAAMQ,eAAeC,4BAA4BF;IAEjD,MAAMG,eAAe,MAAMC,qBACzBR,QAAQ,gBAAgB,EACxBA,QAAQ,IAAI,EACZG;IAGFJ,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKQ;gBACP;YACF;SACD;IACH;IAEA,MAAME,MAAM,MAAMC,yBAChB;QACE;YACE,MAAM;YACN,SAASL;QACX;WACGN,oBAAoB,QAAQ;KAChC,EACDE;IAEF,MAAMU,gBAAgBC,yBAAyBH,IAAI,OAAO;IAE1D,MAAM,EAAEI,IAAI,EAAE,GAAGb;IACjB,MAAM,EAAEc,MAAM,EAAE,GAAGC,aAAa;QAC9B,YAAYJ;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOE,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUV;IACZ;IAEAf,MACE,oBACAe,oBACA,YACAa,KAAK,SAAS,CAACF;IAGjB,MAAMG,mBAAqC,EAAE;IAC7C,MAAMC,mBAA6D,EAAE;IACrE,IAAIC,iBAAiB;IACrBL,OAAO,OAAO,CAAC,CAACM;QACd,MAAMC,aAAcD,AAAAA,CAAAA,OAAO,WAAW,IAAI,EAAC,EAAG,WAAW;QACzD,IAAIC,AAAe,YAAfA,YAAwB;YAC1BC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM5B,QAAQ+B,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YACvDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM7B,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BqB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;YACF;QACF,OAAO,IAAIQ,AAAe,kBAAfA,YAA8B;YACvCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM5B,QAAQ+B,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YACvDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM7B,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BqB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASO,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,mBAAfA,YAA+B;YACxCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM5B,QAAQ+B,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YACvDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM7B,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BqB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASO,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YAAuB;YAChCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCE,OAAOF,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMI,aAAaD,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAC5D,MAAMY,WAAWF,SAASH,OAAO,aAAa,CAAC,OAAO,EAAEP;YACxDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,MAAM;wBACJ,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM7B,YACJ;4BAAE,GAAGiC,UAAU,CAAC,EAAE;4BAAE,GAAGA,UAAU,CAAC,EAAE;wBAAC,GACrCX,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;oBACA,IAAI;wBACF,QAAQO,OAAO,OAAO,IAAI;wBAC1B,MAAM7B,YACJ;4BAAE,GAAGkC,QAAQ,CAAC,EAAE;4BAAE,GAAGA,QAAQ,CAAC,EAAE;wBAAC,GACjCZ,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASO,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,OAAOG,OAAO,aAAa,CAAC,OAAO;YACrC;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,aAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWG,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,eAAfA,YAA2B;YACpCF,iBAAiB;YACjBF,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO,CAAC;gBACR,SAASG,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,aAAfA,YACT,IAAKD,OAAO,aAAa,CAAC,GAAG,EAItB;YACL,MAAMM,OAAOC,qBAAqBP,OAAO,aAAa,CAAC,GAAG;YAE1DH,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,SAASS,KAAK,IAAI,CAAC;gBACrB;gBACA,SAASN,OAAO,OAAO,IAAI;YAC7B;QACF,OAbEQ,QAAQ,IAAI,CACV;aAaC,IAAIP,AAAe,WAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,SAASG,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,YAAY;YAErBH,iBAAiB,IAAI,CAAC;gBACpB,MAAMG;gBACN,SAASD,OAAO,OAAO,IAAI;YAC7B;YACAhC,MAAM,0BAA0BiC,YAAY,YAAYD,OAAO,OAAO;QACxE;IACF;IAEA,IAAIH,AAA4B,MAA5BA,iBAAiB,MAAM,EAAQ;QACjC,MAAMY,eAAyB,EAAE;QAGjC,IAAIf,AAAkB,MAAlBA,OAAO,MAAM,EAAQ;YACvBe,aAAa,IAAI,CAAC;YAGlB,IACEpB,IAAI,OAAO,CAAC,QAAQ,CAAC,eACrB,CAACA,IAAI,OAAO,CAAC,QAAQ,CAAC,YAEtBoB,aAAa,IAAI,CACf;iBAGFA,aAAa,IAAI,CAAC;QAEtB;QAGA,IAAIX,iBAAiB,MAAM,GAAG,GAAG;YAC/B,MAAMY,QAAQZ,iBAAiB,GAAG,CAAC,CAACa,IAAMA,EAAE,IAAI,EAAE,IAAI,CAAC;YACvDF,aAAa,IAAI,CAAC,CAAC,wBAAwB,EAAEC,OAAO;QACtD;QAEA,MAAME,eAAe;YACnB;eACGH;YACH,CAAC,gBAAgB,EAAEpB,IAAI,OAAO,EAAE;SACjC,CAAC,IAAI,CAAC;QAEP,MAAM,IAAIwB,MAAMD,cAAc;YAC5B,OAAO;gBACL,YAAYvB,IAAI,OAAO;gBACvBK;gBACAI;gBACAP;YACF;QACF;IACF;IAEAvB,MAAM,oBAAoB4B,KAAK,SAAS,CAACC,kBAAkB,MAAM;IACjE,MAAMiB,MAAMC,WAAW1B,IAAI,OAAO;IAElCV,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAASmC;IACX;IAEA,OAAO;QACL,SAASjB;QACTiB;QACA,OAAOzB,IAAI,KAAK;QAChB,aAAaO,KAAK,SAAS,CAACP,IAAI,OAAO,EAAE2B,QAAW;QACpD,oCAAoCjB;IACtC;AACF;AAOA,SAASP,yBAAyByB,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,IAAIvD,KAAK,KAAK,CAAEkD,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMG,IAAIxD,KAAK,KAAK,CAAEoD,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,SAAShB,SAAS8B,QAAgB,EAAExC,IAAuC;IACzE,MAAM,CAACqC,GAAGC,EAAE,GAAGnC,KAAK,KAAK,CAACqC;IAC1B,OAAO;QAACH,IAAIrC,KAAK,KAAK;QAAEsC,IAAItC,KAAK,MAAM;KAAC;AAC1C;AAkFO,eAAeL,qBACpB8C,WAAmB,EACnBzC,IAAU,EACV0C,aAA6C;IAE7C,IAAIA,kBAAkBC,mBAAmB,IAAI,EAAE;QAC7CpE,MAAM,uCAAuCyB;QAC7C,MAAM4C,gBAAgB5C,KAAK,KAAK,GAAGA,KAAK,MAAM;QAC9C,MAAM6C,YAAY;QAClB,IAAID,gBAAgBC,WAAW;YAC7B,MAAMC,eAAehE,KAAK,IAAI,CAAC+D,YAAYD;YAC3C,MAAMG,WAAWjE,KAAK,KAAK,CAACkB,KAAK,KAAK,GAAG8C;YACzC,MAAME,YAAYlE,KAAK,KAAK,CAACkB,KAAK,MAAM,GAAG8C;YAC3CvE,MACE,2DACAwE,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":["../../../src/ai-model/ui-tars-planning.ts"],"sourcesContent":["import type {\n  PlanningAIResponse,\n  PlanningAction,\n  Size,\n  UIContext,\n} from '@/types';\nimport { type IModelConfig, UITarsModelVersion } 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 { ConversationHistory } from './conversation-history';\nimport { LatestLocateRecorder } from './latest-locate-recorder';\nimport { getSummary, getUiTarsPlanningPrompt } from './prompt/ui-tars-planning';\nimport { callAIWithStringResponse } from './service-caller/index';\n\ntype ActionType =\n  | 'click'\n  | 'left_double'\n  | 'right_single'\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\nconst lastLocateRecorder = new LatestLocateRecorder();\n\nexport async function uiTarsPlanning(\n  userInstruction: string,\n  options: {\n    conversationHistory: ConversationHistory;\n    context: UIContext;\n    modelConfig: IModelConfig;\n    actionContext?: string;\n  },\n): Promise<PlanningAIResponse> {\n  const { conversationHistory, context, modelConfig, actionContext } = options;\n  const { uiTarsModelVersion } = modelConfig;\n\n  let instruction = userInstruction;\n  if (actionContext) {\n    instruction = `<high_priority_knowledge>${actionContext}</high_priority_knowledge>\\n<user_instruction>${userInstruction}</user_instruction>`;\n  }\n\n  const systemPrompt = getUiTarsPlanningPrompt() + instruction;\n\n  const imagePayload = await resizeImageForUiTars(\n    context.screenshotBase64,\n    context.size,\n    uiTarsModelVersion,\n  );\n\n  conversationHistory.append({\n    role: 'user',\n    content: [\n      {\n        type: 'image_url',\n        image_url: {\n          url: imagePayload,\n        },\n      },\n    ],\n  });\n\n  const res = await callAIWithStringResponse(\n    [\n      {\n        role: 'user',\n        content: systemPrompt,\n      },\n      ...conversationHistory.snapshot(),\n    ],\n    modelConfig,\n  );\n  const convertedText = convertBboxToCoordinates(res.content);\n\n  const { size } = context;\n  const { parsed } = actionParser({\n    prediction: convertedText,\n    factor: [1000, 1000],\n    screenContext: {\n      width: size.width,\n      height: size.height,\n    },\n    modelVer: uiTarsModelVersion,\n  });\n\n  debug(\n    'ui-tars modelVer',\n    uiTarsModelVersion,\n    ', parsed',\n    JSON.stringify(parsed),\n  );\n\n  const transformActions: PlanningAction[] = [];\n  const unhandledActions: Array<{ type: string; thought: string }> = [];\n  let shouldContinue = true;\n  parsed.forEach((action) => {\n    const actionType = (action.action_type || '').toLowerCase();\n    if (actionType === 'click') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n\n      const 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      lastLocateRecorder.recordLocate(locate, 'click');\n      transformActions.push({\n        type: 'Tap',\n        param: {\n          locate: locate,\n        },\n      });\n    } else if (actionType === 'left_double') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n\n      const 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      lastLocateRecorder.recordLocate(locate, 'left_double');\n      transformActions.push({\n        type: 'DoubleClick',\n        param: {\n          locate: locate,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'right_single') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n\n      const 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      lastLocateRecorder.recordLocate(locate, 'right_single');\n      transformActions.push({\n        type: 'RightClick',\n        param: {\n          locate: locate,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'drag') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      assert(action.action_inputs.end_box, 'end_box is required');\n      const startPoint = getPoint(action.action_inputs.start_box, size);\n      const endPoint = getPoint(action.action_inputs.end_box, size);\n      transformActions.push({\n        type: 'DragAndDrop',\n        param: {\n          from: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: startPoint[0], y: startPoint[1] },\n              size.width,\n              size.height,\n            ),\n          },\n          to: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: endPoint[0], y: endPoint[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'type') {\n      const { locate: latestLocate, source } =\n        lastLocateRecorder.getLatestLocate();\n      debug(\n        `use latestLocate from ${source} as locate when Input`,\n        latestLocate,\n      );\n\n      transformActions.push({\n        type: 'Input',\n        param: {\n          value: action.action_inputs.content,\n          locate: latestLocate,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'scroll') {\n      transformActions.push({\n        type: 'Scroll',\n        param: {\n          direction: action.action_inputs.direction,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'finished') {\n      shouldContinue = false;\n      transformActions.push({\n        type: 'Finished',\n        param: {},\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'hotkey') {\n      if (!action.action_inputs.key) {\n        console.warn(\n          'No key found in action: hotkey. Will not perform action.',\n        );\n      } else {\n        const keys = transformHotkeyInput(action.action_inputs.key);\n\n        transformActions.push({\n          type: 'KeyboardPress',\n          param: {\n            keyName: keys.join('+'),\n          },\n          thought: action.thought || '',\n        });\n      }\n    } else if (actionType === 'wait') {\n      transformActions.push({\n        type: 'Sleep',\n        param: {\n          timeMs: 1000,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType) {\n      // Track unhandled action types\n      unhandledActions.push({\n        type: actionType,\n        thought: action.thought || '',\n      });\n      debug('Unhandled action type:', actionType, 'thought:', action.thought);\n    }\n  });\n\n  if (transformActions.length === 0) {\n    const errorDetails: string[] = [];\n\n    // Check if parsing failed\n    if (parsed.length === 0) {\n      errorDetails.push('Action parser returned no actions');\n\n      // Check if response has Thought but no Action\n      if (\n        res.content.includes('Thought:') &&\n        !res.content.includes('Action:')\n      ) {\n        errorDetails.push(\n          'Response contains \"Thought:\" but missing \"Action:\" line',\n        );\n      } else {\n        errorDetails.push('Response may be malformed or empty');\n      }\n    }\n\n    // Check if we have unhandled action types\n    if (unhandledActions.length > 0) {\n      const types = unhandledActions.map((a) => a.type).join(', ');\n      errorDetails.push(`Unhandled action types: ${types}`);\n    }\n\n    const errorMessage = [\n      'No actions found in UI-TARS response.',\n      ...errorDetails,\n      `\\nRaw response: ${res.content}`,\n    ].join('\\n');\n\n    throw new Error(errorMessage, {\n      cause: {\n        prediction: res.content,\n        parsed,\n        unhandledActions,\n        convertedText,\n      },\n    });\n  }\n\n  debug('transformActions', JSON.stringify(transformActions, null, 2));\n  const log = getSummary(res.content);\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: log,\n  });\n\n  return {\n    actions: transformActions,\n    log,\n    usage: res.usage,\n    rawResponse: JSON.stringify(res.content, undefined, 2),\n    more_actions_needed_by_instruction: shouldContinue,\n  };\n}\n\n/**\n * Converts bounding box notation to coordinate points\n * @param text - The text containing bbox tags to be converted\n * @returns The text with bbox tags replaced by coordinate points\n */\nfunction convertBboxToCoordinates(text: string): string {\n  // Match the four numbers after <bbox>\n  const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n  function replaceMatch(\n    match: string,\n    x1: string,\n    y1: string,\n    x2: string,\n    y2: string,\n  ): string {\n    // Convert strings to numbers and calculate center point\n    const x1Num = Number.parseInt(x1, 10);\n    const y1Num = Number.parseInt(y1, 10);\n    const x2Num = Number.parseInt(x2, 10);\n    const y2Num = Number.parseInt(y2, 10);\n\n    // Use Math.floor to truncate and calculate center point\n    const x = Math.floor((x1Num + x2Num) / 2);\n    const y = Math.floor((y1Num + y2Num) / 2);\n\n    // Return formatted coordinate string\n    return `(${x},${y})`;\n  }\n\n  // Remove [EOS] and replace <bbox> coordinates\n  const cleanedText = text.replace(/\\[EOS\\]/g, '');\n  return cleanedText.replace(pattern, replaceMatch).trim();\n}\n\nfunction getPoint(startBox: string, size: { width: number; height: number }) {\n  const [x, y] = JSON.parse(startBox);\n  return [x * size.width, y * size.height];\n}\n\ninterface BaseAction {\n  action_type: ActionType;\n  action_inputs: Record<string, any>;\n  reflection: string | null;\n  thought: string | null;\n}\n\ninterface ClickAction extends BaseAction {\n  action_type: 'click';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface DragAction extends BaseAction {\n  action_type: 'drag';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n    end_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface WaitAction extends BaseAction {\n  action_type: 'wait';\n  action_inputs: {\n    time: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface LeftDoubleAction extends BaseAction {\n  action_type: 'left_double';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface RightSingleAction extends BaseAction {\n  action_type: 'right_single';\n  action_inputs: {\n    start_box: 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  | LeftDoubleAction\n  | RightSingleAction\n  | DragAction\n  | TypeAction\n  | HotkeyAction\n  | ScrollAction\n  | FinishedAction\n  | WaitAction;\n\nexport async function resizeImageForUiTars(\n  imageBase64: string,\n  size: Size,\n  uiTarsVersion: UITarsModelVersion | undefined,\n) {\n  if (uiTarsVersion === UITarsModelVersion.V1_5) {\n    debug('ui-tars-v1.5, will check image size', size);\n    const currentPixels = size.width * size.height;\n    const maxPixels = 16384 * 28 * 28; //\n    if (currentPixels > maxPixels) {\n      const resizeFactor = Math.sqrt(maxPixels / currentPixels);\n      const newWidth = Math.floor(size.width * resizeFactor);\n      const newHeight = Math.floor(size.height * resizeFactor);\n      debug(\n        'resize image for ui-tars, new width: %s, new height: %s',\n        newWidth,\n        newHeight,\n      );\n      const resizedImage = await resizeImgBase64(imageBase64, {\n        width: newWidth,\n        height: newHeight,\n      });\n      return resizedImage;\n    }\n  }\n  return imageBase64;\n}\n"],"names":["debug","getDebug","bboxSize","pointToBbox","point","width","height","Math","lastLocateRecorder","LatestLocateRecorder","uiTarsPlanning","userInstruction","options","conversationHistory","context","modelConfig","actionContext","uiTarsModelVersion","instruction","systemPrompt","getUiTarsPlanningPrompt","imagePayload","resizeImageForUiTars","res","callAIWithStringResponse","convertedText","convertBboxToCoordinates","size","parsed","actionParser","JSON","transformActions","unhandledActions","shouldContinue","action","actionType","assert","getPoint","locate","startPoint","endPoint","latestLocate","source","keys","transformHotkeyInput","console","errorDetails","types","a","errorMessage","Error","log","getSummary","undefined","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","y","cleanedText","startBox","imageBase64","uiTarsVersion","UITarsModelVersion","currentPixels","maxPixels","resizeFactor","newWidth","newHeight","resizedImage","resizeImgBase64"],"mappings":";;;;;;;;;AA4BA,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;AAGH,MAAME,qBAAqB,IAAIC;AAExB,eAAeC,eACpBC,eAAuB,EACvBC,OAKC;IAED,MAAM,EAAEC,mBAAmB,EAAEC,OAAO,EAAEC,WAAW,EAAEC,aAAa,EAAE,GAAGJ;IACrE,MAAM,EAAEK,kBAAkB,EAAE,GAAGF;IAE/B,IAAIG,cAAcP;IAClB,IAAIK,eACFE,cAAc,CAAC,yBAAyB,EAAEF,cAAc,8CAA8C,EAAEL,gBAAgB,mBAAmB,CAAC;IAG9I,MAAMQ,eAAeC,4BAA4BF;IAEjD,MAAMG,eAAe,MAAMC,qBACzBR,QAAQ,gBAAgB,EACxBA,QAAQ,IAAI,EACZG;IAGFJ,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKQ;gBACP;YACF;SACD;IACH;IAEA,MAAME,MAAM,MAAMC,yBAChB;QACE;YACE,MAAM;YACN,SAASL;QACX;WACGN,oBAAoB,QAAQ;KAChC,EACDE;IAEF,MAAMU,gBAAgBC,yBAAyBH,IAAI,OAAO;IAE1D,MAAM,EAAEI,IAAI,EAAE,GAAGb;IACjB,MAAM,EAAEc,MAAM,EAAE,GAAGC,aAAa;QAC9B,YAAYJ;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOE,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUV;IACZ;IAEAjB,MACE,oBACAiB,oBACA,YACAa,KAAK,SAAS,CAACF;IAGjB,MAAMG,mBAAqC,EAAE;IAC7C,MAAMC,mBAA6D,EAAE;IACrE,IAAIC,iBAAiB;IACrBL,OAAO,OAAO,CAAC,CAACM;QACd,MAAMC,aAAcD,AAAAA,CAAAA,OAAO,WAAW,IAAI,EAAC,EAAG,WAAW;QACzD,IAAIC,AAAe,YAAfA,YAAwB;YAC1BC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM9B,QAAQiC,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAEvD,MAAMW,SAAS;gBACb,QAAQJ,OAAO,OAAO,IAAI;gBAC1B,MAAM/B,YACJ;oBAAE,GAAGC,KAAK,CAAC,EAAE;oBAAE,GAAGA,KAAK,CAAC,EAAE;gBAAC,GAC3BuB,KAAK,KAAK,EACVA,KAAK,MAAM;YAEf;YAEAnB,mBAAmB,YAAY,CAAC8B,QAAQ;YACxCP,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQO;gBACV;YACF;QACF,OAAO,IAAIH,AAAe,kBAAfA,YAA8B;YACvCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM9B,QAAQiC,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAEvD,MAAMW,SAAS;gBACb,QAAQJ,OAAO,OAAO,IAAI;gBAC1B,MAAM/B,YACJ;oBAAE,GAAGC,KAAK,CAAC,EAAE;oBAAE,GAAGA,KAAK,CAAC,EAAE;gBAAC,GAC3BuB,KAAK,KAAK,EACVA,KAAK,MAAM;YAEf;YAEAnB,mBAAmB,YAAY,CAAC8B,QAAQ;YACxCP,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQO;gBACV;gBACA,SAASJ,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,mBAAfA,YAA+B;YACxCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM9B,QAAQiC,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAEvD,MAAMW,SAAS;gBACb,QAAQJ,OAAO,OAAO,IAAI;gBAC1B,MAAM/B,YACJ;oBAAE,GAAGC,KAAK,CAAC,EAAE;oBAAE,GAAGA,KAAK,CAAC,EAAE;gBAAC,GAC3BuB,KAAK,KAAK,EACVA,KAAK,MAAM;YAEf;YAEAnB,mBAAmB,YAAY,CAAC8B,QAAQ;YACxCP,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQO;gBACV;gBACA,SAASJ,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YAAuB;YAChCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCE,OAAOF,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMK,aAAaF,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAC5D,MAAMa,WAAWH,SAASH,OAAO,aAAa,CAAC,OAAO,EAAEP;YACxDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,MAAM;wBACJ,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM/B,YACJ;4BAAE,GAAGoC,UAAU,CAAC,EAAE;4BAAE,GAAGA,UAAU,CAAC,EAAE;wBAAC,GACrCZ,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;oBACA,IAAI;wBACF,QAAQO,OAAO,OAAO,IAAI;wBAC1B,MAAM/B,YACJ;4BAAE,GAAGqC,QAAQ,CAAC,EAAE;4BAAE,GAAGA,QAAQ,CAAC,EAAE;wBAAC,GACjCb,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASO,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YAAuB;YAChC,MAAM,EAAE,QAAQM,YAAY,EAAEC,MAAM,EAAE,GACpClC,mBAAmB,eAAe;YACpCR,MACE,CAAC,sBAAsB,EAAE0C,OAAO,qBAAqB,CAAC,EACtDD;YAGFV,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,OAAOG,OAAO,aAAa,CAAC,OAAO;oBACnC,QAAQO;gBACV;gBACA,SAASP,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,aAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWG,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,eAAfA,YAA2B;YACpCF,iBAAiB;YACjBF,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO,CAAC;gBACR,SAASG,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,aAAfA,YACT,IAAKD,OAAO,aAAa,CAAC,GAAG,EAItB;YACL,MAAMS,OAAOC,qBAAqBV,OAAO,aAAa,CAAC,GAAG;YAE1DH,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,SAASY,KAAK,IAAI,CAAC;gBACrB;gBACA,SAAST,OAAO,OAAO,IAAI;YAC7B;QACF,OAbEW,QAAQ,IAAI,CACV;aAaC,IAAIV,AAAe,WAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,SAASG,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,YAAY;YAErBH,iBAAiB,IAAI,CAAC;gBACpB,MAAMG;gBACN,SAASD,OAAO,OAAO,IAAI;YAC7B;YACAlC,MAAM,0BAA0BmC,YAAY,YAAYD,OAAO,OAAO;QACxE;IACF;IAEA,IAAIH,AAA4B,MAA5BA,iBAAiB,MAAM,EAAQ;QACjC,MAAMe,eAAyB,EAAE;QAGjC,IAAIlB,AAAkB,MAAlBA,OAAO,MAAM,EAAQ;YACvBkB,aAAa,IAAI,CAAC;YAGlB,IACEvB,IAAI,OAAO,CAAC,QAAQ,CAAC,eACrB,CAACA,IAAI,OAAO,CAAC,QAAQ,CAAC,YAEtBuB,aAAa,IAAI,CACf;iBAGFA,aAAa,IAAI,CAAC;QAEtB;QAGA,IAAId,iBAAiB,MAAM,GAAG,GAAG;YAC/B,MAAMe,QAAQf,iBAAiB,GAAG,CAAC,CAACgB,IAAMA,EAAE,IAAI,EAAE,IAAI,CAAC;YACvDF,aAAa,IAAI,CAAC,CAAC,wBAAwB,EAAEC,OAAO;QACtD;QAEA,MAAME,eAAe;YACnB;eACGH;YACH,CAAC,gBAAgB,EAAEvB,IAAI,OAAO,EAAE;SACjC,CAAC,IAAI,CAAC;QAEP,MAAM,IAAI2B,MAAMD,cAAc;YAC5B,OAAO;gBACL,YAAY1B,IAAI,OAAO;gBACvBK;gBACAI;gBACAP;YACF;QACF;IACF;IAEAzB,MAAM,oBAAoB8B,KAAK,SAAS,CAACC,kBAAkB,MAAM;IACjE,MAAMoB,MAAMC,WAAW7B,IAAI,OAAO;IAElCV,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAASsC;IACX;IAEA,OAAO;QACL,SAASpB;QACToB;QACA,OAAO5B,IAAI,KAAK;QAChB,aAAaO,KAAK,SAAS,CAACP,IAAI,OAAO,EAAE8B,QAAW;QACpD,oCAAoCpB;IACtC;AACF;AAOA,SAASP,yBAAyB4B,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,IAAI5D,KAAK,KAAK,CAAEuD,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMG,IAAI7D,KAAK,KAAK,CAAEyD,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,SAASnB,SAASiC,QAAgB,EAAE3C,IAAuC;IACzE,MAAM,CAACwC,GAAGC,EAAE,GAAGtC,KAAK,KAAK,CAACwC;IAC1B,OAAO;QAACH,IAAIxC,KAAK,KAAK;QAAEyC,IAAIzC,KAAK,MAAM;KAAC;AAC1C;AAkFO,eAAeL,qBACpBiD,WAAmB,EACnB5C,IAAU,EACV6C,aAA6C;IAE7C,IAAIA,kBAAkBC,mBAAmB,IAAI,EAAE;QAC7CzE,MAAM,uCAAuC2B;QAC7C,MAAM+C,gBAAgB/C,KAAK,KAAK,GAAGA,KAAK,MAAM;QAC9C,MAAMgD,YAAY;QAClB,IAAID,gBAAgBC,WAAW;YAC7B,MAAMC,eAAerE,KAAK,IAAI,CAACoE,YAAYD;YAC3C,MAAMG,WAAWtE,KAAK,KAAK,CAACoB,KAAK,KAAK,GAAGiD;YACzC,MAAME,YAAYvE,KAAK,KAAK,CAACoB,KAAK,MAAM,GAAGiD;YAC3C5E,MACE,2DACA6E,UACAC;YAEF,MAAMC,eAAe,MAAMC,gBAAgBT,aAAa;gBACtD,OAAOM;gBACP,QAAQC;YACV;YACA,OAAOC;QACT;IACF;IACA,OAAOR;AACT"}

package/dist/es/service/index.mjs

@@ -1,3 +1,4 @@
+import { isAutoGLM } from "../ai-model/auto-glm/util.mjs";
 import { AiExtractElementInfo, AiLocateElement, callAIWithObjectResponse } from "../ai-model/index.mjs";
 import { AiLocateSection } from "../ai-model/inspect.mjs";
 import { elementDescriberInstruction } from "../ai-model/prompt/describe.mjs";
@@ -33,6 +34,10 @@ class Service {
             console.warn('The "deepThink" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/model-config');
             searchAreaPrompt = void 0;
         }
+        if (searchAreaPrompt && isAutoGLM(vlMode)) {
+            console.warn('The "deepThink" feature is not supported with AutoGLM.');
+            searchAreaPrompt = void 0;
+        }
         const context = opt?.context || await this.contextRetrieverFn();
         let searchArea;
         let searchAreaRawResponse;

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

@@ -1 +1 @@
-{"version":3,"file":"service/index.mjs","sources":["../../../src/service/index.ts"],"sourcesContent":["import {\n  AiExtractElementInfo,\n  AiLocateElement,\n  callAIWithObjectResponse,\n} from '@/ai-model/index';\nimport { AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport { type AIArgs, expandSearchArea } from '@/common';\nimport type {\n  AIDescribeElementResponse,\n  AIUsageInfo,\n  DetailedLocateParam,\n  LocateResultWithDump,\n  PartialServiceDumpFromSDK,\n  Rect,\n  ServiceExtractOption,\n  ServiceExtractParam,\n  ServiceExtractResult,\n  ServiceTaskInfo,\n  UIContext,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport {\n  type IModelConfig,\n  MIDSCENE_FORCE_DEEP_THINK,\n  globalConfigManager,\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 '../common';\nimport { createServiceDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\ninterface ServiceOptions {\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n  aiVendorFn?: typeof callAIWithObjectResponse;\n}\n\nconst debug = getDebug('ai:service');\nexport default class Service {\n  contextRetrieverFn: () => Promise<UIContext> | UIContext;\n\n  aiVendorFn: Exclude<ServiceOptions['aiVendorFn'], undefined> =\n    callAIWithObjectResponse;\n\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n\n  constructor(\n    context: UIContext | (() => Promise<UIContext> | UIContext),\n    opt?: ServiceOptions,\n  ) {\n    assert(context, 'context is required for Service');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    // just for unit test, aiVendorFn is callAIWithObjectResponse by default\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    modelConfig: IModelConfig,\n  ): Promise<LocateResultWithDump> {\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    const globalDeepThinkSwitch = globalConfigManager.getEnvConfigInBoolean(\n      MIDSCENE_FORCE_DEEP_THINK,\n    );\n    if (globalDeepThinkSwitch) {\n      debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n    }\n    let searchAreaPrompt;\n    if (query.deepThink || globalDeepThinkSwitch) {\n      searchAreaPrompt = query.prompt;\n    }\n\n    const { vlMode } = modelConfig;\n\n    if (searchAreaPrompt && !vlMode) {\n      console.warn(\n        'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/model-config',\n      );\n      searchAreaPrompt = undefined;\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn());\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        modelConfig,\n      });\n      assert(\n        searchAreaResponse.rect,\n        `cannot find search area for \"${searchAreaPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const { parseResult, rect, rawResponse, usage, reasoning_content } =\n      await AiLocateElement({\n        callAIFn: this.aiVendorFn,\n        context,\n        targetElementDescription: queryPrompt,\n        searchConfig: searchAreaResponse,\n        modelConfig,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `failed to locate element: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\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 = parseResult.elements || [];\n\n    const dump = createServiceDump({\n      ...dumpData,\n      matchedElement: elements,\n    });\n\n    if (errorLog) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    if (elements.length > 1) {\n      throw new ServiceError(\n        `locate: multiple elements found, length = ${elements.length}`,\n        dump,\n      );\n    }\n\n    if (elements.length === 1) {\n      return {\n        element: {\n          center: elements[0]!.center,\n          rect: elements[0]!.rect,\n          description: elements[0]!.description,\n        },\n        rect,\n        dump,\n      };\n    }\n\n    return {\n      element: null,\n      rect,\n      dump,\n    };\n  }\n\n  async extract<T>(\n    dataDemand: ServiceExtractParam,\n    modelConfig: IModelConfig,\n    opt?: ServiceExtractOption,\n    pageDescription?: string,\n    multimodalPrompt?: TMultimodalPrompt,\n  ): Promise<ServiceExtractResult<T>> {\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n    const context = await this.contextRetrieverFn();\n\n    const startTime = Date.now();\n\n    const { parseResult, usage, reasoning_content } =\n      await AiExtractElementInfo<T>({\n        context,\n        dataQuery: dataDemand,\n        multimodalPrompt,\n        extractOption: opt,\n        modelConfig,\n        pageDescription,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(parseResult),\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\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    const dump = createServiceDump({\n      ...dumpData,\n      data,\n    });\n\n    if (errorLog && !data) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    return {\n      data,\n      thought,\n      usage,\n      reasoning_content,\n      dump,\n    };\n  }\n\n  async describe(\n    target: Rect | [number, number],\n    modelConfig: IModelConfig,\n    opt?: {\n      deepThink?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for service.describe');\n    const context = await this.contextRetrieverFn();\n    const { screenshotBase64, size } = context;\n    assert(screenshotBase64, 'screenshot is required for service.describe');\n    // The result of the \"describe\" function will be used for positioning, so essentially it is a form of grounding.\n    const { vlMode } = modelConfig;\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      size,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepThink) {\n      const searchArea = expandSearchArea(targetRect, context.size, vlMode);\n      debug('describe: set searchArea', searchArea);\n      const croppedResult = await cropByRect(\n        imagePayload,\n        searchArea,\n        vlMode === 'qwen2.5-vl',\n      );\n      imagePayload = croppedResult.imageBase64;\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 = this\n      .aiVendorFn as typeof callAIWithObjectResponse<AIDescribeElementResponse>;\n\n    const res = await callAIFn(msgs, modelConfig);\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","Service","query","opt","modelConfig","queryPrompt","assert","globalDeepThinkSwitch","globalConfigManager","MIDSCENE_FORCE_DEEP_THINK","searchAreaPrompt","vlMode","console","undefined","context","searchArea","searchAreaRawResponse","searchAreaUsage","searchAreaResponse","AiLocateSection","startTime","Date","parseResult","rect","rawResponse","usage","reasoning_content","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","elements","dump","createServiceDump","ServiceError","dataDemand","pageDescription","multimodalPrompt","AiExtractElementInfo","data","thought","target","screenshotBase64","size","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","croppedResult","cropByRect","msgs","callAIFn","res","content","callAIWithObjectResponse","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;;AA8CA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IA4BnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,WAAyB,EACM;QAC/B,MAAMC,cAAc,AAAiB,YAAjB,OAAOH,QAAqBA,QAAQA,MAAM,MAAM;QACpEI,OAAOD,aAAa;QAEpBC,OAAO,AAAiB,YAAjB,OAAOJ,OAAoB;QAElC,MAAMK,wBAAwBC,oBAAoB,qBAAqB,CACrEC;QAEF,IAAIF,uBACFR,MAAM,yBAAyBQ;QAEjC,IAAIG;QACJ,IAAIR,MAAM,SAAS,IAAIK,uBACrBG,mBAAmBR,MAAM,MAAM;QAGjC,MAAM,EAAES,MAAM,EAAE,GAAGP;QAEnB,IAAIM,oBAAoB,CAACC,QAAQ;YAC/BC,QAAQ,IAAI,CACV;YAEFF,mBAAmBG;QACrB;QAEA,MAAMC,UAAUX,KAAK,WAAY,MAAM,IAAI,CAAC,kBAAkB;QAE9D,IAAIY;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAGJ,IAAIR,kBAAkB;YACpBQ,qBAAqB,MAAMC,gBAAgB;gBACzCL;gBACA,oBAAoBJ;gBACpBN;YACF;YACAE,OACEY,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,EAAEC,WAAW,EAAEC,IAAI,EAAEC,WAAW,EAAEC,KAAK,EAAEC,iBAAiB,EAAE,GAChE,MAAMC,gBAAgB;YACpB,UAAU,IAAI,CAAC,UAAU;YACzBb;YACA,0BAA0BT;YAC1B,cAAca;YACdd;QACF;QAEF,MAAMwB,WAAWP,KAAK,GAAG,KAAKD;QAC9B,MAAMS,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACN;YAC5B,gBAAgBM,KAAK,SAAS,CAACR;YAC/BG;YACAV;YACAC;YACAC;YACAS;QACF;QAEA,IAAIK;QACJ,IAAIT,YAAY,MAAM,EAAE,QACtBS,WAAW,CAAC,4BAA4B,EAAET,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG3E,MAAMU,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAAS3B;YACX;YACA,gBAAgB,EAAE;YAClB,aAAakB;YACb,MAAM;YACNM;YACA,WAAW,CAAC,CAACd;YACb,OAAOgB;QACT;QAEA,MAAME,WAAWX,YAAY,QAAQ,IAAI,EAAE;QAE3C,MAAMY,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACX,gBAAgBC;QAClB;QAEA,IAAIF,UACF,MAAM,IAAIK,aAAaL,UAAUG;QAGnC,IAAID,SAAS,MAAM,GAAG,GACpB,MAAM,IAAIG,aACR,CAAC,0CAA0C,EAAEH,SAAS,MAAM,EAAE,EAC9DC;QAIJ,IAAID,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,SAAS;gBACP,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM;gBAC3B,MAAMA,QAAQ,CAAC,EAAE,CAAE,IAAI;gBACvB,aAAaA,QAAQ,CAAC,EAAE,CAAE,WAAW;YACvC;YACAV;YACAW;QACF;QAGF,OAAO;YACL,SAAS;YACTX;YACAW;QACF;IACF;IAEA,MAAM,QACJG,UAA+B,EAC/BjC,WAAyB,EACzBD,GAA0B,EAC1BmC,eAAwB,EACxBC,gBAAoC,EACF;QAClCjC,OACE,AAAsB,YAAtB,OAAO+B,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAEvE,MAAMvB,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAE7C,MAAMM,YAAYC,KAAK,GAAG;QAE1B,MAAM,EAAEC,WAAW,EAAEG,KAAK,EAAEC,iBAAiB,EAAE,GAC7C,MAAMc,qBAAwB;YAC5B1B;YACA,WAAWuB;YACXE;YACA,eAAepC;YACfC;YACAkC;QACF;QAEF,MAAMV,WAAWP,KAAK,GAAG,KAAKD;QAC9B,MAAMS,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACR;YAC5BI;QACF;QAEA,IAAIK;QACJ,IAAIT,YAAY,MAAM,EAAE,QACtBS,WAAW,CAAC,qBAAqB,EAAET,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMU,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTK;YACF;YACA,gBAAgB,EAAE;YAClB,MAAM;YACNR;YACA,OAAOE;QACT;QAEA,MAAM,EAAEU,IAAI,EAAEC,OAAO,EAAE,GAAGpB,eAAe,CAAC;QAG1C,MAAMY,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACXS;QACF;QAEA,IAAIV,YAAY,CAACU,MACf,MAAM,IAAIL,aAAaL,UAAUG;QAGnC,OAAO;YACLO;YACAC;YACAjB;YACAC;YACAQ;QACF;IACF;IAEA,MAAM,SACJS,MAA+B,EAC/BvC,WAAyB,EACzBD,GAEC,EACwD;QACzDG,OAAOqC,QAAQ;QACf,MAAM7B,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAC7C,MAAM,EAAE8B,gBAAgB,EAAEC,IAAI,EAAE,GAAG/B;QACnCR,OAAOsC,kBAAkB;QAEzB,MAAM,EAAEjC,MAAM,EAAE,GAAGP;QACnB,MAAM0C,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,IAAI9C,KAAK,WAAW;YAClB,MAAMY,aAAauC,iBAAiBL,YAAYnC,QAAQ,IAAI,EAAEH;YAC9DZ,MAAM,4BAA4BgB;YAClC,MAAMwC,gBAAgB,MAAMC,WAC1BJ,cACArC,YACAJ,AAAW,iBAAXA;YAEFyC,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,WAAW,IAAI,CAClB,UAAU;QAEb,MAAMC,MAAM,MAAMD,SAASD,MAAMrD;QAEjC,MAAM,EAAEwD,OAAO,EAAE,GAAGD;QACpBrD,OAAO,CAACsD,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1DtD,OAAOsD,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IAxSA,YACE9C,OAA2D,EAC3DX,GAAoB,CACpB;QAVF;QAEA,qCACE0D;QAEF;QAMEvD,OAAOQ,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMgD,QAAQ,OAAO,CAAChD;QAIlD,IAAI,AAA2B,WAApBX,KAAK,YACd,IAAI,CAAC,UAAU,GAAGA,IAAI,UAAU;QAElC,IAAI,AAAyB,WAAlBA,KAAK,UACd,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AAuRF"}
+{"version":3,"file":"service/index.mjs","sources":["../../../src/service/index.ts"],"sourcesContent":["import { isAutoGLM } from '@/ai-model/auto-glm/util';\nimport {\n  AiExtractElementInfo,\n  AiLocateElement,\n  callAIWithObjectResponse,\n} from '@/ai-model/index';\nimport { AiLocateSection } from '@/ai-model/inspect';\nimport { elementDescriberInstruction } from '@/ai-model/prompt/describe';\nimport { type AIArgs, expandSearchArea } from '@/common';\nimport type {\n  AIDescribeElementResponse,\n  AIUsageInfo,\n  DetailedLocateParam,\n  LocateResultWithDump,\n  PartialServiceDumpFromSDK,\n  Rect,\n  ServiceExtractOption,\n  ServiceExtractParam,\n  ServiceExtractResult,\n  ServiceTaskInfo,\n  UIContext,\n} from '@/types';\nimport { ServiceError } from '@/types';\nimport {\n  type IModelConfig,\n  MIDSCENE_FORCE_DEEP_THINK,\n  globalConfigManager,\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 '../common';\nimport { createServiceDump } from './utils';\n\nexport interface LocateOpts {\n  context?: UIContext;\n}\n\nexport type AnyValue<T> = {\n  [K in keyof T]: unknown extends T[K] ? any : T[K];\n};\n\ninterface ServiceOptions {\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n  aiVendorFn?: typeof callAIWithObjectResponse;\n}\n\nconst debug = getDebug('ai:service');\nexport default class Service {\n  contextRetrieverFn: () => Promise<UIContext> | UIContext;\n\n  aiVendorFn: Exclude<ServiceOptions['aiVendorFn'], undefined> =\n    callAIWithObjectResponse;\n\n  taskInfo?: Omit<ServiceTaskInfo, 'durationMs'>;\n\n  constructor(\n    context: UIContext | (() => Promise<UIContext> | UIContext),\n    opt?: ServiceOptions,\n  ) {\n    assert(context, 'context is required for Service');\n    if (typeof context === 'function') {\n      this.contextRetrieverFn = context;\n    } else {\n      this.contextRetrieverFn = () => Promise.resolve(context);\n    }\n\n    // just for unit test, aiVendorFn is callAIWithObjectResponse by default\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    modelConfig: IModelConfig,\n  ): Promise<LocateResultWithDump> {\n    const queryPrompt = typeof query === 'string' ? query : query.prompt;\n    assert(queryPrompt, 'query is required for locate');\n\n    assert(typeof query === 'object', 'query should be an object for locate');\n\n    const globalDeepThinkSwitch = globalConfigManager.getEnvConfigInBoolean(\n      MIDSCENE_FORCE_DEEP_THINK,\n    );\n    if (globalDeepThinkSwitch) {\n      debug('globalDeepThinkSwitch', globalDeepThinkSwitch);\n    }\n    let searchAreaPrompt;\n    if (query.deepThink || globalDeepThinkSwitch) {\n      searchAreaPrompt = query.prompt;\n    }\n\n    const { vlMode } = modelConfig;\n\n    if (searchAreaPrompt && !vlMode) {\n      console.warn(\n        'The \"deepThink\" feature is not supported with multimodal LLM. Please config VL model for Midscene. https://midscenejs.com/model-config',\n      );\n      searchAreaPrompt = undefined;\n    }\n\n    if (searchAreaPrompt && isAutoGLM(vlMode)) {\n      console.warn('The \"deepThink\" feature is not supported with AutoGLM.');\n      searchAreaPrompt = undefined;\n    }\n\n    const context = opt?.context || (await this.contextRetrieverFn());\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        modelConfig,\n      });\n      assert(\n        searchAreaResponse.rect,\n        `cannot find search area for \"${searchAreaPrompt}\"${\n          searchAreaResponse.error ? `: ${searchAreaResponse.error}` : ''\n        }`,\n      );\n      searchAreaRawResponse = searchAreaResponse.rawResponse;\n      searchAreaUsage = searchAreaResponse.usage;\n      searchArea = searchAreaResponse.rect;\n    }\n\n    const startTime = Date.now();\n    const { parseResult, rect, rawResponse, usage, reasoning_content } =\n      await AiLocateElement({\n        callAIFn: this.aiVendorFn,\n        context,\n        targetElementDescription: queryPrompt,\n        searchConfig: searchAreaResponse,\n        modelConfig,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(rawResponse),\n      formatResponse: JSON.stringify(parseResult),\n      usage,\n      searchArea,\n      searchAreaRawResponse,\n      searchAreaUsage,\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `failed to locate element: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\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 = parseResult.elements || [];\n\n    const dump = createServiceDump({\n      ...dumpData,\n      matchedElement: elements,\n    });\n\n    if (errorLog) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    if (elements.length > 1) {\n      throw new ServiceError(\n        `locate: multiple elements found, length = ${elements.length}`,\n        dump,\n      );\n    }\n\n    if (elements.length === 1) {\n      return {\n        element: {\n          center: elements[0]!.center,\n          rect: elements[0]!.rect,\n          description: elements[0]!.description,\n        },\n        rect,\n        dump,\n      };\n    }\n\n    return {\n      element: null,\n      rect,\n      dump,\n    };\n  }\n\n  async extract<T>(\n    dataDemand: ServiceExtractParam,\n    modelConfig: IModelConfig,\n    opt?: ServiceExtractOption,\n    pageDescription?: string,\n    multimodalPrompt?: TMultimodalPrompt,\n  ): Promise<ServiceExtractResult<T>> {\n    assert(\n      typeof dataDemand === 'object' || typeof dataDemand === 'string',\n      `dataDemand should be object or string, but get ${typeof dataDemand}`,\n    );\n    const context = await this.contextRetrieverFn();\n\n    const startTime = Date.now();\n\n    const { parseResult, usage, reasoning_content } =\n      await AiExtractElementInfo<T>({\n        context,\n        dataQuery: dataDemand,\n        multimodalPrompt,\n        extractOption: opt,\n        modelConfig,\n        pageDescription,\n      });\n\n    const timeCost = Date.now() - startTime;\n    const taskInfo: ServiceTaskInfo = {\n      ...(this.taskInfo ? this.taskInfo : {}),\n      durationMs: timeCost,\n      rawResponse: JSON.stringify(parseResult),\n      reasoning_content,\n    };\n\n    let errorLog: string | undefined;\n    if (parseResult.errors?.length) {\n      errorLog = `AI response error: \\n${parseResult.errors.join('\\n')}`;\n    }\n\n    const dumpData: PartialServiceDumpFromSDK = {\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    const dump = createServiceDump({\n      ...dumpData,\n      data,\n    });\n\n    if (errorLog && !data) {\n      throw new ServiceError(errorLog, dump);\n    }\n\n    return {\n      data,\n      thought,\n      usage,\n      reasoning_content,\n      dump,\n    };\n  }\n\n  async describe(\n    target: Rect | [number, number],\n    modelConfig: IModelConfig,\n    opt?: {\n      deepThink?: boolean;\n    },\n  ): Promise<Pick<AIDescribeElementResponse, 'description'>> {\n    assert(target, 'target is required for service.describe');\n    const context = await this.contextRetrieverFn();\n    const { screenshotBase64, size } = context;\n    assert(screenshotBase64, 'screenshot is required for service.describe');\n    // The result of the \"describe\" function will be used for positioning, so essentially it is a form of grounding.\n    const { vlMode } = modelConfig;\n    const systemPrompt = elementDescriberInstruction();\n\n    // Convert [x,y] center point to Rect if needed\n    const defaultRectSize = 30;\n    const targetRect: Rect = Array.isArray(target)\n      ? {\n          left: Math.floor(target[0] - defaultRectSize / 2),\n          top: Math.floor(target[1] - defaultRectSize / 2),\n          width: defaultRectSize,\n          height: defaultRectSize,\n        }\n      : target;\n\n    let imagePayload = await compositeElementInfoImg({\n      inputImgBase64: screenshotBase64,\n      size,\n      elementsPositionInfo: [\n        {\n          rect: targetRect,\n        },\n      ],\n      borderThickness: 3,\n    });\n\n    if (opt?.deepThink) {\n      const searchArea = expandSearchArea(targetRect, context.size, vlMode);\n      debug('describe: set searchArea', searchArea);\n      const croppedResult = await cropByRect(\n        imagePayload,\n        searchArea,\n        vlMode === 'qwen2.5-vl',\n      );\n      imagePayload = croppedResult.imageBase64;\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 = this\n      .aiVendorFn as typeof callAIWithObjectResponse<AIDescribeElementResponse>;\n\n    const res = await callAIFn(msgs, modelConfig);\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","Service","query","opt","modelConfig","queryPrompt","assert","globalDeepThinkSwitch","globalConfigManager","MIDSCENE_FORCE_DEEP_THINK","searchAreaPrompt","vlMode","console","undefined","isAutoGLM","context","searchArea","searchAreaRawResponse","searchAreaUsage","searchAreaResponse","AiLocateSection","startTime","Date","parseResult","rect","rawResponse","usage","reasoning_content","AiLocateElement","timeCost","taskInfo","JSON","errorLog","dumpData","elements","dump","createServiceDump","ServiceError","dataDemand","pageDescription","multimodalPrompt","AiExtractElementInfo","data","thought","target","screenshotBase64","size","systemPrompt","elementDescriberInstruction","defaultRectSize","targetRect","Array","Math","imagePayload","compositeElementInfoImg","expandSearchArea","croppedResult","cropByRect","msgs","callAIFn","res","content","callAIWithObjectResponse","Promise"],"mappings":";;;;;;;;;;;;;;;;;;;;;AA+CA,MAAMA,QAAQC,SAAS;AACR,MAAMC;IA4BnB,MAAM,OACJC,KAA0B,EAC1BC,GAAe,EACfC,WAAyB,EACM;QAC/B,MAAMC,cAAc,AAAiB,YAAjB,OAAOH,QAAqBA,QAAQA,MAAM,MAAM;QACpEI,OAAOD,aAAa;QAEpBC,OAAO,AAAiB,YAAjB,OAAOJ,OAAoB;QAElC,MAAMK,wBAAwBC,oBAAoB,qBAAqB,CACrEC;QAEF,IAAIF,uBACFR,MAAM,yBAAyBQ;QAEjC,IAAIG;QACJ,IAAIR,MAAM,SAAS,IAAIK,uBACrBG,mBAAmBR,MAAM,MAAM;QAGjC,MAAM,EAAES,MAAM,EAAE,GAAGP;QAEnB,IAAIM,oBAAoB,CAACC,QAAQ;YAC/BC,QAAQ,IAAI,CACV;YAEFF,mBAAmBG;QACrB;QAEA,IAAIH,oBAAoBI,UAAUH,SAAS;YACzCC,QAAQ,IAAI,CAAC;YACbF,mBAAmBG;QACrB;QAEA,MAAME,UAAUZ,KAAK,WAAY,MAAM,IAAI,CAAC,kBAAkB;QAE9D,IAAIa;QACJ,IAAIC;QACJ,IAAIC;QACJ,IAAIC;QAGJ,IAAIT,kBAAkB;YACpBS,qBAAqB,MAAMC,gBAAgB;gBACzCL;gBACA,oBAAoBL;gBACpBN;YACF;YACAE,OACEa,mBAAmB,IAAI,EACvB,CAAC,6BAA6B,EAAET,iBAAiB,CAAC,EAChDS,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,EAAEC,WAAW,EAAEC,IAAI,EAAEC,WAAW,EAAEC,KAAK,EAAEC,iBAAiB,EAAE,GAChE,MAAMC,gBAAgB;YACpB,UAAU,IAAI,CAAC,UAAU;YACzBb;YACA,0BAA0BV;YAC1B,cAAcc;YACdf;QACF;QAEF,MAAMyB,WAAWP,KAAK,GAAG,KAAKD;QAC9B,MAAMS,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACN;YAC5B,gBAAgBM,KAAK,SAAS,CAACR;YAC/BG;YACAV;YACAC;YACAC;YACAS;QACF;QAEA,IAAIK;QACJ,IAAIT,YAAY,MAAM,EAAE,QACtBS,WAAW,CAAC,4BAA4B,EAAET,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAG3E,MAAMU,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACT,SAAS5B;YACX;YACA,gBAAgB,EAAE;YAClB,aAAamB;YACb,MAAM;YACNM;YACA,WAAW,CAAC,CAACd;YACb,OAAOgB;QACT;QAEA,MAAME,WAAWX,YAAY,QAAQ,IAAI,EAAE;QAE3C,MAAMY,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACX,gBAAgBC;QAClB;QAEA,IAAIF,UACF,MAAM,IAAIK,aAAaL,UAAUG;QAGnC,IAAID,SAAS,MAAM,GAAG,GACpB,MAAM,IAAIG,aACR,CAAC,0CAA0C,EAAEH,SAAS,MAAM,EAAE,EAC9DC;QAIJ,IAAID,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,SAAS;gBACP,QAAQA,QAAQ,CAAC,EAAE,CAAE,MAAM;gBAC3B,MAAMA,QAAQ,CAAC,EAAE,CAAE,IAAI;gBACvB,aAAaA,QAAQ,CAAC,EAAE,CAAE,WAAW;YACvC;YACAV;YACAW;QACF;QAGF,OAAO;YACL,SAAS;YACTX;YACAW;QACF;IACF;IAEA,MAAM,QACJG,UAA+B,EAC/BlC,WAAyB,EACzBD,GAA0B,EAC1BoC,eAAwB,EACxBC,gBAAoC,EACF;QAClClC,OACE,AAAsB,YAAtB,OAAOgC,cAA2B,AAAsB,YAAtB,OAAOA,YACzC,CAAC,+CAA+C,EAAE,OAAOA,YAAY;QAEvE,MAAMvB,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAE7C,MAAMM,YAAYC,KAAK,GAAG;QAE1B,MAAM,EAAEC,WAAW,EAAEG,KAAK,EAAEC,iBAAiB,EAAE,GAC7C,MAAMc,qBAAwB;YAC5B1B;YACA,WAAWuB;YACXE;YACA,eAAerC;YACfC;YACAmC;QACF;QAEF,MAAMV,WAAWP,KAAK,GAAG,KAAKD;QAC9B,MAAMS,WAA4B;YAChC,GAAI,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,GAAG,CAAC,CAAC;YACtC,YAAYD;YACZ,aAAaE,KAAK,SAAS,CAACR;YAC5BI;QACF;QAEA,IAAIK;QACJ,IAAIT,YAAY,MAAM,EAAE,QACtBS,WAAW,CAAC,qBAAqB,EAAET,YAAY,MAAM,CAAC,IAAI,CAAC,OAAO;QAGpE,MAAMU,WAAsC;YAC1C,MAAM;YACN,WAAW;gBACTK;YACF;YACA,gBAAgB,EAAE;YAClB,MAAM;YACNR;YACA,OAAOE;QACT;QAEA,MAAM,EAAEU,IAAI,EAAEC,OAAO,EAAE,GAAGpB,eAAe,CAAC;QAG1C,MAAMY,OAAOC,kBAAkB;YAC7B,GAAGH,QAAQ;YACXS;QACF;QAEA,IAAIV,YAAY,CAACU,MACf,MAAM,IAAIL,aAAaL,UAAUG;QAGnC,OAAO;YACLO;YACAC;YACAjB;YACAC;YACAQ;QACF;IACF;IAEA,MAAM,SACJS,MAA+B,EAC/BxC,WAAyB,EACzBD,GAEC,EACwD;QACzDG,OAAOsC,QAAQ;QACf,MAAM7B,UAAU,MAAM,IAAI,CAAC,kBAAkB;QAC7C,MAAM,EAAE8B,gBAAgB,EAAEC,IAAI,EAAE,GAAG/B;QACnCT,OAAOuC,kBAAkB;QAEzB,MAAM,EAAElC,MAAM,EAAE,GAAGP;QACnB,MAAM2C,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,IAAI/C,KAAK,WAAW;YAClB,MAAMa,aAAauC,iBAAiBL,YAAYnC,QAAQ,IAAI,EAAEJ;YAC9DZ,MAAM,4BAA4BiB;YAClC,MAAMwC,gBAAgB,MAAMC,WAC1BJ,cACArC,YACAL,AAAW,iBAAXA;YAEF0C,eAAeG,cAAc,WAAW;QAC1C;QAEA,MAAME,OAAe;YACnB;gBAAE,MAAM;gBAAU,SAASX;YAAa;YACxC;gBACE,MAAM;gBACN,SAAS;oBACP;wBACE,MAAM;wBACN,WAAW;4BACT,KAAKM;4BACL,QAAQ;wBACV;oBACF;iBACD;YACH;SACD;QAED,MAAMM,WAAW,IAAI,CAClB,UAAU;QAEb,MAAMC,MAAM,MAAMD,SAASD,MAAMtD;QAEjC,MAAM,EAAEyD,OAAO,EAAE,GAAGD;QACpBtD,OAAO,CAACuD,QAAQ,KAAK,EAAE,CAAC,iBAAiB,EAAEA,QAAQ,KAAK,EAAE;QAC1DvD,OAAOuD,QAAQ,WAAW,EAAE;QAC5B,OAAOA;IACT;IA7SA,YACE9C,OAA2D,EAC3DZ,GAAoB,CACpB;QAVF;QAEA,qCACE2D;QAEF;QAMExD,OAAOS,SAAS;QAChB,IAAI,AAAmB,cAAnB,OAAOA,SACT,IAAI,CAAC,kBAAkB,GAAGA;aAE1B,IAAI,CAAC,kBAAkB,GAAG,IAAMgD,QAAQ,OAAO,CAAChD;QAIlD,IAAI,AAA2B,WAApBZ,KAAK,YACd,IAAI,CAAC,UAAU,GAAGA,IAAI,UAAU;QAElC,IAAI,AAAyB,WAAlBA,KAAK,UACd,IAAI,CAAC,QAAQ,GAAGA,IAAI,QAAQ;IAEhC;AA4RF"}