빠른 시작: Azure AI 검색에서 에이전트 검색 사용

이 빠른 시작에서는 에이전트 검색을 사용하여 Azure AI 검색에서 인덱싱된 문서와 Azure AI Foundry 모델의 Azure OpenAI에서 LLM(대규모 언어 모델)으로 구동되는 대화형 검색 환경을 만듭니다.

지식 에이전트는 복잡한 쿼리를 하위 쿼리로 분해하고, 하나 이상의 지식 원본에 대해 하위 쿼리를 실행하고, 메타데이터를 사용하여 결과를 반환하여 에이전트 검색을 오케스트레이션합니다. 기본적으로 에이전트는 원본에서 원시 콘텐츠를 출력하지만 이 빠른 시작에서는 자연어 응답 생성을 위해 응답 합성 형식을 사용합니다.

사용자 고유의 데이터를 제공할 수 있지만, 이 빠른 시작에서는 NASA의 어스 앳 나이트 전자책의 샘플 JSON 문서를 사용합니다. 문서는 우주에서 관찰 된 바와 같이 밤에 지구의 일반적인 과학 주제와 이미지를 설명합니다.

필수 조건

• 활성 구독이 있는 Azure 계정. 무료로 계정을 만듭니다.

• 의미 체계 순위가 설정된 기본 계층 이상의 Azure AI Search 서비스입니다.

• Azure AI Foundry 프로젝트 및 Azure AI Foundry 리소스. 프로젝트를 만들면 리소스가 자동으로 만들어집니다.

• Microsoft Entra ID를 사용한 키 없는 인증을 위한 Azure CLI 입니다.

액세스 구성

시작하기 전에 콘텐츠 및 작업에 액세스할 수 있는 권한이 있는지 확인합니다. 인증에는 Microsoft Entra ID를, 권한 부여에는 역할 기반 액세스에 사용하는 것이 좋습니다. 역할을 할당하려면 소유자 또는 사용자 액세스 관리자여야 합니다. 역할이 가능하지 않은 경우 대신 키 기반 인증을 사용합니다.

이 빠른 시작에 대한 액세스를 구성하려면 다음 탭을 모두 선택합니다.

엔드포인트 가져오기

각 Azure AI 검색 서비스와 Azure AI Foundry 리소스에는 리소스에 대한 네트워크 액세스를 식별하고 제공하는 고유한 URL인 엔드포인트가 있습니다. 이후 섹션에서는 이러한 엔드포인트를 지정하여 프로그래밍 방식으로 리소스에 연결합니다.

이 빠른 시작의 엔드포인트를 얻으려면 다음 탭을 모두 선택합니다.

모델 배포

에이전트 검색을 사용하려면 Azure AI Foundry 리소스에 두 개의 Azure OpenAI 모델을 배포해야 합니다.

• 텍스트-벡터 변환을 위한 포함 모델입니다. 이 빠른 시작에서는 text-embedding-3-large를 사용하지만, 모든 text-embedding 모델을 사용할 수 있습니다.

• 쿼리 계획 및 응답 생성을 위한 LLM입니다. 이 빠른 시작에서는 gpt-5-mini을 사용하지만, 사용자는 에이전트 검색에 지원되는 모든 LLM을 사용할 수 있습니다.

배포 지침은 Azure AI Foundry를 사용하여 Azure OpenAI 모델 배포를 참조하세요.

환경 설정

이 빠른 시작에 대한 콘솔 애플리케이션을 설정하려면 다음을 수행합니다.

1. 애플리케이션을 포함할 폴더 quickstart-agentic-retrieval 를 만듭니다.

2. Visual Studio Code에서 해당 폴더를 엽니다.

3. 터미널>새 터미널을 선택한 다음, 다음 명령을 실행하여 콘솔 애플리케이션을 만듭니다.

   dotnet new console
   

4. .NET용 Azure AI Search 클라이언트 라이브러리 를 설치합니다.

   dotnet add package Azure.Search.Documents --version 11.7.0-beta.7
   

5. dotenv.net 파일에서 .env 환경 변수를 로드하는 패키지를 설치합니다.

   dotnet add package dotenv.net
   

6. Microsoft Entra ID를 사용한 키 없는 인증의 경우 Azure.Identity 패키지를 설치합니다.

   dotnet add package Azure.Identity
   

7. Microsoft Entra ID를 사용한 키 없는 인증의 경우 Azure 계정에 로그인합니다. 구독이 여러 개 있는 경우 Azure AI Search 서비스 및 Azure AI Foundry 프로젝트가 포함된 구독을 선택합니다.

   az login
   

코드 실행

에이전트 검색 파이프라인을 만들고 실행하려면 다음을 수행합니다.

1. 폴더에 .env 이름이 지정된 quickstart-agentic-retrieval 파일을 만듭니다.

2. 파일에 다음 환경 변수를 .env 추가합니다.

   SEARCH_ENDPOINT = PUT-YOUR-SEARCH-SERVICE-URL-HERE
   AOAI_ENDPOINT = PUT-YOUR-AOAI-FOUNDRY-URL-HERE
   

3. Set SEARCH_ENDPOINT 및 AOAI_ENDPOINT를 엔드포인트 가져오기에서 얻은 값으로 설정합니다.

4. 다음 코드를 파일에 붙여넣습니다 Program.cs .

   using dotenv.net;
   using System.Text.Json;
   using Azure.Identity;
   using Azure.Search.Documents;
   using Azure.Search.Documents.Indexes;
   using Azure.Search.Documents.Indexes.Models;
   using Azure.Search.Documents.Models;
   using Azure.Search.Documents.Agents;
   using Azure.Search.Documents.Agents.Models;
   
   namespace AzureSearch.Quickstart
   {
       class Program
       {
           static async Task Main(string[] args)
           {
               // Load environment variables from the .env file
               // Ensure your .env file is in the same directory with the required variables
               DotEnv.Load();
   
               string searchEndpoint = Environment.GetEnvironmentVariable("SEARCH_ENDPOINT")
                   ?? throw new InvalidOperationException("SEARCH_ENDPOINT isn't set.");
               string aoaiEndpoint = Environment.GetEnvironmentVariable("AOAI_ENDPOINT")
                   ?? throw new InvalidOperationException("AOAI_ENDPOINT isn't set.");
   
               string aoaiEmbeddingModel = "text-embedding-3-large";
               string aoaiEmbeddingDeployment = "text-embedding-3-large";
               string aoaiGptModel = "gpt-5-mini";
               string aoaiGptDeployment = "gpt-5-mini";
   
               string indexName = "earth-at-night";
               string knowledgeSourceName = "earth-knowledge-source";
               string knowledgeAgentName = "earth-knowledge-agent";
   
               var credential = new DefaultAzureCredential();
   
               // Define fields for the index
               var fields = new List<SearchField>
               {
                   new SimpleField("id", SearchFieldDataType.String) { IsKey = true, IsFilterable = true, IsSortable = true, IsFacetable = true },
                   new SearchField("page_chunk", SearchFieldDataType.String) { IsFilterable = false, IsSortable = false, IsFacetable = false },
                   new SearchField("page_embedding_text_3_large", SearchFieldDataType.Collection(SearchFieldDataType.Single)) { VectorSearchDimensions = 3072, VectorSearchProfileName = "hnsw_text_3_large" },
                   new SimpleField("page_number", SearchFieldDataType.Int32) { IsFilterable = true, IsSortable = true, IsFacetable = true }
               };
   
               // Define a vectorizer
               var vectorizer = new AzureOpenAIVectorizer(vectorizerName: "azure_openai_text_3_large")
               {
                   Parameters = new AzureOpenAIVectorizerParameters
                   {
                       ResourceUri = new Uri(aoaiEndpoint),
                       DeploymentName = aoaiEmbeddingDeployment,
                       ModelName = aoaiEmbeddingModel
                   }
               };
   
               // Define a vector search profile and algorithm
               var vectorSearch = new VectorSearch()
               {
                   Profiles =
                   {
                       new VectorSearchProfile(
                           name: "hnsw_text_3_large",
                           algorithmConfigurationName: "alg"
                       )
                       {
                           VectorizerName = "azure_openai_text_3_large"
                       }
                   },
                   Algorithms =
                   {
                       new HnswAlgorithmConfiguration(name: "alg")
                   },
                   Vectorizers =
                   {
                       vectorizer
                   }
               };
   
               // Define a semantic configuration
               var semanticConfig = new SemanticConfiguration(
                   name: "semantic_config",
                   prioritizedFields: new SemanticPrioritizedFields
                   {
                       ContentFields = { new SemanticField("page_chunk") }
                   }
               );
   
               var semanticSearch = new SemanticSearch()
               {
                   DefaultConfigurationName = "semantic_config",
                   Configurations = { semanticConfig }
               };
   
               // Create the index
               var index = new SearchIndex(indexName)
               {
                   Fields = fields,
                   VectorSearch = vectorSearch,
                   SemanticSearch = semanticSearch
               };
   
               // Create the index client, deleting and recreating the index if it exists
               var indexClient = new SearchIndexClient(new Uri(searchEndpoint), credential);
               await indexClient.CreateOrUpdateIndexAsync(index);
               Console.WriteLine($"Index '{indexName}' created or updated successfully.");
   
               // Upload sample documents from the GitHub URL
               string url = "https://raw.githubusercontent.com/Azure-Samples/azure-search-sample-data/refs/heads/main/nasa-e-book/earth-at-night-json/documents.json";
               var httpClient = new HttpClient();
               var response = await httpClient.GetAsync(url);
               response.EnsureSuccessStatusCode();
               var json = await response.Content.ReadAsStringAsync();
               var documents = JsonSerializer.Deserialize<List<Dictionary<string, object>>>(json);
               var searchClient = new SearchClient(new Uri(searchEndpoint), indexName, credential);
               var searchIndexingBufferedSender = new SearchIndexingBufferedSender<Dictionary<string, object>>(
                   searchClient,
                   new SearchIndexingBufferedSenderOptions<Dictionary<string, object>>
                   {
                       KeyFieldAccessor = doc => doc["id"].ToString(),
                   }
               );
               await searchIndexingBufferedSender.UploadDocumentsAsync(documents);
               await searchIndexingBufferedSender.FlushAsync();
               Console.WriteLine($"Documents uploaded to index '{indexName}' successfully.");
   
               // Create a knowledge source
               var indexKnowledgeSource = new SearchIndexKnowledgeSource(
                   name: knowledgeSourceName,
                   searchIndexParameters: new SearchIndexKnowledgeSourceParameters(searchIndexName: indexName)
                   {
                       SourceDataSelect = "id,page_chunk,page_number"
                   }
               );
               await indexClient.CreateOrUpdateKnowledgeSourceAsync(indexKnowledgeSource);
               Console.WriteLine($"Knowledge source '{knowledgeSourceName}' created or updated successfully.");
   
               // Create a knowledge agent
               var openAiParameters = new AzureOpenAIVectorizerParameters
               {
                   ResourceUri = new Uri(aoaiEndpoint),
                   DeploymentName = aoaiGptDeployment,
                   ModelName = aoaiGptModel
               };
   
               var agentModel = new KnowledgeAgentAzureOpenAIModel(azureOpenAIParameters: openAiParameters);
               var outputConfig = new KnowledgeAgentOutputConfiguration
               {
                   Modality = KnowledgeAgentOutputConfigurationModality.AnswerSynthesis,
                   IncludeActivity = true
               };
   
               var agent = new KnowledgeAgent(
                   name: knowledgeAgentName,
                   models: new[] { agentModel },
                   knowledgeSources: new KnowledgeSourceReference[] {
                   new KnowledgeSourceReference(knowledgeSourceName) {
                           IncludeReferences = true,
                           IncludeReferenceSourceData = true,
                           RerankerThreshold = (float?)2.5
                       }
                   }
               )
   
               {
                   OutputConfiguration = outputConfig
               };
   
               await indexClient.CreateOrUpdateKnowledgeAgentAsync(agent);
               Console.WriteLine($"Knowledge agent '{knowledgeAgentName}' created or updated successfully.");
   
               // Set up messages
               string instructions = @"A Q&A agent that can answer questions about the Earth at night.
               If you don't have the answer, respond with ""I don't know"".";
   
               var messages = new List<Dictionary<string, string>>
               {
                   new Dictionary<string, string>
                   {
                       { "role", "system" },
                       { "content", instructions }
                   }
               };
   
               // Use agentic retrieval to fetch results
               var agentClient = new KnowledgeAgentRetrievalClient(
                   endpoint: new Uri(searchEndpoint),
                   agentName: knowledgeAgentName,
                   tokenCredential: new DefaultAzureCredential()
               );
   
               messages.Add(new Dictionary<string, string>
               {
                   { "role", "user" },
                   { "content", @"Why do suburban belts display larger December brightening than urban cores even though absolute light levels are higher downtown?
                   Why is the Phoenix nighttime street grid is so sharply visible from space, whereas large stretches of the interstate between midwestern cities remain comparatively dim?" }
               });
   
               var retrievalResult = await agentClient.RetrieveAsync(
                   retrievalRequest: new KnowledgeAgentRetrievalRequest(
                       messages: messages
                           .Where(message => message["role"] != "system")
                           .Select(
                               message => new KnowledgeAgentMessage(content: new[] { new KnowledgeAgentMessageTextContent(message["content"]) }) { Role = message["role"] }
                           )
                           .ToList()
                   )
               );
   
               messages.Add(new Dictionary<string, string>
               {
                   { "role", "assistant" },
                   { "content", (retrievalResult.Value.Response[0].Content[0] as KnowledgeAgentMessageTextContent).Text }
               });
   
               // Print the response, activity, and results
               Console.WriteLine("Response:");
               Console.WriteLine((retrievalResult.Value.Response[0].Content[0] as KnowledgeAgentMessageTextContent).Text);
   
               Console.WriteLine("Activity:");
               foreach (var activity in retrievalResult.Value.Activity)
               {
                   Console.WriteLine($"Activity Type: {activity.GetType().Name}");
                   string activityJson = JsonSerializer.Serialize(
                       activity,
                       activity.GetType(),
                       new JsonSerializerOptions { WriteIndented = true }
                   );
                   Console.WriteLine(activityJson);
               }
   
               Console.WriteLine("Results:");
               foreach (var reference in retrievalResult.Value.References)
               {
                   Console.WriteLine($"Reference Type: {reference.GetType().Name}");
                   string referenceJson = JsonSerializer.Serialize(
                       reference,
                       reference.GetType(),
                       new JsonSerializerOptions { WriteIndented = true }
                   );
                   Console.WriteLine(referenceJson);
               }
   
               // Continue the conversation
               messages.Add(new Dictionary<string, string>
               {
                   { "role", "user" },
                   { "content", "How do I find lava at night?" }
               });
   
               retrievalResult = await agentClient.RetrieveAsync(
                   retrievalRequest: new KnowledgeAgentRetrievalRequest(
                       messages: messages
                           .Where(message => message["role"] != "system")
                           .Select(
                               message => new KnowledgeAgentMessage(content: new[] { new KnowledgeAgentMessageTextContent(message["content"]) }) { Role = message["role"] }
                           )
                           .ToList()
                   )
               );
   
               messages.Add(new Dictionary<string, string>
               {
                   { "role", "assistant" },
                   { "content", (retrievalResult.Value.Response[0].Content[0] as KnowledgeAgentMessageTextContent).Text }
               });
   
               // Print the new response, activity, and results
               Console.WriteLine("Response:");
               Console.WriteLine((retrievalResult.Value.Response[0].Content[0] as KnowledgeAgentMessageTextContent).Text);
   
               Console.WriteLine("Activity:");
               foreach (var activity in retrievalResult.Value.Activity)
               {
                   Console.WriteLine($"Activity Type: {activity.GetType().Name}");
                   string activityJson = JsonSerializer.Serialize(
                       activity,
                       activity.GetType(),
                       new JsonSerializerOptions { WriteIndented = true }
                   );
                   Console.WriteLine(activityJson);
               }
   
               Console.WriteLine("Results:");
               foreach (var reference in retrievalResult.Value.References)
               {
                   Console.WriteLine($"Reference Type: {reference.GetType().Name}");
                   string referenceJson = JsonSerializer.Serialize(
                       reference,
                       reference.GetType(),
                       new JsonSerializerOptions { WriteIndented = true }
                   );
                   Console.WriteLine(referenceJson);
               }
   
               // Clean up resources
               await indexClient.DeleteKnowledgeAgentAsync(knowledgeAgentName);
               Console.WriteLine($"Knowledge agent '{knowledgeAgentName}' deleted successfully.");
   
               await indexClient.DeleteKnowledgeSourceAsync(knowledgeSourceName);
               Console.WriteLine($"Knowledge source '{knowledgeSourceName}' deleted successfully.");
   
               await indexClient.DeleteIndexAsync(indexName);
               Console.WriteLine($"Index '{indexName}' deleted successfully.");
           }
       }
   }
   

5. 애플리케이션을 빌드하고 실행합니다.

   dotnet run
   

출력

애플리케이션의 출력은 다음과 유사해야 합니다.

Index 'earth-at-night' created or updated successfully.
Documents uploaded to index 'earth-at-night' successfully.
Knowledge source 'earth-knowledge-source' created or updated successfully.
Knowledge agent 'earth-knowledge-agent' created or updated successfully.
Response:
Suburban belts display larger December brightening than urban cores because holiday lights increase most dramatically in the suburbs and outskirts of major cities, where there is more yard space and a prevalence of single-family homes. Central urban areas, despite having higher absolute light levels, do not see as large an increase in lighting but still experience a brightening of 20 to 30 percent during the holidays [ref_id:2][ref_id:7].

The Phoenix nighttime street grid is sharply visible from space because the metropolitan area is laid out along a regular grid of city blocks and streets, with street lighting clearly visible from low-Earth orbit. The grid pattern is especially evident at night, with major street grids oriented north-south and diagonal corridors like Grand Avenue cutting across cities. The urban grid encourages outward growth along city borders, with extensive surface streets and freeways linking multiple municipalities. In contrast, large stretches of interstate highways between Midwestern cities remain comparatively dim because, although the interstate highways are major transportation corridors, the lighting along these highways is less intense and less continuous than the dense urban street lighting seen in Phoenix. Additionally, navigable rivers and less densely populated areas show less light, indicating that the brightness corresponds closely to urban density and street lighting patterns rather than just the presence of transportation routes [ref_id:0][ref_id:1][ref_id:4].
Activity:
Activity type: KnowledgeAgentModelQueryPlanningActivityRecord
{
  "InputTokens": 2062,
  "OutputTokens": 121,
  "Id": 0,
  "ElapsedMs": 2435
}
Activity type: KnowledgeAgentSearchIndexActivityRecord
{
  "SearchIndexArguments": {
    "Search": "Reasons for larger December brightening in suburban belts compared to urban cores despite higher downtown light levels",      
    "Filter": null
  },
  "KnowledgeSourceName": "earth-knowledge-source",
  "QueryTime": "2025-09-22T15:54:56.528+00:00",
  "Count": 4,
  "Id": 1,
  "ElapsedMs": 1921
}
Activity type: KnowledgeAgentSearchIndexActivityRecord
{
  "SearchIndexArguments": {
    "Search": "Factors making Phoenix nighttime street grid sharply visible from space",
    "Filter": null
  },
  "KnowledgeSourceName": "earth-knowledge-source",
  "QueryTime": "2025-09-22T15:55:06.991+00:00",
  "Count": 5,
  "Id": 2,
  "ElapsedMs": 10451
}
Activity type: KnowledgeAgentSearchIndexActivityRecord
{
  "SearchIndexArguments": {
    "Search": "Reasons why large stretches of interstate between Midwestern cities appear comparatively dim at night from space",
    "Filter": null
  },
  "KnowledgeSourceName": "earth-knowledge-source",
  "QueryTime": "2025-09-22T15:55:07.504+00:00",
  "Count": 13,
  "Id": 3,
  "ElapsedMs": 512
}
Activity type: KnowledgeAgentSemanticRerankerActivityRecord
{
  "InputTokens": 68754,
  "Id": 4,
  "ElapsedMs": null
}
Activity type: KnowledgeAgentModelAnswerSynthesisActivityRecord
{
  "InputTokens": 7231,
  "OutputTokens": 279,
  "Id": 5,
  "ElapsedMs": 6429
}
Results:
Reference type: KnowledgeAgentSearchIndexReference
{
  "DocKey": "earth_at_night_508_page_104_verbalized",
  "Id": "0",
  "ActivitySource": 2,
  "SourceData": {
    "id": "earth_at_night_508_page_104_verbalized",
    "page_chunk": "\u003C!-- PageHeader=\u0022Urban Structure\u0022 --\u003E\n\n### Location of Phoenix, Arizona\n\nThe image depicts a globe highlighting the ___location of Phoenix, Arizona, in the southwestern United States, marked with a blue pinpoint on the map of North America. Phoenix is situated in the central part of Arizona, which is in the southwestern region of the United States.\n\n---\n\n### Grid of City Blocks-Phoenix, Arizona\n\nLike many large urban areas of the central and western United States, the Phoenix metropolitan area is laid out along a regular grid of city blocks and streets. While visible during the day, this grid is most evident at night, when the pattern of street lighting is clearly visible from the low-Earth-orbit vantage point of the ISS.\n\nThis astronaut photograph, taken on March 16, 2013, includes parts of several cities in the metropolitan area, including Phoenix (image right), Glendale (center), and Peoria (left). While the major street grid is oriented north-south, the northwest-southeast oriented Grand Avenue cuts across the three cities at image center. Grand Avenue is a major transportation corridor through the western metropolitan area; the lighting patterns of large industrial and commercial properties are visible along its length. Other brightly lit properties include large shopping centers, strip malls, and gas stations, which tend to be located at the intersections of north-south and east-west trending streets.\n\nThe urban grid encourages growth outwards along a city\u0027s borders by providing optimal access to new real estate. Fueled by the adoption of widespread personal automobile use during the twentieth century, the Phoenix metropolitan area today includes 25 other municipalities (many of them largely suburban and residential) linked by a network of surface streets and freeways.\n\nWhile much of the land area highlighted in this image is urbanized, there are several noticeably dark areas. The Phoenix Mountains are largely public parks and recreational land. To the west, agricultural fields provide a sharp contrast to the lit streets of residential developments. The Salt River channel appears as a dark ribbon within the urban grid.\n\n\n\u003C!-- PageFooter=\u0022Earth at Night\u0022 --\u003E\n\u003C!-- PageNumber=\u002288\u0022 --\u003E",
    "page_number": 104
  },
  "RerankerScore": 2.6642752
}
Reference type: KnowledgeAgentSearchIndexReference
{
  "DocKey": "earth_at_night_508_page_105_verbalized",
  "Id": "3",
  "ActivitySource": 2,
  "SourceData": {
    "id": "earth_at_night_508_page_105_verbalized",
    "page_chunk": "# Urban Structure\n\n## March 16, 2013\n\n### Phoenix Metropolitan Area at Night\n\nThis figure presents a nighttime satellite view of the Phoenix metropolitan area, highlighting urban structure and transport corridors. City lights illuminate the layout of several cities and major thoroughfares.\n\n**Labeled Urban Features:**\n\n- **Phoenix:** Central and brightest area in the right-center of the image.\n- **Glendale:** Located to the west of Phoenix, this city is also brightly lit.\n- **Peoria:** Further northwest, this area is labeled and its illuminated grid is seen.\n- **Grand Avenue:** Clearly visible as a diagonal, brightly lit thoroughfare running from Phoenix through Glendale and Peoria.\n- **Salt River Channel:** Identified in the southeast portion, running through illuminated sections.\n- **Phoenix Mountains:** Dark, undeveloped region to the northeast of Phoenix.\n- **Agricultural Fields:** Southwestern corner of the image, grid patterns are visible but with much less illumination, indicating agricultural land use.\n\n**Additional Notes:**\n\n- The overall pattern shows a grid-like urban development typical of western U.S. cities, with scattered bright nodes at major intersections or city centers.\n- There is a clear transition from dense urban development to sparsely populated or agricultural land, particularly evident towards the bottom and left of the image.\n- The illuminated areas follow the existing road and street grids, showcasing the extensive spread of the metropolitan area.\n\n**Figure Description:**  \nA satellite nighttime image captured on March 16, 2013, showing Phoenix and surrounding areas (including Glendale and Peoria). Major landscape and infrastructural features, such as the Phoenix Mountains, Grand Avenue, the Salt River Channel, and agricultural fields, are labeled. The image reveals the extent of urbanization and the characteristic street grid illuminated by city lights.\n\n---\n\nPage 89",
    "page_number": 105
  },
  "RerankerScore": 2.5905457
}
... // Trimmed for brevity
Response:
Lava can be found at night by using satellite imagery that captures thermal infrared and near-infrared wavelengths, which highlight the heat emitted by active lava flows. For example, the Landsat 8 satellite's night view combines thermal, shortwave infrared, and near-infrared data to distinguish very hot lava (appearing bright white), cooling lava (red), and lava flows obscured by clouds (purple), as demonstrated in the monitoring of Kilauea's lava flows in Hawaii [ref_id:0]. Similarly, the Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS) on Landsat 8 have been used to detect the thermal infrared signature of lava flows during Mount Etna's flank eruption in Italy, highlighting active vents and lava flows at night [ref_id:1]. Additionally, the VIIRS Day/Night Band (DNB) on polar-orbiting satellites can detect faint light sources such as moonlight, which, combined with thermal data, allows for the observation of glowing lava flows at active volcanoes during nighttime [ref_id:1][ref_id:3]. Thus, by using satellite instruments sensitive to thermal and near-infrared wavelengths and leveraging natural illumination sources like moonlight, lava can be effectively located and monitored at night from space.
Activity:
Activity type: KnowledgeAgentModelQueryPlanningActivityRecord
{
  "InputTokens": 2357,
  "OutputTokens": 88,
  "Id": 0,
  "ElapsedMs": 1917
}
Activity type: KnowledgeAgentSearchIndexActivityRecord
{
  "SearchIndexArguments": {
    "Search": "How to locate lava flows at night",
    "Filter": null
  },
  "KnowledgeSourceName": "earth-knowledge-source",
  "QueryTime": "2025-09-22T15:55:16.919+00:00",
  "Count": 16,
  "Id": 1,
  "ElapsedMs": 433
}
Activity type: KnowledgeAgentSearchIndexActivityRecord
{
  "SearchIndexArguments": {
    "Search": "Methods for detecting lava at night",
    "Filter": null
  },
  "KnowledgeSourceName": "earth-knowledge-source",
  "QueryTime": "2025-09-22T15:55:17.389+00:00",
  "Count": 13,
  "Id": 2,
  "ElapsedMs": 468
}
Activity type: KnowledgeAgentSearchIndexActivityRecord
{
  "SearchIndexArguments": {
    "Search": "Safety tips for finding lava at night",
    "Filter": null
  },
  "KnowledgeSourceName": "earth-knowledge-source",
  "QueryTime": "2025-09-22T15:55:17.801+00:00",
  "Count": 3,
  "Id": 3,
  "ElapsedMs": 411
}
Activity type: KnowledgeAgentSemanticRerankerActivityRecord
{
  "InputTokens": 67218,
  "Id": 4,
  "ElapsedMs": null
}
Activity type: KnowledgeAgentModelAnswerSynthesisActivityRecord
{
  "InputTokens": 7345,
  "OutputTokens": 267,
  "Id": 5,
  "ElapsedMs": 6044
}
Results:
Reference type: KnowledgeAgentSearchIndexReference
{
  "DocKey": "earth_at_night_508_page_60_verbalized",
  "Id": "0",
  "ActivitySource": 1,
  "SourceData": {
    "id": "earth_at_night_508_page_60_verbalized",
    "page_chunk": "\u003C!-- PageHeader=\u0022Volcanoes\u0022 --\u003E\n\n## Volcanoes\n\n### The Infrared Glows of Kilauea\u0027s Lava Flows\u2014Hawaii\n\nIn early May 2018, an eruption on Hawaii\u0027s Kilauea volcano began to unfold. The eruption took a dangerous turn on May 3, 2018, when new fissures opened in the residential neighborhood of Leilani Estates. During the summer-long eruptive event, other fissures emerged along the East Rift Zone. Lava from vents along the rift zone flowed downslope, reaching the ocean in several areas, and filling in Kapoho Bay.\n\nA time series of Landsat 8 imagery shows the progression of the lava flows from May 16 to August 13. The night view combines thermal, shortwave infrared, and near-infrared wavelengths to tease out the very hot lava (bright white), cooling lava (red), and lava flows obstructed by clouds (purple).\n\n#### Figure: Location of Kilauea Volcano, Hawaii\n\nA globe is shown centered on North America, with a marker placed in the Pacific Ocean indicating the ___location of Hawaii, to the southwest of the mainland United States.\n\n\u003C!-- PageFooter=\u0022Earth at Night\u0022 --\u003E\n\u003C!-- PageNumber=\u002244\u0022 --\u003E",
    "page_number": 60
  },
  "RerankerScore": 2.779123
}
Reference type: KnowledgeAgentSearchIndexReference
{
  "DocKey": "earth_at_night_508_page_64_verbalized",
  "Id": "2",
  "ActivitySource": 1,
  "SourceData": {
    "id": "earth_at_night_508_page_64_verbalized",
    "page_chunk": "\u003C!-- PageHeader=\u0022Volcanoes\u0022 --\u003E\n\n### Nighttime Glow at Mount Etna - Italy\n\nAt about 2:30 a.m. local time on March 16, 2017, the VIIRS DNB on the Suomi NPP satellite captured this nighttime image of lava flowing on Mount Etna in Sicily, Italy. Etna is one of the world\u0027s most active volcanoes.\n\n#### Figure: Location of Mount Etna\nA world globe is depicted, with a marker indicating the ___location of Mount Etna in Sicily, Italy, in southern Europe near the center of the Mediterranean Sea.\n\n\u003C!-- PageFooter=\u0022Earth at Night\u0022 --\u003E\n\u003C!-- PageNumber=\u002248\u0022 --\u003E",
    "page_number": 64
  },
  "RerankerScore": 2.7684891
}
... // Trimmed for brevity
Knowledge agent 'earth-knowledge-agent' deleted successfully.
Knowledge source 'earth-knowledge-source' deleted successfully.
Index 'earth-at-night' deleted successfully.

코드 이해

이제 코드를 실행했으므로 다음 주요 단계를 세분화해 보겠습니다.

1. 검색 인덱스 만들기
2. 인덱스로 문서 업로드
3. 지식 원본 만들기
4. 지식 에이전트 만들기
5. 메시지 설정
6. 검색 파이프라인 실행
7. 대화 계속

검색 인덱스 만들기

Azure AI Search에서 인덱스는 구조화된 데이터 컬렉션입니다. 다음 코드는 이전에 earth-at-night 변수를 사용하여 지정한 이름인 indexName 인덱스를 정의합니다.

인덱스 스키마에는 문서 식별 및 페이지 콘텐츠, 포함 및 숫자에 대한 필드가 포함됩니다. 또한 스키마에는 의미론적 순위 지정 및 벡터 검색에 대한 구성이 포함되어 있으며, text-embedding-3-large 배포를 사용하여 텍스트를 벡터화하고 의미론적 또는 개념적 유사성에 따라 문서를 일치시킬 수 있습니다.

// Define fields for the index
var fields = new List<SearchField>
{
    new SimpleField("id", SearchFieldDataType.String) { IsKey = true, IsFilterable = true, IsSortable = true, IsFacetable = true },
    new SearchField("page_chunk", SearchFieldDataType.String) { IsFilterable = false, IsSortable = false, IsFacetable = false },
    new SearchField("page_embedding_text_3_large", SearchFieldDataType.Collection(SearchFieldDataType.Single)) { VectorSearchDimensions = 3072, VectorSearchProfileName = "hnsw_text_3_large" },
    new SimpleField("page_number", SearchFieldDataType.Int32) { IsFilterable = true, IsSortable = true, IsFacetable = true }
};

// Define a vectorizer
var vectorizer = new AzureOpenAIVectorizer(vectorizerName: "azure_openai_text_3_large")
{
    Parameters = new AzureOpenAIVectorizerParameters
    {
        ResourceUri = new Uri(aoaiEndpoint),
        DeploymentName = aoaiEmbeddingDeployment,
        ModelName = aoaiEmbeddingModel
    }
};

// Define a vector search profile and algorithm
var vectorSearch = new VectorSearch()
{
    Profiles =
    {
        new VectorSearchProfile(
            name: "hnsw_text_3_large",
            algorithmConfigurationName: "alg"
        )
        {
            VectorizerName = "azure_openai_text_3_large"
        }
    },
    Algorithms =
    {
        new HnswAlgorithmConfiguration(name: "alg")
    },
    Vectorizers =
    {
        vectorizer
    }
};

// Define a semantic configuration
var semanticConfig = new SemanticConfiguration(
    name: "semantic_config",
    prioritizedFields: new SemanticPrioritizedFields
    {
        ContentFields = { new SemanticField("page_chunk") }
    }
);

var semanticSearch = new SemanticSearch()
{
    DefaultConfigurationName = "semantic_config",
    Configurations = { semanticConfig }
};

// Create the index
var index = new SearchIndex(indexName)
{
    Fields = fields,
    VectorSearch = vectorSearch,
    SemanticSearch = semanticSearch
};

// Create the index client, deleting and recreating the index if it exists
var indexClient = new SearchIndexClient(new Uri(searchEndpoint), credential);
await indexClient.CreateOrUpdateIndexAsync(index);
Console.WriteLine($"Index '{indexName}' created or updated successfully.");

인덱스로 문서 업로드

현재 인덱스는 earth-at-night 비어 있습니다. 다음 코드는 NASA의 어스 앳 나이트 전자책의 JSON 문서로 인덱스를 채웁니다. Azure AI Search에서 요구하는 대로 각 문서는 인덱스 스키마에 정의된 필드 및 데이터 형식을 준수합니다.

// Upload sample documents from the GitHub URL
string url = "https://raw.githubusercontent.com/Azure-Samples/azure-search-sample-data/refs/heads/main/nasa-e-book/earth-at-night-json/documents.json";
var httpClient = new HttpClient();
var response = await httpClient.GetAsync(url);
response.EnsureSuccessStatusCode();
var json = await response.Content.ReadAsStringAsync();
var documents = JsonSerializer.Deserialize<List<Dictionary<string, object>>>(json);
var searchClient = new SearchClient(new Uri(searchEndpoint), indexName, credential);
var searchIndexingBufferedSender = new SearchIndexingBufferedSender<Dictionary<string, object>>(
    searchClient,
    new SearchIndexingBufferedSenderOptions<Dictionary<string, object>>
    {
        KeyFieldAccessor = doc => doc["id"].ToString(),
    }
);
await searchIndexingBufferedSender.UploadDocumentsAsync(documents);
await searchIndexingBufferedSender.FlushAsync();
Console.WriteLine($"Documents uploaded to index '{indexName}' successfully.");

지식 원본 만들기

지식 원본은 원본 데이터에 대한 재사용 가능한 참조입니다. 다음 코드는 earth-knowledge-source라는 지식 원본을 정의하며, 이는 earth-at-night 인덱스를 대상으로 합니다.

SourceDataSelect는 검색 및 인용을 위해 액세스할 수 있는 인덱스 필드를 지정합니다. 이 예제에는 응답에 길고 해석할 수 없는 포함이 발생하는 것을 방지하기 위해 사람이 읽을 수 있는 필드만 포함됩니다.

// Create a knowledge source
var indexKnowledgeSource = new SearchIndexKnowledgeSource(
    name: knowledgeSourceNames,
    searchIndexParameters: new SearchIndexKnowledgeSourceParameters(searchIndexName: indexName)
    {
        SourceDataSelect = "id,page_chunk,page_number"
    }
);
await indexClient.CreateOrUpdateKnowledgeSourceAsync(indexKnowledgeSource);
Console.WriteLine($"Knowledge source '{knowledgeSourceName}' created or updated successfully.");

지식 에이전트 만들기

쿼리 시 earth-knowledge-source 및 gpt-5-mini 배포를 대상으로 지정하려면 지식 에이전트가 필요합니다. 다음 코드를 사용하여 코드 셀을 추가하고 실행하여 이전에 earth-knowledge-agent 변수를 사용하여 지정한 knowledgeAgentName라는 지식 에이전트를 정의합니다.

RerankerThreshold는 순위 재지정 점수가 2.5이거나 이보다 낮은 응답을 제외하여 의미론적 관련성을 보장합니다. 한편, Modality는 검색된 문서를 인용하는 자연어 응답을 사용하도록 설정하도록 AnswerSynthesis로 설정됩니다.

// Create a knowledge agent
var openAiParameters = new AzureOpenAIVectorizerParameters
{
    ResourceUri = new Uri(aoaiEndpoint),
    DeploymentName = aoaiGptDeployment,
    ModelName = aoaiGptModel
};

var agentModel = new KnowledgeAgentAzureOpenAIModel(azureOpenAIParameters: openAiParameters);
var outputConfig = new KnowledgeAgentOutputConfiguration
{
    Modality = KnowledgeAgentOutputConfigurationModality.AnswerSynthesis,
    IncludeActivity = true
};

var agent = new KnowledgeAgent(
    name: knowledgeAgentName,
    models: new[] { agentModel },
    knowledgeSources: new KnowledgeSourceReference[] {
        new KnowledgeSourceReference(knowledgeSourceName) {
            IncludeReferences = true,
            IncludeReferenceSourceData = true,
            RerankerThreshold = (float?)2.5
        }
    }
)
{
    OutputConfiguration = outputConfig
};

await indexClient.CreateOrUpdateKnowledgeAgentAsync(agent);
Console.WriteLine($"Knowledge agent '{knowledgeAgentName}' created or updated successfully.");

메시지 설정

메시지는 검색 경로에 대한 입력이며 대화 기록을 포함합니다. 각 메시지에는 원본(예: system 및 user)을 나타내는 역할과 자연어 콘텐츠가 포함됩니다. 사용하는 LLM은 유효한 역할을 결정합니다.

다음 코드는 밤에 지구에 대한 질문에 대답하고 답변을 사용할 수 없을 때 "나는 모른다"로 응답하도록 지시 earth-knowledge-agent 하는 시스템 메시지를 만듭니다.

// Set up messages
string instructions = @"A Q&A agent that can answer questions about the Earth at night.
If you don't have the answer, respond with ""I don't know"".";

var messages = new List<Dictionary<string, string>>
{
    new Dictionary<string, string>
    {
        { "role", "system" },
        { "content", instructions }
    }
};

검색 파이프라인 실행

두 부분으로 구성된 사용자 쿼리를 earth-knowledge-agent로 보내 에이전트 검색을 실행할 준비가 완료되었습니다. 대화 내용 및 검색 매개 변수가 지정된 경우 에이전트는 다음을 수행합니다.

1. 전체 대화를 분석하여 사용자에게 어떤 정보가 필요한지를 유추합니다.
2. 복합 쿼리를 중요 하위 쿼리로 분해합니다.
3. 지식 원본에 대해 하위 쿼리를 동시에 실행합니다.
4. 의미 순위매기기를 사용하여 결과의 순위를 다시 매기고 필터링합니다.
5. 상위 결과를 자연어 응답으로 합성합니다.

// Use agentic retrieval to fetch results
var agentClient = new KnowledgeAgentRetrievalClient(
    endpoint: new Uri(searchEndpoint),
    agentName: knowledgeAgentName,
    tokenCredential: new DefaultAzureCredential()
);

messages.Add(new Dictionary<string, string>
{
    { "role", "user" },
    { "content", @"Why do suburban belts display larger December brightening than urban cores even though absolute light levels are higher downtown?
    Why is the Phoenix nighttime street grid is so sharply visible from space, whereas large stretches of the interstate between midwestern cities remain comparatively dim?" }
});

var retrievalResult = await agentClient.RetrieveAsync(
    retrievalRequest: new KnowledgeAgentRetrievalRequest(
        messages: messages
            .Where(message => message["role"] != "system")
            .Select(
                message => new KnowledgeAgentMessage(content: new[] { new KnowledgeAgentMessageTextContent(message["content"]) }) { Role = message["role"] }
            )
            .ToList()
    )
);

messages.Add(new Dictionary<string, string>
{
    { "role", "assistant" },
    { "content", (retrievalResult.Value.Response[0].Content[0] as KnowledgeAgentMessageTextContent).Text }
});

응답, 활동 및 결과 검토

다음 코드는 검색 파이프라인의 응답, 활동 및 결과를 표시합니다. 여기서는 다음과 같습니다.

• Response는 검색된 문서를 인용하는 쿼리에 대한 합성된 LLM 생성 답변을 제공합니다. 응답 합성을 사용할 수 없는 경우 이 섹션에는 문서에서 직접 추출된 콘텐츠가 포함됩니다.

• Activity는 gpt-5-mini 배포에서 생성된 하위 쿼리와 의미론적 순위 지정, 쿼리 계획 및 응답 합성에 사용되는 토큰을 포함하여 검색 프로세스 중에 수행된 단계를 추적합니다.

• Results 에는 응답에 기여한 문서가 나열되고 각 문서는 해당 DocKey문서로 식별됩니다.

// Print the response, activity, and results
Console.WriteLine("Response:");
Console.WriteLine((retrievalResult.Value.Response[0].Content[0] as KnowledgeAgentMessageTextContent).Text);

Console.WriteLine("Activity:");
foreach (var activity in retrievalResult.Value.Activity)
{
    Console.WriteLine($"Activity Type: {activity.GetType().Name}");
    string activityJson = JsonSerializer.Serialize(
        activity,
        activity.GetType(),
        new JsonSerializerOptions { WriteIndented = true }
    );
    Console.WriteLine(activityJson);
}

Console.WriteLine("Results:");
foreach (var reference in retrievalResult.Value.References)
{
    Console.WriteLine($"Reference Type: {reference.GetType().Name}");
    string referenceJson = JsonSerializer.Serialize(
        reference,
        reference.GetType(),
        new JsonSerializerOptions { WriteIndented = true }
    );
    Console.WriteLine(referenceJson);
}

대화 계속

다음 코드는 .와 earth-knowledge-agent대화를 계속합니다. 이 사용자 쿼리를 보낸 후 에이전트는 관련 콘텐츠를 earth-knowledge-source 가져오고 메시지 목록에 응답을 추가합니다.

// Continue the conversation
messages.Add(new Dictionary<string, string>
{
    { "role", "user" },
    { "content", "How do I find lava at night?" }
});

retrievalResult = await agentClient.RetrieveAsync(
    retrievalRequest: new KnowledgeAgentRetrievalRequest(
        messages: messages
            .Where(message => message["role"] != "system")
            .Select(
                message => new KnowledgeAgentMessage(content: new[] { new KnowledgeAgentMessageTextContent(message["content"]) }) { Role = message["role"] }
            )
            .ToList()
    )
);

messages.Add(new Dictionary<string, string>
{
    { "role", "assistant" },
    { "content", (retrievalResult.Value.Response[0].Content[0] as KnowledgeAgentMessageTextContent).Text }
});

새 응답, 활동 및 결과 검토

다음 코드는 검색 파이프라인의 새 응답, 활동 및 결과를 표시합니다.

// Print the response, activity, and results
Console.WriteLine("Response:");
Console.WriteLine((retrievalResult.Value.Response[0].Content[0] as KnowledgeAgentMessageTextContent).Text);

Console.WriteLine("Activities:");
foreach (var activity in retrievalResult.Value.Activity)
{
    Console.WriteLine($"Activity Type: {activity.GetType().Name}");
    string activityJson = JsonSerializer.Serialize(
        activity,
        activity.GetType(),
        new JsonSerializerOptions { WriteIndented = true }
    );
    Console.WriteLine(activityJson);
}

Console.WriteLine("Results:");
foreach (var reference in retrievalResult.Value.References)
{
    Console.WriteLine($"Reference Type: {reference.GetType().Name}");
    string referenceJson = JsonSerializer.Serialize(
        reference,
        reference.GetType(),
        new JsonSerializerOptions { WriteIndented = true }
    );
    Console.WriteLine(referenceJson);
}

자원을 정리하세요

본인 소유의 구독에서 작업하는 경우 자신이 만든 리소스가 여전히 필요한지 여부를 결정하여 프로젝트를 완료하는 것이 좋습니다. 실행 상태로 방치된 리소스는 비용이 들 수 있습니다.

Azure Portal의 왼쪽 창에서 모든 리소스 또는 리소스 그룹을 선택하여 Azure AI Search 및 Azure AI Foundry 리소스 를 관리할 수 있습니다.

그렇지 않으면, Program.cs의 다음 코드가 이 빠른 시작 가이드에서 만든 개체를 삭제합니다.

지식 에이전트 삭제

await indexClient.DeleteKnowledgeAgentAsync(knowledgeAgentName);
Console.WriteLine($"Knowledge agent '{knowledgeAgentName}' deleted successfully.");

지식 원본 삭제

await indexClient.DeleteKnowledgeSourceAsync(knowledgeSourceName);
Console.WriteLine($"Knowledge source '{knowledgeSourceName}' deleted successfully.");

검색 인덱스 삭제

await indexClient.DeleteIndexAsync(indexName);
Console.WriteLine($"Index '{indexName}' deleted successfully.");     

이 빠른 시작에서는 에이전트 검색 을 사용하여 LLM(대규모 언어 모델) 및 독점 데이터로 구동되는 대화형 검색 환경을 만듭니다. 에이전트 검색은 복잡한 사용자 쿼리를 하위 쿼리로 분해하고, 하위 쿼리를 병렬로 실행하고, Azure AI Search에서 인덱싱된 문서에서 접지 데이터를 추출합니다. 출력은 에이전트 및 사용자 지정 채팅 솔루션과 통합하기 위한 것입니다.

사용자 고유의 데이터를 제공할 수 있지만, 이 빠른 시작에서는 NASA의 어스 앳 나이트 전자책의 샘플 JSON 문서를 사용합니다. 문서는 우주에서 관찰 된 바와 같이 밤에 지구의 일반적인 과학 주제와 이미지를 설명합니다.

필수 조건

• 활성 구독이 있는 Azure 계정. 무료로 계정을 만듭니다.

• 의미 체계 순위가 설정된 기본 계층 이상의 Azure AI Search 서비스입니다.

• Azure AI Foundry 프로젝트입니다. Azure AI Foundry 프로젝트를 만들 때 모델 배포에 필요한 Azure AI Foundry 리소스를 가져옵니다.

• Microsoft Entra ID를 사용한 키 없는 인증을 위한 Azure CLI 입니다.

액세스 구성

시작하기 전에 콘텐츠 및 작업에 액세스할 수 있는 권한이 있는지 확인합니다. 인증에는 Microsoft Entra ID를, 권한 부여에는 역할 기반 액세스에 사용하는 것이 좋습니다. 역할을 할당하려면 소유자 또는 사용자 액세스 관리자여야 합니다. 역할이 가능하지 않은 경우 대신 키 기반 인증을 사용합니다.

이 빠른 시작에 대한 액세스를 구성하려면 다음 탭을 모두 선택합니다.

엔드포인트 가져오기

각 Azure AI 검색 서비스와 Azure AI Foundry 리소스에는 리소스에 대한 네트워크 액세스를 식별하고 제공하는 고유한 URL인 엔드포인트가 있습니다. 이후 섹션에서는 이러한 엔드포인트를 지정하여 프로그래밍 방식으로 리소스에 연결합니다.

이 빠른 시작의 엔드포인트를 얻으려면 다음 탭을 모두 선택합니다.

모델 배포

에이전트 검색을 사용하려면 Azure AI Foundry 리소스에 두 개의 Azure OpenAI 모델을 배포해야 합니다.

• 텍스트-벡터 변환을 위한 포함 모델입니다. 이 빠른 시작에서는 text-embedding-3-large을 사용하지만, 모든 text-embedding 모델을 사용할 수 있습니다.

• 쿼리 계획 및 응답 생성을 위한 LLM입니다. 이 퀵스타트에서는 gpt-5-mini를 사용하지만, 에이전트 검색에 지원되는 모든 LLM을 사용할 수 있습니다.

배포 지침은 Azure AI Foundry를 사용하여 Azure OpenAI 모델 배포를 참조하세요.

환경 설정

이 빠른 시작의 샘플은 Java 런타임에서 작동합니다. Azul Zulu OpenJDK와 같은 Java 개발 키트를 설치해야 합니다. OpenJDK의 Microsoft 빌드 또는 선호하는 JDK도 작동해야 합니다.

1. Apache Maven을 설치합니다. 그런 다음 mvn -v을(를) 실행하여 성공적인 설치를 확인합니다.

2. 애플리케이션을 포함할 새 폴더 quickstart-agentic-retrieval을 만들고 다음 명령을 사용하여 해당 폴더에서 Visual Studio Code를 엽니다.

   mkdir quickstart-agentic-retrieval && cd quickstart-agentic-retrieval
   

3. 새 pom.xml 파일을 프로젝트의 루트에 만들고, 다음 코드를 복사합니다.

   <project xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
       <modelVersion>4.0.0</modelVersion>
       <groupId>azure.search.sample</groupId>
       <artifactId>azuresearchquickstart</artifactId>
       <version>1.0.0-SNAPSHOT</version>
       <build>
           <sourceDirectory>src</sourceDirectory>
           <plugins>
           <plugin>
               <artifactId>maven-compiler-plugin</artifactId>
               <version>3.7.0</version>
               <configuration>
               <source>1.8</source>
               <target>1.8</target>
               </configuration>
           </plugin>
           </plugins>
       </build>
       <dependencies>
           <dependency>
               <groupId>junit</groupId>
               <artifactId>junit</artifactId>
               <version>4.11</version>
               <scope>test</scope>
           </dependency>
           <dependency>
               <groupId>com.azure</groupId>
               <artifactId>azure-search-documents</artifactId>
               <version>11.8.0-beta.7</version>
           </dependency>
           <dependency>
               <groupId>com.azure</groupId>
               <artifactId>azure-core</artifactId>
               <version>1.53.0</version>
           </dependency>
           <dependency>
               <groupId>com.azure</groupId>
               <artifactId>azure-identity</artifactId>
               <version>1.15.1</version>
           </dependency>
           <dependency>
               <groupId>com.azure</groupId>
               <artifactId>azure-ai-openai</artifactId>
               <version>1.0.0-beta.16</version>
           </dependency>
           <dependency>
               <groupId>com.fasterxml.jackson.core</groupId>
               <artifactId>jackson-databind</artifactId>
               <version>2.16.1</version>
           </dependency>
           <dependency>
               <groupId>io.github.cdimascio</groupId>
               <artifactId>dotenv-java</artifactId>
               <version>3.0.0</version>
           </dependency>
           <dependency>
               <groupId>org.apache.httpcomponents.client5</groupId>
               <artifactId>httpclient5</artifactId>
               <version>5.3.1</version>
           </dependency>
       </dependencies>
   </project>
   

4. 다음을 사용하여 Java용 Azure AI 검색 클라이언트 라이브러리(Azure.Search.Documents) 및 Java용 Azure ID 클라이언트 라이브러리를 포함한 종속성을 설치합니다.

   mvn clean dependency:copy-dependencies
   

코드 실행

1. 폴더에 명명된 .envquickstart-agentic-retrieval 새 파일을 만들고 다음 환경 변수를 추가합니다.

   AZURE_OPENAI_ENDPOINT=https://<your-ai-foundry-resource-name>.openai.azure.com/
   AZURE_OPENAI_GPT_DEPLOYMENT=gpt-5-mini
   AZURE_OPENAI_EMBEDDING_DEPLOYMENT=text-embedding-3-large
   AZURE_SEARCH_ENDPOINT=https://<your-search-service-name>.search.windows.net
   AZURE_SEARCH_INDEX_NAME=agentic-retrieval-sample
   

   <your-search-service-name> 및 <your-ai-foundry-resource-name>을 실제 Azure AI Search 서비스 이름과 Azure AI Foundry 리소스 이름으로 대체하십시오.

2. 폴더에 명명 AgenticRetrievalQuickstart.java 된 새 파일에 다음 코드를 붙여넣습니다 quickstart-agentic-retrieval .

   import com.azure.ai.openai.OpenAIAsyncClient;
   import com.azure.ai.openai.OpenAIClientBuilder;
   import com.azure.ai.openai.models.*;
   import com.azure.core.credential.TokenCredential;
   import com.azure.core.http.HttpClient;
   import com.azure.core.http.HttpHeaders;
   import com.azure.core.http.HttpMethod;
   import com.azure.core.http.HttpRequest;
   import com.azure.core.http.HttpResponse;
   import com.azure.core.util.BinaryData;
   import com.azure.identity.DefaultAzureCredential;
   import com.azure.identity.DefaultAzureCredentialBuilder;
   import com.azure.search.documents.SearchClient;
   import com.azure.search.documents.SearchClientBuilder;
   import com.azure.search.documents.SearchDocument;
   import com.azure.search.documents.indexes.SearchIndexClient;
   import com.azure.search.documents.indexes.SearchIndexClientBuilder;
   import com.azure.search.documents.indexes.models.*;
   import com.azure.search.documents.agents.SearchKnowledgeAgentClient;
   import com.azure.search.documents.agents.SearchKnowledgeAgentClientBuilder;
   import com.azure.search.documents.agents.models.*;
   import com.fasterxml.jackson.databind.JsonNode;
   import com.fasterxml.jackson.databind.ObjectMapper;
   import com.fasterxml.jackson.databind.node.ObjectNode;
   import io.github.cdimascio.dotenv.Dotenv;
   
   import java.io.IOException;
   import java.net.URI;
   import java.net.http.HttpRequest.Builder;
   import java.time.Duration;
   import java.util.*;
   import java.util.concurrent.TimeUnit;
   
   public class AgenticRetrievalQuickstart {
   
       // Configuration - Update these values for your environment
       private static final String SEARCH_ENDPOINT;
       private static final String AZURE_OPENAI_ENDPOINT;
       private static final String AZURE_OPENAI_GPT_DEPLOYMENT;
       private static final String AZURE_OPENAI_GPT_MODEL = "gpt-5-mini";
       private static final String AZURE_OPENAI_EMBEDDING_DEPLOYMENT;
       private static final String AZURE_OPENAI_EMBEDDING_MODEL = "text-embedding-3-large";
       private static final String INDEX_NAME = "earth_at_night";
       private static final String AGENT_NAME = "earth-search-agent";
       private static final String SEARCH_API_VERSION = "2025-05-01-Preview";
   
       static {
           // Load environment variables from .env file
           Dotenv dotenv = Dotenv.configure().ignoreIfMissing().load();
   
           SEARCH_ENDPOINT = getEnvVar(dotenv, "AZURE_SEARCH_ENDPOINT", 
               "https://contoso-agentic-search-service.search.windows.net");
           AZURE_OPENAI_ENDPOINT = getEnvVar(dotenv, "AZURE_OPENAI_ENDPOINT",
               "https://contoso-proj-agentic-foundry-res.openai.azure.com/");
           AZURE_OPENAI_GPT_DEPLOYMENT = getEnvVar(dotenv, "AZURE_OPENAI_GPT_DEPLOYMENT", "gpt-5-mini");
           AZURE_OPENAI_EMBEDDING_DEPLOYMENT = getEnvVar(dotenv, "AZURE_OPENAI_EMBEDDING_DEPLOYMENT", "text-embedding-3-large");
       }
   
       private static String getEnvVar(Dotenv dotenv, String key, String defaultValue) {
           String value = dotenv.get(key);
           return (value != null && !value.isEmpty()) ? value : defaultValue;
       }
   
       public static void main(String[] args) {
           try {
               System.out.println("Starting Azure AI Search agentic retrieval quickstart...\n");
   
               // Initialize Azure credentials using managed identity (recommended)
               TokenCredential credential = new DefaultAzureCredentialBuilder().build();
   
               // Create search clients
               SearchIndexClient searchIndexClient = new SearchIndexClientBuilder()
                   .endpoint(SEARCH_ENDPOINT)
                   .credential(credential)
                   .buildClient();
   
               SearchClient searchClient = new SearchClientBuilder()
                   .endpoint(SEARCH_ENDPOINT)
                   .indexName(INDEX_NAME)
                   .credential(credential)
                   .buildClient();
   
               // Create Azure OpenAI client
               OpenAIAsyncClient openAIClient = new OpenAIClientBuilder()
                   .endpoint(AZURE_OPENAI_ENDPOINT)
                   .credential(credential)
                   .buildAsyncClient();
   
               // Step 1: Create search index with vector and semantic capabilities
               createSearchIndex(searchIndexClient);
   
               // Step 2: Upload documents
               uploadDocuments(searchClient);
   
               // Step 3: Create knowledge agent
               createKnowledgeAgent(credential);
   
               // Step 4: Run agentic retrieval with conversation
               runAgenticRetrieval(credential, openAIClient);
   
               // Step 5: Clean up - Delete knowledge agent and search index
               deleteKnowledgeAgent(credential);
               deleteSearchIndex(searchIndexClient);
   
               System.out.println("[DONE] Quickstart completed successfully!");
   
           } catch (Exception e) {
               System.err.println("[ERROR] Error in main execution: " + e.getMessage());
               e.printStackTrace();
           }
       }
   
       private static void createSearchIndex(SearchIndexClient indexClient) {
           System.out.println("[WAIT] Creating search index...");
   
           try {
               // Delete index if it exists
               try {
                   indexClient.deleteIndex(INDEX_NAME);
                   System.out.println("[DELETE] Deleted existing index '" + INDEX_NAME + "'");
               } catch (Exception e) {
                   // Index doesn't exist, which is fine
               }
   
               // Define fields
               List<SearchField> fields = Arrays.asList(
                   new SearchField("id", SearchFieldDataType.STRING)
                       .setKey(true)
                       .setFilterable(true)
                       .setSortable(true)
                       .setFacetable(true),
                   new SearchField("page_chunk", SearchFieldDataType.STRING)
                       .setSearchable(true)
                       .setFilterable(false)
                       .setSortable(false)
                       .setFacetable(false),
                   new SearchField("page_embedding_text_3_large", SearchFieldDataType.collection(SearchFieldDataType.SINGLE))
                       .setSearchable(true)
                       .setFilterable(false)
                       .setSortable(false)
                       .setFacetable(false)
                       .setVectorSearchDimensions(3072)
                       .setVectorSearchProfileName("hnsw_text_3_large"),
                   new SearchField("page_number", SearchFieldDataType.INT32)
                       .setFilterable(true)
                       .setSortable(true)
                       .setFacetable(true)
               );
   
               // Create vectorizer
               AzureOpenAIVectorizer vectorizer = new AzureOpenAIVectorizer("azure_openai_text_3_large")
                   .setParameters(new AzureOpenAIVectorizerParameters()
                       .setResourceUrl(AZURE_OPENAI_ENDPOINT)
                       .setDeploymentName(AZURE_OPENAI_EMBEDDING_DEPLOYMENT)
                       .setModelName(AzureOpenAIModelName.TEXT_EMBEDDING_3_LARGE));
   
               // Create vector search configuration
               VectorSearch vectorSearch = new VectorSearch()
                   .setProfiles(Arrays.asList(
                       new VectorSearchProfile("hnsw_text_3_large", "alg")
                           .setVectorizerName("azure_openai_text_3_large")
                   ))
                   .setAlgorithms(Arrays.asList(
                       new HnswAlgorithmConfiguration("alg")
                   ))
                   .setVectorizers(Arrays.asList(vectorizer));
   
               // Create semantic search configuration
               SemanticSearch semanticSearch = new SemanticSearch()
                   .setDefaultConfigurationName("semantic_config")
                   .setConfigurations(Arrays.asList(
                       new SemanticConfiguration("semantic_config",
                           new SemanticPrioritizedFields()
                               .setContentFields(Arrays.asList(
                                   new SemanticField("page_chunk")
                               ))
                       )
                   ));
   
               // Create the index
               SearchIndex index = new SearchIndex(INDEX_NAME)
                   .setFields(fields)
                   .setVectorSearch(vectorSearch)
                   .setSemanticSearch(semanticSearch);
   
               indexClient.createOrUpdateIndex(index);
               System.out.println("[DONE] Index '" + INDEX_NAME + "' created successfully.");
   
           } catch (Exception e) {
               System.err.println("[ERROR] Error creating index: " + e.getMessage());
               throw new RuntimeException(e);
           }
       }
   
       private static void uploadDocuments(SearchClient searchClient) {
           System.out.println("[WAIT] Uploading documents...");
   
           try {
               // Fetch documents from GitHub
               List<SearchDocument> documents = fetchEarthAtNightDocuments();
   
               searchClient.uploadDocuments(documents);
               System.out.println("[DONE] Uploaded " + documents.size() + " documents successfully.");
   
               // Wait for indexing to complete
               System.out.println("[WAIT] Waiting for document indexing to complete...");
               Thread.sleep(5000);
               System.out.println("[DONE] Document indexing completed.");
   
           } catch (Exception e) {
               System.err.println("[ERROR] Error uploading documents: " + e.getMessage());
               throw new RuntimeException(e);
           }
       }
   
       private static List<SearchDocument> fetchEarthAtNightDocuments() {
           System.out.println("[WAIT] Fetching Earth at Night documents from GitHub...");
   
           String documentsUrl = "https://raw.githubusercontent.com/Azure-Samples/azure-search-sample-data/refs/heads/main/nasa-e-book/earth-at-night-json/documents.json";
   
           try {
               java.net.http.HttpClient httpClient = java.net.http.HttpClient.newHttpClient();
               java.net.http.HttpRequest request = java.net.http.HttpRequest.newBuilder()
                   .uri(URI.create(documentsUrl))
                   .build();
   
               java.net.http.HttpResponse<String> response = httpClient.send(request, 
                   java.net.http.HttpResponse.BodyHandlers.ofString());
   
               if (response.statusCode() != 200) {
                   throw new IOException("Failed to fetch documents: " + response.statusCode());
               }
   
               ObjectMapper mapper = new ObjectMapper();
               JsonNode jsonArray = mapper.readTree(response.body());
   
               List<SearchDocument> documents = new ArrayList<>();
               for (int i = 0; i < jsonArray.size(); i++) {
                   JsonNode doc = jsonArray.get(i);
                   SearchDocument searchDoc = new SearchDocument();
   
                   searchDoc.put("id", doc.has("id") ? doc.get("id").asText() : String.valueOf(i + 1));
                   searchDoc.put("page_chunk", doc.has("page_chunk") ? doc.get("page_chunk").asText() : "");
   
                   // Handle embeddings
                   if (doc.has("page_embedding_text_3_large") && doc.get("page_embedding_text_3_large").isArray()) {
                       List<Double> embeddings = new ArrayList<>();
                       for (JsonNode embedding : doc.get("page_embedding_text_3_large")) {
                           embeddings.add(embedding.asDouble());
                       }
                       searchDoc.put("page_embedding_text_3_large", embeddings);
                   } else {
                       // Fallback embeddings
                       List<Double> fallbackEmbeddings = new ArrayList<>();
                       for (int j = 0; j < 3072; j++) {
                           fallbackEmbeddings.add(0.1);
                       }
                       searchDoc.put("page_embedding_text_3_large", fallbackEmbeddings);
                   }
   
                   searchDoc.put("page_number", doc.has("page_number") ? doc.get("page_number").asInt() : i + 1);
   
                   documents.add(searchDoc);
               }
   
               System.out.println("[DONE] Fetched " + documents.size() + " documents from GitHub");
               return documents;
   
           } catch (Exception e) {
               System.err.println("[ERROR] Error fetching documents from GitHub: " + e.getMessage());
               System.out.println("🔄 Falling back to sample documents...");
   
               // Fallback to sample documents
               List<SearchDocument> fallbackDocs = new ArrayList<>();
   
               SearchDocument doc1 = new SearchDocument();
               doc1.put("id", "1");
               doc1.put("page_chunk", "The Earth at night reveals the patterns of human settlement and economic activity. City lights trace the contours of civilization, creating a luminous map of where people live and work.");
               List<Double> embeddings1 = new ArrayList<>();
               for (int i = 0; i < 3072; i++) {
                   embeddings1.add(0.1);
               }
               doc1.put("page_embedding_text_3_large", embeddings1);
               doc1.put("page_number", 1);
   
               SearchDocument doc2 = new SearchDocument();
               doc2.put("id", "2");
               doc2.put("page_chunk", "From space, the aurora borealis appears as shimmering curtains of green and blue light dancing across the polar regions.");
               List<Double> embeddings2 = new ArrayList<>();
               for (int i = 0; i < 3072; i++) {
                   embeddings2.add(0.2);
               }
               doc2.put("page_embedding_text_3_large", embeddings2);
               doc2.put("page_number", 2);
   
               fallbackDocs.add(doc1);
               fallbackDocs.add(doc2);
   
               return fallbackDocs;
           }
       }
   
       private static void createKnowledgeAgent(TokenCredential credential) {
           System.out.println("[WAIT] Creating knowledge agent...");
   
           // Delete agent if it exists
           deleteKnowledgeAgent(credential);
   
           try {
               ObjectMapper mapper = new ObjectMapper();
               ObjectNode agentDefinition = mapper.createObjectNode();
               agentDefinition.put("name", AGENT_NAME);
               agentDefinition.put("description", "Knowledge agent for Earth at Night e-book content");
   
               ObjectNode model = mapper.createObjectNode();
               model.put("kind", "azureOpenAI");
               ObjectNode azureOpenAIParams = mapper.createObjectNode();
               azureOpenAIParams.put("resourceUri", AZURE_OPENAI_ENDPOINT);
               azureOpenAIParams.put("deploymentId", AZURE_OPENAI_GPT_DEPLOYMENT);
               azureOpenAIParams.put("modelName", AZURE_OPENAI_GPT_MODEL);
               model.set("azureOpenAIParameters", azureOpenAIParams);
               agentDefinition.set("models", mapper.createArrayNode().add(model));
   
               ObjectNode targetIndex = mapper.createObjectNode();
               targetIndex.put("indexName", INDEX_NAME);
               targetIndex.put("defaultRerankerThreshold", 2.5);
               agentDefinition.set("targetIndexes", mapper.createArrayNode().add(targetIndex));
   
               String token = getAccessToken(credential, "https://search.azure.com/.default");
   
               java.net.http.HttpClient httpClient = java.net.http.HttpClient.newHttpClient();
               java.net.http.HttpRequest request = java.net.http.HttpRequest.newBuilder()
                   .uri(URI.create(SEARCH_ENDPOINT + "/agents/" + AGENT_NAME + "?api-version=" + SEARCH_API_VERSION))
                   .header("Content-Type", "application/json")
                   .header("Authorization", "Bearer " + token)
                   .PUT(java.net.http.HttpRequest.BodyPublishers.ofString(mapper.writeValueAsString(agentDefinition)))
                   .build();
   
               java.net.http.HttpResponse<String> response = httpClient.send(request,
                   java.net.http.HttpResponse.BodyHandlers.ofString());
   
               if (response.statusCode() >= 400) {
                   throw new RuntimeException("Failed to create knowledge agent: " + response.statusCode() + " " + response.body());
               }
   
               System.out.println("[DONE] Knowledge agent '" + AGENT_NAME + "' created successfully.");
   
           } catch (Exception e) {
               System.err.println("[ERROR] Error creating knowledge agent: " + e.getMessage());
               throw new RuntimeException(e);
           }
       }
   
       private static void runAgenticRetrieval(TokenCredential credential, OpenAIAsyncClient openAIClient) {
           System.out.println("[SEARCH] Running agentic retrieval...");
   
           // Initialize messages with system instructions
           List<Map<String, String>> messages = new ArrayList<>();
   
           Map<String, String> systemMessage = new HashMap<>();
           systemMessage.put("role", "system");
           systemMessage.put("content", "A Q&A agent that can answer questions about the Earth at night.\n" +
               "Sources have a JSON format with a ref_id that must be cited in the answer.\n" +
               "If you do not have the answer, respond with \"I don't know\".");
           messages.add(systemMessage);
   
           Map<String, String> userMessage = new HashMap<>();
           userMessage.put("role", "user");
           userMessage.put("content", "Why do suburban belts display larger December brightening than urban cores even though absolute light levels are higher downtown? Why is the Phoenix nighttime street grid is so sharply visible from space, whereas large stretches of the interstate between midwestern cities remain comparatively dim?");
           messages.add(userMessage);
   
           try {
               // Call agentic retrieval API (excluding system message)
               List<Map<String, String>> userMessages = messages.stream()
                   .filter(m -> !"system".equals(m.get("role")))
                   .collect(java.util.stream.Collectors.toList());
   
               String retrievalResponse = callAgenticRetrieval(credential, userMessages);
   
               // Add assistant response to conversation history
               Map<String, String> assistantMessage = new HashMap<>();
               assistantMessage.put("role", "assistant");
               assistantMessage.put("content", retrievalResponse);
               messages.add(assistantMessage);
   
               System.out.println(retrievalResponse);
   
               // Now do chat completion with full conversation history
               generateFinalAnswer(openAIClient, messages);
   
               // Continue conversation with second question
               continueConversation(credential, openAIClient, messages);
   
           } catch (Exception e) {
               System.err.println("[ERROR] Error in agentic retrieval: " + e.getMessage());
               throw new RuntimeException(e);
           }
       }
   
       private static String callAgenticRetrieval(TokenCredential credential, List<Map<String, String>> messages) {
           try {
               ObjectMapper mapper = new ObjectMapper();
               ObjectNode retrievalRequest = mapper.createObjectNode();
   
               // Convert messages to the correct format expected by the Knowledge agent
               com.fasterxml.jackson.databind.node.ArrayNode agentMessages = mapper.createArrayNode();
               for (Map<String, String> msg : messages) {
                   ObjectNode agentMessage = mapper.createObjectNode();
                   agentMessage.put("role", msg.get("role"));
   
                   com.fasterxml.jackson.databind.node.ArrayNode content = mapper.createArrayNode();
                   ObjectNode textContent = mapper.createObjectNode();
                   textContent.put("type", "text");
                   textContent.put("text", msg.get("content"));
                   content.add(textContent);
                   agentMessage.set("content", content);
   
                   agentMessages.add(agentMessage);
               }
               retrievalRequest.set("messages", agentMessages);
   
               com.fasterxml.jackson.databind.node.ArrayNode targetIndexParams = mapper.createArrayNode();
               ObjectNode indexParam = mapper.createObjectNode();
               indexParam.put("indexName", INDEX_NAME);
               indexParam.put("rerankerThreshold", 2.5);
               indexParam.put("maxDocsForReranker", 100);
               indexParam.put("includeReferenceSourceData", true);
               targetIndexParams.add(indexParam);
               retrievalRequest.set("targetIndexParams", targetIndexParams);
   
               String token = getAccessToken(credential, "https://search.azure.com/.default");
   
               java.net.http.HttpClient httpClient = java.net.http.HttpClient.newHttpClient();
               java.net.http.HttpRequest request = java.net.http.HttpRequest.newBuilder()
                   .uri(URI.create(SEARCH_ENDPOINT + "/agents/" + AGENT_NAME + "/retrieve?api-version=" + SEARCH_API_VERSION))
                   .header("Content-Type", "application/json")
                   .header("Authorization", "Bearer " + token)
                   .POST(java.net.http.HttpRequest.BodyPublishers.ofString(mapper.writeValueAsString(retrievalRequest)))
                   .build();
   
               java.net.http.HttpResponse<String> response = httpClient.send(request,
                   java.net.http.HttpResponse.BodyHandlers.ofString());
   
               if (response.statusCode() >= 400) {
                   throw new RuntimeException("Agentic retrieval failed: " + response.statusCode() + " " + response.body());
               }
   
               JsonNode responseJson = mapper.readTree(response.body());
   
               // Log activities and results
               logActivitiesAndResults(responseJson);
   
               // Extract response content
               if (responseJson.has("response") && responseJson.get("response").isArray()) {
                   com.fasterxml.jackson.databind.node.ArrayNode responseArray = (com.fasterxml.jackson.databind.node.ArrayNode) responseJson.get("response");
                   if (responseArray.size() > 0) {
                       JsonNode firstResponse = responseArray.get(0);
                       if (firstResponse.has("content") && firstResponse.get("content").isArray()) {
                           com.fasterxml.jackson.databind.node.ArrayNode contentArray = (com.fasterxml.jackson.databind.node.ArrayNode) firstResponse.get("content");
                           if (contentArray.size() > 0) {
                               JsonNode textContent = contentArray.get(0);
                               if (textContent.has("text")) {
                                   return textContent.get("text").asText();
                               }
                           }
                       }
                   }
               }
   
               return "No response content available";
   
           } catch (Exception e) {
               System.err.println("[ERROR] Error in agentic retrieval call: " + e.getMessage());
               throw new RuntimeException(e);
           }
       }
   
       private static void logActivitiesAndResults(JsonNode responseJson) {
           ObjectMapper mapper = new ObjectMapper();
   
           // Log activities
           System.out.println("\nActivities:");
           if (responseJson.has("activity") && responseJson.get("activity").isArray()) {
               for (JsonNode activity : responseJson.get("activity")) {
                   String activityType = "UnknownActivityRecord";
                   if (activity.has("InputTokens")) {
                       activityType = "KnowledgeAgentModelQueryPlanningActivityRecord";
                   } else if (activity.has("TargetIndex")) {
                       activityType = "KnowledgeAgentSearchActivityRecord";
                   } else if (activity.has("QueryTime")) {
                       activityType = "KnowledgeAgentSemanticRankerActivityRecord";
                   }
   
                   System.out.println("Activity Type: " + activityType);
                   try {
                       System.out.println(mapper.writerWithDefaultPrettyPrinter().writeValueAsString(activity));
                   } catch (Exception e) {
                       System.out.println(activity.toString());
                   }
               }
           }
   
           // Log results
           System.out.println("Results");
           if (responseJson.has("references") && responseJson.get("references").isArray()) {
               for (JsonNode reference : responseJson.get("references")) {
                   String referenceType = "KnowledgeAgentAzureSearchDocReference";
   
                   System.out.println("Reference Type: " + referenceType);
                   try {
                       System.out.println(mapper.writerWithDefaultPrettyPrinter().writeValueAsString(reference));
                   } catch (Exception e) {
                       System.out.println(reference.toString());
                   }
               }
           }
       }
   
       private static void generateFinalAnswer(OpenAIAsyncClient openAIClient, List<Map<String, String>> messages) {
           System.out.println("\n[ASSISTANT]: ");
   
           try {
               List<ChatRequestMessage> chatMessages = new ArrayList<>();
               for (Map<String, String> msg : messages) {
                   String role = msg.get("role");
                   String content = msg.get("content");
   
                   switch (role) {
                       case "system":
                           chatMessages.add(new ChatRequestSystemMessage(content));
                           break;
                       case "user":
                           chatMessages.add(new ChatRequestUserMessage(content));
                           break;
                       case "assistant":
                           chatMessages.add(new ChatRequestAssistantMessage(content));
                           break;
                   }
               }
   
               ChatCompletionsOptions chatOptions = new ChatCompletionsOptions(chatMessages)
                   .setMaxTokens(1000)
                   .setTemperature(0.7);
   
               ChatCompletions completion = openAIClient.getChatCompletions(AZURE_OPENAI_GPT_DEPLOYMENT, chatOptions).block();
   
               if (completion != null && completion.getChoices() != null && !completion.getChoices().isEmpty()) {
                   String answer = completion.getChoices().get(0).getMessage().getContent();
                   System.out.println(answer.replace(".", "\n"));
   
                   // Add this response to conversation history
                   Map<String, String> assistantResponse = new HashMap<>();
                   assistantResponse.put("role", "assistant");
                   assistantResponse.put("content", answer);
                   messages.add(assistantResponse);
               }
   
           } catch (Exception e) {
               System.err.println("[ERROR] Error generating final answer: " + e.getMessage());
               throw new RuntimeException(e);
           }
       }
   
       private static void continueConversation(TokenCredential credential, OpenAIAsyncClient openAIClient, List<Map<String, String>> messages) {
           System.out.println("\n === Continuing Conversation ===");
   
           // Add follow-up question
           String followUpQuestion = "How do I find lava at night?";
           System.out.println("[QUESTION] Follow-up question: " + followUpQuestion);
   
           Map<String, String> userMessage = new HashMap<>();
           userMessage.put("role", "user");
           userMessage.put("content", followUpQuestion);
           messages.add(userMessage);
   
           try {
               // FILTER OUT SYSTEM MESSAGE - only send user/assistant messages to agentic retrieval
               List<Map<String, String>> userAssistantMessages = messages.stream()
                   .filter(m -> !"system".equals(m.get("role")))
                   .collect(java.util.stream.Collectors.toList());
   
               String newRetrievalResponse = callAgenticRetrieval(credential, userAssistantMessages);
   
               // Add assistant response to conversation history
               Map<String, String> assistantMessage = new HashMap<>();
               assistantMessage.put("role", "assistant");
               assistantMessage.put("content", newRetrievalResponse);
               messages.add(assistantMessage);
   
               System.out.println(newRetrievalResponse);
   
               // Generate final answer for follow-up
               generateFinalAnswer(openAIClient, messages);
   
               System.out.println("\n === Conversation Complete ===");
   
           } catch (Exception e) {
               System.err.println("[ERROR] Error in conversation continuation: " + e.getMessage());
               throw new RuntimeException(e);
           }
       }
   
       private static void deleteKnowledgeAgent(TokenCredential credential) {
           System.out.println("[DELETE] Deleting knowledge agent...");
   
           try {
               String token = getAccessToken(credential, "https://search.azure.com/.default");
   
               java.net.http.HttpClient httpClient = java.net.http.HttpClient.newHttpClient();
               java.net.http.HttpRequest request = java.net.http.HttpRequest.newBuilder()
                   .uri(URI.create(SEARCH_ENDPOINT + "/agents/" + AGENT_NAME + "?api-version=" + SEARCH_API_VERSION))
                   .header("Authorization", "Bearer " + token)
                   .DELETE()
                   .build();
   
               java.net.http.HttpResponse<String> response = httpClient.send(request,
                   java.net.http.HttpResponse.BodyHandlers.ofString());
   
               if (response.statusCode() == 404) {
                   System.out.println("[INFO] Knowledge agent '" + AGENT_NAME + "' does not exist or was already deleted.");
                   return;
               }
   
               if (response.statusCode() >= 400) {
                   throw new RuntimeException("Failed to delete knowledge agent: " + response.statusCode() + " " + response.body());
               }
   
               System.out.println("[DONE] Knowledge agent '" + AGENT_NAME + "' deleted successfully.");
   
           } catch (Exception e) {
               System.err.println("[ERROR] Error deleting knowledge agent: " + e.getMessage());
               // Don't throw - this is cleanup
           }
       }
   
       private static void deleteSearchIndex(SearchIndexClient indexClient) {
           System.out.println("[DELETE] Deleting search index...");
   
           try {
               indexClient.deleteIndex(INDEX_NAME);
               System.out.println("[DONE] Search index '" + INDEX_NAME + "' deleted successfully.");
   
           } catch (Exception e) {
               if (e.getMessage() != null && (e.getMessage().contains("404") || e.getMessage().contains("IndexNotFound"))) {
                   System.out.println("[INFO] Search index '" + INDEX_NAME + "' does not exist or was already deleted.");
                   return;
               }
               System.err.println("[ERROR] Error deleting search index: " + e.getMessage());
               // Don't throw - this is cleanup
           }
       }
   
       private static String getAccessToken(TokenCredential credential, String scope) {
           try {
               return credential.getToken(new com.azure.core.credential.TokenRequestContext().addScopes(scope)).block().getToken();
           } catch (Exception e) {
               throw new RuntimeException("Failed to get access token", e);
           }
       }
   }
   

3. 다음 명령을 사용하여 Azure에 로그인합니다.

   az login
   

4. 새 콘솔 애플리케이션 실행:

   javac Address.java App.java Hotel.java -cp ".;target\dependency\*"
   java -cp ".;target\dependency\*" App
   

출력

애플리케이션의 출력은 다음과 유사하게 표시됩니다.

Starting Azure AI Search agentic retrieval quickstart...

[WAIT] Creating search index...
[DELETE] Deleted existing index 'earth_at_night'
[DONE] Index 'earth_at_night' created successfully.
[WAIT] Uploading documents...
[WAIT] Fetching Earth at Night documents from GitHub...
[DONE] Fetched 194 documents from GitHub
[DONE] Uploaded 194 documents successfully.
[WAIT] Waiting for document indexing to complete...
[DONE] Document indexing completed.
[WAIT] Creating knowledge agent...
[DELETE] Deleting knowledge agent...
[INFO] Knowledge agent 'earth-search-agent' does not exist or was already deleted.
[DONE] Knowledge agent 'earth-search-agent' created successfully.
[SEARCH] Running agentic retrieval...

Activities:
Activity Type: UnknownActivityRecord
{
  "type" : "ModelQueryPlanning",
  "id" : 0,
  "inputTokens" : 1379,
  "outputTokens" : 545
}
Activity Type: UnknownActivityRecord
{
  "type" : "AzureSearchQuery",
  "id" : 1,
  "targetIndex" : "earth_at_night",
  "query" : {
    "search" : "Why do suburban areas show greater December brightening compared to urban cores despite higher absolute light levels downtown?",
    "filter" : null
  },
  "queryTime" : "2025-07-21T15:07:04.024Z",
  "count" : 0,
  "elapsedMs" : 2609
}
Activity Type: UnknownActivityRecord
{
  "type" : "AzureSearchQuery",
  "id" : 2,
  "targetIndex" : "earth_at_night",
  "query" : {
    "search" : "Why is the Phoenix nighttime street grid sharply visible from space, while large stretches of interstate highways between Midwestern cities appear comparatively dim?",
    "filter" : null
  },
  "queryTime" : "2025-07-21T15:07:04.267Z",
  "count" : 0,
  "elapsedMs" : 243
}
Activity Type: UnknownActivityRecord
{
  "type" : "AzureSearchSemanticRanker",
  "id" : 3,
  "inputTokens" : 48602
}
Results
[]

[ASSISTANT]: 
The suburban belts display larger December brightening than urban cores despite higher absolute light levels downtown likely because suburban areas have more seasonal variation in lighting usage, such as increased outdoor and holiday lighting in December
 Urban cores, being brightly lit throughout the year, show less relative change


Regarding Phoenix's nighttime street grid visibility, it is sharply visible from space due to the structured and continuous lighting of the city's streets
 In contrast, large stretches of interstate highways between Midwestern cities are comparatively dim because highways typically have less intense and less frequent lighting compared to urban street grids


[Note: This explanation is based on general knowledge; no specific source with ref_id was provided
]

 === Continuing Conversation ===
[QUESTION] Follow-up question: How do I find lava at night?

Activities:
Activity Type: UnknownActivityRecord
{
  "type" : "ModelQueryPlanning",
  "id" : 0,
  "inputTokens" : 1545,
  "outputTokens" : 127
}
Activity Type: UnknownActivityRecord
{
  "type" : "AzureSearchQuery",
  "id" : 1,
  "targetIndex" : "earth_at_night",
  "query" : {
    "search" : "How can I find lava at night?",
    "filter" : null
  },
  "queryTime" : "2025-07-21T15:07:15.445Z",
  "count" : 6,
  "elapsedMs" : 370
}
Activity Type: UnknownActivityRecord
{
  "type" : "AzureSearchSemanticRanker",
  "id" : 2,
  "inputTokens" : 22994
}
Results
Reference Type: KnowledgeAgentAzureSearchDocReference
{
  "type" : "AzureSearchDoc",
  "id" : "0",
  "activitySource" : 1,
  "docKey" : "earth_at_night_508_page_44_verbalized",
  "sourceData" : {
    "id" : "earth_at_night_508_page_44_verbalized",
    "page_chunk" : "## Nature's Light Shows\n\nAt night, with the light of the Sun removed, nature's brilliant glow from Earth's surface becomes visible to the naked eye from space. Some of Earth's most spectacular light shows are natural, like the aurora borealis, or Northern Lights, in the Northern Hemisphere (aurora australis, or Southern Lights, in the Southern Hemisphere). The auroras are natural electrical phenomena caused by charged particles that race from the Sun toward Earth, inducing chemical reactions in the upper atmosphere and creating the appearance of streamers of reddish or greenish light in the sky, usually near the northern or southern magnetic pole. Other natural lights can indicate danger, like a raging forest fire encroaching on a city, town, or community, or lava spewing from an erupting volcano.\n\nWhatever the source, the ability of humans to monitor nature's light shows at night has practical applications for society. For example, tracking fires during nighttime hours allows for continuous monitoring and enhances our ability to protect humans and other animals, plants, and infrastructure. Combined with other data sources, our ability to observe the light of fires at night allows emergency managers to more efficiently and accurately issue warnings and evacuation orders and allows firefighting efforts to continue through the night. With enough moonlight (e.g., full-Moon phase), it's even possible to track the movement of smoke plumes at night, which can impact air quality, regardless of time of day.\n\nAnother natural source of light at night is emitted from glowing lava flows at the site of active volcanoes. Again, with enough moonlight, these dramatic scenes can be tracked and monitored for both scientific research and public safety.\n\n\n### Figure: The Northern Lights Viewed from Space\n\n**September 17, 2011**\n\nThis photo, taken from the International Space Station on September 17, 2011, shows a spectacular display of the aurora borealis (Northern Lights) as green and reddish light in the night sky above Earth. In the foreground, part of a Soyuz spacecraft is visible, silhouetted against the bright auroral light. The green glow is generated by energetic charged particles from the Sun interacting with Earth's upper atmosphere, exciting oxygen and nitrogen atoms, and producing characteristic colors. The image demonstrates the vividness and grandeur of natural night-time light phenomena as seen from orbit."
  }
}
Reference Type: KnowledgeAgentAzureSearchDocReference
{
  "type" : "AzureSearchDoc",
  "id" : "1",
  "activitySource" : 1,
  "docKey" : "earth_at_night_508_page_65_verbalized",
  "sourceData" : {
    "id" : "earth_at_night_508_page_65_verbalized",
    "page_chunk" : "# Volcanoes\n\n## Figure: Satellite Image of Sicily and Mount Etna Lava, March 16, 2017\n\nThe annotated satellite image below shows the island of Sicily and the surrounding region at night, highlighting city lights and volcanic activity.\n\n**Description:**\n\n- **Date of image:** March 16, 2017\n- **Geographical locations labeled:**\n    - Major cities: Palermo (northwest Sicily), Marsala (western Sicily), Catania (eastern Sicily)\n    - Significant feature: Mount Etna, labeled with an adjacent \"hot lava\" region showing the glow from active lava flows\n    - Surrounding water body: Mediterranean Sea\n    - Island: Malta to the south of Sicily\n- **Other details:** \n    - The image is shown at night, with bright spots indicating city lights.\n    - The position of \"hot lava\" near Mount Etna is distinctly visible as a bright spot different from other city lights, indicating volcanic activity.\n    - A scale bar is included showing a reference length of 50 km.\n    - North direction is indicated with an arrow.\n    - Cloud cover is visible in the southwest part of the image, partially obscuring the view near Marsala and Malta.\n\n**Summary of Features Visualized:**\n\n| Feature          | Description                                           |\n|------------------|------------------------------------------------------|\n| Cities           | Bright clusters indicating locations: Palermo, Marsala, Catania |\n| Mount Etna       | Marked on the map, located on the eastern side of Sicily, with visible hot lava activity |\n| Malta            | Clearly visible to the south of Sicily               |\n| Water bodies     | Mediterranean Sea labeled                            |\n| Scale & Direction| 50 km scale bar and North indicator                  |\n| Date             | March 16, 2017                                       |\n| Cloud Cover      | Visible in the lower left (southern) part of the image |\n\nThis figure demonstrates the visibility of volcanic activity at Mount Etna from space at night, distinguishing the light from hot lava against the background city lights of Sicily and Malta."
  }
}
Reference Type: KnowledgeAgentAzureSearchDocReference
{
  "type" : "AzureSearchDoc",
  "id" : "2",
  "activitySource" : 1,
  "docKey" : "earth_at_night_508_page_64_verbalized",
  "sourceData" : {
    "id" : "earth_at_night_508_page_64_verbalized",
    "page_chunk" : "<!-- PageHeader=\"Volcanoes\" -->\n\n### Nighttime Glow at Mount Etna - Italy\n\nAt about 2:30 a.m. local time on March 16, 2017, the VIIRS DNB on the Suomi NPP satellite captured this nighttime image of lava flowing on Mount Etna in Sicily, Italy. Etna is one of the world's most active volcanoes.\n\n#### Figure: Location of Mount Etna\nA world globe is depicted, with a marker indicating the ___location of Mount Etna in Sicily, Italy, in southern Europe near the center of the Mediterranean Sea.\n\n<!-- PageFooter=\"Earth at Night\" -->\n<!-- PageNumber=\"48\" -->"
  }
}
Reference Type: KnowledgeAgentAzureSearchDocReference
{
  "type" : "AzureSearchDoc",
  "id" : "3",
  "activitySource" : 1,
  "docKey" : "earth_at_night_508_page_66_verbalized",
  "sourceData" : {
    "id" : "earth_at_night_508_page_66_verbalized",
    "page_chunk" : "# Volcanoes\n\n---\n\n### Mount Etna Erupts - Italy\n\nThe highly active Mount Etna in Italy sent red lava rolling down its flank on March 19, 2017. An astronaut onboard the ISS took the photograph below of the volcano and its environs that night. City lights surround the mostly dark volcanic area.\n\n---\n\n#### Figure 1: Location of Mount Etna, Italy\n\nA world map highlighting the ___location of Mount Etna in southern Italy. The marker indicates its geographic placement on the east coast of Sicily, Italy, in the Mediterranean region, south of mainland Europe and north of northern Africa.\n\n---\n\n#### Figure 2: Nighttime View of Mount Etna's Eruption and Surrounding Cities\n\nThis is a nighttime satellite image taken on March 19, 2017, showing the eruption of Mount Etna (southeastern cone) with visible bright red and orange coloring indicating flowing lava from a lateral vent. The surrounding areas are illuminated by city lights, with the following geographic references labeled:\n\n| Location        | Position in Image         | Visible Characteristics                    |\n|-----------------|--------------------------|--------------------------------------------|\n| Mt. Etna (southeastern cone) | Top center-left | Bright red/orange lava flow                |\n| Lateral vent    | Left of the volcano       | Faint red/orange flow extending outwards   |\n| Resort          | Below the volcano, to the left   | Small cluster of lights                    |\n| Giarre          | Top right                 | Bright cluster of city lights              |\n| Acireale        | Center right              | Large, bright area of city lights          |\n| Biancavilla     | Bottom left               | Smaller cluster of city lights             |\n\nAn arrow pointing north is shown on the image for orientation.\n\n---\n\n<!-- Earth at Night Page Footer -->\n<!-- Page Number: 50 -->"
  }
}
Reference Type: KnowledgeAgentAzureSearchDocReference
{
  "type" : "AzureSearchDoc",
  "id" : "4",
  "activitySource" : 1,
  "docKey" : "earth_at_night_508_page_46_verbalized",
  "sourceData" : {
    "id" : "earth_at_night_508_page_46_verbalized",
    "page_chunk" : "For the first time in perhaps a decade, Mount Etna experienced a \"flank eruption\"�erupting from its side instead of its summit�on December 24, 2018. The activity was accompanied by 130 earthquakes occurring over three hours that morning. Mount Etna, Europe�s most active volcano, has seen periodic activity on this part of the mountain since 2013. The Operational Land Imager (OLI) on the Landsat 8 satellite acquired the main image of Mount Etna on December 28, 2018.\n\nThe inset image highlights the active vent and thermal infrared signature from lava flows, which can be seen near the newly formed fissure on the southeastern side of the volcano. The inset was created with data from OLI and the Thermal Infrared Sensor (TIRS) on Landsat 8. Ash spewing from the fissure cloaked adjacent villages and delayed aircraft from landing at the nearby Catania airport. Earthquakes occurred in the subsequent days after the initial eruption and displaced hundreds of people from their homes.\n\nFor nighttime images of Mount Etna�s March 2017 eruption, see pages 48�51.\n\n---\n\n### Hazards of Volcanic Ash Plumes and Satellite Observation\n\nWith the help of moonlight, satellite instruments can track volcanic ash plumes, which present significant hazards to airplanes in flight. The volcanic ash�composed of tiny pieces of glass and rock�is abrasive to engine turbine blades, and can melt on the blades and other engine parts, causing damage and even engine stalls. This poses a danger to both the plane�s integrity and passenger safety. Volcanic ash also reduces visibility for pilots and can cause etching of windshields, further reducing pilots� ability to see. Nightlight images can be combined with thermal images to provide a more complete view of volcanic activity on Earth�s surface.\n\nThe VIIRS Day/Night Band (DNB) on polar-orbiting satellites uses faint light sources such as moonlight, airglow (the atmosphere�s self-illumination through chemical reactions), zodiacal light (sunlight scattered by interplanetary dust), and starlight from the Milky Way. Using these dim light sources, the DNB can detect changes in clouds, snow cover, and sea ice:\n\n#### Table: Light Sources Used by VIIRS DNB\n\n| Light Source         | Description                                                                  |\n|----------------------|------------------------------------------------------------------------------|\n| Moonlight            | Reflected sunlight from the Moon, illuminating Earth's surface at night      |\n| Airglow              | Atmospheric self-illumination from chemical reactions                        |\n| Zodiacal Light       | Sunlight scattered by interplanetary dust                                    |\n| Starlight/Milky Way  | Faint illumination provided by stars in the Milky Way                        |\n\nGeostationary Operational Environmental Satellites (GOES), managed by NOAA, orbit over Earth�s equator and offer uninterrupted observations of North America. High-latitude areas such as Alaska benefit from polar-orbiting satellites like Suomi NPP, which provide overlapping coverage at the poles, enabling more data collection in these regions. During polar darkness (winter months), VIIRS DNB data allow scientists to:\n\n- Observe sea ice formation\n- Monitor snow cover extent at the highest latitudes\n- Detect open water for ship navigation\n\n#### Table: Satellite Coverage Overview\n\n| Satellite Type          | Orbit           | Coverage Area         | Special Utility                              |\n|------------------------|-----------------|----------------------|----------------------------------------------|\n| GOES                   | Geostationary   | Equatorial/North America | Continuous regional monitoring              |\n| Polar-Orbiting (e.g., Suomi NPP) | Polar-orbiting    | Poles/high latitudes      | Overlapping passes; useful during polar night|\n\n---\n\n### Weather Forecasting and Nightlight Data\n\nThe use of nightlight data by weather forecasters is growing as the VIIRS instrument enables observation of clouds at night illuminated by sources such as moonlight and lightning. Scientists use these data to study the nighttime behavior of weather systems, including severe storms, which can develop and strike populous areas at night as well as during the day. Combined with thermal data, visible nightlight data allow the detection of clouds at various heights in the atmosphere, such as dense marine fog. This capability enables weather forecasters to issue marine advisories with higher confidence, leading to greater utility. (See \"Marine Layer Clouds�California\" on page 56.)\n\nIn this section of the book, you will see how nightlight data are used to observe nature�s spectacular light shows across a wide range of sources.\n\n---\n\n#### Notable Data from Mount Etna Flank Eruption (December 2018)\n\n| Event/Observation                  | Details                                                                    |\n|-------------------------------------|----------------------------------------------------------------------------|\n| Date of Flank Eruption              | December 24, 2018                                                          |\n| Number of Earthquakes               | 130 earthquakes within 3 hours                                              |\n| Image Acquisition                   | December 28, 2018 by Landsat 8 OLI                                         |\n| Location of Eruption                | Southeastern side of Mount Etna                                            |\n| Thermal Imaging Data                | From OLI and TIRS (Landsat 8), highlighting active vent and lava flows     |\n| Impact on Villages/Air Transport    | Ash covered villages; delayed aircraft at Catania airport                  |\n| Displacement                        | Hundreds of residents displaced                                            |\n| Ongoing Seismic Activity            | Earthquakes continued after initial eruption                               |\n\n---\n\n<!-- PageFooter=\"Earth at Night\" -->\n<!-- PageNumber=\"30\" -->"
  }
}
Reference Type: KnowledgeAgentAzureSearchDocReference
{
  "type" : "AzureSearchDoc",
  "id" : "5",
  "activitySource" : 1,
  "docKey" : "earth_at_night_508_page_60_verbalized",
  "sourceData" : {
    "id" : "earth_at_night_508_page_60_verbalized",
    "page_chunk" : "<!-- PageHeader=\"Volcanoes\" -->\n\n## Volcanoes\n\n### The Infrared Glows of Kilauea's Lava Flows�Hawaii\n\nIn early May 2018, an eruption on Hawaii's Kilauea volcano began to unfold. The eruption took a dangerous turn on May 3, 2018, when new fissures opened in the residential neighborhood of Leilani Estates. During the summer-long eruptive event, other fissures emerged along the East Rift Zone. Lava from vents along the rift zone flowed downslope, reaching the ocean in several areas, and filling in Kapoho Bay.\n\nA time series of Landsat 8 imagery shows the progression of the lava flows from May 16 to August 13. The night view combines thermal, shortwave infrared, and near-infrared wavelengths to tease out the very hot lava (bright white), cooling lava (red), and lava flows obstructed by clouds (purple).\n\n#### Figure: Location of Kilauea Volcano, Hawaii\n\nA globe is shown centered on North America, with a marker placed in the Pacific Ocean indicating the ___location of Hawaii, to the southwest of the mainland United States.\n\n<!-- PageFooter=\"Earth at Night\" -->\n<!-- PageNumber=\"44\" -->"
  }
}
[{"ref_id":0,"content":"## Nature's Light Shows\n\nAt night, with the light of the Sun removed, nature's brilliant glow from Earth's surface becomes visible to the naked eye from space. Some of Earth's most spectacular light shows are natural, like the aurora borealis, or Northern Lights, in the Northern Hemisphere (aurora australis, or Southern Lights, in the Southern Hemisphere). The auroras are natural electrical phenomena caused by charged particles that race from the Sun toward Earth, inducing chemical reactions in the upper atmosphere and creating the appearance of streamers of reddish or greenish light in the sky, usually near the northern or southern magnetic pole. Other natural lights can indicate danger, like a raging forest fire encroaching on a city, town, or community, or lava spewing from an erupting volcano.\n\nWhatever the source, the ability of humans to monitor nature's light shows at night has practical applications for society. For example, tracking fires during nighttime hours allows for continuous monitoring and enhances our ability to protect humans and other animals, plants, and infrastructure. Combined with other data sources, our ability to observe the light of fires at night allows emergency managers to more efficiently and accurately issue warnings and evacuation orders and allows firefighting efforts to continue through the night. With enough moonlight (e.g., full-Moon phase), it's even possible to track the movement of smoke plumes at night, which can impact air quality, regardless of time of day.\n\nAnother natural source of light at night is emitted from glowing lava flows at the site of active volcanoes. Again, with enough moonlight, these dramatic scenes can be tracked and monitored for both scientific research and public safety.\n\n\n### Figure: The Northern Lights Viewed from Space\n\n**September 17, 2011**\n\nThis photo, taken from the International Space Station on September 17, 2011, shows a spectacular display of the aurora borealis (Northern Lights) as green and reddish light in the night sky above Earth. In the foreground, part of a Soyuz spacecraft is visible, silhouetted against the bright auroral light. The green glow is generated by energetic charged particles from the Sun interacting with Earth's upper atmosphere, exciting oxygen and nitrogen atoms, and producing characteristic colors. The image demonstrates the vividness and grandeur of natural night-time light phenomena as seen from orbit."},{"ref_id":1,"content":"# Volcanoes\n\n## Figure: Satellite Image of Sicily and Mount Etna Lava, March 16, 2017\n\nThe annotated satellite image below shows the island of Sicily and the surrounding region at night, highlighting city lights and volcanic activity.\n\n**Description:**\n\n- **Date of image:** March 16, 2017\n- **Geographical locations labeled:**\n    - Major cities: Palermo (northwest Sicily), Marsala (western Sicily), Catania (eastern Sicily)\n    - Significant feature: Mount Etna, labeled with an adjacent \"hot lava\" region showing the glow from active lava flows\n    - Surrounding water body: Mediterranean Sea\n    - Island: Malta to the south of Sicily\n- **Other details:** \n    - The image is shown at night, with bright spots indicating city lights.\n    - The position of \"hot lava\" near Mount Etna is distinctly visible as a bright spot different from other city lights, indicating volcanic activity.\n    - A scale bar is included showing a reference length of 50 km.\n    - North direction is indicated with an arrow.\n    - Cloud cover is visible in the southwest part of the image, partially obscuring the view near Marsala and Malta.\n\n**Summary of Features Visualized:**\n\n| Feature          | Description                                           |\n|------------------|------------------------------------------------------|\n| Cities           | Bright clusters indicating locations: Palermo, Marsala, Catania |\n| Mount Etna       | Marked on the map, located on the eastern side of Sicily, with visible hot lava activity |\n| Malta            | Clearly visible to the south of Sicily               |\n| Water bodies     | Mediterranean Sea labeled                            |\n| Scale & Direction| 50 km scale bar and North indicator                  |\n| Date             | March 16, 2017                                       |\n| Cloud Cover      | Visible in the lower left (southern) part of the image |\n\nThis figure demonstrates the visibility of volcanic activity at Mount Etna from space at night, distinguishing the light from hot lava against the background city lights of Sicily and Malta."},{"ref_id":2,"content":"<!-- PageHeader=\"Volcanoes\" -->\n\n### Nighttime Glow at Mount Etna - Italy\n\nAt about 2:30 a.m. local time on March 16, 2017, the VIIRS DNB on the Suomi NPP satellite captured this nighttime image of lava flowing on Mount Etna in Sicily, Italy. Etna is one of the world's most active volcanoes.\n\n#### Figure: Location of Mount Etna\nA world globe is depicted, with a marker indicating the ___location of Mount Etna in Sicily, Italy, in southern Europe near the center of the Mediterranean Sea.\n\n<!-- PageFooter=\"Earth at Night\" -->\n<!-- PageNumber=\"48\" -->"},{"ref_id":3,"content":"# Volcanoes\n\n---\n\n### Mount Etna Erupts - Italy\n\nThe highly active Mount Etna in Italy sent red lava rolling down its flank on March 19, 2017. An astronaut onboard the ISS took the photograph below of the volcano and its environs that night. City lights surround the mostly dark volcanic area.\n\n---\n\n#### Figure 1: Location of Mount Etna, Italy\n\nA world map highlighting the ___location of Mount Etna in southern Italy. The marker indicates its geographic placement on the east coast of Sicily, Italy, in the Mediterranean region, south of mainland Europe and north of northern Africa.\n\n---\n\n#### Figure 2: Nighttime View of Mount Etna's Eruption and Surrounding Cities\n\nThis is a nighttime satellite image taken on March 19, 2017, showing the eruption of Mount Etna (southeastern cone) with visible bright red and orange coloring indicating flowing lava from a lateral vent. The surrounding areas are illuminated by city lights, with the following geographic references labeled:\n\n| Location        | Position in Image         | Visible Characteristics                    |\n|-----------------|--------------------------|--------------------------------------------|\n| Mt. Etna (southeastern cone) | Top center-left | Bright red/orange lava flow                |\n| Lateral vent    | Left of the volcano       | Faint red/orange flow extending outwards   |\n| Resort          | Below the volcano, to the left   | Small cluster of lights                    |\n| Giarre          | Top right                 | Bright cluster of city lights              |\n| Acireale        | Center right              | Large, bright area of city lights          |\n| Biancavilla     | Bottom left               | Smaller cluster of city lights             |\n\nAn arrow pointing north is shown on the image for orientation.\n\n---\n\n<!-- Earth at Night Page Footer -->\n<!-- Page Number: 50 -->"},{"ref_id":4,"content":"For the first time in perhaps a decade, Mount Etna experienced a \"flank eruption\"�erupting from its side instead of its summit�on December 24, 2018. The activity was accompanied by 130 earthquakes occurring over three hours that morning. Mount Etna, Europe�s most active volcano, has seen periodic activity on this part of the mountain since 2013. The Operational Land Imager (OLI) on the Landsat 8 satellite acquired the main image of Mount Etna on December 28, 2018.\n\nThe inset image highlights the active vent and thermal infrared signature from lava flows, which can be seen near the newly formed fissure on the southeastern side of the volcano. The inset was created with data from OLI and the Thermal Infrared Sensor (TIRS) on Landsat 8. Ash spewing from the fissure cloaked adjacent villages and delayed aircraft from landing at the nearby Catania airport. Earthquakes occurred in the subsequent days after the initial eruption and displaced hundreds of people from their homes.\n\nFor nighttime images of Mount Etna�s March 2017 eruption, see pages 48�51.\n\n---\n\n### Hazards of Volcanic Ash Plumes and Satellite Observation\n\nWith the help of moonlight, satellite instruments can track volcanic ash plumes, which present significant hazards to airplanes in flight. The volcanic ash�composed of tiny pieces of glass and rock�is abrasive to engine turbine blades, and can melt on the blades and other engine parts, causing damage and even engine stalls. This poses a danger to both the plane�s integrity and passenger safety. Volcanic ash also reduces visibility for pilots and can cause etching of windshields, further reducing pilots� ability to see. Nightlight images can be combined with thermal images to provide a more complete view of volcanic activity on Earth�s surface.\n\nThe VIIRS Day/Night Band (DNB) on polar-orbiting satellites uses faint light sources such as moonlight, airglow (the atmosphere�s self-illumination through chemical reactions), zodiacal light (sunlight scattered by interplanetary dust), and starlight from the Milky Way. Using these dim light sources, the DNB can detect changes in clouds, snow cover, and sea ice:\n\n#### Table: Light Sources Used by VIIRS DNB\n\n| Light Source         | Description                                                                  |\n|----------------------|------------------------------------------------------------------------------|\n| Moonlight            | Reflected sunlight from the Moon, illuminating Earth's surface at night      |\n| Airglow              | Atmospheric self-illumination from chemical reactions                        |\n| Zodiacal Light       | Sunlight scattered by interplanetary dust                                    |\n| Starlight/Milky Way  | Faint illumination provided by stars in the Milky Way                        |\n\nGeostationary Operational Environmental Satellites (GOES), managed by NOAA, orbit over Earth�s equator and offer uninterrupted observations of North America. High-latitude areas such as Alaska benefit from polar-orbiting satellites like Suomi NPP, which provide overlapping coverage at the poles, enabling more data collection in these regions. During polar darkness (winter months), VIIRS DNB data allow scientists to:\n\n- Observe sea ice formation\n- Monitor snow cover extent at the highest latitudes\n- Detect open water for ship navigation\n\n#### Table: Satellite Coverage Overview\n\n| Satellite Type          | Orbit           | Coverage Area         | Special Utility                              |\n|------------------------|-----------------|----------------------|----------------------------------------------|\n| GOES                   | Geostationary   | Equatorial/North America | Continuous regional monitoring              |\n| Polar-Orbiting (e.g., Suomi NPP) | Polar-orbiting    | Poles/high latitudes      | Overlapping passes; useful during polar night|\n\n---\n\n### Weather Forecasting and Nightlight Data\n\nThe use of nightlight data by weather forecasters is growing as the VIIRS instrument enables observation of clouds at night illuminated by sources such as moonlight and lightning. Scientists use these data to study the nighttime behavior of weather systems, including severe storms, which can develop and strike populous areas at night as well as during the day. Combined with thermal data, visible nightlight data allow the detection of clouds at various heights in the atmosphere, such as dense marine fog. This capability enables weather forecasters to issue marine advisories with higher confidence, leading to greater utility. (See \"Marine Layer Clouds�California\" on page 56.)\n\nIn this section of the book, you will see how nightlight data are used to observe nature�s spectacular light shows across a wide range of sources.\n\n---\n\n#### Notable Data from Mount Etna Flank Eruption (December 2018)\n\n| Event/Observation                  | Details                                                                    |\n|-------------------------------------|----------------------------------------------------------------------------|\n| Date of Flank Eruption              | December 24, 2018                                                          |\n| Number of Earthquakes               | 130 earthquakes within 3 hours                                              |\n| Image Acquisition                   | December 28, 2018 by Landsat 8 OLI                                         |\n| Location of Eruption                | Southeastern side of Mount Etna                                            |\n| Thermal Imaging Data                | From OLI and TIRS (Landsat 8), highlighting active vent and lava flows     |\n| Impact on Villages/Air Transport    | Ash covered villages; delayed aircraft at Catania airport                  |\n| Displacement                        | Hundreds of residents displaced                                            |\n| Ongoing Seismic Activity            | Earthquakes continued after initial eruption                               |\n\n---\n\n<!-- PageFooter=\"Earth at Night\" -->\n<!-- PageNumber=\"30\" -->"},{"ref_id":5,"content":"<!-- PageHeader=\"Volcanoes\" -->\n\n## Volcanoes\n\n### The Infrared Glows of Kilauea's Lava Flows�Hawaii\n\nIn early May 2018, an eruption on Hawaii's Kilauea volcano began to unfold. The eruption took a dangerous turn on May 3, 2018, when new fissures opened in the residential neighborhood of Leilani Estates. During the summer-long eruptive event, other fissures emerged along the East Rift Zone. Lava from vents along the rift zone flowed downslope, reaching the ocean in several areas, and filling in Kapoho Bay.\n\nA time series of Landsat 8 imagery shows the progression of the lava flows from May 16 to August 13. The night view combines thermal, shortwave infrared, and near-infrared wavelengths to tease out the very hot lava (bright white), cooling lava (red), and lava flows obstructed by clouds (purple).\n\n#### Figure: Location of Kilauea Volcano, Hawaii\n\nA globe is shown centered on North America, with a marker placed in the Pacific Ocean indicating the ___location of Hawaii, to the southwest of the mainland United States.\n\n<!-- PageFooter=\"Earth at Night\" -->\n<!-- PageNumber=\"44\" -->"}]

[ASSISTANT]: 
To find lava at night, you can look for the visible glow of active lava flows from erupting volcanoes, which emit light detectable from space during nighttime
 For example:

- The active lava flows of Mount Etna in Sicily, Italy, have been clearly observed at night by satellites and astronauts aboard the International Space Station
 The bright red and orange glow of lava distinguishes it from surrounding city lights (refs 1, 3)


- Similarly, the Kilauea volcano in Hawaii emits an infrared glow from its lava flows, which can be captured in nighttime satellite imagery combining thermal and near-infrared wavelengths (ref 5)


- Nighttime satellite instruments like the VIIRS Day/Night Band (DNB) on the Suomi NPP satellite use faint light sources such as moonlight to detect the glow of lava and volcanic activity even when direct sunlight is absent (refs 2, 4)


Therefore, to find lava at night, monitoring nighttime satellite imagery over active volcanic regions is effective, as the glowing lava stands out distinctly against the dark landscape and city lights


References: [1], [2], [3], [4], [5]

 === Conversation Complete ===
[DELETE] Deleting knowledge agent...
[DONE] Knowledge agent 'earth-search-agent' deleted successfully.
[DELETE] Deleting search index...
[DONE] Search index 'earth_at_night' deleted successfully.
[DONE] Quickstart completed successfully!

코드 이해

이제 코드가 있으므로 주요 구성 요소를 세분화해 보겠습니다.

• 검색 인덱스 만들기
• 인덱스로 문서 업로드
• 지식 에이전트 만들기
• 메시지 설정
• 검색 파이프라인 실행
• 응답, 활동 및 결과 검토
• Azure OpenAI 클라이언트 만들기
• 채팅 완료 API를 사용하여 답변 생성
• 대화 계속

검색 인덱스 만들기

Azure AI Search에서 인덱스는 구조화된 데이터 컬렉션입니다. 다음 코드는 일반 텍스트 및 벡터 콘텐츠를 포함하도록 명명된 earth_at_night 인덱스를 정의합니다. 기존 인덱스를 사용할 수 있지만 에이전트 검색 워크로드에 대한 조건을 충족해야 합니다.

List<SearchField> fields = Arrays.asList(
    new SearchField("id", SearchFieldDataType.STRING)
        .setKey(true)
        .setFilterable(true)
        .setSortable(true)
        .setFacetable(true),
    new SearchField("page_chunk", SearchFieldDataType.STRING)
        .setSearchable(true)
        .setFilterable(false)
        .setSortable(false)
        .setFacetable(false),
    new SearchField("page_embedding_text_3_large", SearchFieldDataType.collection(SearchFieldDataType.SINGLE))
        .setSearchable(true)
        .setFilterable(false)
        .setSortable(false)
        .setFacetable(false)
        .setVectorSearchDimensions(3072)
        .setVectorSearchProfileName("hnsw_text_3_large"),
    new SearchField("page_number", SearchFieldDataType.INT32)
        .setFilterable(true)
        .setSortable(true)
        .setFacetable(true)
);

// Create vectorizer
AzureOpenAIVectorizer vectorizer = new AzureOpenAIVectorizer("azure_openai_text_3_large")
    .setParameters(new AzureOpenAIVectorizerParameters()
        .setResourceUrl(AZURE_OPENAI_ENDPOINT)
        .setDeploymentName(AZURE_OPENAI_EMBEDDING_DEPLOYMENT)
        .setModelName(AzureOpenAIModelName.TEXT_EMBEDDING_3_LARGE));

// Create vector search configuration
VectorSearch vectorSearch = new VectorSearch()
    .setProfiles(Arrays.asList(
        new VectorSearchProfile("hnsw_text_3_large", "alg")
            .setVectorizerName("azure_openai_text_3_large")
    ))
    .setAlgorithms(Arrays.asList(
        new HnswAlgorithmConfiguration("alg")
    ))
    .setVectorizers(Arrays.asList(vectorizer));

// Create semantic search configuration
SemanticSearch semanticSearch = new SemanticSearch()
    .setDefaultConfigurationName("semantic_config")
    .setConfigurations(Arrays.asList(
        new SemanticConfiguration("semantic_config",
            new SemanticPrioritizedFields()
                .setContentFields(Arrays.asList(
                    new SemanticField("page_chunk")
                ))
        )
    ));

// Create the index
SearchIndex index = new SearchIndex(INDEX_NAME)
    .setFields(fields)
    .setVectorSearch(vectorSearch)
    .setSemanticSearch(semanticSearch);

indexClient.createOrUpdateIndex(index);

인덱스 스키마에는 문서 식별 및 페이지 콘텐츠, 포함 및 숫자에 대한 필드가 포함됩니다. 또한 이전에 배포한 text-embedding-3-large 모델을 사용하는 의미 순위 및 벡터 쿼리에 대한 구성도 포함됩니다.

인덱스로 문서 업로드

현재 인덱스는 earth_at_night 비어 있습니다. 다음 코드를 실행하여 인덱스를 NASA의 어스 앳 나이트 전자책의 JSON 문서로 채웁니다. Azure AI Search에서 요구하는 대로 각 문서는 인덱스 스키마에 정의된 필드 및 데이터 형식을 준수합니다.

String documentsUrl = "https://raw.githubusercontent.com/Azure-Samples/azure-search-sample-data/refs/heads/main/nasa-e-book/earth-at-night-json/documents.json";
        
try {
    java.net.http.HttpClient httpClient = java.net.http.HttpClient.newHttpClient();
    java.net.http.HttpRequest request = java.net.http.HttpRequest.newBuilder()
        .uri(URI.create(documentsUrl))
        .build();
    
    java.net.http.HttpResponse<String> response = httpClient.send(request, 
        java.net.http.HttpResponse.BodyHandlers.ofString());
    
    if (response.statusCode() != 200) {
        throw new IOException("Failed to fetch documents: " + response.statusCode());
    }
    
    ObjectMapper mapper = new ObjectMapper();
    JsonNode jsonArray = mapper.readTree(response.body());
    
    List<SearchDocument> documents = new ArrayList<>();
    for (int i = 0; i < jsonArray.size(); i++) {
        JsonNode doc = jsonArray.get(i);
        SearchDocument searchDoc = new SearchDocument();
        
        searchDoc.put("id", doc.has("id") ? doc.get("id").asText() : String.valueOf(i + 1));
        searchDoc.put("page_chunk", doc.has("page_chunk") ? doc.get("page_chunk").asText() : "");
        
        // Handle embeddings
        if (doc.has("page_embedding_text_3_large") && doc.get("page_embedding_text_3_large").isArray()) {
            List<Double> embeddings = new ArrayList<>();
            for (JsonNode embedding : doc.get("page_embedding_text_3_large")) {
                embeddings.add(embedding.asDouble());
            }
            searchDoc.put("page_embedding_text_3_large", embeddings);
        } else {
            // Fallback embeddings
            List<Double> fallbackEmbeddings = new ArrayList<>();
            for (int j = 0; j < 3072; j++) {
                fallbackEmbeddings.add(0.1);
            }
            searchDoc.put("page_embedding_text_3_large", fallbackEmbeddings);
        }
        
        searchDoc.put("page_number", doc.has("page_number") ? doc.get("page_number").asInt() : i + 1);
        
        documents.add(searchDoc);
    }
    
    System.out.println("[DONE] Fetched " + documents.size() + " documents from GitHub");
    return documents;
    
}

지식 에이전트 만들기

Azure AI Search를 gpt-5-mini 배포에 연결하고 쿼리 시 earth_at_night 인덱스를 대상으로 지정하려면, 지식 에이전트가 필요합니다. 다음 코드는 에이전트 정의를 사용하여 쿼리를 처리하고 earth-search-agent 인덱스에서 관련 문서를 검색하는 earth_at_night라는 이름의 지식 에이전트를 정의합니다.

관련성이 있고 의미 있는 응답을 defaultRerankerThreshold 보장하기 위해 재랜커 점수 2.5 가 낮거나 낮은 응답을 제외하도록 설정됩니다.

ObjectMapper mapper = new ObjectMapper();
ObjectNode agentDefinition = mapper.createObjectNode();
agentDefinition.put("name", AGENT_NAME);
agentDefinition.put("description", "Knowledge agent for Earth at Night e-book content");

ObjectNode model = mapper.createObjectNode();
model.put("kind", "azureOpenAI");
ObjectNode azureOpenAIParams = mapper.createObjectNode();
azureOpenAIParams.put("resourceUri", AZURE_OPENAI_ENDPOINT);
azureOpenAIParams.put("deploymentId", AZURE_OPENAI_GPT_DEPLOYMENT);
azureOpenAIParams.put("modelName", AZURE_OPENAI_GPT_MODEL);
model.set("azureOpenAIParameters", azureOpenAIParams);
agentDefinition.set("models", mapper.createArrayNode().add(model));

ObjectNode targetIndex = mapper.createObjectNode();
targetIndex.put("indexName", INDEX_NAME);
targetIndex.put("defaultRerankerThreshold", 2.5);
agentDefinition.set("targetIndexes", mapper.createArrayNode().add(targetIndex));

String token = getAccessToken(credential, "https://search.azure.com/.default");

java.net.http.HttpClient httpClient = java.net.http.HttpClient.newHttpClient();
java.net.http.HttpRequest request = java.net.http.HttpRequest.newBuilder()
    .uri(URI.create(SEARCH_ENDPOINT + "/agents/" + AGENT_NAME + "?api-version=" + SEARCH_API_VERSION))
    .header("Content-Type", "application/json")
    .header("Authorization", "Bearer " + token)
    .PUT(java.net.http.HttpRequest.BodyPublishers.ofString(mapper.writeValueAsString(agentDefinition)))
    .build();

java.net.http.HttpResponse<String> response = httpClient.send(request,
    java.net.http.HttpResponse.BodyHandlers.ofString());

메시지 설정

메시지는 검색 경로에 대한 입력이며 대화 기록을 포함합니다. 각 메시지에는 도우미 또는 사용자와 같은 원본을 나타내는 역할과 자연어로 된 콘텐츠가 포함됩니다. 사용하는 LLM은 유효한 역할을 결정합니다.

사용자 메시지는 처리할 쿼리를 나타내고 보조 메시지는 기술 에이전트에 응답하는 방법을 안내합니다. 검색 프로세스 중에 이러한 메시지는 인덱싱된 문서에서 관련 응답을 추출하기 위해 LLM으로 전송됩니다.

이 보조 메시지는 밤에 지구에 대한 질문에 대답하고, 소스를 earth-search-agent사용하여 인용하고, 답변을 사용할 수 없을 때 "나는 모른다"로 응답하도록 지시 ref_id 합니다.

List<Map<String, String>> messages = new ArrayList<>();

Map<String, String> systemMessage = new HashMap<>();
systemMessage.put("role", "system");
systemMessage.put("content", "A Q&A agent that can answer questions about the Earth at night.\n" +
    "Sources have a JSON format with a ref_id that must be cited in the answer.\n" +
    "If you do not have the answer, respond with \"I don't know\".");
messages.add(systemMessage);

Map<String, String> userMessage = new HashMap<>();
userMessage.put("role", "user");
userMessage.put("content", "Why do suburban belts display larger December brightening than urban cores even though absolute light levels are higher downtown? Why is the Phoenix nighttime street grid is so sharply visible from space, whereas large stretches of the interstate between midwestern cities remain comparatively dim?");
messages.add(userMessage);

검색 파이프라인 실행

이 단계에서는 검색 파이프라인을 실행하여 검색 인덱스에서 관련 정보를 추출합니다. 검색 요청의 메시지 및 매개 변수에 따라 LLM은 다음을 수행합니다.

1. 전체 대화 기록을 분석하여 기본 정보가 필요한지 확인합니다.
2. 복합 사용자 쿼리를 포커스가 있는 하위 쿼리로 구분합니다.
3. 인덱스의 텍스트 필드 및 벡터 포함에 대해 각 하위 쿼리를 동시에 실행합니다.
4. 의미 체계 순위자를 사용하여 모든 하위 쿼리의 결과를 다시 표시합니다.
5. 결과를 단일 문자열로 병합합니다.

다음 코드는 쿼리를 하위 쿼리로 분해하고, 인덱스의 텍스트 필드와 벡터 포함에 대해 하위 쿼리를 실행하고, 결과의 순위를 지정하고 병합하는 earth-search-agent 두 부분으로 구성된 사용자 쿼리earth_at_night를 보냅니다. 그런 다음 응답이 목록에 추가됩니다 messages .

ObjectMapper mapper = new ObjectMapper();
ObjectNode retrievalRequest = mapper.createObjectNode();

// Convert messages to the correct format expected by the Knowledge agent
com.fasterxml.jackson.databind.node.ArrayNode agentMessages = mapper.createArrayNode();
for (Map<String, String> msg : messages) {
    ObjectNode agentMessage = mapper.createObjectNode();
    agentMessage.put("role", msg.get("role"));
    
    com.fasterxml.jackson.databind.node.ArrayNode content = mapper.createArrayNode();
    ObjectNode textContent = mapper.createObjectNode();
    textContent.put("type", "text");
    textContent.put("text", msg.get("content"));
    content.add(textContent);
    agentMessage.set("content", content);
    
    agentMessages.add(agentMessage);
}
retrievalRequest.set("messages", agentMessages);

com.fasterxml.jackson.databind.node.ArrayNode targetIndexParams = mapper.createArrayNode();
ObjectNode indexParam = mapper.createObjectNode();
indexParam.put("indexName", INDEX_NAME);
indexParam.put("rerankerThreshold", 2.5);
indexParam.put("maxDocsForReranker", 100);
indexParam.put("includeReferenceSourceData", true);
targetIndexParams.add(indexParam);
retrievalRequest.set("targetIndexParams", targetIndexParams);

String token = getAccessToken(credential, "https://search.azure.com/.default");

java.net.http.HttpClient httpClient = java.net.http.HttpClient.newHttpClient();
java.net.http.HttpRequest request = java.net.http.HttpRequest.newBuilder()
    .uri(URI.create(SEARCH_ENDPOINT + "/agents/" + AGENT_NAME + "/retrieve?api-version=" + SEARCH_API_VERSION))
    .header("Content-Type", "application/json")
    .header("Authorization", "Bearer " + token)
    .POST(java.net.http.HttpRequest.BodyPublishers.ofString(mapper.writeValueAsString(retrievalRequest)))
    .build();

java.net.http.HttpResponse<String> response = httpClient.send(request,
    java.net.http.HttpResponse.BodyHandlers.ofString());

응답, 활동 및 결과 검토

이제 검색 파이프라인의 응답, 활동 및 결과를 표시하려고 합니다.

Azure AI Search의 각 검색 응답에는 다음이 포함됩니다.

• 검색 결과의 접지 데이터를 나타내는 통합 문자열입니다.

• 쿼리 계획.

• 통합 문자열에 기여한 원본 문서의 청크를 보여 주는 참조 데이터입니다.

ObjectMapper mapper = new ObjectMapper();
        
// Log activities
System.out.println("\nActivities:");
if (responseJson.has("activity") && responseJson.get("activity").isArray()) {
    for (JsonNode activity : responseJson.get("activity")) {
        String activityType = "UnknownActivityRecord";
        if (activity.has("InputTokens")) {
            activityType = "KnowledgeAgentModelQueryPlanningActivityRecord";
        } else if (activity.has("TargetIndex")) {
            activityType = "KnowledgeAgentSearchActivityRecord";
        } else if (activity.has("QueryTime")) {
            activityType = "KnowledgeAgentSemanticRankerActivityRecord";
        }
        
        System.out.println("Activity Type: " + activityType);
        try {
            System.out.println(mapper.writerWithDefaultPrettyPrinter().writeValueAsString(activity));
        } catch (Exception e) {
            System.out.println(activity.toString());
        }
    }
}

// Log results
System.out.println("Results");
if (responseJson.has("references") && responseJson.get("references").isArray()) {
    for (JsonNode reference : responseJson.get("references")) {
        String referenceType = "KnowledgeAgentAzureSearchDocReference";
        
        System.out.println("Reference Type: " + referenceType);
        try {
            System.out.println(mapper.writerWithDefaultPrettyPrinter().writeValueAsString(reference));
        } catch (Exception e) {
            System.out.println(reference.toString());
        }
    }
}

출력에는 다음이 포함되어야 합니다.

• Response 는 사용자 쿼리를 기반으로 검색 인덱스에 가장 관련성이 큰 문서(또는 청크)의 텍스트 문자열을 제공합니다. 이 빠른 시작의 뒷부분에서와 같이 응답 생성을 위해 이 문자열을 LLM에 전달할 수 있습니다.

• Activity 는 배포에서 생성된 하위 쿼리 및 쿼리 계획 및 실행에 사용되는 토큰을 포함하여 검색 프로세스 중에 수행된 gpt-5-mini 단계를 추적합니다.

• Results 에는 응답에 기여한 문서가 나열되고 각 문서는 해당 DocKey문서로 식별됩니다.

Azure OpenAI 클라이언트 만들기

응답 추출에서 응답 생성으로 검색 처리 단계를 확장하려면 배포와 상호 작용할 수 있도록 Azure OpenAI 클라이언트를 설정합니다.

OpenAIAsyncClient openAIClient = new OpenAIClientBuilder()
    .endpoint(AZURE_OPENAI_ENDPOINT)
    .credential(credential)
    .buildAsyncClient();

채팅 완료 API를 사용하여 답변 생성

응답 생성을 위한 한 가지 옵션은 대화를 처리하기 위해 LLM에 대화 기록을 전달하는 채팅 완료 API입니다.

List<ChatRequestMessage> chatMessages = new ArrayList<>();
for (Map<String, String> msg : messages) {
    String role = msg.get("role");
    String content = msg.get("content");
    
    switch (role) {
        case "system":
            chatMessages.add(new ChatRequestSystemMessage(content));
            break;
        case "user":
            chatMessages.add(new ChatRequestUserMessage(content));
            break;
        case "assistant":
            chatMessages.add(new ChatRequestAssistantMessage(content));
            break;
    }
}

ChatCompletionsOptions chatOptions = new ChatCompletionsOptions(chatMessages)
    .setMaxTokens(1000)
    .setTemperature(0.7);

ChatCompletions completion = openAIClient.getChatCompletions(AZURE_OPENAI_GPT_DEPLOYMENT, chatOptions).block();

대화 계속

다른 사용자에게 쿼리를 보내 대화를 계속하세요 earth-search-agent. 다음 코드는 검색 파이프라인을 다시 실행하여 인덱스에서 earth_at_night 관련 콘텐츠를 가져오고 목록에 응답을 messages 추가합니다. 그러나 이전과 달리 이제 Azure OpenAI 클라이언트를 사용하여 검색된 콘텐츠에 따라 답변을 생성할 수 있습니다.

String followUpQuestion = "How do I find lava at night?";
System.out.println("[QUESTION] Follow-up question: " + followUpQuestion);

Map<String, String> userMessage = new HashMap<>();
userMessage.put("role", "user");
userMessage.put("content", followUpQuestion);
messages.add(userMessage);

자원을 정리하세요

자체 구독에서 작업할 때 만든 리소스가 여전히 필요한지 여부를 결정하여 프로젝트를 완료하는 것이 좋습니다. 실행 상태로 방치된 리소스는 비용이 들 수 있습니다. 리소스를 개별적으로 삭제하거나 리소스 그룹을 삭제하여 전체 리소스 집합을 삭제할 수 있습니다.

Azure Portal의 왼쪽 창에서 모든 리소스 또는 리소스 그룹을 선택하여 리소스 를 찾고 관리할 수 있습니다. 다음 코드를 실행하여 이 빠른 시작에서 만든 개체를 삭제할 수도 있습니다.

지식 에이전트 삭제

이 빠른 시작에서 만든 지식 에이전트는 다음 코드 조각을 사용하여 삭제되었습니다.

String token = getAccessToken(credential, "https://search.azure.com/.default");
            
java.net.http.HttpClient httpClient = java.net.http.HttpClient.newHttpClient();
java.net.http.HttpRequest request = java.net.http.HttpRequest.newBuilder()
    .uri(URI.create(SEARCH_ENDPOINT + "/agents/" + AGENT_NAME + "?api-version=" + SEARCH_API_VERSION))
    .header("Authorization", "Bearer " + token)
    .DELETE()
    .build();

java.net.http.HttpResponse<String> response = httpClient.send(request,
    java.net.http.HttpResponse.BodyHandlers.ofString());

검색 인덱스 삭제

이 빠른 시작에서 만든 검색 인덱스가 다음 코드 조각을 사용하여 삭제되었습니다.

indexClient.deleteIndex(INDEX_NAME);
System.out.println("[DONE] Search index '" + INDEX_NAME + "' deleted successfully.");

이 빠른 시작에서는 에이전트 검색 을 사용하여 LLM(대규모 언어 모델) 및 독점 데이터로 구동되는 대화형 검색 환경을 만듭니다. 에이전트 검색은 복잡한 사용자 쿼리를 하위 쿼리로 분해하고, 하위 쿼리를 병렬로 실행하고, Azure AI Search에서 인덱싱된 문서에서 접지 데이터를 추출합니다. 출력은 에이전트 및 사용자 지정 채팅 솔루션과 통합하기 위한 것입니다.

사용자 고유의 데이터를 제공할 수 있지만, 이 빠른 시작에서는 NASA의 어스 앳 나이트 전자책의 샘플 JSON 문서를 사용합니다. 문서는 우주에서 관찰 된 바와 같이 밤에 지구의 일반적인 과학 주제와 이미지를 설명합니다.

필수 조건

• 활성 구독이 있는 Azure 계정. 무료로 계정을 만듭니다.

• 의미 체계 순위가 설정된 기본 계층 이상의 Azure AI Search 서비스입니다.

• Azure AI Foundry 프로젝트입니다. Azure AI Foundry 프로젝트를 만들 때 모델 배포에 필요한 Azure AI Foundry 리소스를 가져옵니다.

• Microsoft Entra ID를 사용한 키 없는 인증을 위한 Azure CLI 입니다.

액세스 구성

시작하기 전에 콘텐츠 및 작업에 액세스할 수 있는 권한이 있는지 확인합니다. 인증에는 Microsoft Entra ID를, 권한 부여에는 역할 기반 액세스에 사용하는 것이 좋습니다. 역할을 할당하려면 소유자 또는 사용자 액세스 관리자여야 합니다. 역할이 가능하지 않은 경우 대신 키 기반 인증을 사용합니다.

이 빠른 시작에 대한 액세스를 구성하려면 다음 탭을 모두 선택합니다.

엔드포인트 가져오기

각 Azure AI 검색 서비스와 Azure AI Foundry 리소스에는 리소스에 대한 네트워크 액세스를 식별하고 제공하는 고유한 URL인 엔드포인트가 있습니다. 이후 섹션에서는 이러한 엔드포인트를 지정하여 프로그래밍 방식으로 리소스에 연결합니다.

이 빠른 시작의 엔드포인트를 얻으려면 다음 탭을 모두 선택합니다.

모델 배포

에이전트 검색을 사용하려면 Azure AI Foundry 리소스에 두 개의 Azure OpenAI 모델을 배포해야 합니다.

• 텍스트-벡터 변환을 위한 포함 모델입니다. 이 빠른 시작에서는 text-embedding-3-large를 사용하지만, 모든 text-embedding 모델을 사용할 수 있습니다.

• 쿼리 계획 및 응답 생성을 위한 LLM입니다. 이 빠른 시작에서는 gpt-5-mini를 사용하지만, 에이전트 검색을 위해 지원되는 어떠한 LLM도 사용할 수 있습니다.

배포 지침은 Azure AI Foundry를 사용하여 Azure OpenAI 모델 배포를 참조하세요.

환경 설정

1. 애플리케이션을 포함할 새 폴더 quickstart-agentic-retrieval을 만들고 다음 명령을 사용하여 해당 폴더에서 Visual Studio Code를 엽니다.

   mkdir quickstart-agentic-retrieval && cd quickstart-agentic-retrieval
   

2. 다음 명령을 사용하여 package.json 만듭니다.

   npm init -y
   

3. JavaScript용 Azure AI 검색 클라이언트 라이브러리(Azure.Search.Documents)를 설치하려면 다음을 사용합니다.

   npm install @azure/search-documents --version 12.2.0-alpha.20250606.1
   

4. 다음을 사용하여 Azure OpenAI 클라이언트 라이브러리를 설치합니다.

   npm install @azure/openai --version 5.10.1
   

5. dotenv 패키지를 설치하여 .env 파일에서 환경 변수를 로드합니다.

   npm install dotenv
   

6. Microsoft Entra ID로 권장되는 키 없는 인증의 경우 다음을 사용하여 Azure ID 클라이언트 라이브러리를 설치합니다.

   npm install @azure/identity
   

코드 실행

1. 폴더에 명명된 .envquickstart-agentic-retrieval 새 파일을 만들고 다음 환경 변수를 추가합니다.

   AZURE_OPENAI_ENDPOINT=https://<your-ai-foundry-resource-name>.openai.azure.com/
   AZURE_OPENAI_GPT_DEPLOYMENT=gpt-5-mini
   AZURE_OPENAI_EMBEDDING_DEPLOYMENT=text-embedding-3-large
   AZURE_SEARCH_ENDPOINT=https://<your-search-service-name>.search.windows.net
   AZURE_SEARCH_INDEX_NAME=agentic-retrieval-sample
   

   <your-search-service-name> 및 <your-ai-foundry-resource-name>을 실제 Azure AI Search 서비스 이름과 Azure AI Foundry 리소스 이름으로 대체하십시오.

2. 다음 코드를 다음과 같은 새 index.js파일에 붙여넣습니다.

   import { DefaultAzureCredential, getBearerTokenProvider } from '@azure/identity';
   import { SearchIndexClient, SearchClient } from '@azure/search-documents';
   import { AzureOpenAI } from "openai/index.mjs";
   // Load the .env file if it exists
   import * as dotenv from "dotenv";
   dotenv.config();
   // Configuration - Update these values for your environment
   const config = {
       searchEndpoint: process.env.AZURE_SEARCH_ENDPOINT || "https://your-search-service.search.windows.net",
       azureOpenAIEndpoint: process.env.AZURE_OPENAI_ENDPOINT || "https://your-ai-foundry-resource.openai.azure.com/",
       azureOpenAIGptDeployment: process.env.AZURE_OPENAI_GPT_DEPLOYMENT || "gpt-5-mini",
       azureOpenAIGptModel: "gpt-5-mini",
       azureOpenAIApiVersion: process.env.OPENAI_API_VERSION || "2025-03-01-preview",
       azureOpenAIEmbeddingDeployment: process.env.AZURE_OPENAI_EMBEDDING_DEPLOYMENT || "text-embedding-3-large",
       azureOpenAIEmbeddingModel: "text-embedding-3-large",
       indexName: "earth_at_night",
       agentName: "earth-search-agent",
       searchApiVersion: "2025-05-01-Preview"
   };
   async function main() {
       try {
           console.log("🚀 Starting Azure AI Search agentic retrieval quickstart...\n");
           // Initialize Azure credentials using managed identity (recommended)
           const credential = new DefaultAzureCredential();
           // Create search clients
           const searchIndexClient = new SearchIndexClient(config.searchEndpoint, credential);
           const searchClient = new SearchClient(config.searchEndpoint, config.indexName, credential);
           // Create Azure OpenAI client
           const scope = "https://cognitiveservices.azure.com/.default";
           const azureADTokenProvider = getBearerTokenProvider(credential, scope);
           const openAIClient = new AzureOpenAI({
               endpoint: config.azureOpenAIEndpoint,
               apiVersion: config.azureOpenAIApiVersion,
               azureADTokenProvider,
           });
           // Create search index with vector and semantic capabilities
           await createSearchIndex(searchIndexClient);
           // Upload sample documents
           await uploadDocuments(searchClient);
           // Create knowledge agent for agentic retrieval
           await createKnowledgeAgent(credential);
           // Run agentic retrieval with conversation
           await runAgenticRetrieval(credential, openAIClient);
           // Clean up - Delete knowledge agent and search index
           await deleteKnowledgeAgent(credential);
           await deleteSearchIndex(searchIndexClient);
           console.log("✅ Quickstart completed successfully!");
       }
       catch (error) {
           console.error("❌ Error in main execution:", error);
           throw error;
       }
   }
   async function createSearchIndex(indexClient) {
       console.log("📊 Creating search index...");
       const index = {
           name: config.indexName,
           fields: [
               {
                   name: "id",
                   type: "Edm.String",
                   key: true,
                   filterable: true,
                   sortable: true,
                   facetable: true
               },
               {
                   name: "page_chunk",
                   type: "Edm.String",
                   searchable: true,
                   filterable: false,
                   sortable: false,
                   facetable: false
               },
               {
                   name: "page_embedding_text_3_large",
                   type: "Collection(Edm.Single)",
                   searchable: true,
                   filterable: false,
                   sortable: false,
                   facetable: false,
                   vectorSearchDimensions: 3072,
                   vectorSearchProfileName: "hnsw_text_3_large"
               },
               {
                   name: "page_number",
                   type: "Edm.Int32",
                   filterable: true,
                   sortable: true,
                   facetable: true
               }
           ],
           vectorSearch: {
               profiles: [
                   {
                       name: "hnsw_text_3_large",
                       algorithmConfigurationName: "alg",
                       vectorizerName: "azure_openai_text_3_large"
                   }
               ],
               algorithms: [
                   {
                       name: "alg",
                       kind: "hnsw"
                   }
               ],
               vectorizers: [
                   {
                       vectorizerName: "azure_openai_text_3_large",
                       kind: "azureOpenAI",
                       parameters: {
                           resourceUrl: config.azureOpenAIEndpoint,
                           deploymentId: config.azureOpenAIEmbeddingDeployment,
                           modelName: config.azureOpenAIEmbeddingModel
                       }
                   }
               ]
           },
           semanticSearch: {
               defaultConfigurationName: "semantic_config",
               configurations: [
                   {
                       name: "semantic_config",
                       prioritizedFields: {
                           contentFields: [
                               { name: "page_chunk" }
                           ]
                       }
                   }
               ]
           }
       };
       try {
           await indexClient.createOrUpdateIndex(index);
           console.log(`✅ Index '${config.indexName}' created or updated successfully.`);
       }
       catch (error) {
           console.error("❌ Error creating index:", error);
           throw error;
       }
   }
   async function deleteSearchIndex(indexClient) {
       console.log("🗑️ Deleting search index...");
       try {
           await indexClient.deleteIndex(config.indexName);
           console.log(`✅ Search index '${config.indexName}' deleted successfully.`);
       }
       catch (error) {
           if (error?.statusCode === 404 || error?.code === 'IndexNotFound') {
               console.log(`ℹ️ Search index '${config.indexName}' does not exist or was already deleted.`);
               return;
           }
           console.error("❌ Error deleting search index:", error);
           throw error;
       }
   }
   // Fetch Earth at Night documents from GitHub
   async function fetchEarthAtNightDocuments() {
       console.log("📡 Fetching Earth at Night documents from GitHub...");
       const documentsUrl = "https://raw.githubusercontent.com/Azure-Samples/azure-search-sample-data/refs/heads/main/nasa-e-book/earth-at-night-json/documents.json";
       try {
           const response = await fetch(documentsUrl);
           if (!response.ok) {
               throw new Error(`Failed to fetch documents: ${response.status} ${response.statusText}`);
           }
           const documents = await response.json();
           console.log(`✅ Fetched ${documents.length} documents from GitHub`);
           // Validate and transform documents to match our interface
           const transformedDocuments = documents.map((doc, index) => {
               return {
                   id: doc.id || String(index + 1),
                   page_chunk: doc.page_chunk || doc.content || '',
                   page_embedding_text_3_large: doc.page_embedding_text_3_large || new Array(3072).fill(0.1),
                   page_number: doc.page_number || index + 1
               };
           });
           return transformedDocuments;
       }
       catch (error) {
           console.error("❌ Error fetching documents from GitHub:", error);
           console.log("🔄 Falling back to sample documents...");
           // Fallback to sample documents if fetch fails
           return [
               {
                   id: "1",
                   page_chunk: "The Earth at night reveals the patterns of human settlement and economic activity. City lights trace the contours of civilization, creating a luminous map of where people live and work.",
                   page_embedding_text_3_large: new Array(3072).fill(0.1),
                   page_number: 1
               },
               {
                   id: "2",
                   page_chunk: "From space, the aurora borealis appears as shimmering curtains of green and blue light dancing across the polar regions.",
                   page_embedding_text_3_large: new Array(3072).fill(0.2),
                   page_number: 2
               }
               // Add more fallback documents as needed
           ];
       }
   }
   async function uploadDocuments(searchClient) {
       console.log("📄 Uploading documents...");
       try {
           // Fetch documents from GitHub
           const documents = await fetchEarthAtNightDocuments();
           const result = await searchClient.uploadDocuments(documents);
           console.log(`✅ Uploaded ${result.results.length} documents successfully.`);
           // Wait for indexing to complete
           console.log("⏳ Waiting for document indexing to complete...");
           await new Promise(resolve => setTimeout(resolve, 5000));
           console.log("✅ Document indexing completed.");
       }
       catch (error) {
           console.error("❌ Error uploading documents:", error);
           throw error;
       }
   }
   async function createKnowledgeAgent(credential) {
       // In case the agent already exists, delete it first
       await deleteKnowledgeAgent(credential);
       console.log("🤖 Creating knowledge agent...");
       const agentDefinition = {
           name: config.agentName,
           description: "Knowledge agent for Earth at Night e-book content",
           models: [
               {
                   kind: "azureOpenAI",
                   azureOpenAIParameters: {
                       resourceUri: config.azureOpenAIEndpoint,
                       deploymentId: config.azureOpenAIGptDeployment,
                       modelName: config.azureOpenAIGptModel
                   }
               }
           ],
           targetIndexes: [
               {
                   indexName: config.indexName,
                   defaultRerankerThreshold: 2.5
               }
           ]
       };
       try {
           const token = await getAccessToken(credential, "https://search.azure.com/.default");
           const response = await fetch(`${config.searchEndpoint}/agents/${config.agentName}?api-version=${config.searchApiVersion}`, {
               method: 'PUT',
               headers: {
                   'Content-Type': 'application/json',
                   'Authorization': `Bearer ${token}`
               },
               body: JSON.stringify(agentDefinition)
           });
           if (!response.ok) {
               const errorText = await response.text();
               throw new Error(`Failed to create knowledge agent: ${response.status} ${response.statusText}\n${errorText}`);
           }
           console.log(`✅ Knowledge agent '${config.agentName}' created successfully.`);
       }
       catch (error) {
           console.error("❌ Error creating knowledge agent:", error);
           throw error;
       }
   }
   async function runAgenticRetrieval(credential, openAIClient) {
       console.log("🔍 Running agentic retrieval...");
       const messages = [
           {
               role: "system",
               content: `A Q&A agent that can answer questions about the Earth at night.
   Sources have a JSON format with a ref_id that must be cited in the answer.
   If you do not have the answer, respond with "I don't know".`
           },
           {
               role: "user",
               content: "Why do suburban belts display larger December brightening than urban cores even though absolute light levels are higher downtown? Why is the Phoenix nighttime street grid is so sharply visible from space, whereas large stretches of the interstate between midwestern cities remain comparatively dim?"
           }
       ];
       try {
           // Call agentic retrieval API
           const userMessages = messages.filter(m => m.role !== "system");
           const retrievalResponse = await callAgenticRetrieval(credential, userMessages);
           // Extract the assistant response from agentic retrieval
           let assistantContent = '';
           if (typeof retrievalResponse.response === 'string') {
               assistantContent = retrievalResponse.response;
           }
           else if (Array.isArray(retrievalResponse.response)) {
               assistantContent = JSON.stringify(retrievalResponse.response);
           }
           // Add assistant response to conversation history
           messages.push({
               role: "assistant",
               content: assistantContent
           });
           console.log(assistantContent);
           // Log activities and results...
           console.log("\nActivities:");
           if (retrievalResponse.activity && Array.isArray(retrievalResponse.activity)) {
               retrievalResponse.activity.forEach((activity) => {
                   const activityType = activity.activityType || activity.type || 'UnknownActivityRecord';
                   console.log(`Activity Type: ${activityType}`);
                   console.log(JSON.stringify(activity, null, 2));
               });
           }
           console.log("Results");
           if (retrievalResponse.references && Array.isArray(retrievalResponse.references)) {
               retrievalResponse.references.forEach((reference) => {
                   const referenceType = reference.referenceType || reference.type || 'AzureSearchDoc';
                   console.log(`Reference Type: ${referenceType}`);
                   console.log(JSON.stringify(reference, null, 2));
               });
           }
           // Now do chat completion with full conversation history
           await generateFinalAnswer(openAIClient, messages);
           // Continue conversation with second question
           await continueConversation(credential, openAIClient, messages);
       }
       catch (error) {
           console.error("❌ Error in agentic retrieval:", error);
           throw error;
       }
   }
   async function generateFinalAnswer(openAIClient, messages) {
       console.log("\n[ASSISTANT]: ");
       try {
           const completion = await openAIClient.chat.completions.create({
               model: config.azureOpenAIGptDeployment,
               messages: messages.map(m => ({ role: m.role, content: m.content })),
               max_tokens: 1000,
               temperature: 0.7
           });
           const answer = completion.choices[0].message.content;
           console.log(answer?.replace(/\./g, "\n"));
           // Add this response to conversation history
           if (answer) {
               messages.push({
                   role: "assistant",
                   content: answer
               });
           }
       }
       catch (error) {
           console.error("❌ Error generating final answer:", error);
           throw error;
       }
   }
   async function callAgenticRetrieval(credential, messages) {
       // Convert messages to the correct format expected by the Knowledge agent
       const agentMessages = messages.map(msg => ({
           role: msg.role,
           content: [
               {
                   type: "text",
                   text: msg.content
               }
           ]
       }));
       const retrievalRequest = {
           messages: agentMessages,
           targetIndexParams: [
               {
                   indexName: config.indexName,
                   rerankerThreshold: 2.5,
                   maxDocsForReranker: 100,
                   includeReferenceSourceData: true
               }
           ]
       };
       const token = await getAccessToken(credential, "https://search.azure.com/.default");
       const response = await fetch(`${config.searchEndpoint}/agents/${config.agentName}/retrieve?api-version=${config.searchApiVersion}`, {
           method: 'POST',
           headers: {
               'Content-Type': 'application/json',
               'Authorization': `Bearer ${token}`
           },
           body: JSON.stringify(retrievalRequest)
       });
       if (!response.ok) {
           const errorText = await response.text();
           throw new Error(`Agentic retrieval failed: ${response.status} ${response.statusText}\n${errorText}`);
       }
       return await response.json();
   }
   async function deleteKnowledgeAgent(credential) {
       console.log("🗑️ Deleting knowledge agent...");
       try {
           const token = await getAccessToken(credential, "https://search.azure.com/.default");
           const response = await fetch(`${config.searchEndpoint}/agents/${config.agentName}?api-version=${config.searchApiVersion}`, {
               method: 'DELETE',
               headers: {
                   'Authorization': `Bearer ${token}`
               }
           });
           if (!response.ok) {
               if (response.status === 404) {
                   console.log(`ℹ️ Knowledge agent '${config.agentName}' does not exist or was already deleted.`);
                   return;
               }
               const errorText = await response.text();
               throw new Error(`Failed to delete knowledge agent: ${response.status} ${response.statusText}\n${errorText}`);
           }
           console.log(`✅ Knowledge agent '${config.agentName}' deleted successfully.`);
       }
       catch (error) {
           console.error("❌ Error deleting knowledge agent:", error);
           throw error;
       }
   }
   async function continueConversation(credential, openAIClient, messages) {
       console.log("\n💬 === Continuing Conversation ===");
       // Add follow-up question
       const followUpQuestion = "How do I find lava at night?";
       console.log(`❓ Follow-up question: ${followUpQuestion}`);
       messages.push({
           role: "user",
           content: followUpQuestion
       });
       try {
           // Don't include system messages in this retrieval
           const userAssistantMessages = messages.filter((m) => m.role !== "system");
           const newRetrievalResponse = await callAgenticRetrieval(credential, userAssistantMessages);
           // Extract assistant response and add to conversation
           let assistantContent = '';
           if (typeof newRetrievalResponse.response === 'string') {
               assistantContent = newRetrievalResponse.response;
           }
           else if (Array.isArray(newRetrievalResponse.response)) {
               assistantContent = JSON.stringify(newRetrievalResponse.response);
           }
           // Add assistant response to conversation history
           messages.push({
               role: "assistant",
               content: assistantContent
           });
           console.log(assistantContent);
           // Log activities and results like the first retrieval
           console.log("\nActivities:");
           if (newRetrievalResponse.activity && Array.isArray(newRetrievalResponse.activity)) {
               newRetrievalResponse.activity.forEach((activity) => {
                   const activityType = activity.activityType || activity.type || 'UnknownActivityRecord';
                   console.log(`Activity Type: ${activityType}`);
                   console.log(JSON.stringify(activity, null, 2));
               });
           }
           console.log("Results");
           if (newRetrievalResponse.references && Array.isArray(newRetrievalResponse.references)) {
               newRetrievalResponse.references.forEach((reference) => {
                   const referenceType = reference.referenceType || reference.type || 'AzureSearchDoc';
                   console.log(`Reference Type: ${referenceType}`);
                   console.log(JSON.stringify(reference, null, 2));
               });
           }
           // Generate final answer for follow-up
           await generateFinalAnswer(openAIClient, messages);
           console.log("\n🎉 === Conversation Complete ===");
       }
       catch (error) {
           console.error("❌ Error in conversation continuation:", error);
           throw error;
       }
   }
   async function getAccessToken(credential, scope) {
       const tokenResponse = await credential.getToken(scope);
       return tokenResponse.token;
   }
   // Error handling wrapper
   async function runWithErrorHandling() {
       try {
           await main();
       }
       catch (error) {
           console.error("💥 Application failed:", error);
           process.exit(1);
       }
   }
   // Execute the application - ES module style
   runWithErrorHandling();
   export { main, createSearchIndex, deleteSearchIndex, fetchEarthAtNightDocuments, uploadDocuments, createKnowledgeAgent, deleteKnowledgeAgent, runAgenticRetrieval };
   

3. 다음 명령을 사용하여 Azure에 로그인합니다.

   az login
   

4. 다음 명령으로 JavaScript 코드를 실행합니다.

   node index.js
   

출력

애플리케이션의 출력은 다음과 유사하게 표시됩니다.

[dotenv@17.2.0] injecting env (0) from .env (tip: ⚙️  override existing env vars with { override: true })
🚀 Starting Azure AI Search agentic retrieval quickstart...

📊 Creating search index...
✅ Index 'earth_at_night' created or updated successfully.
📄 Uploading documents...
📡 Fetching Earth at Night documents from GitHub...
✅ Fetched 194 documents from GitHub
✅ Uploaded 194 documents successfully.
⏳ Waiting for document indexing to complete...
✅ Document indexing completed.
🗑️ Deleting knowledge agent...
ℹ️ Knowledge agent 'earth-search-agent' does not exist or was already deleted.
🤖 Creating knowledge agent...
✅ Knowledge agent 'earth-search-agent' created successfully.
🔍 Running agentic retrieval...
[{"role":"assistant","content":[{"type":"text","text":"[]"}]}]

Activities:
Activity Type: ModelQueryPlanning
{
  "type": "ModelQueryPlanning",
  "id": 0,
  "inputTokens": 1379,
  "outputTokens": 551
}
Activity Type: AzureSearchQuery
{
  "type": "AzureSearchQuery",
  "id": 1,
  "targetIndex": "earth_at_night",
  "query": {
    "search": "Why do suburban areas show greater December brightening compared to urban cores despite higher absolute light levels downtown?",
    "filter": null
  },
  "queryTime": "2025-07-20T16:12:59.804Z",
  "count": 0,
  "elapsedMs": 549
}
Activity Type: AzureSearchQuery
{
  "type": "AzureSearchQuery",
  "id": 2,
  "targetIndex": "earth_at_night",
  "query": {
    "search": "Why is the Phoenix nighttime street grid sharply visible from space, while large stretches of interstate highways between Midwestern cities appear comparatively dim?",
    "filter": null
  },
  "queryTime": "2025-07-20T16:13:00.061Z",
  "count": 0,
  "elapsedMs": 256
}
Activity Type: AzureSearchSemanticRanker
{
  "type": "AzureSearchSemanticRanker",
  "id": 3,
  "inputTokens": 47630
}
Results

[ASSISTANT]: 
Suburban belts show larger December brightening than urban cores despite higher absolute light levels downtown because suburban areas often have more seasonal variation in lighting usage, such as increased decorative and outdoor lighting during the holiday season in December
 Urban cores typically have more constant and dense lighting throughout the year, so the relative increase in brightness during December is less pronounced compared to suburban areas
\n\nThe Phoenix nighttime street grid is sharply visible from space because the city has a well-planned, extensive grid of streets with consistent and bright street lighting
 In contrast, large stretches of interstate highways between Midwestern cities appear comparatively dim because these highways have less continuous lighting and lower intensity lights, making them less visible from space
\n\n(Note: These explanations are based on general knowledge about urban lighting patterns and visibility from space; specific studies or sources were not provided
)

💬 === Continuing Conversation ===
❓ Follow-up question: How do I find lava at night?
[{"role":"assistant","content":[{"type":"text","text":"[{\"ref_id\":0,\"content\":\"<!-- PageHeader=\\\"Volcanoes\\\" -->\\n\\n### Nighttime Glow at Mount Etna - Italy\\n\\nAt about 2:30 a.m. local time on March 16, 2017, the VIIRS DNB on the Suomi NPP satellite captured this nighttime image of lava flowing on Mount Etna in Sicily, Italy. Etna is one of the world's most active volcanoes.\\n\\n#### Figure: Location of Mount Etna\\nA world globe is depicted, with a marker indicating the ___location of Mount Etna in Sicily, Italy, in southern Europe near the center of the Mediterranean Sea.\\n\\n<!-- PageFooter=\\\"Earth at Night\\\" -->\\n<!-- PageNumber=\\\"48\\\" -->\"},{\"ref_id\":1,\"content\":\"<!-- PageHeader=\\\"Volcanoes\\\" -->\\n\\n## Volcanoes\\n\\n### The Infrared Glows of Kilauea's Lava Flows—Hawaii\\n\\nIn early May 2018, an eruption on Hawaii's Kilauea volcano began to unfold. The eruption took a dangerous turn on May 3, 2018, when new fissures opened in the residential neighborhood of Leilani Estates. During the summer-long eruptive event, other fissures emerged along the East Rift Zone. Lava from vents along the rift zone flowed downslope, reaching the ocean in several areas, and filling in Kapoho Bay.\\n\\nA time series of Landsat 8 imagery shows the progression of the lava flows from May 16 to August 13. The night view combines thermal, shortwave infrared, and near-infrared wavelengths to tease out the very hot lava (bright white), cooling lava (red), and lava flows obstructed by clouds (purple).\\n\\n#### Figure: Location of Kilauea Volcano, Hawaii\\n\\nA globe is shown centered on North America, with a marker placed in the Pacific Ocean indicating the ___location of Hawaii, to the southwest of the mainland United States.\\n\\n<!-- PageFooter=\\\"Earth at Night\\\" -->\\n<!-- PageNumber=\\\"44\\\" -->\"},{\"ref_id\":2,\"content\":\"For the first time in perhaps a decade, Mount Etna experienced a \\\"flank eruption\\\"—erupting from its side instead of its summit—on December 24, 2018. The activity was accompanied by 130 earthquakes occurring over three hours that morning. Mount Etna, Europe’s most active volcano, has seen periodic activity on this part of the mountain since 2013. The Operational Land Imager (OLI) on the Landsat 8 satellite acquired the main image of Mount Etna on December 28, 2018.\\n\\nThe inset image highlights the active vent and thermal infrared signature from lava flows, which can be seen near the newly formed fissure on the southeastern side of the volcano. The inset was created with data from OLI and the Thermal Infrared Sensor (TIRS) on Landsat 8. Ash spewing from the fissure cloaked adjacent villages and delayed aircraft from landing at the nearby Catania airport. Earthquakes occurred in the subsequent days after the initial eruption and displaced hundreds of people from their homes.\\n\\nFor nighttime images of Mount Etna’s March 2017 eruption, see pages 48–51.\\n\\n---\\n\\n### Hazards of Volcanic Ash Plumes and Satellite Observation\\n\\nWith the help of moonlight, satellite instruments can track volcanic ash plumes, which present significant hazards to airplanes in flight. The volcanic ash—composed of tiny pieces of glass and rock—is abrasive to engine turbine blades, and can melt on the blades and other engine parts, causing damage and even engine stalls. This poses a danger to both the plane’s integrity and passenger safety. Volcanic ash also reduces visibility for pilots and can cause etching of windshields, further reducing pilots’ ability to see. Nightlight images can be combined with thermal images to provide a more complete view of volcanic activity on Earth’s surface.\\n\\nThe VIIRS Day/Night Band (DNB) on polar-orbiting satellites uses faint light sources such as moonlight, airglow (the atmosphere’s self-illumination through chemical reactions), zodiacal light (sunlight scattered by interplanetary dust), and starlight from the Milky Way. Using these dim light sources, the DNB can detect changes in clouds, snow cover, and sea ice:\\n\\n#### Table: Light Sources Used by VIIRS DNB\\n\\n| Light Source         | Description                                                                  |\\n|----------------------|------------------------------------------------------------------------------|\\n| Moonlight            | Reflected sunlight from the Moon, illuminating Earth's surface at night      |\\n| Airglow              | Atmospheric self-illumination from chemical reactions                        |\\n| Zodiacal Light       | Sunlight scattered by interplanetary dust                                    |\\n| Starlight/Milky Way  | Faint illumination provided by stars in the Milky Way                        |\\n\\nGeostationary Operational Environmental Satellites (GOES), managed by NOAA, orbit over Earth’s equator and offer uninterrupted observations of North America. High-latitude areas such as Alaska benefit from polar-orbiting satellites like Suomi NPP, which provide overlapping coverage at the poles, enabling more data collection in these regions. During polar darkness (winter months), VIIRS DNB data allow scientists to:\\n\\n- Observe sea ice formation\\n- Monitor snow cover extent at the highest latitudes\\n- Detect open water for ship navigation\\n\\n#### Table: Satellite Coverage Overview\\n\\n| Satellite Type          | Orbit           | Coverage Area         | Special Utility                              |\\n|------------------------|-----------------|----------------------|----------------------------------------------|\\n| GOES                   | Geostationary   | Equatorial/North America | Continuous regional monitoring              |\\n| Polar-Orbiting (e.g., Suomi NPP) | Polar-orbiting    | Poles/high latitudes      | Overlapping passes; useful during polar night|\\n\\n---\\n\\n### Weather Forecasting and Nightlight Data\\n\\nThe use of nightlight data by weather forecasters is growing as the VIIRS instrument enables observation of clouds at night illuminated by sources such as moonlight and lightning. Scientists use these data to study the nighttime behavior of weather systems, including severe storms, which can develop and strike populous areas at night as well as during the day. Combined with thermal data, visible nightlight data allow the detection of clouds at various heights in the atmosphere, such as dense marine fog. This capability enables weather forecasters to issue marine advisories with higher confidence, leading to greater utility. (See \\\"Marine Layer Clouds—California\\\" on page 56.)\\n\\nIn this section of the book, you will see how nightlight data are used to observe nature’s spectacular light shows across a wide range of sources.\\n\\n---\\n\\n#### Notable Data from Mount Etna Flank Eruption (December 2018)\\n\\n| Event/Observation                  | Details                                                                    |\\n|-------------------------------------|----------------------------------------------------------------------------|\\n| Date of Flank Eruption              | December 24, 2018                                                          |\\n| Number of Earthquakes               | 130 earthquakes within 3 hours                                              |\\n| Image Acquisition                   | December 28, 2018 by Landsat 8 OLI                                         |\\n| Location of Eruption                | Southeastern side of Mount Etna                                            |\\n| Thermal Imaging Data                | From OLI and TIRS (Landsat 8), highlighting active vent and lava flows     |\\n| Impact on Villages/Air Transport    | Ash covered villages; delayed aircraft at Catania airport                  |\\n| Displacement                        | Hundreds of residents displaced                                            |\\n| Ongoing Seismic Activity            | Earthquakes continued after initial eruption                               |\\n\\n---\\n\\n<!-- PageFooter=\\\"Earth at Night\\\" -->\\n<!-- PageNumber=\\\"30\\\" -->\"},{\"ref_id\":3,\"content\":\"# Volcanoes\\n\\n---\\n\\n### Mount Etna Erupts - Italy\\n\\nThe highly active Mount Etna in Italy sent red lava rolling down its flank on March 19, 2017. An astronaut onboard the ISS took the photograph below of the volcano and its environs that night. City lights surround the mostly dark volcanic area.\\n\\n---\\n\\n#### Figure 1: Location of Mount Etna, Italy\\n\\nA world map highlighting the ___location of Mount Etna in southern Italy. The marker indicates its geographic placement on the east coast of Sicily, Italy, in the Mediterranean region, south of mainland Europe and north of northern Africa.\\n\\n---\\n\\n#### Figure 2: Nighttime View of Mount Etna's Eruption and Surrounding Cities\\n\\nThis is a nighttime satellite image taken on March 19, 2017, showing the eruption of Mount Etna (southeastern cone) with visible bright red and orange coloring indicating flowing lava from a lateral vent. The surrounding areas are illuminated by city lights, with the following geographic references labeled:\\n\\n| Location        | Position in Image         | Visible Characteristics                    |\\n|-----------------|--------------------------|--------------------------------------------|\\n| Mt. Etna (southeastern cone) | Top center-left | Bright red/orange lava flow                |\\n| Lateral vent    | Left of the volcano       | Faint red/orange flow extending outwards   |\\n| Resort          | Below the volcano, to the left   | Small cluster of lights                    |\\n| Giarre          | Top right                 | Bright cluster of city lights              |\\n| Acireale        | Center right              | Large, bright area of city lights          |\\n| Biancavilla     | Bottom left               | Smaller cluster of city lights             |\\n\\nAn arrow pointing north is shown on the image for orientation.\\n\\n---\\n\\n<!-- Earth at Night Page Footer -->\\n<!-- Page Number: 50 -->\"},{\"ref_id\":4,\"content\":\"## Nature's Light Shows\\n\\nAt night, with the light of the Sun removed, nature's brilliant glow from Earth's surface becomes visible to the naked eye from space. Some of Earth's most spectacular light shows are natural, like the aurora borealis, or Northern Lights, in the Northern Hemisphere (aurora australis, or Southern Lights, in the Southern Hemisphere). The auroras are natural electrical phenomena caused by charged particles that race from the Sun toward Earth, inducing chemical reactions in the upper atmosphere and creating the appearance of streamers of reddish or greenish light in the sky, usually near the northern or southern magnetic pole. Other natural lights can indicate danger, like a raging forest fire encroaching on a city, town, or community, or lava spewing from an erupting volcano.\\n\\nWhatever the source, the ability of humans to monitor nature's light shows at night has practical applications for society. For example, tracking fires during nighttime hours allows for continuous monitoring and enhances our ability to protect humans and other animals, plants, and infrastructure. Combined with other data sources, our ability to observe the light of fires at night allows emergency managers to more efficiently and accurately issue warnings and evacuation orders and allows firefighting efforts to continue through the night. With enough moonlight (e.g., full-Moon phase), it's even possible to track the movement of smoke plumes at night, which can impact air quality, regardless of time of day.\\n\\nAnother natural source of light at night is emitted from glowing lava flows at the site of active volcanoes. Again, with enough moonlight, these dramatic scenes can be tracked and monitored for both scientific research and public safety.\\n\\n\\n### Figure: The Northern Lights Viewed from Space\\n\\n**September 17, 2011**\\n\\nThis photo, taken from the International Space Station on September 17, 2011, shows a spectacular display of the aurora borealis (Northern Lights) as green and reddish light in the night sky above Earth. In the foreground, part of a Soyuz spacecraft is visible, silhouetted against the bright auroral light. The green glow is generated by energetic charged particles from the Sun interacting with Earth's upper atmosphere, exciting oxygen and nitrogen atoms, and producing characteristic colors. The image demonstrates the vividness and grandeur of natural night-time light phenomena as seen from orbit.\"}]"}]}]

Activities:
Activity Type: ModelQueryPlanning
{
  "type": "ModelQueryPlanning",
  "id": 0,
  "inputTokens": 1598,
  "outputTokens": 159
}
Activity Type: AzureSearchQuery
{
  "type": "AzureSearchQuery",
  "id": 1,
  "targetIndex": "earth_at_night",
  "query": {
    "search": "How can I locate lava flows during nighttime?",
    "filter": null
  },
  "queryTime": "2025-07-20T16:13:10.659Z",
  "count": 5,
  "elapsedMs": 260
}
Activity Type: AzureSearchSemanticRanker
{
  "type": "AzureSearchSemanticRanker",
  "id": 2,
  "inputTokens": 24146
}
Results
Reference Type: AzureSearchDoc
{
  "type": "AzureSearchDoc",
  "id": "0",
  "activitySource": 1,
  "docKey": "earth_at_night_508_page_64_verbalized",
  "sourceData": {
    "id": "earth_at_night_508_page_64_verbalized",
    "page_chunk": "<!-- PageHeader=\"Volcanoes\" -->\n\n### Nighttime Glow at Mount Etna - Italy\n\nAt about 2:30 a.m. local time on March 16, 2017, the VIIRS DNB on the Suomi NPP satellite captured this nighttime image of lava flowing on Mount Etna in Sicily, Italy. Etna is one of the world's most active volcanoes.\n\n#### Figure: Location of Mount Etna\nA world globe is depicted, with a marker indicating the ___location of Mount Etna in Sicily, Italy, in southern Europe near the center of the Mediterranean Sea.\n\n<!-- PageFooter=\"Earth at Night\" -->\n<!-- PageNumber=\"48\" -->"
  }
}
Reference Type: AzureSearchDoc
{
  "type": "AzureSearchDoc",
  "id": "1",
  "activitySource": 1,
  "docKey": "earth_at_night_508_page_60_verbalized",
  "sourceData": {
    "id": "earth_at_night_508_page_60_verbalized",
    "page_chunk": "<!-- PageHeader=\"Volcanoes\" -->\n\n## Volcanoes\n\n### The Infrared Glows of Kilauea's Lava Flows—Hawaii\n\nIn early May 2018, an eruption on Hawaii's Kilauea volcano began to unfold. The eruption took a dangerous turn on May 3, 2018, when new fissures opened in the residential neighborhood of Leilani Estates. During the summer-long eruptive event, other fissures emerged along the East Rift Zone. Lava from vents along the rift zone flowed downslope, reaching the ocean in several areas, and filling in Kapoho Bay.\n\nA time series of Landsat 8 imagery shows the progression of the lava flows from May 16 to August 13. The night view combines thermal, shortwave infrared, and near-infrared wavelengths to tease out the very hot lava (bright white), cooling lava (red), and lava flows obstructed by clouds (purple).\n\n#### Figure: Location of Kilauea Volcano, Hawaii\n\nA globe is shown centered on North America, with a marker placed in the Pacific Ocean indicating the ___location of Hawaii, to the southwest of the mainland United States.\n\n<!-- PageFooter=\"Earth at Night\" -->\n<!-- PageNumber=\"44\" -->"
  }
}
Reference Type: AzureSearchDoc
{
  "type": "AzureSearchDoc",
  "id": "2",
  "activitySource": 1,
  "docKey": "earth_at_night_508_page_46_verbalized",
  "sourceData": {
    "id": "earth_at_night_508_page_46_verbalized",
    "page_chunk": "For the first time in perhaps a decade, Mount Etna experienced a \"flank eruption\"—erupting from its side instead of its summit—on December 24, 2018. The activity was accompanied by 130 earthquakes occurring over three hours that morning. Mount Etna, Europe’s most active volcano, has seen periodic activity on this part of the mountain since 2013. The Operational Land Imager (OLI) on the Landsat 8 satellite acquired the main image of Mount Etna on December 28, 2018.\n\nThe inset image highlights the active vent and thermal infrared signature from lava flows, which can be seen near the newly formed fissure on the southeastern side of the volcano. The inset was created with data from OLI and the Thermal Infrared Sensor (TIRS) on Landsat 8. Ash spewing from the fissure cloaked adjacent villages and delayed aircraft from landing at the nearby Catania airport. Earthquakes occurred in the subsequent days after the initial eruption and displaced hundreds of people from their homes.\n\nFor nighttime images of Mount Etna’s March 2017 eruption, see pages 48–51.\n\n---\n\n### Hazards of Volcanic Ash Plumes and Satellite Observation\n\nWith the help of moonlight, satellite instruments can track volcanic ash plumes, which present significant hazards to airplanes in flight. The volcanic ash—composed of tiny pieces of glass and rock—is abrasive to engine turbine blades, and can melt on the blades and other engine parts, causing damage and even engine stalls. This poses a danger to both the plane’s integrity and passenger safety. Volcanic ash also reduces visibility for pilots and can cause etching of windshields, further reducing pilots’ ability to see. Nightlight images can be combined with thermal images to provide a more complete view of volcanic activity on Earth’s surface.\n\nThe VIIRS Day/Night Band (DNB) on polar-orbiting satellites uses faint light sources such as moonlight, airglow (the atmosphere’s self-illumination through chemical reactions), zodiacal light (sunlight scattered by interplanetary dust), and starlight from the Milky Way. Using these dim light sources, the DNB can detect changes in clouds, snow cover, and sea ice:\n\n#### Table: Light Sources Used by VIIRS DNB\n\n| Light Source         | Description                                                                  |\n|----------------------|------------------------------------------------------------------------------|\n| Moonlight            | Reflected sunlight from the Moon, illuminating Earth's surface at night      |\n| Airglow              | Atmospheric self-illumination from chemical reactions                        |\n| Zodiacal Light       | Sunlight scattered by interplanetary dust                                    |\n| Starlight/Milky Way  | Faint illumination provided by stars in the Milky Way                        |\n\nGeostationary Operational Environmental Satellites (GOES), managed by NOAA, orbit over Earth’s equator and offer uninterrupted observations of North America. High-latitude areas such as Alaska benefit from polar-orbiting satellites like Suomi NPP, which provide overlapping coverage at the poles, enabling more data collection in these regions. During polar darkness (winter months), VIIRS DNB data allow scientists to:\n\n- Observe sea ice formation\n- Monitor snow cover extent at the highest latitudes\n- Detect open water for ship navigation\n\n#### Table: Satellite Coverage Overview\n\n| Satellite Type          | Orbit           | Coverage Area         | Special Utility                              |\n|------------------------|-----------------|----------------------|----------------------------------------------|\n| GOES                   | Geostationary   | Equatorial/North America | Continuous regional monitoring              |\n| Polar-Orbiting (e.g., Suomi NPP) | Polar-orbiting    | Poles/high latitudes      | Overlapping passes; useful during polar night|\n\n---\n\n### Weather Forecasting and Nightlight Data\n\nThe use of nightlight data by weather forecasters is growing as the VIIRS instrument enables observation of clouds at night illuminated by sources such as moonlight and lightning. Scientists use these data to study the nighttime behavior of weather systems, including severe storms, which can develop and strike populous areas at night as well as during the day. Combined with thermal data, visible nightlight data allow the detection of clouds at various heights in the atmosphere, such as dense marine fog. This capability enables weather forecasters to issue marine advisories with higher confidence, leading to greater utility. (See \"Marine Layer Clouds—California\" on page 56.)\n\nIn this section of the book, you will see how nightlight data are used to observe nature’s spectacular light shows across a wide range of sources.\n\n---\n\n#### Notable Data from Mount Etna Flank Eruption (December 2018)\n\n| Event/Observation                  | Details                                                                    |\n|-------------------------------------|----------------------------------------------------------------------------|\n| Date of Flank Eruption              | December 24, 2018                                                          |\n| Number of Earthquakes               | 130 earthquakes within 3 hours                                              |\n| Image Acquisition                   | December 28, 2018 by Landsat 8 OLI                                         |\n| Location of Eruption                | Southeastern side of Mount Etna                                            |\n| Thermal Imaging Data                | From OLI and TIRS (Landsat 8), highlighting active vent and lava flows     |\n| Impact on Villages/Air Transport    | Ash covered villages; delayed aircraft at Catania airport                  |\n| Displacement                        | Hundreds of residents displaced                                            |\n| Ongoing Seismic Activity            | Earthquakes continued after initial eruption                               |\n\n---\n\n<!-- PageFooter=\"Earth at Night\" -->\n<!-- PageNumber=\"30\" -->"
  }
}
Reference Type: AzureSearchDoc
{
  "type": "AzureSearchDoc",
  "id": "3",
  "activitySource": 1,
  "docKey": "earth_at_night_508_page_66_verbalized",
  "sourceData": {
    "id": "earth_at_night_508_page_66_verbalized",
    "page_chunk": "# Volcanoes\n\n---\n\n### Mount Etna Erupts - Italy\n\nThe highly active Mount Etna in Italy sent red lava rolling down its flank on March 19, 2017. An astronaut onboard the ISS took the photograph below of the volcano and its environs that night. City lights surround the mostly dark volcanic area.\n\n---\n\n#### Figure 1: Location of Mount Etna, Italy\n\nA world map highlighting the ___location of Mount Etna in southern Italy. The marker indicates its geographic placement on the east coast of Sicily, Italy, in the Mediterranean region, south of mainland Europe and north of northern Africa.\n\n---\n\n#### Figure 2: Nighttime View of Mount Etna's Eruption and Surrounding Cities\n\nThis is a nighttime satellite image taken on March 19, 2017, showing the eruption of Mount Etna (southeastern cone) with visible bright red and orange coloring indicating flowing lava from a lateral vent. The surrounding areas are illuminated by city lights, with the following geographic references labeled:\n\n| Location        | Position in Image         | Visible Characteristics                    |\n|-----------------|--------------------------|--------------------------------------------|\n| Mt. Etna (southeastern cone) | Top center-left | Bright red/orange lava flow                |\n| Lateral vent    | Left of the volcano       | Faint red/orange flow extending outwards   |\n| Resort          | Below the volcano, to the left   | Small cluster of lights                    |\n| Giarre          | Top right                 | Bright cluster of city lights              |\n| Acireale        | Center right              | Large, bright area of city lights          |\n| Biancavilla     | Bottom left               | Smaller cluster of city lights             |\n\nAn arrow pointing north is shown on the image for orientation.\n\n---\n\n<!-- Earth at Night Page Footer -->\n<!-- Page Number: 50 -->"
  }
}
Reference Type: AzureSearchDoc
{
  "type": "AzureSearchDoc",
  "id": "4",
  "activitySource": 1,
  "docKey": "earth_at_night_508_page_44_verbalized",
  "sourceData": {
    "id": "earth_at_night_508_page_44_verbalized",
    "page_chunk": "## Nature's Light Shows\n\nAt night, with the light of the Sun removed, nature's brilliant glow from Earth's surface becomes visible to the naked eye from space. Some of Earth's most spectacular light shows are natural, like the aurora borealis, or Northern Lights, in the Northern Hemisphere (aurora australis, or Southern Lights, in the Southern Hemisphere). The auroras are natural electrical phenomena caused by charged particles that race from the Sun toward Earth, inducing chemical reactions in the upper atmosphere and creating the appearance of streamers of reddish or greenish light in the sky, usually near the northern or southern magnetic pole. Other natural lights can indicate danger, like a raging forest fire encroaching on a city, town, or community, or lava spewing from an erupting volcano.\n\nWhatever the source, the ability of humans to monitor nature's light shows at night has practical applications for society. For example, tracking fires during nighttime hours allows for continuous monitoring and enhances our ability to protect humans and other animals, plants, and infrastructure. Combined with other data sources, our ability to observe the light of fires at night allows emergency managers to more efficiently and accurately issue warnings and evacuation orders and allows firefighting efforts to continue through the night. With enough moonlight (e.g., full-Moon phase), it's even possible to track the movement of smoke plumes at night, which can impact air quality, regardless of time of day.\n\nAnother natural source of light at night is emitted from glowing lava flows at the site of active volcanoes. Again, with enough moonlight, these dramatic scenes can be tracked and monitored for both scientific research and public safety.\n\n\n### Figure: The Northern Lights Viewed from Space\n\n**September 17, 2011**\n\nThis photo, taken from the International Space Station on September 17, 2011, shows a spectacular display of the aurora borealis (Northern Lights) as green and reddish light in the night sky above Earth. In the foreground, part of a Soyuz spacecraft is visible, silhouetted against the bright auroral light. The green glow is generated by energetic charged particles from the Sun interacting with Earth's upper atmosphere, exciting oxygen and nitrogen atoms, and producing characteristic colors. The image demonstrates the vividness and grandeur of natural night-time light phenomena as seen from orbit."
  }
}

[ASSISTANT]: 
To find lava at night, satellite instruments like the VIIRS Day/Night Band (DNB) and thermal infrared sensors are used to detect the glow of very hot lava flows on the Earth's surface
 For example, nighttime satellite images have captured lava flowing from active volcanoes such as Mount Etna in Italy and Kilauea in Hawaii, where the hot lava emits bright light visible from space even at night
 Scientists combine thermal, shortwave infrared, and near-infrared data to distinguish very hot lava (bright white), cooling lava (red), and areas obscured by clouds
 Additionally, moonlight and other faint natural light sources help illuminate the surroundings to improve observation of volcanic activity at night
 Monitoring lava flow at night is important for scientific research and public safety, as it helps track volcanic eruptions and associated hazards such as ash plumes that can affect air travel and nearby communities [refs 0,1,2,3,4]


🎉 === Conversation Complete ===
🗑️ Deleting knowledge agent...
✅ Knowledge agent 'earth-search-agent' deleted successfully.
🗑️ Deleting search index...
✅ Search index 'earth_at_night' deleted successfully.
✅ Quickstart completed successfully!

코드 이해

이제 코드가 있으므로 주요 구성 요소를 세분화해 보겠습니다.

• 검색 인덱스 만들기
• 인덱스로 문서 업로드
• 지식 에이전트 만들기
• 메시지 설정
• 검색 파이프라인 실행
• 응답, 활동 및 결과 검토
• Azure OpenAI 클라이언트 만들기
• 채팅 완료 API를 사용하여 답변 생성
• 대화 계속

검색 인덱스 만들기

Azure AI Search에서 인덱스는 구조화된 데이터 컬렉션입니다. 다음 코드는 일반 텍스트 및 벡터 콘텐츠를 포함하도록 명명된 earth_at_night 인덱스를 정의합니다. 기존 인덱스를 사용할 수 있지만 에이전트 검색 워크로드에 대한 조건을 충족해야 합니다.

const index = {
    name: config.indexName,
    fields: [
        {
            name: "id",
            type: "Edm.String",
            key: true,
            filterable: true,
            sortable: true,
            facetable: true
        },
        {
            name: "page_chunk",
            type: "Edm.String",
            searchable: true,
            filterable: false,
            sortable: false,
            facetable: false
        },
        {
            name: "page_embedding_text_3_large",
            type: "Collection(Edm.Single)",
            searchable: true,
            filterable: false,
            sortable: false,
            facetable: false,
            vectorSearchDimensions: 3072,
            vectorSearchProfileName: "hnsw_text_3_large"
        },
        {
            name: "page_number",
            type: "Edm.Int32",
            filterable: true,
            sortable: true,
            facetable: true
        }
    ],
    vectorSearch: {
        profiles: [
            {
                name: "hnsw_text_3_large",
                algorithmConfigurationName: "alg",
                vectorizerName: "azure_openai_text_3_large"
            }
        ],
        algorithms: [
            {
                name: "alg",
                kind: "hnsw"
            }
        ],
        vectorizers: [
            {
                vectorizerName: "azure_openai_text_3_large",
                kind: "azureOpenAI",
                parameters: {
                    resourceUrl: config.azureOpenAIEndpoint,
                    deploymentId: config.azureOpenAIEmbeddingDeployment,
                    modelName: config.azureOpenAIEmbeddingModel
                }
            }
        ]
    },
    semanticSearch: {
        defaultConfigurationName: "semantic_config",
        configurations: [
            {
                name: "semantic_config",
                prioritizedFields: {
                    contentFields: [
                        { name: "page_chunk" }
                    ]
                }
            }
        ]
    }
};

인덱스 스키마에는 문서 식별 및 페이지 콘텐츠, 포함 및 숫자에 대한 필드가 포함됩니다. 또한 이전에 배포한 text-embedding-3-large 모델을 사용하는 의미 순위 및 벡터 쿼리에 대한 구성도 포함됩니다.

인덱스로 문서 업로드

현재 인덱스는 earth_at_night 비어 있습니다. 다음 코드를 실행하여 인덱스를 NASA의 어스 앳 나이트 전자책의 JSON 문서로 채웁니다. Azure AI Search에서 요구하는 대로 각 문서는 인덱스 스키마에 정의된 필드 및 데이터 형식을 준수합니다.

console.log("📡 Fetching Earth at Night documents from GitHub...");
const documentsUrl = "https://raw.githubusercontent.com/Azure-Samples/azure-search-sample-data/refs/heads/main/nasa-e-book/earth-at-night-json/documents.json";
try {
    const response = await fetch(documentsUrl);
    if (!response.ok) {
        throw new Error(`Failed to fetch documents: ${response.status} ${response.statusText}`);
    }
    const documents = await response.json();
    console.log(`✅ Fetched ${documents.length} documents from GitHub`);
    // Validate and transform documents to match our interface
    const transformedDocuments = documents.map((doc, index) => {
        return {
            id: doc.id || String(index + 1),
            page_chunk: doc.page_chunk || doc.content || '',
            page_embedding_text_3_large: doc.page_embedding_text_3_large || new Array(3072).fill(0.1),
            page_number: doc.page_number || index + 1
        };
    });
    return transformedDocuments;
}

지식 에이전트 만들기

Azure AI Search를 gpt-5-mini 배포에 연결하고 쿼리 시 earth_at_night 인덱스를 대상으로 지정하려면, 지식 에이전트가 필요합니다. 다음 코드는 에이전트 정의를 사용하여 쿼리를 처리하고 earth-search-agent 인덱스에서 관련 문서를 검색하는 earth_at_night라는 이름의 지식 에이전트를 정의합니다.

관련성이 있고 의미 있는 응답을 defaultRerankerThreshold 보장하기 위해 재랜커 점수 2.5 가 낮거나 낮은 응답을 제외하도록 설정됩니다.

const agentDefinition = {
    name: config.agentName,
    description: "Knowledge agent for Earth at Night e-book content",
    models: [
        {
            kind: "azureOpenAI",
            azureOpenAIParameters: {
                resourceUri: config.azureOpenAIEndpoint,
                deploymentId: config.azureOpenAIGptDeployment,
                modelName: config.azureOpenAIGptModel
            }
        }
    ],
    targetIndexes: [
        {
            indexName: config.indexName,
            defaultRerankerThreshold: 2.5
        }
    ]
};

메시지 설정

메시지는 검색 경로에 대한 입력이며 대화 기록을 포함합니다. 각 메시지에는 도우미 또는 사용자와 같은 원본을 나타내는 역할과 자연어로 된 콘텐츠가 포함됩니다. 사용하는 LLM은 유효한 역할을 결정합니다.

사용자 메시지는 처리할 쿼리를 나타내고 보조 메시지는 기술 에이전트에 응답하는 방법을 안내합니다. 검색 프로세스 중에 이러한 메시지는 인덱싱된 문서에서 관련 응답을 추출하기 위해 LLM으로 전송됩니다.

이 보조 메시지는 밤에 지구에 대한 질문에 대답하고, 소스를 earth-search-agent사용하여 인용하고, 답변을 사용할 수 없을 때 "나는 모른다"로 응답하도록 지시 ref_id 합니다.

const messages = [
    {
        role: "system",
        content: `A Q&A agent that can answer questions about the Earth at night.
Sources have a JSON format with a ref_id that must be cited in the answer.
If you do not have the answer, respond with "I don't know".`
    },
    {
        role: "user",
        content: "Why do suburban belts display larger December brightening than urban cores even though absolute light levels are higher downtown? Why is the Phoenix nighttime street grid is so sharply visible from space, whereas large stretches of the interstate between midwestern cities remain comparatively dim?"
    }
];

검색 파이프라인 실행

이 단계에서는 검색 파이프라인을 실행하여 검색 인덱스에서 관련 정보를 추출합니다. 검색 요청의 메시지 및 매개 변수에 따라 LLM은 다음을 수행합니다.

1. 전체 대화 기록을 분석하여 기본 정보가 필요한지 확인합니다.
2. 복합 사용자 쿼리를 포커스가 있는 하위 쿼리로 구분합니다.
3. 인덱스의 텍스트 필드 및 벡터 포함에 대해 각 하위 쿼리를 동시에 실행합니다.
4. 의미 체계 순위자를 사용하여 모든 하위 쿼리의 결과를 다시 표시합니다.
5. 결과를 단일 문자열로 병합합니다.

다음 코드는 쿼리를 하위 쿼리로 분해하고, 인덱스의 텍스트 필드와 벡터 포함에 대해 하위 쿼리를 실행하고, 결과의 순위를 지정하고 병합하는 earth-search-agent 두 부분으로 구성된 사용자 쿼리earth_at_night를 보냅니다. 그런 다음 응답이 목록에 추가됩니다 messages .

const agentMessages = messages.map(msg => ({
    role: msg.role,
    content: [
        {
            type: "text",
            text: msg.content
        }
    ]
}));
const retrievalRequest = {
    messages: agentMessages,
    targetIndexParams: [
        {
            indexName: config.indexName,
            rerankerThreshold: 2.5,
            maxDocsForReranker: 100,
            includeReferenceSourceData: true
        }
    ]
};
const token = await getAccessToken(credential, "https://search.azure.com/.default");
const response = await fetch(`${config.searchEndpoint}/agents/${config.agentName}/retrieve?api-version=${config.searchApiVersion}`, {
    method: 'POST',
    headers: {
        'Content-Type': 'application/json',
        'Authorization': `Bearer ${token}`
    },
    body: JSON.stringify(retrievalRequest)
});
if (!response.ok) {
    const errorText = await response.text();
    throw new Error(`Agentic retrieval failed: ${response.status} ${response.statusText}\n${errorText}`);
}
return await response.json();

응답, 활동 및 결과 검토

이제 검색 파이프라인의 응답, 활동 및 결과를 표시하려고 합니다.

Azure AI Search의 각 검색 응답에는 다음이 포함됩니다.

• 검색 결과의 접지 데이터를 나타내는 통합 문자열입니다.

• 쿼리 계획.

• 통합 문자열에 기여한 원본 문서의 청크를 보여 주는 참조 데이터입니다.

console.log("\nActivities:");
if (retrievalResponse.activity && Array.isArray(retrievalResponse.activity)) {
    retrievalResponse.activity.forEach((activity) => {
        const activityType = activity.activityType || activity.type || 'UnknownActivityRecord';
        console.log(`Activity Type: ${activityType}`);
        console.log(JSON.stringify(activity, null, 2));
    });
}
console.log("Results");
if (retrievalResponse.references && Array.isArray(retrievalResponse.references)) {
    retrievalResponse.references.forEach((reference) => {
        const referenceType = reference.referenceType || reference.type || 'AzureSearchDoc';
        console.log(`Reference Type: ${referenceType}`);
        console.log(JSON.stringify(reference, null, 2));
    });
}

출력에는 다음이 포함되어야 합니다.

• Response 는 사용자 쿼리를 기반으로 검색 인덱스에 가장 관련성이 큰 문서(또는 청크)의 텍스트 문자열을 제공합니다. 이 빠른 시작의 뒷부분에서와 같이 응답 생성을 위해 이 문자열을 LLM에 전달할 수 있습니다.

• Activity 는 배포에서 생성된 하위 쿼리 및 쿼리 계획 및 실행에 사용되는 토큰을 포함하여 검색 프로세스 중에 수행된 gpt-5-mini 단계를 추적합니다.

• Results 에는 응답에 기여한 문서가 나열되고 각 문서는 해당 DocKey문서로 식별됩니다.

Azure OpenAI 클라이언트 만들기

응답 추출에서 응답 생성으로 검색 처리 단계를 확장하려면 배포와 상호 작용할 수 있도록 Azure OpenAI 클라이언트를 설정합니다.

const scope = "https://cognitiveservices.azure.com/.default";
const azureADTokenProvider = getBearerTokenProvider(credential, scope);
const openAIClient = new AzureOpenAI({
    endpoint: config.azureOpenAIEndpoint,
    apiVersion: config.azureOpenAIApiVersion,
    azureADTokenProvider,
});

채팅 완료 API를 사용하여 답변 생성

응답 생성을 위한 한 가지 옵션은 대화를 처리하기 위해 LLM에 대화 기록을 전달하는 채팅 완료 API입니다.

const completion = await openAIClient.chat.completions.create({
    model: config.azureOpenAIGptDeployment,
    messages: messages.map(m => ({ role: m.role, content: m.content })),
    max_tokens: 1000,
    temperature: 0.7
});
const answer = completion.choices[0].message.content;
console.log(answer?.replace(/\./g, "\n"));

대화 계속

다른 사용자에게 쿼리를 보내 대화를 계속하세요 earth-search-agent. 다음 코드는 검색 파이프라인을 다시 실행하여 인덱스에서 earth_at_night 관련 콘텐츠를 가져오고 목록에 응답을 messages 추가합니다. 그러나 이전과 달리 이제 Azure OpenAI 클라이언트를 사용하여 검색된 콘텐츠에 따라 답변을 생성할 수 있습니다.

const followUpQuestion = "How do I find lava at night?";
console.log(`❓ Follow-up question: ${followUpQuestion}`);
messages.push({
    role: "user",
    content: followUpQuestion
});

자원을 정리하세요

자체 구독에서 작업할 때 만든 리소스가 여전히 필요한지 여부를 결정하여 프로젝트를 완료하는 것이 좋습니다. 실행 상태로 방치된 리소스는 비용이 들 수 있습니다. 리소스를 개별적으로 삭제하거나 리소스 그룹을 삭제하여 전체 리소스 집합을 삭제할 수 있습니다.

Azure Portal의 왼쪽 창에서 모든 리소스 또는 리소스 그룹을 선택하여 리소스 를 찾고 관리할 수 있습니다. 다음 코드를 실행하여 이 빠른 시작에서 만든 개체를 삭제할 수도 있습니다.

지식 에이전트 삭제

이 빠른 시작에서 만든 지식 에이전트는 다음 코드 조각을 사용하여 삭제되었습니다.

const token = await getAccessToken(credential, "https://search.azure.com/.default");
const response = await fetch(`${config.searchEndpoint}/agents/${config.agentName}?api-version=${config.searchApiVersion}`, {
    method: 'DELETE',
    headers: {
        'Authorization': `Bearer ${token}`
    }
});

검색 인덱스 삭제

이 빠른 시작에서 만든 검색 인덱스가 다음 코드 조각을 사용하여 삭제되었습니다.

await indexClient.deleteIndex(config.indexName);
console.log(`✅ Search index '${config.indexName}' deleted successfully.`);

이 빠른 시작에서는 에이전트 검색을 사용하여 Azure AI 검색에서 인덱싱된 문서와 Azure AI Foundry 모델의 Azure OpenAI에서 LLM(대규모 언어 모델)으로 구동되는 대화형 검색 환경을 만듭니다.

지식 에이전트는 복잡한 쿼리를 하위 쿼리로 분해하고, 하나 이상의 지식 원본에 대해 하위 쿼리를 실행하고, 메타데이터를 사용하여 결과를 반환하여 에이전트 검색을 오케스트레이션합니다. 기본적으로 에이전트는 원본에서 원시 콘텐츠를 출력하지만 이 빠른 시작에서는 자연어 응답 생성을 위해 응답 합성 형식을 사용합니다.

사용자 고유의 데이터를 제공할 수 있지만, 이 빠른 시작에서는 NASA의 어스 앳 나이트 전자책의 샘플 JSON 문서를 사용합니다. 문서는 우주에서 관찰 된 바와 같이 밤에 지구의 일반적인 과학 주제와 이미지를 설명합니다.

필수 조건

• 활성 구독이 있는 Azure 계정. 무료로 계정을 만듭니다.

• 의미 체계 순위가 설정된 기본 계층 이상의 Azure AI Search 서비스입니다.

• Azure AI Foundry 프로젝트 및 Azure AI Foundry 리소스. 프로젝트를 만들면 리소스가 자동으로 만들어집니다.

• Microsoft Entra ID를 사용한 키 없는 인증을 위한 Azure CLI 입니다.

• Python 확장 및 Jupyter 패키지가 있는 Visual Studio Code.

액세스 구성

시작하기 전에 콘텐츠 및 작업에 액세스할 수 있는 권한이 있는지 확인합니다. 인증에는 Microsoft Entra ID를, 권한 부여에는 역할 기반 액세스에 사용하는 것이 좋습니다. 역할을 할당하려면 소유자 또는 사용자 액세스 관리자여야 합니다. 역할이 가능하지 않은 경우 대신 키 기반 인증을 사용합니다.

이 빠른 시작에 대한 액세스를 구성하려면 다음 탭을 모두 선택합니다.

엔드포인트 가져오기

각 Azure AI 검색 서비스와 Azure AI Foundry 리소스에는 리소스에 대한 네트워크 액세스를 식별하고 제공하는 고유한 URL인 엔드포인트가 있습니다. 이후 섹션에서는 이러한 엔드포인트를 지정하여 프로그래밍 방식으로 리소스에 연결합니다.

이 빠른 시작의 엔드포인트를 얻으려면 다음 탭을 모두 선택합니다.

모델 배포

에이전트 검색을 사용하려면 Azure AI Foundry 리소스에 두 개의 Azure OpenAI 모델을 배포해야 합니다.

• 텍스트-벡터 변환을 위한 포함 모델입니다. 이 빠른 시작 가이드에서는 text-embedding-3-large를 사용하지만, 모든 text-embedding 모델을 사용할 수 있습니다.

• 쿼리 계획 및 응답 생성을 위한 LLM입니다. 이 빠른 시작에서는 gpt-5-mini를 사용하지만, 에이전트 검색에 지원되는 모든 LLM을 사용할 수 있습니다.

배포 지침은 Azure AI Foundry를 사용하여 Azure OpenAI 모델 배포를 참조하세요.

로컬 시스템에서 연결

Azure AI Foundry에서 Azure AI 검색 및 Azure OpenAI와 상호 작용하도록 역할 기반 액세스를 구성했습니다. Azure CLI를 사용하여 두 리소스에 대해 동일한 구독 및 테넌트에 로그인합니다. 자세한 내용은 빠른 시작: 키 없이 연결하세요.

로컬 시스템에서 연결하려면 다음을 수행합니다.

1. Visual Studio Code에서 파일을 저장할 폴더를 엽니다.

2. 터미널>새 터미널을 선택합니다.

3. 다음 명령을 실행하여 Azure 계정에 로그인합니다. 구독이 여러 개 있는 경우 Azure AI Search 서비스 및 Azure AI Foundry 프로젝트가 포함된 구독을 선택합니다.

   az login
   

패키지 설치 및 연결 로드

코드를 실행하기 전에 Python 패키지를 설치하고 Azure AI Foundry에서 Azure AI 검색 및 Azure OpenAI에 대한 연결을 위한 엔드포인트, 자격 증명 및 배포 세부 정보를 정의합니다. 이러한 값은 다음 섹션에서 사용됩니다.

패키지를 설치하고 연결을 로드하려면 다음을 수행합니다.

1. Visual Studio Code의 동일한 폴더에서 이름이 quickstart-agentic-retrieval.ipynb인 파일을 만듭니다.

2. 코드 셀을 추가한 후 다음 pip install 명령을 붙여넣습니다.

   ! pip install azure-search-documents==11.7.0b1 --quiet
   ! pip install azure-identity --quiet
   ! pip install openai --quiet
   ! pip install aiohttp --quiet
   ! pip install ipykernel --quiet
   ! pip install requests --quiet
   

3. 셀 실행을 선택하여 패키지를 설치합니다.

4. 다른 코드 셀을 추가한 후 다음 import 문 및 변수를 붙여넣습니다.

   from azure.identity import DefaultAzureCredential, get_bearer_token_provider
   import os
   
   search_endpoint = "PUT-YOUR-SEARCH-SERVICE-URL-HERE"
   credential = DefaultAzureCredential()
   token_provider = get_bearer_token_provider(credential, "https://search.azure.com/.default")
   aoai_endpoint = "PUT-YOUR-AOAI-FOUNDRY-URL-HERE"
   aoai_embedding_model = "text-embedding-3-large"
   aoai_embedding_deployment = "text-embedding-3-large"
   aoai_gpt_model = "gpt-5-mini"
   aoai_gpt_deployment = "gpt-5-mini"
   index_name = "earth-at-night"
   knowledge_source_name = "earth-knowledge-source"
   knowledge_agent_name = "earth-knowledge-agent"
   search_api_version = "2025-08-01-preview"
   

5. Set search_endpoint 및 aoai_endpoint를 엔드포인트 가져오기에서 얻은 값으로 설정합니다.

6. 셀 실행을 선택하여 변수를 로드합니다.

검색 인덱스 만들기

Azure AI Search에서 인덱스는 구조화된 데이터 컬렉션입니다. 다음 코드를 사용하여 코드 셀을 추가하고 실행하여 이전에 earth-at-night 변수를 사용하여 지정한 인덱스인 index_name라는 인덱스를 정의합니다.

인덱스 스키마에는 문서 식별 및 페이지 콘텐츠, 포함 및 숫자에 대한 필드가 포함됩니다. 또한 스키마에는 의미론적 순위 지정 및 벡터 검색에 대한 구성이 포함되어 있으며, text-embedding-3-large 배포를 사용하여 텍스트를 벡터화하고 의미론적 유사성에 따라 문서를 일치시킬 수 있습니다.

from azure.search.documents.indexes.models import SearchIndex, SearchField, VectorSearch, VectorSearchProfile, HnswAlgorithmConfiguration, AzureOpenAIVectorizer, AzureOpenAIVectorizerParameters, SemanticSearch, SemanticConfiguration, SemanticPrioritizedFields, SemanticField
from azure.search.documents.indexes import SearchIndexClient
from openai import AzureOpenAI
from azure.identity import get_bearer_token_provider

azure_openai_token_provider = get_bearer_token_provider(credential, "https://cognitiveservices.azure.com/.default")
index = SearchIndex(
    name=index_name,
    fields=[
        SearchField(name="id", type="Edm.String", key=True, filterable=True, sortable=True, facetable=True),
        SearchField(name="page_chunk", type="Edm.String", filterable=False, sortable=False, facetable=False),
        SearchField(name="page_embedding_text_3_large", type="Collection(Edm.Single)", stored=False, vector_search_dimensions=3072, vector_search_profile_name="hnsw_text_3_large"),
        SearchField(name="page_number", type="Edm.Int32", filterable=True, sortable=True, facetable=True)
    ],
    vector_search=VectorSearch(
        profiles=[VectorSearchProfile(name="hnsw_text_3_large", algorithm_configuration_name="alg", vectorizer_name="azure_openai_text_3_large")],
        algorithms=[HnswAlgorithmConfiguration(name="alg")],
        vectorizers=[
            AzureOpenAIVectorizer(
                vectorizer_name="azure_openai_text_3_large",
                parameters=AzureOpenAIVectorizerParameters(
                    resource_url=aoai_endpoint,
                    deployment_name=aoai_embedding_deployment,
                    model_name=aoai_embedding_model
                )
            )
        ]
    ),
    semantic_search=SemanticSearch(
        default_configuration_name="semantic_config",
        configurations=[
            SemanticConfiguration(
                name="semantic_config",
                prioritized_fields=SemanticPrioritizedFields(
                    content_fields=[
                        SemanticField(field_name="page_chunk")
                    ]
                )
            )
        ]
    )
)

index_client = SearchIndexClient(endpoint=search_endpoint, credential=credential)
index_client.create_or_update_index(index)
print(f"Index '{index_name}' created or updated successfully.")

인덱스로 문서 업로드

현재 인덱스는 earth-at-night 비어 있습니다. 다음 코드를 사용하여 코드 셀을 추가하고 실행하여 NASA의 Earth at Night 전자책의 JSON 문서로 인덱스를 채웁니다. Azure AI Search에서 요구하는 대로 각 문서는 인덱스 스키마에 정의된 필드 및 데이터 형식을 준수합니다.

import requests
from azure.search.documents import SearchIndexingBufferedSender

url = "https://raw.githubusercontent.com/Azure-Samples/azure-search-sample-data/refs/heads/main/nasa-e-book/earth-at-night-json/documents.json"
documents = requests.get(url).json()

with SearchIndexingBufferedSender(endpoint=search_endpoint, index_name=index_name, credential=credential) as client:
    client.upload_documents(documents=documents)

print(f"Documents uploaded to index '{index_name}' successfully.")

지식 원본 만들기

지식 원본은 원본 데이터에 대한 재사용 가능한 참조입니다. 다음 코드를 사용하여 코드 셀을 추가하고 실행하여 earth-knowledge-source 인덱스를 대상으로 하는 earth-at-night라는 지식 원본을 정의합니다.

source_data_select는 검색 및 인용을 위해 액세스할 수 있는 인덱스 필드를 지정합니다. 이 예제에는 응답에 길고 해석할 수 없는 포함이 발생하는 것을 방지하기 위해 사람이 읽을 수 있는 필드만 포함됩니다.

from azure.search.documents.indexes.models import SearchIndexKnowledgeSource, SearchIndexKnowledgeSourceParameters
from azure.search.documents.indexes import SearchIndexClient

ks = SearchIndexKnowledgeSource(
    name=knowledge_source_name,
    description="Knowledge source for Earth at night data",
    search_index_parameters=SearchIndexKnowledgeSourceParameters(
        search_index_name=index_name,
        source_data_select="id,page_chunk,page_number",
    ),
)

index_client = SearchIndexClient(endpoint=search_endpoint, credential=credential)
index_client.create_or_update_knowledge_source(knowledge_source=ks, api_version=search_api_version)
print(f"Knowledge source '{knowledge_source_name}' created or updated successfully.")

지식 에이전트 만들기

쿼리 시 earth-knowledge-source 및 gpt-5-mini 배포를 대상으로 지정하려면 지식 에이전트가 필요합니다. 다음 코드를 사용하여 코드 셀을 추가하고 실행하여 이전에 earth-knowledge-agent 변수를 사용하여 지정한 knowledge_agent_name라는 지식 에이전트를 정의합니다.

reranker_threshold는 순위 재지정 점수가 2.5이거나 이보다 낮은 응답을 제외하여 의미론적 관련성을 보장합니다. 한편, modality는 검색된 문서를 인용하는 자연어 응답을 사용하도록 설정하도록 ANSWER_SYNTHESIS로 설정됩니다.

from azure.search.documents.indexes.models import KnowledgeAgent, KnowledgeAgentAzureOpenAIModel, KnowledgeSourceReference, AzureOpenAIVectorizerParameters, KnowledgeAgentOutputConfiguration, KnowledgeAgentOutputConfigurationModality
from azure.search.documents.indexes import SearchIndexClient

aoai_params = AzureOpenAIVectorizerParameters(
    resource_url=aoai_endpoint,
    deployment_name=aoai_gpt_deployment,
    model_name=aoai_gpt_model,
)

output_cfg = KnowledgeAgentOutputConfiguration(
    modality=KnowledgeAgentOutputConfigurationModality.ANSWER_SYNTHESIS,
    include_activity=True,
)

agent = KnowledgeAgent(
    name=knowledge_agent_name,
    models=[KnowledgeAgentAzureOpenAIModel(azure_open_ai_parameters=aoai_params)],
    knowledge_sources=[
        KnowledgeSourceReference(
            name=knowledge_source_name,
            reranker_threshold=2.5,
        )
    ],
    output_configuration=output_cfg,
)

index_client = SearchIndexClient(endpoint=search_endpoint, credential=credential)
index_client.create_or_update_agent(agent, api_version=search_api_version)
print(f"Knowledge agent '{knowledge_agent_name}' created or updated successfully.")

메시지 설정

메시지는 검색 경로에 대한 입력이며 대화 기록을 포함합니다. 각 메시지에는 원본(예: system 및 user)을 나타내는 역할과 자연어 콘텐츠가 포함됩니다. 사용하는 LLM은 유효한 역할을 결정합니다.

다음 코드를 사용하여 코드 셀을 추가하고 실행하여 Earth at night에 대한 질문에 답변하고 답변을 사용할 수 없는 경우 "알 수 없음"으로 응답하도록 earth-knowledge-agent에 지시하는 시스템 메시지를 만듭니다.

instructions = """
A Q&A agent that can answer questions about the Earth at night.
If you don't have the answer, respond with "I don't know".
"""

messages = [
    {
        "role": "system",
        "content": instructions
    }
]

검색 파이프라인 실행

에이전트 검색을 실행할 준비가 완료되었습니다. 다음 코드를 사용하여 코드 셀을 추가하고 실행하여 두 부분으로 구성된 사용자 쿼리를 earth-knowledge-agent로 보냅니다.

대화 내용 및 검색 매개 변수가 지정된 경우 에이전트는 다음을 수행합니다.

1. 전체 대화를 분석하여 사용자에게 어떤 정보가 필요한지를 유추합니다.
2. 복합 쿼리를 중요 하위 쿼리로 분해합니다.
3. 지식 원본에 대해 하위 쿼리를 동시에 실행합니다.
4. 의미 순위매기기를 사용하여 결과의 순위를 다시 매기고 필터링합니다.
5. 상위 결과를 자연어 응답으로 합성합니다.

from azure.search.documents.agent import KnowledgeAgentRetrievalClient
from azure.search.documents.agent.models import KnowledgeAgentRetrievalRequest, KnowledgeAgentMessage, KnowledgeAgentMessageTextContent, SearchIndexKnowledgeSourceParams

agent_client = KnowledgeAgentRetrievalClient(endpoint=search_endpoint, agent_name=knowledge_agent_name, credential=credential)
query_1 = """
    Why do suburban belts display larger December brightening than urban cores even though absolute light levels are higher downtown?
    Why is the Phoenix nighttime street grid is so sharply visible from space, whereas large stretches of the interstate between midwestern cities remain comparatively dim?
    """

messages.append({
    "role": "user",
    "content": query_1
})

req = KnowledgeAgentRetrievalRequest(
    messages=[
        KnowledgeAgentMessage(
            role=m["role"],
            content=[KnowledgeAgentMessageTextContent(text=m["content"])]
        ) for m in messages if m["role"] != "system"
    ],
    knowledge_source_params=[
        SearchIndexKnowledgeSourceParams(
            knowledge_source_name=knowledge_source_name,
            kind="searchIndex"
        )
    ]
)

result = agent_client.retrieve(retrieval_request=req, api_version=search_api_version)
print(f"Retrieved content from '{knowledge_source_name}' successfully.")

응답, 활동 및 결과 검토

다음 코드를 사용하여 코드 셀을 추가하고 실행하여 검색 파이프라인의 응답, 활동 및 결과를 표시합니다.

import textwrap
import json

print("Response")
print(textwrap.fill(result.response[0].content[0].text, width=120))

print("Activity")
print(json.dumps([a.as_dict() for a in result.activity], indent=2))

print("Results")
print(json.dumps([r.as_dict() for r in result.references], indent=2))

출력은 다음 예제와 유사해야 합니다. 여기서는 다음과 같습니다.

• Response는 검색된 문서를 인용하는 쿼리에 대한 합성된 LLM 생성 답변을 제공합니다. 응답 합성을 사용할 수 없는 경우 이 섹션에는 문서에서 직접 추출된 콘텐츠가 포함됩니다.

• Activity는 gpt-5-mini 배포에서 생성된 하위 쿼리와 의미론적 순위 지정, 쿼리 계획 및 응답 합성에 사용되는 토큰을 포함하여 검색 프로세스 중에 수행된 단계를 추적합니다.

• Results 에는 응답에 기여한 문서가 나열되고 각 문서는 해당 doc_key문서로 식별됩니다.

Response
Suburban belts display larger December brightening than urban cores despite higher absolute light levels downtown
because the urban grid encourages outward growth along city borders, fueled by widespread personal automobile use,
leading to extensive suburban and residential municipalities linked by surface streets and freeways. This expansion
results in increased lighting in suburban areas during December, reflecting growth and development patterns rather than
just absolute light intensity downtown [ref_id:0].  The Phoenix nighttime street grid is sharply visible from space
because the metropolitan area is laid out along a regular grid of city blocks and streets, with major street lighting
clearly visible from low-Earth orbit. The grid pattern is especially evident at night due to street lighting, and major
transportation corridors like Grand Avenue and brightly lit commercial properties enhance this visibility. In contrast,
large stretches of interstate highways between Midwestern cities remain comparatively dim because, although the United
States has extensive road networks, the lighting along interstate highways is less intense and continuous than the dense
urban street grids. Additionally, navigable rivers and less urbanized areas show less light, indicating that lighting
intensity correlates with urban density and development patterns rather than just the presence of transportation
corridors [ref_id:0][ref_id:1][ref_id:2].
Activity
[
  {
    "id": 0,
    "type": "modelQueryPlanning",
    "elapsed_ms": 4572,
    "input_tokens": 2071,
    "output_tokens": 166
  },
  {
    "id": 1,
    "type": "searchIndex",
    "elapsed_ms": 608,
    "knowledge_source_name": "earth-knowledge-source",
    "query_time": "2025-09-05T17:38:49.330Z",
    "count": 0,
    "search_index_arguments": {
      "search": "Reasons for larger December brightening in suburban belts compared to urban cores despite higher downtown light levels"
    }
  },
    ... // Trimmed for brevity
  {
    "id": 4,
    "type": "semanticReranker",
    "input_tokens": 68989
  },
  {
    "id": 5,
    "type": "modelAnswerSynthesis",
    "elapsed_ms": 5619,
    "input_tokens": 3931,
    "output_tokens": 249
  }
]
Results
[
  {
    "type": "searchIndex",
    "id": "0",
    "activity_source": 2,
    "reranker_score": 2.6642752,
    "doc_key": "earth_at_night_508_page_104_verbalized"
  },
  ... // Trimmed for brevity
]

대화 계속

다음 코드를 사용하여 코드 셀을 추가하고 실행하여 earth-knowledge-agent와의 대화를 계속합니다. 이 사용자 쿼리를 보낸 후 에이전트는 관련 콘텐츠를 earth-knowledge-source에서 가져와서 messages 목록에 응답을 추가합니다.

query_2 = "How do I find lava at night?"
messages.append({
    "role": "user",
    "content": query_2
})

req = KnowledgeAgentRetrievalRequest(
    messages=[
        KnowledgeAgentMessage(
            role=m["role"],
            content=[KnowledgeAgentMessageTextContent(text=m["content"])]
        ) for m in messages if m["role"] != "system"
    ],
    knowledge_source_params=[
        SearchIndexKnowledgeSourceParams(
            knowledge_source_name=knowledge_source_name,
            kind="searchIndex"
        )
    ]
)

result = agent_client.retrieve(retrieval_request=req, api_version=search_api_version)
print(f"Retrieved content from '{knowledge_source_name}' successfully.")

새 응답, 활동 및 결과 검토

다음 코드를 사용하여 코드 셀을 추가하고 실행하여 검색 파이프라인의 새 응답, 활동 및 결과를 표시합니다.

import textwrap
import json

print("Response")
print(textwrap.fill(result.response[0].content[0].text, width=120))

print("Activity")
print(json.dumps([a.as_dict() for a in result.activity], indent=2))

print("Results")
print(json.dumps([r.as_dict() for r in result.references], indent=2))

자원을 정리하세요

본인 소유의 구독에서 작업하는 경우 자신이 만든 리소스가 여전히 필요한지 여부를 결정하여 프로젝트를 완료하는 것이 좋습니다. 실행 상태로 방치된 리소스는 비용이 들 수 있습니다.

Azure Portal의 왼쪽 창에서 모든 리소스 또는 리소스 그룹을 선택하여 Azure AI 검색 및 Azure AI Foundry 리소스를 관리할 수 있습니다.

그렇지 않은 경우 다음 코드와 함께 코드 셀을 추가하고 실행하여 이 빠른 시작에서 만든 개체를 삭제합니다.

지식 에이전트 삭제

from azure.search.documents.indexes import SearchIndexClient

index_client = SearchIndexClient(endpoint=search_endpoint, credential=credential)
index_client.delete_agent(knowledge_agent_name)
print(f"Knowledge agent '{knowledge_agent_name}' deleted successfully.")

지식 원본 삭제

from azure.search.documents.indexes import SearchIndexClient

index_client = SearchIndexClient(endpoint=search_endpoint, credential=credential)
index_client.delete_knowledge_source(knowledge_source=knowledge_source_name)
print(f"Knowledge source '{knowledge_source_name}' deleted successfully.")

검색 인덱스 삭제

from azure.search.documents.indexes import SearchIndexClient

index_client = SearchIndexClient(endpoint=search_endpoint, credential=credential)
index_client.delete_index(index_name)
print(f"Index '{index_name}' deleted successfully.")

이 빠른 시작에서는 에이전트 검색 을 사용하여 LLM(대규모 언어 모델) 및 독점 데이터로 구동되는 대화형 검색 환경을 만듭니다. 에이전트 검색은 복잡한 사용자 쿼리를 하위 쿼리로 분해하고, 하위 쿼리를 병렬로 실행하고, Azure AI Search에서 인덱싱된 문서에서 접지 데이터를 추출합니다. 출력은 에이전트 및 사용자 지정 채팅 솔루션과 통합하기 위한 것입니다.

사용자 고유의 데이터를 제공할 수 있지만, 이 빠른 시작에서는 NASA의 어스 앳 나이트 전자책의 샘플 JSON 문서를 사용합니다. 문서는 우주에서 관찰 된 바와 같이 밤에 지구의 일반적인 과학 주제와 이미지를 설명합니다.

필수 조건

• 활성 구독이 있는 Azure 계정. 무료로 계정을 만듭니다.

• 의미 체계 순위가 설정된 기본 계층 이상의 Azure AI Search 서비스입니다.

• Azure AI Foundry 프로젝트입니다. Azure AI Foundry 프로젝트를 만들 때 모델 배포에 필요한 Azure AI Foundry 리소스를 가져옵니다.

• Microsoft Entra ID를 사용한 키 없는 인증을 위한 Azure CLI 입니다.

액세스 구성

시작하기 전에 콘텐츠 및 작업에 액세스할 수 있는 권한이 있는지 확인합니다. 인증에는 Microsoft Entra ID를, 권한 부여에는 역할 기반 액세스에 사용하는 것이 좋습니다. 역할을 할당하려면 소유자 또는 사용자 액세스 관리자여야 합니다. 역할이 가능하지 않은 경우 대신 키 기반 인증을 사용합니다.

이 빠른 시작에 대한 액세스를 구성하려면 다음 탭을 모두 선택합니다.

엔드포인트 가져오기

각 Azure AI 검색 서비스와 Azure AI Foundry 리소스에는 리소스에 대한 네트워크 액세스를 식별하고 제공하는 고유한 URL인 엔드포인트가 있습니다. 이후 섹션에서는 이러한 엔드포인트를 지정하여 프로그래밍 방식으로 리소스에 연결합니다.

이 빠른 시작의 엔드포인트를 얻으려면 다음 탭을 모두 선택합니다.

모델 배포

에이전트 검색을 사용하려면 Azure AI Foundry 리소스에 두 개의 Azure OpenAI 모델을 배포해야 합니다.

• 텍스트-벡터 변환을 위한 포함 모델입니다. 이 빠른 시작에서는 text-embedding-3-large을 사용하지만, 다른 모든 text-embedding 모델을 사용할 수 있습니다.

• 쿼리 계획 및 응답 생성을 위한 LLM입니다. 빠른 시작은 gpt-5-mini를 사용합니다. 하지만, 에이전트 검색을 위해 지원되는 모든 LLM을 사용할 수 있습니다.

배포 지침은 Azure AI Foundry를 사용하여 Azure OpenAI 모델 배포를 참조하세요.

환경 설정

1. 애플리케이션을 포함할 새 폴더 quickstart-agentic-retrieval을 만들고 다음 명령을 사용하여 해당 폴더에서 Visual Studio Code를 엽니다.

   mkdir quickstart-agentic-retrieval && cd quickstart-agentic-retrieval
   

2. 다음 명령을 사용하여 package.json 만듭니다.

   npm init -y
   

3. package.json 다음 명령을 사용하여 ECMAScript로 업데이트합니다.

   npm pkg set type=module
   

4. JavaScript용 Azure AI 검색 클라이언트 라이브러리(Azure.Search.Documents)를 설치하려면 다음을 사용합니다.

   npm install @azure/search-documents --version 12.2.0-alpha.20250606.1
   

5. 다음을 사용하여 Azure OpenAI 클라이언트 라이브러리를 설치합니다.

   npm install @azure/openai --version 5.10.1
   

6. dotenv 패키지를 설치하여 .env 파일에서 환경 변수를 로드합니다.

   npm install dotenv
   

7. Microsoft Entra ID로 권장되는 키 없는 인증의 경우 다음을 사용하여 Azure ID 클라이언트 라이브러리를 설치합니다.

   npm install @azure/identity
   

코드 실행

1. 폴더에 명명된 .envquickstart-agentic-retrieval 새 파일을 만들고 다음 환경 변수를 추가합니다.

   AZURE_OPENAI_ENDPOINT=https://<your-ai-foundry-resource-name>.openai.azure.com/
   AZURE_OPENAI_GPT_DEPLOYMENT=gpt-5-mini
   AZURE_OPENAI_EMBEDDING_DEPLOYMENT=text-embedding-3-large
   AZURE_SEARCH_ENDPOINT=https://<your-search-service-name>.search.windows.net
   AZURE_SEARCH_INDEX_NAME=agentic-retrieval-sample
   

   <your-search-service-name> 및 <your-ai-foundry-resource-name>을 실제 Azure AI Search 서비스 이름과 Azure AI Foundry 리소스 이름으로 대체하십시오.

2. 다음 코드를 다음과 같은 새 index.ts파일에 붙여넣습니다.

   import { DefaultAzureCredential, getBearerTokenProvider } from '@azure/identity';
   import { 
       SearchIndexClient, 
       SearchClient,
       SearchIndex,
       SearchField,
       VectorSearch,
       VectorSearchProfile,
       HnswAlgorithmConfiguration,
       AzureOpenAIVectorizer,
       AzureOpenAIParameters,
       SemanticSearch,
       SemanticConfiguration,
       SemanticPrioritizedFields,
       SemanticField
   } from '@azure/search-documents';
   import { AzureOpenAI } from "openai/index.mjs";
   
   // Load the .env file if it exists
   import * as dotenv from "dotenv";
   dotenv.config();
   
   // Configuration - Update these values for your environment
   const config = {
       searchEndpoint: process.env.AZURE_SEARCH_ENDPOINT || "https://your-search-service.search.windows.net",
       azureOpenAIEndpoint: process.env.AZURE_OPENAI_ENDPOINT || "https://your-ai-foundry-resource.openai.azure.com/",
       azureOpenAIGptDeployment: process.env.AZURE_OPENAI_GPT_DEPLOYMENT || "gpt-5-mini",
       azureOpenAIGptModel: "gpt-5-mini",
       azureOpenAIApiVersion: process.env.OPENAI_API_VERSION || "2025-03-01-preview",
       azureOpenAIEmbeddingDeployment: process.env.AZURE_OPENAI_EMBEDDING_DEPLOYMENT || "text-embedding-3-large",
       azureOpenAIEmbeddingModel: "text-embedding-3-large",
       indexName: "earth_at_night",
       agentName: "earth-search-agent",
       searchApiVersion: "2025-05-01-Preview"
   };
   
   // Earth at Night document interface
   interface EarthAtNightDocument {
       id: string;
       page_chunk: string;
       page_embedding_text_3_large: number[];
       page_number: number;
   }
   
   // Knowledge agent message interface
   interface KnowledgeAgentMessage {
       role: 'user' | 'assistant' | 'system';
       content: string;
   }
   
   // Agentic retrieval response interface
   interface AgenticRetrievalResponse {
       response?: string | any[];
       references?: Array<{
           docKey?: string;
           content?: string;
           score?: number;
           referenceType?: string;
           type?: string;
           SourceData?: any;
           Id?: string;
           ActivitySource?: number;
           // Allow any additional properties
           [key: string]: any;
       }>;
       activity?: Array<{
           step?: string;
           description?: string;
           tokensUsed?: number;
           activityType?: string;
           type?: string;
           InputTokens?: number;
           OutputTokens?: number;
           TargetIndex?: string;
           QueryTime?: string;
           Query?: any;
           Count?: number;
           ElapsedMs?: number | null;
           Id?: number;
           // Allow any additional properties
           [key: string]: any;
       }>;
       // Add any other possible response fields
       [key: string]: any;
   }
   
   async function main(): Promise<void> {
       try {
           console.log("🚀 Starting Azure AI Search agentic retrieval quickstart...\n");
   
           // Initialize Azure credentials using managed identity (recommended)
           const credential = new DefaultAzureCredential();
   
           // Create search clients
           const searchIndexClient = new SearchIndexClient(config.searchEndpoint, credential);
           const searchClient = new SearchClient<EarthAtNightDocument>(config.searchEndpoint, config.indexName, credential);
   
           // Create Azure OpenAI client
           const scope = "https://cognitiveservices.azure.com/.default";
           const azureADTokenProvider = getBearerTokenProvider(credential, scope);
           const openAIClient = new AzureOpenAI({
               endpoint: config.azureOpenAIEndpoint,
               apiVersion: config.azureOpenAIApiVersion,
               azureADTokenProvider,
           });
   
           // Create search index with vector and semantic capabilities
           await createSearchIndex(searchIndexClient);
   
           // Upload sample documents
           await uploadDocuments(searchClient);
   
           // Create knowledge agent for agentic retrieval
           await createKnowledgeAgent(credential);
   
           // Run agentic retrieval with conversation
           await runAgenticRetrieval(credential, openAIClient);
   
           // Clean up - Delete knowledge agent and search index
           await deleteKnowledgeAgent(credential);
           await deleteSearchIndex(searchIndexClient);
   
           console.log("✅ Quickstart completed successfully!");
   
       } catch (error) {
           console.error("❌ Error in main execution:", error);
           throw error;
       }
   }
   
   async function createSearchIndex(indexClient: SearchIndexClient): Promise<void> {
       console.log("📊 Creating search index...");
   
       const index: SearchIndex = {
           name: config.indexName,
           fields: [
               {
                   name: "id",
                   type: "Edm.String",
                   key: true,
                   filterable: true,
                   sortable: true,
                   facetable: true
               } as SearchField,
               {
                   name: "page_chunk",
                   type: "Edm.String",
                   searchable: true,
                   filterable: false,
                   sortable: false,
                   facetable: false
               } as SearchField,
               {
                   name: "page_embedding_text_3_large",
                   type: "Collection(Edm.Single)",
                   searchable: true,
                   filterable: false,
                   sortable: false,
                   facetable: false,
                   vectorSearchDimensions: 3072,
                   vectorSearchProfileName: "hnsw_text_3_large"
               } as SearchField,
               {
                   name: "page_number",
                   type: "Edm.Int32",
                   filterable: true,
                   sortable: true,
                   facetable: true
               } as SearchField
           ],
           vectorSearch: {
               profiles: [
                   {
                       name: "hnsw_text_3_large",
                       algorithmConfigurationName: "alg",
                       vectorizerName: "azure_openai_text_3_large"
                   } as VectorSearchProfile
               ],
               algorithms: [
                   {
                       name: "alg",
                       kind: "hnsw"
                   } as HnswAlgorithmConfiguration
               ],
               vectorizers: [
                   {
                       vectorizerName: "azure_openai_text_3_large",
                       kind: "azureOpenAI",
                       parameters: {
                           resourceUrl: config.azureOpenAIEndpoint,
                           deploymentId: config.azureOpenAIEmbeddingDeployment,
                           modelName: config.azureOpenAIEmbeddingModel
                       } as AzureOpenAIParameters
                   } as AzureOpenAIVectorizer
               ]
           } as VectorSearch,
           semanticSearch: {
               defaultConfigurationName: "semantic_config",
               configurations: [
                   {
                       name: "semantic_config",
                       prioritizedFields: {
                           contentFields: [
                               { name: "page_chunk" } as SemanticField
                           ]
                       } as SemanticPrioritizedFields
                   } as SemanticConfiguration
               ]
           } as SemanticSearch
       };
   
       try {
           await indexClient.createOrUpdateIndex(index);
           console.log(`✅ Index '${config.indexName}' created or updated successfully.`);
       } catch (error) {
           console.error("❌ Error creating index:", error);
           throw error;
       }
   }
   
   async function deleteSearchIndex(indexClient: SearchIndexClient): Promise<void> {
       console.log("🗑️ Deleting search index...");
   
       try {
           await indexClient.deleteIndex(config.indexName);
           console.log(`✅ Search index '${config.indexName}' deleted successfully.`);
   
       } catch (error: any) {
           if (error?.statusCode === 404 || error?.code === 'IndexNotFound') {
               console.log(`ℹ️ Search index '${config.indexName}' does not exist or was already deleted.`);
               return;
           }
           console.error("❌ Error deleting search index:", error);
           throw error;
       }
   }
   
   // Fetch Earth at Night documents from GitHub
   async function fetchEarthAtNightDocuments(): Promise<EarthAtNightDocument[]> {
       console.log("📡 Fetching Earth at Night documents from GitHub...");
   
       const documentsUrl = "https://raw.githubusercontent.com/Azure-Samples/azure-search-sample-data/refs/heads/main/nasa-e-book/earth-at-night-json/documents.json";
   
       try {
           const response = await fetch(documentsUrl);
   
           if (!response.ok) {
               throw new Error(`Failed to fetch documents: ${response.status} ${response.statusText}`);
           }
   
           const documents = await response.json();
           console.log(`✅ Fetched ${documents.length} documents from GitHub`);
   
           // Validate and transform documents to match our interface
           const transformedDocuments: EarthAtNightDocument[] = documents.map((doc: any, index: number) => {
               return {
                   id: doc.id || String(index + 1),
                   page_chunk: doc.page_chunk || doc.content || '',
                   page_embedding_text_3_large: doc.page_embedding_text_3_large || new Array(3072).fill(0.1),
                   page_number: doc.page_number || index + 1
               };
           });
   
           return transformedDocuments;
   
       } catch (error) {
           console.error("❌ Error fetching documents from GitHub:", error);
           console.log("🔄 Falling back to sample documents...");
   
           // Fallback to sample documents if fetch fails
           return [
               {
                   id: "1",
                   page_chunk: "The Earth at night reveals the patterns of human settlement and economic activity. City lights trace the contours of civilization, creating a luminous map of where people live and work.",
                   page_embedding_text_3_large: new Array(3072).fill(0.1),
                   page_number: 1
               },
               {
                   id: "2", 
                   page_chunk: "From space, the aurora borealis appears as shimmering curtains of green and blue light dancing across the polar regions.",
                   page_embedding_text_3_large: new Array(3072).fill(0.2),
                   page_number: 2
               }
               // Add more fallback documents as needed
           ];
       }
   }
   
   async function uploadDocuments(searchClient: SearchClient<EarthAtNightDocument>): Promise<void> {
       console.log("📄 Uploading documents...");
   
       try {
           // Fetch documents from GitHub
           const documents = await fetchEarthAtNightDocuments();
   
           const result = await searchClient.uploadDocuments(documents);
           console.log(`✅ Uploaded ${result.results.length} documents successfully.`);
   
           // Wait for indexing to complete
           console.log("⏳ Waiting for document indexing to complete...");
           await new Promise(resolve => setTimeout(resolve, 5000));
           console.log("✅ Document indexing completed.");
   
       } catch (error) {
           console.error("❌ Error uploading documents:", error);
           throw error;
       }
   }
   
   async function createKnowledgeAgent(credential: DefaultAzureCredential): Promise<void> {
   
       // In case the agent already exists, delete it first
       await deleteKnowledgeAgent(credential);
   
       console.log("🤖 Creating knowledge agent...");
   
       const agentDefinition = {
           name: config.agentName,
           description: "Knowledge agent for Earth at Night e-book content",
           models: [
               {
                   kind: "azureOpenAI",
                   azureOpenAIParameters: {
                       resourceUri: config.azureOpenAIEndpoint,
                       deploymentId: config.azureOpenAIGptDeployment,
                       modelName: config.azureOpenAIGptModel
                   }
               }
           ],
           targetIndexes: [
               {
                   indexName: config.indexName,
                   defaultRerankerThreshold: 2.5
               }
           ]
       };
   
       try {
           const token = await getAccessToken(credential, "https://search.azure.com/.default");
           const response = await fetch(`${config.searchEndpoint}/agents/${config.agentName}?api-version=${config.searchApiVersion}`, {
               method: 'PUT',
               headers: {
                   'Content-Type': 'application/json',
                   'Authorization': `Bearer ${token}`
               },
               body: JSON.stringify(agentDefinition)
           });
   
           if (!response.ok) {
               const errorText = await response.text();
               throw new Error(`Failed to create knowledge agent: ${response.status} ${response.statusText}\n${errorText}`);
           }
   
           console.log(`✅ Knowledge agent '${config.agentName}' created successfully.`);
   
       } catch (error) {
           console.error("❌ Error creating knowledge agent:", error);
           throw error;
       }
   }
   
   async function runAgenticRetrieval(credential: DefaultAzureCredential, openAIClient: AzureOpenAI): Promise<void> {
       console.log("🔍 Running agentic retrieval...");
   
       const messages: KnowledgeAgentMessage[] = [
           {
               role: "system",
               content: `A Q&A agent that can answer questions about the Earth at night.
   Sources have a JSON format with a ref_id that must be cited in the answer.
   If you do not have the answer, respond with "I don't know".`
           },
           {
               role: "user",
               content: "Why do suburban belts display larger December brightening than urban cores even though absolute light levels are higher downtown? Why is the Phoenix nighttime street grid is so sharply visible from space, whereas large stretches of the interstate between midwestern cities remain comparatively dim?"
           }
       ];
   
       try {
           // Call agentic retrieval API
           const userMessages = messages.filter(m => m.role !== "system");
           const retrievalResponse = await callAgenticRetrieval(credential, userMessages);
   
           // Extract the assistant response from agentic retrieval
           let assistantContent = '';
           if (typeof retrievalResponse.response === 'string') {
               assistantContent = retrievalResponse.response;
           } else if (Array.isArray(retrievalResponse.response)) {
               assistantContent = JSON.stringify(retrievalResponse.response);
           }
   
           // Add assistant response to conversation history
           messages.push({
               role: "assistant",
               content: assistantContent
           });
   
           console.log(assistantContent);
   
           // Log activities and results...
           console.log("\nActivities:");
           if (retrievalResponse.activity && Array.isArray(retrievalResponse.activity)) {
               retrievalResponse.activity.forEach((activity) => {
                   const activityType = activity.activityType || activity.type || 'UnknownActivityRecord';
                   console.log(`Activity Type: ${activityType}`);
                   console.log(JSON.stringify(activity, null, 2));
               });
           }
   
           console.log("Results");
           if (retrievalResponse.references && Array.isArray(retrievalResponse.references)) {
               retrievalResponse.references.forEach((reference) => {
                   const referenceType = reference.referenceType || reference.type || 'AzureSearchDoc';
                   console.log(`Reference Type: ${referenceType}`);
                   console.log(JSON.stringify(reference, null, 2));
               });
           }
   
           // Now do chat completion with full conversation history
           await generateFinalAnswer(openAIClient, messages);
   
           // Continue conversation with second question
           await continueConversation(credential, openAIClient, messages);
   
       } catch (error) {
           console.error("❌ Error in agentic retrieval:", error);
           throw error;
       }
   }
   
   async function generateFinalAnswer(
       openAIClient: AzureOpenAI,
       messages: KnowledgeAgentMessage[]
   ): Promise<void> {
   
       console.log("\n[ASSISTANT]: ");
   
       try {
           const completion = await openAIClient.chat.completions.create({
               model: config.azureOpenAIGptDeployment,
               messages: messages.map(m => ({ role: m.role, content: m.content })) as any,
               max_tokens: 1000,
               temperature: 0.7
           });
   
           const answer = completion.choices[0].message.content;
           console.log(answer?.replace(/\./g, "\n"));
   
           // Add this response to conversation history
           if (answer) {
               messages.push({
                   role: "assistant",
                   content: answer
               });
           }
   
       } catch (error) {
           console.error("❌ Error generating final answer:", error);
           throw error;
       }
   }
   
   async function callAgenticRetrieval(
       credential: DefaultAzureCredential, 
       messages: KnowledgeAgentMessage[]
   ): Promise<AgenticRetrievalResponse> {
   
       // Convert messages to the correct format expected by the Knowledge agent
       const agentMessages = messages.map(msg => ({
           role: msg.role,
           content: [
               {
                   type: "text",
                   text: msg.content
               }
           ]
       }));
   
       const retrievalRequest = {
           messages: agentMessages,
           targetIndexParams: [
               {
                   indexName: config.indexName,
                   rerankerThreshold: 2.5,
                   maxDocsForReranker: 100,
                   includeReferenceSourceData: true
               }
           ]
       };
   
       const token = await getAccessToken(credential, "https://search.azure.com/.default");
       const response = await fetch(
           `${config.searchEndpoint}/agents/${config.agentName}/retrieve?api-version=${config.searchApiVersion}`,
           {
               method: 'POST',
               headers: {
                   'Content-Type': 'application/json',
                   'Authorization': `Bearer ${token}`
               },
               body: JSON.stringify(retrievalRequest)
           }
       );
   
       if (!response.ok) {
           const errorText = await response.text();
           throw new Error(`Agentic retrieval failed: ${response.status} ${response.statusText}\n${errorText}`);
       }
   
       return await response.json() as AgenticRetrievalResponse;
   }
   
   async function deleteKnowledgeAgent(credential: DefaultAzureCredential): Promise<void> {
       console.log("🗑️ Deleting knowledge agent...");
   
       try {
           const token = await getAccessToken(credential, "https://search.azure.com/.default");
           const response = await fetch(`${config.searchEndpoint}/agents/${config.agentName}?api-version=${config.searchApiVersion}`, {
               method: 'DELETE',
               headers: {
                   'Authorization': `Bearer ${token}`
               }
           });
   
           if (!response.ok) {
               if (response.status === 404) {
                   console.log(`ℹ️ Knowledge agent '${config.agentName}' does not exist or was already deleted.`);
                   return;
               }
               const errorText = await response.text();
               throw new Error(`Failed to delete knowledge agent: ${response.status} ${response.statusText}\n${errorText}`);
           }
   
           console.log(`✅ Knowledge agent '${config.agentName}' deleted successfully.`);
   
       } catch (error) {
           console.error("❌ Error deleting knowledge agent:", error);
           throw error;
       }
   }
   
   async function continueConversation(
       credential: DefaultAzureCredential,
       openAIClient: AzureOpenAI,
       messages: KnowledgeAgentMessage[]
   ): Promise<void> {
       console.log("\n💬 === Continuing Conversation ===");
   
       // Add follow-up question
       const followUpQuestion = "How do I find lava at night?"; 
       console.log(`❓ Follow-up question: ${followUpQuestion}`);
   
       messages.push({
           role: "user",
           content: followUpQuestion
       });
   
       try {
           // Don't include system messages in this retrieval
           const userAssistantMessages = messages.filter((m: KnowledgeAgentMessage) => m.role !== "system");
           const newRetrievalResponse = await callAgenticRetrieval(credential, userAssistantMessages);
   
           // Extract assistant response and add to conversation
           let assistantContent = '';
           if (typeof newRetrievalResponse.response === 'string') {
               assistantContent = newRetrievalResponse.response;
           } else if (Array.isArray(newRetrievalResponse.response)) {
               assistantContent = JSON.stringify(newRetrievalResponse.response);
           }
   
           // Add assistant response to conversation history
           messages.push({
               role: "assistant",
               content: assistantContent
           });
   
           console.log(assistantContent);
   
           // Log activities and results like the first retrieval
           console.log("\nActivities:");
           if (newRetrievalResponse.activity && Array.isArray(newRetrievalResponse.activity)) {
               newRetrievalResponse.activity.forEach((activity) => {
                   const activityType = activity.activityType || activity.type || 'UnknownActivityRecord';
                   console.log(`Activity Type: ${activityType}`);
                   console.log(JSON.stringify(activity, null, 2));
               });
           }
   
           console.log("Results");
           if (newRetrievalResponse.references && Array.isArray(newRetrievalResponse.references)) {
               newRetrievalResponse.references.forEach((reference) => {
                   const referenceType = reference.referenceType || reference.type || 'AzureSearchDoc';
                   console.log(`Reference Type: ${referenceType}`);
                   console.log(JSON.stringify(reference, null, 2));
               });
           }
   
           // Generate final answer for follow-up
           await generateFinalAnswer(openAIClient, messages);
   
           console.log("\n🎉 === Conversation Complete ===");
   
       } catch (error) {
           console.error("❌ Error in conversation continuation:", error);
           throw error;
       }
   }
   
   async function getAccessToken(credential: DefaultAzureCredential, scope: string): Promise<string> {
       const tokenResponse = await credential.getToken(scope);
       return tokenResponse.token;
   }
   
   // Error handling wrapper
   async function runWithErrorHandling(): Promise<void> {
       try {
           await main();
       } catch (error) {
           console.error("💥 Application failed:", error);
           process.exit(1);
       }
   }
   
   // Execute the application - ES module style
   runWithErrorHandling();
   
   export {
       main,
       createSearchIndex,
       deleteSearchIndex,
       fetchEarthAtNightDocuments,
       uploadDocuments,
       createKnowledgeAgent,
       deleteKnowledgeAgent,
       runAgenticRetrieval,
       EarthAtNightDocument,
       KnowledgeAgentMessage,
       AgenticRetrievalResponse
   };
   

3. tsconfig.json TypeScript 코드를 전환하기 위한 파일을 만들고, ECMAScript용으로 다음 코드를 복사하세요.

   {
       "compilerOptions": {
         "module": "NodeNext",
         "target": "ES2022", // Supports top-level await
         "moduleResolution": "NodeNext",
         "skipLibCheck": true, // Avoid type errors from node_modules
         "strict": true // Enable strict type-checking options
       },
       "include": ["*.ts"]
   }
   

4. TypeScript에서 JavaScript로 변환합니다.

   tsc
   

5. 다음 명령을 사용하여 Azure에 로그인합니다.

   az login
   

6. 다음 명령으로 JavaScript 코드를 실행합니다.

   node index.js
   

출력

애플리케이션의 출력은 다음과 유사하게 표시됩니다.

[dotenv@17.2.0] injecting env (0) from .env (tip: ⚙️  override existing env vars with { override: true })
🚀 Starting Azure AI Search agentic retrieval quickstart...

📊 Creating search index...
✅ Index 'earth_at_night' created or updated successfully.
📄 Uploading documents...
📡 Fetching Earth at Night documents from GitHub...
✅ Fetched 194 documents from GitHub
✅ Uploaded 194 documents successfully.
⏳ Waiting for document indexing to complete...
✅ Document indexing completed.
🗑️ Deleting knowledge agent...
ℹ️ Knowledge agent 'earth-search-agent' does not exist or was already deleted.
🤖 Creating knowledge agent...
✅ Knowledge agent 'earth-search-agent' created successfully.
🔍 Running agentic retrieval...
[{"role":"assistant","content":[{"type":"text","text":"[]"}]}]

Activities:
Activity Type: ModelQueryPlanning
{
  "type": "ModelQueryPlanning",
  "id": 0,
  "inputTokens": 1379,
  "outputTokens": 551
}
Activity Type: AzureSearchQuery
{
  "type": "AzureSearchQuery",
  "id": 1,
  "targetIndex": "earth_at_night",
  "query": {
    "search": "Why do suburban areas show greater December brightening compared to urban cores despite higher absolute light levels downtown?",
    "filter": null
  },
  "queryTime": "2025-07-20T16:12:59.804Z",
  "count": 0,
  "elapsedMs": 549
}
Activity Type: AzureSearchQuery
{
  "type": "AzureSearchQuery",
  "id": 2,
  "targetIndex": "earth_at_night",
  "query": {
    "search": "Why is the Phoenix nighttime street grid sharply visible from space, while large stretches of interstate highways between Midwestern cities appear comparatively dim?",
    "filter": null
  },
  "queryTime": "2025-07-20T16:13:00.061Z",
  "count": 0,
  "elapsedMs": 256
}
Activity Type: AzureSearchSemanticRanker
{
  "type": "AzureSearchSemanticRanker",
  "id": 3,
  "inputTokens": 47630
}
Results

[ASSISTANT]: 
Suburban belts show larger December brightening than urban cores despite higher absolute light levels downtown because suburban areas often have more seasonal variation in lighting usage, such as increased decorative and outdoor lighting during the holiday season in December
 Urban cores typically have more constant and dense lighting throughout the year, so the relative increase in brightness during December is less pronounced compared to suburban areas
\n\nThe Phoenix nighttime street grid is sharply visible from space because the city has a well-planned, extensive grid of streets with consistent and bright street lighting
 In contrast, large stretches of interstate highways between Midwestern cities appear comparatively dim because these highways have less continuous lighting and lower intensity lights, making them less visible from space
\n\n(Note: These explanations are based on general knowledge about urban lighting patterns and visibility from space; specific studies or sources were not provided
)

💬 === Continuing Conversation ===
❓ Follow-up question: How do I find lava at night?
[{"role":"assistant","content":[{"type":"text","text":"[{\"ref_id\":0,\"content\":\"<!-- PageHeader=\\\"Volcanoes\\\" -->\\n\\n### Nighttime Glow at Mount Etna - Italy\\n\\nAt about 2:30 a.m. local time on March 16, 2017, the VIIRS DNB on the Suomi NPP satellite captured this nighttime image of lava flowing on Mount Etna in Sicily, Italy. Etna is one of the world's most active volcanoes.\\n\\n#### Figure: Location of Mount Etna\\nA world globe is depicted, with a marker indicating the ___location of Mount Etna in Sicily, Italy, in southern Europe near the center of the Mediterranean Sea.\\n\\n<!-- PageFooter=\\\"Earth at Night\\\" -->\\n<!-- PageNumber=\\\"48\\\" -->\"},{\"ref_id\":1,\"content\":\"<!-- PageHeader=\\\"Volcanoes\\\" -->\\n\\n## Volcanoes\\n\\n### The Infrared Glows of Kilauea's Lava Flows—Hawaii\\n\\nIn early May 2018, an eruption on Hawaii's Kilauea volcano began to unfold. The eruption took a dangerous turn on May 3, 2018, when new fissures opened in the residential neighborhood of Leilani Estates. During the summer-long eruptive event, other fissures emerged along the East Rift Zone. Lava from vents along the rift zone flowed downslope, reaching the ocean in several areas, and filling in Kapoho Bay.\\n\\nA time series of Landsat 8 imagery shows the progression of the lava flows from May 16 to August 13. The night view combines thermal, shortwave infrared, and near-infrared wavelengths to tease out the very hot lava (bright white), cooling lava (red), and lava flows obstructed by clouds (purple).\\n\\n#### Figure: Location of Kilauea Volcano, Hawaii\\n\\nA globe is shown centered on North America, with a marker placed in the Pacific Ocean indicating the ___location of Hawaii, to the southwest of the mainland United States.\\n\\n<!-- PageFooter=\\\"Earth at Night\\\" -->\\n<!-- PageNumber=\\\"44\\\" -->\"},{\"ref_id\":2,\"content\":\"For the first time in perhaps a decade, Mount Etna experienced a \\\"flank eruption\\\"—erupting from its side instead of its summit—on December 24, 2018. The activity was accompanied by 130 earthquakes occurring over three hours that morning. Mount Etna, Europe’s most active volcano, has seen periodic activity on this part of the mountain since 2013. The Operational Land Imager (OLI) on the Landsat 8 satellite acquired the main image of Mount Etna on December 28, 2018.\\n\\nThe inset image highlights the active vent and thermal infrared signature from lava flows, which can be seen near the newly formed fissure on the southeastern side of the volcano. The inset was created with data from OLI and the Thermal Infrared Sensor (TIRS) on Landsat 8. Ash spewing from the fissure cloaked adjacent villages and delayed aircraft from landing at the nearby Catania airport. Earthquakes occurred in the subsequent days after the initial eruption and displaced hundreds of people from their homes.\\n\\nFor nighttime images of Mount Etna’s March 2017 eruption, see pages 48–51.\\n\\n---\\n\\n### Hazards of Volcanic Ash Plumes and Satellite Observation\\n\\nWith the help of moonlight, satellite instruments can track volcanic ash plumes, which present significant hazards to airplanes in flight. The volcanic ash—composed of tiny pieces of glass and rock—is abrasive to engine turbine blades, and can melt on the blades and other engine parts, causing damage and even engine stalls. This poses a danger to both the plane’s integrity and passenger safety. Volcanic ash also reduces visibility for pilots and can cause etching of windshields, further reducing pilots’ ability to see. Nightlight images can be combined with thermal images to provide a more complete view of volcanic activity on Earth’s surface.\\n\\nThe VIIRS Day/Night Band (DNB) on polar-orbiting satellites uses faint light sources such as moonlight, airglow (the atmosphere’s self-illumination through chemical reactions), zodiacal light (sunlight scattered by interplanetary dust), and starlight from the Milky Way. Using these dim light sources, the DNB can detect changes in clouds, snow cover, and sea ice:\\n\\n#### Table: Light Sources Used by VIIRS DNB\\n\\n| Light Source         | Description                                                                  |\\n|----------------------|------------------------------------------------------------------------------|\\n| Moonlight            | Reflected sunlight from the Moon, illuminating Earth's surface at night      |\\n| Airglow              | Atmospheric self-illumination from chemical reactions                        |\\n| Zodiacal Light       | Sunlight scattered by interplanetary dust                                    |\\n| Starlight/Milky Way  | Faint illumination provided by stars in the Milky Way                        |\\n\\nGeostationary Operational Environmental Satellites (GOES), managed by NOAA, orbit over Earth’s equator and offer uninterrupted observations of North America. High-latitude areas such as Alaska benefit from polar-orbiting satellites like Suomi NPP, which provide overlapping coverage at the poles, enabling more data collection in these regions. During polar darkness (winter months), VIIRS DNB data allow scientists to:\\n\\n- Observe sea ice formation\\n- Monitor snow cover extent at the highest latitudes\\n- Detect open water for ship navigation\\n\\n#### Table: Satellite Coverage Overview\\n\\n| Satellite Type          | Orbit           | Coverage Area         | Special Utility                              |\\n|------------------------|-----------------|----------------------|----------------------------------------------|\\n| GOES                   | Geostationary   | Equatorial/North America | Continuous regional monitoring              |\\n| Polar-Orbiting (e.g., Suomi NPP) | Polar-orbiting    | Poles/high latitudes      | Overlapping passes; useful during polar night|\\n\\n---\\n\\n### Weather Forecasting and Nightlight Data\\n\\nThe use of nightlight data by weather forecasters is growing as the VIIRS instrument enables observation of clouds at night illuminated by sources such as moonlight and lightning. Scientists use these data to study the nighttime behavior of weather systems, including severe storms, which can develop and strike populous areas at night as well as during the day. Combined with thermal data, visible nightlight data allow the detection of clouds at various heights in the atmosphere, such as dense marine fog. This capability enables weather forecasters to issue marine advisories with higher confidence, leading to greater utility. (See \\\"Marine Layer Clouds—California\\\" on page 56.)\\n\\nIn this section of the book, you will see how nightlight data are used to observe nature’s spectacular light shows across a wide range of sources.\\n\\n---\\n\\n#### Notable Data from Mount Etna Flank Eruption (December 2018)\\n\\n| Event/Observation                  | Details                                                                    |\\n|-------------------------------------|----------------------------------------------------------------------------|\\n| Date of Flank Eruption              | December 24, 2018                                                          |\\n| Number of Earthquakes               | 130 earthquakes within 3 hours                                              |\\n| Image Acquisition                   | December 28, 2018 by Landsat 8 OLI                                         |\\n| Location of Eruption                | Southeastern side of Mount Etna                                            |\\n| Thermal Imaging Data                | From OLI and TIRS (Landsat 8), highlighting active vent and lava flows     |\\n| Impact on Villages/Air Transport    | Ash covered villages; delayed aircraft at Catania airport                  |\\n| Displacement                        | Hundreds of residents displaced                                            |\\n| Ongoing Seismic Activity            | Earthquakes continued after initial eruption                               |\\n\\n---\\n\\n<!-- PageFooter=\\\"Earth at Night\\\" -->\\n<!-- PageNumber=\\\"30\\\" -->\"},{\"ref_id\":3,\"content\":\"# Volcanoes\\n\\n---\\n\\n### Mount Etna Erupts - Italy\\n\\nThe highly active Mount Etna in Italy sent red lava rolling down its flank on March 19, 2017. An astronaut onboard the ISS took the photograph below of the volcano and its environs that night. City lights surround the mostly dark volcanic area.\\n\\n---\\n\\n#### Figure 1: Location of Mount Etna, Italy\\n\\nA world map highlighting the ___location of Mount Etna in southern Italy. The marker indicates its geographic placement on the east coast of Sicily, Italy, in the Mediterranean region, south of mainland Europe and north of northern Africa.\\n\\n---\\n\\n#### Figure 2: Nighttime View of Mount Etna's Eruption and Surrounding Cities\\n\\nThis is a nighttime satellite image taken on March 19, 2017, showing the eruption of Mount Etna (southeastern cone) with visible bright red and orange coloring indicating flowing lava from a lateral vent. The surrounding areas are illuminated by city lights, with the following geographic references labeled:\\n\\n| Location        | Position in Image         | Visible Characteristics                    |\\n|-----------------|--------------------------|--------------------------------------------|\\n| Mt. Etna (southeastern cone) | Top center-left | Bright red/orange lava flow                |\\n| Lateral vent    | Left of the volcano       | Faint red/orange flow extending outwards   |\\n| Resort          | Below the volcano, to the left   | Small cluster of lights                    |\\n| Giarre          | Top right                 | Bright cluster of city lights              |\\n| Acireale        | Center right              | Large, bright area of city lights          |\\n| Biancavilla     | Bottom left               | Smaller cluster of city lights             |\\n\\nAn arrow pointing north is shown on the image for orientation.\\n\\n---\\n\\n<!-- Earth at Night Page Footer -->\\n<!-- Page Number: 50 -->\"},{\"ref_id\":4,\"content\":\"## Nature's Light Shows\\n\\nAt night, with the light of the Sun removed, nature's brilliant glow from Earth's surface becomes visible to the naked eye from space. Some of Earth's most spectacular light shows are natural, like the aurora borealis, or Northern Lights, in the Northern Hemisphere (aurora australis, or Southern Lights, in the Southern Hemisphere). The auroras are natural electrical phenomena caused by charged particles that race from the Sun toward Earth, inducing chemical reactions in the upper atmosphere and creating the appearance of streamers of reddish or greenish light in the sky, usually near the northern or southern magnetic pole. Other natural lights can indicate danger, like a raging forest fire encroaching on a city, town, or community, or lava spewing from an erupting volcano.\\n\\nWhatever the source, the ability of humans to monitor nature's light shows at night has practical applications for society. For example, tracking fires during nighttime hours allows for continuous monitoring and enhances our ability to protect humans and other animals, plants, and infrastructure. Combined with other data sources, our ability to observe the light of fires at night allows emergency managers to more efficiently and accurately issue warnings and evacuation orders and allows firefighting efforts to continue through the night. With enough moonlight (e.g., full-Moon phase), it's even possible to track the movement of smoke plumes at night, which can impact air quality, regardless of time of day.\\n\\nAnother natural source of light at night is emitted from glowing lava flows at the site of active volcanoes. Again, with enough moonlight, these dramatic scenes can be tracked and monitored for both scientific research and public safety.\\n\\n\\n### Figure: The Northern Lights Viewed from Space\\n\\n**September 17, 2011**\\n\\nThis photo, taken from the International Space Station on September 17, 2011, shows a spectacular display of the aurora borealis (Northern Lights) as green and reddish light in the night sky above Earth. In the foreground, part of a Soyuz spacecraft is visible, silhouetted against the bright auroral light. The green glow is generated by energetic charged particles from the Sun interacting with Earth's upper atmosphere, exciting oxygen and nitrogen atoms, and producing characteristic colors. The image demonstrates the vividness and grandeur of natural night-time light phenomena as seen from orbit.\"}]"}]}]

Activities:
Activity Type: ModelQueryPlanning
{
  "type": "ModelQueryPlanning",
  "id": 0,
  "inputTokens": 1598,
  "outputTokens": 159
}
Activity Type: AzureSearchQuery
{
  "type": "AzureSearchQuery",
  "id": 1,
  "targetIndex": "earth_at_night",
  "query": {
    "search": "How can I locate lava flows during nighttime?",
    "filter": null
  },
  "queryTime": "2025-07-20T16:13:10.659Z",
  "count": 5,
  "elapsedMs": 260
}
Activity Type: AzureSearchSemanticRanker
{
  "type": "AzureSearchSemanticRanker",
  "id": 2,
  "inputTokens": 24146
}
Results
Reference Type: AzureSearchDoc
{
  "type": "AzureSearchDoc",
  "id": "0",
  "activitySource": 1,
  "docKey": "earth_at_night_508_page_64_verbalized",
  "sourceData": {
    "id": "earth_at_night_508_page_64_verbalized",
    "page_chunk": "<!-- PageHeader=\"Volcanoes\" -->\n\n### Nighttime Glow at Mount Etna - Italy\n\nAt about 2:30 a.m. local time on March 16, 2017, the VIIRS DNB on the Suomi NPP satellite captured this nighttime image of lava flowing on Mount Etna in Sicily, Italy. Etna is one of the world's most active volcanoes.\n\n#### Figure: Location of Mount Etna\nA world globe is depicted, with a marker indicating the ___location of Mount Etna in Sicily, Italy, in southern Europe near the center of the Mediterranean Sea.\n\n<!-- PageFooter=\"Earth at Night\" -->\n<!-- PageNumber=\"48\" -->"
  }
}
Reference Type: AzureSearchDoc
{
  "type": "AzureSearchDoc",
  "id": "1",
  "activitySource": 1,
  "docKey": "earth_at_night_508_page_60_verbalized",
  "sourceData": {
    "id": "earth_at_night_508_page_60_verbalized",
    "page_chunk": "<!-- PageHeader=\"Volcanoes\" -->\n\n## Volcanoes\n\n### The Infrared Glows of Kilauea's Lava Flows—Hawaii\n\nIn early May 2018, an eruption on Hawaii's Kilauea volcano began to unfold. The eruption took a dangerous turn on May 3, 2018, when new fissures opened in the residential neighborhood of Leilani Estates. During the summer-long eruptive event, other fissures emerged along the East Rift Zone. Lava from vents along the rift zone flowed downslope, reaching the ocean in several areas, and filling in Kapoho Bay.\n\nA time series of Landsat 8 imagery shows the progression of the lava flows from May 16 to August 13. The night view combines thermal, shortwave infrared, and near-infrared wavelengths to tease out the very hot lava (bright white), cooling lava (red), and lava flows obstructed by clouds (purple).\n\n#### Figure: Location of Kilauea Volcano, Hawaii\n\nA globe is shown centered on North America, with a marker placed in the Pacific Ocean indicating the ___location of Hawaii, to the southwest of the mainland United States.\n\n<!-- PageFooter=\"Earth at Night\" -->\n<!-- PageNumber=\"44\" -->"
  }
}
Reference Type: AzureSearchDoc
{
  "type": "AzureSearchDoc",
  "id": "2",
  "activitySource": 1,
  "docKey": "earth_at_night_508_page_46_verbalized",
  "sourceData": {
    "id": "earth_at_night_508_page_46_verbalized",
    "page_chunk": "For the first time in perhaps a decade, Mount Etna experienced a \"flank eruption\"—erupting from its side instead of its summit—on December 24, 2018. The activity was accompanied by 130 earthquakes occurring over three hours that morning. Mount Etna, Europe’s most active volcano, has seen periodic activity on this part of the mountain since 2013. The Operational Land Imager (OLI) on the Landsat 8 satellite acquired the main image of Mount Etna on December 28, 2018.\n\nThe inset image highlights the active vent and thermal infrared signature from lava flows, which can be seen near the newly formed fissure on the southeastern side of the volcano. The inset was created with data from OLI and the Thermal Infrared Sensor (TIRS) on Landsat 8. Ash spewing from the fissure cloaked adjacent villages and delayed aircraft from landing at the nearby Catania airport. Earthquakes occurred in the subsequent days after the initial eruption and displaced hundreds of people from their homes.\n\nFor nighttime images of Mount Etna’s March 2017 eruption, see pages 48–51.\n\n---\n\n### Hazards of Volcanic Ash Plumes and Satellite Observation\n\nWith the help of moonlight, satellite instruments can track volcanic ash plumes, which present significant hazards to airplanes in flight. The volcanic ash—composed of tiny pieces of glass and rock—is abrasive to engine turbine blades, and can melt on the blades and other engine parts, causing damage and even engine stalls. This poses a danger to both the plane’s integrity and passenger safety. Volcanic ash also reduces visibility for pilots and can cause etching of windshields, further reducing pilots’ ability to see. Nightlight images can be combined with thermal images to provide a more complete view of volcanic activity on Earth’s surface.\n\nThe VIIRS Day/Night Band (DNB) on polar-orbiting satellites uses faint light sources such as moonlight, airglow (the atmosphere’s self-illumination through chemical reactions), zodiacal light (sunlight scattered by interplanetary dust), and starlight from the Milky Way. Using these dim light sources, the DNB can detect changes in clouds, snow cover, and sea ice:\n\n#### Table: Light Sources Used by VIIRS DNB\n\n| Light Source         | Description                                                                  |\n|----------------------|------------------------------------------------------------------------------|\n| Moonlight            | Reflected sunlight from the Moon, illuminating Earth's surface at night      |\n| Airglow              | Atmospheric self-illumination from chemical reactions                        |\n| Zodiacal Light       | Sunlight scattered by interplanetary dust                                    |\n| Starlight/Milky Way  | Faint illumination provided by stars in the Milky Way                        |\n\nGeostationary Operational Environmental Satellites (GOES), managed by NOAA, orbit over Earth’s equator and offer uninterrupted observations of North America. High-latitude areas such as Alaska benefit from polar-orbiting satellites like Suomi NPP, which provide overlapping coverage at the poles, enabling more data collection in these regions. During polar darkness (winter months), VIIRS DNB data allow scientists to:\n\n- Observe sea ice formation\n- Monitor snow cover extent at the highest latitudes\n- Detect open water for ship navigation\n\n#### Table: Satellite Coverage Overview\n\n| Satellite Type          | Orbit           | Coverage Area         | Special Utility                              |\n|------------------------|-----------------|----------------------|----------------------------------------------|\n| GOES                   | Geostationary   | Equatorial/North America | Continuous regional monitoring              |\n| Polar-Orbiting (e.g., Suomi NPP) | Polar-orbiting    | Poles/high latitudes      | Overlapping passes; useful during polar night|\n\n---\n\n### Weather Forecasting and Nightlight Data\n\nThe use of nightlight data by weather forecasters is growing as the VIIRS instrument enables observation of clouds at night illuminated by sources such as moonlight and lightning. Scientists use these data to study the nighttime behavior of weather systems, including severe storms, which can develop and strike populous areas at night as well as during the day. Combined with thermal data, visible nightlight data allow the detection of clouds at various heights in the atmosphere, such as dense marine fog. This capability enables weather forecasters to issue marine advisories with higher confidence, leading to greater utility. (See \"Marine Layer Clouds—California\" on page 56.)\n\nIn this section of the book, you will see how nightlight data are used to observe nature’s spectacular light shows across a wide range of sources.\n\n---\n\n#### Notable Data from Mount Etna Flank Eruption (December 2018)\n\n| Event/Observation                  | Details                                                                    |\n|-------------------------------------|----------------------------------------------------------------------------|\n| Date of Flank Eruption              | December 24, 2018                                                          |\n| Number of Earthquakes               | 130 earthquakes within 3 hours                                              |\n| Image Acquisition                   | December 28, 2018 by Landsat 8 OLI                                         |\n| Location of Eruption                | Southeastern side of Mount Etna                                            |\n| Thermal Imaging Data                | From OLI and TIRS (Landsat 8), highlighting active vent and lava flows     |\n| Impact on Villages/Air Transport    | Ash covered villages; delayed aircraft at Catania airport                  |\n| Displacement                        | Hundreds of residents displaced                                            |\n| Ongoing Seismic Activity            | Earthquakes continued after initial eruption                               |\n\n---\n\n<!-- PageFooter=\"Earth at Night\" -->\n<!-- PageNumber=\"30\" -->"
  }
}
Reference Type: AzureSearchDoc
{
  "type": "AzureSearchDoc",
  "id": "3",
  "activitySource": 1,
  "docKey": "earth_at_night_508_page_66_verbalized",
  "sourceData": {
    "id": "earth_at_night_508_page_66_verbalized",
    "page_chunk": "# Volcanoes\n\n---\n\n### Mount Etna Erupts - Italy\n\nThe highly active Mount Etna in Italy sent red lava rolling down its flank on March 19, 2017. An astronaut onboard the ISS took the photograph below of the volcano and its environs that night. City lights surround the mostly dark volcanic area.\n\n---\n\n#### Figure 1: Location of Mount Etna, Italy\n\nA world map highlighting the ___location of Mount Etna in southern Italy. The marker indicates its geographic placement on the east coast of Sicily, Italy, in the Mediterranean region, south of mainland Europe and north of northern Africa.\n\n---\n\n#### Figure 2: Nighttime View of Mount Etna's Eruption and Surrounding Cities\n\nThis is a nighttime satellite image taken on March 19, 2017, showing the eruption of Mount Etna (southeastern cone) with visible bright red and orange coloring indicating flowing lava from a lateral vent. The surrounding areas are illuminated by city lights, with the following geographic references labeled:\n\n| Location        | Position in Image         | Visible Characteristics                    |\n|-----------------|--------------------------|--------------------------------------------|\n| Mt. Etna (southeastern cone) | Top center-left | Bright red/orange lava flow                |\n| Lateral vent    | Left of the volcano       | Faint red/orange flow extending outwards   |\n| Resort          | Below the volcano, to the left   | Small cluster of lights                    |\n| Giarre          | Top right                 | Bright cluster of city lights              |\n| Acireale        | Center right              | Large, bright area of city lights          |\n| Biancavilla     | Bottom left               | Smaller cluster of city lights             |\n\nAn arrow pointing north is shown on the image for orientation.\n\n---\n\n<!-- Earth at Night Page Footer -->\n<!-- Page Number: 50 -->"
  }
}
Reference Type: AzureSearchDoc
{
  "type": "AzureSearchDoc",
  "id": "4",
  "activitySource": 1,
  "docKey": "earth_at_night_508_page_44_verbalized",
  "sourceData": {
    "id": "earth_at_night_508_page_44_verbalized",
    "page_chunk": "## Nature's Light Shows\n\nAt night, with the light of the Sun removed, nature's brilliant glow from Earth's surface becomes visible to the naked eye from space. Some of Earth's most spectacular light shows are natural, like the aurora borealis, or Northern Lights, in the Northern Hemisphere (aurora australis, or Southern Lights, in the Southern Hemisphere). The auroras are natural electrical phenomena caused by charged particles that race from the Sun toward Earth, inducing chemical reactions in the upper atmosphere and creating the appearance of streamers of reddish or greenish light in the sky, usually near the northern or southern magnetic pole. Other natural lights can indicate danger, like a raging forest fire encroaching on a city, town, or community, or lava spewing from an erupting volcano.\n\nWhatever the source, the ability of humans to monitor nature's light shows at night has practical applications for society. For example, tracking fires during nighttime hours allows for continuous monitoring and enhances our ability to protect humans and other animals, plants, and infrastructure. Combined with other data sources, our ability to observe the light of fires at night allows emergency managers to more efficiently and accurately issue warnings and evacuation orders and allows firefighting efforts to continue through the night. With enough moonlight (e.g., full-Moon phase), it's even possible to track the movement of smoke plumes at night, which can impact air quality, regardless of time of day.\n\nAnother natural source of light at night is emitted from glowing lava flows at the site of active volcanoes. Again, with enough moonlight, these dramatic scenes can be tracked and monitored for both scientific research and public safety.\n\n\n### Figure: The Northern Lights Viewed from Space\n\n**September 17, 2011**\n\nThis photo, taken from the International Space Station on September 17, 2011, shows a spectacular display of the aurora borealis (Northern Lights) as green and reddish light in the night sky above Earth. In the foreground, part of a Soyuz spacecraft is visible, silhouetted against the bright auroral light. The green glow is generated by energetic charged particles from the Sun interacting with Earth's upper atmosphere, exciting oxygen and nitrogen atoms, and producing characteristic colors. The image demonstrates the vividness and grandeur of natural night-time light phenomena as seen from orbit."
  }
}

[ASSISTANT]: 
To find lava at night, satellite instruments like the VIIRS Day/Night Band (DNB) and thermal infrared sensors are used to detect the glow of very hot lava flows on the Earth's surface
 For example, nighttime satellite images have captured lava flowing from active volcanoes such as Mount Etna in Italy and Kilauea in Hawaii, where the hot lava emits bright light visible from space even at night
 Scientists combine thermal, shortwave infrared, and near-infrared data to distinguish very hot lava (bright white), cooling lava (red), and areas obscured by clouds
 Additionally, moonlight and other faint natural light sources help illuminate the surroundings to improve observation of volcanic activity at night
 Monitoring lava flow at night is important for scientific research and public safety, as it helps track volcanic eruptions and associated hazards such as ash plumes that can affect air travel and nearby communities [refs 0,1,2,3,4]


🎉 === Conversation Complete ===
🗑️ Deleting knowledge agent...
✅ Knowledge agent 'earth-search-agent' deleted successfully.
🗑️ Deleting search index...
✅ Search index 'earth_at_night' deleted successfully.
✅ Quickstart completed successfully!

코드 이해

이제 코드가 있으므로 주요 구성 요소를 세분화해 보겠습니다.

• 검색 인덱스 만들기
• 인덱스로 문서 업로드
• 지식 에이전트 만들기
• 메시지 설정
• 검색 파이프라인 실행
• 응답, 활동 및 결과 검토
• Azure OpenAI 클라이언트 만들기
• 채팅 완료 API를 사용하여 답변 생성
• 대화 계속

검색 인덱스 만들기

Azure AI Search에서 인덱스는 구조화된 데이터 컬렉션입니다. 다음 코드는 일반 텍스트 및 벡터 콘텐츠를 포함하도록 명명된 earth_at_night 인덱스를 정의합니다. 기존 인덱스를 사용할 수 있지만 에이전트 검색 워크로드에 대한 조건을 충족해야 합니다.

const index: SearchIndex = {
    name: config.indexName,
    fields: [
        {
            name: "id",
            type: "Edm.String",
            key: true,
            filterable: true,
            sortable: true,
            facetable: true
        } as SearchField,
        {
            name: "page_chunk",
            type: "Edm.String",
            searchable: true,
            filterable: false,
            sortable: false,
            facetable: false
        } as SearchField,
        {
            name: "page_embedding_text_3_large",
            type: "Collection(Edm.Single)",
            searchable: true,
            filterable: false,
            sortable: false,
            facetable: false,
            vectorSearchDimensions: 3072,
            vectorSearchProfileName: "hnsw_text_3_large"
        } as SearchField,
        {
            name: "page_number",
            type: "Edm.Int32",
            filterable: true,
            sortable: true,
            facetable: true
        } as SearchField
    ],
    vectorSearch: {
        profiles: [
            {
                name: "hnsw_text_3_large",
                algorithmConfigurationName: "alg",
                vectorizerName: "azure_openai_text_3_large"
            } as VectorSearchProfile
        ],
        algorithms: [
            {
                name: "alg",
                kind: "hnsw"
            } as HnswAlgorithmConfiguration
        ],
        vectorizers: [
            {
                vectorizerName: "azure_openai_text_3_large",
                kind: "azureOpenAI",
                parameters: {
                    resourceUrl: config.azureOpenAIEndpoint,
                    deploymentId: config.azureOpenAIEmbeddingDeployment,
                    modelName: config.azureOpenAIEmbeddingModel
                } as AzureOpenAIParameters
            } as AzureOpenAIVectorizer
        ]
    } as VectorSearch,
    semanticSearch: {
        defaultConfigurationName: "semantic_config",
        configurations: [
            {
                name: "semantic_config",
                prioritizedFields: {
                    contentFields: [
                        { name: "page_chunk" } as SemanticField
                    ]
                } as SemanticPrioritizedFields
            } as SemanticConfiguration
        ]
    } as SemanticSearch
};

try {
    await indexClient.createOrUpdateIndex(index);
    console.log(`✅ Index '${config.indexName}' created or updated successfully.`);
} catch (error) {
    console.error("❌ Error creating index:", error);
    throw error;
}

인덱스 스키마에는 문서 식별 및 페이지 콘텐츠, 포함 및 숫자에 대한 필드가 포함됩니다. 또한 이전에 배포한 text-embedding-3-large 모델을 사용하는 의미 순위 및 벡터 쿼리에 대한 구성도 포함됩니다.

인덱스로 문서 업로드

현재 인덱스는 earth_at_night 비어 있습니다. 다음 코드를 실행하여 인덱스를 NASA의 어스 앳 나이트 전자책의 JSON 문서로 채웁니다. Azure AI Search에서 요구하는 대로 각 문서는 인덱스 스키마에 정의된 필드 및 데이터 형식을 준수합니다.

const documentsUrl = "https://raw.githubusercontent.com/Azure-Samples/azure-search-sample-data/refs/heads/main/nasa-e-book/earth-at-night-json/documents.json";
    
try {
    const response = await fetch(documentsUrl);
    
    if (!response.ok) {
        throw new Error(`Failed to fetch documents: ${response.status} ${response.statusText}`);
    }
    
    const documents = await response.json();
    console.log(`✅ Fetched ${documents.length} documents from GitHub`);
    
    // Validate and transform documents to match our interface
    const transformedDocuments: EarthAtNightDocument[] = documents.map((doc: any, index: number) => {
        return {
            id: doc.id || String(index + 1),
            page_chunk: doc.page_chunk || doc.content || '',
            page_embedding_text_3_large: doc.page_embedding_text_3_large || new Array(3072).fill(0.1),
            page_number: doc.page_number || index + 1
        };
    });
    
    return transformedDocuments;
    
}

지식 에이전트 만들기

Azure AI Search를 gpt-5-mini 배포에 연결하고 쿼리 시 earth_at_night 인덱스를 대상으로 지정하려면, 지식 에이전트가 필요합니다. 다음 코드는 에이전트 정의를 사용하여 쿼리를 처리하고 earth-search-agent 인덱스에서 관련 문서를 검색하는 earth_at_night라는 이름의 지식 에이전트를 정의합니다.

관련성이 있고 의미 있는 응답을 defaultRerankerThreshold 보장하기 위해 재랜커 점수 2.5 가 낮거나 낮은 응답을 제외하도록 설정됩니다.

const agentDefinition = {
    name: config.agentName,
    description: "Knowledge agent for Earth at Night e-book content",
    models: [
        {
            kind: "azureOpenAI",
            azureOpenAIParameters: {
                resourceUri: config.azureOpenAIEndpoint,
                deploymentId: config.azureOpenAIGptDeployment,
                modelName: config.azureOpenAIGptModel
            }
        }
    ],
    targetIndexes: [
        {
            indexName: config.indexName,
            defaultRerankerThreshold: 2.5
        }
    ]
};   

메시지 설정

메시지는 검색 경로에 대한 입력이며 대화 기록을 포함합니다. 각 메시지에는 도우미 또는 사용자와 같은 원본을 나타내는 역할과 자연어로 된 콘텐츠가 포함됩니다. 사용하는 LLM은 유효한 역할을 결정합니다.

사용자 메시지는 처리할 쿼리를 나타내고 보조 메시지는 기술 에이전트에 응답하는 방법을 안내합니다. 검색 프로세스 중에 이러한 메시지는 인덱싱된 문서에서 관련 응답을 추출하기 위해 LLM으로 전송됩니다.

이 보조 메시지는 밤에 지구에 대한 질문에 대답하고, 소스를 earth-search-agent사용하여 인용하고, 답변을 사용할 수 없을 때 "나는 모른다"로 응답하도록 지시 ref_id 합니다.

const messages: KnowledgeAgentMessage[] = [
    {
        role: "system",
        content: `A Q&A agent that can answer questions about the Earth at night.
Sources have a JSON format with a ref_id that must be cited in the answer.
If you do not have the answer, respond with "I don't know".`
    },
    {
        role: "user",
        content: "Why do suburban belts display larger December brightening than urban cores even though absolute light levels are higher downtown? Why is the Phoenix nighttime street grid is so sharply visible from space, whereas large stretches of the interstate between midwestern cities remain comparatively dim?"
    }
];

검색 파이프라인 실행

이 단계에서는 검색 파이프라인을 실행하여 검색 인덱스에서 관련 정보를 추출합니다. 검색 요청의 메시지 및 매개 변수에 따라 LLM은 다음을 수행합니다.

1. 전체 대화 기록을 분석하여 기본 정보가 필요한지 확인합니다.
2. 복합 사용자 쿼리를 포커스가 있는 하위 쿼리로 구분합니다.
3. 인덱스의 텍스트 필드 및 벡터 포함에 대해 각 하위 쿼리를 동시에 실행합니다.
4. 의미 체계 순위자를 사용하여 모든 하위 쿼리의 결과를 다시 표시합니다.
5. 결과를 단일 문자열로 병합합니다.

다음 코드는 쿼리를 하위 쿼리로 분해하고, 인덱스의 텍스트 필드와 벡터 포함에 대해 하위 쿼리를 실행하고, 결과의 순위를 지정하고 병합하는 earth-search-agent 두 부분으로 구성된 사용자 쿼리earth_at_night를 보냅니다. 그런 다음 응답이 목록에 추가됩니다 messages .

const agentMessages = messages.map(msg => ({
    role: msg.role,
    content: [
        {
            type: "text",
            text: msg.content
        }
    ]
}));

const retrievalRequest = {
    messages: agentMessages,
    targetIndexParams: [
        {
            indexName: config.indexName,
            rerankerThreshold: 2.5,
            maxDocsForReranker: 100,
            includeReferenceSourceData: true
        }
    ]
};

const token = await getAccessToken(credential, "https://search.azure.com/.default");
const response = await fetch(
    `${config.searchEndpoint}/agents/${config.agentName}/retrieve?api-version=${config.searchApiVersion}`,
    {
        method: 'POST',
        headers: {
            'Content-Type': 'application/json',
            'Authorization': `Bearer ${token}`
        },
        body: JSON.stringify(retrievalRequest)
    }
);

if (!response.ok) {
    const errorText = await response.text();
    throw new Error(`Agentic retrieval failed: ${response.status} ${response.statusText}\n${errorText}`);
}

return await response.json() as AgenticRetrievalResponse;

응답, 활동 및 결과 검토

이제 검색 파이프라인의 응답, 활동 및 결과를 표시하려고 합니다.

Azure AI Search의 각 검색 응답에는 다음이 포함됩니다.

• 검색 결과의 접지 데이터를 나타내는 통합 문자열입니다.

• 쿼리 계획.

• 통합 문자열에 기여한 원본 문서의 청크를 보여 주는 참조 데이터입니다.

console.log("\nActivities:");
if (retrievalResponse.activity && Array.isArray(retrievalResponse.activity)) {
    retrievalResponse.activity.forEach((activity) => {
        const activityType = activity.activityType || activity.type || 'UnknownActivityRecord';
        console.log(`Activity Type: ${activityType}`);
        console.log(JSON.stringify(activity, null, 2));
    });
}

console.log("Results");
if (retrievalResponse.references && Array.isArray(retrievalResponse.references)) {
    retrievalResponse.references.forEach((reference) => {
        const referenceType = reference.referenceType || reference.type || 'AzureSearchDoc';
        console.log(`Reference Type: ${referenceType}`);
        console.log(JSON.stringify(reference, null, 2));
    });
}

출력에는 다음이 포함되어야 합니다.

• Response 는 사용자 쿼리를 기반으로 검색 인덱스에 가장 관련성이 큰 문서(또는 청크)의 텍스트 문자열을 제공합니다. 이 빠른 시작의 뒷부분에서와 같이 응답 생성을 위해 이 문자열을 LLM에 전달할 수 있습니다.

• Activity 는 배포에서 생성된 하위 쿼리 및 쿼리 계획 및 실행에 사용되는 토큰을 포함하여 검색 프로세스 중에 수행된 gpt-5-mini 단계를 추적합니다.

• Results 에는 응답에 기여한 문서가 나열되고 각 문서는 해당 DocKey문서로 식별됩니다.

Azure OpenAI 클라이언트 만들기

응답 추출에서 응답 생성으로 검색 처리 단계를 확장하려면 배포와 상호 작용할 수 있도록 Azure OpenAI 클라이언트를 설정합니다.

const scope = "https://cognitiveservices.azure.com/.default";
const azureADTokenProvider = getBearerTokenProvider(credential, scope);
const openAIClient = new AzureOpenAI({
    endpoint: config.azureOpenAIEndpoint,
    apiVersion: config.azureOpenAIApiVersion,
    azureADTokenProvider,
});

채팅 완료 API를 사용하여 답변 생성

응답 생성을 위한 한 가지 옵션은 대화를 처리하기 위해 LLM에 대화 기록을 전달하는 채팅 완료 API입니다.

const completion = await openAIClient.chat.completions.create({
    model: config.azureOpenAIGptDeployment,
    messages: messages.map(m => ({ role: m.role, content: m.content })) as any,
    max_tokens: 1000,
    temperature: 0.7
});

const answer = completion.choices[0].message.content;
console.log(answer?.replace(/\./g, "\n"));

대화 계속

다른 사용자에게 쿼리를 보내 대화를 계속하세요 earth-search-agent. 다음 코드는 검색 파이프라인을 다시 실행하여 인덱스에서 earth_at_night 관련 콘텐츠를 가져오고 목록에 응답을 messages 추가합니다. 그러나 이전과 달리 이제 Azure OpenAI 클라이언트를 사용하여 검색된 콘텐츠에 따라 답변을 생성할 수 있습니다.

const followUpQuestion = "How do I find lava at night?"; 
console.log(`❓ Follow-up question: ${followUpQuestion}`);

messages.push({
    role: "user",
    content: followUpQuestion
}); 

자원을 정리하세요

자체 구독에서 작업할 때 만든 리소스가 여전히 필요한지 여부를 결정하여 프로젝트를 완료하는 것이 좋습니다. 실행 상태로 방치된 리소스는 비용이 들 수 있습니다. 리소스를 개별적으로 삭제하거나 리소스 그룹을 삭제하여 전체 리소스 집합을 삭제할 수 있습니다.

Azure Portal의 왼쪽 창에서 모든 리소스 또는 리소스 그룹을 선택하여 리소스 를 찾고 관리할 수 있습니다. 다음 코드를 실행하여 이 빠른 시작에서 만든 개체를 삭제할 수도 있습니다.

지식 에이전트 삭제

이 빠른 시작에서 만든 지식 에이전트는 다음 코드 조각을 사용하여 삭제되었습니다.

const token = await getAccessToken(credential, "https://search.azure.com/.default");
const response = await fetch(`${config.searchEndpoint}/agents/${config.agentName}?api-version=${config.searchApiVersion}`, {
    method: 'DELETE',
    headers: {
        'Authorization': `Bearer ${token}`
    }
});

검색 인덱스 삭제

이 빠른 시작에서 만든 검색 인덱스가 다음 코드 조각을 사용하여 삭제되었습니다.

await indexClient.deleteIndex(config.indexName);
console.log(`✅ Search index '${config.indexName}' deleted successfully.`);

이 빠른 시작에서는 에이전트 검색을 사용하여 Azure AI 검색에서 인덱싱된 문서와 Azure AI Foundry 모델의 Azure OpenAI에서 LLM(대규모 언어 모델)으로 구동되는 대화형 검색 환경을 만듭니다.

지식 에이전트는 복잡한 쿼리를 하위 쿼리로 분해하고, 하나 이상의 지식 원본에 대해 하위 쿼리를 실행하고, 메타데이터를 사용하여 결과를 반환하여 에이전트 검색을 오케스트레이션합니다. 기본적으로 에이전트는 원본에서 원시 콘텐츠를 출력하지만 이 빠른 시작에서는 자연어 응답 생성을 위해 응답 합성 형식을 사용합니다.

사용자 고유의 데이터를 제공할 수 있지만, 이 빠른 시작에서는 NASA의 어스 앳 나이트 전자책의 샘플 JSON 문서를 사용합니다. 문서는 우주에서 관찰 된 바와 같이 밤에 지구의 일반적인 과학 주제와 이미지를 설명합니다.

필수 조건

• 활성 구독이 있는 Azure 계정. 무료로 계정을 만듭니다.

• 의미 체계 순위가 설정된 기본 계층 이상의 Azure AI Search 서비스입니다.

• Azure AI Foundry 프로젝트 및 Azure AI Foundry 리소스. 프로젝트를 만들면 리소스가 자동으로 만들어집니다.

• Microsoft Entra ID를 사용한 키 없는 인증을 위한 Azure CLI 입니다.

• REST 클라이언트 확장이 있는 Visual Studio Code입니다.

액세스 구성

시작하기 전에 콘텐츠 및 작업에 액세스할 수 있는 권한이 있는지 확인합니다. 인증에는 Microsoft Entra ID를, 권한 부여에는 역할 기반 액세스에 사용하는 것이 좋습니다. 역할을 할당하려면 소유자 또는 사용자 액세스 관리자여야 합니다. 역할이 가능하지 않은 경우 대신 키 기반 인증을 사용합니다.

이 빠른 시작에 대한 액세스를 구성하려면 다음 탭을 모두 선택합니다.

엔드포인트 가져오기

각 Azure AI 검색 서비스와 Azure AI Foundry 리소스에는 리소스에 대한 네트워크 액세스를 식별하고 제공하는 고유한 URL인 엔드포인트가 있습니다. 이후 섹션에서는 이러한 엔드포인트를 지정하여 프로그래밍 방식으로 리소스에 연결합니다.

이 빠른 시작의 엔드포인트를 얻으려면 다음 탭을 모두 선택합니다.

모델 배포

에이전트 검색을 사용하려면 Azure AI Foundry 리소스에 두 개의 Azure OpenAI 모델을 배포해야 합니다.

• 텍스트-벡터 변환을 위한 포함 모델입니다. 이 빠른 시작에서 text-embedding-3-large를 사용하지만, 모든 text-embedding 모델을 사용할 수 있습니다.

• 쿼리 계획 및 응답 생성을 위한 LLM입니다. 이 빠른 시작에서는 gpt-5-mini을 사용하지만, 에이전트 검색에 지원되는 모든 LLM을 사용할 수 있습니다.

배포 지침은 Azure AI Foundry를 사용하여 Azure OpenAI 모델 배포를 참조하세요.

로컬 시스템에서 연결

Azure AI Foundry에서 Azure AI 검색 및 Azure OpenAI와 상호 작용하도록 역할 기반 액세스를 구성했습니다. 명령줄에서 Azure CLI를 사용하여 두 리소스에 대해 동일한 구독 및 테넌트에 로그인합니다. 자세한 내용은 빠른 시작: 키 없이 연결하세요.

로컬 시스템에서 연결하려면 다음을 수행합니다.

1. PowerShell과 같은 명령줄 도구를 엽니다.

2. Azure 계정에 로그인합니다. 구독이 여러 개 있는 경우 Azure AI Search 서비스 및 Azure AI Foundry 프로젝트가 포함된 구독을 선택합니다.

   az login
   

3. Microsoft Entra ID 토큰을 생성합니다.

   az account get-access-token --scope https://search.azure.com/.default --query accessToken --output tsv
   

4. 다음 섹션에서 사용할 토큰을 기록해 둡니다.

연결을 불러오기

요청을 보내기 전에 Azure AI Foundry에서 Azure AI 검색 및 Azure OpenAI에 대한 연결을 위한 엔드포인트, 자격 증명 및 배포 세부 정보를 정의합니다. 이러한 값은 다음 섹션에서 사용됩니다.

연결을 로드하려면 다음을 수행합니다.

1. Visual Studio Code에서 agentic-retrieval.rest라는 파일을 만듭니다.

2. 다음 변수 및 HTTP 요청을 파일에 붙여넣습니다.

   @search-url = PUT-YOUR-SEARCH-SERVICE-URL-HERE
   @token = PUT-YOUR-MICROSOFT-ENTRA-TOKEN-HERE
   @aoai-url = PUT-YOUR-AOAI-FOUNDRY-URL-HERE
   @aoai-embedding-model = text-embedding-3-large
   @aoai-embedding-deployment = text-embedding-3-large
   @aoai-gpt-model = gpt-5-mini
   @aoai-gpt-deployment = gpt-5-mini
   @index-name = earth-at-night
   @knowledge-source-name = earth-knowledge-source
   @knowledge-agent-name = earth-knowledge-agent
   @api-version = 2025-08-01-Preview
   
   ### List existing indexes by name
   GET {{search-url}}/indexes?api-version={{api-version}}  HTTP/1.1
       Content-Type: application/json
       Authorization: Bearer {{token}}
   

3. Set @search-url 및 aoai-url를 엔드포인트 가져오기에서 얻은 값으로 설정합니다.

4. @token을 로컬 시스템에서 연결에서 얻은 값으로 설정합니다.

5. ### List existing indexes by name 아래에서 요청 보내기를 선택하여 검색 서비스에 대한 연결을 확인합니다.

   인접한 창에 응답이 표시됩니다. 기존 인덱스가 있는 경우 인덱스가 나열됩니다. 그러지 않으면 목록은 비어 있습니다. HTTP 코드가 200 OK일 경우, 계속 진행할 준비가 된 것입니다.

검색 인덱스 만들기

Azure AI Search에서 인덱스는 구조화된 데이터 컬렉션입니다. 인덱스 - 만들기(REST API)를 사용하여 이전에 earth-at-night 변수를 사용하여 지정한 @index-name라는 인덱스를 정의합니다.

인덱스 스키마에는 문서 식별 및 페이지 콘텐츠, 포함 및 숫자에 대한 필드가 포함됩니다. 또한 스키마에는 의미론적 순위 지정 및 벡터 검색에 대한 구성이 포함되어 있으며, text-embedding-3-large 배포를 사용하여 텍스트를 벡터화하고 의미론적 유사성에 따라 문서를 일치시킬 수 있습니다.

### Create an index
PUT {{search-url}}/indexes/{{index-name}}?api-version={{api-version}}  HTTP/1.1
    Content-Type: application/json
    Authorization: Bearer {{token}}

    {
        "name": "{{index-name}}",
        "fields": [
            {
                "name": "id",
                "type": "Edm.String",
                "key": true
            },
            {
                "name": "page_chunk",
                "type": "Edm.String",
                "searchable": true
            },
            {
                "name": "page_embedding_text_3_large",
                "type": "Collection(Edm.Single)",
                "stored": false,
                "dimensions": 3072,
                "vectorSearchProfile": "hnsw_text_3_large"
            },
            {
                "name": "page_number",
                "type": "Edm.Int32",
                "filterable": true
            }
        ],
        "semantic": {
            "defaultConfiguration": "semantic_config",
            "configurations": [
                {
                    "name": "semantic_config",
                    "prioritizedFields": {
                    "prioritizedContentFields": [
                        {
                            "fieldName": "page_chunk"
                        }
                    ]
                    }
                }
            ]
        },
        "vectorSearch": {
            "profiles": [
                {
                    "name": "hnsw_text_3_large",
                    "algorithm": "alg",
                    "vectorizer": "azure_openai_text_3_large"
                }
            ],
            "algorithms": [
                {
                    "name": "alg",
                    "kind": "hnsw"
                }
            ],
            "vectorizers": [
                {
                    "name": "azure_openai_text_3_large",
                    "kind": "azureOpenAI",
                    "azureOpenAIParameters": {
                    "resourceUri": "{{aoai-url}}",
                    "deploymentId": "{{aoai-embedding-deployment}}",
                    "modelName": "{{aoai-embedding-model}}"
                    }
                }
            ]
        }
    }

인덱스로 문서 업로드

현재 인덱스는 earth-at-night 비어 있습니다. 문서 - 인덱스(REST API)를 사용하여 NASA의 Earth at Night 전자책에서 JSON 문서로 인덱스를 채웁니다. Azure AI Search에서 요구하는 대로 각 문서는 인덱스 스키마에 정의된 필드 및 데이터 형식을 준수합니다.

### Upload documents
POST {{search-url}}/indexes/{{index-name}}/docs/index?api-version={{api-version}}  HTTP/1.1
    Content-Type: application/json
    Authorization: Bearer {{token}}

    {
        "value": [
            {
                "@search.action": "upload",
                "id": "earth_at_night_508_page_104_verbalized",
                "page_chunk": "<!-- PageHeader=\"Urban Structure\" -->\n\n### Location of Phoenix, Arizona\n\nThe image depicts a globe highlighting the ___location of Phoenix, Arizona, in the southwestern United States, marked with a blue pinpoint on the map of North America. Phoenix is situated in the central part of Arizona, which is in the southwestern region of the United States.\n\n---\n\n### Grid of City Blocks-Phoenix, Arizona\n\nLike many large urban areas of the central and western United States, the Phoenix metropolitan area is laid out along a regular grid of city blocks and streets. While visible during the day, this grid is most evident at night, when the pattern of street lighting is clearly visible from the low-Earth-orbit vantage point of the ISS.\n\nThis astronaut photograph, taken on March 16, 2013, includes parts of several cities in the metropolitan area, including Phoenix (image right), Glendale (center), and Peoria (left). While the major street grid is oriented north-south, the northwest-southeast oriented Grand Avenue cuts across the three cities at image center. Grand Avenue is a major transportation corridor through the western metropolitan area; the lighting patterns of large industrial and commercial properties are visible along its length. Other brightly lit properties include large shopping centers, strip malls, and gas stations, which tend to be located at the intersections of north-south and east-west trending streets.\n\nThe urban grid encourages growth outwards along a city's borders by providing optimal access to new real estate. Fueled by the adoption of widespread personal automobile use during the twentieth century, the Phoenix metropolitan area today includes 25 other municipalities (many of them largely suburban and residential) linked by a network of surface streets and freeways.\n\nWhile much of the land area highlighted in this image is urbanized, there are several noticeably dark areas. The Phoenix Mountains are largely public parks and recreational land. To the west, agricultural fields provide a sharp contrast to the lit streets of residential developments. The Salt River channel appears as a dark ribbon within the urban grid.\n\n\n<!-- PageFooter=\"Earth at Night\" -->\n<!-- PageNumber=\"88\" -->",
                "page_embedding_text_3_large": [
                -0.002984904684126377, 0.0007500237552449107, -0.004803949501365423, 0.010587676428258419, -0.008392670191824436, -0.043565936386585236, 0.05432070791721344, 0.024532422423362732, -0.03305421024560928, -0.011362385004758835, 0.0029678153805434704, 0.0520421527326107, 0.019276559352874756, -0.05398651957511902, -0.025550175458192825, 0.018592992797493935, -0.02951485849916935, 0.036365706473588943, -0.02734263800084591, 0.028664197772741318, 0.027874300256371498, 0.008255957625806332, -0.05046235769987106, 0.01759042963385582, -0.003096933476626873, 0.03682141751050949, -0.002149434993043542, 0.009190164506435394, 0.0026716035790741444, -0.0031633912585675716, -0.014354884624481201, 0.004758378490805626, 0.01637520082294941, -0.010299060493707657, 0.004705212078988552, 0.016587866470217705, 0.0440824069082737, 0.019033513963222504, 0.039130352437496185, 0.04028481990098953, 0.018760086968541145, -0.05720687285065651, 0.030608562752604485, 0.010526915080845356, 0.020431026816368103, -0.04772809147834778, 0.03262887895107269, -0.02760087326169014, 0.03305421024560928, 0.009068641811609268, -0.003104528645053506, -0.035727713257074356, -0.04490268602967262, -0.039403777569532394, 0.026491977274417877, 0.01214468851685524, 0.037732839584350586, 0.08652425557374954, 0.005525491200387478, -0.0031994683668017387, 0.04684705287218094, 0.02872495912015438, 0.010481344535946846, 0.024076711386442184, 0.015813158825039864, 0.023180481046438217, -0.015949871391057968, 0.014749834313988686, -0.0006285008857958019, 0.0005739105399698019, 0.007192632649093866, 0.05787524953484535, 0.0043748221360147, 0.00038687934284098446, 0.04110509902238846, -0.0028273046482354403, 0.01397512573748827, -0.009106617420911789, -0.00770910456776619, -0.015304281376302242, 0.002582360291853547, -0.01092945970594883, -0.008552169427275658, 0.06665527075529099, 0.04657362401485443, -0.012197853997349739, -0.028861673548817635, -0.08925852179527283, -0.003831766778603196, 0.056538499891757965, -0.023453906178474426, -0.03083641827106476, -0.022223487496376038, -0.010253489017486572, 0.010807937011122704, 0.00313301058486104, 0.033904869109392166, -0.010116775520145893, 0.01742333546280861, 0.02594512514770031, -0.007777461316436529, -0.0002520649286452681, 0.005202696193009615, -0.02594512514770031, 0.010648438706994057, -0.01740814559161663, 0.0031254154164344072, -0.007542010862380266, 0.026142599061131477, 0.01731700450181961, 0.013504224829375744, -0.036183424293994904, -0.006357163190841675, -0.010428178124129772, -0.061338648200035095, 0.005457134917378426, 0.05316624045372009, 0.007861007936298847, 0.04311022534966469, 0.03867464140057564, -0.0021570303943008184, 0.016496725380420685, 0.05246748402714729, 0.007200228050351143, -0.003930503968149424, 0.007785056717693806, 0.017484096810221672, -0.003875439055263996, -0.0031538973562419415, 0.03180859982967377, -0.02497294172644615, 0.03627456724643707, 0.02533750981092453, 0.008385075256228447, -0.007298965007066727, 0.009866135194897652, -0.030168043449521065, 0.02594512514770031, 0.026233742013573647, 0.02079559490084648, -0.03396563231945038, -0.012607993558049202, -0.016162537038326263, -0.03378334641456604, -0.020582929253578186, -0.013846008107066154, 0.010215513408184052, 0.03317573294043541, 0.015706826001405716, -0.015296686440706253, 0.008491408079862595, 0.014727048575878143, 0.021828539669513702, 0.009942086413502693, -0.014096648432314396, -0.00913699809461832, -0.014354884624481201, 0.01672457903623581, -0.06118674576282501, 0.009212949313223362, 0.029970569536089897, -0.016572676599025726, 0.013071299530565739, -0.015828348696231842, 0.0012218741467222571, -0.04663438722491264, 0.01722586154937744, -0.02793506160378456, -0.035909995436668396, 0.007386309560388327, 0.04283680021762848, -0.051252253353595734, -0.036608751863241196, 0.006281211506575346, 0.029043957591056824, -0.022405771538615227, 0.011878857389092445, -0.0073141553439199924, -0.028785720467567444, -0.009904110804200172, -0.023013386875391006, -0.03527200222015381, 0.019534794613718987, -0.005905250087380409, -0.020491788163781166, 0.00045927087194286287, 0.0038450583815574646, -0.013435868546366692, 0.03840121626853943, -0.0059508210979402065, -0.023453906178474426, 0.004492547363042831, 0.05404727905988693, -0.01075477059930563, 0.04760656878352165, -0.04028481990098953, 0.03411753475666046, -0.008878761902451515, -0.02558055706322193, -0.013785246759653091, -0.010071204975247383, -0.01092945970594883, -0.04396088421344757, 0.017909428104758263, 0.03317573294043541, 0.03742903098464012, 0.02349947765469551, -0.013557391241192818, -0.004367226734757423, 0.03970758616924286, -0.002141839824616909, -0.032780785113573074, -0.008324313908815384, -0.025702079758048058, -0.02767682448029518, 0.02166144549846649, 0.03369220718741417, -0.043839361518621445, 0.011871261522173882, -0.024608373641967773, 0.015296686440706253, 0.02942371554672718, -0.015737207606434822, 0.017620811238884926, -0.01663343794643879, -0.03126174956560135, -0.02532231993973255, 0.018334757536649704, -0.04927751049399376, -0.03894806653261185, -0.02002088725566864, -0.025140035897493362, 0.016056204214692116, 0.02898319624364376, 0.029271813109517097, 0.020567739382386208, -0.006436912342905998, 0.022603247314691544, -0.023712143301963806, -0.004386214539408684, 0.030243994668126106, 0.0013244090368971229, -0.019276559352874756, -0.017043577507138252, 0.06234121322631836, -0.01757523976266384, -0.02829962968826294, 0.027099592611193657, 0.02088673785328865, 0.030168043449521065, 0.01005601417273283, -0.01537263859063387, 0.015737207606434822, 0.027904679998755455, 0.05744991824030876, -0.002301338594406843, -0.0022975411266088486, 0.004716604948043823, -0.0006194816087372601, 0.01985379308462143, -0.0403759628534317, 0.03612266108393669, 0.003028576960787177, 0.022694388404488564, 0.014218171127140522, 0.006710338871926069, -0.0023374157026410103, 0.0069951582700014114, 0.011202885769307613, -0.023195670917630196, 0.029742714017629623, -0.057753726840019226, 0.025747649371623993, -0.024000760167837143, 0.015395424328744411, -0.0019073388539254665, 0.019899364560842514, -0.009987657889723778, 0.004492547363042831, -0.018137283623218536, -0.002177916932851076, 0.004283679649233818, 0.03211240842938423, -0.03039589896798134, -0.04830532521009445, 0.037034083157777786, -0.016208108514547348, -0.018349947407841682, -0.010716794990003109, -0.0410747192800045, -0.022846292704343796, -0.08069115877151489, -0.008126839064061642, 0.024532422423362732, 0.03244659677147865, -0.010663628578186035, 0.01184847578406334, -0.05781448632478714, -0.04894332215189934, 0.002551979385316372, -0.008635716512799263, 0.028573056682944298, -0.06471090763807297, 0.033206112682819366, 0.0027589481323957443, 0.0271755438297987, -0.03211240842938423, 0.026598310098052025, -0.04472040385007858, -0.0648931935429573, -0.0012171270791441202, -0.012288996949791908, 0.0015370739856734872, -0.019200608134269714, -0.002876673359423876, 0.011954808607697487, 0.03196050599217415, -0.005316623952239752, -0.011932022869586945, -0.02916548028588295, 0.025534985587000847, -0.044446974992752075, -0.016344821080565453, 0.0257780309766531, -0.02141840010881424, 0.01109655387699604, 0.0007789803203195333, -0.022238679230213165, 0.0008444887353107333, -0.036791037768125534, -0.03806702792644501, 0.008347099646925926, 0.0020070255268365145, -0.021114591509103775, 0.05814867466688156, -0.028512295335531235, 0.031140226870775223, 0.03402639180421829, -0.0044887494295835495, 0.030517421662807465, 0.02401595003902912, -0.018000569194555283, -0.02106902189552784, -0.009676255285739899, 0.02673502266407013, -0.03305421024560928, 0.004750783089548349, 0.03314535319805145, -0.024274185299873352, -0.0007766068447381258, 0.0010823127813637257, 0.016177726909518242, -0.000631823786534369, -0.026507167145609856, 0.025200797244906425, 0.016162537038326263, -0.03621380403637886, -0.015813158825039864, 0.032598499208688736, -0.024775467813014984, 0.03305421024560928, -0.03387448936700821, -0.031565554440021515, 0.006262223701924086, -0.0447811633348465, 0.03232507407665253, -0.014878951944410801, 0.0027342636603862047, 0.005290040746331215, 0.020825974643230438, -0.0506446398794651, 0.030335137620568275, -0.04447735846042633, -0.013299155049026012, -0.01180290523916483, -0.022512104362249374, 0.003322890028357506, -0.004298869986087084, -0.008643311448395252, -0.003376056207343936, -0.0018057533307000995, 0.07036171853542328, 0.03445172309875488, -0.010640842840075493, 0.04554068297147751, 0.045662205666303635, -0.003444412723183632, -0.02305895835161209, 0.018395518884062767, 0.011939618736505508, -0.021175352856516838, -0.03445172309875488, 0.021874109283089638, -0.03867464140057564, 0.0188968013972044, -0.014408051036298275, -0.0021646255627274513, 0.006604006513953209, 0.03663913533091545, 0.022755149751901627, 0.00563562149181962, 0.05122187361121178, 0.0026165384333580732, 0.042624134570360184, -0.016056204214692116, 0.06070065498352051, 0.02384885586798191, -0.04630019888281822, 0.0049976264126598835, -0.038340453058481216, -0.014742238447070122, -0.0049368650652468204, -0.002063989406451583, -0.01803095079958439, -0.009342067874968052, 0.019534794613718987, -0.019868982955813408, 0.023742523044347763, 0.0024361531250178814, -0.006436912342905998, 0.005582455080002546, 0.0036969524808228016, -0.0536523312330246, 0.03132250905036926, -0.0433836504817009, -0.0010073103476315737, 0.012623184360563755, 0.0250792745500803, -0.01018513273447752, 0.017043577507138252, 0.0026279313024133444, -0.011962403543293476, -0.006569828372448683, 0.03332763537764549, -0.03091237135231495, 0.0039039209950715303, 0.014643501490354538, 0.010511725209653378, 0.013481439091265202, -0.03855311870574951, -0.022618437185883522, 0.03882654383778572, -0.010785151273012161, -0.024745086207985878, -0.01646634377539158, 0.05407766252756119, -0.003926706500351429, 0.01502326037734747, 0.03265926241874695, -0.034968193620443344, -0.037489794194698334, 0.04219880327582359, -0.031474415212869644, -0.0060381656512618065, 0.017043577507138252, -0.013921959325671196, -0.018395518884062767, -0.009061045944690704, 0.015486566349864006, -0.02646159753203392, 0.033114973455667496, -0.02116016298532486, 0.005882464814931154, -0.0690857321023941, 0.007568594068288803, -0.003814677707850933, -0.010823126882314682, 0.02915029041469097, 0.012600398622453213, -0.021372828632593155, 0.029408525675535202, 0.014590335078537464, -0.013390297070145607, 0.062280453741550446, -0.011180100962519646, 0.014438431710004807, -0.01636001095175743, -0.033388398587703705, 0.03888730704784393, -0.02839077264070511, -0.039312634617090225, 0.035120099782943726, 0.026051457971334457, -0.01792461797595024, 0.011742143891751766, 0.02456280216574669, 0.0014563752338290215, 0.0029070540331304073, -0.035818856209516525, 0.0275249220430851, 0.041317764669656754, 0.004484951961785555, 0.005028007086366415, -0.01323839370161295, 0.0003873540263157338, 0.01275230199098587, -0.04572296515107155, -8.188550418708473e-05, -0.008278743363916874, 0.0322035513818264, -0.05887781083583832, 0.01584353856742382, -0.014240956865251064, 0.0069951582700014114, -0.01022310834378004, 0.006417924538254738, 0.01923098787665367, 0.00792176928371191, -0.01127124298363924, -0.010777556337416172, 0.02018798142671585, -0.009744612500071526, -0.006653374992311001, -0.01602582447230816, 0.01602582447230816, -0.02673502266407013, 0.011514288373291492, 0.004579891916364431, 0.020415835082530975, -0.012995348311960697, -0.0016813823021948338, -0.009805373847484589, -0.00036077090771868825, -0.01827399618923664, 0.0027969239745289087, 0.003070350270718336, 0.01828918606042862, -0.013648533262312412, 0.003320991061627865, -0.009539542719721794, -0.02699325978755951, 0.03603151813149452, 0.026355264708399773, -0.019382892176508904, 0.021570302546024323, -0.008316718973219395, -0.024669134989380836, 0.01271432638168335, 0.039130352437496185, -0.005605240818113089, -0.03551504760980606, 0.0018370834877714515, 0.0001732649834593758, 0.025702079758048058, 0.010352225974202156, 0.018258806318044662, -0.008461027406156063, -0.002624133601784706, 0.008400266058743, 0.0012892812956124544, -0.005757144186645746, -0.005077376030385494, -0.0036342921666800976, 0.010443368926644325, -0.013830817304551601, -0.031292129307985306, 0.006797683425247669, 0.00988132506608963, -0.016663817688822746, 0.026598310098052025, -0.002910851500928402, -0.016496725380420685, -0.01913984678685665, -0.01593468151986599, -0.017438527196645737, -0.007093895226716995, -0.027874300256371498, 0.028405962511897087, 0.0023583024740219116, 0.02081078477203846, -0.01214468851685524, -0.008134434930980206, 0.023985568434000015, 0.02281591109931469, 0.018395518884062767, 0.019079085439443588, -0.020066456869244576, -0.050614260137081146, -0.012091522105038166, -0.006638184655457735, -0.0011829488212242723, 0.007397702429443598, 0.01698281615972519, -0.0028197094798088074, 0.0017298015300184488, -0.020506978034973145, 0.004435583483427763, 0.0005658406880684197, 0.009919301606714725, 0.012349758297204971, 0.030365517362952232, 0.026233742013573647, 0.04630019888281822, 0.03414791449904442, 0.011347194202244282, 0.029028765857219696, -0.0015256812330335379, 0.0027494539972394705, 0.0026962878182530403, -0.02627931348979473, 0.026005886495113373, 0.02027912251651287, 0.011248457245528698, -0.02561093680560589, 0.008278743363916874, 0.016253678128123283, 0.07868603616952896, -0.001408905372954905, 0.03284154459834099, -0.004644450731575489, -0.011164910160005093, -0.011233266443014145, 0.024577993899583817, -0.02395518869161606, 0.013846008107066154, -0.03505933657288551, -0.004386214539408684, 0.028861673548817635, 0.013982720673084259, -0.05587012320756912, 0.01092945970594883, -0.009911705739796162, 0.01775752380490303, -0.00600398750975728, -0.035484667867422104, -0.0010338934371247888, 0.00200322805903852, 0.020127220079302788, -0.01231937762349844, -0.055930882692337036, 0.015038450248539448, -0.02673502266407013, 0.008992689661681652, -0.05103959143161774, 0.008719263598322868, -0.008514193817973137, -0.013709294609725475, 0.00500522181391716, -0.01453716866672039, -0.045297637581825256, -0.013040918856859207, 0.0412873812019825, 0.009463590569794178, -0.02254248596727848, 0.0054685273207724094, 0.007697712164372206, -0.012425709515810013, -0.02732744626700878, 0.023286812007427216, 0.022922243922948837, -0.0006603057263419032, 0.004731795284897089, 0.0024855216033756733, -0.003024779260158539, 0.01537263859063387, -0.03091237135231495, -0.012045950628817081, -0.021266495808959007, -0.024836229160428047, -0.03755055367946625, -0.017620811238884926, 0.027843918651342392, 0.0030399695970118046, -0.014127029106020927, 0.017803095281124115, 0.010595272295176983, -0.005001423880457878, 0.005407765973359346, 0.024274185299873352, 0.0004153612535446882, 0.012395328842103481, 0.015456185676157475, 0.032082028687000275, 0.008331908844411373, 0.024866608902812004, -0.033479541540145874, 0.008916737511754036, 0.008947118185460567, -0.006923004053533077, 0.011947213672101498, 0.015220735222101212, 0.009326877072453499, 0.013686508871614933, -0.02594512514770031, 0.007234406191855669, -0.013504224829375744, 0.038613881915807724, -0.014544764533638954, 0.03244659677147865, -0.0011525681475177407, -0.01838032901287079, -0.020856356248259544, -0.014954904094338417, 0.0023222253657877445, -0.009425614960491657, 0.01035982184112072, -0.006714136339724064, -0.0026279313024133444, 0.01626886986196041, -0.02037026546895504, -0.015949871391057968, -0.022314630448818207, 0.014430836774408817, -0.00010496772301848978, -0.018562613055109978, 0.04137852415442467, -0.012175069190561771, 0.010268679820001125, -0.028330011293292046, -0.0020241145975887775, 0.003621000563725829, -0.004329251125454903, -0.005065983161330223, 0.034087155014276505, -7.624845602549613e-05, -0.009326877072453499, -0.04611791670322418, -0.0033817526418715715, -0.007936960086226463, -0.0006455900729633868, 0.012137092649936676, -0.00012140415492467582, -0.03183898329734802, -0.01626886986196041, -0.011407955549657345, -0.02899838611483574, -0.01838032901287079, -0.007219215855002403, -4.57490750704892e-05, 0.004815342370420694, -0.022329820320010185, -0.009653470478951931, 0.016846103593707085, -0.005700180307030678, -0.008559764362871647, -0.020431026816368103, -0.019291749224066734, 0.009714231826364994, -0.0012645969400182366, -0.020142409950494766, -0.002806417876854539, -0.01898794248700142, 0.026233742013573647, -0.02134244702756405, -0.010435773059725761, 0.040163297206163406, 0.01838032901287079, -0.0038716415874660015, -0.006736922077834606, 0.007219215855002403, 0.0035735308192670345, -0.02489699050784111, -0.0037842970341444016, -0.034087155014276505, 0.008536978624761105, 0.009592708200216293, -0.0002598974679131061, -0.03039589896798134, -0.0035811259876936674, 0.01219025906175375, 0.004606474656611681, 0.01323079876601696, -0.03998101130127907, 0.04469002038240433, -0.010769961401820183, 0.0019633532501757145, -0.0002748504630289972, 0.004454571288079023, 0.02664388157427311, -0.0019177822396159172, 0.012387733906507492, 0.0025671699550002813, -0.023013386875391006, -0.020598120987415314, -0.005992594640702009, 0.0157523974776268, 0.0038203741423785686, 0.013671319000422955, -0.005859679076820612, -0.013678913936018944, -0.004496344830840826, -0.021722206845879555, -0.0014782113721594214, 0.004564701579511166, 0.006919206120073795, -0.03250735625624657, 0.039555683732032776, -0.026188170537352562, -0.06938953697681427, 0.007678723894059658, 0.02097787894308567, 0.010975031182169914, -0.0006498623406514525, -0.027813538908958435, 0.011749738827347755, -0.010207917541265488, 0.01358777191489935, -0.007576189003884792, -0.009630684740841389, 0.012782682664692402, 0.044811543077230453, 0.010131966322660446, 0.003269723616540432, 0.009402829222381115, -0.012600398622453213, -0.03518085926771164, 0.015205544419586658, -0.014757429249584675, 0.01705876737833023, 0.014240956865251064, 0.022952625527977943, -0.004268489312380552, -0.001107946503907442, 0.03755055367946625, -0.016603056341409683, 0.0009769296739250422, -0.010542105883359909, 0.028603436425328255, 0.011149720288813114, -0.01792461797595024, -0.009197759442031384, 0.02412228286266327, 0.00500522181391716, 0.0014297920279204845, -0.004929270129650831, -0.015691636130213737, -0.011461121961474419, -0.015691636130213737, 0.012068736366927624, 0.007185037713497877, -0.0030304756946861744, 0.014476407319307327, 0.0034159307833760977, 0.05626507103443146, 0.0014782113721594214, 0.025793220847845078, -0.008833191357553005, 0.029271813109517097, -0.012630779296159744, 0.013291560113430023, 0.020005695521831512, 0.010853508487343788, 0.027221115306019783, -0.019079085439443588, 0.015858730301260948, 0.019276559352874756, -0.0007253393996506929, -0.011468717828392982, 0.015813158825039864, 0.032264310866594315, 0.04241146892309189, 0.03864426165819168, -0.019428463652729988, 0.04271527752280235, -0.0178486667573452, 0.0076141650788486, -0.008020507171750069, 0.00018252160225529224, -0.021965252235531807, -0.00827114749699831, -0.003489983966574073, -0.0001274565584026277, 0.005100161302834749, 0.011407955549657345, 0.00624703336507082, 0.007496439851820469, 0.02410709112882614, 0.019215798005461693, -0.019428463652729988, -0.016299249604344368, -0.01705876737833023, 0.0013680813135579228, 0.02986423671245575, -0.009410424157977104, 0.009592708200216293, -0.007196430116891861, 0.01201556995511055, 0.01541820913553238, -0.023028576746582985, 0.013656128197908401, -0.0010490837739780545, 0.011537074111402035, 0.028330011293292046, 0.020871546119451523, -0.02778315730392933, -0.007750878110527992, -0.011787714436650276, -0.02202601358294487, -0.003393145278096199, -0.027828728780150414, -0.013694103807210922, 0.0005321371136233211, -0.039494920521974564, 0.008453432470560074, 0.008954714052379131, -0.02175258658826351, 0.00085967913037166, -0.016511915251612663, 0.0049748411402106285, 0.029757903888821602, 0.014689072035253048, 0.012121902778744698, 0.011073768138885498, 0.021281685680150986, -0.007758473511785269, 0.030152853578329086, -0.028861673548817635, 0.027737585827708244, 0.009030665270984173, 0.04830532521009445, 0.04189499840140343, -0.010731984861195087, 0.006212854757905006, 0.003949492238461971, -0.021433589980006218, 0.020415835082530975, -0.0033342826645821333, -0.006193866953253746, -0.009083831682801247, -0.033996012061834335, 0.009752207435667515, 0.019717080518603325, -0.026947688311338425, 0.012091522105038166, 0.007929365150630474, -0.014119434170424938, 0.009623088873922825, -0.007488844450563192, -0.018182853236794472, 0.02662869170308113, -0.034087155014276505, 1.7727024896885268e-06, -0.0016813823021948338, 0.008757239207625389, 0.010215513408184052, -0.013496629893779755, -0.022481724619865417, -0.03341877833008766, -0.007678723894059658, -0.014453621581196785, 0.014939713291823864, -0.0037368270568549633, 0.010154752060770988, -0.01681572198867798, -0.032598499208688736, 0.010162346996366978, -0.0094484006986022, -0.002876673359423876, 0.01882084831595421, -0.008400266058743, 0.023529859259724617, -0.039221495389938354, 0.0031159212812781334, 0.03797588497400284, -0.000209816760616377, 0.0016424570931121707, -0.02236020192503929, -0.01838032901287079, -0.004241906572133303, -0.014240956865251064, -0.0061027249321341515, -0.0022690591868013144, -0.018592992797493935, 0.012000380083918571, -0.025383081287145615, -0.008529383689165115, -0.029469287022948265, -0.043657079339027405, -0.01005601417273283, 0.006000190041959286, 0.02456280216574669, -0.01838032901287079, 0.021919680759310722, -0.018304375931620598, 0.01257761288434267, 0.014795404858887196, 0.0023545047733932734, -0.005187505856156349, 0.015349852852523327, 0.010823126882314682, 0.020962689071893692, 0.006915408652275801, 0.00277413846924901, 0.038705021142959595, 0.03527200222015381, -0.03524162247776985, 0.01757523976266384, -0.010488939471542835, 0.017803095281124115, 0.01687648333609104, -0.004872306250035763, -0.04074053093791008, 0.004982436075806618, -0.06264501810073853, -0.01836513727903366, -0.012691540643572807, 0.014370075426995754, 0.008453432470560074, -0.008643311448395252, -0.006592613644897938, 0.023712143301963806, 0.00448115449398756, -0.029879426583647728, -0.01035982184112072, -0.006505269091576338, 0.012243425473570824, -0.015965061262249947, -0.0006550840334966779, 0.004652046132832766, -0.013519415631890297, 0.03323649615049362, 0.0034254249185323715, 0.0174992885440588, -0.006858444772660732, 0.007428083103150129, 0.029226241633296013, -0.024502040818333626, 0.01882084831595421, 0.003157694824039936, 0.03314535319805145, -0.016162537038326263, -0.009949682280421257, -0.008780024945735931, -0.00687743304297328, 0.0018978448351845145, 0.012152283452451229, -0.003020981792360544, -0.007936960086226463, -0.0016899269539862871, 0.021099401637911797, -0.0005853033508174121, 0.028846481814980507, -0.011187695898115635, -0.028193296864628792, -0.021433589980006218, 0.03797588497400284, -0.005202696193009615, -0.00019106616673525423, 0.00047090096632018685, 0.02046140655875206, 0.01376246102154255, 0.0012864330783486366, -0.015509351156651974, 0.03448210284113884, 0.0009873730596154928, -0.03089717961847782, -0.01932213082909584, 0.03335801884531975, -0.02829962968826294, 0.00448115449398756, -0.018425898626446724, -0.0054191588424146175, 0.02204120345413685, -0.011081363074481487, -0.012076331302523613, -0.002903256332501769, -0.0012494066031649709, 0.02281591109931469, 0.024623563513159752, 0.023985568434000015, -0.0017649292713031173, 0.017119528725743294, -0.0025918541941791773, 0.010549700818955898, 0.023970378562808037, 0.0015807462623342872, 0.0002632203686516732, -0.027570493519306183, 0.01992974430322647, 0.0029089527670294046, 0.006596411112695932, 0.0006773948553018272, -0.0038526535499840975, 0.010086394846439362, 0.005905250087380409, 0.005134339910000563, -0.0005159973516128957, 0.01070920005440712, -0.017377765849232674, -0.01288142055273056, 0.010762365534901619, 0.011575049720704556, -0.03718598559498787, 0.03882654383778572, -0.003734928322955966, -0.0017468907171860337, 0.011499098502099514, -0.005833095870912075, -0.008468622341752052, 0.006417924538254738, -0.006797683425247669, 0.022846292704343796, 0.022071585059165955, 0.011726953089237213, -0.004446976352483034, -0.01882084831595421, -0.01698281615972519, 0.08925852179527283, -0.00021812399791087955, 0.011377574875950813, 0.01362574752420187, -0.002124750753864646, -0.00016377100837416947, -0.011400360614061356, 0.011149720288813114, 0.01058008149266243, -0.01958036608994007, 0.0257780309766531, 0.014810595661401749, -0.008985094726085663, 0.0010661729611456394, 0.027373017743229866, -0.029028765857219696, -0.013944745063781738, 0.037216369062662125, -0.002675401046872139, 0.019398082047700882, 0.012152283452451229, 0.014977688901126385, -0.00391911156475544, -0.01345865335315466, 0.0028216082137078047, -0.015076426789164543, 0.03776321932673454, 0.011537074111402035, -0.0031538973562419415, -0.012304186820983887, -0.017271433025598526, -0.019018324092030525, -0.008362289518117905, -0.007644545752555132, -0.022420963272452354, -0.00122946931514889, -0.018957562744617462, 0.004439380951225758, 0.012615589424967766, 0.009342067874968052, -0.0181980449706316, 0.028770530596375465, 0.0033475742675364017, -0.02515522576868534, -0.005020412150770426, -0.013162441551685333, -0.01414221990853548, 0.020947499200701714, 0.013732080347836018, -0.005760941654443741, -0.007010348606854677, -0.02114497311413288, -0.0006152093410491943, -0.024699516594409943, -0.018592992797493935, 0.003148200921714306, -0.012607993558049202, -0.004184942692518234, 0.02270958013832569, 0.017301812767982483, -0.014043482020497322, -0.00896990392357111, -0.008559764362871647, -0.021995631977915764, -0.009174973703920841, -0.009524351917207241, -0.013428272679448128, -0.008635716512799263, -0.0040824078023433685, 0.026431215927004814, -0.03284154459834099, -0.004086205270141363, 0.011514288373291492, -0.012213044799864292, 0.007485046982765198, 0.027388207614421844, 0.0049102818593382835, 0.01602582447230816, 0.007150859106332064, 0.017438527196645737, -0.00228424952365458, 0.00759517727419734, -0.011369979940354824, -0.004302667919546366, -0.0073901074938476086, -0.025033703073859215, 0.004720402415841818, 0.007371119223535061, -0.01967150904238224, 0.001319661969318986, -0.03177822008728981, 0.018699325621128082, -0.030076900497078896, 0.011901642195880413, -0.008088863454759121, -0.011780119501054287, 0.0382796935737133, -0.010982626117765903, 0.008088863454759121, 0.008802809752523899, 0.029013575986027718, -0.00255577708594501, 0.01775752380490303, -0.028330011293292046, -0.001745941350236535, -0.021312067285180092, -0.009425614960491657, 0.03083641827106476, -0.017362574115395546, 0.003930503968149424, -0.024243805557489395, -0.007211620453745127, 0.002403873484581709, 0.02204120345413685, 0.02166144549846649, 0.013109276071190834, 0.021129783242940903, -0.021388018503785133, -0.02097787894308567, 0.02125130593776703, -0.010124371387064457, 0.027129972353577614, 0.008529383689165115, -0.0078002470545470715, 0.02421342395246029, 0.020339883863925934, 0.005612835753709078, -0.010526915080845356, 0.008430646732449532, 0.014119434170424938, -0.025048894807696342, -0.0014620715519413352, 0.014871357008814812, -0.0010709199123084545, 0.009964872151613235, -0.01375486608594656, 0.018046140670776367, -0.026081837713718414, 0.006292604375630617, 0.028254058212041855, -0.008681287057697773, -0.006535649765282869, 0.021813347935676575, 0.011233266443014145, -0.0028216082137078047, 0.005795120261609554, -0.03666951507329941, 0.010428178124129772, 0.004834330175071955, 0.0018304376862943172, -0.0008891104371286929, -0.03232507407665253, -0.020506978034973145, -0.022527296096086502, 0.009919301606714725, 0.012866229750216007, -0.007378714624792337, -0.005939428694546223, 0.012129497714340687, -0.014909332618117332, 0.005126744508743286, -0.017711952328681946, 0.0015019462443888187, -0.00339504424482584, 0.020506978034973145, 0.02184372954070568, 0.012152283452451229, -0.012395328842103481, -0.0033456755336374044, -0.00840786099433899, 0.0066078039817512035, -0.024441279470920563, -0.014073862694203854, 0.0019244280410930514, 0.03587961569428444, 0.006387543864548206, 0.008362289518117905, -0.012509256601333618, 0.02236020192503929, -0.009790183044970036, -0.03423905745148659, -0.01593468151986599, -0.009228140115737915, 0.011909238062798977, 0.02854267507791519, 0.016572676599025726, -0.0039001235272735357, 0.009699041023850441, 0.012463685125112534, -0.00021171556727495044, -0.0019671509508043528, 0.025459034368395805, -0.006596411112695932, -0.014392860233783722, 0.0011364283272996545, -0.00900787953287363, -0.006448305211961269, -0.02360581047832966, 0.02401595003902912, -0.0031975696329027414, -0.01127883791923523, -0.010162346996366978, 0.0069040157832205296, 0.008878761902451515, -0.019960125908255577, -0.003357068169862032, -0.02778315730392933, 0.0020677868742495775, -0.0024171650875359774, -0.015000474639236927, 0.03153517469763756, 0.005791322328150272, -0.004891294054687023, -0.03575809299945831, 0.03335801884531975, -0.00223108334466815, -0.0022861482575535774, 0.0013899174518883228, 0.008726858533918858, 0.011286432854831219, 0.0036779644433408976, -0.007238203659653664, 0.019868982955813408, 0.001202886109240353, -0.036608751863241196, 0.0026735023129731417, 0.004993828944861889, -0.022588057443499565, -0.006543245166540146, -0.007990126498043537, 0.02517041750252247, 0.005331814289093018, 0.020522167906165123, -0.007731890305876732, -0.019261369481682777, -0.01079274620860815, -0.00421912083402276, 0.025383081287145615, -0.010093990713357925, 0.04222918301820755, -0.006660970393568277, 0.0013832716504111886, -0.00664198212325573, 0.006193866953253746, -0.02646159753203392, -8.930266631068662e-05, -0.006368556059896946, -0.007553403731435537, -0.015091616660356522, -0.005647014360874891, -0.029104718938469887, 0.013291560113430023, -0.035302381962537766, -0.007602772209793329, -0.006577423308044672, -0.019382892176508904, -0.01566125452518463, -0.015152378007769585, -0.0064862812869250774, -0.012737112119793892, -0.019443653523921967, 0.016056204214692116, 0.0001465631794417277, -0.010314250364899635, 0.004503939766436815, 0.011324409395456314, -0.03515047952532768, -0.010207917541265488, 0.024137472733855247, 0.002639323938637972, -0.02237539179623127, -0.02515522576868534, -0.009501566179096699, -0.01109655387699604, 0.006520459428429604, -0.001006360980682075, -0.00272097229026258, 0.0031444032210856676, -0.001634861808270216, 0.0027589481323957443, 0.004841925576329231, 0.011294028721749783, 0.012737112119793892, 0.01053451094776392, 0.01619291678071022, 0.014590335078537464, -0.021266495808959007, 0.019443653523921967, 0.030426278710365295, 0.014172600582242012, 0.000182165575097315, 0.017286622896790504, 0.007686319295316935, 0.028147725388407707, -0.0031425044871866703, -0.008818000555038452, -0.03414791449904442, 0.010147156193852425, -0.01637520082294941, -0.005369790364056826, -0.0017791702412068844, 0.026947688311338425, 0.042168423533439636, -0.009418019093573093, -0.02229944057762623, -0.018577802926301956, 0.00013884932559449226, 0.021129783242940903, -0.017894236370921135, -0.008901547640562057, -0.009539542719721794, -0.003169087693095207, -0.0014582739677280188, 0.006041963584721088, 0.009828158654272556, -0.01888160966336727, -0.017833475023508072, 0.020613310858607292, -0.023803284391760826, -0.013025728985667229, -0.002684895182028413, -0.019990505650639534, -0.01906389370560646, -0.0012266210978850722, -0.012683945707976818, 0.006049558520317078, -0.0031823792960494757, 0.001622519688680768, -0.02924143150448799, -0.020704451948404312, -0.03275040164589882, -0.009197759442031384, 0.0009256622288376093, -0.024577993899583817, 0.010101585648953915, 0.017651190981268883, 0.019170226529240608, -0.0074128927662968636, -0.03247697651386261, 0.017438527196645737, -0.01057248655706644, -0.015494161285459995, -0.03100351244211197, -0.02038545534014702, -0.007424285635352135, 0.015873920172452927, -0.010861103422939777, -0.010845912620425224, 0.014863761141896248, 0.013359916396439075, -0.028178106993436813, 0.017043577507138252, -0.03545428812503815, 0.0034406152553856373, 0.006455900613218546, 0.014438431710004807, 0.011020601727068424, -0.007891388610005379, 0.02447166107594967, 0.021114591509103775, 0.007526820525527, 0.012220639735460281, -0.03305421024560928, -0.007424285635352135, -0.01363334245979786, 0.01983860321342945, -0.006524256896227598, 0.0060381656512618065, 0.004473559092730284, -0.016694199293851852, 0.0008981297141872346, -0.0038089812733232975, -0.0017155606765300035, -0.007260989397764206, 0.011871261522173882, 0.0055672647431492805, -0.00826355256140232, 0.04125700145959854, -0.0012361151166260242, -0.017651190981268883, -0.017636001110076904, 0.014461217448115349, 0.01200797501951456, -0.005768537055701017, -0.026659071445465088, -0.03475553169846535, -0.012509256601333618, -0.011195290833711624, 0.03193012252449989, 0.015509351156651974, 0.02097787894308567, -0.011195290833711624, -0.02559574693441391, -0.021220924332737923, -0.0161017756909132, 0.024608373641967773, 0.02942371554672718, 0.0014013102045282722, -0.006566030438989401, 0.04839646816253662, -0.023180481046438217, -0.012805468402802944, 0.003148200921714306, 0.0069305989891290665, -4.082407758687623e-05, 0.006497674155980349, -0.0042001330293715, 0.007067312486469746, -0.02079559490084648, 0.01201556995511055, 0.009980062954127789, -0.03551504760980606, -0.023529859259724617, 0.0012085825437679887, -0.0011041489196941257, -0.012554828077554703, -0.01740814559161663, -0.0017991075292229652, 0.015600494109094143, -0.0089243333786726, 0.0054457420483231544, 0.02523117884993553, 0.024395707994699478, -0.0017544858856126666, 0.007306560408324003, -0.026355264708399773, 0.013739675283432007, -0.012858634814620018, -0.009630684740841389, 0.0001252017536899075, 0.0020943700801581144, 0.01096743531525135, 0.010139561258256435, -0.00827114749699831, 0.012851039879024029, -0.013291560113430023, -0.012509256601333618, 0.004697617143392563, 0.011073768138885498, 0.015114402398467064, -0.010147156193852425, -0.012129497714340687, -0.00840786099433899, 0.005290040746331215, -0.015114402398467064, -0.0012313680490478873, 0.003875439055263996, 0.016162537038326263, 0.0040254439227283, 0.001309218700043857, -0.0060381656512618065, -0.00448115449398756, 0.0032070635352283716, 0.01379284169524908, 0.017392955720424652, -4.346458808868192e-05, -0.021281685680150986, -0.0012664957903325558, 0.006763505283743143, -0.019185416400432587, 0.02037026546895504, -0.0016823316691443324, -0.008704072795808315, 0.018137283623218536, -0.011081363074481487, 0.0040444317273795605, 0.03256811946630478, -0.008886356838047504, 0.010853508487343788, -0.006000190041959286, 0.009645874612033367, 0.01180290523916483, 0.00502420961856842, 0.00598499970510602, -0.004781163763254881, 0.0015275799669325352, -0.005738156381994486, -0.013823222368955612, 0.008468622341752052, -0.012266211211681366, 0.010040824301540852, -0.005290040746331215, -0.017347384244203568, -0.00592423789203167, 0.019428463652729988, 0.03177822008728981, 0.007473654113709927, 0.0058027151972055435, 0.028679389506578445, -0.018243614584207535, -0.004371024202555418, 0.01602582447230816, 0.01502326037734747, -0.014218171127140522, 0.00858255010098219, -0.02239058166742325, -0.0044052028097212315, 0.010352225974202156, -0.007272382266819477, -0.011028196662664413, 0.0007091996376402676, -0.014172600582242012, -0.0007509731221944094, 0.025124846026301384, -0.002405772218480706, -0.004484951961785555, 0.009524351917207241, 0.0029754105489701033, -0.002367796376347542, -0.0010243995347991586, 0.002379189245402813, 0.00827114749699831, -0.03065413422882557, 0.016906864941120148, 0.01628405973315239, 0.010769961401820183, -0.022253869101405144, -0.01184088084846735, 0.011377574875950813, -0.0087496442720294, 0.012159878388047218, 0.006019177846610546, 0.023894427344202995, 0.01001044362783432, 0.017089148983359337, 0.007553403731435537, -0.0028292033821344376, 0.008567359298467636, 0.006600209046155214, 0.032264310866594315, -0.01661824807524681, 0.005028007086366415, 0.0022462736815214157, -0.006345770321786404, -0.010694009251892567, -0.013527010567486286, 0.011134529486298561, 0.004090002737939358, 0.002563372254371643, 0.0071166809648275375, 0.004424190614372492, -0.0027228710241615772, -0.011901642195880413, 0.01253963727504015, 0.005806512665003538, 0.009205354377627373, -0.000719168339855969, 0.002172220731154084, 0.017377765849232674, -0.022086774930357933, 0.01061805710196495, 0.0010557295754551888, 0.0025481819175183773, -0.002544384216889739, -0.012106711976230145, 0.008020507171750069, 0.01288142055273056, -0.011856071650981903, 0.016679009422659874, 0.007997721433639526, -0.014924522489309311, -0.008043292909860611, 0.051677584648132324, -0.01593468151986599, -0.0032678248826414347, -0.006436912342905998, 0.004268489312380552, -0.011407955549657345, -0.008514193817973137, -0.006919206120073795, -0.0014326402451843023, 0.012380138970911503, 6.770388426957652e-05, -0.03232507407665253, -0.008977498859167099, 0.008780024945735931, -0.013253583572804928, -0.006436912342905998, -0.003717839252203703, -0.0044014048762619495, 0.010823126882314682, -0.01950441487133503, -0.014628310687839985, 0.012463685125112534, 0.006960979662835598, 0.021296877413988113, 0.004667236469686031, 0.00827114749699831, -0.014066267758607864, -0.015949871391057968, -0.007006550673395395, 0.007868603803217411, -0.01593468151986599, -0.026172980666160583, 0.008855976164340973, -0.000833095982670784, -0.0031273141503334045, -0.018243614584207535, -0.01345865335315466, 0.007568594068288803, 0.008772429078817368, -0.0015028956113383174, -0.0010111079318448901, 0.001840881071984768, 0.007185037713497877, -0.004902686923742294, 0.0018836038652807474, 0.011149720288813114, -0.020613310858607292, -0.014658691361546516, -0.021281685680150986, -0.0030152853578329086, 0.016998006030917168, 0.017089148983359337, 0.0008340454078279436, -0.01257761288434267, 0.0026962878182530403, 0.0023469096049666405, -0.0004457419563550502, 0.02160068415105343, 0.0036760657094419003, 0.00949397124350071, 0.005939428694546223, 0.019003132358193398, -0.03135289251804352, -0.011757333762943745, 0.009023070335388184, 0.016344821080565453, -0.007504034787416458, 0.002474128967151046, -0.0008796164183877409, -0.007777461316436529, -0.009212949313223362, -0.0036779644433408976, -0.011187695898115635, 0.0031633912585675716, -0.008164815604686737, -0.010595272295176983, 0.005472325254231691, 0.010610462166368961, 0.004690021742135286, -0.011711763218045235, 0.0012769391760230064, -0.006216652225703001, -0.01803095079958439, 0.005282445810735226, -0.009774993173778057, -0.012235830537974834, -0.0011867464054375887, -0.01844109036028385, -0.009402829222381115, -0.006167283747345209, -0.0010177537333220243, -0.001499098027125001, 0.003470995929092169, -0.013170037418603897, -0.002641222905367613, 0.0003273046750109643, 0.013185227289795876, -0.005848286207765341, -0.00753441546112299, -0.013739675283432007, 0.007671128958463669, 0.017195479944348335, 0.0215854924172163, 0.013595366850495338, 0.005658406764268875, 0.003121617715805769, 0.019625937566161156, 0.021524731069803238, 0.011468717828392982, -0.009630684740841389, -0.017879046499729156, -0.008491408079862595, -0.005100161302834749, 0.002193107269704342, 0.03560619056224823, 0.0027095794212073088, -0.0016851798864081502, 0.026081837713718414, -0.02517041750252247, -0.005138137377798557, 0.02169182524085045, -0.013185227289795876, -0.01436247956007719, -0.011020601727068424, 0.011688977479934692, -0.025140035897493362, 0.015137188136577606, 0.02071964368224144, -0.013375107198953629, -0.005821703001856804, 0.001578847412019968, 0.024076711386442184, 0.009243330918252468, -0.01950441487133503, 0.004811544436961412, -0.011932022869586945, 0.0031026299111545086, -0.010671223513782024, 0.016830911859869957, -0.0032735213171690702, 0.010458558797836304, -0.004686224274337292, 0.021388018503785133, 0.01898794248700142, 0.005126744508743286, -0.008172410540282726, 0.013504224829375744, 0.010337036103010178, -0.0034652994945645332, -0.0006337225786410272, 0.010602867230772972, 0.009714231826364994, -0.010678819380700588, -0.006026772782206535, -0.011438336223363876, 0.015152378007769585, 0.006395139265805483, -0.0017089148750528693, 0.014453621581196785, 0.007967340759932995, -0.013511819764971733, -0.004086205270141363, -0.009600304067134857, -0.010777556337416172, -0.00014217222633305937, -0.012448495253920555, -0.005282445810735226, -0.006979967933148146, 0.016329631209373474, -0.0013234595535323024, -0.03697332367300987, -0.016694199293851852, -3.245751577196643e-05, 0.00021824266877956688, -0.020051266998052597, -2.18509685510071e-05, 0.0002836323983501643, 0.01200797501951456, -0.014058672823011875, -0.017742333933711052, -0.012038355693221092, 0.00391911156475544, 0.0072647868655622005, 0.01740814559161663, -0.009061045944690704, -0.007576189003884792, -0.0011554163647815585, 0.013656128197908401, 0.022496914491057396, -0.031033894047141075, 0.008301528170704842, 0.0011411753948777914, 0.03457324579358101, -0.00792176928371191, 0.012311781756579876, 0.009364853613078594, -0.01872970722615719, 0.0009209152194671333, 0.02229944057762623, 0.00762176001444459, 0.017256243154406548, -0.006828064098954201, -0.010937054641544819, 0.0038336655125021935, 0.005612835753709078, 0.008430646732449532, -0.02038545534014702, 0.010283869691193104, 0.011696572415530682, 0.017089148983359337, -0.00379758863709867, 0.009463590569794178, -0.004834330175071955, -0.011810500174760818, -0.022739959880709648, -0.01235735323280096, -0.002682996215298772, -0.011810500174760818, -0.011985189281404018, -0.03265926241874695, -0.009524351917207241, -0.00552928913384676, -0.005498907994478941, 0.002538688015192747, 0.03463400900363922, 0.020309504121541977, 0.009873730130493641, 0.004777366295456886, -0.004314060788601637, 0.012030760757625103, 0.02559574693441391, 0.0022899459581822157, -0.020233551040291786, -0.029727522283792496, 0.0019633532501757145, -0.0011734548024833202, 0.003332383930683136, -0.0019633532501757145, -0.01271432638168335, -0.00788379367440939, 0.008734453469514847, 0.00019616918871179223, 0.00029241430456750095, 0.0010690211784094572, -0.020157599821686745, -0.018182853236794472, 0.019990505650639534, 0.01567644625902176, 0.011483907699584961, -0.004371024202555418, -0.016830911859869957, -0.0023393144365400076, -0.030927561223506927, -0.01053451094776392, -0.014757429249584675, 0.013094085268676281, -0.008157219737768173, -0.02011202834546566, 0.0075609986670315266, -0.009699041023850441, -0.0025045096408575773, 0.0031595935579389334, -0.0017117629759013653, 0.005191303323954344, 0.013306749984622002, 0.015782777220010757, 0.01219025906175375, -0.004014051053673029, -0.008248362690210342, 0.0062242476269602776, 0.013329535722732544, 0.014446026645600796, -0.010261083953082561, -0.006885027978569269, -0.008499003015458584, 0.003005791222676635, -0.00201272196136415, -0.011210481636226177, 0.016557486727833748, -0.012418114580214024, 0.013694103807210922, 0.00672932667657733, 0.0090990224853158, -0.004796354100108147, -0.021266495808959007, -0.01305610965937376, 0.0005330864805728197, 0.003987467847764492, -0.003070350270718336, 0.015327067114412785, -0.02524636872112751, -0.035909995436668396, 0.009569923393428326, -0.018258806318044662, 0.005734358914196491, -0.045996394008398056, -0.01541061419993639, 0.01202316489070654, 0.032780785113573074, -0.008893952704966068, 0.011066173203289509, 0.013808031566441059, 0.01467388216406107, -0.010139561258256435, -0.025140035897493362, -1.90324462892022e-05, 0.021782968193292618, -0.0075002373196184635, 0.007454666309058666, 0.001183898188173771, 0.0012883319286629558, -0.004450773820281029, 0.00017468907753936946, -0.005624228622764349, -0.018866419792175293, 0.020750023424625397, -0.006774898152798414, -0.008468622341752052, -0.005191303323954344, -0.020780405029654503, 0.0040254439227283, -0.013640938326716423, 0.020066456869244576, 0.0027057817205786705, 0.0025481819175183773, 0.003041868330910802, -0.007659736089408398, 0.027054021134972572, -0.024274185299873352, 0.0016054305015131831, -0.0015038450947031379, 0.012106711976230145, -0.005020412150770426, 0.00918256863951683, 0.004967245738953352, -0.01941327191889286, 0.006786290556192398, 0.010640842840075493, 0.003960884641855955, -0.012076331302523613, 0.0035545427817851305, 0.01144593209028244, 0.030243994668126106, 0.008802809752523899, -0.010466153733432293, 0.01058008149266243, -0.027281876653432846, -0.003262128448113799, 0.019443653523921967, 0.0035127694718539715, 0.006649577524513006, -0.02592993527650833, 0.008331908844411373, 0.02726668491959572, -0.009304092265665531, -0.020522167906165123, 0.026856545358896255, -0.003953289706259966, -0.01074717566370964, 0.01897275261580944, -0.021448779851198196, 0.0048874965868890285, 0.020947499200701714, -0.002352606039494276, 0.011924427933990955, -0.0002855311904568225, 0.005255862604826689, 0.002405772218480706, -0.0005995442625135183, -0.002310832729563117, -0.01235735323280096, -0.007340738549828529, -0.014947308227419853, 0.007823032326996326, -0.015266305766999722, -0.0003391721402294934, -0.004735592752695084, -0.01236494816839695, -0.01453716866672039, -0.0064217220060527325, 0.01626886986196041, 0.018258806318044662, 0.004876103717833757, -0.006520459428429604, -0.00391911156475544, 0.00948637630790472, -0.00517991092056036, -0.0008554068044759333, -0.004731795284897089, -0.004090002737939358, 0.0019462641794234514, 0.0027494539972394705, -0.0036266969982534647, 0.016162537038326263, -0.003227950306609273, 0.012220639735460281, -0.00014668185031041503, 0.003191873198375106, 0.0014487800654023886, 0.003829868044704199, 0.018957562744617462, 0.007188835181295872, 0.0037501186598092318, 0.011073768138885498, 0.010040824301540852, -0.008931928314268589, -0.040862053632736206, -0.01127883791923523, -0.02384885586798191, 0.006360960658639669, -0.007014146074652672, 0.01179531030356884, -0.0017715750727802515, 0.000940852565690875, 0.016086585819721222, 0.006957182195037603, 0.018304375931620598, 0.0026545142754912376, -0.008073673583567142, -0.00589765515178442, 0.026613499969244003, -0.009805373847484589, 0.012425709515810013, -0.02185891941189766, 0.017742333933711052, 0.0171499103307724, -0.02778315730392933, -0.012152283452451229, -0.0025159025099128485, -0.010253489017486572, -0.002141839824616909, 0.01994493417441845, -0.01681572198867798, -0.0037767018657177687, -0.009121807292103767, 0.021023450419306755, -0.006000190041959286, -0.00448115449398756, 0.005305231083184481, -0.007944555021822453, -0.004819139838218689, 0.0023298205342143774, 0.003227950306609273, 0.009721826761960983, -0.01687648333609104, 0.007469856645911932, 0.006767302751541138, -0.008871166966855526, 0.002476027701050043, -0.018942371010780334, 0.006087534595280886, -0.019170226529240608, 0.014825785532593727, 0.007743283174932003, 0.007128073833882809, 0.011263648048043251, 0.0063989367336034775, -0.009045856073498726, 0.020248742774128914, 0.006793885957449675, 0.015889110043644905, 0.0004813443520106375, -0.010853508487343788, 0.0021285484544932842, 0.010944650508463383, -0.00913699809461832, -0.0034539068583399057, -0.0016073293518275023, -0.007090097758919001, 0.0041051930747926235, 0.0005345105892047286, -0.008536978624761105, -0.02471470646560192, 0.013656128197908401, -0.0047242003493011, 0.0016377100255340338, 0.009812968783080578, -0.011164910160005093, -0.011590240523219109, -0.0028443937189877033, 0.005096363835036755, 0.03594037890434265, 0.016056204214692116, 0.0039039209950715303, 0.012911801226437092, 0.004165954422205687, -0.004678628873080015, 0.013048513792455196, -0.017787905409932137, 0.000356498610926792, 0.014377670362591743, -0.009471185505390167, -0.0033361816313117743, -0.003700749948620796, 0.01958036608994007, -0.0013766258489340544, -0.027221115306019783, -0.010952245444059372, 0.005009019281715155, -0.01092945970594883, 0.007405297830700874, -0.017529668286442757, -0.011476312763988972, 0.012433304451406002, 0.021023450419306755, -0.007131871301680803, -0.001692775054834783, -0.00231842789798975, -0.01742333546280861, 0.0020772810094058514, 0.005783727392554283, -0.005335611756891012, 0.014392860233783722, 0.010245894081890583, 0.036608751863241196, 0.00528624327853322, -0.019033513963222504, 0.01341308280825615, 0.0036760657094419003, 0.005009019281715155, -0.01801575906574726, -0.007215418387204409, 0.006045761052519083, -0.00037073958083055913, -0.004777366295456886, -0.002088673645630479, 0.01880565844476223, 0.0004727997584268451, 0.02037026546895504, -0.006569828372448683, 0.0008805658435449004, 0.015722015872597694, 0.003767207730561495, -0.00809645839035511, -0.00044052026350982487, 0.011240862309932709, 0.0008473369525745511, -0.01793980784714222, 0.0002572866214904934, -0.0027418588288128376, -0.015030855312943459, -0.0012313680490478873, 0.006600209046155214, 0.0006560334004461765, 0.006152093410491943, 0.019398082047700882, 0.0073901074938476086, 0.011126934550702572, -0.02272477000951767, -0.004014051053673029, 0.0011164910392835736, 0.007511630188673735, 0.011681382544338703, 0.004895091522485018, 0.010800342075526714, -0.0018741099629551172, 0.008947118185460567, 0.015220735222101212, -0.0038431596476584673, -0.0036532802041620016, 0.016846103593707085, 0.006941991858184338, 0.019808221608400345, 0.0052330768667161465, -0.006482483819127083, -0.014278932474553585, -0.0060153803788125515, -0.022846292704343796, 0.014127029106020927, -0.023803284391760826, -0.0009550935355946422, 0.0013395993737503886, -0.01836513727903366, 0.0069078137166798115, 0.01792461797595024, -0.020522167906165123, 0.005962213966995478, 0.0014041583053767681, 0.02195006236433983, -0.009076236747205257, -0.0073521314188838005, -0.010610462166368961, 0.00519889872521162, 0.02011202834546566, -0.02088673785328865, -0.0028728756587952375, -7.945267134346068e-05, -0.000648912915494293, 0.017651190981268883, 0.01882084831595421, 0.025990696623921394, -0.0020848761778324842, 0.006566030438989401, 0.014506787993013859, 0.03350992128252983, -0.01619291678071022, 0.029909808188676834, -0.000627076777163893, 0.0008145827450789511, 0.0028026204090565443, -0.02925662137567997, 0.002639323938637972, -0.0012456090189516544, -0.021296877413988113, -0.008423051796853542, 0.008369885385036469, -0.007967340759932995, -9.212119039148092e-06, -0.009812968783080578, -0.022238679230213165, -0.006030570715665817, 0.006436912342905998, 0.013344726525247097, 0.0015968859661370516, -0.004295072518289089, 0.012000380083918571, -0.0073141553439199924, 0.005650811828672886, -0.014749834313988686, -0.0322035513818264, 0.012448495253920555, -0.014559954404830933, 0.010071204975247383, -0.00681667122989893, -0.028952814638614655, -0.0157523974776268, 0.005377385299652815, -0.006945789325982332, -0.011088958941400051, -0.014924522489309311, -0.016937244683504105, 0.0033817526418715715, 0.004359631799161434, 0.0035735308192670345, -0.0012579512549564242, 0.01344346348196268, -0.006239437963813543, 0.0045760939829051495, 0.013420677743852139, -0.01162062119692564, -0.003655178938060999, 0.008187600411474705, -0.010261083953082561, 0.0031007309444248676, 0.004921674728393555, -0.02071964368224144, 0.008073673583567142, -0.018486661836504936, 0.0014297920279204845, 0.002538688015192747, 0.02219310775399208, -0.012843444012105465, 0.009737016633152962, 0.006706541404128075, -0.0023374157026410103, 0.018091712146997452, 0.0006826165481470525, -0.0075002373196184635, -0.003619101829826832, 0.0027171745896339417, -0.015349852852523327, -0.005248267203569412, 0.0174992885440588, -0.03475553169846535, 0.012797873467206955, -0.006216652225703001, 0.01670938916504383, 0.021205734461545944, -0.013557391241192818, -0.0013747269986197352, -0.03082122839987278, -0.009433209896087646, 0.006774898152798414, 0.015737207606434822, 0.015706826001405716, 0.007302762940526009, -0.007162251975387335, -0.002536789048463106, 0.021433589980006218, 0.004302667919546366, -0.004671033937484026, 0.011286432854831219, -0.006174879148602486, -0.0005644165794365108, -0.010701604187488556, -0.0067824930883944035, 0.008992689661681652, -0.017438527196645737, 0.006068546324968338, -0.013990316540002823, -0.014385265298187733, -0.002001329092308879, 0.014134624972939491, 0.011939618736505508, 0.004803949501365423, 0.00047730939695611596, -0.011688977479934692, 0.008711667731404305, 0.011567454785108566, -0.005498907994478941, 0.0049748411402106285, -0.029484476894140244, -0.008673692122101784, 0.026689453050494194, -0.004659641068428755, -0.005347004625946283, 0.0003961359616369009, 0.0036494825035333633, 0.001312066800892353, 0.02855786494910717, 0.019306940957903862, 0.008552169427275658, 0.05304471775889397, 0.015889110043644905, -0.010633247904479504, -0.005081173498183489, 0.0002791227598208934, 0.0026564132422208786, 0.010944650508463383, 0.0059508210979402065, 0.021965252235531807, 0.012547232210636139, 0.007815437391400337, 0.004538118373602629, 0.017347384244203568, -0.015478970482945442, -0.018167663365602493, 0.0028235071804374456, 0.013116871006786823, -0.0019918351899832487, 0.007321750745177269, 0.014096648432314396, -0.01018513273447752, -0.009904110804200172, 0.016694199293851852, -0.010595272295176983, -0.01005601417273283, -0.008977498859167099, -0.00483053270727396, -0.014757429249584675, 0.010610462166368961, 0.011597835458815098, 0.0073141553439199924, 0.002544384216889739, -0.032082028687000275, 0.012782682664692402, -0.0062508308328688145, -0.006136903073638678, -0.011878857389092445, 0.01544858980923891, -0.005039399955421686, -0.007629355415701866, -0.004055824596434832, -0.009797777980566025, 0.005703977774828672, -0.008704072795808315, -0.002508307108655572, 0.00983575452119112, 0.005981201771646738, -0.002367796376347542, -0.007967340759932995, 0.007724294904619455, -0.008423051796853542, 0.01271432638168335, -0.009266115725040436, -0.018349947407841682, -0.0028728756587952375, 0.002962118946015835, -0.012213044799864292, -0.007074907422065735, -0.00228424952365458, -0.010952245444059372, 0.006053355988115072, -0.009053451009094715, 0.011514288373291492, 0.019352510571479797, 0.009258520789444447, -0.0039001235272735357, 0.0007324598846025765, -0.00792176928371191, 0.011180100962519646, -0.01558530330657959, -0.0054457420483231544, 0.008719263598322868, 0.0003204215317964554, -0.011673787608742714, 0.008590145036578178, 0.009387638419866562, -0.0019481629133224487, -0.03463400900363922, 0.016208108514547348, -0.020248742774128914, -0.0033893478102982044, -0.0012655464233830571, -0.0056546092964708805, 0.011947213672101498, 0.00826355256140232, 0.012167473323643208, 0.007355928886681795, 0.013344726525247097, -0.004260894376784563, -0.0051229470409452915, 0.01810690201818943, 0.0069040157832205296, -0.004564701579511166, -0.005453336983919144, -0.0011601633159443736, 0.036699894815683365, 0.01023070327937603, 0.005422956310212612, -0.013352321460843086, 0.0025652709882706404, 0.0012380138505250216, -0.012592803686857224, 0.0031102250795811415, -0.005438146647065878, 0.008134434930980206, -0.014780214987695217, -0.01092945970594883, 0.013640938326716423, 0.0023924808483570814, 0.0001991360477404669, 0.004033038858324289, -0.0016130257863551378, -0.01271432638168335, -0.026932498440146446, -0.01663343794643879, 0.013732080347836018, -0.014932118356227875, 0.00509256636723876, -0.02760087326169014, 0.00949397124350071, -0.0026450203731656075, -0.019990505650639534, 0.005290040746331215, 0.004705212078988552, -0.007238203659653664, -0.01509921159595251, 0.01035982184112072, -0.01001044362783432, -0.0029792082495987415, -0.011050982400774956, -0.03168707713484764, -0.02088673785328865, 0.010033228434622288, 0.011742143891751766, -0.0007091996376402676, -0.03013766184449196, 0.008673692122101784, 0.0009432260412722826, -0.01733219437301159, 0.015691636130213737, -0.00421912083402276, -0.004226716235280037, -0.0038013861048966646, -0.005434349179267883, 0.008719263598322868, -0.00456849904730916, 0.01897275261580944, -0.0058900597505271435, 0.02743377909064293, 0.014871357008814812, -0.022223487496376038, 0.004116585943847895, 0.013777650892734528, 0.01992974430322647, -0.0050583877600729465, 0.0038963258266448975, 0.003560239216312766, -0.02009683847427368, -0.001692775054834783, -0.004496344830840826, 0.0036001140251755714, 0.008544574491679668, -0.02541346289217472, -0.01654229499399662, -0.0056318240240216255, 0.009683851152658463, -0.029469287022948265, -0.004674831405282021, -0.007238203659653664, 0.005324219353497028, 0.006676160730421543, -0.002755150431767106, 0.010868698358535767, 0.0019918351899832487, -0.0045533087104558945, -0.009121807292103767, -0.0006308744195848703, -0.012873824685811996, -0.005582455080002546, -0.007340738549828529, -0.002474128967151046, -0.0062508308328688145, 0.020339883863925934, -0.007321750745177269, 0.00689642084762454, -0.0033912465441972017, -0.01672457903623581, -0.021555112674832344, 0.022952625527977943, -0.023636190220713615, -0.002037406200543046, 0.008134434930980206, -0.012813063338398933, 0.003096933476626873, -0.0001596173970028758, -0.006945789325982332, 0.00195955578237772, -0.011559859849512577, -0.004234311170876026, 0.022481724619865417, -0.02114497311413288, 0.016329631209373474, -0.007511630188673735, 0.014172600582242012, -0.013564986176788807, -0.0076483432203531265, -0.03054780140519142, 0.0035450488794595003, -0.011218076571822166, 0.01827399618923664, -0.005886262282729149, -0.026248931884765625, 0.021220924332737923, 0.0018105002818629146, 0.0005307130049914122, -0.029712332412600517, -0.001513338997028768, 0.0018617677269503474, -0.014689072035253048, 0.01035982184112072, -0.010344631038606167, 0.022436153143644333, 0.03475553169846535, -0.021023450419306755, 0.017377765849232674, -0.011605430394411087, 0.007074907422065735, -0.008567359298467636, -0.011263648048043251, -0.004355833865702152, -0.00988132506608963, 0.013071299530565739, 0.01253204233944416, -0.005403968505561352, -0.02176777832210064, -0.02907433733344078, 0.007747080642729998, 0.009342067874968052, 0.0005378334899432957, 0.010382606647908688, 0.026583120226860046, 0.00163201370742172, 0.011195290833711624, 0.005753346718847752, -0.018091712146997452, -0.004173549823462963, 0.0005098262336105108, -0.013010538183152676, 0.011704168282449245, -0.01363334245979786, 0.015615683980286121, -0.02132725715637207, -0.02655273862183094, 0.022071585059165955, 0.0049102818593382835, 0.008385075256228447, 0.009410424157977104, -0.00918256863951683, 0.0007709104684181511, 0.005225481931120157, -0.016208108514547348, -0.01637520082294941, -0.007249596528708935, -0.017210671678185463, -0.008559764362871647, -0.011088958941400051, -0.0036874585784971714, 0.009243330918252468, 0.007750878110527992, 0.0045267255045473576, -0.006406531669199467, -0.006748314946889877, -0.016253678128123283, 0.012828254140913486, -0.0004716130206361413, 0.022648818790912628, 0.003839361947029829, 0.02734263800084591, 0.028937624767422676, 0.003846957115456462, -0.01619291678071022, -0.002626032568514347, -0.0005259659956209362, 0.0002826830022968352, 0.00037952151615172625, -0.007363524287939072, 0.00826355256140232, 0.002903256332501769, -0.013549796305596828, -0.0015835943631827831, 0.008893952704966068, -0.004276084713637829, 0.0008929080213420093, -0.022876672446727753, 0.009372448548674583, 0.009174973703920841, -0.011172505095601082, 0.023377954959869385, 0.006019177846610546, -0.009888920933008194, -0.01913984678685665, 0.016496725380420685, -0.007853413000702858, 0.0038811354897916317, 0.006463495548814535, 0.003831766778603196, 0.02134244702756405, 0.006566030438989401, 0.0014364378293976188, -0.011461121961474419, 0.011088958941400051, 0.016663817688822746, 0.009615493938326836, -0.005662204697728157, -0.0027570491656661034, 0.023621000349521637, -0.020263932645320892, -0.0012361151166260242, -0.0024266589898616076, 0.009395234286785126, -0.0024855216033756733, -0.019109465181827545, -0.004697617143392563, 0.015965061262249947, 0.00456849904730916, 0.00983575452119112, -0.007724294904619455, 0.0037842970341444016, 0.012911801226437092, -0.017013195902109146, 0.01358777191489935, 0.0017668281216174364, -0.015737207606434822, -0.018213234841823578, -0.003848856082186103, 0.0046140700578689575, -0.009114212356507778, -0.022071585059165955, -0.0032773190177977085, 0.004416595678776503, 0.014453621581196785, 0.017620811238884926, -0.008947118185460567, 0.008878761902451515, 0.03350992128252983, -0.019079085439443588, 0.02671983279287815, -0.0029640179127454758, 0.010033228434622288, -0.007306560408324003, -0.016329631209373474, 0.0018731605960056186, -0.014438431710004807, -0.025534985587000847, 0.01396753080189228, 0.02629450336098671, -0.003941896837204695, 0.013390297070145607, 0.014453621581196785, 0.011871261522173882, 0.0028045191429555416, 0.010071204975247383, 0.006653374992311001, 0.03293268755078316, -0.014939713291823864, 0.013306749984622002, -0.00788379367440939, -0.001630114857107401, 0.010982626117765903, -0.0035203646402806044, -0.01376246102154255, 0.027190733700990677, -0.016405582427978516, -0.0018399315886199474, -0.002897560130804777, 0.0033285864628851414, 0.010420583188533783, -0.02079559490084648, -0.01201556995511055, -0.00953194685280323, -0.007101490627974272, 0.039312634617090225, -0.006064748857170343, -0.0014202981255948544, -0.008559764362871647, -0.006114117335528135, -0.025641318410634995, 0.002684895182028413, -0.010428178124129772, -0.007807841990143061, 9.054875408764929e-05, 0.006638184655457735, 0.0007400550530292094, -0.017620811238884926, -0.0226336270570755, -0.01376246102154255, -0.012410519644618034, -0.02018798142671585, -0.01731700450181961, 0.010861103422939777, 0.014020697213709354, 0.014233361929655075, 0.002946928609162569, -0.02795025147497654, 0.005525491200387478, -0.016132155433297157, -0.0085749551653862, -0.01536504365503788, 0.017134718596935272, 0.0016557485796511173, 0.007325548212975264, -0.03976834565401077, -0.0030646538361907005, -0.005973606836050749, -0.0078761987388134, -0.004671033937484026, 0.0018038545968011022, -0.012106711976230145, 0.0037767018657177687, 0.0005872021429240704, -0.011818095110356808, 0.0039343019016087055, 0.006972372531890869, -0.011742143891751766, 0.0072268107905983925, -0.018562613055109978, -0.025793220847845078, 0.0034690971951931715, 0.015889110043644905, -0.013420677743852139, -0.003786195768043399, 0.0007870502304285765, -0.010526915080845356, -0.016663817688822746, -0.005183708388358355, 0.006022975314408541, -0.024775467813014984, -0.02421342395246029, 0.009896515868604183, -0.02002088725566864, 0.007792651653289795, 0.02793506160378456, -0.0014839076902717352, 0.005411563441157341, 0.012478875927627087, 0.006938194390386343, -0.0006546093500219285, 0.0016358112916350365, 0.00025277698296122253, 0.010428178124129772, -0.010086394846439362, -0.004853317979723215, 0.01362574752420187, -0.02176777832210064, 0.01601063273847103, -0.005248267203569412, -0.016481533646583557, -0.005411563441157341, 0.009061045944690704, 0.01005601417273283, 0.004424190614372492, 0.019048703834414482, -0.030061710625886917, 0.009790183044970036, 0.013846008107066154, -0.005711573176085949, -0.007982530631124973, -0.01366372313350439, 0.010633247904479504, -0.009212949313223362, 0.007762270979583263, 0.01558530330657959, -0.008726858533918858, -0.004959650803357363, -0.012797873467206955, -0.019306940957903862, 0.004769771359860897, 0.005240672267973423, -0.009288901463150978, 0.003507073037326336, -0.02210196480154991, -0.012167473323643208, 0.02125130593776703, 0.005878666881471872, -0.015258710831403732, 0.015501756221055984, -0.02592993527650833, -0.02664388157427311, -0.017560049891471863, 0.016041014343500137, 0.016922054812312126, 0.008658502250909805, -0.018349947407841682, -0.021479161456227303, 0.015722015872597694, 0.012380138970911503, 0.0031633912585675716, 0.05371309071779251, 0.0004884648369625211, -4.657979661715217e-05, -0.014575145207345486, 0.0026450203731656075, -0.001096553634852171, 0.0014174499083310366, -0.01932213082909584, -0.003262128448113799, -0.014514382928609848, 0.0015940377488732338, -0.001965251984074712, -0.021919680759310722, 0.0016139751533046365, -0.008893952704966068, 0.028223678469657898, 0.0014250450767576694, -0.007279977202415466, -0.00988132506608963, -0.013701699674129486, -0.021782968193292618, -0.005628026090562344, 0.010838317684829235, -0.011651001870632172, 0.019914554432034492, 0.01879046857357025, -0.009752207435667515, -0.002563372254371643, -0.000522643094882369, -0.015319472178816795, 0.014651096425950527, -0.012782682664692402, 0.00138232228346169, -0.0018342352705076337, 0.0020506978034973145, 0.0006356213707476854, -0.04441659525036812, -0.006281211506575346, 0.013109276071190834, -0.016314439475536346, 0.006030570715665817, -0.017453717067837715, 0.005301433615386486, -0.0017326497472822666, 0.0010614260099828243, -0.014529573731124401, 0.0001401547488057986, -0.009220545180141926, -0.015182758681476116, -0.030593372881412506, -0.007553403731435537, 0.004648248199373484, -0.019048703834414482, 0.019534794613718987, -0.020825974643230438, 0.00792176928371191, 0.00349188270047307, 0.010071204975247383, -0.002132345922291279, 0.008278743363916874, 0.02204120345413685, 0.0007970188744366169, -0.018228424713015556, 0.015030855312943459, 0.00010728187771746889, 0.018046140670776367, -0.010724389925599098, -0.002833001082763076, 0.01366372313350439, 0.0038792367558926344, 0.014727048575878143, 0.0029943985864520073, 0.0022500711493194103, 0.009896515868604183, -0.002897560130804777, 0.0021038639824837446, -0.01271432638168335, -0.005392575636506081, 0.017818285152316093, -0.0044052028097212315, 0.00984334945678711, 0.011909238062798977, -0.010481344535946846, -0.006539447233080864, -0.01184088084846735, -0.016557486727833748, -0.009767397306859493, -0.011681382544338703, 0.011992784217000008, -0.008772429078817368, 0.009258520789444447, 0.007298965007066727, 0.007131871301680803, -0.02090192772448063, 0.013549796305596828, 0.0034425139892846346, -0.022071585059165955, -0.0076255579479038715, 0.007386309560388327, 0.0013101680669933558, 0.007724294904619455, 0.01127124298363924, -0.01231937762349844, -0.012790278531610966, 0.02307414822280407, -0.009759802371263504, 0.013306749984622002, 0.005666002165526152, 0.006923004053533077, -0.002937434706836939, -0.009288901463150978, -0.01626886986196041, -0.007891388610005379, -0.021403208374977112, 0.00858255010098219, 0.010382606647908688, -0.0007993924082256854, -0.008559764362871647, -0.0034690971951931715, 0.008461027406156063, -0.016056204214692116, 0.011894047260284424, 0.019109465181827545, 0.005882464814931154, -0.026431215927004814, -0.014218171127140522, 0.0014848571736365557, 0.0071432641707360744, 0.005806512665003538, -0.0004839551984332502, -0.01144593209028244, 0.013853603042662144, 0.02125130593776703, 0.018501851707696915, 0.014225766994059086, 0.016557486727833748, 0.00843824166804552, -0.030927561223506927, 0.015630874782800674, 0.012964967638254166, -0.03803664818406105, 0.012721921317279339, -0.009957277216017246, -0.01768157258629799, 0.0018997436854988337, 0.0005406816489994526, 0.0030190828256309032, -0.015828348696231842, 0.0021665242966264486, -0.0067559098824858665, -0.008536978624761105, 0.010458558797836304, 0.0315047949552536, -0.010131966322660446, -0.012205449864268303, -0.0329023078083992, 0.0007523972308263183, -0.008278743363916874, 0.02099306881427765, 0.011050982400774956, -0.015600494109094143, 0.01363334245979786, -0.003907718695700169, 0.007409095298498869, -0.001107946503907442, -0.011544669046998024, -0.01932213082909584, -0.003300104523077607, -0.018836038187146187, 0.011043387465178967, -0.003972277510911226, 0.003315294859930873, -0.002624133601784706, 0.018167663365602493, 0.009509162046015263, -0.006679958198219538, 0.009630684740841389, -0.0005800816579721868, -0.004283679649233818, -0.01501566544175148, 0.00511914910748601, 0.0029450298752635717, 0.0174992885440588, -0.02594512514770031, 0.009114212356507778, -0.007693914230912924, -0.003619101829826832, 0.011681382544338703, 0.016861293464899063, -0.0014354884624481201, 0.007652140688151121, -0.02497294172644615, 0.010443368926644325, 0.008909142576158047, -0.018911991268396378, 4.767753853229806e-05, -0.01305610965937376, 0.004283679649233818, -0.02656792849302292, -0.02550460398197174, -0.002367796376347542, 0.00517991092056036, -0.01853223145008087, 0.007378714624792337
                ],
                "page_number": 104
            },
            {
                "@search.action": "upload",
                "id": "earth_at_night_508_page_105_verbalized",
                "page_chunk": "# Urban Structure\n\n## March 16, 2013\n\n### Phoenix Metropolitan Area at Night\n\nThis figure presents a nighttime satellite view of the Phoenix metropolitan area, highlighting urban structure and transport corridors. City lights illuminate the layout of several cities and major thoroughfares.\n\n**Labeled Urban Features:**\n\n- **Phoenix:** Central and brightest area in the right-center of the image.\n- **Glendale:** Located to the west of Phoenix, this city is also brightly lit.\n- **Peoria:** Further northwest, this area is labeled and its illuminated grid is seen.\n- **Grand Avenue:** Clearly visible as a diagonal, brightly lit thoroughfare running from Phoenix through Glendale and Peoria.\n- **Salt River Channel:** Identified in the southeast portion, running through illuminated sections.\n- **Phoenix Mountains:** Dark, undeveloped region to the northeast of Phoenix.\n- **Agricultural Fields:** Southwestern corner of the image, grid patterns are visible but with much less illumination, indicating agricultural land use.\n\n**Additional Notes:**\n\n- The overall pattern shows a grid-like urban development typical of western U.S. cities, with scattered bright nodes at major intersections or city centers.\n- There is a clear transition from dense urban development to sparsely populated or agricultural land, particularly evident towards the bottom and left of the image.\n- The illuminated areas follow the existing road and street grids, showcasing the extensive spread of the metropolitan area.\n\n**Figure Description:**  \nA satellite nighttime image captured on March 16, 2013, showing Phoenix and surrounding areas (including Glendale and Peoria). Major landscape and infrastructural features, such as the Phoenix Mountains, Grand Avenue, the Salt River Channel, and agricultural fields, are labeled. The image reveals the extent of urbanization and the characteristic street grid illuminated by city lights.\n\n---\n\nPage 89",
                "page_embedding_text_3_large": [
                -0.012408008798956871, -0.010935738682746887, -0.01799791119992733, 0.021761255338788033, 0.008125041611492634, -0.04487668350338936, 0.03457866609096527, 0.03738148882985115, -0.025697806850075722, -0.0032535595819354057, -0.00041063150274567306, 0.07577073574066162, 0.032972551882267, -0.049852482974529266, -0.020564543083310127, 0.003302766475826502, -0.040751177817583084, 0.030327189713716507, -0.015344676561653614, 0.03243718296289444, 0.027981005609035492, -0.01735231839120388, -0.02837466076016426, 0.020958198234438896, -0.004117632284760475, 0.02560332790017128, 0.020596034824848175, 0.015486392192542553, 0.004263285081833601, 0.009408357553184032, -0.01991894841194153, 0.006778741255402565, 0.021336106583476067, -0.02295796573162079, -0.003273242386057973, 0.02432788535952568, 0.019604025408625603, 0.008589554578065872, 0.041003115475177765, 0.019037161022424698, 0.0077431960962712765, -0.06295332312583923, 0.02824869193136692, 0.008188026025891304, -0.00022856600116938353, -0.039743419736623764, 0.018722238019108772, -0.010534211061894894, 0.027303919196128845, 0.0054796794429421425, -0.010565703734755516, -0.02750862017273903, -0.049411591142416, -0.02887853980064392, 0.025902505964040756, 0.01876947656273842, 0.04966352880001068, 0.07766028493642807, 0.00472386134788394, -0.005298597738146782, 0.03256314992904663, 0.012911887839436531, 0.0014516032533720136, 0.018155373632907867, 0.017887689173221588, 0.0418219193816185, -0.012439501471817493, -0.0007174364291131496, -0.010384622029960155, -0.008920225314795971, -0.009266640990972519, 0.0633942186832428, 0.00955794658511877, 0.009030448272824287, 0.06909434497356415, -0.017824703827500343, 0.025414373725652695, -0.007003124337643385, -0.012974872253835201, -0.005278915166854858, 0.017824703827500343, -0.016139859333634377, -0.009014702402055264, 0.07192866504192352, 0.015620235353708267, -0.010211413726210594, -0.03596433252096176, -0.09422528743743896, 0.016895677894353867, 0.03662567213177681, -0.03854670748114586, -0.03634224086999893, -0.013478751294314861, 0.009683915413916111, 0.0032594643998891115, 0.0016395736020058393, 0.05630842596292496, 0.00037963115028105676, 0.02637489326298237, 0.02623317763209343, 0.006991314701735973, 0.008613173849880695, 0.0029150161426514387, -0.028689585626125336, 0.03369687870144844, -0.021257376298308372, -0.010305890813469887, 0.0011199488071724772, 0.0005688316305167973, 0.03750745952129364, 0.01686418429017067, -0.030075250193476677, -0.007231444586068392, -0.02017088793218136, -0.03240568935871124, -0.0022910728584975004, 0.018722238019108772, 0.00937686488032341, 0.06676390767097473, 0.03259464353322983, 0.019194623455405235, -0.0023501210380345583, 0.0367831327021122, -0.007833736948668957, -0.01837582141160965, 0.003137431340292096, 0.01836007460951805, -0.014628224074840546, -0.0023383114021271467, 0.04487668350338936, -0.04878174141049385, 0.02837466076016426, 0.009455596096813679, 0.01572258584201336, -0.009353245608508587, -0.014392031356692314, -0.0132976695895195, 0.008668285794556141, 0.021745508536696434, 0.0123450243845582, -0.020076410844922066, -0.014415650628507137, 0.0002598123683128506, -0.028595108538866043, -0.034893590956926346, 0.006046542432159185, 0.02141483873128891, 0.044561758637428284, 0.013911771588027477, -0.01040036790072918, -0.011274282820522785, -0.005522981286048889, 0.04538056254386902, -0.004034964833408594, -0.019966186955571175, -0.006751185283064842, -0.0053025344386696815, -0.00039709958946332335, -0.043522510677576065, 0.0038656932301819324, 0.015549377538263798, 0.004452239256352186, -0.0015155721921473742, -0.027918020263314247, 0.013163827359676361, -0.04002685099840164, 0.026674071326851845, -0.04966352880001068, -0.04487668350338936, 0.010738911107182503, 0.047962937504053116, -0.022280879318714142, -0.03508254513144493, -0.003668865654617548, 0.02865809202194214, -0.040593717247247696, 0.011974988505244255, 0.027697574347257614, -0.03939700499176979, 0.0065386113710701466, -0.024800272658467293, -0.041506994515657425, 0.0040743304416537285, -0.0155257573351264, 0.0019062749342992902, -0.00855806190520525, 0.03015398234128952, -0.004231792408972979, 0.028815554454922676, -0.005263168830424547, -0.00288155535236001, 0.03473612666130066, 0.06181959807872772, -0.02966585010290146, 0.031098755076527596, -0.028563614934682846, 0.027807798236608505, -0.02686302550137043, -0.01597452536225319, -0.012722933664917946, -0.01924186199903488, -0.007144840434193611, -0.021084168925881386, 0.037696413695812225, 0.02571355178952217, 0.02270602621138096, 0.01977723278105259, -0.0039660753682255745, -0.008447838947176933, 0.01623433642089367, 0.0011051867622882128, -0.061882585287094116, -0.009904362261295319, -0.026894517242908478, -0.021761255338788033, 0.01760425604879856, 0.016013890504837036, -0.02840615250170231, 0.013541735708713531, -0.03213800489902496, 0.028185706585645676, 0.029193462803959846, 0.007758942432701588, 0.017462540417909622, -0.0062551796436309814, -0.027020487934350967, -0.028579361736774445, 0.010841261595487595, -0.06071736663579941, -0.018218358978629112, -0.027713319286704063, -0.022863488644361496, 0.01273867953568697, 0.025398628786206245, 0.025697806850075722, 0.012612709775567055, 0.004385318141430616, 0.02028111182153225, -0.011447491124272346, -0.019037161022424698, 0.016753962263464928, -0.005393075291067362, -0.026658324524760246, -0.026579594239592552, 0.049600545316934586, -0.019320592284202576, -0.019415071234107018, 0.03347643092274666, 0.014612478204071522, 0.035271499305963516, 0.009164291433990002, -0.020202379673719406, 0.0329095683991909, 0.018564775586128235, 0.028957270085811615, 0.017667241394519806, -0.003540927777066827, 0.008077803067862988, 0.005456059705466032, 0.007652655243873596, -0.03243718296289444, 0.042010873556137085, -0.01273867953568697, 0.027225187048316002, 0.0006318164523690939, 0.015313183888792992, -0.02317841351032257, 0.005873334128409624, 0.003970012068748474, -0.03911357372999191, 0.03253166005015373, -0.06770867854356766, 0.01850179024040699, -0.023792514577507973, 0.012518232688307762, -0.0032791472040116787, 0.015895793214440346, -0.016801200807094574, -0.0006948012742213905, -0.002169039798900485, 0.001939735608175397, -0.007530622184276581, 0.025650566443800926, -0.04424683377146721, -0.03829476982355118, 0.026768548414111137, -0.017383810132741928, -0.013982629403471947, -0.015029752627015114, -0.04103460907936096, -0.015919413417577744, -0.06701584905385971, 0.006707882974296808, 0.006877155043184757, 0.032626137137413025, -0.005306471139192581, 0.004995483439415693, -0.05983557924628258, -0.04777398332953453, 0.007298365700989962, -0.008337615057826042, 0.04134953394532204, -0.052403368055820465, 0.010242906399071217, 0.013502370566129684, 0.04345952346920967, -0.018564775586128235, 0.04638832062482834, -0.016753962263464928, -0.046199362725019455, -0.002481995616108179, -0.026815786957740784, 0.008935971185564995, -0.00938473828136921, 0.01760425604879856, 0.008046310395002365, 0.030342936515808105, 0.006621278822422028, -0.008408472873270512, -0.017131870612502098, 0.02017088793218136, -0.04843532666563988, -0.0032437180634588003, 0.006892900913953781, -0.019178876653313637, 0.010660180822014809, -0.00824313797056675, -0.008211645297706127, -0.02369803749024868, -0.02165103144943714, -0.01280953735113144, 0.008605300448834896, 0.001155377714894712, -0.0020037044305354357, 0.03479911386966705, -0.009180037304759026, 0.026689816266298294, 0.027461381629109383, -0.00021737143106292933, 0.01607687585055828, 0.015895793214440346, -0.02155655436217785, -0.031114500015974045, -0.020580289885401726, 0.008935971185564995, -0.02382400818169117, -0.010998724028468132, 0.015029752627015114, -0.00855806190520525, 0.003346068551763892, 0.0010677895043045282, 0.020816482603549957, -0.020060664042830467, -0.018470298498868942, 0.009597311727702618, 0.021509315818548203, -0.03117748536169529, -0.03602731600403786, 0.0557415634393692, -0.01291976124048233, 0.02902025543153286, -0.055174700915813446, -0.04160147160291672, 0.018045149743556976, -0.03690910339355469, 0.035680901259183884, -0.012313531711697578, 0.003662960836663842, -0.00025538375484757125, 0.009148544631898403, -0.04361698776483536, 0.023146919906139374, -0.03637373447418213, -0.009282387793064117, 0.004153061658143997, -0.031618379056453705, 0.0380113385617733, -0.01057357620447874, -0.012431629002094269, 0.007203888613730669, -0.024753034114837646, 0.05234038457274437, 0.027020487934350967, 0.0038617567624896765, 0.011597080156207085, 0.05794603377580643, -0.00849507749080658, -0.025965491309762, 0.018942683935165405, 0.004700242076069117, -0.0008862160611897707, -0.016407545655965805, 0.033633891493082047, -0.020092157647013664, 0.034925080835819244, 0.0014319204492494464, -0.008282504044473171, -0.00969178881496191, 0.016171352937817574, 0.04131804034113884, -0.0030783829279243946, 0.04245176911354065, 0.028075482696294785, 0.06487436592578888, -0.002304850611835718, 0.05051382631063461, 0.021493569016456604, -0.03451567888259888, 0.0068495990708470345, -0.01876947656273842, 0.0035743885673582554, -0.007971515879034996, -0.008534442633390427, -0.02102118358016014, 0.0013266177847981453, 0.01476206723600626, -0.011904130689799786, 0.035775378346443176, 0.003956234082579613, 0.016061129048466682, -0.0011622667079791427, -0.00256466306746006, -0.055552609264850616, 0.024264901876449585, -0.013218938373029232, 0.0024583761114627123, -0.00038430580752901733, 0.01475419383496046, -0.0048025925643742085, 0.00835336185991764, 0.03977491334080696, -0.014714828692376614, 0.009636676870286465, -0.0017389714485034347, -0.0025764727033674717, -0.005121453199535608, 0.018202612176537514, 0.013069350272417068, 0.007967579178512096, -0.030138235539197922, -0.03634224086999893, 0.020722005516290665, 0.013793675228953362, -0.012085212394595146, -0.013707071542739868, 0.03942849487066269, 0.0027929830830544233, -0.0077353231608867645, 0.012321405112743378, -0.030468905344605446, -0.03407478705048561, 0.042010873556137085, -0.05104919523000717, -0.02141483873128891, 0.021115660667419434, -0.021477822214365005, -0.005873334128409624, -0.019430816173553467, 0.009605185128748417, -0.022658787667751312, 0.014478635042905807, -0.022611549124121666, 0.004660876467823982, -0.07274746894836426, 0.0070661092177033424, 0.013588974252343178, -0.011093201115727425, 0.029980773106217384, 0.01607687585055828, 0.0012419818667694926, 0.04446728155016899, -6.22774678049609e-05, -0.04978949949145317, 0.07281044870615005, -0.027587350457906723, 0.0034385775215923786, -0.022580057382583618, -0.04128654673695564, 0.02065902017056942, -0.026296161115169525, -0.03665716573596001, 0.03501955792307854, 0.016675230115652084, -0.014132218435406685, -0.0018472266383469105, 0.024264901876449585, 0.0009073750698007643, 0.014549492858350277, -0.04471922293305397, 0.01760425604879856, 0.035428959876298904, 0.006845662370324135, -0.006511055398732424, -0.0072747464291751385, -0.012203308753669262, 0.00535764591768384, -0.03470463678240776, 0.0006490388768725097, -0.014305426739156246, 0.046703241765499115, -0.06474839150905609, 0.03508254513144493, -0.001677954918704927, -0.01071529183536768, -0.003456291975453496, 0.006416578311473131, 0.003907027188688517, 0.012400136329233646, -0.030232712626457214, -0.007621163036674261, 0.014699081890285015, -0.00040325045119971037, -0.007825863547623158, -0.0354919470846653, 0.015872174873948097, -0.02026536501944065, 0.030216965824365616, 0.0033874022774398327, 0.006861408706754446, -0.004420747049152851, 0.00862891972064972, -0.02346184477210045, -0.002019450766965747, -0.01248674001544714, -0.012526106089353561, 0.013092969544231892, 0.014360538683831692, -0.02713070996105671, 0.005495425313711166, -0.03949148207902908, -0.037570443004369736, 0.04106610268354416, 0.028028244152665138, -0.003505498869344592, 0.004212109837681055, -0.022989459335803986, -0.02587101422250271, 0.03391732648015022, 0.0158485546708107, -0.004861640743911266, -0.03281509131193161, 0.019855964928865433, 0.0018245914252474904, 0.033381953835487366, 0.018706491217017174, 0.023099681362509727, 0.012785918079316616, 0.012061593122780323, -0.0012439502170309424, -0.006270925980061293, -0.005404884926974773, -0.0016661452827975154, -0.0017173204105347395, 0.029429657384753227, -0.02560332790017128, -0.011101074516773224, 0.00409007677808404, -0.0004212109779473394, -0.015754077583551407, 0.03640522435307503, -0.015313183888792992, -0.014966767281293869, -0.031508155167102814, 0.011959242634475231, -0.017525525763630867, -0.0008035484934225678, -0.01599027030169964, 0.02700474113225937, -0.004822275135666132, 0.028280183672904968, -0.022721773013472557, 0.0072235711850225925, 0.011935623362660408, 0.024705795571208, 0.023414606228470802, 0.031004276126623154, -0.012289912439882755, -0.02242259494960308, -0.015069117769598961, -0.013604721054434776, 0.004767163656651974, 0.0019613865297287703, 0.033130016177892685, 0.012896141968667507, -0.00424753874540329, -0.01273867953568697, -0.00267882295884192, 0.0035625786986202, 0.026563847437500954, 0.0021651030983775854, 0.03081532195210457, 0.01545489951968193, 0.0532536655664444, 0.004940371494740248, 0.01659649983048439, 0.014407777227461338, -0.014360538683831692, 0.005727681796997786, 0.010227159596979618, -0.026658324524760246, 0.017887689173221588, 0.022973712533712387, 0.01900566928088665, -0.012707186862826347, 0.018816715106368065, 0.029114732518792152, 0.07564476877450943, 0.0007440081681124866, 0.029949281364679337, -0.006794487126171589, -0.03081532195210457, -0.001780305290594697, 0.02776055969297886, -0.011974988505244255, 0.02092670649290085, -0.03917655721306801, -0.0028205388225615025, 0.04027879238128662, 0.016187097877264023, -0.03001226671040058, 0.0136204669252038, 0.0017753845313563943, 0.02991778962314129, -0.00721963495016098, -0.029429657384753227, -0.01451800111681223, -0.016659485176205635, 0.0057867299765348434, -0.027225187048316002, -0.06726778298616409, 0.012982745654881, -0.018580522388219833, 0.018596267327666283, -0.051269643008708954, -0.007215698249638081, -0.019950442016124725, -0.02320990525186062, 0.0018442742293700576, 0.006369339767843485, -0.013447258621454239, -0.022847743704915047, 0.03464164957404137, -0.016565008088946342, -0.006570104043930769, 0.009502834640443325, 0.0008581681759096682, -0.027571603655815125, -0.040089838206768036, 0.015116356313228607, -0.008833620697259903, -0.014242442324757576, 0.01810813508927822, -0.0028854920528829098, 0.010235032998025417, 0.015447027049958706, -0.03190181031823158, -0.016832692548632622, -0.020706258714199066, -0.004857704043388367, -0.01082551572471857, -0.02609146200120449, 0.02457982487976551, -0.02330438233911991, -0.004286904353648424, 0.010746784508228302, 0.032106511294841766, -0.008014817722141743, -0.010628688149154186, 0.012911887839436531, -0.020485812798142433, 0.013321288861334324, 0.004456175956875086, 0.03220098838210106, -0.006026859860867262, 0.02242259494960308, -0.02878406271338463, -0.01216394267976284, 0.009983093477785587, -0.0018954493571072817, 0.014651843346655369, 0.0016622086986899376, 0.0008586602052673697, 0.0017625908367335796, -0.028343169018626213, 0.007821926847100258, -0.01620284467935562, 0.05363157391548157, -0.005932382773607969, 0.00022364531469065696, -0.0015362390549853444, -0.011841146275401115, -0.0354604534804821, -0.0030350808519870043, 0.007373160216957331, -0.008069929666817188, -0.0016100493958219886, -0.004188490565866232, 0.012533978559076786, 0.049222636967897415, -0.04679771885275841, -0.005578092765063047, -0.026579594239592552, -0.010975104756653309, 0.0026768548414111137, -0.029240701347589493, 0.035932838916778564, -0.011045962572097778, 0.00803450122475624, -0.034011803567409515, 0.008920225314795971, -0.005889080464839935, 0.003944424446672201, 0.0007213730132207274, 0.01710037887096405, 0.026138700544834137, -0.015021879225969315, -0.02862660028040409, 0.005751301068812609, 0.0003437101258896291, -0.00023643911117687821, 0.02091095969080925, -0.002879587234929204, -0.022123416885733604, 9.939791925717145e-05, -0.015746204182505608, -0.033759862184524536, -0.014864416792988777, -0.000627879926469177, -0.007046426646411419, 0.006266989279538393, -0.027666080743074417, -0.03265762701630592, 0.006207941100001335, -0.0013423638883978128, -0.017289333045482635, -0.022485580295324326, -0.019714247435331345, -0.0016543356468901038, -0.007459764368832111, -0.026390638202428818, -0.0050505949184298515, -0.013659832067787647, 0.01724209450185299, -0.029177717864513397, -0.029429657384753227, 0.019178876653313637, 0.023146919906139374, -0.00020580780983436853, -0.0171161238104105, 0.0020548796746879816, -0.0027063789311796427, -0.016250083222985268, -0.012463120743632317, -0.02017088793218136, 0.0148880360648036, 0.0023678354918956757, 0.015061244368553162, -0.00969178881496191, 0.002747712656855583, -0.009211529977619648, -0.005152945406734943, 0.01750977896153927, -0.050860241055488586, 0.05038785561919212, 0.003877502866089344, -0.002647330751642585, -0.003251591231673956, -0.009739027358591557, 0.023587815463542938, 2.2681300833937712e-05, 0.0052277399227023125, 0.0024012962821871042, -0.008376981131732464, -0.009243021719157696, 0.025304151698946953, 0.0246900487691164, 0.02317841351032257, 0.02028111182153225, -0.017147617414593697, -0.006680327467620373, -0.0010392494732514024, -0.022721773013472557, 0.0055190445855259895, 0.018438804894685745, 0.007503066677600145, -0.01153409481048584, 0.03464164957404137, -0.025288404896855354, -0.05520619451999664, 0.015124229714274406, 0.017037393525242805, 0.011864765547215939, 0.01812388189136982, -0.035554930567741394, 0.026815786957740784, -0.008329742588102818, 0.008636793121695518, 0.018029404804110527, -0.011187678202986717, 0.00938473828136921, 0.024753034114837646, 0.023666545748710632, 0.017320824787020683, -0.007987262681126595, 0.0007868181564845145, -0.03750745952129364, 0.003698389744386077, -0.025052212178707123, 0.015596616081893444, 0.01000671274960041, 0.018045149743556976, -0.0027634589932858944, 0.0031315265223383904, 0.021383345127105713, -0.02229662612080574, 0.008069929666817188, -0.010683800093829632, 0.006605532951653004, -0.012896141968667507, -0.0040231551975011826, -0.003849947126582265, 0.004160934593528509, 0.005117516499012709, -0.004412873648107052, -0.025571836158633232, 0.0011347108520567417, 0.007073982618749142, -0.005830032285302877, 0.004944308195263147, 0.009431976824998856, -0.006869281642138958, 0.027902275323867798, -0.020312603563070297, 0.04282967746257782, 0.00012609265104401857, 0.02522541955113411, -0.009526453912258148, 0.034263741225004196, -0.014691208489239216, 0.014273934066295624, 0.014336919412016869, 0.004306586924940348, 0.022784758359193802, -0.00962880440056324, 0.012022227048873901, -0.006514992099255323, 0.01026652567088604, -0.013903899118304253, 0.02763458900153637, 0.027272425591945648, 0.027650335803627968, 0.037822384387254715, -0.008731270208954811, 0.0316498726606369, -0.010250778868794441, 0.0055190445855259895, -0.008613173849880695, 0.018722238019108772, -0.019037161022424698, 0.006432324647903442, 0.0012734743067994714, 0.014966767281293869, -0.007471574004739523, 0.0042003002017736435, -0.010392495431005955, 0.004361698869615793, 0.02612295374274254, 0.013494497165083885, -0.015494265593588352, -0.020737752318382263, -0.02269028127193451, -0.010927866213023663, 0.03116173855960369, 0.0002306573005625978, 0.0005560378776863217, 0.0043656351044774055, 0.04068819433450699, 0.017840450629591942, -0.005499362014234066, 0.0036531195510178804, 0.004121568985283375, 0.009109179489314556, 0.002728029852733016, 0.0183915663510561, -0.020107902586460114, -0.01887969858944416, -0.012069465592503548, -0.003096097381785512, 3.210749491699971e-05, -0.02635914646089077, -0.00372397736646235, 0.013588974252343178, -0.019588278606534004, 0.018407313153147697, -0.00324568641372025, 0.008723397739231586, -0.00987287051975727, -0.01001458615064621, -0.0015283660031855106, 0.011730922386050224, 0.0064953095279634, 0.009487087838351727, 0.02902025543153286, 0.017651494592428207, -0.0050112297758460045, 0.02154080756008625, -0.018706491217017174, 0.01801365800201893, -0.008266757242381573, 0.0329410582780838, 0.014147965237498283, -0.011478982865810394, 0.020470065996050835, -0.003619658760726452, -0.002357994206249714, 0.02026536501944065, 3.678214852698147e-05, 0.005467869341373444, -0.002413105918094516, -0.02459557168185711, 0.023493336513638496, 0.012148196808993816, -0.025697806850075722, 0.01773022674024105, 0.006022923160344362, 0.005393075291067362, 0.012274166569113731, -0.016816945746541023, -0.004275094717741013, 0.03741298243403435, -0.025052212178707123, -0.0023127237800508738, 0.023115428164601326, 0.011045962572097778, -0.013911771588027477, 0.0028913968708366156, -0.011093201115727425, -0.020344097167253494, 0.003117748536169529, -0.016313068568706512, 0.014281807467341423, 0.0009152481216005981, 0.03564940765500069, -0.03196479380130768, -0.03637373447418213, 0.015793442726135254, 0.00481046549975872, 0.013250431045889854, -0.008565935306251049, -0.015415534377098083, 0.005798540078103542, -0.03624776378273964, 0.004475858528167009, 0.03448418900370598, 0.001021535019390285, 0.0027103153988718987, -0.016265830025076866, -0.020989689975976944, 0.009959474205970764, -0.0032909568399190903, 0.004050711169838905, 0.018549028784036636, -0.034673143178224564, 0.023729531094431877, -0.03040592186152935, -0.021509315818548203, -0.019871709868311882, -0.03816879913210869, -0.030894054099917412, 0.0057788570411503315, 8.349794370587915e-06, -0.007932150736451149, 0.0069243935868144035, -0.023162666708230972, 0.013242557644844055, 0.035680901259183884, -0.015943031758069992, -0.005885143764317036, 0.014714828692376614, 0.01659649983048439, 0.028091229498386383, -0.002649298869073391, 0.017824703827500343, 0.041884902864694595, 0.04462474212050438, -0.02698899433016777, 0.006416578311473131, -0.0001847718667704612, 0.004318396560847759, 0.013770055957138538, 0.0076605286449193954, -0.032012034207582474, 0.0006962774787098169, -0.05470231547951698, -0.018974175676703453, -0.022501327097415924, 0.01927335374057293, 0.015565123409032822, 0.004983673803508282, -0.013809421099722385, -0.0018452584045007825, -0.004849831108003855, -0.012274166569113731, 0.009809885174036026, -0.01032163668423891, -0.001655319705605507, -0.011967115104198456, -0.02295796573162079, -0.01812388189136982, -0.02294222079217434, 0.03574388474225998, 0.005408821161836386, 0.013793675228953362, -0.00930600706487894, 0.0018718300852924585, 0.024044454097747803, -0.01388027984648943, 0.009817758575081825, 0.015580869279801846, 0.025241166353225708, -0.026453623548150063, 0.01090424694120884, -0.031020022928714752, 0.0048025925643742085, 0.01007757056504488, 0.022737519815564156, -0.0183915663510561, -0.00835336185991764, 0.0017586542526260018, 0.020611781626939774, 0.008062057197093964, 0.021194390952587128, -0.003353941487148404, -0.02520967274904251, -0.01083338912576437, 0.02687877044081688, -0.0030252395663410425, -0.02080073580145836, -0.002733934670686722, 0.031508155167102814, 0.005959938280284405, -0.00803843792527914, -0.014171584509313107, 0.04160147160291672, -0.015517884865403175, -0.005467869341373444, -0.024107439443469048, 0.025571836158633232, -0.031712856143713, 0.01887969858944416, -0.017194855958223343, -0.001075662556104362, 0.03253166005015373, -0.024422364309430122, -0.018423059955239296, -0.005306471139192581, -0.0012734743067994714, 0.019194623455405235, 0.01914738491177559, 0.036310747265815735, -0.0015391914639621973, 0.009746900759637356, 0.0017891625175252557, 0.0033992119133472443, 0.010203540325164795, 0.014565239660441875, 0.011951369233429432, -0.008447838947176933, 0.011014469899237156, 0.01033738348633051, 0.0034425139892846346, 0.010494844987988472, -0.013415765948593616, 0.02180849388241768, -0.007085792254656553, 0.006444134283810854, -0.0032122258562594652, -0.006605532951653004, -0.03204352781176567, 0.006026859860867262, 0.0038401056081056595, -0.010864880867302418, -0.02432788535952568, 0.030500398948788643, -0.008010881952941418, 0.0034129898995161057, 0.011770287528634071, -0.01323468517512083, -0.015635980293154716, 0.0005309423431754112, -0.013714944012463093, 0.023288637399673462, 0.023650798946619034, 0.006570104043930769, 0.004546716809272766, -0.019855964928865433, -0.010990850627422333, 0.096303790807724, 0.013470877893269062, 0.01051059179008007, 0.015242326073348522, 0.007388906553387642, -0.0029799691401422024, -0.030594876036047935, 0.009683915413916111, 0.028689585626125336, -0.007211761549115181, 0.014171584509313107, 0.004763226956129074, -0.018832460045814514, -0.009431976824998856, 0.013533863238990307, -0.04349101707339287, -0.005058468319475651, 0.03198054060339928, -0.0027378713712096214, 0.026327654719352722, 0.007684147916734219, 0.003531086491420865, -0.01965126395225525, -0.0031020021997392178, -0.010628688149154186, -0.02015514113008976, 0.04569548740983009, 0.004345952533185482, 0.006133146584033966, -0.018470298498868942, -0.021714016795158386, -0.027162203565239906, -0.002909111324697733, -0.009613057598471642, -0.02650086209177971, 0.01153409481048584, -0.002015514299273491, 0.01697440817952156, 0.01999768055975437, 0.010250778868794441, -0.017651494592428207, 0.028075482696294785, -0.005223803222179413, -0.020580289885401726, -0.002773300278931856, -0.005082087591290474, -0.006322101224213839, 0.018564775586128235, 0.004593955352902412, -0.004778973292559385, -0.00424753874540329, -0.02155655436217785, 0.014210949651896954, -0.030972784385085106, -0.01773022674024105, 0.02902025543153286, -0.014911655336618423, 0.006778741255402565, 0.023036697879433632, 0.008872986771166325, 0.011943495832383633, 0.002852031262591481, -0.010526337660849094, -0.022249387577176094, 0.00994372833520174, 0.009857123717665672, -0.0048419577069580555, -0.005751301068812609, 0.010030332021415234, 0.009998840279877186, -0.03479911386966705, 0.002857936080545187, 0.0054363771341741085, -0.00408614007756114, -0.002852031262591481, 0.023477591574192047, -0.003580293385311961, -0.0016061129281297326, -0.006637025158852339, 0.00930600706487894, 0.0021670714486390352, -8.002270624274388e-05, -0.007884912192821503, 0.0075345588847994804, -0.008534442633390427, -0.013541735708713531, 0.0030823196284472942, -0.001230172230862081, -0.03341344743967056, -0.010486972518265247, -0.02965010330080986, 0.03716104477643967, -0.03243718296289444, 0.00913279876112938, -0.01813962683081627, -0.009117052890360355, 0.041003115475177765, -0.01977723278105259, -0.0035980078391730785, 0.009109179489314556, 0.025288404896855354, -0.0011898225639015436, 0.029051747173070908, -0.011998607777059078, 0.0005225771456025541, -0.02393423020839691, -0.021005436778068542, 0.019541040062904358, -0.000627879926469177, -0.0029465085826814175, -0.0056528872810304165, -0.003536991309374571, 0.002296977676451206, 0.0234303530305624, 0.029492640867829323, 0.002674886491149664, 0.008699778467416763, -0.0036787071730941534, -0.021241629496216774, -0.0007046426762826741, -0.0133134163916111, 0.02799675241112709, 0.014321173541247845, 0.003098065732046962, 0.015053371898829937, 0.008904478512704372, 0.00955794658511877, -0.003798771882429719, 0.012313531711697578, 0.0014899845700711012, -0.015195087529718876, -0.004420747049152851, 0.006593723315745592, 0.031224723905324936, -0.0015332866460084915, -0.025697806850075722, 0.034925080835819244, -0.027445634827017784, 0.011636445298790932, 0.011557714082300663, -0.02054879628121853, -0.00204307003878057, 0.022217893972992897, 0.02332012914121151, -0.0027693638112396, 0.0046175746247172356, -0.031020022928714752, 0.013998376205563545, 0.014407777227461338, 0.009077686816453934, 0.022973712533712387, -0.010943612083792686, -0.007664464879781008, -0.009164291433990002, -4.846878437092528e-05, 0.007455827668309212, -0.0006869281642138958, 0.01432904601097107, 0.020060664042830467, -0.021855732426047325, 0.0015844618901610374, -0.018060896545648575, -0.013431512750685215, -0.01400624867528677, 0.024753034114837646, 0.0297760721296072, 0.011014469899237156, 0.030311444774270058, -0.005841841921210289, -0.020816482603549957, 0.01999768055975437, -0.018344327807426453, -0.004515224136412144, 0.010959358885884285, 0.0158406812697649, 0.027193695306777954, -0.006845662370324135, 0.0027634589932858944, 0.02394997701048851, -0.005326153710484505, -0.021729761734604836, -0.023005204275250435, -0.027036232873797417, 0.027776304632425308, 0.0065897866152226925, 0.009361119009554386, 0.0018747824942693114, -0.006554357707500458, 0.005621395073831081, -0.008290376514196396, 0.005684379953891039, 0.0316498726606369, 0.003436609171330929, -0.0009708519210107625, 0.006310291588306427, -0.0024249155540019274, -0.0061646392568945885, -0.03234270587563515, 0.027335410937666893, -0.01597452536225319, 0.005345836281776428, -0.005499362014234066, 0.0007016902673058212, 0.02698899433016777, -0.009400484152138233, 0.0036019443068653345, -0.025020718574523926, 0.010794023051857948, -0.005495425313711166, -0.013392146676778793, 0.02875256910920143, 0.012274166569113731, -0.009581565856933594, -0.031744349747896194, 0.03133494779467583, -0.010518464259803295, 0.00706217298284173, -0.010471225716173649, 0.0024741224478930235, 0.020942451432347298, 0.003137431340292096, -0.011392379179596901, 0.011061708442866802, -0.001651383237913251, -0.04204236716032028, 0.007829800248146057, 0.025571836158633232, -0.012195435352623463, -0.009927982464432716, -0.012589090503752232, 0.030090996995568275, 0.005802476312965155, 0.010250778868794441, -8.752675785217434e-05, -0.02736690454185009, 0.011085327714681625, -0.02357206866145134, 0.008817874826490879, 0.0007420398760586977, 0.036688655614852905, 0.011069581843912601, -0.020060664042830467, 0.009613057598471642, 0.028059735894203186, -0.026910264045000076, -0.00012154101568739861, -0.007440081797540188, -0.02015514113008976, -0.013856659643352032, -0.0014112535864114761, -0.020706258714199066, 0.017068885266780853, -0.02583952248096466, 0.004385318141430616, -0.0008222471224144101, -0.02256431058049202, -0.014132218435406685, -0.008896606042981148, -0.01823410578072071, -0.011683683842420578, -0.02459557168185711, 0.004153061658143997, 0.019367830827832222, 0.0009191847057081759, 0.0038125498685985804, -0.0019771328661590815, -0.02650086209177971, -0.012604836374521255, 0.026658324524760246, -0.006601596251130104, -0.0183915663510561, 0.0034936892334371805, -0.011683683842420578, -0.006310291588306427, 0.008857239969074726, 0.0037318505346775055, -0.008439965546131134, -0.021477822214365005, 0.004558526445180178, 0.0027851099148392677, -0.008132915012538433, -0.004597891587764025, 0.02080073580145836, 0.0021375473588705063, 0.0033933070953935385, 0.019037161022424698, -0.017021646723151207, 0.008644666522741318, 0.025918252766132355, 0.02127312310039997, -0.013242557644844055, 0.026768548414111137, 0.011360886506736279, 0.03514552861452103, 0.0064244517125189304, -0.020233873277902603, -0.044404298067092896, 0.01570683903992176, -0.0032102575059980154, -0.0056961895897984505, 0.013959010131657124, 0.012510359287261963, 0.04727010801434517, -0.004853767808526754, -0.021777000278234482, -0.025902505964040756, 0.0012291880557313561, 0.014951021410524845, -0.011967115104198456, -0.012659948319196701, -0.01573833078145981, 0.004700242076069117, -0.003635405097156763, 0.03385433927178383, -0.0016415418358519673, -0.012124577537178993, -0.010864880867302418, 0.02089521288871765, -0.007896721363067627, -0.031208977103233337, -0.01900566928088665, -0.025335643440485, 0.001469317707233131, 0.010518464259803295, -0.03284658119082451, 0.011069581843912601, -0.0107625313103199, -0.007255063857883215, -0.014880163595080376, -0.020092157647013664, -0.027823543176054955, -0.010723165236413479, -0.0022359611466526985, -0.011478982865810394, 0.0152108334004879, 0.02939816378057003, 0.01008544396609068, -0.0005265137297101319, -0.02483176440000534, 0.018926937133073807, -0.00408614007756114, -0.005589902866631746, -0.0316341258585453, -0.01248674001544714, 0.003739723702892661, 0.006955885794013739, -0.002747712656855583, 0.0028402216266840696, 0.0034425139892846346, -0.0022418659646064043, -0.03432672470808029, 0.01146323699504137, -0.037696413695812225, 0.012045846320688725, -0.00018329566228203475, 0.024705795571208, 0.013896025717258453, -0.0037220090162009, 0.021777000278234482, 0.0004303142486605793, 0.022280879318714142, 0.0005304502556100488, -0.035177022218704224, -0.013447258621454239, -0.0032063208054751158, 0.01177816092967987, -0.005818222649395466, 0.0005151961231604218, 0.006365403067320585, -0.00481440220028162, -0.0018964335322380066, -7.448693213518709e-05, -0.014029867947101593, -0.01160495262593031, 0.009014702402055264, 0.001914147986099124, -0.009644550271332264, 0.030185474082827568, -0.004389254376292229, -0.008140787482261658, -0.01606900244951248, 0.0010451542912051082, -0.001415190170519054, 0.015069117769598961, -0.01008544396609068, -0.033161506056785583, -0.00288155535236001, -0.004145188257098198, 0.026422131806612015, 0.019320592284202576, 0.03464164957404137, -0.014045614749193192, -0.007503066677600145, -0.010998724028468132, -0.01965126395225525, 0.02067476697266102, 0.015273818746209145, -0.004184553865343332, -0.010282271541655064, 0.01824985072016716, -0.00976264663040638, -0.0031236533541232347, -0.005711935926228762, 0.004235729109495878, 0.008203772827982903, -0.00251939264126122, 0.006983441766351461, 0.006310291588306427, -0.03128770738840103, 0.014376284554600716, 0.020596034824848175, -0.018281344324350357, -0.018045149743556976, -0.001103218412026763, 0.00027039184351451695, -0.0030705099925398827, -0.01573045924305916, -0.014226695522665977, 0.005959938280284405, -0.013108715415000916, -0.008778508752584457, 0.023005204275250435, 0.04273520037531853, -0.010534211061894894, 0.002596155507490039, -0.01645478419959545, -0.00388340768404305, -0.026815786957740784, 0.002192659070715308, -0.010408241301774979, -0.00943984929472208, 0.0158485546708107, -0.003322449279949069, -0.009274514392018318, 0.011912004090845585, -0.01539191510528326, -0.0152108334004879, 0.012211182154715061, 0.02583952248096466, 0.017415301874279976, -0.00987287051975727, -0.01773022674024105, -0.0038164863362908363, 0.0037377553526312113, -0.011982861906290054, -0.0012262356467545033, 0.013888152316212654, 0.010298017412424088, -0.009148544631898403, -0.004672686103731394, 0.0009678995120339096, -0.0013167763827368617, 0.0054875523783266544, 0.007085792254656553, 0.017399556934833527, 0.009069813415408134, -0.02294222079217434, 0.0028166023548692465, 0.007022807374596596, -0.013636212795972824, 0.024170424789190292, -0.009644550271332264, -0.01242375560104847, 0.029996519908308983, -0.0058969538658857346, -0.00048198149306699634, 0.03204352781176567, -0.0003114796127192676, 0.00913279876112938, -0.02952413447201252, 0.01051059179008007, 0.0005535774980671704, 0.005223803222179413, 0.002968159504234791, -0.005731618497520685, -0.008857239969074726, 0.0024544396437704563, -0.011274282820522785, 0.011408125050365925, -0.007259000558406115, 0.005920573137700558, -0.012242673896253109, 0.0009324705461040139, 0.009298133663833141, 0.014958894811570644, 0.04736458510160446, 0.004483731929212809, 0.008408472873270512, 0.019336339086294174, -0.005810349714010954, -0.009990966878831387, 0.023335875943303108, 0.012494613416492939, -0.0035507690627127886, 0.007581797428429127, -0.018407313153147697, -0.0018147501396015286, 0.0008438981603831053, -0.01736806333065033, -0.005711935926228762, -0.00697163213044405, -0.017462540417909622, -0.0033657513558864594, 0.04015282168984413, -0.014565239660441875, 0.008928097784519196, 0.011447491124272346, 0.0017084631836041808, 0.003300798125565052, -0.0023638990242034197, 0.0012656012549996376, 0.01280953735113144, -0.011227044276893139, 0.015510011464357376, 0.018454551696777344, 0.009502834640443325, -0.02295796573162079, -0.015265945345163345, 0.01475419383496046, -0.006668517831712961, 0.00016607325233053416, 0.00810929574072361, 0.020202379673719406, 0.014242442324757576, 0.00576704740524292, 0.0067984238266944885, -0.008935971185564995, 0.0009368991595692933, 0.016911424696445465, 0.022485580295324326, -0.029051747173070908, 0.00031295581720769405, 0.005530854221433401, -0.0017980197444558144, 0.00026399496709927917, -0.005263168830424547, 0.013329162262380123, 0.013069350272417068, -0.004963991232216358, 0.0017350349808111787, -0.003409053198993206, 0.014840797521173954, -0.009400484152138233, 0.00937686488032341, 0.006133146584033966, 0.00032427339465357363, -0.00721963495016098, 0.001130774267949164, -0.002411137567833066, -0.023776769638061523, -0.0042081731371581554, -0.015250199474394321, 0.004227856174111366, 0.010526337660849094, -0.0035488009452819824, 0.0008935971418395638, 0.03144517168402672, 0.000392670975998044, 0.023886991664767265, 0.0027693638112396, -0.02621743083000183, -0.01861201412975788, 0.0404992401599884, -0.011990734376013279, -0.006853535771369934, 0.008416346274316311, 0.012431629002094269, -0.010282271541655064, -0.013848787173628807, -0.011447491124272346, 0.0008148660999722779, 0.009345372207462788, 0.012502486817538738, -0.023509083315730095, 0.0019200528040528297, 0.0058969538658857346, 0.0018550996901467443, 0.002113928087055683, -0.0032279719598591328, -0.012360770255327225, -0.0016996059566736221, -0.015801316127181053, -0.013423639349639416, 0.007243254221975803, 0.0026315844152122736, 0.02776055969297886, -0.0014801432844251394, 0.012053719721734524, -0.011053835973143578, -0.015935158357024193, -0.011415998451411724, 0.019367830827832222, -0.008010881952941418, -0.009904362261295319, 0.0171161238104105, 0.006416578311473131, 0.006546484772115946, -0.014392031356692314, -0.014848670922219753, 0.0027181885670870543, 0.002127705840393901, -0.007436145097017288, 0.00035945631680078804, 0.003690516809001565, -0.004397127777338028, -0.004944308195263147, -0.0017911307513713837, 0.017903434112668037, -0.02941391058266163, -0.012589090503752232, -0.006219750735908747, 0.006522865500301123, 0.004464048892259598, 0.0034523552749305964, 0.0034425139892846346, 0.0025508850812911987, 0.00983350444585085, -0.000918200530577451, -0.0076605286449193954, 0.028043990954756737, 0.005566283129155636, -0.0026847277767956257, 0.02267453446984291, 0.012904014438390732, -0.030374428257346153, -0.0008872002363204956, 0.00424360204488039, 0.02051730453968048, -0.023036697879433632, 0.016627991572022438, -0.011486856266856194, -0.015761950984597206, -0.021099913865327835, -0.022280879318714142, -0.008282504044473171, -0.0004652511270251125, -0.015628108754754066, -0.014998259954154491, -0.004940371494740248, 0.006625215522944927, -0.013179573230445385, 0.0018560838652774692, 0.004700242076069117, 0.0023717719595879316, -0.024674301967024803, -0.0031098753679543734, -0.022107671946287155, -0.007932150736451149, 0.008298249915242195, -0.01861201412975788, -0.01355748251080513, -0.0142660615965724, -0.00160119216889143, -0.00032205908792093396, 0.0019200528040528297, -0.012022227048873901, -0.012360770255327225, -0.0006455943803302944, 0.017320824787020683, 0.0009792171185836196, 0.004353825468569994, -0.02368229255080223, 0.022658787667751312, 0.011762415058910847, 0.02431214042007923, 0.013770055957138538, 0.006436261348426342, 0.008518696762621403, 0.027729066088795662, 0.022737519815564156, 0.02051730453968048, -0.0285321231931448, -0.02292647399008274, 0.005735555198043585, 0.0010067729745060205, 0.0049679274670779705, 0.036184776574373245, 0.0014584922464564443, 0.003905058838427067, 0.025918252766132355, -0.0202496200799942, -0.011667937971651554, 0.01991894841194153, -0.010423987172544003, -0.0022398976143449545, -0.007936087436974049, 0.007507002912461758, -0.013911771588027477, 0.01572258584201336, 0.020123649388551712, -0.0017606224864721298, -0.014021995477378368, 0.003580293385311961, 0.022391103208065033, 0.008912351913750172, -0.01645478419959545, 0.00440106401219964, -0.004188490565866232, 0.0023422478698194027, -0.00888085924088955, 0.023272890597581863, 0.0017143680015578866, 0.014872290194034576, 0.00799513515084982, 0.014140091836452484, 0.01762000285089016, 0.003962138667702675, -0.011203425005078316, -0.005503298714756966, 0.014565239660441875, 0.001233124639838934, -0.0023796451278030872, -0.006148892920464277, 0.003847978776320815, -0.027319664135575294, -0.02089521288871765, -0.01153409481048584, 0.031350694596767426, 0.0020863721147179604, 0.003328354097902775, 0.0152108334004879, 0.0058615244925022125, 0.006259116344153881, 0.002174944616854191, -0.007329858373850584, -0.007873102091252804, -0.016816945746541023, -0.008195899426937103, 0.0026866961270570755, 4.560863453662023e-05, 0.006129210349172354, 0.003879471216350794, -0.023367367684841156, -0.02544586732983589, -3.60748017556034e-05, -0.014880163595080376, -0.016627991572022438, -0.001232140464708209, 0.010313764214515686, 0.016753962263464928, -0.010038205422461033, -0.02105267532169819, -0.019352085888385773, -0.0011593142990022898, 0.015903666615486145, 0.02785503678023815, -0.009187910705804825, -0.00535764591768384, 0.005782793741673231, 0.015297438018023968, 0.02609146200120449, -0.038704171776771545, 0.009927982464432716, 0.007203888613730669, 0.03536597639322281, -0.01064443401992321, 0.013927518390119076, 0.02155655436217785, -0.01636030711233616, -0.006700010038912296, 0.03700358048081398, -0.0012744584819301963, 0.022391103208065033, -0.016139859333634377, 0.00043646511039696634, -0.002877618884667754, -0.001989926677197218, 0.011132566258311272, -0.01121917087584734, 0.012124577537178993, 0.0023678354918956757, 0.02801249735057354, -0.005755237769335508, 0.021887224167585373, -0.01007757056504488, -0.011297902092337608, -0.014486508443951607, -0.007455827668309212, -0.011935623362660408, -0.00976264663040638, -0.02332012914121151, -0.018297089263796806, -0.00288352370262146, -0.003544864244759083, -0.011164058931171894, -0.0142581881955266, 0.02394997701048851, 0.001074678497388959, 0.00697163213044405, -0.007333794608712196, 0.011187678202986717, -0.001207537017762661, 0.014203076250851154, 0.00497580086812377, -0.0025548217818140984, -0.02927219495177269, 0.011069581843912601, -0.006613405887037516, 0.007105474825948477, -0.01007757056504488, -0.002544980263337493, -0.005881207529455423, -0.004613637924194336, -0.007542431820183992, 0.00816440675407648, -0.007770752068608999, -0.011242790147662163, -0.026280416175723076, 0.011242790147662163, 0.026170192286372185, 0.007034617010504007, 0.00624337000772357, -0.0071290940977633, -0.0008114216034300625, -0.028721077367663383, -0.010786150582134724, -0.010353129357099533, -0.0008596443803980947, -0.008975336328148842, -0.034673143178224564, -0.008518696762621403, -0.01634456031024456, 0.004582145716995001, 0.008313995786011219, 0.015218706801533699, 0.012935507111251354, 0.0002050697075901553, 0.026784293353557587, -0.006314228288829327, -0.0025331706274300814, -0.003727913834154606, 7.030434062471613e-05, 0.011376633308827877, 0.02621743083000183, 0.0043656351044774055, 0.0002908127207774669, 0.016029635444283485, 0.01876947656273842, 0.004586081951856613, 0.00449554156512022, 0.01621859148144722, -0.0025981238577514887, 0.015620235353708267, -0.003826327621936798, 0.021603792905807495, -0.00401725061237812, -0.021336106583476067, -0.006700010038912296, -0.010179921053349972, 0.011132566258311272, -0.015092737041413784, 0.010557830333709717, -0.02812272123992443, -0.04623085632920265, 0.018674999475479126, -0.028327422216534615, 0.005790666677057743, -0.033507924526929855, -0.020344097167253494, 0.004767163656651974, 0.016565008088946342, 0.014722701162099838, 0.003176796715706587, 0.015769824385643005, 0.014053487218916416, -0.019792979583144188, -0.013738563284277916, -0.005089960526674986, 0.018848206847906113, 0.010384622029960155, 0.016021763905882835, 0.00017529954493511468, 0.011006597429513931, -0.0028146340046077967, -0.0060701617039740086, 0.004448303021490574, -0.014951021410524845, 0.01710037887096405, -0.007101538125425577, -0.012392262928187847, -0.012408008798956871, -0.015092737041413784, 0.024611318483948708, -0.004279030952602625, 0.014234568923711777, 0.009085560217499733, 0.011801780201494694, 0.013431512750685215, 0.00608197133988142, 0.03662567213177681, -0.025146689265966415, -0.007121221162378788, 0.0036767388228327036, 0.00737709691748023, -0.017194855958223343, 0.00045122718438506126, 0.010502718389034271, -0.0014968735631555319, -0.006408705376088619, 0.020470065996050835, -0.001258712261915207, -0.00471992464736104, -0.0011967115569859743, 0.013045731000602245, 0.024044454097747803, 0.0005452123587019742, -0.012880395166575909, 0.011880511417984962, -0.029996519908308983, 0.0067157563753426075, 0.0015323025872930884, 0.006290608551353216, 0.007495193276554346, -0.029445402324199677, 0.0069322665221989155, 0.03391732648015022, -0.009431976824998856, -0.019714247435331345, 0.02229662612080574, -0.008266757242381573, -0.006278798915445805, 0.012045846320688725, -0.00236193067394197, 0.002426883904263377, 0.00955794658511877, 0.0036826436407864094, 0.006892900913953781, -0.01210095826536417, 0.004901006352156401, 0.00020408557611517608, -0.003981821704655886, -0.0027713319286704063, -0.02114715240895748, -0.012951252982020378, -0.005735555198043585, 0.006078035105019808, -0.016250083222985268, -0.0001238168333657086, 0.011250663548707962, -0.010951485484838486, -0.0070070610381662846, 6.34153766441159e-05, 0.014738447964191437, 0.020186634734272957, -0.01000671274960041, -0.013415765948593616, 0.0016307163750752807, -0.003613753942772746, 0.007573924493044615, -0.009101306088268757, 0.0016927169635891914, 0.00874701701104641, 0.00037101993802934885, 0.007428272161632776, 0.006601596251130104, 0.009707535617053509, -0.012565471231937408, 0.011770287528634071, -0.00037913909181952477, 0.003224035492166877, 0.0013531894655898213, 0.005940255708992481, 0.018785221502184868, 0.009266640990972519, 0.013588974252343178, -0.007510939612984657, -0.0005147040355950594, -0.0139353908598423, -0.020092157647013664, -0.013337035663425922, -0.03684611991047859, 0.014706955291330814, -0.009487087838351727, 0.008707650937139988, 6.821304850745946e-05, 0.006207941100001335, 0.018533281981945038, -0.003033112734556198, 0.010361002758145332, 0.011030216701328754, -0.009888616390526295, -0.00535764591768384, 0.016753962263464928, -0.015084863640367985, 0.009156418032944202, -0.010046078823506832, 0.009392610751092434, 0.008644666522741318, -0.027729066088795662, -0.005593839101493359, -0.01273867953568697, -0.020312603563070297, -0.017210600897669792, 0.014919528737664223, -0.009219402447342873, -0.011675810441374779, -0.0021946271881461143, 0.010982978157699108, -0.01572258584201336, -0.011935623362660408, 0.014210949651896954, -0.006416578311473131, -0.0183915663510561, 0.0018491948721930385, -0.005806413013488054, 0.014470761641860008, -0.01337640080600977, 0.0007833736599422991, -0.010298017412424088, -0.014478635042905807, 0.012699314393103123, -0.021084168925881386, 0.002639457583427429, -0.010935738682746887, 0.024642810225486755, 0.002143452176824212, 0.007869165390729904, 0.0048419577069580555, 0.007636909373104572, -0.011195551604032516, 0.01606900244951248, 0.011305774562060833, 0.003475974779576063, 0.010486972518265247, -0.004227856174111366, 0.004621510859578848, 0.01191987656056881, -0.015407660976052284, 0.0018167183734476566, -0.007329858373850584, -0.010951485484838486, 0.012573344632983208, 0.0005240533500909805, -0.01837582141160965, -0.02141483873128891, 0.026800040155649185, -0.010242906399071217, -0.008022691123187542, 0.01773022674024105, -0.0049679274670779705, 0.0006849599303677678, 0.0015618266770616174, 0.0004130918241571635, 0.025335643440485, 0.013833040371537209, 0.009715408086776733, 0.008392727002501488, 0.024879002943634987, 0.002192659070715308, 0.0029720962047576904, -0.015549377538263798, -0.010211413726210594, 0.006727566011250019, -0.019304847344756126, 0.0023205969482660294, -0.0015815094811841846, 0.02042282745242119, -0.003586198203265667, -0.003515340154990554, -0.005680443253368139, 0.010526337660849094, -0.01606900244951248, 0.016423290595412254, -0.004117632284760475, -0.010794023051857948, 0.016753962263464928, 0.010368876159191132, 0.005664696916937828, -0.0013315384276211262, -0.01464396994560957, -0.004302650224417448, -0.0017744004726409912, 0.0007735323160886765, -0.013006364926695824, 0.0177617184817791, 0.015998143702745438, 0.02114715240895748, 0.016690976917743683, -0.010841261595487595, 0.019210370257496834, 0.0036373732145875692, -0.0006273878389038146, -0.012762298807501793, 0.001337443245574832, 0.006133146584033966, -0.0024386935401707888, 0.0048025925643742085, -0.005959938280284405, 0.00791640393435955, -0.002619774779304862, 0.024249155074357986, -0.017667241394519806, 0.006865345407277346, 0.016250083222985268, 0.013470877893269062, -0.007821926847100258, 0.001833448652178049, 0.007625099737197161, 0.0008655491983518004, -0.0074007161892950535, 0.004025123547762632, 0.011290028691291809, 0.0012409978080540895, -0.0009590422851033509, -0.007404652889817953, 0.006282735615968704, 0.0011130598140880466, 0.017525525763630867, 0.0075345588847994804, 0.019541040062904358, -0.002749681007117033, -0.016502022743225098, -0.011597080156207085, 0.001962370704859495, -0.0011475046630948782, -0.0016464624786749482, 0.007262936793267727, -0.002913047792389989, 0.009975221008062363, 0.012573344632983208, 0.009014702402055264, 0.006148892920464277, 0.00033337666536681354, 0.021997448056936264, 0.027335410937666893, -0.003720040898770094, 0.0056961895897984505, -0.01064443401992321, -0.010919992811977863, -0.040247298777103424, -0.0056135221384465694, -0.0120301004499197, 0.011345140635967255, -0.008376981131732464, -0.02267453446984291, -0.009754774160683155, 0.0164862759411335, -0.04194789007306099, 0.010038205422461033, 0.010612942278385162, 0.022107671946287155, -0.0074440180324018, 0.003639341564849019, -0.014273934066295624, -0.0003132018609903753, 0.012691440992057323, -0.028799807652831078, -0.006829916033893824, -8.531244384357706e-05, 0.003216162323951721, 0.02042282745242119, -0.003962138667702675, 0.027303919196128845, -0.002586314221844077, 0.00240720110014081, 0.01160495262593031, 0.030059505254030228, -0.0032102575059980154, 0.013218938373029232, -0.002176912734284997, -0.018218358978629112, -0.0019357990240678191, -0.029461149126291275, -0.006644898559898138, 0.006125273648649454, -0.020706258714199066, -0.013092969544231892, 0.010880627669394016, -0.010778277181088924, -7.942299816932064e-06, -0.026044221594929695, -0.010935738682746887, -0.0068102334626019, -0.005869397893548012, 0.02053305134177208, 0.004168807528913021, 0.0003776628873310983, 0.012447374872863293, 0.0026827596593648195, 0.00021269677381496876, -0.012329278513789177, -0.030327189713716507, 0.009219402447342873, -0.0034976257011294365, 0.010620814748108387, -0.005503298714756966, -0.017147617414593697, -0.026059968397021294, -0.01184901874512434, -0.012872522696852684, -0.013116588816046715, -0.010738911107182503, -0.009786265902221203, 0.004897069651633501, 0.011187678202986717, 0.004621510859578848, 0.003826327621936798, 0.012321405112743378, -0.007388906553387642, -0.014392031356692314, 0.02240685001015663, -0.02520967274904251, 0.003460228443145752, 0.008526570163667202, 0.00393064646050334, -0.0018797031370922923, 0.0059560020454227924, -0.02650086209177971, -3.210749491699971e-05, -0.013470877893269062, -0.0025528534315526485, -0.01938357762992382, 0.028280183672904968, -0.0026276479475200176, 0.019446562975645065, 0.010628688149154186, 0.005314344074577093, 0.015746204182505608, 0.00976264663040638, -0.011951369233429432, -0.007636909373104572, 0.005530854221433401, -0.02028111182153225, -0.005030912347137928, 0.005298597738146782, -0.021210137754678726, 0.01210095826536417, 0.012297785840928555, 0.015927286818623543, 0.020989689975976944, -0.019588278606534004, -0.0066724540665745735, -0.022343864664435387, -0.0016464624786749482, 0.004361698869615793, 0.011455363593995571, 0.004070393741130829, 0.0008817874477244914, 0.0013138239737600088, -0.004448303021490574, 0.022123416885733604, 0.00440500071272254, -0.0014968735631555319, 0.006452007219195366, -0.0030449223704636097, 0.006554357707500458, 0.0031059389002621174, -0.013683452270925045, 0.0005565299070440233, -0.006353593431413174, 0.013864533044397831, -0.0034779428970068693, -0.023650798946619034, 0.01596665196120739, 0.010471225716173649, 0.010841261595487595, 0.0025981238577514887, 0.0003894725232385099, 0.0020588161423802376, 0.0035291181411594152, 0.0073495409451425076, -0.01058144960552454, 0.019352085888385773, -0.007424335461109877, 0.0011721081100404263, 0.016816945746541023, 0.009274514392018318, -0.0031827015336602926, 0.007018870674073696, 0.0006490388768725097, -0.004310523625463247, 0.01861201412975788, 0.008432092145085335, 0.002704410580918193, 0.036058809608221054, 0.005582029465585947, -0.0014014121843501925, 0.0028756505344063044, 0.0029150161426514387, 0.012368643656373024, 0.0014329046243801713, 0.009487087838351727, 0.011140439659357071, 0.006522865500301123, 0.0012862681178376079, 0.013179573230445385, 0.016171352937817574, -0.014116472564637661, -0.009258768521249294, -0.0024012962821871042, 0.008581681177020073, -0.004735670983791351, 0.008258884772658348, -2.0021052478114143e-05, -0.01813962683081627, 0.004397127777338028, 0.009920109063386917, -0.007943959906697273, -0.007676274981349707, -0.011486856266856194, -0.005574156530201435, -0.015887919813394547, 0.01152622140944004, 0.0021414838265627623, 0.014226695522665977, -0.0026315844152122736, -0.023146919906139374, 0.019415071234107018, -0.00664883479475975, 0.0028185707051306963, -0.00803450122475624, 0.017399556934833527, -0.004204236436635256, -0.008054183796048164, -0.0164705291390419, -0.013667705468833447, 0.015628108754754066, -0.004042838234454393, -0.02042282745242119, 0.003989694640040398, -0.0029189526103436947, 0.017194855958223343, -0.0018678935011848807, -0.00983350444585085, 0.0003562578931450844, 0.021210137754678726, -0.0272566806524992, -0.023351620882749557, -0.018438804894685745, 0.010998724028468132, -0.010486972518265247, -0.0038637248799204826, -0.008085676468908787, 0.00880212802439928, 0.017147617414593697, -0.004948244895786047, 0.004294777289032936, 0.007471574004739523, 0.007959706708788872, 0.0133055429905653, -0.0028185707051306963, -0.0013905867235735059, 0.009400484152138233, -0.011408125050365925, -0.007739259395748377, 0.005204120650887489, 0.0007090712897479534, -0.00471992464736104, -0.0008822795352898538, -0.003905058838427067, -0.00931388046592474, -0.0253513902425766, 0.007821926847100258, -0.03347643092274666, -0.012447374872863293, -0.0028402216266840696, 0.007093665190041065, 0.02801249735057354, 0.0015401756390929222, 0.002277294872328639, -0.00029548737802542746, 0.008125041611492634, 0.008282504044473171, -0.007892784662544727, 0.02420191653072834, 0.013604721054434776, -0.0025252974592149258, -0.0070661092177033424, 0.005818222649395466, 0.028847046196460724, 0.0002920428814832121, 0.015431280247867107, -0.006078035105019808, -0.002102118218317628, 0.00535764591768384, -0.01121917087584734, -0.00912492536008358, 0.004897069651633501, 0.009510708041489124, -0.005054531618952751, -0.02017088793218136, 0.013014238327741623, -0.011447491124272346, -0.0009236133191734552, 0.002621743129566312, -0.0020607844926416874, -0.017840450629591942, -0.01599027030169964, -0.014840797521173954, 0.007668401580303907, -0.005743428133428097, 0.019336339086294174, -0.02812272123992443, 0.003115780185908079, -0.003513371804729104, -0.026579594239592552, 0.006766931619495153, 0.021084168925881386, -0.01950954832136631, -0.011124693788588047, 0.013344908133149147, -0.008321869187057018, 0.0001590612664585933, -0.012360770255327225, -0.01990320347249508, -0.017966419458389282, -0.0005516092060133815, 0.005999303888529539, 0.0031571141444146633, -0.016407545655965805, 0.0037613746244460344, -0.0005063388962298632, -0.018958430737257004, 0.011549840681254864, -0.0034917208831757307, 0.006644898559898138, -0.005365519318729639, 0.004641193896532059, 0.009676042944192886, 0.0003653611638583243, 0.004042838234454393, -0.0007774688419885933, 0.018155373632907867, 0.001312839798629284, 0.0029740643221884966, -0.00017726782243698835, 0.0133134163916111, 0.022627295926213264, -0.014951021410524845, 0.008699778467416763, -0.004507351201027632, -0.026926010847091675, 0.004479795228689909, -0.0010569640435278416, -0.006577976979315281, 0.005306471139192581, -0.02862660028040409, -0.016549261286854744, -0.016171352937817574, 0.0148959094658494, -0.027209442108869553, -0.013463005423545837, 0.005593839101493359, -0.0027969195507466793, 0.011415998451411724, -0.010140555910766125, 0.013754310086369514, 0.014281807467341423, -0.005385201890021563, -0.019210370257496834, -0.00345038715749979, -0.005105706863105297, -0.009723281487822533, -0.015376169234514236, -0.009117052890360355, -0.010148429311811924, 0.01636030711233616, -0.018659252673387527, 0.005459996405988932, -0.015289564616978168, -0.006822043098509312, -0.03253166005015373, 0.013659832067787647, -0.023225652053952217, 0.006141019985079765, 0.001732082455419004, -0.011431744322180748, 0.009518580511212349, -0.0005845778505317867, 0.002171007916331291, 0.005546600557863712, -0.003430704353377223, 0.011707303114235401, 0.013100842013955116, 0.00440500071272254, 0.012722933664917946, 0.0026827596593648195, -0.0007922309450805187, -0.0068062967620790005, -0.009266640990972519, -0.016769707202911377, -0.003844042308628559, 0.011171932332217693, 0.01661224663257599, 0.006274862680584192, -0.03621627017855644, 0.023603560402989388, -0.0031256217043846846, 0.011801780201494694, -0.027666080743074417, -0.0025036465376615524, 0.0006913567776791751, -0.03180733323097229, 0.009542199783027172, 0.006089844740927219, 0.020958198234438896, 0.022202149033546448, -0.026406385004520416, 0.012565471231937408, -0.007664464879781008, 0.003198447870090604, 0.004916752222925425, -0.013037857599556446, -0.0023698038421571255, -0.009408357553184032, 0.01407710649073124, 0.0129906190559268, 0.004597891587764025, -0.019966186955571175, -0.027146456763148308, -0.004397127777338028, 0.011376633308827877, 0.006668517831712961, 0.018060896545648575, 0.026453623548150063, 0.0033362270332872868, 0.015494265593588352, 0.0063811494037508965, -0.017194855958223343, 0.00024111375387292355, -0.0002543996088206768, -0.014864416792988777, 0.01057357620447874, -0.01684843935072422, 0.011376633308827877, -0.02269028127193451, -0.029476895928382874, 0.013588974252343178, -0.0018039245624095201, 0.009888616390526295, 0.007203888613730669, -0.007688084617257118, -0.00576311070472002, 0.0031827015336602926, 0.0013699198607355356, -0.02179274708032608, -0.0054442500695586205, -0.027540111914277077, -0.026800040155649185, -0.0023481526877731085, -0.0011189646320417523, 0.0037101993802934885, 0.007975452579557896, 0.0053025344386696815, -0.003753501456230879, -0.009164291433990002, -0.026910264045000076, 0.00936899147927761, 0.013415765948593616, 0.016879931092262268, -0.005263168830424547, 0.02483176440000534, 0.026437876746058464, -0.00497186416760087, -0.012581217102706432, -0.0008970415801741183, 0.004731734283268452, -0.006436261348426342, -0.0046766228042542934, -0.021194390952587128, 0.0008581681759096682, 0.012636329047381878, -0.02623317763209343, -0.002853999612852931, 0.016391798853874207, -0.0012272198218852282, -0.009636676870286465, -0.02925644814968109, 0.0072747464291751385, 0.003113812068477273, -0.012589090503752232, 0.01798216626048088, 0.01750977896153927, -0.011927749961614609, -0.005747364833950996, 0.017714479938149452, 0.0031236533541232347, -0.003505498869344592, 0.006459880620241165, 0.003844042308628559, 0.009164291433990002, -0.009266640990972519, -0.004786846227943897, -0.02114715240895748, -0.01128215529024601, 0.022233640775084496, 0.017431048676371574, -0.0014023963594809175, 0.0018462424632161856, 0.021588046103715897, -0.020485812798142433, -0.007546368520706892, -0.005416694562882185, 0.013392146676778793, 0.0024741224478930235, -0.020501557737588882, -0.00018526393978390843, 0.01146323699504137, -0.0008822795352898538, 0.018297089263796806, -0.011085327714681625, 0.007841610349714756, 0.018470298498868942, -0.008644666522741318, 0.01991894841194153, -0.0073495409451425076, -0.003745628520846367, -0.003430704353377223, -0.005566283129155636, 0.01032951008528471, -0.014990386553108692, -0.023855499923229218, -0.005743428133428097, 0.0004792750987689942, 0.026406385004520416, 0.012211182154715061, -0.0018472266383469105, 0.015328929759562016, 0.02143058367073536, 0.0008926129667088389, 0.0126756951212883, -0.00969966221600771, -0.003688548458740115, -0.0064638168551027775, -0.01621859148144722, -0.011447491124272346, -0.0013276018435135484, -0.019100146368145943, 0.0018747824942693114, 0.01686418429017067, 0.0026119016110897064, 0.016313068568706512, 0.01798216626048088, 0.008872986771166325, 0.006105591077357531, 0.015439153648912907, -0.0076605286449193954, 0.018706491217017174, -0.015998143702745438, 0.002387518296018243, -0.010731038637459278, -0.011266409419476986, 0.0139353908598423, -0.006613405887037516, -0.018076643347740173, 0.03382284939289093, -0.002485932083800435, 0.005208057351410389, 0.0013423638883978128, 0.01621859148144722, 0.013636212795972824, -0.026658324524760246, -0.01418733038008213, 0.005499362014234066, -0.0107625313103199, 0.03220098838210106, 0.0062394337728619576, -0.0042081731371581554, -0.0029563498683273792, -0.01597452536225319, -0.015336803160607815, 0.005235612858086824, -0.012612709775567055, -0.014021995477378368, -0.010872754268348217, 0.005853651557117701, 0.012746552936732769, -0.012360770255327225, -0.015021879225969315, 0.0012459184508770704, -0.013722817413508892, -0.021131407469511032, -0.024658557027578354, 0.008479331620037556, 0.020470065996050835, 0.010022459551692009, -0.003324417397379875, -0.04541205242276192, 0.011486856266856194, -0.009022574871778488, -0.0019505610689520836, -0.00874701701104641, 0.032626137137413025, 0.0013561418745666742, 0.003351973369717598, -0.025130942463874817, 0.0017409397987648845, 0.008723397739231586, 0.00267882295884192, 0.0018462424632161856, -0.0021257377229630947, -0.005707999225705862, 0.0005073230131529272, -0.0011602984741330147, -0.009746900759637356, 0.009707535617053509, 0.01724209450185299, -0.015163594856858253, 0.0022103735245764256, -0.006078035105019808, -0.034043293446302414, 0.004704178776592016, 0.004263285081833601, -0.017147617414593697, -0.02089521288871765, -0.001467349473387003, -0.02839040756225586, -0.013494497165083885, -0.007987262681126595, -0.0005826095584779978, -0.0228005051612854, -0.01388027984648943, 0.005648951046168804, -0.017305077984929085, 0.009739027358591557, 0.03265762701630592, 0.010794023051857948, 0.012935507111251354, 0.009077686816453934, 0.010534211061894894, 0.0009934870759025216, 0.005223803222179413, -0.006577976979315281, 0.011612826026976109, -0.014840797521173954, -0.013392146676778793, 0.016942916437983513, -0.018942683935165405, 0.013337035663425922, -0.004349889233708382, -0.010872754268348217, -0.00409007677808404, -0.006637025158852339, 0.010235032998025417, 0.012242673896253109, 0.020706258714199066, -0.02432788535952568, 0.0077668153680861, 0.01569896563887596, 0.011156185530126095, 0.003113812068477273, -0.007247190456837416, 0.004897069651633501, 0.0015116356080397964, 0.0038066450506448746, -0.0015047467313706875, -0.025256913155317307, 0.0030606684740632772, 0.00044384613283909857, -0.010534211061894894, 0.005641077645123005, 0.024028709158301353, -0.006617342587560415, 0.0011839177459478378, -0.012187561951577663, -0.015344676561653614, 0.011982861906290054, 0.0027359030209481716, -0.023855499923229218, 0.007522749248892069, -0.036814626306295395, -0.015289564616978168, -0.01670672371983528, 0.01873798295855522, 0.009904362261295319, 0.00994372833520174, -0.004275094717741013, -0.01887969858944416, 0.003540927777066827, -0.0023934231139719486, 0.00471992464736104, 0.05882782116532326, -0.01001458615064621, 0.0009196767932735384, -0.021761255338788033, 0.008652538992464542, 0.0008680095197632909, 0.0019564658869057894, -0.024296393617987633, -0.0054875523783266544, -0.00045836216304451227, 0.001520492834970355, -0.0001845258375396952, -0.008298249915242195, -0.005641077645123005, -0.004385318141430616, 0.011297902092337608, -0.006282735615968704, 0.004523097071796656, -0.00930600706487894, -0.0171161238104105, -0.009746900759637356, -0.004590018652379513, 0.009006829001009464, 0.001047122641466558, 0.024422364309430122, -0.0053025344386696815, -0.012510359287261963, 0.006050479132682085, -0.008518696762621403, -0.03079957701265812, 0.021635284647345543, -0.024154677987098694, 0.002143452176824212, -6.864361057523638e-05, -0.01064443401992321, 0.003464164910838008, -0.03435821831226349, -0.0048773870803415775, 0.016391798853874207, -0.009613057598471642, 0.0142581881955266, 0.003554705763235688, 0.011959242634475231, 0.010864880867302418, -0.014368412084877491, -0.006829916033893824, 0.005830032285302877, -0.015132103115320206, -0.0016100493958219886, -0.023335875943303108, 0.000774024345446378, 0.005168691743165255, -0.005369455553591251, 0.03240568935871124, -0.021950209513306618, -0.004530970472842455, -0.0025882823392748833, 0.021603792905807495, -0.0055387276224792, 0.014053487218916416, 0.005908763501793146, -0.0026611085049808025, -0.00881000142544508, 0.0003259956429246813, 0.005841841921210289, 0.017588511109352112, -0.011297902092337608, -0.008518696762621403, 0.004818338435143232, -0.015494265593588352, 0.02484751120209694, -0.01000671274960041, 0.01242375560104847, 0.01089637354016304, 0.02064327336847782, -0.0016740183345973492, -0.020580289885401726, 0.0016405576607212424, 0.025697806850075722, -0.008770636282861233, 0.0164705291390419, -0.0020568480249494314, -0.000957073993049562, -0.00943984929472208, -0.018690744414925575, -0.02950838766992092, -0.006731502711772919, -0.0016622086986899376, 0.0310515146702528, -0.006483499892055988, 0.01952529326081276, 0.009794139303267002, -0.003668865654617548, -0.003046890487894416, 0.004901006352156401, -0.0035763566847890615, -0.011691557243466377, -0.002745744539424777, 0.00855806190520525, 0.0048773870803415775, -0.0005811333539895713, 0.011549840681254864, -0.01887969858944416, 0.0007700878195464611, 0.03131920099258423, -0.00640083197504282, 0.008888732641935349, 0.0071251573972404, 0.005971747916191816, -0.009101306088268757, -0.006680327467620373, -0.011723048985004425, -0.006558294408023357, -0.014478635042905807, 0.0017724321223795414, 0.001913163810968399, 0.002068657660856843, 0.003753501456230879, -0.002586314221844077, 0.007566051557660103, -0.02711496502161026, 0.008896606042981148, 0.0164705291390419, -0.0009664233075454831, -0.018060896545648575, -0.015801316127181053, -0.0016877963207662106, 0.004830148071050644, 0.0060386694967746735, -0.010620814748108387, 0.002643394051119685, 0.01450225431472063, 0.02938241697847843, 0.028453391045331955, 0.025036465376615524, 0.00873914361000061, 0.006066225469112396, -0.012219054624438286, 0.009203656576573849, 0.002048974856734276, -0.023398859426379204, 0.0060386694967746735, -0.005424567498266697, -0.010872754268348217, 0.015683220699429512, 0.0011061708210036159, -0.0019239893881604075, -0.009715408086776733, 0.006322101224213839, -0.0145809855312109, -0.0046687498688697815, 0.011990734376013279, 0.03867267817258835, -0.0028146340046077967, -0.005853651557117701, -0.01762000285089016, -0.01147111039608717, -0.020044919103384018, 0.01419520378112793, 0.013274050317704678, -0.010723165236413479, 0.017210600897669792, 0.007168459706008434, 0.0046254475601017475, 0.0142660615965724, -0.010518464259803295, -0.012951252982020378, -0.00671969261020422, 0.00034986098762601614, 0.013966883532702923, 0.0016070969868451357, 0.014203076250851154, -0.001782273524440825, 0.0068810912780463696, 0.01412434596568346, -0.015478518791496754, 0.02128886803984642, -0.015903666615486145, 0.008251011371612549, -0.021225884556770325, 0.005334026645869017, -4.161057586316019e-05, 0.03602731600403786, -0.013762182556092739, 0.00440500071272254, -0.0038401056081056595, -0.003954265732318163, 0.009014702402055264, -0.0002566139155533165, 0.014691208489239216, -0.0002895825309678912, -0.028185706585645676, -0.002828411990776658, 0.005818222649395466, -0.015171468257904053, 0.008006945252418518, -0.020627528429031372, 0.016407545655965805, -0.018029404804110527, -0.024170424789190292, 0.014533746987581253, 0.0011229012161493301, -0.02028111182153225, 0.0004876402672380209
                ],
                "page_number": 105
            }
        ]
    }

지식 원본 만들기

지식 원본은 원본 데이터에 대한 재사용 가능한 참조입니다. 지식 원본 - 만들기(REST API)를 사용하여 earth-knowledge-source 인덱스를 대상으로 하는 earth-at-night라는 지식 원본을 정의합니다.

searchIndexParameters.sourceDataSelect는 검색 및 인용을 위해 액세스할 수 있는 인덱스 필드를 지정합니다. 이 예제에는 응답에 길고 해석할 수 없는 포함이 발생하는 것을 방지하기 위해 사람이 읽을 수 있는 필드만 포함됩니다.

### Create a knowledge source
POST {{search-url}}/knowledgesources?api-version={{api-version}}  HTTP/1.1
    Content-Type: application/json
    Authorization: Bearer {{token}}

    {
        "name": "{{knowledge-source-name}}",
        "description": "This knowledge source pulls from a search index that contains pages from the Earth at Night e-book.",
        "kind": "searchIndex",
        "searchIndexParameters": {
            "searchIndexName": "{{index-name}}",
            "sourceDataSelect": "id, page_chunk, page_number"
        }
    }

지식 에이전트 만들기

쿼리 시 earth-knowledge-source 및 gpt-5-mini 배포를 대상으로 지정하려면 지식 에이전트가 필요합니다. 지식 에이전트 - 만들기(REST API)를 사용하여 이전에 earth-knowledge-agent 변수를 사용하여 지정한 @knowledge-agent-name라는 에이전트를 정의합니다.

knowledgeSources.rerankerThreshold는 순위 재지정 점수가 2.5이거나 이보다 낮은 응답을 제외하여 의미론적 관련성을 보장합니다. 한편, outputConfiguration.modality는 검색된 문서를 인용하는 자연어 응답을 사용하도록 설정하도록 answerSynthesis로 설정됩니다.

### Create a knowledge agent
PUT {{search-url}}/agents/{{knowledge-agent-name}}?api-version={{api-version}}  HTTP/1.1
    Content-Type: application/json
    Authorization: Bearer {{token}}

    {
        "name": "{{knowledge-agent-name}}",
        "knowledgeSources": [
            {
            "name": "{{knowledge-source-name}}",
            "rerankerThreshold": 2.5
            }
        ],
        "models": [
            {
                "kind": "azureOpenAI",
                "azureOpenAIParameters": {
                    "resourceUri": "{{aoai-url}}",
                    "deploymentId": "{{aoai-gpt-deployment}}",
                    "modelName": "{{aoai-gpt-model}}"
                }
            }
        ],
        "outputConfiguration": {
            "modality": "answerSynthesis"
        }
    }

검색 파이프라인 실행

에이전트 검색을 실행할 준비가 완료되었습니다. 지식 검색 - 검색(REST API)을 사용하여 쿼리를 하위 쿼리로 분해하고, earth-knowledge-agent 인덱스의 텍스트 및 벡터 필드에 대해 하위 쿼리를 실행하고, 결과의 순위를 지정하고 병합하는 두 부분으로 구성된 사용자 쿼리를 earth-at-night로 보냅니다.

### Run agentic retrieval
POST {{search-url}}/agents('{{knowledge-agent-name}}')/retrieve?api-version={{api-version}}  HTTP/1.1
    Content-Type: application/json
    Authorization: Bearer {{token}}

    {
        "messages": [
            {
                "role": "user",
                "content": [
                    {
                        "type": "text",
                        "text": "Why do suburban belts display larger December brightening than urban cores even though absolute light levels are higher downtown? Why is the Phoenix nighttime street grid is so sharply visible from space, whereas large stretches of the interstate between midwestern cities remain comparatively dim?"
                    }
                ]
            }
        ],
        "knowledgeSourceParams": [
            {
                "knowledgeSourceName": "{{knowledge-source-name}}",
                "kind": "searchIndex"
            }
        ]
    }

출력은 다음 JSON과 유사해야 합니다. 여기서는 다음과 같습니다.

• response는 검색된 문서를 인용하는 쿼리에 대한 합성된 LLM 생성 답변을 제공합니다. 응답 합성을 사용할 수 없는 경우 이 섹션에는 문서에서 직접 추출된 콘텐츠가 포함됩니다.

• activity는 gpt-5-mini 배포에서 생성된 하위 쿼리와 의미론적 순위 지정, 쿼리 계획 및 응답 합성에 사용되는 토큰을 포함하여 검색 프로세스 중에 수행된 단계를 추적합니다.

• references 에는 응답에 기여한 문서가 나열되고 각 문서는 해당 docKey문서로 식별됩니다.

{
  "response": [
    {
      "content": [
        {
          "type": "text",
          "text": "Suburban belts display larger December brightening than urban cores despite higher absolute light levels downtown because the urban grid encourages outward growth along city borders, fueled by widespread personal automobile use, leading to extensive suburban and residential municipalities linked by surface streets and freeways. This expansion results in increased lighting in suburban areas during December. The Phoenix nighttime street grid is sharply visible from space due to its regular grid layout of city blocks and streets, with major transportation corridors like Grand Avenue and brightly lit commercial properties at intersections, which create distinct lighting patterns. In contrast, large stretches of interstate highways between Midwestern cities remain comparatively dim because they lack the dense, grid-like urban development and associated lighting found in metropolitan areas like Phoenix [ref_id:0][ref_id:1]."
        }
      ]
    }
  ],
  "activity": [
    {
      "type": "modelQueryPlanning",
      "id": 0,
      "inputTokens": 2079,
      "outputTokens": 121,
      "elapsedMs": 2887
    },
    {
      "type": "searchIndex",
      "id": 1,
      "knowledgeSourceName": "earth-knowledge-source",
      "queryTime": "2025-08-25T16:23:17.832Z",
      "count": 0,
      "elapsedMs": 1065,
      "searchIndexArguments": {
        "search": "Reasons for larger December brightening in suburban belts compared to urban cores despite higher downtown light levels",
        "filter": null
      }
    },
    {
      "type": "searchIndex",
      "id": 2,
      "knowledgeSourceName": "earth-knowledge-source",
      "queryTime": "2025-08-25T16:23:18.139Z",
      "count": 2,
      "elapsedMs": 298,
      "searchIndexArguments": {
        "search": "Factors making Phoenix nighttime street grid sharply visible from space",
        "filter": null
      }
    },
    {
      "type": "searchIndex",
      "id": 3,
      "knowledgeSourceName": "earth-knowledge-source",
      "queryTime": "2025-08-25T16:23:18.332Z",
      "count": 0,
      "elapsedMs": 189,
      "searchIndexArguments": {
        "search": "Reasons why large stretches of interstate between Midwestern cities are comparatively dim at night from space",
        "filter": null
      }
    },
    {
      "type": "semanticReranker",
      "id": 4,
      "inputTokens": 2349
    },
    {
      "type": "modelAnswerSynthesis",
      "id": 5,
      "inputTokens": 3216,
      "outputTokens": 155,
      "elapsedMs": 2274
    }
  ],
  "references": [
    {
      "type": "searchIndex",
      "id": "0",
      "activitySource": 2,
      "sourceData": null,
      "rerankerScore": 2.6642752,
      "docKey": "earth_at_night_508_page_104_verbalized"
    },
    {
      "type": "searchIndex",
      "id": "1",
      "activitySource": 2,
      "sourceData": null,
      "rerankerScore": 2.5905457,
      "docKey": "earth_at_night_508_page_105_verbalized"
    }
  ]
}

자원을 정리하세요

본인 소유의 구독에서 작업하는 경우 자신이 만든 리소스가 여전히 필요한지 여부를 결정하여 프로젝트를 완료하는 것이 좋습니다. 실행 상태로 방치된 리소스는 비용이 들 수 있습니다.

Azure Portal의 왼쪽 창에서 모든 리소스 또는 리소스 그룹을 선택하여 Azure AI 검색 및 Azure AI Foundry 리소스를 관리할 수 있습니다.

그렇지 않은 경우 다음 코드를 실행하여 이 빠른 시작에서 만든 개체를 삭제합니다.

지식 에이전트 삭제

### Delete the knowledge agent
DELETE {{search-url}}/agents/{{knowledge-agent-name}}?api-version={{api-version}}  HTTP/1.1
    Content-Type: application/json
    Authorization: Bearer {{token}}

지식 원본 삭제

### Delete the knowledge source
DELETE {{search-url}}/knowledgesources('{{knowledge-source-name}}')?api-version={{api-version}}  HTTP/1.1
    Content-Type: application/json
    Authorization: Bearer {{token}}

검색 인덱스 삭제

### Delete the index
DELETE {{search-url}}//indexes/{{index-name}}?api-version={{api-version}}  HTTP/1.1
    Content-Type: application/json
    Authorization: Bearer {{token}}