메모
Azure AI 검색 Azure 포털, REST API 및 Azure SDK 통해 사용할 수 있습니다. 또한 엔터프라이즈 콘텐츠를 Microsoft Foundry 포털의 에이전트에 대해 재사용 가능한 사용 권한 인식 기술 자료로 변환하는 관리되는 기술 계층인 Foundry IQ를 뒷받침합니다.
Important
이러한 기능과 기능은 2026-05-01-preview REST API의 일부입니다. 2026-05-01-preview는 Azure 구독의 일부로 사용자에게 라이선스가 부여되며, Microsoft Product Terms의 "미리 보기"에 적용되는 약관, Microsoft 제품 및 서비스 데이터 보호 부록("DPA") 및 Microsoft Azure 미리 보기용 추가 사용 약관의 적용을 받습니다.
2026-05-01-preview는 다른 Microsoft 서비스 및 타사 서비스에 대한 연결을 지원합니다. 이러한 서비스의 사용은 해당 약관의 적용을 받으며 Azure 규정 준수 경계 외부의 데이터 처리 또는 스토리지뿐만 아니라 Azure 규정 준수 경계로 데이터가 유입될 수 있습니다.
Fabric IQ에 연결하면 비용이 발생할 수 있으며, 데이터는 Azure 규정 준수 경계 외부로 전송되고 해당 서비스 약관 및 데이터 처리 정책에 따라 처리될 수 있습니다. 데이터가 조직의 규정 준수 및 지리적 경계와 관련된 의미를 벗어나는지 여부와 적절한 권한, 경계 및 승인이 프로비전되는지를 관리하는 것은 사용자의 책임입니다.
특정 사용 사례의 컨텍스트에서 빌드한 애플리케이션을 신중하게 검토하고 테스트하고 모든 적절한 결정 및 사용자 지정을 수행할 책임이 있습니다. 여기에는 메타프롬프트, 콘텐츠 필터 또는 기타 안전 시스템과 같은 책임 있는 AI 완화를 구현하고 애플리케이션이 적절한 품질, 안정성, 보안 및 신뢰성 표준을 충족하도록 보장하는 것이 포함됩니다. 자세한 내용은 Azure AI 검색 투명성 정보를 참고하세요.
Fabric 데이터 에이전트 지식 소스(미리 보기)는 Microsoft Fabric 데이터 에이전트를 Azure AI 검색의 에이전트형 검색 파이프라인에 연결합니다. 기술 원본은 독립적으로 생성되고, 기술 자료에서 참조되며, 런타임에 기술 자료를 쿼리할 때 접지 데이터로 사용됩니다.
데이터 에이전트는 가상 분석가 역할을 하며 Microsoft Fabric 데이터에 대해 쿼리를 생성하고 실행하여 자연어 답변, 테이블 및 차트를 반환합니다. 이렇게 하면 Fabric 데이터 에이전트의 지식 소스를 변화가 잦은 데이터와 분석용 쿼리에 유용하게 활용할 수 있습니다.
인덱싱된 지식 원본과 달리 Fabric 데이터 에이전트 지식 원본은 검색 시 라이브 데이터를 직접 쿼리합니다. 수집 파이프라인이 필요하지 않습니다. 쿼리에는 최종 사용자를 대신하여 검색 엔진이 Fabric 데이터 에이전트를 쿼리하는 데 사용하는 최종 사용자 액세스 토큰이 필요합니다.
사용량 지원
| Azure Portal | Microsoft Foundry 포털 | .NET SDK | Python SDK | Java SDK | JavaScript SDK | REST API |
|---|---|---|---|---|---|---|
| ❌ | ✔️ | ✔️ | ✔️ | ✔️ | ✔️ | ✔️ |
사전 요구 사항
Azure AI 검색 서비스는 모든 지역에서 에이전트 검색 기능을 제공합니다.
데이타 에이전트 있는 Microsoft Fabric 작업 영역입니다.
검색 서비스 및 작업 영역은 동일한 Microsoft Entra ID 테넌트에 있어야 합니다.
지식 소스를 생성할 수 있는 권한 사용자 계정에 할당된 Search Service 기여자 역할을 사용하여 키 없는 인증을 구성하거나 API 키를 사용합니다.
- 최신
Azure.Search.Documents미리 보기 패키지:dotnet add package Azure.Search.Documents --prerelease
- 최신
azure-search-documents미리 보기 패키지:pip install --pre azure-search-documents
- Search Service REST API의 2026-05-01-preview 버전입니다.
데이터 거버넌스 및 규정 준수
Fabric IQ는 작업 영역의 Microsoft Fabric 규정 준수 경계 내에서 요청을 처리합니다. Fabric IQ를 통해 에이전트 쿼리를 라우팅하는 경우 다음 약정이 적용됩니다.
데이터 저장 위치
Fabric IQ는 Microsoft Fabric 작업 영역이 있는 지역 내에서 데이터를 검색하고 처리합니다. 쿼리를 실행하는 동안 데이터는 지역 경계를 넘지 않습니다. 해당 지역 및 해당 규정 준수 범위는 작업 영역 위치에 따라 결정됩니다. 각 지역이 충족하는 지원되는 지역 및 규정 준수 프레임워크 목록은 Microsoft Fabric 지역 가용성 참조하세요.
메모
Azure AI 검색 서비스가 Fabric 작업 영역과 다른 Azure 지역에 있는 경우 쿼리 결과는 지역 간으로 반환됩니다. 다른 지역의 Fabric 작업 영역을 연결하기 전에 Microsoft Fabric 지역 가용성 및 조직의 데이터 상주 요구 사항을 검토합니다.
규정 준수 인증
Fabric IQ는 작업 영역 영역에 대한 Microsoft Fabric 규정 준수 인증을 상속합니다. 규정 준수 설명서, 감사 보고서 및 각 지역에 적용되는 프레임워크는 Microsoft Fabric 지역 가용성 참조하세요.
기존 지식 원본 확인
기술 자료는 재사용 가능한 최상위 개체입니다. 기존 지식 원본에 대해 아는 것은 새 개체를 다시 사용하거나 이름을 지정하는 데 유용합니다.
다음 코드를 실행하여 이름 및 형식별로 기술 자료를 나열합니다.
// List knowledge sources by name and type
using Azure.Search.Documents.Indexes;
var indexClient = new SearchIndexClient(new Uri(searchEndpoint), credential);
var knowledgeSources = indexClient.GetKnowledgeSourcesAsync();
Console.WriteLine("Knowledge Sources:");
await foreach (var ks in knowledgeSources)
{
Console.WriteLine($" Name: {ks.Name}, Type: {ks.GetType().Name}");
}
# List knowledge sources by name and type
from azure.core.credentials import AzureKeyCredential
from azure.search.documents.indexes import SearchIndexClient
index_client = SearchIndexClient(endpoint = "search_url", credential = AzureKeyCredential("api_key"))
for ks in index_client.list_knowledge_sources():
print(f" - {ks.name} ({ks.kind})")
### List knowledge sources by name and type
GET {{search-url}}/knowledgesources?api-version={{api-version}}&$select=name,kind
api-key: {{api-key}}
참조:기술 자료 - 목록
이름으로 단일 기술 원본을 반환하여 JSON 정의를 검토할 수도 있습니다.
using Azure.Search.Documents.Indexes;
using System.Text.Json;
var indexClient = new SearchIndexClient(new Uri(searchEndpoint), credential);
// Specify the knowledge source name to retrieve
string ksNameToGet = "earth-knowledge-source";
// Get its definition
var knowledgeSourceResponse = await indexClient.GetKnowledgeSourceAsync(ksNameToGet);
var ks = knowledgeSourceResponse.Value;
// Serialize to JSON for display
var jsonOptions = new JsonSerializerOptions
{
WriteIndented = true,
DefaultIgnoreCondition = System.Text.Json.Serialization.JsonIgnoreCondition.Never
};
Console.WriteLine(JsonSerializer.Serialize(ks, ks.GetType(), jsonOptions));
# Get a knowledge source definition
from azure.core.credentials import AzureKeyCredential
from azure.search.documents.indexes import SearchIndexClient
import json
index_client = SearchIndexClient(endpoint = "search_url", credential = AzureKeyCredential("api_key"))
ks = index_client.get_knowledge_source("knowledge_source_name")
print(json.dumps(ks.as_dict(), indent = 2))
### Get a knowledge source definition
GET {{search-url}}/knowledgesources/{{knowledge-source-name}}?api-version={{api-version}}
api-key: {{api-key}}
참조:기술 자료 - 가져오기
다음 JSON은 Fabric 데이터 에이전트 지식 원본에 대한 예제 응답입니다.
{
"name": "my-fabric-data-agent-ks",
"kind": "fabricDataAgent",
"description": "A Fabric Data Agent knowledge source.",
"encryptionKey": null,
"fabricDataAgentParameters": {
"workspaceId": "00000000-0000-0000-0000-000000000000",
"dataAgentId": "00000000-0000-0000-0000-000000000001"
}
}
기술 자료 만들기
다음 코드를 실행하여 Fabric 데이터 에이전트 지식 원본을 만듭니다.
using Azure;
using Azure.Search.Documents.Indexes;
using Azure.Search.Documents.Indexes.Models;
Uri searchEndpoint = new Uri("<search-service-url>");
AzureKeyCredential credential = new AzureKeyCredential("<api-key>");
var indexClient = new SearchIndexClient(searchEndpoint, credential);
var fabricDataAgent = new FabricDataAgentKnowledgeSource(
"<knowledge-source-name>",
new FabricDataAgentKnowledgeSourceParameters("<fabric-workspace-id>", "<fabric-data-agent-id>"))
{
Description = "A Fabric Data Agent knowledge source."
};
await indexClient.CreateOrUpdateKnowledgeSourceAsync(fabricDataAgent);
from azure.core.credentials import AzureKeyCredential
from azure.search.documents.indexes import SearchIndexClient
from azure.search.documents.indexes.models import (
FabricDataAgentKnowledgeSource,
FabricDataAgentKnowledgeSourceParameters,
)
index_client = SearchIndexClient(
endpoint="<search-service-url>",
credential=AzureKeyCredential("<api-key>")
)
knowledge_source = FabricDataAgentKnowledgeSource(
name="<knowledge-source-name>",
description="A Fabric Data Agent knowledge source.",
fabric_data_agent_parameters=FabricDataAgentKnowledgeSourceParameters(
workspace_id="<fabric-workspace-id>",
data_agent_id="<fabric-data-agent-id>"
)
)
index_client.create_or_update_knowledge_source(knowledge_source)
### Create a Fabric Data Agent knowledge source
PUT {{search-url}}/knowledgesources/my-fabric-data-agent-ks?api-version=2026-05-01-preview
api-key: {{api-key}}
Content-Type: application/json
{
"name": "my-fabric-data-agent-ks",
"kind": "fabricDataAgent",
"description": "A Fabric Data Agent knowledge source.",
"fabricDataAgentParameters": {
"workspaceId": "{fabric-workspace-id}",
"dataAgentId": "{fabric-data-agent-id}"
}
}
기술 자료에 할당
기술 자료에 만족하는 경우 기술 자료에 추가합니다.
Important
Fabric Data Agent 지식 원본은 minimal검색 추론 작업을 지원하지 않습니다. 대신 low 또는 medium를 사용하십시오.
지식 베이스 질의
기술 자료가 구성되면 검색 작업 또는 MCP 엔드포인트를 호출하여 Fabric 데이터 에이전트 콘텐츠를 쿼리합니다. 이 기술 원본에는 고유한 쿼리 시간 권한 적용 및 응답 특성이 있습니다.
쿼리 시 사용 권한 적용
Fabric 데이터 에이전트 참조 자료 원본은 OBO(On-Behalf-of) 토큰 흐름을 사용합니다. 검색 요청의 헤더에 x-ms-query-source-authorization 최종 사용자 액세스 토큰을 전달합니다. 토큰은 Azure AI 검색 대상 그룹(https://search.azure.com/.default)으로 범위가 지정되어야 합니다. 검색 엔진은 이 토큰을 Microsoft Fabric 범위의 토큰으로 교환하고 이를 사용하여 최종 사용자를 대신하여 Fabric 데이터 에이전트를 쿼리합니다.
검색 요청에도 표준 Azure AI 검색 인증이 필요합니다. 토큰은 x-ms-query-source-authorization 별도로 전달되며 서비스 인증을 대체하지 않습니다.
토큰 전달에 대한 지침은 쿼리 시 권한 적용(미리 보기)을 참조하세요.
Fabric Data Agent 전용 응답 필드
Fabric 데이터 에이전트 지식 원본은 각 sourceData 항목의 references 개체에 결과를 반환하고, activity 배열에 쿼리 진단을 반환합니다.
sourceData 포함되어 있습니다:
-
fabricAnswer: 데이터 에이전트의 자연어 답변입니다. -
fabricEmbeddedResources: 쿼리에서 구조화된 출력을 생성할 때 테이블 및 차트와 같은 포함된 모든 리소스입니다.
다음 예제에서는 Fabric 데이터 에이전트 기술 자료 참조 및 해당 활동 레코드를 포함하는 검색 응답을 보여 있습니다. 검색 응답 해석에 대한 광범위한 지침은 응답 검토를 참조하세요.
Tip
참조용 sourceData를 받으려면 검색 요청의 includeReferenceSourceData 내에 있는 지식 소스 항목에서 true를 knowledgeSourceParams(으)로 설정하세요.
{
"response": [
// ... Response omitted for brevity
],
"activity": [
{
"type": "fabricDataAgent",
"id": 1,
"knowledgeSourceName": "my-fabric-data-agent-ks",
"queryTime": "2026-05-11T19:37:11.600Z",
"count": 1,
"fabricDataAgentArguments": {
"search": "my query"
}
},
{
// ... Additional activity records omitted for brevity
}
],
"references": [
{
"type": "fabricDataAgent",
"id": "0",
"activitySource": 1,
"rerankerScore": 2.57,
"workspaceId": "00000000-0000-0000-0000-000000000000",
"dataAgentId": "00000000-0000-0000-0000-000000000001",
"sourceData": {
"fabricAnswer": "There were 47 airline delays in Q1 2026. The majority occurred on transcontinental routes.",
"fabricEmbeddedResources": [
{
"title": "Q1 2026 Airline Operations Report",
"mimeType": "text/csv",
"content": "Route | Delays | Avg. Delay (min)\n--- | --- | ---\nJFK–LAX | 18 | 42\nORD–MIA | 12 | 31"
}
]
}
},
{
// ... Additional references omitted for brevity
}
]
}
지식 원본 삭제
기술 자료를 삭제하려면 먼저 이를 참조하는 기술 자료를 삭제하거나 기술 자료 정의를 업데이트하여 참조를 제거해야 합니다. 인덱스 및 인덱서 파이프라인을 생성하는 기술 원본의 경우 생성된 모든 개체 도 삭제됩니다. 그러나 기존 인덱스로 기술 원본을 만든 경우 인덱스가 삭제되지 않습니다.
사용 중인 기술 원본을 삭제하려고 하면 작업이 실패하고 영향을 받는 기술 자료 목록이 반환됩니다.
기술 자료를 삭제하려면 다음을 수행합니다.
검색 서비스의 모든 기술 자료 목록을 가져옵니다.
using Azure.Search.Documents.Indexes; var indexClient = new SearchIndexClient(new Uri(searchEndpoint), credential); var knowledgeBases = indexClient.GetKnowledgeBasesAsync(); Console.WriteLine("Knowledge Bases:"); await foreach (var kb in knowledgeBases) { Console.WriteLine($" - {kb.Name}"); }예제 응답은 다음과 같을 수 있습니다.
{ "@odata.context": "https://my-search-service.search.windows.net/$metadata#knowledgebases(name)", "value": [ { "name": "my-kb" }, { "name": "my-kb-2" } ] }개별 기술 자료 정의를 가져와서 기술 자료 참조를 확인합니다.
using Azure.Search.Documents.Indexes; using System.Text.Json; var indexClient = new SearchIndexClient(new Uri(searchEndpoint), credential); // Specify the knowledge base name to retrieve string kbNameToGet = "earth-knowledge-base"; // Get a specific knowledge base definition var knowledgeBaseResponse = await indexClient.GetKnowledgeBaseAsync(kbNameToGet); var kb = knowledgeBaseResponse.Value; // Serialize to JSON for display string json = JsonSerializer.Serialize(kb, new JsonSerializerOptions { WriteIndented = true }); Console.WriteLine(json);예제 응답은 다음과 같을 수 있습니다.
{ "Name": "earth-knowledge-base", "KnowledgeSources": [ { "Name": "earth-knowledge-source" } ], "Models": [ {} ], "RetrievalReasoningEffort": {}, "OutputMode": {}, "ETag": "\u00220x8DE278629D782B3\u0022", "EncryptionKey": null, "Description": null, "RetrievalInstructions": null, "AnswerInstructions": null }기술 자료를 삭제하거나 여러 기술 자료가 있는 경우 기술 자료를 업데이트하여 원본을 제거합니다. 이 예제에서는 삭제를 보여줍니다.
using Azure.Search.Documents.Indexes; var indexClient = new SearchIndexClient(new Uri(searchEndpoint), credential); await indexClient.DeleteKnowledgeBaseAsync(knowledgeBaseName); System.Console.WriteLine($"Knowledge base '{knowledgeBaseName}' deleted successfully.");기술 자료를 삭제합니다.
await indexClient.DeleteKnowledgeSourceAsync(knowledgeSourceName); System.Console.WriteLine($"Knowledge source '{knowledgeSourceName}' deleted successfully.");
검색 서비스의 모든 기술 자료 목록을 가져옵니다.
# Get knowledge bases from azure.core.credentials import AzureKeyCredential from azure.search.documents.indexes import SearchIndexClient index_client = SearchIndexClient(endpoint = "search_url", credential = AzureKeyCredential("api_key")) print("Knowledge Bases:") for kb in index_client.list_knowledge_bases(): print(f" - {kb.name}")예제 응답은 다음과 같을 수 있습니다.
{ "@odata.context": "https://my-search-service.search.windows.net/$metadata#knowledgebases(name)", "value": [ { "name": "my-kb" }, { "name": "my-kb-2" } ] }개별 기술 자료 정의를 가져와서 기술 자료 참조를 확인합니다.
# Get a knowledge base definition from azure.core.credentials import AzureKeyCredential from azure.search.documents.indexes import SearchIndexClient index_client = SearchIndexClient(endpoint = "search_url", credential = AzureKeyCredential("api_key")) kb = index_client.get_knowledge_base("knowledge_base_name") print(kb)예제 응답은 다음과 같을 수 있습니다.
{ "name": "my-kb", "description": null, "retrievalInstructions": null, "answerInstructions": null, "outputMode": null, "knowledgeSources": [ { "name": "my-blob-ks", } ], "models": [], "encryptionKey": null, "retrievalReasoningEffort": { "kind": "low" } }기술 자료를 삭제하거나 여러 기술 자료가 있는 경우 기술 자료를 업데이트하여 원본을 제거합니다. 이 예제에서는 삭제를 보여줍니다.
# Delete a knowledge base from azure.core.credentials import AzureKeyCredential from azure.search.documents.indexes import SearchIndexClient index_client = SearchIndexClient(endpoint = "search_url", credential = AzureKeyCredential("api_key")) index_client.delete_knowledge_base("knowledge_base_name") print(f"Knowledge base deleted successfully.")기술 자료를 삭제합니다.
# Delete a knowledge source from azure.core.credentials import AzureKeyCredential from azure.search.documents.indexes import SearchIndexClient index_client = SearchIndexClient(endpoint = "search_url", credential = AzureKeyCredential("api_key")) index_client.delete_knowledge_source("knowledge_source_name") print(f"Knowledge source deleted successfully.")
검색 서비스의 모든 기술 자료 목록을 가져옵니다.
### Get knowledge bases GET {{search-url}}/knowledgebases?api-version={{api-version}}&$select=name api-key: {{api-key}}참조:기술 자료 - 목록
예제 응답은 다음과 같을 수 있습니다.
{ "@odata.context": "https://my-search-service.search.windows.net/$metadata#knowledgebases(name)", "value": [ { "name": "my-kb" }, { "name": "my-kb-2" } ] }개별 기술 자료 정의를 가져와서 기술 자료 참조를 확인합니다.
### Get a knowledge base definition GET {{search-url}}/knowledgebases/{{knowledge-base-name}}?api-version={{api-version}} api-key: {{api-key}}참조:기술 자료 - 가져오기
예제 응답은 다음과 같을 수 있습니다.
{ "name": "my-kb", "description": null, "retrievalInstructions": null, "answerInstructions": null, "outputMode": null, "knowledgeSources": [ { "name": "my-blob-ks", } ], "models": [], "encryptionKey": null, "retrievalReasoningEffort": { "kind": "low" } }기술 자료를 삭제하거나 여러 기술 자료가 있는 경우 기술 자료를 업데이트하여 원본을 제거합니다. 이 예제에서는 삭제를 보여줍니다.
### Delete a knowledge base DELETE {{search-url}}/knowledgebases/{{knowledge-base-name}}?api-version={{api-version}} api-key: {{api-key}}참조:기술 자료 - 삭제
기술 자료를 삭제합니다.
### Delete a knowledge source DELETE {{search-url}}/knowledgesources/{{knowledge-source-name}}?api-version={{api-version}} api-key: {{api-key}}참조:기술 자료 - 삭제