빠른 시작: Developer Knowledge API와 함께 클라이언트 라이브러리 사용

이 가이드에서는 공식 클라이언트 라이브러리를 사용하여 Developer Knowledge API를 시작하는 방법을 보여줍니다. 환경을 설정하고, 원하는 언어의 클라이언트 라이브러리를 설치하고, API를 호출하여 개발자 문서를 검색하고 가져오는 방법을 알아봅니다.

시작하기 전에

Developer Knowledge API 클라이언트 라이브러리를 사용하기 전에 다음 섹션을 완료하세요.

API 사용 설정

1. Google API 라이브러리에서 Developer Knowledge API 페이지를 엽니다.
2. API를 사용할 올바른 프로젝트가 선택되어 있는지 확인합니다.
3. 사용 설정을 클릭합니다. API를 사용 설정하거나 사용하는 데 특정 IAM 역할이 필요하지 않습니다.

인증 설정

Developer Knowledge API 클라이언트 라이브러리는 애플리케이션 기본 사용자 인증 정보 (ADC)를 사용하여 요청을 인증합니다.

로컬 인증 사용자 인증 정보를 설정하려면 다음 명령어를 실행합니다.

gcloud auth application-default login

서비스 계정과 같은 사용자 인증 정보 옵션에 대해 자세히 알아보려면 애플리케이션 기본 사용자 인증 정보 문서를 참고하세요.

클라이언트 라이브러리 설치

공식 Developer Knowledge API 클라이언트 라이브러리를 설치하려면 프로그래밍 언어를 선택하세요.

Python

pip install google-developer-knowledge

Node.js 및 TypeScript

npm install @google/developer-knowledge

Go

go get cloud.google.com/go/developerknowledge/apiv1

자바

<!-- Maven dependency -->
<dependency>
  <groupId>com.google.cloud</groupId>
  <artifactId>google-cloud-developer-knowledge</artifactId>
  <version>0.3.0</version>
</dependency>

문서에서 답변 생성

AnswerQuery 엔드포인트는 공식 문서 소스에서 정보를 가져와 코드 설정, 문제 해결 절차, 제품 기능과 관련된 복잡한 자연어 질문에 답변합니다.

언어 탭을 선택하여 AnswerQuery를 호출하는 방법의 예를 확인하세요.

Python

from google.cloud import developer_knowledge_v1


def answer_query(
    query: str = "How do I create a Google Cloud Storage bucket?",
) -> developer_knowledge_v1.AnswerQueryResponse:
    """Answers a developer question grounded in Google developer documentation.

    Args:
        query: The technical question to answer.

    Returns:
        The AnswerQueryResponse containing the grounded answer,
        citations, and references.
    """
    client = developer_knowledge_v1.DeveloperKnowledgeClient()

    request = developer_knowledge_v1.AnswerQueryRequest(
        query=query,
    )

    response = client.answer_query(request=request)

    print(f"Answer:\n{response.answer.answer_text}\n")
    print(f"Citations count: {len(response.answer.citations)}")
    print(f"References count: {len(response.answer.references)}")

    return response

Node.js 및 TypeScript

const {DeveloperKnowledgeClient} = require('@google/developer-knowledge');

/**
 * Answers a developer question grounded in Google developer documentation.
 *
 * @param {string} query The technical question to answer.
 */
async function answerQuery(
  query = 'How do I create a Google Cloud Storage bucket?'
) {
  const client = new DeveloperKnowledgeClient();

  const request = {
    query,
  };

  const [response] = await client.answerQuery(request);

  console.log(`Answer:\n${response.answer.answerText}\n`);
  const citationsCount = response.answer.citations
    ? response.answer.citations.length
    : 0;
  const referencesCount = response.answer.references
    ? response.answer.references.length
    : 0;
  console.log(`Citations count: ${citationsCount}`);
  console.log(`References count: ${referencesCount}`);

  return response;
}

Go

import (
	"context"
	"fmt"
	"io"

	developerknowledge "cloud.google.com/go/developerknowledge/apiv1"
	developerknowledgepb "cloud.google.com/go/developerknowledge/apiv1/developerknowledgepb"
)

// answerQuery answers a developer question grounded in Google developer documentation.
func answerQuery(w io.Writer, query string) (*developerknowledgepb.AnswerQueryResponse, error) {
	ctx := context.Background()

	client, err := developerknowledge.NewDeveloperKnowledgeClient(ctx)
	if err != nil {
		return nil, fmt.Errorf("developerknowledge.NewDeveloperKnowledgeClient: %w", err)
	}
	defer client.Close()

	req := &developerknowledgepb.AnswerQueryRequest{
		Query: query,
	}

	resp, err := client.AnswerQuery(ctx, req)
	if err != nil {
		return nil, fmt.Errorf("AnswerQuery: %w", err)
	}

	if resp.GetAnswer() != nil {
		fmt.Fprintf(w, "Answer:\n%s\n\n", resp.GetAnswer().GetAnswerText())
		fmt.Fprintf(w, "Citations count: %d\n", len(resp.GetAnswer().GetCitations()))
		fmt.Fprintf(w, "References count: %d\n", len(resp.GetAnswer().GetReferences()))
	}

	return resp, nil
}

자바

import com.google.developers.knowledge.v1.AnswerQueryRequest;
import com.google.developers.knowledge.v1.AnswerQueryResponse;
import com.google.developers.knowledge.v1.DeveloperKnowledgeClient;
import java.io.IOException;

public class AnswerQuery {

  public static void main(String[] args) throws IOException {
    // TODO(developer): Replace these variables before running the sample.
    String query = "How do I create a Google Cloud Storage bucket?";
    answerQuery(query);
  }

  // Answers a developer question grounded in Google developer documentation.
  public static AnswerQueryResponse answerQuery(String query) throws IOException {
    // Initialize client that will be used to send requests. This client only needs to be created
    // once, and can be reused for multiple requests. After completing all of your requests, call
    // the "close" method on the client to safely clean up any remaining background resources.
    try (DeveloperKnowledgeClient client = DeveloperKnowledgeClient.create()) {
      AnswerQueryRequest request =
          AnswerQueryRequest.newBuilder().setQuery(query).build();

      AnswerQueryResponse response = client.answerQuery(request);

      System.out.println("Answer:\n" + response.getAnswer().getAnswerText() + "\n");
      System.out.println("Citations count: " + response.getAnswer().getCitationsCount());
      System.out.println("References count: " + response.getAnswer().getReferencesCount());

      return response;
    }
  }
}

문서 청크 검색

생성된 대답이 아닌 문서 내에서 정확한 현지화된 텍스트 세그먼트를 찾으려면 SearchDocumentChunks 엔드포인트를 사용하세요. 이 메서드는 코퍼스를 스캔하고 전체 문서 콘텐츠를 검색하는 데 사용할 수 있는 상위 문서 식별자와 함께 개별 콘텐츠 스니펫 (청크)을 반환합니다.

언어 탭을 선택하여 문서 청크를 검색하는 방법의 예를 확인하세요.

Python

from google.cloud import developer_knowledge_v1


def search_document_chunks(
    query: str = "How to create a Cloud Storage bucket",
    page_size: int = 5,
) -> (
    developer_knowledge_v1.services.developer_knowledge.pagers.SearchDocumentChunksPager
):
    """Searches developer documentation chunks for a given query.

    Args:
        query: The natural language search query.
        page_size: The maximum number of document chunks to return.

    Returns:
        The SearchDocumentChunksPager containing relevant document chunks.
    """
    client = developer_knowledge_v1.DeveloperKnowledgeClient()

    request = developer_knowledge_v1.SearchDocumentChunksRequest(
        query=query,
        page_size=page_size,
    )

    response = client.search_document_chunks(request=request)

    count = 0
    for chunk in response:
        print(f"Parent Document: {chunk.parent}")
        print(f"Chunk ID: {chunk.id}")
        print(f"Content: {chunk.content[:100]}...\n")
        count += 1
        if page_size > 0 and count >= page_size:
            break

    return response

Node.js 및 TypeScript

const {DeveloperKnowledgeClient} = require('@google/developer-knowledge');

/**
 * Searches developer documentation chunks for a given query.
 *
 * @param {string} query The search query string.
 * @param {number} pageSize The maximum number of document chunks to return.
 */
async function searchDocumentChunks(
  query = 'How to create a Cloud Storage bucket',
  pageSize = 5
) {
  const client = new DeveloperKnowledgeClient();

  const request = {
    query,
    pageSize,
  };

  // Warning: Should always disable autoPaginate to avoid iterating through all pages.
  // By default NodeJS SDK returns an iterable where you can iterate through all
  // search results instead of only the limited number of results requested on pageSize.
  const [chunks] = await client.searchDocumentChunks(request, {
    autoPaginate: false,
  });

  for (const chunk of chunks) {
    console.log(`Parent Document: ${chunk.parent}`);
    console.log(`Chunk ID: ${chunk.id}`);
    console.log(`Content Preview: ${chunk.content.substring(0, 100)}...\n`);
  }

  return chunks;
}

Go

import (
	"context"
	"fmt"
	"io"

	developerknowledge "cloud.google.com/go/developerknowledge/apiv1"
	developerknowledgepb "cloud.google.com/go/developerknowledge/apiv1/developerknowledgepb"
	"google.golang.org/api/iterator"
)

// searchDocumentChunks searches developer documentation chunks for a given query.
func searchDocumentChunks(w io.Writer, query string, pageSize int32) ([]*developerknowledgepb.DocumentChunk, error) {
	ctx := context.Background()

	client, err := developerknowledge.NewDeveloperKnowledgeClient(ctx)
	if err != nil {
		return nil, fmt.Errorf("developerknowledge.NewDeveloperKnowledgeClient: %w", err)
	}
	defer client.Close()

	req := &developerknowledgepb.SearchDocumentChunksRequest{
		Query:    query,
		PageSize: pageSize,
	}

	var results []*developerknowledgepb.DocumentChunk
	it := client.SearchDocumentChunks(ctx, req)
	for {
		chunk, err := it.Next()
		if err == iterator.Done {
			break
		}
		if err != nil {
			return nil, fmt.Errorf("SearchDocumentChunks: %w", err)
		}
		results = append(results, chunk)
		fmt.Fprintf(w, "Parent Document: %s\n", chunk.GetParent())
		fmt.Fprintf(w, "Chunk ID: %s\n", chunk.GetId())
		fmt.Fprintf(w, "Content: %s\n\n", chunk.GetContent())

		if pageSize > 0 && len(results) >= int(pageSize) {
			break
		}
	}

	return results, nil
}

자바

import com.google.developers.knowledge.v1.DeveloperKnowledgeClient;
import com.google.developers.knowledge.v1.DeveloperKnowledgeClient.SearchDocumentChunksPagedResponse;
import com.google.developers.knowledge.v1.DocumentChunk;
import com.google.developers.knowledge.v1.SearchDocumentChunksRequest;
import java.io.IOException;

public class SearchDocumentChunks {

  public static void main(String[] args) throws IOException {
    // TODO(developer): Replace these variables before running the sample.
    String query = "How to create a Cloud Storage bucket";
    int pageSize = 5;
    searchDocumentChunks(query, pageSize);
  }

  // Searches developer documentation chunks for a given query.
  public static SearchDocumentChunksPagedResponse searchDocumentChunks(
      String query, int pageSize) throws IOException {
    // Initialize client that will be used to send requests. This client only needs to be created
    // once, and can be reused for multiple requests. After completing all of your requests, call
    // the "close" method on the client to safely clean up any remaining background resources.
    try (DeveloperKnowledgeClient client = DeveloperKnowledgeClient.create()) {
      SearchDocumentChunksRequest request =
          SearchDocumentChunksRequest.newBuilder()
              .setQuery(query)
              .setPageSize(pageSize)
              .build();

      SearchDocumentChunksPagedResponse response = client.searchDocumentChunks(request);

      for (DocumentChunk chunk : response.getPage().getValues()) {
        System.out.println("Parent Document: " + chunk.getParent());
        System.out.println("Chunk ID: " + chunk.getId());
        String preview = chunk.getContent();
        if (preview.length() > 100) {
          preview = preview.substring(0, 100) + "...";
        }
        System.out.println("Content: " + preview + "\n");
      }

      return response;
    }
  }
}

문서 가져오기

각 문서 청크에는 상위 문서의 리소스 이름이 포함된 parent 필드가 포함됩니다. 해당 리소스 이름과 함께 GetDocument를 사용하여 전체 문서를 가져옵니다.

다음 샘플은 샘플 문서를 가져옵니다. 샘플 문서 이름을 검색 결과에서 반환된 parent 리소스 이름으로 바꿀 수 있습니다.

언어 탭을 선택하여 GetDocument를 호출하는 방법의 예를 확인하세요.

Python

from google.cloud import developer_knowledge_v1


def get_document(
    name: str = "documents/docs.cloud.google.com/storage/docs/creating-buckets",
) -> developer_knowledge_v1.Document:
    """Retrieves a single developer documentation page by its resource name.

    Args:
        name: The resource name of the document in format
            'documents/{uri_without_scheme}'.

    Returns:
        The Document containing the full Markdown content and metadata.
    """
    client = developer_knowledge_v1.DeveloperKnowledgeClient()

    request = developer_knowledge_v1.GetDocumentRequest(
        name=name,
    )

    document = client.get_document(request=request)

    print(f"Title: {document.title}")
    print(f"URI: {document.uri}")
    print(f"Data Source: {document.data_source}")
    print(f"Content Length: {document.content_length_bytes} bytes")
    print(f"Content Preview: {document.content[:150]}...\n")

    return document

Node.js 및 TypeScript

const {DeveloperKnowledgeClient} = require('@google/developer-knowledge');

/**
 * Retrieves a single developer documentation page by its resource name.
 *
 * @param {string} name The resource name in format 'documents/{uri_without_scheme}'.
 */
async function getDocument(
  name = 'documents/docs.cloud.google.com/storage/docs/creating-buckets'
) {
  const client = new DeveloperKnowledgeClient();

  const request = {
    name,
  };

  const [document] = await client.getDocument(request);

  console.log(`Title: ${document.title}`);
  console.log(`URI: ${document.uri}`);
  console.log(`Data Source: ${document.dataSource}`);
  console.log(`Content Length: ${document.contentLengthBytes} bytes`);
  console.log(`Content Preview: ${document.content.substring(0, 150)}...\n`);

  return document;
}

Go

import (
	"context"
	"fmt"
	"io"

	developerknowledge "cloud.google.com/go/developerknowledge/apiv1"
	developerknowledgepb "cloud.google.com/go/developerknowledge/apiv1/developerknowledgepb"
)

// getDocument retrieves a single developer documentation page by its resource name.
func getDocument(w io.Writer, name string) (*developerknowledgepb.Document, error) {
	ctx := context.Background()

	client, err := developerknowledge.NewDeveloperKnowledgeClient(ctx)
	if err != nil {
		return nil, fmt.Errorf("developerknowledge.NewDeveloperKnowledgeClient: %w", err)
	}
	defer client.Close()

	req := &developerknowledgepb.GetDocumentRequest{
		Name: name,
	}

	doc, err := client.GetDocument(ctx, req)
	if err != nil {
		return nil, fmt.Errorf("GetDocument: %w", err)
	}

	fmt.Fprintf(w, "Title: %s\n", doc.GetTitle())
	fmt.Fprintf(w, "URI: %s\n", doc.GetUri())
	fmt.Fprintf(w, "Data Source: %s\n", doc.GetDataSource())
	fmt.Fprintf(w, "Content Length: %d bytes\n\n", doc.GetContentLengthBytes())

	return doc, nil
}

자바

import com.google.developers.knowledge.v1.DeveloperKnowledgeClient;
import com.google.developers.knowledge.v1.Document;
import com.google.developers.knowledge.v1.GetDocumentRequest;
import java.io.IOException;

public class GetDocument {

  public static void main(String[] args) throws IOException {
    // TODO(developer): Replace these variables before running the sample.
    String name = "documents/docs.cloud.google.com/storage/docs/creating-buckets";
    getDocument(name);
  }

  // Retrieves a single developer documentation page by its resource name.
  public static Document getDocument(String name) throws IOException {
    // Initialize client that will be used to send requests. This client only needs to be created
    // once, and can be reused for multiple requests. After completing all of your requests, call
    // the "close" method on the client to safely clean up any remaining background resources.
    try (DeveloperKnowledgeClient client = DeveloperKnowledgeClient.create()) {
      GetDocumentRequest request = GetDocumentRequest.newBuilder().setName(name).build();

      Document document = client.getDocument(request);

      System.out.println("Title: " + document.getTitle());
      System.out.println("URI: " + document.getUri());
      System.out.println("Data Source: " + document.getDataSource());
      System.out.println("Content Length: " + document.getContentLengthBytes() + " bytes");
      String preview = document.getContent();
      if (preview.length() > 150) {
        preview = preview.substring(0, 150) + "...";
      }
      System.out.println("Content Preview: " + preview + "\n");

      return document;
    }
  }
}

여러 문서 가져오기

BatchGetDocuments를 사용하여 단일 API 호출에서 리소스 이름별로 최대 20개의 문서를 검색합니다.

언어 탭을 선택하여 BatchGetDocuments를 호출하는 방법의 예를 확인하세요.

Python

from typing import List, Optional

from google.cloud import developer_knowledge_v1


def batch_get_documents(
    names: Optional[List[str]] = None,
) -> developer_knowledge_v1.BatchGetDocumentsResponse:
    """Retrieves multiple developer documentation pages in a single request.

    Args:
        names: A list of resource names in format 'documents/{uri_without_scheme}'.

    Returns:
        The BatchGetDocumentsResponse containing the retrieved documents.
    """
    if names is None:
        names = [
            "documents/docs.cloud.google.com/storage/docs/creating-buckets",
            "documents/docs.cloud.google.com/storage/docs/deleting-buckets",
        ]

    client = developer_knowledge_v1.DeveloperKnowledgeClient()

    request = developer_knowledge_v1.BatchGetDocumentsRequest(
        names=names,
    )

    response = client.batch_get_documents(request=request)

    for doc in response.documents:
        print(f"Title: {doc.title}")
        print(f"URI: {doc.uri}")
        print(f"Content Length: {doc.content_length_bytes} bytes\n")

    return response

Node.js 및 TypeScript

const {DeveloperKnowledgeClient} = require('@google/developer-knowledge');

/**
 * Retrieves multiple developer documentation pages in a single request.
 *
 * @param {string[]} names Array of resource names in format 'documents/{uri_without_scheme}'.
 */
async function batchGetDocuments(
  names = [
    'documents/docs.cloud.google.com/storage/docs/creating-buckets',
    'documents/docs.cloud.google.com/storage/docs/deleting-buckets',
  ]
) {
  const client = new DeveloperKnowledgeClient();

  const request = {
    names,
  };

  const [response] = await client.batchGetDocuments(request);

  if (response.documents) {
    for (const doc of response.documents) {
      console.log(`Title: ${doc.title}`);
      console.log(`URI: ${doc.uri}`);
      console.log(`Content Length: ${doc.contentLengthBytes} bytes\n`);
    }
  }

  return response;
}

Go

import (
	"context"
	"fmt"
	"io"

	developerknowledge "cloud.google.com/go/developerknowledge/apiv1"
	developerknowledgepb "cloud.google.com/go/developerknowledge/apiv1/developerknowledgepb"
)

// batchGetDocuments retrieves multiple developer documentation pages in a single request.
func batchGetDocuments(w io.Writer, names []string) (*developerknowledgepb.BatchGetDocumentsResponse, error) {
	ctx := context.Background()

	client, err := developerknowledge.NewDeveloperKnowledgeClient(ctx)
	if err != nil {
		return nil, fmt.Errorf("developerknowledge.NewDeveloperKnowledgeClient: %w", err)
	}
	defer client.Close()

	req := &developerknowledgepb.BatchGetDocumentsRequest{
		Names: names,
	}

	resp, err := client.BatchGetDocuments(ctx, req)
	if err != nil {
		return nil, fmt.Errorf("BatchGetDocuments: %w", err)
	}

	for _, doc := range resp.GetDocuments() {
		fmt.Fprintf(w, "Title: %s\n", doc.GetTitle())
		fmt.Fprintf(w, "\tURI: %s\n", doc.GetUri())
		fmt.Fprintf(w, "\tContent Length: %d bytes\n\n", doc.GetContentLengthBytes())
	}

	return resp, nil
}

Java

import com.google.developers.knowledge.v1.BatchGetDocumentsRequest;
import com.google.developers.knowledge.v1.BatchGetDocumentsResponse;
import com.google.developers.knowledge.v1.DeveloperKnowledgeClient;
import com.google.developers.knowledge.v1.Document;
import java.io.IOException;
import java.util.Arrays;
import java.util.List;

public class BatchGetDocuments {

  public static void main(String[] args) throws IOException {
    // TODO(developer): Replace these variables before running the sample.
    List<String> names =
        Arrays.asList(
            "documents/docs.cloud.google.com/storage/docs/creating-buckets",
            "documents/docs.cloud.google.com/storage/docs/deleting-buckets");
    batchGetDocuments(names);
  }

  // Retrieves multiple developer documentation pages in a single request.
  public static BatchGetDocumentsResponse batchGetDocuments(List<String> names) throws IOException {
    // Initialize client that will be used to send requests. This client only needs to be created
    // once, and can be reused for multiple requests. After completing all of your requests, call
    // the "close" method on the client to safely clean up any remaining background resources.
    try (DeveloperKnowledgeClient client = DeveloperKnowledgeClient.create()) {
      BatchGetDocumentsRequest request =
          BatchGetDocumentsRequest.newBuilder().addAllNames(names).build();

      BatchGetDocumentsResponse response = client.batchGetDocuments(request);

      for (Document doc : response.getDocumentsList()) {
        System.out.println("Title: " + doc.getTitle());
        System.out.println("URI: " + doc.getUri());
        System.out.println("Content Length: " + doc.getContentLengthBytes() + " bytes\n");
      }

      return response;
    }
  }
}

다음 단계

• 코퍼스 참조를 살펴보고 포함된 문서의 전체 목록을 확인하세요.
• API 메서드 및 매개변수에 관한 자세한 내용은 API 참조 문서를 참고하세요.
• Google Antigravity에서 MCP 서버를 설정하는 방법을 알아봅니다.