クイックスタート: Developer Knowledge API でクライアント ライブラリを使用する

このガイドでは、公式クライアント ライブラリを使用して Developer Knowledge API の使用を開始する方法について説明します。環境の設定方法、使用する言語のクライアント ライブラリのインストール方法、API 呼び出しを行ってデベロッパー ドキュメントを検索して取得する方法について説明します。

始める前に

Developer Knowledge API クライアント ライブラリの使用を開始する前に、次のセクションを完了してください。

API を有効にする

1. Google API ライブラリで Developer Knowledge API ページを開きます。
2. API を使用するプロジェクトが正しく選択されていることを確認します。
3. [有効にする] をクリックします。API を有効または使用するために、特定の IAM ロールは必要ありません。

認証を設定する

Developer Knowledge API クライアント ライブラリは、アプリケーションのデフォルト認証情報（ADC）を使用してリクエストを認証します。

ローカル認証情報を設定するには、次のコマンドを実行します。

gcloud auth application-default login

サービス アカウントなどの認証情報オプションの詳細については、アプリケーションのデフォルト認証情報のドキュメントをご覧ください。

クライアント ライブラリをインストールする

公式の Developer Knowledge API クライアント ライブラリをインストールするには、プログラミング言語を選択します。

Python

pip install google-developer-knowledge

Node.js と TypeScript

npm install @google/developer-knowledge

Go

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

Java

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

ドキュメントから回答を生成する

AnswerQuery エンドポイントは、コード設定、トラブルシューティング手順、プロダクト機能などに関する複雑な自然言語の質問に、公式ドキュメント ソースから情報を取得して回答します。

言語タブを選択すると、AnswerQuery を呼び出す方法の例が表示されます。

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 と 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
}

Java

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;
    }
  }
}

ドキュメント チャンクを検索する

生成された回答ではなく、ドキュメント内の正確なローカライズされたテキスト セグメントを見つけるには、SearchDocumentChunks エンドポイントを使用します。このメソッドは、コーパスをスキャンし、個々のコンテンツ スニペット（チャンク）と親ドキュメント ID を返します。この ID を使用して、ドキュメントのコンテンツ全体を取得できます。

言語タブを選択すると、ドキュメント チャンクを検索する方法の例が表示されます。

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 と 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
}

Java

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;
    }
  }
}

ドキュメントを取得する

各ドキュメント チャンクには、親ドキュメントのリソース名を含む parent フィールドが含まれています。そのリソース名で GetDocument を使用して、ドキュメント全体を取得します。

次のサンプルは、サンプル ドキュメントを取得します。サンプル ドキュメント名は、検索結果から返された任意の parent リソース名に置き換えることができます。

言語タブを選択すると、GetDocument を呼び出す方法の例が表示されます。

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 と 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
}

Java

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;
    }
  }
}

複数のドキュメントを取得する

BatchGetDocuments を使用すると、1 回の API 呼び出しでリソース名で最大 20 個のドキュメントを取得できます。

言語タブを選択すると、BatchGetDocuments を呼び出す方法の例が表示されます。

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 と 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;
    }
  }
}

次のステップ

• コーパス リファレンスで、含まれているドキュメントの完全なリストを確認してください。
• API メソッドとパラメータの詳細については、API リファレンス ドキュメントをご覧ください。
• Google Antigravity で MCP サーバーを設定する方法を確認する。