一方、生成 AI のソリューションが成熟してきたら、エンドツーエンドで生成 AI のアプリケーションやソリューションを構築し、デプロイするためのプラットフォームが必要になることがあります。Google Cloud は、アプリ開発の初期段階からアプリのデプロイ、アプリのホスティング、複雑なデータの大規模な管理まで、デベロッパーが生成 AI の力を活用できるツールの包括的なエコシステムを提供します。
Google Cloud の Vertex AI Platform には、効率と信頼性のために AI モデルの使用、デプロイ、モニタリングを合理化する一連の MLOps ツールが用意されています。さらに、データベース、DevOps ツール、ロギング、モニタリング、IAM とのインテグレーションにより、生成 AI のライフサイクル全体を包括的に管理できます。
左側のナビゲーション パネルの下部にある [Build with Vertex AI on Google Cloud] をクリックします。
[アカウントを作成して Google Cloud の使用を開始する] ページが開きます。
[同意して続行] をクリックします。
[本人確認を行います] ページが表示されます。
[無料で利用開始] をクリックします。
[Get Started with Vertex AI studio] ダイアログが表示されます。
Vertex AI の実行に必要な API を有効にするには、[同意して続行] をクリックします。
省略可: Google AI Studio からデータを移行する方法については、このページのプロンプトを移行するをご覧ください。
Python: Vertex AI の Gemini API に移行する
以下のセクションでは、Python コードを移行して Vertex AI の Gemini API を使用する場合に役立つコード スニペットを示します。
Vertex AI Python SDK の設定
Vertex AI では、API キーは必要ありません。代わりに、Vertex AI の Gemini は、IAM アクセスを使用して管理されます。IAM アクセスは、Vertex AI SDK を介して Gemini API を呼び出すユーザー、グループ、またはサービス アカウントの権限を制御します。
Vertex AI に対して推論呼び出しを行うには、ユーザーまたはサービス アカウントに Vertex AI ユーザーロールが割り当てられていることも確認する必要があります。
クライアントをインストールするコードサンプル
Gemini Developer API
Vertex AI の Gemini API
# To install the Python SDK, use this CLI command:# pip install google-generativeaiimportgoogle.generativeaiasgenaifromgoogle.generativeaiimportGenerativeModelAPI_KEY="API_KEY"genai.configure(api_key=API_KEY)
# To install the Python SDK, use this CLI command:# pip install google-genaifromgoogleimportgenaiPROJECT_ID="PROJECT_ID"LOCATION="LOCATION"# e.g. us-central1client=genai.Client(project=PROJECT_ID,location=LOCATION,vertexai=True)
テキスト プロンプトからテキストを生成するコードサンプル
Gemini Developer API
Vertex AI の Gemini API
model=GenerativeModel("gemini-2.0-flash")response=model.generate_content("The opposite of hot is")print(response.text)# The opposite of hot is cold.
fromgoogleimportgenaifromgoogle.genai.typesimportHttpOptionsclient=genai.Client(http_options=HttpOptions(api_version="v1"))response=client.models.generate_content(model="gemini-2.5-flash",contents="How does AI work?",)print(response.text)# Example response:# Okay, let's break down how AI works. It's a broad field, so I'll focus on the ...## Here's a simplified overview:# ...
テキストと画像からテキストを生成するコードサンプル
Gemini Developer API
Vertex AI の Gemini API
importPIL.Imagemultimodal_model=GenerativeModel("gemini-2.0-flash")image=PIL.Image.open("image.jpg")response=multimodal_model.generate_content(["What is this picture?",image])print(response.text)# A cat is shown in this picture.
fromgoogleimportgenaifromgoogle.genai.typesimportHttpOptions,Partclient=genai.Client(http_options=HttpOptions(api_version="v1"))response=client.models.generate_content(model="gemini-2.5-flash",contents=["What is shown in this image?",Part.from_uri(file_uri="gs://cloud-samples-data/generative-ai/image/scones.jpg",mime_type="image/jpeg",),],)print(response.text)# Example response:# The image shows a flat lay of blueberry scones arranged on parchment paper. There are ...
マルチターン チャットを生成するコードサンプル
Gemini Developer API
Vertex AI の Gemini API
model=GenerativeModel("gemini-2.0-flash")chat=model.start_chat()print(chat.send_message("How are you?").text)print(chat.send_message("What can you do?").text)
fromgoogleimportgenaifromgoogle.genai.typesimportHttpOptions,ModelContent,Part,UserContentclient=genai.Client(http_options=HttpOptions(api_version="v1"))chat_session=client.chats.create(model="gemini-2.5-flash",history=[UserContent(parts=[Part(text="Hello")]),ModelContent(parts=[Part(text="Great to meet you. What would you like to know?")],),],)response=chat_session.send_message("Tell me a story.")print(response.text)# Example response:# Okay, here's a story for you:# ...
プロンプトを Vertex AI Studio に移行する
Google AI Studio のプロンプト データは、Google ドライブ フォルダに保存されます。このセクションでは、プロンプトを Vertex AI Studio に移行する方法について説明します。
[[["わかりやすい","easyToUnderstand","thumb-up"],["問題の解決に役立った","solvedMyProblem","thumb-up"],["その他","otherUp","thumb-up"]],[["わかりにくい","hardToUnderstand","thumb-down"],["情報またはサンプルコードが不正確","incorrectInformationOrSampleCode","thumb-down"],["必要な情報 / サンプルがない","missingTheInformationSamplesINeed","thumb-down"],["翻訳に関する問題","translationIssue","thumb-down"],["その他","otherDown","thumb-down"]],["最終更新日 2025-08-28 UTC。"],[],[],null,["As your [Gemini API](https://ai.google.dev/gemini-api/docs)\napplications mature, you might find that you need a more expansive platform for\nbuilding and deploying generative AI applications and solutions end-to-end.\nVertex AI provides a comprehensive ecosystem of tools to enable\ndevelopers to harness the power of generative AI, from the initial stages of app\ndevelopment to app deployment, app hosting, and managing complex data at scale.\n\nWith Vertex AI, you get access to a suite of Machine Learning\nOperations (MLOps) tools to streamline usage, deployment, and monitoring of AI\nmodels for efficiency and reliability. Additionally, integrations with\ndatabases, Development Operations (DevOps) tools, logging, monitoring, and\nIAM offer a comprehensive approach to managing the entire\ngenerative AI lifecycle.\n\nDifferences between using the Gemini API on its own and Vertex AI\n\nThe following table summarizes the main differences between the\nGemini API and Vertex AI to help you decide which option is\nright for your use case:\n\n| **Feature** | **Gemini API** | **Vertex AI** |\n| Endpoint names | `generativelanguage.googleapis.com` | `aiplatform.googleapis.com` |\n| Sign up | Google Account | Google Cloud account (with terms agreement and billing) |\n| Authentication | API key | Google Cloud service account |\n| User interface playground | Google AI Studio | Vertex AI Studio |\n|-----------------------------|---------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|\n| API \\& SDK | Server and mobile/web client SDKs - Server: Python, Node.js, Go, Dart, ABAP - Mobile/Web client: Android (Kotlin/Java), Swift, Web, Flutter | Server and mobile/web client SDKs - Server: Python, Node.js, Go, Java, ABAP - Mobile/Web client (via [Vertex AI in Firebase](https://firebase.google.com/docs/vertex-ai)): Android (Kotlin/Java), Swift, Web, Flutter |\n| No-cost usage of API \\& SDK | Yes, [where applicable](https://ai.google.dev/gemini-api/docs/billing#is-Gemini-free-in-EEA-UK-CH) | $300 Google Cloud credit for new users |\n| Quota (requests per minute) | Varies based on model and pricing plan (see [detailed information](https://ai.google.dev/pricing)) | Varies based on model and region (see [detailed information](/vertex-ai/generative-ai/docs/quotas)) |\n| Enterprise support | No | - Customer encryption key - Virtual private cloud - Data residency - Access transparency - Scalable infrastructure for application hosting - Databases and data storage |\n| MLOps | No | Full MLOps on Vertex AI (examples: model evaluation, Model Monitoring, Model Registry) |\n\nMigration steps\n\nThe following sections cover the steps required to migrate your Gemini\nAPI code to Vertex AI. These steps assume you have prompt data from\nGoogle AI Studio saved in Google Drive.\n\nWhen migrating to Vertex AI:\n\n- You can use your existing Google Cloud project (the same one you used to generate your Gemini API key) or you can create a new [Google Cloud project](/resource-manager/docs/creating-managing-projects).\n- Supported regions might differ between the Gemini API and Vertex AI. See the list of [supported regions for generative\n AI on Google Cloud](/vertex-ai/generative-ai/docs/learn/locations).\n- Any models you created in Google AI Studio need to be retrained in Vertex AI.\n\n1. Migrate your prompts to Vertex AI Studio\n\nYour Google AI Studio prompt data is saved in a Google Drive folder. This\nsection shows how to migrate your prompts to Vertex AI Studio.\n\n1. Open [Google Drive](https://drive.google.com).\n2. Navigate to the **AI_Studio** folder where the prompts are stored.\n3. Download your prompts from Google Drive to a local directory.\n\n | **Note:** Prompts downloaded from Google Drive are in the text (`txt`) format. Before you upload them to Vertex AI Studio, change the file extensions from `.txt` to `.json` to convert them to JSON files.\n4. Open [Vertex AI Studio](https://console.cloud.google.com/vertex-ai/generative) in the Google Cloud console.\n\n5. In the **Vertex AI** menu, click **Recents \\\u003e View all** to open the\n **Prompt management** menu.\n\n6. Click download**Import prompt**.\n\n7. Next to the **Prompt file** field, click **Browse** and select a prompt\n from your local directory.\n\n To upload prompts in bulk, you must manually combine your prompts into a\n single JSON file.\n8. Click **Upload**.\n\n2. Upload training data to Vertex AI Studio\n\nTo migrate your training data to Vertex AI, you need to upload your\ndata to a Cloud Storage bucket. For more information, see\n[Introduction to tuning](/vertex-ai/generative-ai/docs/models/tune-models).\n\n3. Delete unused API Keys\n\nIf you no longer need to use your Gemini API key for the\nGemini Developer API, then follow security best practices and delete\nit.\n\nTo delete an API key:\n\n1. Open the [Google Cloud API Credentials](https://console.cloud.google.com/apis/credentials)\n page.\n\n2. Find the API key that you want to delete and click the **Actions** icon.\n\n3. Select **Delete API key**.\n\n4. In the **Delete credential** modal, select **Delete**.\n\n Deleting an API key takes a few minutes to propagate. After propagation\n completes, any traffic using the deleted API key is rejected.\n\n| **Important:** If you delete a key that's still used in production and need to recover it, see [`gcloud beta services api-keys\n| undelete`](/sdk/gcloud/reference/beta/services/api-keys/undelete).\n\nWhat's next\n\n- Try a quickstart tutorial using [Vertex AI Studio](/vertex-ai/generative-ai/docs/start/quickstarts/quickstart) or the [Vertex AI API](/vertex-ai/generative-ai/docs/start/quickstarts/quickstart-multimodal)."]]
