AI と Azure Cosmos DB の使用に関するベスト プラクティスとソリューション

アプリケーションの成長に合わせてデータベースを拡張できるように、AI 搭載アプリケーション用のデータベースとして Azure Cosmos DB for NoSQL を使います。 また、Azure Cosmos DB の速度と組み込みの信頼性を利用し、時間が経過してニーズが変化してもソリューションを高速で使用可能な状態に維持することもできます。

AI アプリケーションを最新化する

Azure Cosmos DB for NoSQL、Azure OpenAI、Azure Kubernetes Service、Azure AI Search を使って、ベクトル検索と AI アシスタントを実装します。

Screenshot of an AI assistants application responding to queries about various bikes for a retail shop.

Diagram of the architecture of the application modernization solution accelerator.

Azure AI Search、Azure OpenAI、Azure Storage、Azure Cosmos DB、バッキング サービスを使用する、Kubernetes に支えられた Web アプリケーションを示す図。 ベクトルと項目は Azure Cosmos DB に保存され、ファイルは Azure Storage に保存されます。

リンク
ソリューション アクセラレータ https://github.com/Azure/Vector-Search-AI-Assistant/tree/cognitive-search-vector
ハッカソン https://github.com/Azure/Build-Modern-AI-Apps-Hackathon

支払いとトランザクション処理

支払い追跡プロセスを実装するには、Azure Front Door、Azure OpenAI、Azure Kubernetes Service、Azure Static Web Apps、Azure Cosmos DB for NoSQL を使います。

Diagram of the architecture of the payment processing solution accelerator.

Azure Static Web アプリと Azure Front Door を顧客インターフェイスとして使用するサービスを示す図。 この場合、ソリューションは支払い API と worker サービスの組み合わせをホストし、Azure Kubernetes Service で支払いトランザクションを処理します。 最後に、Kubernetes コンテナーは Azure Cosmos DB にデータを格納し、Azure OpenAI から AI 入力候補を取得します。

リンク
ソリューション アクセラレータ https://github.com/Azure/Real-time-Payment-Transaction-Processing-at-Scale
ハッカソン https://github.com/Azure/Real-Time-Transactions-Hackathon

医療費請求トランザクション処理

Azure Event Hubs、Azure Static Web Apps、Azure Kubernetes Service、Azure OpenAI、Azure Cosmos DB for NoSQL で構築されたソリューションを使って、複雑な医療費請求を処理します。

Diagram of the architecture of the claims processing solution accelerator.

Azure Event Hubs を使って請求を取り込む外部システムを示す図。 同時に、エージェントは Azure Static Web アプリに関心を示しています。 worker サービスと API は、Azure Kubernetes Service でホストされます。 コンテナーでは、入力候補に Azure OpenAI が使われます。 コンテナーは Azure Cosmos DB for NoSQL にもデータを格納し、それは Azure Synapse Analytics を使って分析および管理されます。

リンク
ソリューション アクセラレータ https://github.com/Azure/Medical-Claims-Transaction-Processing-at-scale
ハッカソン https://github.com/Azure/Medical-Claims-Processing-Hackathon

AI ソリューションを自動化する

新しい Azure Developer CLI などのツールを使って、AI 搭載ソリューションのデプロイを自動化します。 この自動化を使って、最新の開発者および運用ワークフローを作成します。

リンク
サンプル チャット アプリケーション https://github.com/Azure-Samples/cosmosdb-chatgpt
トレーニング モジュール https://learn.microsoft.com/training/modules/build-chat-bot-azure-cosmos-db-openai-blazor

アプリケーションの成長に合わせてデータベースを拡張できるように、AI 搭載アプリケーション用のデータベースとして Azure Cosmos DB for MongoDB 仮想コアを使います。 また、Azure Cosmos DB の速度と組み込みの信頼性を利用し、時間が経過してニーズが変化してもソリューションを高速で使用可能な状態に維持することもできます。

取得拡張生成

Azure Cosmos DB for MongoDB 仮想コア、Azure OpenAI、Azure Functions、Azure Web Apps の組み合わせを使って RAG パターンを実装します。

リンク
ソリューション アクセラレータ https://github.com/Azure/Vector-Search-AI-Assistant-MongoDBvCore
Python ノートブック https://github.com/Microsoft/AzureDataRetrievalAugmentedGenerationSamples/tree/main/Python/CosmosDB-MongoDB-vCore

次のステップ