本文為英文版的機器翻譯版本，如內容有任何歧義或不一致之處，概以英文版為準。 # 使用 AWS CloudFormation 管理專案 Amazon Bedrock 已與 AWS CloudFormation 整合，可讓您在基礎設施範本中定義和管理專案。您可以使用 JSON 或 YAML 範本，在多個 AWS 帳戶和區域中一致且重複地佈建專案。 ## AWS::BedrockMantle::Project 使用 `AWS::BedrockMantle::Project` 資源在 CloudFormation 範本中建立和管理 Bedrock 專案。透過 CloudFormation 建立的專案支援與透過 API 建立的相同功能，包括 IAM 政策連接、標記和可觀測性。 ### 語法 若要在您的 CloudFormation 範本中宣告此實體，請使用下列語法： **Example CloudFormation 語法** ``` { "Type": "AWS::BedrockMantle::Project", "Properties": { "Name": String, "Tags": [ { "Key": String, "Value": String }, { "Key": String, "Value": String }, { "Key": String, "Value": String }, { "Key": String, "Value": String } ] } } ``` ``` Type: AWS::BedrockMantle::Project Properties: Name: String Tags: Key: Value ``` ### Properties 名稱 必要. 專案的名稱。在您的 AWS 帳戶中必須是唯一的。 類型：字串 下限：1 上限：64 模式：`^([0-9a-zA-Z][ _-]?)+$` 更新要求：替換 Tags (標籤) 要與專案建立關聯的鍵值對映射，用於成本分配和存取控制。 類型：字串的映射 更新要求：不中斷 **標籤更新注意事項** 上的 CloudFormation 標籤更新會在內部`AWS::BedrockMantle::Project`使用個別的新增和移除操作。沒有原子完整標籤取代。如果堆疊更新在操作中失敗，專案的標籤集可能處於部分更新狀態。在修改標籤的堆疊更新之後，請務必驗證最終標籤狀態。 ### 傳回值 #### Ref 當您將此資源的邏輯 ID 傳遞至內部 `Ref`函數時， 便會`Ref`傳回專案 ID （例如 `proj_abc123`)。 #### Fn::GetAtt ProjectId 專案的唯一識別符 （例如 `proj_abc123`)。 ProjectArn 專案的 Amazon Resource Name (ARN) （例如 `arn:aws:bedrock-mantle:us-east-1:123456789012:project/proj_abc123`)。 狀態 專案的狀態。 `ACTIVE`表示專案已就緒可供使用。 `ARCHIVED`表示專案已封存，無法接受新的推論請求。 CreatedAt 建立專案的時間戳記。 UpdatedAt 上次更新專案的時間戳記。 ## 範例 ### 建立基本專案 下列範例會為生產聊天機器人應用程式建立專案： **Example 基本專案** ``` AWSTemplateFormatVersion: '2010-09-09' Description: Amazon Bedrock Project for Production Chatbot Resources: CustomerChatbotProject: Type: AWS::BedrockMantle::Project Properties: Name: CustomerChatbot-Production Tags: - Key: Project Value: CustomerChatbot - Key: Environment Value: Production - Key: Owner Value: TeamAlpha - Key: CostCenter Value: "21524" Outputs: ProjectId: Description: The ID of the created project Value: !Ref CustomerChatbotProject ProjectArn: Description: The ARN of the created project Value: !GetAtt CustomerChatbotProject.ProjectArn ``` ``` { "AWSTemplateFormatVersion": "2010-09-09", "Resources": { "CustomerChatbotProject": { "Type": "AWS::BedrockMantle::Project", "Properties": { "Name": "CustomerChatbot-Production", "Tags": [ { "Key": "Project", "Value": "CustomerChatbot" }, { "Key": "Environment", "Value": "Production" }, { "Key": "Owner", "Value": "TeamAlpha" }, { "Key": "CostCenter", "Value": "21524" } ] } } }, "Outputs": { "ProjectId": { "Description": "The ID of the created project", "Value": { "Ref": "CustomerChatbotProject" } }, "ProjectArn": { "Description": "The ARN of the created project", "Value": { "Fn::GetAtt": ["CustomerChatbotProject", "ProjectArn"] } } } } ``` ### 為不同的環境建立多個專案 下列範例為單一堆疊中的開發、預備和生產環境佈建個別專案： ``` AWSTemplateFormatVersion: '2010-09-09' Description: Amazon Bedrock Projects for Multi-Environment Deployment Parameters: ApplicationName: Type: String Default: InternalSearch Description: Name of the application CostCenter: Type: String Description: Cost center for billing allocation Resources: DevelopmentProject: Type: AWS::BedrockMantle::Project Properties: Name: !Sub "${ApplicationName}-Development" Tags: - Key: Project Value: !Ref ApplicationName - Key: Environment Value: Development - Key: CostCenter Value: !Ref CostCenter StagingProject: Type: AWS::BedrockMantle::Project Properties: Name: !Sub "${ApplicationName}-Staging" Tags: - Key: Project Value: !Ref ApplicationName - Key: Environment Value: Staging - Key: CostCenter Value: !Ref CostCenter ProductionProject: Type: AWS::BedrockMantle::Project Properties: Name: !Sub "${ApplicationName}-Production" Tags: - Key: Project Value: !Ref ApplicationName - Key: Environment Value: Production - Key: CostCenter Value: !Ref CostCenter Outputs: DevelopmentProjectArn: Value: !GetAtt DevelopmentProject.ProjectArn Export: Name: !Sub "${ApplicationName}-Dev-ProjectArn" StagingProjectArn: Value: !GetAtt StagingProject.ProjectArn Export: Name: !Sub "${ApplicationName}-Staging-ProjectArn" ProductionProjectArn: Value: !GetAtt ProductionProject.ProjectArn Export: Name: !Sub "${ApplicationName}-Prod-ProjectArn" ``` ### 建立具有 IAM 角色存取權的專案 下列範例會建立專案並連接 IAM 政策，授予特定角色調用模型的存取權： ``` AWSTemplateFormatVersion: '2010-09-09' Description: Amazon Bedrock Project with IAM Access Control Resources: ProductionProject: Type: AWS::BedrockMantle::Project Properties: Name: CustomerChatbot-Production Tags: - Key: Environment Value: Production - Key: CostCenter Value: "21524" ProductionAppRole: Type: AWS::IAM::Role Properties: RoleName: BedrockProjectProductionRole AssumeRolePolicyDocument: Version: '2012-10-17' Statement: - Effect: Allow Principal: Service: lambda.amazonaws.com Action: sts:AssumeRole Policies: - PolicyName: BedrockProjectInvokeAccess PolicyDocument: Version: '2012-10-17' Statement: - Effect: Allow Action: - bedrock-mantle:CreateInference - bedrock-mantle:GetProject Resource: !GetAtt ProductionProject.ProjectArn Outputs: ProjectArn: Value: !GetAtt ProductionProject.ProjectArn RoleArn: Value: !GetAtt ProductionAppRole.Arn ``` ## 搭配專案 API 使用 CloudFormation 輸出 部署 CloudFormation 堆疊之後，您可以使用堆疊輸出在應用程式程式碼中參考專案 ARN 和 ID： ``` import boto3 from openai import OpenAI # Retrieve project details from CloudFormation stack outputs cfn = boto3.client('cloudformation', region_name='us-east-1') response = cfn.describe_stacks(StackName='my-bedrock-projects-stack') outputs = {o['OutputKey']: o['OutputValue'] for o in response['Stacks'][0]['Outputs']} production_project_arn = outputs['ProductionProjectArn'] # Extract project ID from ARN # ARN format: arn:aws:bedrock-mantle:us-east-1:123456789012:project/proj_abc123 project_id = production_project_arn.split('/')[-1] print(f"Using project: {project_id}") # Use the project for inference client = OpenAI(project=project_id) response = client.responses.create( model="openai.gpt-oss-120b", input="Hello from a CloudFormation-managed project!" ) print(response) ``` ## 進一步了解 如需搭配 Amazon Bedrock 資源使用 CloudFormation 的詳細資訊，請參閱： + [使用 AWS CloudFormation 建立 Amazon Bedrock 資源](creating-resources-with-cloudformation.md) + [AWS CloudFormation 使用者指南](https://docs.aws.amazon.com/AWSCloudFormation/latest/UserGuide/Welcome.html) + [Amazon Bedrock 資源類型參考](https://docs.aws.amazon.com/AWSCloudFormation/latest/TemplateReference/AWS_Bedrock.html)