Fix bug in deployment procedure in AWS CDK and support HTTPS

cdk.Dockerfile

@@ -15,6 +15,7 @@ COPY ./ ./
 # Install dependencies
 RUN poetry config virtualenvs.create false && poetry install --no-interaction --no-ansi
 
-RUN poetry add pymysql==1.0.2
+RUN poetry add botocore
+RUN poetry add pymysql
 
 CMD ["sh", "./container-cmd-cdk.sh"]

container-cmd-cdk.sh

@@ -1,3 +1,5 @@
 export LANGFLOW_DATABASE_URL="mysql+pymysql://${username}:${password}@${host}:3306/${dbname}"
 # echo $LANGFLOW_DATABASE_URL
-uvicorn --factory src.backend.langflow.main:create_app --host 0.0.0.0 --port 7860 --reload --log-level debug
+uvicorn --factory src.backend.langflow.main:create_app --host 0.0.0.0 --port 7860 --reload --log-level debug
+
+# python -m langflow run --host 0.0.0.0 --port 7860

scripts/aws/README.ja.md

@@ -8,10 +8,9 @@ Langflow on AWS では、 [AWS Cloud Development Kit](https://aws.amazon.com/cdk
 
 作成するアプリケーションのアーキテクチャです。
 ![langflow-archi](./img/langflow-archi.png)
-AWS CDK によって [Application Load Balancer](https://aws.amazon.com/elasticloadbalancing/application-load-balancer/?nc1=h_ls)、[AWS Fargate](https://aws.amazon.com/fargate/?nc2=type_a)、[Amazon Aurora](https://aws.amazon.com/rds/aurora/?nc2=type_a) を作成します。
+AWS CDK によって Langflow のアプリケーションをデプロイします。アプリケーションは [Amazon CloudFront](https://aws.amazon.com/cloudfront/?nc1=h_ls) を介して配信されます。CloudFront は 2 つのオリジンを有しています。1 つ目は静的な Web サイトを配信するための [Amazon Simple Storage Service](https://aws.amazon.com/s3/?nc1=h_ls) (S3)、2 つ目は バックエンドと通信するための [Application Load Balancer](https://aws.amazon.com/elasticloadbalancing/application-load-balancer/?nc1=h_ls) (ALB) です。ALB の背後には FastAPI が動作する [AWS Fargate](https://aws.amazon.com/fargate/?nc2=type_a) 、データベースの [Amazon Aurora](https://aws.amazon.com/rds/aurora/?nc2=type_a) が作成されます。
+Fargate は [Amazon Elastic Container Registry](https://aws.amazon.com/ecr/?nc1=h_ls) (ECR) に保存された Docker イメージを使用します。
 Auroraのシークレットは [AWS Secrets Manager](https://aws.amazon.com/secrets-manager/?nc2=type_a) によって管理されます。
-Fargate のタスクはフロントエンドとバックエンドに分かれており、サービス検出によって通信します。
-リソースをデプロイするだけであれば、上記の各サービスについて深い知識は必要ありません。
 
 # 環境構築とデプロイ方法
 1. [AWS CloudShell](https://us-east-1.console.aws.amazon.com/cloudshell/home?region=us-east-1)を開きます。
@@ -28,7 +27,7 @@ Fargate のタスクはフロントエンドとバックエンドに分かれて
 
 1. 以下のコマンドを実行します。
  ```shell
- git clone -b aws-cdk https://github.com/logspace-ai/langflow.git
+ git clone https://github.com/logspace-ai/langflow.git
  cd langflow/scripts/aws
  cp .env.example .env # 環境設定を変える場合はこのファイル(.env)を編集してください。
  npm ci
@@ -38,7 +37,7 @@ Fargate のタスクはフロントエンドとバックエンドに分かれて
 1. 表示される URL にアクセスします。
  ```shell
  Outputs:
- LangflowAppStack.NetworkURLXXXXXX = http://alb-XXXXXXXXXXX.elb.amazonaws.com
+ LangflowAppStack.frontendURLXXXXXX = https://XXXXXXXXXXX.cloudfront.net
  ```
 1. サインイン画面でユーザー名とパスワードを入力します。`.env`ファイルでユーザー名とパスワードを設定していない場合、ユーザー名は`admin`、パスワードは`123456`で設定されます。
  ![make-cloud9](./img/langflow-signin.png)

scripts/aws/README.md

@@ -9,11 +9,9 @@ This tutorial assumes you have an AWS account and basic knowledge of AWS.
 
 The architecture of the application to be created:
 ![langflow-archi](./img/langflow-archi.png)
-
-[Application Load Balancer](https://aws.amazon.com/elasticloadbalancing/application-load-balancer/?nc1=h_ls), [AWS Fargate](https://aws.amazon.com/fargate/?nc2=type_a) and [Amazon Aurora](https://aws.amazon.com/rds/aurora/?nc2=type_a) are created by AWS CDK.
-The aurora's secrets are managed by [AWS Secrets Manager](https://aws.amazon.com/secrets-manager/?nc2=type_a).
-The Fargate task is divided into a frontend and a backend, which communicate through service discovery.
-If you just want to deploy resources, you do not need in-depth knowledge of each of the above services.
+Langflow is deployed using AWS CDK. The application is distributed via [Amazon CloudFront](https://aws.amazon.com/cloudfront/?nc1=h_ls), which has two origins: the first is [Amazon Simple Storage Service](https://aws.amazon.com/s3/?nc1=h_ls) (S3) for serving a static website, and the second is an [Application Load Balancer](https://aws.amazon.com/elasticloadbalancing/application-load-balancer/?nc1=h_ls) (ALB) for communicating with the backend. [AWS Fargate](https://aws.amazon.com/fargate/?nc2=type_a), where FastAPI runs and [Amazon Aurora](https://aws.amazon.com/rds/aurora/?nc2=type_a), the database, are created behind the ALB.
+Fargate uses a Docker image stored in [Amazon Elastic Container Registry](https://aws.amazon.com/ecr/?nc1=h_ls) (ECR).
+Aurora's secret is managed by [AWS Secrets Manager](https://aws.amazon.com/secrets-manager/?nc2=type_a).
 
 # How to set up your environment and deploy langflow
 
@@ -22,13 +20,14 @@ If you just want to deploy resources, you do not need in-depth knowledge of each
  ```shell
  git clone https://github.com/aws-samples/cloud9-setup-for-prototyping
  cd cloud9-setup-for-prototyping
+ cat params.json | jq '.name |= "c9-for-langflow"'
  ./bin/bootstrap
  ```
-1. When you see `Done!` in Cloudshell, open `cloud9-for-prototyping` from [AWS Cloud9](https://us-east-1.console.aws.amazon.com/cloud9control/home?region=us-east-1#/).
+1. When you see `Done!` in Cloudshell, open `c9-for-langflow` from [AWS Cloud9](https://us-east-1.console.aws.amazon.com/cloud9control/home?region=us-east-1#/).
  ![make-cloud9](./img/langflow-cloud9-en.png)
 1. Run the following command in the Cloud9 terminal.
  ```shell
- git clone -b aws-cdk https://github.com/logspace-ai/langflow.git
+ git clone https://github.com/logspace-ai/langflow.git
  cd langflow/scripts/aws
  cp .env.example .env # Edit this file if you need environment settings
  npm ci
@@ -38,7 +37,7 @@ If you just want to deploy resources, you do not need in-depth knowledge of each
 1. Access the URL displayed.
  ```shell
  Outputs:
- LangflowAppStack.NetworkURLXXXXXX = http://alb-XXXXXXXXXXX.elb.amazonaws.com
+ LangflowAppStack.frontendURLXXXXXX = https://XXXXXXXXXXX.cloudfront.net
  ```
 1. Enter your user name and password to sign in. If you have not set a user name and password in your `.env` file, the user name will be set to `admin` and the password to `123456`.
  ![signin-langflow](./img/langflow-signin.png)
@@ -49,5 +48,6 @@ If you just want to deploy resources, you do not need in-depth knowledge of each
  ```shell
  bash delete-resources.sh
  ```
-1. Open [AWS CloudFormation](https://us-east-1.console.aws.amazon.com/cloudformation/home?region=us-east-1#/getting-started), select `aws-cloud9-cloud9-for-prototyping-XXXX` and delete it.
+1. Open [AWS CloudFormation](https://us-east-1.console.aws.amazon.com/cloudformation/home?region=us-east-1#/getting-started), select `aws-cloud9-c9-for-langflow-XXXX` and delete it.
  ![delete-cfn](./img/langflow-cfn.png)
+s

scripts/aws/bin/cdk.ts

@@ -4,6 +4,7 @@ import * as cdk from 'aws-cdk-lib';
 import { LangflowAppStack } from '../lib/cdk-stack';
 
 const app = new cdk.App();
+
 new LangflowAppStack(app, 'LangflowAppStack', {
  /* If you don't specify 'env', this stack will be environment-agnostic.
  * Account/Region-dependent features and context lookups will not work,

scripts/aws/cdk.json

@@ -17,6 +17,8 @@
  ]
  },
  "context": {
+ "ragEnabled": false,
+ "kendraIndexArn": null,
  "@aws-cdk/aws-lambda:recognizeLayerVersion": true,
  "@aws-cdk/core:checkSecretUsage": true,
  "@aws-cdk/core:target-partitions": [

scripts/aws/delete-resources.sh

@@ -1,4 +1,4 @@
-aws cloudformation delete-stack --stack-name LangflowAppStack
+# aws cloudformation delete-stack --stack-name LangflowAppStack
 aws ecr delete-repository --repository-name langflow-backend-repository --force
-aws ecr delete-repository --repository-name langflow-frontend-repository --force
+# aws ecr delete-repository --repository-name langflow-frontend-repository --force
 # aws ecr describe-repositories --output json | jq -re ".repositories[].repositoryName"

scripts/aws/lib/cdk-stack.ts

@@ -2,21 +2,38 @@ import * as cdk from 'aws-cdk-lib';
 import { Construct } from 'constructs';
 import * as ecs from 'aws-cdk-lib/aws-ecs'
 
-import { Network, EcrRepository, FrontEndCluster, BackEndCluster, Rds, EcsIAM } from './construct';
+import { Network, EcrRepository, Web, BackEndCluster, Rds, EcsIAM, Rag} from './construct';
 // import * as sqs from 'aws-cdk-lib/aws-sqs';
 
+const errorMessageForBooleanContext = (key: string) => {
+ return `There was an error setting $ {key}. Possible causes are as follows.
+ - Trying to set it with the -c option instead of changing cdk.json
+ - cdk.json is set to a value that is not a boolean (e.g. “true” double quotes are not required)
+ - no items in cdk.json (unset) `;
+};
+
+
 export class LangflowAppStack extends cdk.Stack {
  constructor(scope: Construct, id: string, props?: cdk.StackProps) {
  super(scope, id, props);
+ // Kendra Enable
+ const ragEnabled: boolean = this.node.tryGetContext('ragEnabled')!;
+ if (typeof ragEnabled !== 'boolean') {
+ throw new Error(errorMessageForBooleanContext('ragEnabled'));
+ }
+ if (ragEnabled) {
+ new Rag(this, 'Rag', {
+ });
+ }
+
  // Arch
  const arch = ecs.CpuArchitecture.X86_64
 
  // VPC
- const { vpc, cluster, alb, targetGroup, cloudmapNamespace, ecsFrontSG, ecsBackSG, dbSG, albSG, backendLogGroup, frontendLogGroup} = new Network(this, 'Network')
+ const { vpc, cluster, ecsBackSG, dbSG, backendLogGroup, alb, albTG, albSG} = new Network(this, 'Network')
 
  // ECR
- const { ecrFrontEndRepository,ecrBackEndRepository} = new EcrRepository(this, 'Ecr', {
- cloudmapNamespace:cloudmapNamespace,
+ const { ecrBackEndRepository } = new EcrRepository(this, 'Ecr', {
  arch:arch
  })
 
@@ -25,7 +42,7 @@ export class LangflowAppStack extends cdk.Stack {
  const { rdsCluster } = new Rds(this, 'Rds', { vpc, dbSG })
 
  // IAM
- const { frontendTaskRole, frontendTaskExecutionRole, backendTaskRole, backendTaskExecutionRole } = new EcsIAM(this, 'EcsIAM',{
+ const { backendTaskRole, backendTaskExecutionRole } = new EcsIAM(this, 'EcsIAM',{
  rdsCluster:rdsCluster
  })
 
@@ -36,29 +53,18 @@ export class LangflowAppStack extends cdk.Stack {
  backendTaskRole:backendTaskRole,
  backendTaskExecutionRole:backendTaskExecutionRole,
  backendLogGroup:backendLogGroup,
- cloudmapNamespace:cloudmapNamespace,
  rdsCluster:rdsCluster,
- alb:alb,
- arch:arch
+ arch:arch,
+ albTG:albTG
  })
  backendService.node.addDependency(rdsCluster);
 
- const frontendService = new FrontEndCluster(this, 'frontend',{
+ const frontendService = new Web(this, 'frontend',{
  cluster:cluster,
- ecsFrontSG:ecsFrontSG,
- ecrFrontEndRepository:ecrFrontEndRepository,
- targetGroup: targetGroup,
- backendServiceName: backendService.backendServiceName,
- frontendTaskRole: frontendTaskRole,
- frontendTaskExecutionRole: frontendTaskExecutionRole,
- frontendLogGroup: frontendLogGroup,
- cloudmapNamespace: cloudmapNamespace,
- arch:arch 
+ alb:alb,
+ albSG:albSG
  })
  frontendService.node.addDependency(backendService);
 
-
- // S3+CloudFront
- // new Web(this,'Cloudfront-S3')
  }
 }

scripts/aws/lib/construct/backend.ts

@@ -21,18 +21,17 @@ interface BackEndProps {
  backendTaskRole: iam.Role;
  backendTaskExecutionRole: iam.Role;
  backendLogGroup: logs.LogGroup;
- cloudmapNamespace: servicediscovery.PrivateDnsNamespace;
  rdsCluster:rds.DatabaseCluster
- alb:elb.IApplicationLoadBalancer
  arch:ecs.CpuArchitecture
+ albTG: elb.ApplicationTargetGroup;
 }
 
 export class BackEndCluster extends Construct {
- readonly backendServiceName: string 
 
  constructor(scope: Construct, id: string, props:BackEndProps) {
  super(scope, id)
- const containerPort = 7860
+ const backendServiceName = 'backend'
+ const backendServicePort = 7860
  // Secrets ManagerからDB認証情報を取ってくる
  const secretsDB = props.rdsCluster.secret!;
 
@@ -59,20 +58,13 @@ export class BackEndCluster extends Construct {
  logGroup: props.backendLogGroup,
  }),
  environment:{
- // user:pass@endpoint:port/dbname
- // "LANGFLOW_DATABASE_URL" : `mysql+pymysql://${username}:${password}@${host}:3306/${dbname}`,
- // "LANGFLOW_DATABASE_URL" : "sqlite:///./langflow.db",
- // "LANGFLOW_LANGCHAIN_CACHE" : "SQLiteCache",
- // "LANGFLOW_AUTO_LOGIN" : "false",
- // "LANGFLOW_SUPERUSER" : "admin",
- // "LANGFLOW_SUPERUSER_PASSWORD" : "1234567"
  "LANGFLOW_AUTO_LOGIN" : process.env.LANGFLOW_AUTO_LOGIN ?? 'false',
  "LANGFLOW_SUPERUSER" : process.env.LANGFLOW_SUPERUSER ?? "admin",
  "LANGFLOW_SUPERUSER_PASSWORD" : process.env.LANGFLOW_SUPERUSER_PASSWORD ?? "123456"
  },
  portMappings: [
  {
- containerPort: containerPort,
+ containerPort: backendServicePort,
  protocol: ecs.Protocol.TCP,
  },
  ],
@@ -84,22 +76,15 @@ export class BackEndCluster extends Construct {
  "password": ecs.Secret.fromSecretsManager(secretsDB, 'password'),
  },
  });
- this.backendServiceName = 'backend'
+
  const backendService = new ecs.FargateService(this, 'BackEndService', {
  cluster: props.cluster,
- serviceName: this.backendServiceName,
+ serviceName: backendServiceName,
  taskDefinition: backendTaskDefinition,
  enableExecuteCommand: true,
  securityGroups: [props.ecsBackSG],
- cloudMapOptions: {
- cloudMapNamespace: props.cloudmapNamespace,
- containerPort: containerPort,
- dnsRecordType: servicediscovery.DnsRecordType.A,
- dnsTtl: Duration.seconds(10),
- name: this.backendServiceName
- },
  vpcSubnets: { subnetType: ec2.SubnetType.PRIVATE_WITH_EGRESS },
  });
-
+ props.albTG.addTarget(backendService);
  }
 }

scripts/aws/lib/construct/ecr.ts

@@ -9,65 +9,41 @@ import { Construct } from 'constructs'
 
 
 interface ECRProps {
- cloudmapNamespace: servicediscovery.PrivateDnsNamespace;
  arch:ecs.CpuArchitecture;
 }
 
 export class EcrRepository extends Construct {
- readonly ecrFrontEndRepository: ecr.Repository
  readonly ecrBackEndRepository: ecr.Repository
 
  constructor(scope: Construct, id: string, props: ECRProps) {
  super(scope, id)
 
  const imagePlatform = props.arch == ecs.CpuArchitecture.ARM64 ? Platform.LINUX_ARM64 : Platform.LINUX_AMD64
  const backendPath = path.join(__dirname, "../../../../../", "langflow")
- const frontendPath = path.join(__dirname, "../../../../src/", "frontend")
  const excludeDir = ['node_modules','.git', 'cdk.out']
  const LifecycleRule = {
  tagStatus: ecr.TagStatus.ANY,
  description: 'Delete more than 30 image',
  maxImageCount: 30,
  }
 
- // リポジトリ作成
- this.ecrFrontEndRepository = new ecr.Repository(scope, 'LangflowFrontEndRepository', {
- repositoryName: 'langflow-frontend-repository',
- removalPolicy: RemovalPolicy.RETAIN,
- imageScanOnPush: true,
- })
+ // Backend ECR リポジトリ作成
  this.ecrBackEndRepository = new ecr.Repository(scope, 'LangflowBackEndRepository', {
  repositoryName: 'langflow-backend-repository',
  removalPolicy: RemovalPolicy.RETAIN,
  imageScanOnPush: true,
  })
  // LifecycleRule作成
- this.ecrFrontEndRepository.addLifecycleRule(LifecycleRule)
  this.ecrBackEndRepository.addLifecycleRule(LifecycleRule)
 
  // Create Docker Image Asset
- const dockerFrontEndImageAsset = new DockerImageAsset(this, "DockerFrontEndImageAsset", {
- directory: frontendPath,
- file:"cdk.Dockerfile",
- buildArgs:{
- "BACKEND_URL":`http://backend.${props.cloudmapNamespace.namespaceName}:7860`
- },
- exclude: excludeDir,
- platform: imagePlatform,
- });
  const dockerBackEndImageAsset = new DockerImageAsset(this, "DockerBackEndImageAsset", {
  directory: backendPath,
  file:"cdk.Dockerfile",
  exclude: excludeDir,
  platform: imagePlatform,
  });
 
- // Deploy Docker Image to ECR Repository
- new ecrdeploy.ECRDeployment(this, "DeployFrontEndImage", {
- src: new ecrdeploy.DockerImageName(dockerFrontEndImageAsset.imageUri),
- dest: new ecrdeploy.DockerImageName(this.ecrFrontEndRepository.repositoryUri)
- });
-
  // Deploy Docker Image to ECR Repository
  new ecrdeploy.ECRDeployment(this, "DeployBackEndImage", {
  src: new ecrdeploy.DockerImageName(dockerBackEndImageAsset.imageUri),