Dify Self-Hosted on AWS

This project provides an External Knowledge Base API designed to bridge Dify with Amazon Bedrock Knowledge Bases. This allows Dify to leverage Bedrock's managed RAG (Retrieval-Augmented Generation) capabilities as an external knowledge source.

To implement this connection, you must follow the Dify guide for connecting to an external knowledge base.

To use Bedrock LLMs in Dify, you must first enable the desired models in the AWS Bedrock management console. Once enabled, follow these steps in the Dify UI:

1. Log in to Dify.
2. Click your profile in the top-right corner.
3. Navigate to WORKSPACE -> Model Provider.
4. Select AWS Bedrock model.
5. Select the AWS region where your models are enabled.
6. Click Save.

IAM policies are pre-configured by this deployment to allow access to Bedrock.

If you are using an environment with limited storage, such as AWS CloudShell, you can use the provided simple-deploy.sh script to automate the process. This script will guide you through the deployment via interactive prompts and eventually output the DifyOnAwsStack.DifyUrl.

git clone https://github.com/aws-samples/dify-self-hosted-on-aws.git
cd dify-self-hosted-on-aws
./simple-deploy.sh

Before deploying, ensure your local environment or deployment instance meets the following requirements:

• Node.js: version v22 or newer
• Docker: installed and running
• AWS CLI: installed and configured with an IAM profile that has the Administrator policy attached

To deploy Dify using the AWS CDK, ensure you have configured your environment and then follow the standard npm and CDK command sequence. You can customize deployment parameters like AWS regions by modifying bin/cdk.ts. For a full list of configurable parameters, refer to the EnvironmentProps interface in lib/environment-props.ts.

# install npm dependencies
npm ci

# bootstrap the AWS account (required only once per account and region)
npx cdk bootstrap

# deploy the CDK stack
npx cdk deploy --all

You can extend the available Python libraries in Dify's code execution feature by editing the requirements file.

1. Edit ./lib/constructs/dify-services/docker/python-requirements.txt using the standard Requirements File Format.
2. If certain libraries require additional system calls, you can enable them by setting the allowAnySysCalls flag in bin/cdk.ts.

WARNING

Enabling allowAnySysCalls allows all system calls in the Dify sandbox. Only do this if you fully trust the code being executed in your Dify tenant.

To run manual commands (e.g., for migrations or maintenance), use the ConsoleConnectToTaskCommand output provided by the CDK deployment. This command uses AWS ECS Execute Command.

# Example command structure
aws ecs execute-command \
  --region <region> \
  --cluster <cluster-name> \
  --container Main \
  --interactive \
  --command "bash" \
  --task <task-id>

If you need to handle breaking changes manually during an upgrade:

1. Set autoMigration: false in lib/dify-on-aws-stack.ts within the ApiService construct.
2. Deploy the project with the new Dify version.
3. Use the ConsoleConnectToTaskCommand to connect to the container.
4. Run the required migration or maintenance commands manually.
5. Set autoMigration: true and deploy again to re-enable automatic migrations.

To avoid ongoing AWS charges, destroy the stack:

npx cdk destroy --force

Note: If you used customEcrRepositoryName and ran the copy-to-ecr.ts script, you must manually delete the ECR repository and its images, as cdk destroy will not remove them.