English  ｜ 简体中文 

Dify AWS Tool

Introduction

This repo provides some sample code to show how to integrate SageMaker Provider and some AWS Service based tools to Dify.

In addition to the reference code, you can also refer to the Dify official guide for more information.

Prerequisites

• Dify Environment

• AWS Account and AWS Experience

• Basic experience with Linux environments

Assets

[Attention]：We welcome contributions of more workflows, SageMaker models, and built-in tools. You can fork this repository and submit a merge request, please also update README.md, you need to add a new row to the corresponding table

Workflow

DSL Name	Description	Link	Owner
Term_based_translate	Translation Workflow with Term mapping Retrieval Tool	DSL	[email protected]
Code_translate	Code Transform between different Program Language	Coming	[email protected]
Basic_RAG_Sample	simple basic rag workflow with rerank tool	DSL	[email protected]
Andrewyng/translation-agent	Andrew Ng's translate agent.	DSL	[email protected]

Builtin_Tools

Tool Name	Tool Type	Description	Deploy_doc	Owner
Rerank	PAAS	Text Similarity Rerank Tool	Notebook	[email protected]
Term_multilingual_mapping	PAAS	Word Segment/ Term mapping Retrieval Tool	Repo	[email protected]
Bedrock Guardrails	SAAS	Text moderation Tool, implemented through the independent assessment API ApplyGuardrail API provided on Amazon Bedrock Guardrail.	Built-in Tool	[email protected]

Model_Provider

Model Name	model_type	Deploy_doc	Owner
Bge-m3-rerank-v2	SageMaker\Rerank	Notebook	[email protected]
Bge-embedding-m3	SageMaker\Embedding	Notebook	[email protected]

How to Install

Below Script is only for SageMaker Model_provider and AWS Builtin Tools, you can import workflows from Web Interface.

dify_path=/home/ec2-user/dify
tag=aws

# step1 - download code
git clone https://github.com/aws-samples/dify-aws-tool/

# step2 - intall code
mv ./dify-aws-tool/builtin_tools/aws ${dify_path}/api/core/tools/provider/builtin/
mv ./dify-aws-tool/model_provider/sagemaker ${dify_path}/api/core/model_runtime/model_providers/

# step3 - build image
cd ${dify_path}/api
sudo docker build -t dify-api:${tag} .

# step4 - restart dify with new image
# [Todo] modify ${dify_path}/docker/docker-compose.yaml, change api and worker service's images to the image you just built
cd ${dify_path}/docker/
sudo docker-compose down
sudo docker-compose up -d

How to deploy SageMaker Endpoint

If you want to add your Embedding/Rerank model to Dify Sagemaker Model Provider, you should deploy them by yourself in AWS/SageMaker at first. Please see the Guide.

Target Audience

• Dify / AWS User
• GenAI Developer