i want to build an rag application where i want to utilize open source embedding model along with the model i write my complete code in sagemaker notebook and make direct call from react

Treat it just like any other std ML workflow , your client code call goes thru RAG model(embedding model) and then goes to the main model. While you can use sagemaker for RAG setup , however please note you will have to setup whole pipeline of your own , including RAG model call , DB setup etc, data sources , packaging deployment etc. I would highly recommend to use Bedrock service where-in you can easily plugin externals RAG platforms/frameworks thru cli/sdk or use inbuild managed RAG experience called Knowledge bases.

Just with few clicks you will have your app ready. Few ref links if it picks your interest: https://aws.amazon.com/bedrock/knowledge-bases/

One nice blog to rea thru for RAG - https://aws.amazon.com/blogs/aws/knowledge-bases-now-delivers-fully-managed-rag-experience-in-amazon-bedrock/