aws-samples / serverless-rag-demo Goto Github PK

It looks like the sagemaker endpoint creation is mandatory and that is quite costly. So why not make that optional and use bedrock service model for model endpoint.

Hello,

I am working with this awesome project and I have everything working from the base image, great work I love it!

I am however wondering if the only current way to update the html page is to make the updates and then re-run creator.sh. I don't see any other mentions of updating the scripts and I am new-ish to code build so I am not sure if there is a better route to take?

I have tried this route and in codebuild, it says the build was successful but I do not see any html changes, now this could be due to errors in the deployment so I have been checking those.

Thank you for any support!

getting this error

"{"success":false,"errorMessage":"Exception occured when querying LLM: An error occurred (validationException) when calling the InvokeModelWithResponseStream operation: "claude-3-sonnet-20240229" is not supported on this API. Please use the Messages API instead.","statusCode":"400"}"

Llama2 supports 4K tokens. We should increase the defaults on the lambda to 4K

Hello,
This is a great project. This was really helpful. I was able to create run this locally. But, creating the docker container is giving me a unknown service error : it is not able to find bedrock-runtime. It would be great if you can help me with this.

Support Conversational Memory with Amazon Bedrock. Opensearch Conversational Memory is in preview.

With Bedrock's upcoming support for Mistral AI models, we have an opportunity to integrate these models into our serverless-rag-demo project. This integration would enable Retrieve and Generate (RAG) functionality over Mistral's AI capabilities. Given Mistral's strengths in summarization and text generation, adding its models to our project could significantly enhance the RAG experience and open up new use cases.
By leveraging Bedrock's serverless architecture, we can plug the Mistral models into our demo with minimal effort