Is better to use AWS Kendra or Pinecone as a vectorial database?

Question

Hi Guys!

I will like to know if it is better to use AWS Kendra or Pinecone as vectorial Database, and also if it is best to use Pinecone from the AWS Marketplace or to use Pinecone from their own site.

I want to build my own chatbot. This would be the use case. The users will upload documents to the given Vectorial DB (Kendra or Pinecone). Then a Lambda function will be called by the user with a question. Then inside the Lambda function I will call Kendra or Pinecone along with the user question to get a reponse and return that response to the user.

I want to to use Bedrock, but Bedrock is only avaialable for now in europe in frankfurt, and Kendra is not available in frankfurt

For my use case maybe is best to use Pinecone?

Thank you guys!

Accepted Answer

As whether one vector db is better than another, it depends on functional and non functional requirements. Nevertheless, given you want to use bedrock, my suggestions would be to use pine cone (marketplace or dockerized in ec2, Ecs or Eks), or either use opensearch serverless or RDS Aurora with pg vector, both available in Frankfurt.

Answer

You should also try Vectara: it's a retrieval-augmented-generation service and super easy to build your own chatbot with. There is an API endpoint to ingest your documents, and then Vectara takes care of the chunking, embedding and storing in the vector store. And from the lambda you simply call the query API to get the response. Super easy to setup and use (see docs for more details).

It's available on the AWS marketplace

It's available on the [AWS marketplace](https://aws.amazon.com/marketplace/pp/prodview-kgovs73t7qsve)

Is better to use AWS Kendra or Pinecone as a vectorial database?

Relevant content