README.md

Serverless RAG with Lambda, S3, and LanceDB - Testing

This data ingestion pipeline allows you to create embeddings from your PDF documents and make them available to LanceDB in your Lambda function.

Prerequisites

As you'll be running this locally, make sure your aws cli is configured with permissions to PUT files on S3 and invoke models on Amazon Bedrock.

Usage

No Authentication

cd testing
chmod u+x test-no-auth.sh # ensure you can execute the file
./test-no-auth.sh <your-stack-name> # find it in ./samconfig.toml

With IAM Authentication

You will need to ensure you have your AWS credentials as environment variables in order for this script to work.

# AWS_ACCESS_KEY_ID
# AWS_SECRET_ACCESS_KEY
# AWS_SESSION_TOKEN
# AWS_REGION
cd testing
chmod u+x test-with-auth.sh # ensure you can execute the file
./test-with-auth.sh <your-stack-name> # find it in ./samconfig.toml

Sample Output

 Based on the context provided, Amazon Bedrock is a fully managed service that makes base models from Amazon and third-party model providers accessible through an API. Some key points about Amazon Bedrock:

- It allows users to access and explore various AI/ML models from Amazon and other providers through APIs, text/image/chat playgrounds in the AWS Management Console, and examples. 

- It supports models from providers like Amazon, Anthropic, Cohere, Stability.ai, etc. for tasks like text generation, image generation, conversations. 

- It provides capabilities like text/image/chat playgrounds, examples library, API access, embeddings generation, provisioned throughput for discounted inference pricing, fine-tuning of models, and model invocation logging. 

- It is available in various AWS regions and users are charged only for the specific services/models they use based on volume of input/output tokens and provisioned throughput purchases.

So in summary, Amazon Bedrock is a managed service that provides unified access and capabilities for various AI/ML models from different providers through its APIs and console for users to explore and utilize for their applications and use cases.

Implementing within a React.js application

If you wish to invoke this function URL within a React application, you will need to use a request library that supports streaming. Inside the React folder, you can see a simple example using the MIT licensed @microsoft/fetch-event-source package and also using AWS IAM auth.

Usage

You will need to ensure you have your AWS credentials as environment variables in order for this script to work.

# AWS_ACCESS_KEY_ID
# AWS_SECRET_ACCESS_KEY
# AWS_SESSION_TOKEN
# AWS_REGION
export STACK_NAME=<your-stack-name>  # find it in ./samconfig.toml
export LAMBDA_ENDPOINT_URL=$(aws cloudformation describe-stacks --stack-name $STACK_NAME --query 'Stacks[0].Outputs[?OutputKey==`StreamingRAGFunctionURL`].OutputValue' --output text)

cd testing/react
npm install
npm run start
# A UI will open at localhost:3000 that you can then ask questions to.