project-bootstrap.md

This is a Next.js project bootstrapped with create-next-app. But using SST as deployment engine.

Next.js with AWS as infrastructure provider using SST and OpenNext

Before start:

1. Configure AWS Credentials with unix

   $ cat ~/.aws/credentials
   
   [default]
    aws_access_key_id=XXXXXXXXXXXXXXXXXXXXX
    aws_secret_access_key=XXXXXXXXXXXXXXXXXXXXXXXXXXX

2. Install the SST CLI
   The CLI helps you manage your SST apps.

   curl -fsSL https://ion.sst.dev/install | bash

Create the project

Create the app:

npm create-next-app@latest multinamespace-rag
pnpm dlx create-next-app@latest multinamespace-rag
cd multinamespace-rag

Check: SST/Next.js

Init SST(infrastructure)

sst init

This will create file sst.config.ts at root, and wraps your dev script as "sst dev next dev"
To run your app in dev mode run pnpm dev that executes under the hood sst dev next dev

App Secrets

In this project we didn't use .env.* files.
we'll use a couple of secret API keys but will set those values as sst.Secret component

• Secret are encrypted when they are stored in your state file or in a function package. Secrets are encrypted and stored in an S3 bucket in your AWS account. If used in your app config, they’ll be encrypted in your state file as well. If used in your function code, they’ll be decrypted and stored in the function package.

Create a secret:
At sst.config.ts

const secret = new sst.Secret("MySecret");

We created two, one for OpenAI and other for PineCone api keys sst.config

...
async run() {
   const openAIApiSecret = new sst.Secret('OpenAIApiSecret');
   const pineconeApiSecret = new sst.Secret('PineconeApiSecret');
...

To set the values for secrets you just run CLI cmd

sst secret set MySecret my-secret-value

Then to use in code the secret just call it as member of Resource object
e.g at: src/lib/embeddings-utils.ts

export function openAIEmbeddings(
  batchSize = 512,
  openaiEmbeddingsModel = constants.embeddingModel
) {
  const embeddings = new OpenAIEmbeddings({
    openAIApiKey: Resource.OpenAIApiSecret.value,
    model: openaiEmbeddingsModel,
    batchSize: batchSize
  });
  return embeddings;
}