Hi @Frank and @Jay, is it possible to build docker image runtime and specify that for the Lambda?

Hey @Dillon Peterson, it’s on the roadmap - https://github.com/serverless-stack/serverless-stack/discussions/177

Would be for ML Model Deployment (inference endpoint) via Lambda.

I see.. are you building the docker image, or reusing an existing public ECR image?

Building a docker image.

We’d need to figure out how to run the lambda code locally when running inside

sst start

I don't currently have a sample repo, just got started with this. 😬

I see. You can actually do what’s described in the blog post in SST. Follow all the steps till

docker push

the image to ECR. And then in ur SST app, do something like this to create a Lambda function with docker runtime: (I haven’t tested this code.)

import * as ecr from '@aws-cdk/aws-ecr';

const repo = ecr.Repository.fromRepositoryName(this, Repo', 'bert-lambda');

new lambda.DockerImageFunction(this, 'ECRFunction', {
  code: lambda.DockerImageCode.fromEcr(repo),
});

Like I mentioned in the discussion,

sst start

won’t be able to auto rebuild (ie. docker build) the docker image on code changes. That is not supported yet.

Many thanks @Frank!! I appreciate it!~