This message was deleted.

codepipeline to run pulumi with an APIG-integrated lambda that can start the pipeline? end users would need to know how make API calls but that's probably a smaller ask than installing the pulumi CLI

Really -- if you are only making an S3 bucket, you might simply make direct calls to the AWS sdk directly. I lot less overhead. https://github.com/aws/aws-sdk-go

@important-belgium-68902 no, it's not possible, the automation API uses the Pulumi CLI, but your Go binary can handle and manage installing the Pulumi CLI and logging in for them. You can also use package dependencies to figure it out too, like

depends_on

in Homebrew @stocky-wire-60053 the problem with using the aws-sdk-go is that you have to use your own logic to decide if the bucket exists, it's imperative and so therefor you need a lot of control logic

"therefor you need a lot of control logic" ... maybe, but checking for bucket existence doesn't seem like much additional logic as opposed to the complexities of Pulumi in this case -- but I may be ignorant of an important detail.

I'll actually be provisioning a lot more than just a S3 bucket. I simplified it down to focus more on the main issue, which is the deployment and distribution story for apps using Automation API. I like @billowy-army-68599’s suggestion of installing the CLI through the Go app. I can use that as a starting point. Amazon CDK is a fallback option for me right now, but we want this to be a multicloud solution later on and didn't want to tie ourselves to AWS CDK from the start.

CDK doesn't support this sort of thing

So technically this is possible. However, it will require quite a bit of refactoring of pulumi code. We have an open issue for it: https://github.com/pulumi/pulumi/issues/7219

That's exactly the functionality I was looking for @bland-chef-60652. As jaxxstorm mentioned, CDK doesn't support the usecase I'm after either and I still have to use a CLI. Thanks for the help everyone, I have clarity on what my options are now.