I'm trying to perform introspection of the AWS resources (

pulumi_aws

) so that we can dynamically define configuration templates. I'm running into the following issue:

Traceback (most recent call last):
  File "~/Documents/workspace/ops/pulumi/pulumi-config.py", line 17, in <module>
    import pulumi_aws as aws
  File "~/.virtualenvs/pulumi/lib/python3.9/site-packages/pulumi_aws/__init__.py", line 5, in <module>
    from . import _utilities
  File "~/.virtualenvs/pulumi/lib/python3.9/site-packages/pulumi_aws/_utilities.py", line 15, in <module>
    import pulumi.runtime
ModuleNotFoundError: No module named 'pulumi.runtime'; 'pulumi' is not a package

Any ideas on how I can get

pulumi.runtime

recognized?

Can you show some code for context?

Definitely. Here's the code:

import boto3
import docopt
import pulumi
import pulumi_aws as aws

def driver(args):
    """Main driver for the management CLI tool
    """
    obj = dir("aws.organizations.Organization")
    print(obj)

if __name__ == '__main__':
    driver(docopt.docopt(__doc__))

It's not clear to me what you mean by "dynamically define configuration templates". Can you explain a little more about what you're trying to build?

That's not the best explanation. I'm trying to build a framework for reviewing the objects with types from

pulumi_aws

in a Pulumi stack and providing users a way to what constructor parameters they'll need to enter for those objects. For example, with

aws.organizations.Organization

, I'd want to know that I'd need to set the parameters of

aws_service_access_principals

,

enabled_policy_types

,

feature_set

are available (source). We want to fill these parameters with values from a data store. So for

aws.organizations.Organization

, we should be able to retrieve the values of

aws_service_access_principals

from a document. We want to have configuration provider similar to the secrets provider that are available with Pulumi. The way that configuration is currently tied to a stack, it seems that the intended association should be environment yaml to stack. This becomes limiting when we want to deploy multiple versions of the stack (for example, multiple RDS/Aurora instances) with different configuration.

Oof. That's rough.