Pulumi AI
Create cloud infrastructure with Pulumi Automation API npx pulumi-ai.
Note: This is an experimental AI experience for interactively building cloud infrastructure using GPT. It will likely do surprising and interesting things, and will make mistakes! You have the option to provide access to deploy infrastructure directly into your cloud account(s), which puts incredible power in the hands of the AI, be sure to use with approporiate caution.
Running
To use the CLI tool, you must configure the following:
- Download and install
pulumi. - Get an OpenAI API Key and make it available as
OPENAI_API_KEY. - Login to Pulumi via
pulumi login, or else by exporting a Pulumi Access Token asPULUMI_ACCESS_TOKEN. - Configure your cloud credentials for your target Cloud environment (AWS, Azure, Google Cloud, Kubernetes, etc.).
Note: The
pulumi-aiCLI works best with AWS today, but can be used with any cloud.
Then run:
npx pulumi-aiYou can ask for any infrastructure you want, then use these commands to take actions outside of interacting with the AI:
!quit: Exit the AI and cleanup the temporary stack.!program: See the Pulumi program that has been developed so far.!stack: See details of the Pulumi stack that has been deployed so far.!verbose: Turn verbose mode on for debugging interactions with GPT and Pulumi.!open <output>: Open a URL from the given stack output name in the system browser.
The following environment variables are also available to configure the GPT AI used:
OPENAI_MODEL: Select one of the valid OpenAI Models, suchas asgpt-4(default, and most accurate but slow) orgpt-3.5-turbo(not as accurate but much faster).OPENAI_TEMPERATURE: Configure the temperature to tune the AI to be more predicatable (lower values) or more creative (higher values).
Examples
Welcome to Pulumi AI.
Your stack: https://app.pulumi.com/luke/pulumi-ai/dev/resources
What cloud infrastructure do you want to build today?
> An AWS VPC
create aws:ec2/vpc:Vpc my-vpc ...
created my-vpc
vpcCidrBlock: 10.0.0.0/16
vpcId: vpc-016f35c7f078ab9e8
> Add three private subnets
create aws:ec2/subnet:Subnet private-subnet-2-2 ...
create aws:ec2/subnet:Subnet private-subnet-1-1 ...
create aws:ec2/subnet:Subnet private-subnet-3-3 ...
created private-subnet-2-2
created private-subnet-3-3
created private-subnet-1-1
privateSubnetIds: subnet-08add1c8ae97e3cfb,subnet-0cb4ea675f7ac64fe,subnet-0f9a22c87e766fa17
vpcId: vpc-016f35c7f078ab9e8
> Remove one of the subnets
delete aws:ec2/subnet:Subnet private-subnet-3-3 ...
deleted private-subnet-3-3
privateSubnetIds: subnet-08add1c8ae97e3cfb,subnet-0cb4ea675f7ac64fe
vpcId: vpc-016f35c7f078ab9e8
>
Welcome to Pulumi AI.
Your stack: https://app.pulumi.com/luke-pulumi-corp/pulumi-ai/dev/resources
What cloud infrastructure do you want to build today?
> an s3 bucket
create aws:s3/bucket:Bucket my-bucket ...
created my-bucket
Stack Outputs:
bucketName: my-bucket-cc63555
> add an index.html file that says "Hello, world!" in three languages
update aws:s3/bucket:Bucket my-bucket ...
updated my-bucket
create aws:s3/bucketObject:BucketObject index ...
created index
Stack Outputs:
bucketName: my-bucket-cc63555
> give me a url for that index.html file
update aws:s3/bucket:Bucket my-bucket ...
updated my-bucket
create aws:s3/bucketPolicy:BucketPolicy bucketPolicy ...
created bucketPolicy
Stack Outputs:
bucketName: my-bucket-cc63555
indexUrl: http://undefined/index.html
> That gave an undefined url. Can you fix it?
update aws:s3/bucket:Bucket my-bucket ...
updated my-bucket
Stack Outputs:
bucketName: my-bucket-cc63555
indexUrl: http://my-bucket-cc63555.s3-website-us-west-2.amazonaws.com/index.html
> Great - that worked! Now make it fancier, and add some color :-)
update aws:s3/bucket:Bucket my-bucket ...
updated my-bucket
update aws:s3/bucketObject:BucketObject index ...
updated index
Stack Outputs:
bucketName: my-bucket-cc63555
indexUrl: http://my-bucket-cc63555.s3-website-us-west-2.amazonaws.com/index.html
> !program
<snip>
> !quit