Manage AI Object Storage with Terraform

Use Terraform to manage CoreWeave AI Object Storage infrastructure as code

Terraform providers are essential plugins that enable Terraform to manage infrastructure resources on various platforms.

You can use Terraform to manage access policies and configure bucket lifecycles and versioning. As CoreWeave AI Object Storage is S3-compatible, you can manage it using either the CoreWeave Terraform provider or the official Terraform AWS Provider. Both providers translate Terraform's declarative syntax into S3-compatible API calls.

If you want to manage data and objects, you can use any S3-compatible tool to manage your data. See How-To: Manage Objects to learn how to use CoreWeave AI Object Storage with s3cmd, Boto3, or the AWS CLI.

Configure Terraform

These examples show how to set up the Terraform provider with CoreWeave AI Object Storage.

Endpoint selection

• Use the LOTA endpoint, http://cwlota.com, when running inside a CoreWeave cluster. The LOTA endpoint routes to the LOTA cache for best performance.
• Use the primary endpoint, https://cwobject.com, when running outside of a CoreWeave cluster.

The AWS provider example configuration includes the required overrides for compatibility with CoreWeave AI Object Storage, while the native CoreWeave provider example configuration does not require these overrides.

CoreWeave Provider

Example

1
terraform {
2
  required_providers {
3
    coreweave = {
4
      source  = "coreweave/coreweave"
5
      version = ">= 0.3.0"
6
    }
7
  }
8
  required_version = ">= 1.2.0"
9
}
10

11
provider "coreweave" {
12
  token = env.COREWEAVE_API_TOKEN
13
  s3_endpoint = env.COREWEAVE_S3_ENDPOINT # If using LOTA, set to http://cwlota.com
14
}
15

Some aspects to note in this example:

• token is set as an environment variable for security. Get your API access token from the CoreWeave Access Tokens page in Cloud Console.
• s3_endpoint defaults to https://cwobject.com, which is the primary CoreWeave AI Object Storage endpoint. If using LOTA, set to http://cwlota.com.
• version is set to >= 0.3.0 because 0.3.0 is the first version to support AI Object Storage.

AWS Provider

Example

1
terraform {
2
  required_providers {
3
    aws = {
4
      source  = "hashicorp/aws"
5
      version = "5.67.0"
6
    }
7
  }
8

9
  required_version = ">= 1.2.0"
10
}
11

12
provider "aws" {
13
  access_key = env.AWS_ACCESS_KEY
14
  secret_key = env.AWS_ACCESS_SECRET
15
  region     = local.region
16

17
  # These two values must be true so Terraform doesn't attempt STS authentication.
18
  skip_credentials_validation = true
19
  skip_requesting_account_id  = true
20

21
  # CoreWeave uses different region names, so validation must be skipped.
22
  skip_region_validation      = true
23
  s3_use_path_style           = false
24

25
  endpoints {
26
    s3 = local.endpoint
27
  }
28
}
29

30
locals {
31
  region   = "US-EAST-03A"
32
  endpoint = "https://cwobject.com"
33
}
34

Some aspects to note in this example, highlighted above:

• access_key and secret_key are set as environment variables for security.
• skip_credentials_validation, skip_requesting_account_id, and skip_region_validation options are set to true to bypass AWS-specific validation checks.
• In this example, region is set to US-EAST-03A. Note that CoreWeave AI Object Storage uses different region names than AWS.
• endpoint is set to https://cwobject.com, which is the CoreWeave AI Object Storage endpoint.
• The s3_use_path_style option is set to false to use virtual-hosted-style URLs.

Object Storage resources

You can use Terraform to manage organization access policies and bucket access policies, and configure bucket lifecycles and versioning. For detailed examples and documentation, see the CoreWeave provider registry documentation.

Additional resources

• To learn more about using the Terraform AWS provider with Object Storage, see the S3 section of the official documentation.
• To learn more about the CoreWeave Terraform provider, see the CoreWeave provider documentation.
• Use the CoreWeave provider in your OpenTofu configuration.