Import and reconcile OpenAI resources

Import existing OpenAI resources instead of recreating them. A safe adoption starts with configuration that matches the remote resource, previews and applies the import, and produces a no-op plan before any intended update.

Import blocks require Terraform 1.5 or later.

Declare and import resources

Declare each existing resource using its current settings, then add an import block with the ID format from the provider reference:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
resource "openai_project" "existing" {
  name = "existing-project"
}

resource "openai_group" "existing" {
  name = "existing-group"
}

resource "openai_project_service_account" "existing" {
  project_id = openai_project.existing.project_id
  name       = "existing-service-account"
}

import {
  to = openai_project.existing
  id = "proj_123"
}

import {
  to = openai_group.existing
  id = "group_123"
}

import {
  to = openai_project_service_account.existing
  id = "proj_123/svc_acct_123"
}

Preview the imports in a saved plan:

terraform plan -out=tfplan
terraform show tfplan

The plan should show the imports without proposing updates to the remote resources. If it proposes updates, make the configuration match the current settings before continuing. Apply the saved plan to perform the imports, then run another plan:

terraform apply tfplan
terraform plan

The second plan should report no changes. You can keep the import blocks in your configuration as a record of how Terraform adopted the resources.

Common import ID formats include:

Check the provider reference for the exact format of every resource.

Read resources without adopting them

Use data sources when Terraform needs current information but another system owns the resource. The provider includes data sources for projects, groups, roles, users, role assignments, rate limits, model permissions, hosted-tool permissions, spend alerts, data retention, and certificates.

For example, read an existing project and its current groups:

1
2
3
4
5
6
7
8
9
10
11
data "openai_project" "existing" {
  project_id = var.project_id
}

data "openai_project_groups" "existing" {
  project_id = data.openai_project.existing.project_id
}

output "project_groups" {
  value = data.openai_project_groups.existing.groups
}

Detect and reconcile drift

Run a normal plan to read the current OpenAI settings and compare them with the desired values in your Terraform configuration:

terraform plan -detailed-exitcode

Exit code 0 means there are no changes, 2 means the plan contains changes, and 1 means Terraform encountered an error.

If the plan shows a setting that changed outside Terraform:

1. Determine whether the change was intentional.
2. To keep the remote change, update the Terraform configuration to match it.
3. To undo the remote change, review and apply the plan to restore the configured value.
4. Run another plan and require a no-op result.

Understand removal behavior

Removing a resource block removes the resource from Terraform state, but it doesn’t always delete or reset the same kind of remote object: