Terraform drift on Genesys AI Bot Language Model version

Architect & Flows
1

Does anyone know why genesyscloud_ai_language_model fails to reconcile version pinning during GitHub Actions deployment?

  1. Define resource with version = "2023-10-01" in HCL.
  2. Run terraform plan.
  3. Output shows no changes despite API returning 2023-11-15.

State file persists old version. Drift detection silent. Using GC CLI v2.1.4. Need force recreation or explicit attribute to lock version.

2

If I remember correctly…

Does anyone know why genesyscloud_ai_language_model fails to reconcile version pinning during GitHub Actions deployment?

The issue likely stems from how the provider handles immutable resource attributes versus mutable metadata. When you pin a version like "2023-10-01", the API may treat subsequent updates to that same logical model as metadata changes rather than a replacement of the underlying resource ID. Terraform often ignores these if the ID remains stable.

Try adding lifecycle { create_before_destroy = true } to your resource block. This forces the provider to recognize the version change as a distinct resource lifecycle event rather than an in-place update. Alternatively, check if the genesyscloud_ai_language_model resource supports a force_new tag or if you need to explicitly reference the id from a data source to ensure state accuracy.

resource "genesyscloud_ai_language_model" "my_model" {
 # ... config ...
 lifecycle {
 create_before_destroy = true
 }
}

This approach mirrors how we handle SIP credential rotations on our BYOC trunks-forcing a clean state transition rather than attempting an atomic update.