Platform API 422 on nice_cxone_ai_bot_dialog_flow intent_mapping binding in provider 1.19.7

The 02:45 JST bot deployment pipeline stalled again. Terraform 1.9.8 paired with nice-cxone 1.19.7 throws a 422 Unprocessable Entity when applying changes to the nice_cxone_ai_bot_dialog_flow resource. The provider is rejecting the intent_mapping block entirely.

Logs show the exact rejection:

Error: POST /api/v2/ai-bots/dialog-flows returned 422
{"errors":[{"detail":"intent_mapping is invalid. Bot reference mismatch or empty payload."}]}

State backup doesn’t show any drift. Local plan output looks clean. The POST to /api/v2/ai-bots/dialog-flows is sending null for the intent_mapping array. The 1.19.7 provider appears to be stripping nested intent objects before the request hits the platform. Rolled back to 1.19.5. Same result. Cleared the .terraform/providers cache. Ran terraform init -upgrade. Still choking on the apply phase.

The dialog flow itself works fine in the CXone UI. Manually adding intents through Architect bypasses the issue completely. Automation won’t handle the binding without manual intervention. The error payload returns a validation error on the intent_mapping field, claiming it requires a valid bot_id reference even though the parent resource already defines it.

Tried flattening the mapping structure. Tried passing the intent IDs as a comma-separated string. Tried wrapping it in a dynamic block. Nothing’s sticking. The console just spits out the 422 and halts the run. Queue sync jobs are backing up because the bot provisioning step won’t complete. Doing jack all with the current provider build.

The state file already tracks the correct bot_id. The provider can’t serialize the nested JSON correctly. Running tf apply -target=nice_cxone_ai_bot_dialog_flow.main just loops the same validation failure. You’ll hit the exact same behavior on a fresh workspace. The API gateway drops the request before it even hits the bot service layer.