Looking for advice on automating bot NLP model retraining via the Genesys Cloud CLI. The deployment pipeline succeeds for bot structure and intents, but the final step to trigger model training consistently fails.
Environment details:
- Region: AU-1 BYOC
- Genesys Cloud CLI: v2.1.0
- Bot: Customer Service Bot (v1.2.0)
- Terraform Provider: v1.45.0
The workflow uses genesyscloud bot nlp train to initiate the training job. The command returns a 500 Internal Server Error immediately. No training job ID is generated in the response.
$ genesyscloud bot nlp train --bot-id 12345
Error: 500 Internal Server Error
Body: {
"errors": [
{
"code": "internal_error",
"message": "An unexpected error occurred while processing the request."
}
]
}
Checked the API documentation for /api/v2/bots/{botId}/nlp/train:
The server will respond with a 202 Accepted and a job ID if the request is valid. A 500 error indicates a backend service failure or invalid payload structure.
The payload is minimal, just the bot ID. Verified the bot ID exists and is active via genesyscloud bot get --id 12345. The bot has 15 intents and 500 utterances. No syntax errors in the intent definitions. Tried reducing the utterance count to 100, same result. Tried creating a new bot with 1 intent, same result.
The issue is specific to the NLP training endpoint. Other bot operations (create, update, deploy) work fine. Checked the Genesys Cloud status page, no incidents reported for AU-1 BYOC.
Is there a known issue with the NLP training service in AU-1 BYOC? Or is there a hidden validation rule causing the 500? The error message is unhelpful. Need a way to debug this or bypass the CLI to trigger training via direct API call with more verbose logging.