Struggling to figure out why the virtual agent intent matching fails when processing historical Zendesk ticket data. We are migrating from Zendesk to Genesys Cloud and trying to use the Virtual Agent to categorize old tickets. The API call to /api/v2/agents/virtual/requests returns a 400 Bad Request with error_code: "INVALID_INPUT". The payload includes the ticket body as input_text, but the system rejects it. In Zendesk, we just tagged tickets manually, but here we want automated classification. The environment is Genesys Cloud EU Central. We are using the default NLP model. The error log shows "message": "Input text exceeds maximum length for intent classification". However, the text is only 200 characters. We tried truncating it, but the error persists. Is there a specific format required for the input_text field? We are following the standard migration guide, but this step is blocking our bulk processing. The goal is to map Zendesk tags to Genesys interaction attributes. Any help would be appreciated. We are stuck on this validation step.
The problem here is the payload structure. Zendesk text often contains HTML or excessive whitespace that breaks the strict schema. Sanitize the input before sending. Use the bulk export API to clean data first. Check this guide for valid JSON formatting.
This looks like a schema validation error. The input_text field in the virtual agent request is strict about character limits and encoding.
Trim the ticket body to under 1000 characters and ensure it’s plain text. Zendesk exports often include hidden HTML tags that break the JSON schema.