“The agent returned an empty TTS” when action is not opened separately to asking an intent

问题

When I invoke the skill with 'okay Google, ask {skillname} to {utterance}' I get the response 'The agent returned an empty tts" and the conversation closes. It doesn't even seem to hit my backend - I've tried hosting the backend code on both a local server and on AWS and the same issue happens in both cases. There's no log of the request being made either on Stackdriver on Google or on CloudWatch on AWS. The only response I get is the following in the debug tab:

{
  "response": "The agent returned an empty TTS.",
  "expectUserResponse": false,
  "conversationToken": "EosDS2o4d0...",
  "audioResponse": "",
  "ssmlMarkList": [],
  "debugInfo": {
    "sharedDebugInfoList": [
      {
        "name": "Response Validation",
        "debugInfo": "The agent returned an empty TTS.",
        "subDebugEntryList": []
      }
    ],
    "conversationBuilderExecutionEventsList": []
  },
  "visualResponse": {
    "visualElementsList": [
      {
        "displayText": {
          "content": "The agent returned an empty TTS."
        }
      }
    ],
    "suggestionsList": [],
    "agentLogoUrl": ""
  },
  "clientError": 0,
  "is3pResponse": true,
  "clientOperationList": [],
  "projectName": "",
  "renderedHtml": ""
}

Nothing in any of the other tabs. When I try invoking the phrases in the same way on my phone, I just see a loading symbol and again nothing in the logs to say it's even hitting the backend.

Weirdly, this doesn't happen when I say 'okay Google, talk to {skillname} {utterance}' - this works absolutely fine.

All the intents work perfectly well when I've opened my action first, and then invoke them. I've built this using Jovo and published on Alexa also and the problem is only happening on Google.

来源：https://stackoverflow.com/questions/59054593/the-agent-returned-an-empty-tts-when-action-is-not-opened-separately-to-asking