'{"transcriber":{"provider":"deepgram","model":"nova-2","keywords":["Qatar Airways:1","Singapore Airlines:1"]}}' {"message":["transcriber.each keyword must be in the format 'word' or 'word:number'. for example, 'companyname' or 'companyname:1'"],"error":"Bad Request","statusCode":400}% why am I getting this error code?

You are getting this error because the

keywords

array only accepts single words (no spaces or punctuation). For multi-word phrases like "Qatar Airways" or "Singapore Airlines", you should use the

keyterm

array instead. Correct example:

json
{
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2",
    "keywords": ["Qatar:1", "Singapore:1"],
    "keyterm": ["Qatar Airways", "Singapore Airlines"]
  }
}

Use

keywords

for single words and

keyterm

for phrases ([see docs](https://docs.vapi.ai/customization/custom-keywords)). Source: - [Deepgram Keywords and Keyterm Prompting](https://docs.vapi.ai/customization/custom-keywords)

The error happens because Vapi expects keywords as plain strings, not with weights like “:1”. Use ["Qatar Airways", "Singapore Airlines"] instead of ["Qatar Airways:1", "Singapore Airlines:1"]. I can help fix your transcriber config and test the right Deepgram setup. Are you setting this directly in your assistant config or sending it via API? @andre_serra