API reference

Voice test

Get a short-lived token to talk to an agent from a browser, without a phone call.

A voice test lets you speak with an agent through a browser microphone instead of a phone line. This endpoint does not start the conversation itself. It returns the WebSocket address of the voice engine and a token that lets one browser open one test call with one agent. The browser voice test guide shows how to use them, and testing without a phone lists the other options.

Note: A voice test needs the voice engine to be running and reachable from the browser at the returned address. The address is operator-configured; when the operator has not set a public one, a local development address is returned.

Create a voice test token

POST /agent-voice-test/{agent_id}

Returns where the browser should connect and a signed token for one test call with this agent. The token is valid for 120 seconds and can be used once. Request a new token for each test call.

The test call is recorded as an ordinary execution: it has a transcript and a recording, and it is charged like a call but without a telephony cost. It is marked with context_details.web_call so you can tell it apart from phone calls. See Executions.

Auth: API key or session token. Role: member and up.

The request has no body.

Request

curl -X POST https://api.sevennadi.si/agent-voice-test/3c9d2f8e-1a2b-4c3d-8e4f-5a6b7c8d9e0f \
  -H "Authorization: Bearer vk-0123456789abcdef0123456789abcdef"

Response 200 OK

{
  "ws_url": "wss://engine.sevennadi.si/web",
  "token": "MDNkMDY0YWEuLi4.0123456789abcdef0123456789abcdef01234567",
  "expires_in": 120
}

Create the token on your server and hand only the token and address to the browser. Your API key should not reach the browser.

Errors