Sent Voice Agents for TypeScript

An AI agent that takes and places phone calls through Sent, on the Sent Voice Agents protocol 1.0.0, for Node.

Status: Preview, with the protocol. Generated and rendered by sent-dm-sdk-pipeline from the protocol at v1.0.0; every release passes the protocol's conformance scenarios in CI.

Read the Voice Agents docs to get started.

Install

npm install @sentdm/voice-agents

Node 20.19 or later. The one dependency is ws. The package is ES modules; require() works on the Node versions it supports.

Quickstart

import { Agent, Session, UserSaid } from '@sentdm/voice-agents';

class Support extends Agent {
  override async onSessionStart(session: Session) {
    session.say('Hi, this is Acme support. How can I help?');
  }

  override async onUserSaid(session: Session, said: UserSaid) {
    if (said.text.includes('person')) {
      session.say('Transferring you now.');
      await session.transfer('+14155550100');
    } else {
      session.say(`You said ${said.text}`);
    }
  }
}

await new Support({ apiKey: process.env.SENT_API_KEY!, agentId: 'agent_...' }).serve();

serve() connects to Sent with the account API key and serves calls until SIGINT or SIGTERM; run() does the same without the signal handlers, and stop() ends it. Each session's handlers run in order on their own chain, so sessions never wait for one another. The runtime answers heartbeats, opens a new connection when one drains or is lost, and resumes the sessions that can move; the handlers only see calls.

| Handler | When | | ------------------------------- | ---------------------------------------------------------------------------------------------- | | onSessionStart(session) | A call reached the agent. In a text session, say() to speak | | onUserSaid(session, said) | Text sessions: the caller said something; said.correction means it replaces the previous one | | onAudio(session, samples) | Audio sessions: the caller's audio, 16-bit mono PCM at session.sampleRate | | onToolResult(session, result) | Sent answered a tool call; the awaiting promise gets the same result | | onSessionEnd(session, end) | The call left the agent; end.reason says why | | onSessionLost(session) | The connection was lost and the session could not be resumed |

On a Session: say(text, { last, interruptible }), clear(), sendAudio(samples), and the tools transfer(to), hangup(), sendDtmf(digits), hold(held), each resolving with Sent's ToolResult. session.from, to, direction, sessionType, language, context, sampleRate carry what session.start said.

@sentdm/voice-agents/frames holds every frame as a type, with decode() and encode() that check the protocol's rules; it is generated from the protocol and not edited by hand.

Develop

npm install
npm run typecheck && npm test && npm run build    # types, the tests, dist/
npm run format                                    # prettier

The conformance scenarios run in CI through the protocol's action, against node dist/conformance.js. Releases are cut by release-please from main and published to npm from GitHub Actions.