Phonic NPM | npm.io

Phonic Node.js SDK

Node.js library for the Phonic API.

Installation

npm i phonic

Setup

Grab an API key from the Phonic API Keys section and pass it to the Phonic constructor.

import { Phonic } from "phonic";

const phonic = new Phonic("ph_...");

Usage

Get voices

const { data, error } = await phonic.voices.list({ model: "shasta" });

if (error === null) {
  console.log(data.voices);
}

Get voice by id

const { data, error } = await phonic.voices.get("meredith");

if (error === null) {
  console.log(data.voice);
}

Get conversation by id

const { data, error } = await phonic.conversations.get("conv_b1804883-5be4-42fe-b1cf-aa84450d5c84");

if (error === null) {
  console.log(data.conversation);
}

Get conversation by external id

const { data, error } = await phonic.conversations.getByExternalId("CAdb9c032c809fec7feb932ea4c96d71e1");

if (error === null) {
  console.log(data.conversation);
}

Speech-to-speech via WebSocket

To start a conversation, open a WebSocket connection:

const { data, error } = await phonic.sts.websocket({
  input_format: "mulaw_8000",

  // Optional fields
  system_prompt: "You are a helpful assistant.",
  welcome_message: "Hello, how can I help you?",
  voice_id: "meredith",
  output_format: "mulaw_8000"
});

if (error !== null) {
  throw new Error(`Failed to start conversation: ${error.message}`);
}

const { phonicWebSocket } = data;

Stream input (user) audio chunks:

phonicWebSocket.audioChunk({
  audio: "...", // base64 encoded audio chunk
});

Process messages that Phonic sends back to you:

phonicWebSocket.onMessage((message) => {
  switch (message.type) {
    case "input_text": {
      console.log(`User: ${message.text}`);
      break;
    }

    case "audio_chunk": {
      // Send the audio chunk to Twilio, for example:
      twilioWebSocket.send(
        JSON.stringify({
          event: "media",
          streamSid: "...",
          media: {
            payload: message.audio,
          },
        }),
      );
      break;
    }
  }
});

Update the system prompt mid-conversation:

phonicWebSocket.updateSystemPrompt({
  systemPrompt: "..."
})

Set an external id for the conversation (can be the Twilio Call SID, for example):

phonicWebSocket.setExternalId({
  externalId: "..."
})

To end the conversation, close the WebSocket:

phonicWebSocket.close();

You can also listen for close and error events:

phonicWebSocket.onClose((event) => {
  console.log(
    `Phonic WebSocket closed with code ${event.code} and reason "${event.reason}"`,
  );
});

phonicWebSocket.onError((event) => {
  console.log(`Error from Phonic WebSocket: ${event.message}`);
});