Flutter Retell

widget.dart

import 'dart:typed_data';
import 'package:record/record.dart';
import 'package:flutter/material.dart';
import 'package:web_socket_channel/web_socket_channel.dart';
import 'package:http/http.dart' as http;
import 'dart:async';
import 'package:mp_audio_stream/mp_audio_stream.dart';

var apiToken = 'ADD API TOKEN HERE';

class AudioTestFinalEcho extends StatefulWidget {
  const AudioTestFinalEcho({super.key});

  @override
  State<AudioTestFinalEcho> createState() => AudioTestFinalEchoState();
}

class AudioTestFinalEchoState extends State<AudioTestFinalEcho> {
  final record = AudioRecorder();
  late WebSocketChannel _channel;
  final audioStream = getAudioStream();
  bool _isPlaying = false;
  StreamSubscription? _audioStreamSubscription;

  @override
  void initState() {
    super.initState();
    _registerCall();
    _initializeAudioStream();
  }

  Future<void> _registerCall() async {
    final response = await http.get(
      Uri.parse('https://safe-gorge-65703-197182a5b4e2.herokuapp.com/call-id'),
      headers: {
        'X-API-TOKEN': apiToken,
        'Content-Type': 'application/json',
      },
    );
    var callId = response.body.replaceAll(RegExp(r'^"|"$'), '');
    print('call id: $callId');
    _initializeWebsocketChannel(callId);
  }

  Future<void> _initializeAudioStream() async {
    // 1 channel since its just speech, we don't care about L/R speakers
    // todo: experiment with higher or lower sample rates. lower is less data, higher is better quality
    // bufferMilliSec is the maximum amount of audio that can be played in one conversational turn
    audioStream.init(channels: 1, sampleRate: 24000, bufferMilliSec: 25000);
  }

  Float32List _convertToFloat32(Uint8List data) {
    // Assuming the input data is in PCM 16-bit signed little-endian format
    final int len = data.length ~/ 2;
    final Float32List float32Data = Float32List(len);
    final ByteData byteData = ByteData.sublistView(data);
    for (int i = 0; i < len; i++) {
      float32Data[i] = byteData.getInt16(i * 2, Endian.little) / 32768.0;
    }
    return float32Data;
  }

  void _initializeWebsocketChannel(String callId) {
    print('Initializing websocket channel');
    _channel = WebSocketChannel.connect(
        Uri.parse('wss://api.retellai.com/audio-websocket/$callId'));
    print('Channel initialized');
    _startPlayerStream();
    _startRecording();
  }

  void _startPlayerStream() {
    _audioStreamSubscription = _channel.stream.listen((data) {
      print('data ${data.length}');
      if (data is List<int>) {
        Float32List convertedData = _convertToFloat32(Uint8List.fromList(data));
        audioStream.push(convertedData);
        if (!_isPlaying) {
          print('playing');
          setState(() {
            _isPlaying = true;
          });
          audioStream.resume();
        }
      } else {
        // todo: clear the audioStream
        // setState(() {
        //   _isPlaying = false;
        // });
      }
    });
  }

  Future<void> _startRecording() async {
    if (!await record.hasPermission()) {
      print('Missing recording permission...');
    }

    const RecordConfig config = RecordConfig(
      encoder: AudioEncoder.pcm16bits,
      sampleRate: 24000,
      numChannels: 1,
    );

    Stream<Uint8List>? stream = await record.startStream(config);

    stream.listen((chunk) async {
      _channel.sink.add(chunk);
    });
  }

  @override
  void dispose() {
    record.dispose();
    _channel.sink.close();
    super.dispose();
    _audioStreamSubscription?.cancel();
    audioStream.uninit();
  }

  @override
  Widget build(BuildContext context) {
    return Scaffold(
      appBar: AppBar(
        title: const Text('Audio Test'),
      ),
      body: const Center(
        child: Column(
          mainAxisAlignment: MainAxisAlignment.center,
          children: [],
        ),
      ),
    );
  }
}