mediasoup/app.js

require('dotenv').config();

const express = require('express');
const app = express();
const Server = require('socket.io');
const path = require('node:path');
const fs = require('node:fs');
let https;
try {
  https = require('node:https');
} catch (err) {
  console.log('https support is disabled!');
}
const mediasoup = require('mediasoup');

let worker;
/**
 *
 * videoCalls - Dictionary of Object(s)
 * '<callId>': {
 *  router: Router,
 *  initiatorAudioProducer: Producer,
 *  initiatorVideoProducer: Producer,
 *  receiverVideoProducer: Producer,
 *  receiverAudioProducer: Producer,
 *  initiatorProducerTransport: Producer Transport,
 *  receiverProducerTransport: Producer Transport,
 *  initiatorConsumerVideo: Consumer,
 *  initiatorConsumerAudio: Consumer,
 *  initiatorConsumerTransport: Consumer Transport
 *  initiatorSocket
 *  receiverSocket
 * }
 *
 **/
let videoCalls = {};
let socketDetails = {};

app.get('/', (_req, res) => {
  res.send('Hello from mediasoup app!');
});

app.use('/sfu', express.static(path.join(__dirname, 'public')));

// SSL cert for HTTPS access
const options = {
  key: fs.readFileSync(process.env.SERVER_KEY, 'utf-8'),
  cert: fs.readFileSync(process.env.SERVER_CERT, 'utf-8'),
};

const httpsServer = https.createServer(options, app);

const io = new Server(httpsServer, {
  allowEIO3: true,
  origins: ['*:*'],
});

httpsServer.listen(process.env.PORT, () => {
  console.log('Video server listening on port:', process.env.PORT);
});

const peers = io.of('/');

const createWorker = async () => {
  try {
    worker = await mediasoup.createWorker({
      rtcMinPort: parseInt(process.env.RTC_MIN_PORT),
      rtcMaxPort: parseInt(process.env.RTC_MAX_PORT),
    });
    console.log(`[createWorker] worker pid ${worker.pid}`);

    worker.on('died', (error) => {
      // This implies something serious happened, so kill the application
      console.error('mediasoup worker has died', error);
      setTimeout(() => process.exit(1), 2000); // exit in 2 seconds
    });
    return worker;
  } catch (error) {
    console.error(`[createWorker] | ERROR | error: ${error.message}`);
  }
};

// We create a Worker as soon as our application starts
worker = createWorker();

// This is an Array of RtpCapabilities
// https://mediasoup.org/documentation/v3/mediasoup/rtp-parameters-and-capabilities/#RtpCodecCapability
// list of media codecs supported by mediasoup ...
// https://github.com/versatica/mediasoup/blob/v3/src/supportedRtpCapabilities.ts
const mediaCodecs = [
  {
    kind: 'audio',
    mimeType: 'audio/opus',
    clockRate: 48000,
    channels: 2,
  },
  {
    kind: 'video',
    mimeType: 'video/VP8',
    clockRate: 90000,
    parameters: {
      'x-google-start-bitrate': 1000,
    },
    channels: 2,
  },
  {
    kind: 'video',
    mimeType: 'video/VP9',
    clockRate: 90000,
    parameters: {
      'profile-id': 2,
      'x-google-start-bitrate': 1000,
    },
  },
  {
    kind: 'video',
    mimeType: 'video/h264',
    clockRate: 90000,
    parameters: {
      'packetization-mode': 1,
      'profile-level-id': '4d0032',
      'level-asymmetry-allowed': 1,
      'x-google-start-bitrate': 1000,
    },
  },
  {
    kind: 'video',
    mimeType: 'video/h264',
    clockRate: 90000,
    parameters: {
      'packetization-mode': 1,
      'profile-level-id': '42e01f',
      'level-asymmetry-allowed': 1,
      'x-google-start-bitrate': 1000,
    },
  },
];

const closeCall = (callId) => {
  try {
    if (callId && videoCalls[callId]) {
      videoCalls[callId].receiverVideoProducer?.close();
      videoCalls[callId].receiverAudioProducer?.close();
      videoCalls[callId].initiatorConsumerVideo?.close();
      videoCalls[callId].initiatorConsumerAudio?.close();

      videoCalls[callId]?.initiatorConsumerTransport?.close();
      videoCalls[callId]?.receiverProducerTransport?.close();
      videoCalls[callId]?.router?.close();
      delete videoCalls[callId];
      console.log(`[closeCall] | callId: ${callId}`);
    }
  } catch (error) {
    console.error(`[closeCall] | ERROR | callId: ${callId} | error: ${error.message}`);
  }
};

/*
  - Handlers for WS events
  - These are created only when we have a connection with a peer
*/
peers.on('connection', async (socket) => {
  console.log('[connection] socketId:', socket.id);

  // After making the connection successfully, we send the client a 'connection-success' event
  socket.emit('connection-success', {
    socketId: socket.id,
  });

  // It is triggered when the peer is disconnected
  socket.on('disconnect', () => {
    const callId = socketDetails[socket.id];
    console.log(`disconnect | socket ${socket.id} | callId ${callId}`);
    delete socketDetails[socket.id];
    closeCall(callId);
  });

  /*
    - This event creates a room with the roomId and the callId sent
    - It will return the rtpCapabilities of that room
    - If the room already exists, it will not create it, but will only return rtpCapabilities
  */
  socket.on('createRoom', async ({ callId }, callback) => {
    let callbackResponse = null;
    try {
      // We can continue with the room creation process only if we have a callId
      if (callId) {
        console.log(`[createRoom] socket.id ${socket.id} callId ${callId}`);
        if (!videoCalls[callId]) {
          videoCalls[callId] = { router: await worker.createRouter({ mediaCodecs }) };
          console.log(`[createRoom] Generate Router ID: ${videoCalls[callId].router.id}`);
          videoCalls[callId].receiverSocket = socket;
        } else {
          videoCalls[callId].initiatorSocket = socket;
        }
        socketDetails[socket.id] = callId;
        // rtpCapabilities is set for callback
        callbackResponse = {
          rtpCapabilities: videoCalls[callId].router.rtpCapabilities,
        };
      } else {
        console.log(`[createRoom] missing callId: ${callId}`);
      }
    } catch (error) {
      console.error(`[createRoom] | ERROR | callId: ${callId} | error: ${error.message}`);
    } finally {
      callback(callbackResponse);
    }
  });

  /*
    - Client emits a request to create server side Transport
    - Depending on the sender, a producer or consumer is created is created on that router
    - It will return parameters, these are required for the client to create the RecvTransport 
    from the client.
      - If the client is producer(sender: true) then it will use parameters for device.createSendTransport(params)
      - If the client is a consumer(sender: false) then it will use parameters for device.createRecvTransport(params)
  */
  socket.on('createWebRtcTransport', async ({ sender }, callback) => {
    try {
      const callId = socketDetails[socket.id];
      console.log(`[createWebRtcTransport] socket ${socket.id} | sender ${sender} | callId ${callId}`);
      if (sender) {
        if (!videoCalls[callId].receiverProducerTransport && !isInitiator(callId, socket.id)) {
          videoCalls[callId].receiverProducerTransport = await createWebRtcTransportLayer(callId, callback);
        } else if (!videoCalls[callId].initiatorProducerTransport && isInitiator(callId, socket.id)) {
          videoCalls[callId].initiatorProducerTransport = await createWebRtcTransportLayer(callId, callback);
        } else {
          console.log(`producerTransport has already been defined | callId ${callId}`);
          callback(null);
        }
      } else if (!sender) {
        if (!videoCalls[callId].receiverConsumerTransport && !isInitiator(callId, socket.id)) {
          videoCalls[callId].receiverConsumerTransport = await createWebRtcTransportLayer(callId, callback);
        } else if (!videoCalls[callId].initiatorConsumerTransport && isInitiator(callId, socket.id)) {
          videoCalls[callId].initiatorConsumerTransport = await createWebRtcTransportLayer(callId, callback);
        }
      }
    } catch (error) {
      console.error(
        `[createWebRtcTransport] | ERROR | callId: ${socketDetails[socket.id]} | sender: ${sender} | error: ${
          error.message
        }`
      );
      callback(error);
    }
  });

  /*
    - The client sends this event after successfully creating a createSendTransport(AS PRODUCER)
    - The connection is made to the created transport
  */
  socket.on('transport-connect', async ({ dtlsParameters }) => {
    try {
      const callId = socketDetails[socket.id];
      if (typeof dtlsParameters === 'string') dtlsParameters = JSON.parse(dtlsParameters);

      console.log(`[transport-connect] socket ${socket.id} | callId ${callId}`);

      isInitiator(callId, socket.id)
        ? await videoCalls[callId].initiatorProducerTransport.connect({ dtlsParameters })
        : await videoCalls[callId].receiverProducerTransport.connect({ dtlsParameters });
    } catch (error) {
      console.error(`[transport-connect] | ERROR | callId: ${socketDetails[socket.id]} | error: ${error.message}`);
    }
  });

  /*
    - The event sent by the client (PRODUCER) after successfully connecting to receiverProducerTransport/initiatorProducerTransport
    - For the router with the id callId, we make produce on receiverProducerTransport/initiatorProducerTransport
    - Create the handler on producer at the 'transportclose' event
    */
  socket.on('transport-produce', async ({ kind, rtpParameters, appData }, callback) => {
    try {
      const callId = socketDetails[socket.id];
      if (typeof rtpParameters === 'string') rtpParameters = JSON.parse(rtpParameters);

      console.log(`[transport-produce] callId: ${callId} | kind: ${kind} | socket: ${socket.id}`);

      if (kind === 'video') {
        if (!isInitiator(callId, socket.id)) {
          videoCalls[callId].receiverVideoProducer = await videoCalls[callId].receiverProducerTransport.produce({
            kind,
            rtpParameters,
          });

          videoCalls[callId].receiverVideoProducer.on('transportclose', () => {
            console.log('transport for this producer closed', callId);
            closeCall(callId);
          });

          // Send back to the client the Producer's id
          callback &&
            callback({
              id: videoCalls[callId].receiverVideoProducer.id,
            });
        } else {
          videoCalls[callId].initiatorVideoProducer = await videoCalls[callId].initiatorProducerTransport.produce({
            kind,
            rtpParameters,
          });

          videoCalls[callId].initiatorVideoProducer.on('transportclose', () => {
            console.log('transport for this producer closed', callId);
            closeCall(callId);
          });

          callback &&
            callback({
              id: videoCalls[callId].initiatorVideoProducer.id,
            });
        }
      } else if (kind === 'audio') {
        if (!isInitiator(callId, socket.id)) {
          videoCalls[callId].receiverAudioProducer = await videoCalls[callId].receiverProducerTransport.produce({
            kind,
            rtpParameters,
          });

          videoCalls[callId].receiverAudioProducer.on('transportclose', () => {
            console.log('transport for this producer closed', callId);
            closeCall(callId);
          });

          // Send back to the client the Producer's id
          callback &&
            callback({
              id: videoCalls[callId].receiverAudioProducer.id,
            });
        } else {
          videoCalls[callId].initiatorAudioProducer = await videoCalls[callId].initiatorProducerTransport.produce({
            kind,
            rtpParameters,
          });

          videoCalls[callId].initiatorAudioProducer.on('transportclose', () => {
            console.log('transport for this producer closed', callId);
            closeCall(callId);
          });

          // Send back to the client the Producer's id
          callback &&
            callback({
              id: videoCalls[callId].initiatorAudioProducer.id,
            });
        }
      }

      const socketToEmit = isInitiator(callId, socket.id)
        ? videoCalls[callId].receiverSocket
        : videoCalls[callId].initiatorSocket;

      // callId - Id of the call
      // kind - producer type: audio/video
      socketToEmit?.emit('new-producer', { callId, kind });
    } catch (error) {
      console.error(`[transport-produce] | ERROR | callId: ${socketDetails[socket.id]} | error: ${error.message}`);
    }
  });

  /*
    - The client sends this event after successfully creating a createRecvTransport(AS CONSUMER)
    - The connection is made to the created consumerTransport
  */
  socket.on('transport-recv-connect', async ({ dtlsParameters }) => {
    try {
      const callId = socketDetails[socket.id];
      console.log(`[transport-recv-connect] socket ${socket.id} | callId ${callId}`);
      if (typeof dtlsParameters === 'string') dtlsParameters = JSON.parse(dtlsParameters);
      // await videoCalls[callId].consumerTransport.connect({ dtlsParameters });
      if (!isInitiator(callId, socket.id)) {
        await videoCalls[callId].receiverConsumerTransport.connect({ dtlsParameters });
      } else if (isInitiator(callId, socket.id)) {
        await videoCalls[callId].initiatorConsumerTransport.connect({ dtlsParameters });
      }
    } catch (error) {
      console.error(`[transport-recv-connect] | ERROR | callId: ${socketDetails[socket.id]} | error: ${error.message}`);
    }
  });

  /*
    - The customer consumes after successfully connecting to consumerTransport
    - The previous step was 'transport-recv-connect', and before that 'createWebRtcTransport'
    - This event is only sent by the consumer
    - The parameters that the consumer consumes are returned
      - The consumer does consumerTransport.consume(params)
  */
  socket.on('consume', async ({ rtpCapabilities }, callback) => {
    const callId = socketDetails[socket.id];
    const socketId = socket.id;

    console.log(`[consume] socket ${socketId} | callId: ${callId}`);

    if (typeof rtpCapabilities === 'string') rtpCapabilities = JSON.parse(rtpCapabilities);

    callback({
      videoParams: await consumeVideo({ callId, socketId, rtpCapabilities }),
      audioParams: await consumeAudio({ callId, socketId, rtpCapabilities }),
    });
  });

  /*
    - Event sent by the consumer after consuming to resume the pause
    - When consuming on consumerTransport, it is initially done with paused: true, here we will resume
    - For the initiator we resume the initiatorConsumerAUDIO/VIDEO and for receiver the receiverConsumerAUDIO/VIDEO
  */
  socket.on('consumer-resume', () => {
    try {
      const callId = socketDetails[socket.id];
      const isInitiatorValue = isInitiator(callId, socket.id);
      console.log(`[consumer-resume] callId: ${callId} | isInitiator: ${isInitiatorValue}`);

      const consumerVideo = isInitiatorValue
        ? videoCalls[callId].initiatorConsumerVideo
        : videoCalls[callId].receiverConsumerVideo;

      const consumerAudio = isInitiatorValue
        ? videoCalls[callId].initiatorConsumerAudio
        : videoCalls[callId].receiverConsumerAudio;

      consumerVideo?.resume();
      consumerAudio?.resume();
    } catch (error) {
      console.error(
        `[consumer-resume] | ERROR | callId: ${socketDetails[socket.id]} | isInitiator: ${isInitiator} | error: ${
          error.message
        }`
      );
    }
  });

  socket.on('close-producer', ({ callId, kind }) => {
    try {
      if (isInitiator(callId, socket.id)) {
        console.log(`[close-producer] initiator --EMIT--> receiver | callId: ${callId} | kind: ${kind}`);
        videoCalls[callId].receiverSocket.emit('close-producer', { callId, kind });
      } else {
        console.log(`[close-producer] receiver --EMIT--> initiator | callId: ${callId} | kind: ${kind}`);
        videoCalls[callId].initiatorSocket.emit('close-producer', { callId, kind });
      }
    } catch (error) {
      console.error(`[close-producer] | ERROR | callId: ${socketDetails[socket.id]} | error: ${error.message}`);
    }
  });
});

const canConsume = ({ callId, producerId, rtpCapabilities }) => {
  return !!videoCalls[callId].router.canConsume({
    producerId,
    rtpCapabilities,
  });
};

const consumeVideo = async ({ callId, socketId, rtpCapabilities }) => {
  // Handlers for consumer transport https://mediasoup.org/documentation/v3/mediasoup/api/#consumer-on-transportclose
  if (isInitiator(callId, socketId) && videoCalls[callId].receiverVideoProducer) {
    const producerId = videoCalls[callId].receiverVideoProducer.id;
    if (!canConsume({ callId, producerId, rtpCapabilities })) return null;

    videoCalls[callId].initiatorConsumerVideo = await videoCalls[callId].initiatorConsumerTransport.consume({
      producerId,
      rtpCapabilities,
      paused: true,
    });

    return {
      id: videoCalls[callId].initiatorConsumerVideo.id,
      producerId,
      kind: 'video',
      rtpParameters: videoCalls[callId].initiatorConsumerVideo.rtpParameters,
    };
  } else if (videoCalls[callId].initiatorVideoProducer) {
    const producerId = videoCalls[callId].initiatorVideoProducer.id;
    if (!canConsume({ callId, producerId, rtpCapabilities })) return null;

    videoCalls[callId].receiverConsumerVideo = await videoCalls[callId].receiverConsumerTransport.consume({
      producerId,
      rtpCapabilities,
      paused: true,
    });

    return {
      id: videoCalls[callId].receiverConsumerVideo.id,
      producerId,
      kind: 'video',
      rtpParameters: videoCalls[callId].receiverConsumerVideo.rtpParameters,
    };
  } else {
    return null;
  }
};

const consumeAudio = async ({ callId, socketId, rtpCapabilities }) => {
  try {
    // Handlers for consumer transport https://mediasoup.org/documentation/v3/mediasoup/api/#consumer-on-transportclose
    if (isInitiator(callId, socketId) && videoCalls[callId].receiverAudioProducer) {
      const producerId = videoCalls[callId].receiverAudioProducer.id;
      if (!canConsume({ callId, producerId, rtpCapabilities })) return null;

      videoCalls[callId].initiatorConsumerAudio = await videoCalls[callId].initiatorConsumerTransport.consume({
        producerId,
        rtpCapabilities,
        paused: true,
      });

      return {
        id: videoCalls[callId].initiatorConsumerAudio.id,
        producerId,
        kind: 'audio',
        rtpParameters: videoCalls[callId].initiatorConsumerAudio.rtpParameters,
      };
    } else if (videoCalls[callId].initiatorAudioProducer) {
      const producerId = videoCalls[callId].initiatorAudioProducer.id;
      if (!canConsume({ callId, producerId, rtpCapabilities })) return null;

      videoCalls[callId].receiverConsumerAudio = await videoCalls[callId].receiverConsumerTransport.consume({
        producerId,
        rtpCapabilities,
        paused: true,
      });

      return {
        id: videoCalls[callId].receiverConsumerAudio.id,
        producerId,
        kind: 'audio',
        rtpParameters: videoCalls[callId].receiverConsumerAudio.rtpParameters,
      };
    } else {
      return null;
    }
  } catch (error) {
    console.error(`[consumeAudio] | ERROR | error: ${error}`);
  }
};

const isInitiator = (callId, socketId) => {
  return videoCalls[callId]?.initiatorSocket?.id === socketId;
};

/*
  - Called from at event 'createWebRtcTransport' and assigned to the consumer or producer transport
  - It will return parameters, these are required for the client to create the RecvTransport 
    from the client.
    - If the client is producer(sender: true) then it will use parameters for device.createSendTransport(params)
    - If the client is a consumer(sender: false) then it will use parameters for device.createRecvTransport(params)
*/
const createWebRtcTransportLayer = async (callId, callback) => {
  try {
    console.log(`[createWebRtcTransportLayer] callId: ${callId}`);
    // https://mediasoup.org/documentation/v3/mediasoup/api/#WebRtcTransportOptions
    const webRtcTransport_options = {
      listenIps: [
        {
          ip: process.env.IP, // Listening IPv4 or IPv6.
          announcedIp: process.env.ANNOUNCED_IP, // Announced IPv4 or IPv6 (useful when running mediasoup behind NAT with private IP).
        },
      ],
      enableUdp: true,
      enableTcp: true,
      preferUdp: true,
    };

    // https://mediasoup.org/documentation/v3/mediasoup/api/#router-createWebRtcTransport
    let transport = await videoCalls[callId].router.createWebRtcTransport(webRtcTransport_options);

    // Handler for when DTLS(Datagram Transport Layer Security) changes
    transport.on('dtlsstatechange', (dtlsState) => {
      console.log(`transport | dtlsstatechange | calldId ${callId} | dtlsState ${dtlsState}`);
      if (dtlsState === 'closed') {
        transport.close();
      }
    });

    // Handler if the transport layer has closed (for various reasons)
    transport.on('close', () => {
      console.log(`transport | closed | calldId ${callId}`);
    });

    const params = {
      id: transport.id,
      iceParameters: transport.iceParameters,
      iceCandidates: transport.iceCandidates,
      dtlsParameters: transport.dtlsParameters,
    };

    // Send back to the client the params
    callback({ params });

    // Set transport to producerTransport or consumerTransport
    return transport;
  } catch (error) {
    console.error(
      `[createWebRtcTransportLayer] | ERROR | callId: ${socketDetails[socket.id]} | error: ${error.message}`
    );
    callback({ params: { error } });
  }
};