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,                                            router
 *  initiatorAudioProducer: Producer,
 *  initiatorVideoProducer: Producer,
 *  receiverVideoProducer: Producer,                           producerVideo
 *  receiverAudioProducer: Producer,                           producerAudio
 *  initiatorProducerTransport: Producer Transport,
 *  receiverProducerTransport: Producer Transport,             producerTransport
 *  initiatorConsumerVideo: Consumer,                          consumerVideo
 *  initiatorConsumerAudio: Consumer,                          consumerAudio
 *  initiatorConsumerTransport: Consumer Transport             consumerTransport
 *  initiatorSockerId
 *  receiverSocketId
 * }
 * 
 **/
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.log(`ERROR | createWorker | ${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];
    } else {
      console.log(`The call with id ${callId} has already been deleted`);
    }
  } catch (error) {
    console.log(`ERROR | closeCall | callid ${callId} | ${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]) {
          console.log('[createRoom] callId', callId);
          videoCalls[callId] = { router: await worker.createRouter({ mediaCodecs }) }
          console.log(`[createRoom] Router ID: ${videoCalls[callId].router.id}`);
          videoCalls[callId].receiverSocketId = socket.id
        } else {
          videoCalls[callId].initiatorSockerId = socket.id
        }
        socketDetails[socket.id] = callId;
        // rtpCapabilities is set for callback
        console.log('[getRtpCapabilities] callId', callId);
        callbackResponse = { 
          rtpCapabilities :videoCalls[callId].router.rtpCapabilities
        };
      } else {
        console.log(`[createRoom] missing callId ${callId}`);
      }
    } catch (error) {
      console.log(`ERROR | createRoom | callId ${callId} | ${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] 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.log(`ERROR | createWebRtcTransport | callId ${socketDetails[socket.id]} | sender ${sender} | ${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.id ${socket.id} | callId ${callId}`);
      if (!isInitiator(callId, socket.id)) {
        await videoCalls[callId].receiverProducerTransport.connect({ dtlsParameters });
      } else {
        await videoCalls[callId].initiatorProducerTransport.connect({ dtlsParameters });
      }
    } catch (error) {
      console.log(`ERROR | transport-connect | callId ${socketDetails[socket.id]} | ${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] kind: ${kind} | socket.id: ${socket.id} | callId: ${callId}`);
      if (kind === 'video') {
        if (!isInitiator(callId, socket.id)) {
          videoCalls[callId].receiverVideoProducer = await videoCalls[callId].receiverProducerTransport.produce({
            kind,
            rtpParameters,
          });

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

          videoCalls[callId].receiverVideoProducer.on('transportclose', () => {
            const callId = socketDetails[socket.id];
            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,
          });

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

          videoCalls[callId].initiatorVideoProducer.on('transportclose', () => {
            const callId = socketDetails[socket.id];
            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,
          });

          console.log(`[transport-produce] receiverAudioProducer Producer ID: ${videoCalls[callId].receiverAudioProducer.id} | kind: ${videoCalls[callId].receiverAudioProducer.kind}`);
      
          videoCalls[callId].receiverAudioProducer.on('transportclose', () => {
            const callId = socketDetails[socket.id];
            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,
          });

          console.log(`[transport-produce] initiatorAudioProducer Producer ID: ${videoCalls[callId].initiatorAudioProducer.id} | kind: ${videoCalls[callId].initiatorAudioProducer.kind}`);
      
          videoCalls[callId].initiatorAudioProducer.on('transportclose', () => {
            const callId = socketDetails[socket.id];
            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
          });
        }
      }
    } catch (error) {
      console.log(`ERROR | transport-produce | callId ${socketDetails[socket.id]} | ${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.id ${socket.id} | callId ${callId}`);
      // 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.log(`ERROR | transport-recv-connect | callId ${socketDetails[socket.id]} | ${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) => {
    try {
      console.log(`[consume] rtpCapabilities: ${JSON.stringify(rtpCapabilities)}`);

      const callId = socketDetails[socket.id];
      console.log('[consume] callId', callId);
      let canConsumeVideo, canConsumeAudio;
      if (isInitiator(callId, socket.id)) {
        canConsumeVideo = !!videoCalls[callId].receiverVideoProducer && !!videoCalls[callId].router.canConsume({
          producerId: videoCalls[callId].receiverVideoProducer.id,
          rtpCapabilities
        });
        canConsumeAudio = !!videoCalls[callId].receiverAudioProducer && !!videoCalls[callId].router.canConsume({
          producerId: videoCalls[callId].receiverAudioProducer.id,
          rtpCapabilities
        });

      } else {
        canConsumeVideo = !!videoCalls[callId].initiatorVideoProducer && !!videoCalls[callId].router.canConsume({
          producerId: videoCalls[callId].initiatorVideoProducer.id,
          rtpCapabilities
        });

        canConsumeAudio = !!videoCalls[callId].initiatorAudioProducer && !!videoCalls[callId].router.canConsume({
          producerId: videoCalls[callId].initiatorAudioProducer.id,
          rtpCapabilities
        });
      }
      console.log('[consume] canConsumeVideo', canConsumeVideo);
      console.log('[consume] canConsumeAudio', canConsumeAudio);

      if (canConsumeVideo && !canConsumeAudio) {
        const videoParams = await consumeVideo(callId, socket.id, rtpCapabilities)
        callback({ videoParams, audioParams: null });
      } else if (canConsumeVideo && canConsumeAudio) {
        const videoParams = await consumeVideo(callId, socket.id, rtpCapabilities)
        const audioParams = await consumeAudio(callId, socket.id, rtpCapabilities)
        callback({ videoParams, audioParams });
      } else if (!canConsumeVideo && canConsumeAudio) {
        const audioParams = await consumeAudio(callId, socket.id, rtpCapabilities)
        callback({ videoParams: null, audioParams });
      } else {
        console.log(`[consume] Can't consume | callId ${callId}`);
        callback(null);
      }
    } catch (error) {
      console.log(`ERROR | consume | callId ${socketDetails[socket.id]} | ${error.message}`)
      callback({ params: { error } });
    }
  });

  /*
    - 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 
  */
  socket.on('consumer-resume', async () => {
    try {
      const callId = socketDetails[socket.id];
      console.log(`[consumer-resume] callId ${callId}`)

      if (isInitiator(callId, socket.id)) {
        await videoCalls[callId].initiatorConsumerVideo.resume();
        await videoCalls[callId].initiatorConsumerAudio.resume();
      } else {
        await videoCalls[callId].receiverConsumerVideo.resume();
        await videoCalls[callId].receiverConsumerAudio.resume();
      }
      // await videoCalls[callId].consumerVideo.resume();
      // await videoCalls[callId].consumerAudio.resume();
    } catch (error) {
      console.log(`ERROR | consumer-resume | callId ${socketDetails[socket.id]} | ${error.message}`);
    }
  });
});

const consumeVideo = async (callId, socketId, rtpCapabilities) => {

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

    // https://mediasoup.org/documentation/v3/mediasoup/api/#consumer-on-transportclose
    videoCalls[callId].initiatorConsumerVideo.on('transportclose', () => {
      const callId = socketDetails[socket.id];
      console.log('transport close from consumer', callId);
      closeCall(callId);
    });

    // https://mediasoup.org/documentation/v3/mediasoup/api/#consumer-on-producerclose
    videoCalls[callId].initiatorConsumerVideo.on('producerclose', () => {
      const callId = socketDetails[socket.id];
      console.log('producer of consumer closed', callId);
      closeCall(callId);
    });

    return {
      id: videoCalls[callId].initiatorConsumerVideo.id,
      producerId: videoCalls[callId].receiverVideoProducer.id,
      kind: 'video',
      rtpParameters: videoCalls[callId].initiatorConsumerVideo.rtpParameters,
    }
  } else {
    videoCalls[callId].receiverConsumerVideo = await videoCalls[callId].receiverConsumerTransport.consume({
      producerId: videoCalls[callId].initiatorVideoProducer.id,
      rtpCapabilities,
      paused: true,
    });

    videoCalls[callId].receiverConsumerVideo.on('transportclose', () => {
      const callId = socketDetails[socket.id];
      console.log('transport close from consumer', callId);
      closeCall(callId);
    });

    videoCalls[callId].receiverConsumerVideo.on('producerclose', () => {
      const callId = socketDetails[socket.id];
      console.log('producer of consumer closed', callId);
      closeCall(callId);
    });

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

const consumeAudio = async (callId, socketId, rtpCapabilities) => {
  if(isInitiator(callId, socketId)) {
    videoCalls[callId].initiatorConsumerAudio = await videoCalls[callId].initiatorConsumerTransport.consume({
      producerId: videoCalls[callId].receiverAudioProducer.id,
      rtpCapabilities,
      paused: true,
    });

    // https://mediasoup.org/documentation/v3/mediasoup/api/#consumer-on-transportclose
    videoCalls[callId].initiatorConsumerAudio.on('transportclose', () => {
      const callId = socketDetails[socket.id];
      console.log('transport close from consumer', callId);
      closeCall(callId);
    });

    // https://mediasoup.org/documentation/v3/mediasoup/api/#consumer-on-producerclose
    videoCalls[callId].initiatorConsumerAudio.on('producerclose', () => {
      const callId = socketDetails[socket.id];
      console.log('producer of consumer closed', callId);
      closeCall(callId);
    });

    return {
      id: videoCalls[callId].initiatorConsumerAudio.id,
      producerId: videoCalls[callId].receiverAudioProducer.id,
      kind: 'audio',
      rtpParameters: videoCalls[callId].initiatorConsumerAudio.rtpParameters,
    }
  } else {
    videoCalls[callId].receiverConsumerAudio = await videoCalls[callId].receiverConsumerTransport.consume({
      producerId: videoCalls[callId].initiatorAudioProducer.id,
      rtpCapabilities,
      paused: true,
    });

    videoCalls[callId].receiverConsumerAudio.on('transportclose', () => {
      const callId = socketDetails[socket.id];
      console.log('transport close from consumer', callId);
      closeCall(callId);
    });

    videoCalls[callId].receiverConsumerAudio.on('producerclose', () => {
      const callId = socketDetails[socket.id];
      console.log('producer of consumer closed', callId);
      closeCall(callId);
    });

    return {
      id: videoCalls[callId].receiverConsumerAudio.id,
      producerId: videoCalls[callId].initiatorAudioProducer.id,
      kind: 'audio',
      rtpParameters: videoCalls[callId].receiverConsumerAudio.rtpParameters,
    }
  }
}

const isInitiator = (callId, socketId) => {
  return (videoCalls[callId].initiatorSockerId === 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)
    console.log(`callId: ${callId} | transport id: ${transport.id}`)

    // 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,
    };

    console.log('[createWebRtcTransportLayer] callback params', params);
    // Send back to the client the params
    callback({ params });

    // Set transport to producerTransport or consumerTransport
    return transport;

  } catch (error) {
    console.log(`ERROR | createWebRtcTransportLayer | callId ${socketDetails[socket.id]} | ${error.message}`);
    callback({ params: { error } });
  }
}