versatica / mediasoup-client-aiortc

mediasoup-client handler for aiortc Python library
https://mediasoup.org
ISC License
56 stars 30 forks source link
client-library nodejs webrtc

mediasoup-client-aiortc

mediasoup-client handler for aiortc Python library. Suitable for building Node.js applications that connect to a mediasoup server using WebRTC and exchange real audio, video and DataChannel messages with it in both directions.

Requirements

Installation

Install mediasoup-client-aiortc within your Node.js application:

npm install mediasoup-client-aiortc

The "postinstall" script in package.json will install the Python libraries (including aiortc). You can override the path to python executable by setting the PYTHON environment variable:

PYTHON=/home/me/bin/python3.13 npm install mediasoup-client-aiortc

Same once you run your Node.js application. mediasoup-client-aiortc will spawn Python processes and communicate with them via UnixSocket. You can override the python executable path by setting the PYTHON environment variable:

PYTHON=/home/me/bin/python3.13 node my_app.js

API

// ES6 style.
import {
    createWorker,
    Worker,
    WorkerSettings,
    WorkerLogLevel,
    AiortcMediaStream,
    AiortcMediaStreamConstraints,
    AiortcMediaTrackConstraints,
} from 'mediasoup-client-aiortc';

// CommonJS style.
const {
    createWorker,
    Worker,
    WorkerSettings,
    WorkerLogLevel,
    AiortcMediaStream,
    AiortcMediaStreamConstraints,
    AiortcMediaTrackConstraints,
} = require('mediasoup-client-aiortc');

async createWorker(settings: WorkerSettings) function

Creates a mediasoup-client-aiortc Worker instance. Each Worker spawns and manages a Python subprocess.

@async

@returns Worker

const worker = await createWorker({
    logLevel: 'warn',
});

Worker class

The Worker class. It represents a separate Python subprocess that can provide the Node.js application with audio/video tracks and mediasoup-client handlers.

worker.pid getter

The Python subprocess PID.

@type String, read only

worker.closed getter

Whether the subprocess is closed.

worker.died getter

Whether the subprocess died unexpectedly (probably a bug somewhere).

worker.subprocessClosed getter

Whether the subprocessed is closed. It becomes true once the worker subprocess is completely closed and 'subprocessclose' event fires.

@type Boolean, read only

worker.close() method

Closes the subprocess and all its open resources (such as audio/video tracks and mediasoup-client handlers).

async worker.getUserMedia(constraints: AiortcMediaStreamConstraints) method

Mimics the navigator.getUserMedia() API. It creates an AiortcMediaStream instance containing audio and/or video tracks. Those tracks can point to different sources such as device microphone, webcam, multimedia files or HTTP streams.

@async

@returns AiortcMediaStream

const stream = await getUserMedia({
    audio: true,
    video: {
        source: 'file',
        file: 'file:///home/foo/media/foo.mp4',
    },
});

const audioTrack = stream.getAudioTracks()[0];
const videoTrack = stream.getVideoTracks()[0];

async worker.createHandlerFactory() method

Creates a mediasoup-client handler factory, suitable for the handlerFactory argument when instantiating a mediasoup-client Device.

@async

@returns HandlerFactory

const device = new mediasoupClient.Device({
    handlerFactory: worker.createHandlerFactory(),
});

Note that all Python resources (such as audio/video) used within the Device must be obtained from the same mediasoup-client-aiortc Worker instance.

worker.on("died", fn(error: Error) event

Emitted if the subprocess abruptly dies. This should not happen. If it happens there is a bug in the Python component.

worker.on("subprocessclose", fn()) event

Emitted when the subprocess has closed completely. This event is emitted asynchronously once worker.close() has been called (or after 'died' event in case the worker subprocess abnormally died).

Await for this event if you can to be sure that no Node handler is still open/running after you close a worker.

WorkerSettings type

type WorkerSettings = {
    /**
     * Logging level for logs generated by the Python subprocess.
     */
    logLevel?: WorkerLogLevel; // If unset it defaults to "error".
};

WorkerLogLevel type

type WorkerLogLevel = 'debug' | 'warn' | 'error' | 'none';

Logs generated by both, Node.js and Python components of this module, are printed using the mediasoup-client debugging system with "mediasoup-client-aiortc" prefix/namespace.

AiortcMediaStream class

A custom implementation of the W3C MediaStream class. An instance of AiortcMediaStream is generated by calling worker.getUserMedia().

Audio and video tracks within an AiortcMediaStream are instances of FakeMediaStreamTrack and reference "native" MediaStreamTracks in the Python subprocess (handled by aiortc library).

AiortcMediaStreamConstraints type

The argument given to worker.getUserMedia().

type AiortcMediaStreamConstraints = {
    audio?: AiortcMediaTrackConstraints | boolean;
    video?: AiortcMediaTrackConstraints | boolean;
};

Setting audio or video to true equals to {source: "device"} (so default microphone or webcam will be used to obtain the track or tracks).

AiortcMediaTrackConstraints type

type AiortcMediaTrackConstraints = {
    source: 'device' | 'file' | 'url';
    device?: string;
    file?: string;
    url?: string;
    format?: string;
    options?: object;
    timeout?: number;
    loop?: boolean;
    decode?: boolean;
};

source

Determines which source aiortc will use to generate the audio or video track. These are the possible values:

device

If source is "device" and this field is given, it specifies the device ID of the microphone or webcam to use. If unset, the default one in the system will be used.

file

Mandatory if source is "file". Must be the absolute path to a multimedia file.

url

Mandatory if source is "url". Must be the URL of an HTTP stream.

format

Specifies the device format used by ffmpeg.

options

Specifies the device options used by ffmpeg.

timeout, loop and decode

See documentation in aiortc site (decode option is not documented but you can figure it out by reading usage examples).

Other considerations

DataChannel

mediasoup-client-aiortc supports sending/receiving string and binary DataChannel messages. However, due to the lack of Blob support in Node.js, dataChannel.binaryType is always "arraybuffer" so received binary messages are always ArrayBuffer instances.

When sending, dataChannel.send() (and hence dataProducer.send()) allows passing a string, a Buffer instance or an ArrayBuffer instance.

Development

Lint

npm run lint

Test

npm run test

Check release

npm run release:check

Make Python log to stdout/stderr while running tests

PYTHON_LOG_TO_STDOUT=true npm run test

Caveats

See the list of open issues.

Authors

License

ISC