chao faf16d266e update | 3 months ago | |
---|---|---|
.. | ||
scripts | 3 months ago | |
src | 3 months ago | |
.gitignore | 3 months ago | |
.npmignore | 3 months ago | |
CHANGELOG.json | 3 months ago | |
CHANGELOG.md | 3 months ago | |
LICENSE | 3 months ago | |
README.md | 3 months ago | |
jest.config.js | 3 months ago | |
package.json | 3 months ago | |
scrcpy.LICENSE | 3 months ago | |
tsconfig.build.json | 3 months ago | |
tsconfig.json | 3 months ago | |
tsconfig.test.json | 3 months ago |
TypeScript implementation of Scrcpy client.
It's compatible with the official Scrcpy server binaries.
WARNING: The public API is UNSTABLE. If you have any questions, please open an issue.
@yume-chan/adb
Although Scrcpy doesn't install any App on the device, it does have a server binary executable to be run on the device.
With the official Scrcpy client and server, the client uses ADB to transfer the server binary file to the device, run it, and communicate with it via ADB tunnel (the bootstrap process).
This package provides types that can serialize and deserialize Scrcpy protocol messages, but it generally requires you to do the bootstrapping and provide the data stream to the server.
If you are also using @yume-chan/adb
, this package has a helper class that can complete the bootstrap process using it, doing what the official client does.
NOTE: @yume-chan/adb
is a peer dependency, you need to install it yourself. Types that named begin with Adb
requires @yume-chan/adb
, and types that named begin with Scrcpy
doesn't.
This package doesn't include the server binary. It's compatible with many versions of the official server binary, but may not work with future versions due to protocol changes.
You can download the server binary from official releases (https://github.com/Genymobile/scrcpy/releases), or use the built-in fetch-scrcpy-server
script to automate the process.
The server binary is subject to Apache License 2.0.
fetch-scrcpy-server
This package also has a script that can download the server binary from official releases for you.
To use it, first you need to install the gh-release-fetch@3
NPM into your project, as it's a peer dependency.
Then you can invoke it in a terminal:
$ npx fetch-scrcpy-server <version>
For example:
$ npx fetch-scrcpy-server 1.24
It can also be added to the postinstall
script in your package.json
, so running npm install
will automatically invoke the script.
"scripts": {
"postinstall": "fetch-scrcpy-server 1.24",
},
The server binary will be named bin/scrcpy-server
in this package's installation directory (usually in node_modules
).
The server binary file needs to be embedded into your application, the exact method depends on the runtime.
To name a few:
const fs = require('fs');
const path: string = require.resolve('@yume-chan/scrcpy/bin/scrcpy-server'); // Or your own server binary path
const buffer: Buffer = fs.readFileSync(path);
import fs from 'node:fs/promises';
import { createRequire } from 'node:module';
const path: string = createRequire(import.meta.url).resolve('@yume-chan/scrcpy/bin/scrcpy-server'); // Or your own server binary path
const buffer: Buffer = await fs.readFile(path);
Currently, ES Module doesn't have a resolve
function like require.resolve
in CommonJS, so createRequire
is used to create a CommonJS resolver.
import.meta.resolve
(https://github.com/whatwg/html/pull/5572) is a proposal that fills this gap. Node.js already has experimental support for it behind a flag. See https://nodejs.org/api/esm.html#importmetaresolvespecifier-parent for more information.
const path: string = import.meta.resolve('@yume-chan/scrcpy/bin/scrcpy-server');
Requires installing and configuring file-loader (https://v4.webpack.js.org/loaders/file-loader/)
import SCRCPY_SERVER_URL from '@yume-chan/scrcpy/bin/scrcpy-server'; // Or your own server binary path
const buffer: ArrayBuffer = await fetch(SCRCPY_SERVER_URL).then(res => res.arrayBuffer());
Requires configuring Asset Modules (https://webpack.js.org/guides/asset-modules/)
const SCRCPY_SERVER_URL = new URL('@yume-chan/scrcpy/bin/scrcpy-server', import.meta.url); // Or your own server binary path
const buffer: ArrayBuffer = await fetch(SCRCPY_SERVER_URL).then(res => res.arrayBuffer());
The correct version number is required to launch the server, so fetch-scrcpy-server
also writes the version number to bin/version.js
.
import SCRCPY_SERVER_VERSION from '@yume-chan/scrcpy/bin/version.js';
console.log(SCRCPY_SERVER_VERSION); // "1.24"
Scrcpy server options change over time, and some of them are not backwards compatible. This package provides option types for each version (or range). Using wrong option version usually results in errors.
The latest one may continue to work for future server versions, but there is no guarantee.
Version | Type |
---|---|
1.16~1.17 | ScrcpyOptions1_16 |
1.18~1.20 | ScrcpyOptions1_18 |
1.21 | ScrcpyOptions1_21 |
1.22 | ScrcpyOptions1_22 |
1.23 | ScrcpyOptions1_23 |
1.24 | ScrcpyOptions1_24 |
When using AdbScrcpyClient
, there are AdbScrcpyOptions
containing @yume-chan/adb
related options:
Version | Type |
---|---|
1.16~1.21 | AdbScrcpyOptions1_16 |
1.22~1.24 | AdbScrcpyOptions1_22 |
@yume-chan/adb
@yume-chan/adb
is a TypeScript ADB implementation that can run on Web browser. It can be used to bootstrap the server on a device.
The Adb#sync()#write()
method can be used to push files to the device. Read more at @yume-chan/adb
's documentation (https://github.com/yume-chan/ya-webadb/tree/main/libraries/adb#readme).
This package also provides the AdbScrcpyClient.pushServer()
static method as a shortcut, plus it will automatically close the AdbSync
object on completion.
Example using write()
:
import { AdbScrcpyClient } from '@yume-chan/scrcpy';
const stream: WritableStream<Uint8Array> = AdbScrcpyClient.pushServer(adb);
const writer = stream.getWriter();
await writer.write(new Uint8Array(buffer));
await writer.close();
Example using pipeTo()
:
import { WrapReadableStream } from '@yume-chan/adb';
import { AdbScrcpyClient } from '@yume-chan/scrcpy';
await fetch(SCRCPY_SERVER_URL)
// `WrapReadableStream` is required because native `ReadableStream` (from `fetch`)
// doesn't support `pipeTo()` non-native `WritableStream`s
// (`@yume-chan/adb` is using `web-streams-polyfill`)
.then(response => new WrapReadableStream(response.body))
.then(stream => stream.pipeTo(AdbScrcpyClient.pushServer(adb)));
To start the server, use the AdbScrcpyClient.start()
method. It automatically sets up port forwarding, launches the server, and connects to it.
import { AdbScrcpyClient, AdbScrcpyOptions1_22, DEFAULT_SERVER_PATH, ScrcpyOptions1_24 } from '@yume-chan/scrcpy';
import SCRCPY_SERVER_VERSION from '@yume-chan/scrcpy/bin/version.js';
const client: AdbScrcpyClient = await AdbScrcpyClient.start(
adb,
DEFAULT_SERVER_PATH,
SCRCPY_SERVER_VERSION, // Or provide your own version number
new AdbScrcpyOptions1_22(ScrcpyOptions1_24({
// options
}))
);
const stdout: ReadableStream<string> = client.stdout;
const videoPacketStream: ReadableStream<ScrcpyVideoStreamPacket> = client.videoStream;
const controlMessageSerializer: ScrcpyControlMessageSerializer | undefined = client.controlMessageSerializer;
const deviceMessageStream: ReadableStream<ScrcpyDeviceMessage> | undefined = client.deviceMessageStream;
// to stop the server
client.close();
If you push, start and connect to the server yourself, you can still use this package to serialize/deserialize packets.
Requires a ReadableStream<Uint8Array>
that reads from the video socket, preserving packet boundaries.
NOTE: Because this package uses web-streams-polyfill
NPM package's Web Streams API implementation, the provided ReadableStream
must also be from web-streams-polyfill
(or another polyfill that doesn't check object prototype when piping).
import { ScrcpyOptions1_24, ScrcpyVideoStreamPacket } from '@yume-chan/scrcpy';
const videoStream: ReadableStream<Uint8Array>; // get the stream yourself
const options = new ScrcpyOptions1_24({
// use the same version and options
});
const videoPacketStream: ReadableStream<ScrcpyVideoStreamPacket> = videoStream.pipeThrough(options.createVideoStreamTransformer());
// Read from `videoPacketStream`
Requires a WritableStream<Uint8Array>
that writes to the control socket.
Control socket is optional if control is not enabled. Video socket and control socket can run completely separately.
import { ScrcpyControlMessageSerializer, ScrcpyOptions1_24 } from '@yume-chan/scrcpy';
const controlStream: ReadableWritablePair<Uint8Array, Uint8Array> | undefined // get the stream yourself
const options = new ScrcpyOptions1_24({
// use the same version and options
});
const controlMessageSerializer = new ScrcpyControlMessageSerializer(controlStream.writable, options);
// Call methods on `controlMessageSerializer`
controlMessageSerializer.injectText("Hello World!");
Requires a ReadableStream<Uint8Array>
that reads from the control socket.
NOTE: Because this package uses web-streams-polyfill
NPM package's Web Streams API implementation, the provided ReadableStream
must also be from web-streams-polyfill
(or another polyfill that doesn't check object prototype when piping).
import { ScrcpyDeviceMessageDeserializeStream, ScrcpyOptions1_24 } from '@yume-chan/scrcpy';
const controlStream: ReadableWritablePair<Uint8Array, Uint8Array> | undefined // get the stream yourself
const deviceMessageStream: ReadableStream<ScrcpyDeviceMessage> = controlStream.readable.pipeThrough(new ScrcpyDeviceMessageDeserializeStream());
In Web Streams API, ReadableStream
will block its upstream when too many chunks are kept not read. If multiple streams are from the same upstream source, block one stream means blocking all of them.
For Scrcpy, usually all streams are originated from the same ADB connection (either USB or TCP), so it's important to always read from all streams, even if you don't care about their data.
// when using `AdbScrcpyClient`
stdout
.pipeTo(
new WritableStream<string>({
write: (line) => {
// Handle or ignore the stdout line
},
}),
)
.catch(() => {})
.then(() => {
// Handle server exit
});
videoPacketStream
.pipeTo(new WritableStream<ScrcpyVideoStreamPacket>({
write: (packet) => {
// Handle or ignore the video packet
},
}))
.catch(() => {});
deviceMessageStream
.pipeTo(new WritableStream<ScrcpyDeviceMessage>({
write: (message) => {
// Handle or ignore the device message
},
}))
.catch(() => {});
The data from videoPacketStream
has two types: configuration
and frame
. Some fields may not be populated depending on the server version and options.
export interface ScrcpyVideoStreamConfigurationPacket {
type: 'configuration';
data: H264Configuration;
}
export interface ScrcpyVideoStreamFramePacket {
type: 'frame';
keyframe?: boolean | undefined;
pts?: bigint | undefined;
data: Uint8Array;
}
When sendFrameMeta: false
is set, videoPacketStream
only contains frame
packets, and only the data
field in it is available. It's commonly used when feeding into decoders like FFmpeg that can parse the H.264 stream itself, or saving to disk directly.
Otherwise, both configuration
and frame
packets are available.
configuration
packets contain the parsed SPS data, and can be used to initialize a video decoder.pts
(and keyframe
field from server version 1.23) fields in frame
packets are available to help decode the video.@yume-chan/scrcpy-decoder-tinyh264
and @yume-chan/scrcpy-decoder-webcodecs
can be used to decode and render the video stream in Browser environments. Refer to their README files for compatibility and usage information.