@dxos/plugin-thread 0.8.4-main.e8ec1fe → 0.8.4-main.effb148878

package/src/calls/media-manager.ts

@@ -1,378 +0,0 @@
-//
-// Copyright 2025 DXOS.org
-//
-
-import { DeferredTask, Event, sleep, synchronized } from '@dxos/async';
-import { type Context, Resource, cancelWithContext } from '@dxos/context';
-import { invariant } from '@dxos/invariant';
-import { log } from '@dxos/log';
-
-import { SpeakingMonitor } from './speaking-monitor';
-import { type EncodedTrackName, TrackNameCodec, type TrackObject } from './types';
-import {
-  type CallsServiceConfig,
-  CallsServicePeer,
-  createBlackCanvasStreamTrack,
-  createInaudibleAudioStreamTrack,
-  getScreenshare,
-  getUserMediaTrack,
-} from './util';
-
-export type MediaState = {
-  audioDeviceId?: string;
-  audioEnabled?: boolean;
-  audioTrack?: MediaStreamTrack;
-
-  videoDeviceId?: string;
-  videoEnabled?: boolean;
-  videoTrack?: MediaStreamTrack;
-  videoStream?: MediaStream;
-
-  screenshareEnabled?: boolean;
-  screenshareTrack?: MediaStreamTrack;
-  screenshareVideoStream?: MediaStream;
-
-  pushedVideoTrack?: TrackObject;
-  pushedAudioTrack?: TrackObject;
-  pushedScreenshareTrack?: TrackObject;
-  peer?: CallsServicePeer;
-
-  pulledAudioTracks: Record<EncodedTrackName, { track: MediaStreamTrack; ctx: Context }>;
-  /**
-   * Includes video feeds and screenshare feeds of other participants.
-   */
-  pulledVideoStreams: Record<EncodedTrackName, { stream: MediaStream; ctx: Context }>;
-};
-
-// TOOD(burdon): Hard coded.
-const VIDEO_WIDTH = 1280;
-const VIDEO_HEIGHT = 720;
-const MAX_WEB_CAM_FRAMERATE = 24;
-const MAX_WEB_CAM_BITRATE = 120_0000;
-const RETRY_INTERVAL = 100;
-
-export type MediaManagerParams = {
-  serviceConfig: CallsServiceConfig;
-};
-
-const USE_INAUDIBLE_AUDIO = true;
-
-export class MediaManager extends Resource {
-  public readonly stateUpdated = new Event<MediaState>();
-
-  private readonly _state: MediaState = {
-    pulledVideoStreams: {},
-    pulledAudioTracks: {},
-  };
-
-  private _speakingMonitor?: SpeakingMonitor = undefined;
-  private _trackToReconcile: EncodedTrackName[] = [];
-  private _blackCanvasStreamTrack?: MediaStreamTrack = undefined;
-  private _inaudibleAudioStreamTrack?: MediaStreamTrack = undefined;
-  private _pushTracksTask?: DeferredTask = undefined;
-  private _pullTracksTask?: DeferredTask = undefined;
-
-  get isSpeaking() {
-    return this._speakingMonitor?.isSpeaking;
-  }
-
-  /**
-   * @internal
-   */
-  _getState(): MediaState {
-    return this._state;
-  }
-
-  protected override async _open(): Promise<void> {
-    this._blackCanvasStreamTrack = await createBlackCanvasStreamTrack({
-      ctx: this._ctx,
-      width: VIDEO_WIDTH,
-      height: VIDEO_HEIGHT,
-    });
-
-    this._state.videoTrack = this._blackCanvasStreamTrack;
-    this._state.videoStream = new MediaStream();
-    this._state.videoStream.addTrack(this._state.videoTrack);
-
-    if (USE_INAUDIBLE_AUDIO) {
-      this._inaudibleAudioStreamTrack = await createInaudibleAudioStreamTrack({ ctx: this._ctx });
-      this._state.audioTrack = this._inaudibleAudioStreamTrack;
-    }
-
-    this._pushTracksTask = new DeferredTask(this._ctx, async () => {
-      await this._pushTracks();
-    });
-    this._pullTracksTask = new DeferredTask(this._ctx, async () => {
-      await this._reconcilePulledMedia();
-    });
-  }
-
-  protected override async _close(): Promise<void> {
-    void this._speakingMonitor?.close();
-    this._state.videoTrack && this._state.videoStream?.removeTrack(this._state.videoTrack);
-    this._state.audioTrack?.stop();
-    this._state.videoTrack?.stop();
-    this._state.screenshareTrack?.stop();
-    this._pushTracksTask = undefined;
-    this._pullTracksTask = undefined;
-  }
-
-  @synchronized
-  async join(serviceConfig: CallsServiceConfig): Promise<void> {
-    this._state.peer = new CallsServicePeer(serviceConfig);
-    await this._state.peer!.open();
-    this._pushTracksTask!.schedule();
-  }
-
-  @synchronized
-  async leave(): Promise<void> {
-    await Promise.all(Object.values(this._state.pulledAudioTracks).map(({ ctx }) => ctx.dispose()));
-    await Promise.all(Object.values(this._state.pulledVideoStreams).map(({ ctx }) => ctx.dispose()));
-    this._trackToReconcile = [];
-    await this._state.peer?.close();
-    this._state.peer = undefined;
-    this._state.pushedVideoTrack = undefined;
-    this._state.pushedAudioTrack = undefined;
-    this._state.pushedScreenshareTrack = undefined;
-    this._state.pulledAudioTracks = {};
-    this._state.pulledVideoStreams = {};
-  }
-
-  async turnVideoOn(): Promise<void> {
-    this._state.videoStream!.removeTrack(this._state.videoTrack!);
-    this._state.videoTrack = await getUserMediaTrack('videoinput', { width: VIDEO_WIDTH, height: VIDEO_HEIGHT });
-    this._state.videoStream!.addTrack(this._state.videoTrack);
-
-    this._state.videoEnabled = true;
-    this.stateUpdated.emit(this._state);
-    this._pushTracksTask!.schedule();
-  }
-
-  async turnVideoOff(): Promise<void> {
-    if (this._state.videoTrack !== this._blackCanvasStreamTrack) {
-      this._state.videoStream!.removeTrack(this._state.videoTrack!);
-      this._state.videoTrack?.stop();
-      this._state.videoTrack = this._blackCanvasStreamTrack;
-      this._state.videoStream!.addTrack(this._state.videoTrack!);
-    }
-
-    this._state.videoEnabled = false;
-    this.stateUpdated.emit(this._state);
-    this._pushTracksTask!.schedule();
-  }
-
-  // TODO(mykola): Change to `setAudioEnabled(enabled: boolean)`.
-  async turnAudioOn(): Promise<void> {
-    void this._speakingMonitor?.close();
-    this._state.audioEnabled = true;
-    this._state.audioTrack = await getUserMediaTrack('audioinput', {
-      echoCancellation: true,
-      noiseSuppression: true,
-      autoGainControl: true,
-      sampleRate: 16_000,
-    });
-    this.stateUpdated.emit(this._state);
-    this._pushTracksTask!.schedule();
-    this._speakingMonitor = new SpeakingMonitor(this._state.audioTrack!);
-    this._speakingMonitor.speakingChanged.on(this._ctx, () => this.stateUpdated.emit(this._state));
-    void this._speakingMonitor.open();
-  }
-
-  async turnAudioOff(): Promise<void> {
-    void this._speakingMonitor?.close();
-    this._speakingMonitor = undefined;
-    this._state.audioEnabled = false;
-    if (USE_INAUDIBLE_AUDIO && this._state.audioTrack !== this._inaudibleAudioStreamTrack) {
-      this._state.audioTrack?.stop();
-      this._state.audioTrack = this._inaudibleAudioStreamTrack;
-    } else {
-      this._state.audioTrack?.stop();
-      this._state.audioTrack = undefined;
-    }
-    this.stateUpdated.emit(this._state);
-    this._pushTracksTask!.schedule();
-  }
-
-  async turnScreenshareOn(): Promise<void> {
-    const ms = await getScreenshare({ contentHint: 'text' });
-    this._state.screenshareVideoStream = ms;
-    this._state.screenshareTrack = ms.getVideoTracks()[0];
-    this._state.screenshareEnabled = true;
-    this.stateUpdated.emit(this._state);
-    this._pushTracksTask!.schedule();
-  }
-
-  async turnScreenshareOff(): Promise<void> {
-    this._state.screenshareEnabled = false;
-    this._state.screenshareTrack?.stop();
-    this._state.screenshareTrack = undefined;
-    this.stateUpdated.emit(this._state);
-    this._pushTracksTask!.schedule();
-  }
-
-  /**
-   * @internal
-   */
-  _schedulePullTracks(tracks?: EncodedTrackName[]): void {
-    if (
-      !tracks ||
-      (this._trackToReconcile.every((name) => tracks.includes(name)) && this._trackToReconcile.length === tracks.length)
-    ) {
-      return;
-    }
-
-    this._trackToReconcile = tracks;
-    this._pullTracksTask!.schedule();
-  }
-
-  private async _reconcilePulledMedia(): Promise<void> {
-    log('reconciling tracks');
-    // TODO(mykola): Currently cloudflare fails if you try to pull track immediately after pushing it.
-    // Add retry logic, remove sleep.
-    // Wait for cloudflare to process the track.
-    await cancelWithContext(this._ctx, sleep(800));
-    const trackNames = this._trackToReconcile;
-    const tracksToPull = trackNames.filter(
-      (name) => !this._state.pulledAudioTracks[name] && !this._state.pulledVideoStreams[name],
-    );
-
-    const audioTracksToClose = Object.entries(this._state.pulledAudioTracks).filter(
-      ([key]) => !trackNames.some((name) => name === key),
-    );
-
-    const videoStreamsToClose = Object.entries(this._state.pulledVideoStreams).filter(
-      ([key]) => !trackNames.some((name) => name === key),
-    );
-
-    log('reconciling tracks', {
-      trackNames,
-      tracksToPull,
-      videoStreamsToClose,
-      audioTracksToClose,
-      currentAudioTracks: Object.keys(this._state.pulledAudioTracks),
-      currentVideoStreams: Object.keys(this._state.pulledVideoStreams),
-    });
-
-    // Pull new tracks.
-    const pullResults = await Promise.all(tracksToPull.map((name) => this._pullTrack(name)));
-
-    // Close old tracks.
-    await Promise.all([...audioTracksToClose, ...videoStreamsToClose].map(([_, { ctx }]) => ctx.dispose()));
-
-    log('reconciled tracks', {
-      currentAudioTracks: Object.keys(this._state.pulledAudioTracks),
-      currentVideoStreams: Object.keys(this._state.pulledVideoStreams),
-    });
-
-    this.stateUpdated.emit(this._state);
-    if (pullResults.some(({ shouldRetry }) => shouldRetry)) {
-      await cancelWithContext(this._ctx, sleep(RETRY_INTERVAL));
-      log.info('retrying pull tracks', { tracksToPull });
-      invariant(this._pullTracksTask);
-      this._pullTracksTask.schedule();
-    }
-  }
-
-  private async _pushTracks(): Promise<void> {
-    if (!this._state.peer?.isOpen) {
-      return;
-    }
-
-    let updated = false;
-    const [pushVideoResult, pushAudioResult, pushScreenshareResult] = await Promise.all([
-      this._maybePushTrack(this._state.videoTrack, this._state.pushedVideoTrack, [
-        { maxFramerate: MAX_WEB_CAM_FRAMERATE, maxBitrate: MAX_WEB_CAM_BITRATE },
-      ]),
-      this._maybePushTrack(this._state.audioTrack, this._state.pushedAudioTrack, [{ networkPriority: 'high' }]),
-      this._maybePushTrack(this._state.screenshareTrack, this._state.pushedScreenshareTrack),
-    ]);
-
-    if (pushVideoResult.track !== this._state.pushedVideoTrack) {
-      this._state.pushedVideoTrack = pushVideoResult.track;
-      updated = true;
-    }
-    if (pushAudioResult.track !== this._state.pushedAudioTrack) {
-      this._state.pushedAudioTrack = pushAudioResult.track;
-      updated = true;
-    }
-    if (pushScreenshareResult.track !== this._state.pushedScreenshareTrack) {
-      this._state.pushedScreenshareTrack = pushScreenshareResult.track;
-      updated = true;
-    }
-
-    const shouldRetry = pushVideoResult.shouldRetry || pushAudioResult.shouldRetry || pushScreenshareResult.shouldRetry;
-    if (updated) {
-      this.stateUpdated.emit(this._state);
-    }
-
-    if (shouldRetry) {
-      await cancelWithContext(this._ctx, sleep(RETRY_INTERVAL));
-      log.info('retrying push tracks', { pushVideoResult, pushAudioResult, pushScreenshareResult });
-      invariant(this._pushTracksTask);
-      this._pushTracksTask.schedule();
-    }
-  }
-
-  private async _pullTrack(name: EncodedTrackName): Promise<{ shouldRetry: boolean }> {
-    const ctx = this._ctx.derive();
-    try {
-      const trackData = TrackNameCodec.decode(name);
-      // We need to set mid here to `undefined`, because mid is peer specific.
-      const track = await this._state.peer!.pullTrack({ trackData: { ...trackData, mid: undefined }, ctx });
-      if (track?.readyState === 'ended') {
-        throw new Error('Pulled track ended immediately');
-      }
-
-      switch (track?.kind) {
-        case 'audio': {
-          this._state.pulledAudioTracks[name] = { track, ctx };
-          ctx.onDispose(() => delete this._state.pulledAudioTracks[name]);
-          return { shouldRetry: false };
-        }
-        case 'video': {
-          const mediaStream = new MediaStream();
-          mediaStream.addTrack(track);
-          this._state.pulledVideoStreams[name] = { stream: mediaStream, ctx };
-          ctx.onDispose(() => {
-            mediaStream.removeTrack(track);
-            track.stop();
-            delete this._state.pulledVideoStreams[name];
-          });
-          return { shouldRetry: false };
-        }
-        default:
-          throw new Error(`Invalid track kind: ${track?.kind}`);
-      }
-    } catch (err) {
-      log.info('failed to pull track', { err, name });
-      void ctx.dispose();
-      return { shouldRetry: true };
-    }
-  }
-
-  private async _maybePushTrack(
-    track?: MediaStreamTrack,
-    previousTrack?: TrackObject,
-    encodings?: RTCRtpEncodingParameters[],
-  ): Promise<{ track?: TrackObject; shouldRetry?: boolean }> {
-    if (!track && !previousTrack) {
-      return { track: undefined };
-    }
-
-    const ctx = this._ctx.derive();
-    try {
-      return {
-        track: await this._state.peer!.pushTrack({
-          ctx,
-          track: track ?? null,
-          previousTrack,
-          encodings,
-        }),
-      };
-    } catch (err) {
-      log.info('failed to push track', { err, track, previousTrack, encodings });
-      void ctx.dispose();
-      return { shouldRetry: true };
-    }
-  }
-}

package/src/calls/speaking-monitor.ts

@@ -1,85 +0,0 @@
-//
-// Copyright 2025 DXOS.org
-//
-
-import { DeferredTask, Event, sleep } from '@dxos/async';
-import { Resource } from '@dxos/context';
-
-import { monitorAudioLevel } from './util';
-
-const NOT_SPEAKING_AFTER = 500; // [ms]
-const START_SPEAKING_THRESHOLD = 0.05;
-const CONTINUE_SPEAKING_THRESHOLD = 0.02;
-const MIN_EVENT_INTERVAL = 100; // [ms]
-
-export class SpeakingMonitor extends Resource {
-  /**
-   * Event that is emitted when the user starts or stops speaking.
-   * It is debounced to prevent rapid state changes.
-   */
-  public readonly speakingChanged = new Event<boolean>();
-
-  /**
-   * Real state of the user speaking.
-   */
-  private _isSpeaking = false;
-  private _lastEventTime = 0;
-  private _deferredUpdate?: DeferredTask | undefined;
-
-  constructor(private readonly _mediaStreamTrack: MediaStreamTrack) {
-    super();
-  }
-
-  get isSpeaking(): boolean {
-    return this._isSpeaking;
-  }
-
-  protected override async _open(): Promise<void> {
-    this._deferredUpdate = new DeferredTask(this._ctx, async () => this._emitSpeakingChanged());
-
-    let timeout: NodeJS.Timeout | undefined;
-    const cleanup = monitorAudioLevel({
-      mediaStreamTrack: this._mediaStreamTrack,
-      onMeasure: (vol) => {
-        // Once the user has been determined to be speaking, we want
-        // to lower the threshold because speech patterns don't always
-        // kick up above 0.05.
-        const audioLevelAboveThreshold =
-          vol > (this._isSpeaking ? CONTINUE_SPEAKING_THRESHOLD : START_SPEAKING_THRESHOLD);
-        if (audioLevelAboveThreshold) {
-          // User is still speaking, clear timeout & reset.
-          clearTimeout(timeout);
-          timeout = undefined;
-          if (!this._isSpeaking) {
-            this._isSpeaking = true;
-            this._deferredUpdate?.schedule();
-          }
-        } else if (timeout === undefined) {
-          // User is not speaking and timeout is not set.
-          timeout = setTimeout(() => {
-            this._isSpeaking = false;
-            timeout = undefined;
-            this._deferredUpdate?.schedule();
-          }, NOT_SPEAKING_AFTER);
-        }
-      },
-    });
-    this._ctx.onDispose(() => {
-      cleanup();
-      clearTimeout(timeout);
-    });
-  }
-
-  /**
-   * Emits the speaking changed event with debouncing based on timing.
-   */
-  private async _emitSpeakingChanged(): Promise<void> {
-    const timeSinceLastEvent = Date.now() - this._lastEventTime;
-    const timeToNextEvent = MIN_EVENT_INTERVAL - timeSinceLastEvent;
-    if (timeToNextEvent > 0) {
-      await sleep(timeToNextEvent);
-    }
-    this._lastEventTime = Date.now();
-    this.speakingChanged.emit(this._isSpeaking);
-  }
-}

package/src/calls/types.ts

@@ -1,74 +0,0 @@
-//
-// Copyright 2025 DXOS.org
-//
-
-import { invariant } from '@dxos/invariant';
-import { buf } from '@dxos/protocols/buf';
-import {
-  type ActivitySchema,
-  type UserState as UserStateProto,
-  UserStateSchema,
-} from '@dxos/protocols/buf/dxos/edge/calls_pb';
-
-/**
- * Endpoint to the calls service.
- */
-export const CALLS_URL = 'https://calls-service.dxos.workers.dev';
-
-export type UserState = buf.MessageInitShape<typeof UserStateSchema>;
-
-export type ActivityState = buf.MessageInitShape<typeof ActivitySchema>;
-
-export const codec = {
-  encode: (message: UserState): Uint8Array => {
-    return buf.toBinary(UserStateSchema, buf.create(UserStateSchema, message));
-  },
-  decode: (message: Uint8Array): UserStateProto => {
-    return buf.fromBinary(UserStateSchema, message);
-  },
-};
-
-export type TrackObject = {
-  location?: 'local' | 'remote';
-  trackName?: string;
-  sessionId?: string;
-  mid?: string | null;
-};
-
-export type SessionDescription = {
-  type: 'offer' | 'answer';
-  sdp: string;
-};
-
-export interface ErrorResponse {
-  errorCode?: string;
-  errorDescription?: string;
-}
-
-export interface SessionResponse extends ErrorResponse {
-  sessionId: string;
-  sessionDescription: SessionDescription;
-}
-
-export interface TracksResponse extends ErrorResponse {
-  sessionDescription: SessionDescription;
-  requiresImmediateRenegotiation: boolean;
-  tracks?: (TrackObject & ErrorResponse)[];
-}
-
-export interface RenegotiationResponse extends ErrorResponse {}
-
-export type EncodedTrackName = string & { __brand: 'EncodedTrackName' };
-
-export const TrackNameCodec = {
-  encode: (trackData: TrackObject): EncodedTrackName => {
-    invariant(trackData.sessionId);
-    invariant(trackData.trackName);
-    invariant(trackData.mid);
-    return (trackData.sessionId + '/' + trackData.trackName + '/' + trackData.mid) as EncodedTrackName;
-  },
-  decode: (name: EncodedTrackName): TrackObject => {
-    const [sessionId, trackName, mid] = name.split('/');
-    return { sessionId, trackName, location: 'remote', mid };
-  },
-};

package/src/calls/util/audio.ts

@@ -1,58 +0,0 @@
-//
-// Copyright 2025 DXOS.org
-//
-
-export const monitorAudioLevel = ({
-  mediaStreamTrack,
-  intervalMs = 1000 / 10, // 12 fps.
-  onMeasure,
-}: {
-  mediaStreamTrack: MediaStreamTrack;
-  intervalMs?: number;
-  onMeasure: (value: number) => void;
-}) => {
-  let timeout = -1;
-  let interval = -1;
-  const audioContext = new AudioContext();
-  const stream = new MediaStream();
-  stream.addTrack(mediaStreamTrack);
-  const mediaStreamAudioSourceNode = audioContext.createMediaStreamSource(stream);
-  const analyserNode = audioContext.createAnalyser();
-  mediaStreamAudioSourceNode.connect(analyserNode);
-  // Since we just need a rough approximation and will be measuring
-  // frequently, we want to drop this down from the default of 2048.
-  analyserNode.fftSize = 32;
-
-  const pcmData = new Float32Array(analyserNode.fftSize);
-  let peak = 0;
-
-  interval = window.setInterval(() => {
-    onMeasure(peak);
-    peak = 0;
-  }, intervalMs);
-
-  const tick = () => {
-    timeout = window.setTimeout(() => {
-      analyserNode.getFloatTimeDomainData(pcmData);
-      let sumSquares = 0.0;
-      for (const amplitude of pcmData) {
-        sumSquares += amplitude * amplitude;
-      }
-      const current = Math.sqrt(sumSquares / pcmData.length);
-      if (current > peak) {
-        peak = current;
-      }
-      tick();
-    }, intervalMs);
-  };
-  tick();
-
-  return () => {
-    mediaStreamAudioSourceNode.disconnect();
-    analyserNode.disconnect();
-    void audioContext.close();
-    clearInterval(interval);
-    clearTimeout(timeout);
-    stream.removeTrack(mediaStreamTrack);
-  };
-};