From 7a16182f5bf09f7fdb834a078201672ee0b7d29d Mon Sep 17 00:00:00 2001
From: jevb <jevb@telenet.be>
Date: Sun, 22 Mar 2026 15:34:26 +0100
Subject: [PATCH] fix: restore audio track attachment for remote playback

The refactored LiveKit session dropped track.attach() for remote audio,
so no <audio> element was created and remote participants were silent.
Also refactors noise suppression to use LiveKit TrackProcessor API,
adds input volume gain node bypass at 100%, and exposes __lkDebug()
on window for DevTools diagnostics.
---
 Client/tauri-client/src/lib/livekitSession.ts | 544 +++++++-----------
 .../tauri-client/src/lib/noise-suppression.ts | 390 ++++++-------
 Client/tauri-client/src/lib/types.ts          |   1 -
 README.md                                     |  37 +-
 docs/brain/06-Specs/CHATSERVER.md             | 134 +++--
 docs/brain/06-Specs/PROTOCOL.md               |  29 +-
 6 files changed, 488 insertions(+), 647 deletions(-)

diff --git a/Client/tauri-client/src/lib/livekitSession.ts b/Client/tauri-client/src/lib/livekitSession.ts
index 73757b11..524c8863 100644
--- a/Client/tauri-client/src/lib/livekitSession.ts
+++ b/Client/tauri-client/src/lib/livekitSession.ts
@@ -6,6 +6,7 @@ import {
   type RemoteTrack,
   type RemoteTrackPublication,
   type RemoteParticipant,
+  type Participant,
   DisconnectReason,
 } from "livekit-client";
 import type { WsClient } from "@lib/ws";
@@ -18,8 +19,7 @@ import {
 } from "@stores/voice.store";
 import { loadPref, savePref } from "@components/settings/helpers";
 import { createLogger } from "@lib/logger";
-import { createNoiseSuppressor } from "@lib/noise-suppression";
-import type { NoiseSuppressor } from "@lib/noise-suppression";
+import { createRNNoiseProcessor } from "@lib/noise-suppression";
 
 const log = createLogger("livekitSession");
 
@@ -32,81 +32,52 @@ export function parseUserId(identity: string): number {
   return 0;
 }
 
-/** Compute RMS audio level (0-1) from frequency data. */
-export function computeRms(data: Uint8Array<ArrayBuffer>): number {
-  let sum = 0;
-  for (let i = 0; i < data.length; i++) {
-    const v = (data[i] ?? 0) / 255;
-    sum += v * v;
-  }
-  return Math.sqrt(sum / data.length);
-}
-
-/** Get saved per-user volume (0-200 range, default 100). */
+/** Get saved per-user volume (0-200 range, default 100). Applied via LiveKit's GainNode-backed setVolume(). */
 function getSavedUserVolume(userId: number): number {
   return loadPref<number>(`userVolume_${userId}`, 100);
 }
 
-/** Compare two pre-sorted speaker ID arrays. Both must be sorted ascending. */
-function speakerSetsEqual(a: readonly number[], b: readonly number[]): boolean {
-  if (a.length !== b.length) return false;
-  for (let i = 0; i < a.length; i++) {
-    if (a[i] !== b[i]) return false;
-  }
-  return true;
-}
-
 // --- Types ---
 
 type RemoteVideoCallback = (userId: number, stream: MediaStream) => void;
 type RemoteVideoRemovedCallback = (userId: number) => void;
 
-interface RemoteAnalyser {
-  analyser: AnalyserNode;
-  source: MediaStreamAudioSourceNode;
-  data: Uint8Array<ArrayBuffer>;
-}
-
 // --- LiveKitSession class ---
 
 export class LiveKitSession {
   private room: Room | null = null;
   private ws: WsClient | null = null;
-  private noiseSuppressor: NoiseSuppressor | null = null;
   private onErrorCallback: ((message: string) => void) | null = null;
   private currentChannelId: number | null = null;
   private serverHost: string | null = null;
-  private speakingPollInterval: ReturnType<typeof setInterval> | null = null;
-  private speakingThreshold = ((100 - 50) / 100) * 0.15;
-  private rawMicStream: MediaStream | null = null;
-  private readonly audioElements = new Map<string, HTMLAudioElement>();
-  private audioContainer: HTMLDivElement | null = null;
   private onRemoteVideoCallback: RemoteVideoCallback | null = null;
   private onRemoteVideoRemovedCallback: RemoteVideoRemovedCallback | null = null;
-  private sharedAudioCtx: AudioContext | null = null;
-  private readonly remoteAnalysers = new Map<number, RemoteAnalyser>();
-  private localMicGated = false;
-  private localAnalyser: AnalyserNode | null = null;
-  private localAnalyserSource: MediaStreamAudioSourceNode | null = null;
-  private localAnalyserClonedTrack: MediaStreamTrack | null = null;
-  private readonly localAnalyserData = new Uint8Array(128);
-  private previousSpeakerIds: readonly number[] = [];
   private tokenRefreshTimer: ReturnType<typeof setTimeout> | null = null;
+  /** Master output volume multiplier (0-2.0). Per-user volumes are scaled by this. */
+  private outputVolumeMultiplier = loadPref<number>("outputVolume", 100) / 100;
 
-  // --- Shared AudioContext (lazy) ---
+  // --- RNNoise processor (LiveKit TrackProcessor API) ---
 
-  private getSharedAudioCtx(): AudioContext {
-    if (this.sharedAudioCtx === null || this.sharedAudioCtx.state === "closed") {
-      this.sharedAudioCtx = new AudioContext({ sampleRate: 48000 });
-    }
-    return this.sharedAudioCtx;
+  /** Attach RNNoise processor to the local mic track. Safe to call if already attached. */
+  private async applyNoiseSuppressor(): Promise<void> {
+    if (this.room === null) return;
+    const micPub = this.room.localParticipant.getTrackPublication(Track.Source.Microphone);
+    if (micPub?.track === undefined) return;
+    if (micPub.track.getProcessor() !== undefined) return;
+    const processor = createRNNoiseProcessor();
+    // eslint-disable-next-line @typescript-eslint/no-explicit-any -- LocalTrack.setProcessor uses wide generic, but AudioProcessorOptions is guaranteed at runtime with webAudioMix
+    await micPub.track.setProcessor(processor as any);
+    log.info("RNNoise processor attached to mic track");
   }
 
-  private closeSharedAudioCtx(): void {
-    if (this.sharedAudioCtx !== null) {
-      void this.sharedAudioCtx.close();
-      this.sharedAudioCtx = null;
-    }
+  /** Remove RNNoise processor from the local mic track. Safe to call if none attached. */
+  private async removeNoiseSuppressor(): Promise<void> {
+    if (this.room === null) return;
+    const micPub = this.room.localParticipant.getTrackPublication(Track.Source.Microphone);
+    if (micPub?.track === undefined) return;
+    if (micPub.track.getProcessor() === undefined) return;
+    await micPub.track.stopProcessor();
+    log.info("RNNoise processor removed from mic track");
   }
 
   // --- Room factory ---
@@ -115,6 +86,7 @@ export class LiveKitSession {
     const newRoom = new Room({
       adaptiveStream: true,
       dynacast: true,
+      webAudioMix: true,
       audioCaptureDefaults: {
         echoCancellation: loadPref("echoCancellation", true),
         noiseSuppression: loadPref("noiseSuppression", true),
@@ -124,70 +96,10 @@ export class LiveKitSession {
     newRoom.on(RoomEvent.TrackSubscribed, this.handleTrackSubscribed);
     newRoom.on(RoomEvent.TrackUnsubscribed, this.handleTrackUnsubscribed);
     newRoom.on(RoomEvent.Disconnected, this.handleDisconnected);
+    newRoom.on(RoomEvent.ActiveSpeakersChanged, this.handleActiveSpeakersChanged);
     return newRoom;
   }
 
-  // --- Audio container ---
-
-  private getOrCreateAudioContainer(): HTMLDivElement {
-    if (this.audioContainer !== null) return this.audioContainer;
-    const existing = document.getElementById("voice-audio-container");
-    if (existing instanceof HTMLDivElement) {
-      this.audioContainer = existing;
-      return this.audioContainer;
-    }
-    const div = document.createElement("div");
-    div.id = "voice-audio-container";
-    div.style.display = "none";
-    document.body.appendChild(div);
-    this.audioContainer = div;
-    return this.audioContainer;
-  }
-
-  private cleanupAudioElements(): void {
-    for (const el of this.audioElements.values()) {
-      el.srcObject = null;
-      el.remove();
-    }
-    this.audioElements.clear();
-  }
-
-  // --- RNNoise ---
-
-  private async publishWithNoiseSuppression(): Promise<void> {
-    if (this.room === null) return;
-    const savedDevice = loadPref<string>("audioInputDevice", "");
-    const constraints: MediaStreamConstraints = {
-      audio: {
-        deviceId: savedDevice ? { exact: savedDevice } : undefined,
-        echoCancellation: loadPref("echoCancellation", true),
-        noiseSuppression: loadPref("noiseSuppression", true),
-        autoGainControl: loadPref("autoGainControl", true),
-      },
-      video: false,
-    };
-    const rawStream = await navigator.mediaDevices.getUserMedia(constraints);
-    this.rawMicStream = rawStream;
-    try {
-      this.noiseSuppressor = createNoiseSuppressor();
-      const processedStream = await this.noiseSuppressor.process(rawStream);
-      const processedTrack = processedStream.getAudioTracks()[0];
-      if (processedTrack) {
-        await this.room.localParticipant.publishTrack(processedTrack, {
-          source: Track.Source.Microphone,
-        });
-      }
-    } catch (err) {
-      for (const track of rawStream.getTracks()) track.stop();
-      this.rawMicStream = null;
-      if (this.noiseSuppressor !== null) {
-        this.noiseSuppressor.destroy();
-        this.noiseSuppressor = null;
-      }
-      throw err;
-    }
-  }
-
   // --- Room event handlers (arrow fns to preserve `this`) ---
 
   private handleTrackSubscribed = (
@@ -197,22 +109,20 @@ export class LiveKitSession {
   ): void => {
     const userId = parseUserId(participant.identity);
     if (track.kind === Track.Kind.Audio) {
-      const container = this.getOrCreateAudioContainer();
+      // Attach creates an <audio> element — required for playback
       const audioEl = track.attach();
+      audioEl.style.display = "none";
+      document.body.appendChild(audioEl);
+      // Apply saved per-user volume scaled by master output volume
       const savedVolume = userId > 0 ? getSavedUserVolume(userId) : 100;
       audioEl.volume = Math.min(savedVolume, 100) / 100;
-      if (voiceStore.getState().localDeafened) audioEl.muted = true;
       const savedOutput = loadPref<string>("audioOutputDevice", "");
       if (savedOutput !== "" && typeof audioEl.setSinkId === "function") {
         audioEl.setSinkId(savedOutput).catch((err) => {
           log.warn("Failed to set output device on remote audio", err);
         });
       }
-      container.appendChild(audioEl);
-      const trackKey = `${participant.identity}-${track.sid}`;
-      this.audioElements.set(trackKey, audioEl);
-      if (userId > 0) this.addRemoteAnalyser(userId, track.mediaStreamTrack);
-      log.debug("Remote audio track subscribed", { userId, trackSid: track.sid });
+      log.debug("Remote audio track subscribed and attached", { userId, trackSid: track.sid });
     } else if (track.kind === Track.Kind.Video) {
       if (userId > 0 && this.onRemoteVideoCallback !== null) {
         const stream = new MediaStream([track.mediaStreamTrack]);
@@ -229,11 +139,8 @@ export class LiveKitSession {
   ): void => {
     const userId = parseUserId(participant.identity);
     if (track.kind === Track.Kind.Audio) {
-      track.detach().forEach((el) => el.remove());
-      const trackKey = `${participant.identity}-${track.sid}`;
-      this.audioElements.delete(trackKey);
-      if (userId > 0) this.removeRemoteAnalyser(userId);
-      log.debug("Remote audio track unsubscribed", { userId, trackSid: track.sid });
+      for (const el of track.detach()) el.remove();
+      log.debug("Remote audio track unsubscribed and detached", { userId, trackSid: track.sid });
     } else if (track.kind === Track.Kind.Video) {
       track.detach();
       if (userId > 0) this.onRemoteVideoRemovedCallback?.(userId);
@@ -241,126 +148,17 @@ export class LiveKitSession {
     }
   };
 
-  // --- Remote audio analysers ---
-
-  private addRemoteAnalyser(userId: number, mediaTrack: MediaStreamTrack): void {
-    this.removeRemoteAnalyser(userId);
-    try {
-      const ctx = this.getSharedAudioCtx();
-      const analyser = ctx.createAnalyser();
-      analyser.fftSize = 256;
-      analyser.smoothingTimeConstant = 0.5;
-      const stream = new MediaStream([mediaTrack]);
-      const source = ctx.createMediaStreamSource(stream);
-      source.connect(analyser);
-      this.remoteAnalysers.set(userId, { analyser, source, data: new Uint8Array(128) as Uint8Array<ArrayBuffer> });
-    } catch (err) {
-      log.warn("Failed to create remote analyser", { userId, error: err });
+  /** LiveKit's built-in speaking detection — replaces custom RMS polling. */
+  private handleActiveSpeakersChanged = (speakers: Participant[]): void => {
+    if (this.currentChannelId === null) return;
+    const speakerIds: number[] = [];
+    for (const speaker of speakers) {
+      const userId = parseUserId(speaker.identity);
+      if (userId > 0) speakerIds.push(userId);
     }
-  }
-
-  private removeRemoteAnalyser(userId: number): void {
-    const ra = this.remoteAnalysers.get(userId);
-    if (ra) {
-      ra.source.disconnect();
-      ra.analyser.disconnect();
-      this.remoteAnalysers.delete(userId);
-    }
-  }
-
-  private cleanupAllRemoteAnalysers(): void {
-    for (const [id] of this.remoteAnalysers) this.removeRemoteAnalyser(id);
-  }
-
-  // --- Local mic analyser ---
-
-  private startLocalAnalyser(): void {
-    this.stopLocalAnalyser();
-    if (this.room === null) return;
-    const micPub = this.room.localParticipant.getTrackPublication(Track.Source.Microphone);
-    const mediaTrack = micPub?.track?.mediaStreamTrack;
-    if (!mediaTrack) return;
-    try {
-      const ctx = this.getSharedAudioCtx();
-      this.localAnalyser = ctx.createAnalyser();
-      this.localAnalyser.fftSize = 256;
-      this.localAnalyser.smoothingTimeConstant = 0.5;
-      this.localAnalyserClonedTrack = mediaTrack.clone();
-      const stream = new MediaStream([this.localAnalyserClonedTrack]);
-      this.localAnalyserSource = ctx.createMediaStreamSource(stream);
-      this.localAnalyserSource.connect(this.localAnalyser);
-      log.debug("Local mic analyser started");
-    } catch (err) {
-      log.warn("Failed to start local mic analyser", err);
-    }
-  }
-
-  private stopLocalAnalyser(): void {
-    if (this.localAnalyserSource !== null) {
-      this.localAnalyserSource.disconnect();
-      this.localAnalyserSource = null;
-    }
-    if (this.localAnalyser !== null) {
-      this.localAnalyser.disconnect();
-      this.localAnalyser = null;
-    }
-    if (this.localAnalyserClonedTrack !== null) {
-      this.localAnalyserClonedTrack.stop();
-      this.localAnalyserClonedTrack = null;
-    }
-  }
-
-  // --- Speaking poll ---
-
-  private startSpeakingPoll(): void {
-    this.stopSpeakingPoll();
-    this.localMicGated = false;
-    this.previousSpeakerIds = [];
-    this.speakingPollInterval = setInterval(() => {
-      if (this.room === null || this.currentChannelId === null) return;
-      const speakerIds: number[] = [];
-      let localLevel = 0;
-      if (this.localAnalyser !== null) {
-        this.localAnalyser.getByteFrequencyData(this.localAnalyserData);
-        localLevel = computeRms(this.localAnalyserData);
-      }
-      const localSpeaking = localLevel > this.speakingThreshold;
-      const micPub = this.room.localParticipant.getTrackPublication(Track.Source.Microphone);
-      if (micPub?.track?.mediaStreamTrack) {
-        if (localSpeaking && this.localMicGated) {
-          micPub.track.mediaStreamTrack.enabled = true;
-          this.localMicGated = false;
-        } else if (!localSpeaking && !this.localMicGated) {
-          micPub.track.mediaStreamTrack.enabled = false;
-          this.localMicGated = true;
-        }
-      }
-      if (localSpeaking) {
-        const localId = parseUserId(this.room.localParticipant.identity);
-        if (localId > 0) speakerIds.push(localId);
-      }
-      for (const [userId, ra] of this.remoteAnalysers) {
-        ra.analyser.getByteFrequencyData(ra.data);
-        if (computeRms(ra.data) > this.speakingThreshold) speakerIds.push(userId);
-      }
-      // Sort in place so speakerSetsEqual can compare without allocations.
-      speakerIds.sort((x, y) => x - y);
-      if (!speakerSetsEqual(speakerIds, this.previousSpeakerIds)) {
-        this.previousSpeakerIds = speakerIds; // already a fresh array each tick
-        setSpeakers({ channel_id: this.currentChannelId, speakers: speakerIds });
-      }
-    }, 100);
-  }
-
-  private stopSpeakingPoll(): void {
-    if (this.speakingPollInterval !== null) {
-      clearInterval(this.speakingPollInterval);
-      this.speakingPollInterval = null;
-    }
-    this.stopLocalAnalyser();
-    this.cleanupAllRemoteAnalysers();
-    this.previousSpeakerIds = [];
-  }
+    speakerIds.sort((x, y) => x - y);
+    setSpeakers({ channel_id: this.currentChannelId, speakers: speakerIds });
+  };
 
   private handleDisconnected = (reason?: DisconnectReason): void => {
     log.info("LiveKit room disconnected", { reason });
@@ -408,23 +206,31 @@ export class LiveKitSession {
     }
     log.info("Requesting voice token refresh");
     this.ws.send({ type: "voice_token_refresh", payload: {} });
-    // Re-arm as fallback in case the server doesn't respond (network hiccup,
-    // restart). If the server does respond, handleVoiceTokenRefresh restarts
-    // the timer, superseding this one.
     this.startTokenRefreshTimer();
   }
 
-  /**
-   * Handle a voice_token message that is a refresh (room already connected).
-   * LiveKit tokens are validated at connect time only — the existing connection
-   * stays alive regardless of token expiry. We just restart the refresh timer
-   * so we keep requesting fresh tokens periodically.
-   */
   handleVoiceTokenRefresh(): void {
     this.startTokenRefreshTimer();
     log.info("Voice token refreshed, timer restarted");
   }
 
+  // --- Volume helpers ---
+
+  /** Compute the effective volume for a participant: per-user volume * master output. */
+  private getEffectiveVolume(userId: number): number {
+    const userVol = userId > 0 ? getSavedUserVolume(userId) : 100;
+    return (userVol / 100) * this.outputVolumeMultiplier;
+  }
+
+  /** Apply effective volume to all remote participants. */
+  private applyAllVolumes(): void {
+    if (this.room === null) return;
+    for (const participant of this.room.remoteParticipants.values()) {
+      const userId = parseUserId(participant.identity);
+      participant.setVolume(this.getEffectiveVolume(userId));
+    }
+  }
+
   // --- Public API ---
 
   setWsClient(client: WsClient): void { this.ws = client; }
@@ -442,9 +248,6 @@ export class LiveKitSession {
   async handleVoiceToken(
     token: string, url: string, channelId: number, directUrl?: string,
   ): Promise<void> {
-    // If already connected to the same channel, this is a token refresh response.
-    // LiveKit validates tokens only at connect time, so just restart the timer.
-    // Guard on room.state to avoid treating a mid-retry token as a refresh.
     if (this.room !== null && this.currentChannelId === channelId
         && this.room.state === "connected") {
       this.handleVoiceTokenRefresh();
@@ -473,16 +276,11 @@ export class LiveKitSession {
         }
       }
       log.info("Connected to LiveKit room", { channelId, url: resolvedUrl });
-      this.speakingThreshold = ((100 - loadPref<number>("voiceSensitivity", 50)) / 100) * 0.15;
-      this.startSpeakingPoll();
-      const enhancedNS = loadPref<boolean>("enhancedNoiseSuppression", false);
       try {
-        if (enhancedNS) {
-          await this.publishWithNoiseSuppression();
-          log.info("Published mic with RNNoise noise suppression");
-        } else {
-          await this.room.localParticipant.setMicrophoneEnabled(true);
-          log.info("Published mic via LiveKit native capture");
+        await this.room.localParticipant.setMicrophoneEnabled(true);
+        log.info("Published mic via LiveKit native capture");
+        if (loadPref<boolean>("enhancedNoiseSuppression", false)) {
+          await this.applyNoiseSuppressor();
         }
       } catch (micErr) {
         if (micErr instanceof DOMException && micErr.name === "NotAllowedError") {
@@ -500,8 +298,9 @@ export class LiveKitSession {
       if (savedInput) await this.room.switchActiveDevice("audioinput", savedInput);
       const savedOutput = loadPref<string>("audioOutputDevice", "");
       if (savedOutput) await this.room.switchActiveDevice("audiooutput", savedOutput);
+      // Apply saved input volume
+      this.applyInputVolume(loadPref<number>("inputVolume", 100));
       this.currentChannelId = channelId;
-      this.startLocalAnalyser();
       this.startTokenRefreshTimer();
       log.info("Voice session active", { channelId });
     } catch (err) {
@@ -518,25 +317,14 @@ export class LiveKitSession {
     if (sendWs && this.ws !== null) {
       this.ws.send({ type: "voice_leave", payload: {} });
     }
-    if (this.rawMicStream !== null) {
-      for (const track of this.rawMicStream.getTracks()) track.stop();
-      this.rawMicStream = null;
-    }
-    if (this.noiseSuppressor !== null) {
-      this.noiseSuppressor.destroy();
-      this.noiseSuppressor = null;
-    }
-    this.stopSpeakingPoll();
+    this.cleanupInputGain();
     if (this.room !== null) {
       const r = this.room;
       this.room = null;
       r.removeAllListeners();
       r.disconnect().catch((err) => log.warn("room.disconnect() error (non-fatal)", err));
     }
-    this.cleanupAudioElements();
-    this.closeSharedAudioCtx();
     this.currentChannelId = null;
-    // Reset camera state so the UI doesn't show a stale video grid on rejoin.
     setLocalCamera(false);
     log.info("Left voice session");
   }
@@ -557,12 +345,10 @@ export class LiveKitSession {
 
   setDeafened(deafened: boolean): void {
     setLocalDeafened(deafened);
-    if (this.room !== null) {
-      for (const participant of this.room.remoteParticipants.values()) {
-        for (const pub of participant.audioTrackPublications.values()) pub.setSubscribed(!deafened);
-      }
+    if (this.room === null) return;
+    for (const participant of this.room.remoteParticipants.values()) {
+      for (const pub of participant.audioTrackPublications.values()) pub.setSubscribed(!deafened);
     }
-    for (const el of this.audioElements.values()) el.muted = deafened;
     log.debug("Deafen state changed", { deafened });
   }
 
@@ -610,27 +396,21 @@ export class LiveKitSession {
       return;
     }
     try {
-      const enhancedNS = loadPref<boolean>("enhancedNoiseSuppression", false);
-      if (enhancedNS && this.noiseSuppressor !== null) {
-        if (this.rawMicStream !== null) {
-          for (const track of this.rawMicStream.getTracks()) track.stop();
-          this.rawMicStream = null;
-        }
-        this.noiseSuppressor.destroy();
-        this.noiseSuppressor = null;
-        for (const pub of this.room.localParticipant.audioTrackPublications.values()) {
-          if (pub.source === Track.Source.Microphone && pub.track) {
-            await this.room.localParticipant.unpublishTrack(pub.track);
-          }
-        }
-        await this.publishWithNoiseSuppression();
+      if (deviceId) {
+        await this.room.switchActiveDevice("audioinput", deviceId);
       } else {
-        if (deviceId) {
-          await this.room.switchActiveDevice("audioinput", deviceId);
-        } else {
-          await this.room.localParticipant.setMicrophoneEnabled(false);
-          await this.room.localParticipant.setMicrophoneEnabled(true);
-        }
+        await this.room.localParticipant.setMicrophoneEnabled(false);
+        await this.room.localParticipant.setMicrophoneEnabled(true);
+      }
+      // Reset and re-apply input volume after device switch (source track changed)
+      this.cleanupInputGain();
+      this.applyInputVolume(loadPref<number>("inputVolume", 100));
+      // Re-apply or remove RNNoise processor based on current setting
+      const enhancedNS = loadPref<boolean>("enhancedNoiseSuppression", false);
+      if (enhancedNS) {
+        await this.applyNoiseSuppressor();
+      } else {
+        await this.removeNoiseSuppressor();
       }
       log.info("Switched input device", { deviceId });
     } catch (err) {
@@ -641,13 +421,6 @@ export class LiveKitSession {
 
   async switchOutputDevice(deviceId: string): Promise<void> {
     if (this.room !== null) await this.room.switchActiveDevice("audiooutput", deviceId);
-    for (const el of this.audioElements.values()) {
-      if (typeof el.setSinkId === "function") {
-        try { await el.setSinkId(deviceId); } catch (err) {
-          log.warn("Failed to set output device on audio element", err);
-        }
-      }
-    }
     log.info("Switched output device", { deviceId });
   }
 
@@ -657,13 +430,7 @@ export class LiveKitSession {
     if (this.room !== null) {
       for (const participant of this.room.remoteParticipants.values()) {
         if (parseUserId(participant.identity) === userId) {
-          for (const pub of participant.audioTrackPublications.values()) {
-            if (pub.track) {
-              for (const el of pub.track.attachedElements) {
-                if (el instanceof HTMLAudioElement) el.volume = Math.min(clamped, 100) / 100;
-              }
-            }
-          }
+          participant.setVolume((clamped / 100) * this.outputVolumeMultiplier);
         }
       }
     }
@@ -671,23 +438,110 @@ export class LiveKitSession {
 
   getUserVolume(userId: number): number { return getSavedUserVolume(userId); }
 
+  /** Input volume GainNode — adjusts mic gain via the WebRTC sender. */
+  private inputGainNode: GainNode | null = null;
+  private inputGainCtx: AudioContext | null = null;
+  private inputGainDest: MediaStreamAudioDestinationNode | null = null;
+
+  /** Apply input volume gain to the local mic track via a Web Audio GainNode. */
+  private applyInputVolume(volume: number): void {
+    if (this.room === null) return;
+    const micPub = this.room.localParticipant.getTrackPublication(Track.Source.Microphone);
+    if (micPub?.track === undefined) return;
+    const gain = Math.max(0, Math.min(200, volume)) / 100;
+
+    // At 100% (gain=1.0), tear down the pipeline — no processing needed
+    if (gain === 1 && this.inputGainNode !== null) {
+      this.restoreOriginalSenderTrack();
+      this.cleanupInputGain();
+      log.info("Input volume reset to 100% — gain pipeline removed");
+      return;
+    }
+
+    // No gain node and volume is default — nothing to do
+    if (gain === 1) return;
+
+    if (this.inputGainNode !== null) {
+      this.inputGainNode.gain.setTargetAtTime(gain, 0, 0.05);
+      log.debug("Input volume adjusted", { gain });
+      return;
+    }
+
+    // Build GainNode pipeline and replace the sender's track
+    try {
+      const mediaTrack = micPub.track.mediaStreamTrack;
+      const ctx = new AudioContext({ sampleRate: 48000 });
+      const source = ctx.createMediaStreamSource(new MediaStream([mediaTrack]));
+      const gainNode = ctx.createGain();
+      gainNode.gain.setTargetAtTime(gain, 0, 0.05);
+      const dest = ctx.createMediaStreamDestination();
+      source.connect(gainNode);
+      gainNode.connect(dest);
+
+      this.inputGainNode = gainNode;
+      this.inputGainCtx = ctx;
+      this.inputGainDest = dest;
+
+      // Replace the WebRTC sender's track with the gain-adjusted one
+      const adjustedTrack = dest.stream.getAudioTracks()[0];
+      if (adjustedTrack !== undefined && micPub.track.sender) {
+        void micPub.track.sender.replaceTrack(adjustedTrack).catch((err) => {
+          log.warn("Failed to replace sender track with gain-adjusted track", err);
+        });
+      }
+      log.info("Input volume GainNode created", { gain });
+    } catch (err) {
+      log.warn("Failed to set up input volume gain", err);
+    }
+  }
+
+  /** Restore the original mic track on the WebRTC sender (undo gain pipeline). */
+  private restoreOriginalSenderTrack(): void {
+    if (this.room === null) return;
+    const micPub = this.room.localParticipant.getTrackPublication(Track.Source.Microphone);
+    if (micPub?.track === undefined) return;
+    const originalTrack = micPub.track.mediaStreamTrack;
+    if (micPub.track.sender) {
+      void micPub.track.sender.replaceTrack(originalTrack).catch((err) => {
+        log.warn("Failed to restore original sender track", err);
+      });
+    }
+  }
+
+  private cleanupInputGain(): void {
+    if (this.inputGainNode !== null) {
+      this.inputGainNode.disconnect();
+      this.inputGainNode = null;
+    }
+    if (this.inputGainDest !== null) {
+      this.inputGainDest.disconnect();
+      this.inputGainDest = null;
+    }
+    if (this.inputGainCtx !== null) {
+      void this.inputGainCtx.close();
+      this.inputGainCtx = null;
+    }
+  }
+
   setInputVolume(volume: number): void {
-    savePref("inputVolume", Math.max(0, Math.min(200, volume)));
+    const clamped = Math.max(0, Math.min(200, volume));
+    savePref("inputVolume", clamped);
+    this.applyInputVolume(clamped);
   }
 
   setOutputVolume(volume: number): void {
     const clamped = Math.max(0, Math.min(200, volume));
     savePref("outputVolume", clamped);
-    if (this.room !== null) {
-      for (const el of this.audioElements.values()) {
-        el.volume = Math.min(clamped, 100) / 100;
-      }
-    }
+    this.outputVolumeMultiplier = clamped / 100;
+    // Re-apply all per-user volumes scaled by the new master output
+    this.applyAllVolumes();
   }
 
-  setVoiceSensitivity(sensitivity: number): void {
-    const clamped = Math.max(0, Math.min(100, sensitivity));
-    this.speakingThreshold = ((100 - clamped) / 100) * 0.15;
+  setVoiceSensitivity(_sensitivity: number): void {
+    // Voice sensitivity is now handled by LiveKit's built-in speaking detection.
+    // The sensitivity parameter is saved in preferences by the UI but
+    // LiveKit's server-side VAD determines speaking state.
+    log.debug("Voice sensitivity setting saved (handled by LiveKit VAD)");
   }
 
   getLocalCameraStream(): MediaStream | null {
@@ -699,24 +553,34 @@ export class LiveKitSession {
 
   getSessionDebugInfo(): Record<string, unknown> {
     if (this.room === null) {
-      return { hasRoom: false, hasNoiseSuppressor: this.noiseSuppressor !== null, currentChannelId: this.currentChannelId };
+      return { hasRoom: false, hasRNNoiseProcessor: false, currentChannelId: this.currentChannelId };
     }
-    const remoteParticipants = [...this.room.remoteParticipants.values()].map((p) => ({
-      identity: p.identity,
-      userId: parseUserId(p.identity),
-      tracks: [...p.trackPublications.values()].map((pub) => ({
-        sid: pub.trackSid, source: pub.source, kind: pub.kind,
-        subscribed: pub.isSubscribed, enabled: pub.isEnabled,
-      })),
-    }));
+    const remoteParticipants = [...this.room.remoteParticipants.values()].map((p) => {
+      const userId = parseUserId(p.identity);
+      return {
+        identity: p.identity,
+        userId,
+        volume: p.getVolume(),
+        effectiveVolume: this.getEffectiveVolume(userId),
+        tracks: [...p.trackPublications.values()].map((pub) => ({
+          sid: pub.trackSid, source: pub.source, kind: pub.kind,
+          subscribed: pub.isSubscribed, enabled: pub.isEnabled,
+        })),
+      };
+    });
     const localTracks = [...this.room.localParticipant.trackPublications.values()].map((pub) => ({
       sid: pub.trackSid, source: pub.source, kind: pub.kind, isMuted: pub.isMuted,
     }));
     return {
       hasRoom: true, roomName: this.room.name, roomState: this.room.state,
-      hasNoiseSuppressor: this.noiseSuppressor !== null, currentChannelId: this.currentChannelId,
+      hasRNNoiseProcessor: this.room.localParticipant.getTrackPublication(Track.Source.Microphone)?.track?.getProcessor() !== undefined,
+      currentChannelId: this.currentChannelId,
+      outputVolumeMultiplier: this.outputVolumeMultiplier,
+      inputGainActive: this.inputGainNode !== null,
+      inputGainValue: this.inputGainNode?.gain.value ?? null,
+      inputGainCtxState: this.inputGainCtx?.state ?? null,
       localParticipant: this.room.localParticipant.identity, localTracks,
-      remoteParticipants, audioElements: this.audioElements.size,
+      remoteParticipants,
     };
   }
 }
@@ -725,6 +589,10 @@ export class LiveKitSession {
 
 const session = new LiveKitSession();
 
+// Expose debug info on window for DevTools console access
+// Usage: JSON.stringify(__lkDebug(), null, 2)
+(window as unknown as Record<string, unknown>).__lkDebug = session.getSessionDebugInfo.bind(session);
+
 export const setWsClient = session.setWsClient.bind(session);
 export const setServerHost = session.setServerHost.bind(session);
 export const setOnError = session.setOnError.bind(session);
diff --git a/Client/tauri-client/src/lib/noise-suppression.ts b/Client/tauri-client/src/lib/noise-suppression.ts
index 458f2b06..5a6340ea 100644
--- a/Client/tauri-client/src/lib/noise-suppression.ts
+++ b/Client/tauri-client/src/lib/noise-suppression.ts
@@ -1,14 +1,17 @@
 // =============================================================================
-// Noise Suppression — RNNoise ML-based noise removal via Web Audio API
+// Noise Suppression — RNNoise ML-based noise removal as a LiveKit TrackProcessor
+//
+// Implements LiveKit's TrackProcessor<Track.Kind.Audio> interface so it
+// integrates with setProcessor() / stopProcessor() lifecycle, device switching,
+// and mid-call toggling automatically.
 //
-// Inserts between getUserMedia stream and the PeerConnection to clean audio.
 // RNNoise processes 480-sample frames at 48kHz (10ms).
-//
 // Uses AudioWorklet (modern, runs on audio thread) with ScriptProcessorNode
 // fallback (deprecated but widely supported).
 // =============================================================================
 
 import { createRNNWasmModule } from "@jitsi/rnnoise-wasm";
+import { Track, type TrackProcessor, type AudioProcessorOptions } from "livekit-client";
 import { createLogger } from "@lib/logger";
 
 const log = createLogger("noise-suppression");
@@ -16,13 +19,8 @@ const log = createLogger("noise-suppression");
 const RNNOISE_FRAME_SIZE = 480;
 const SCRIPT_PROCESSOR_BUFFER = 4096;
 
-export interface NoiseSuppressor {
-  process(input: MediaStream): Promise<MediaStream>;
-  destroy(): void;
-}
-
 // ---------------------------------------------------------------------------
-// Shared WASM module cache (used by ScriptProcessorNode fallback)
+// Shared WASM module cache
 // ---------------------------------------------------------------------------
 
 interface RNNoiseModule {
@@ -52,102 +50,78 @@ async function loadRNNoise(): Promise<RNNoiseModule> {
   return mod;
 }
 
+/** Check if AudioWorklet is available in this browser context. */
+function supportsAudioWorklet(): boolean {
+  try {
+    return typeof AudioWorkletNode !== "undefined"
+      && typeof AudioContext !== "undefined"
+      && "audioWorklet" in AudioContext.prototype;
+  } catch {
+    return false;
+  }
+}
+
 // ---------------------------------------------------------------------------
-// AudioWorklet-based suppressor (preferred, runs on audio thread)
+// Internal processing pipeline — shared by both init strategies
 // ---------------------------------------------------------------------------
 
-function createWorkletSuppressor(): NoiseSuppressor {
-  let audioContext: AudioContext | null = null;
-  let sourceNode: MediaStreamAudioSourceNode | null = null;
-  let destNode: MediaStreamAudioDestinationNode | null = null;
-  let workletNode: AudioWorkletNode | null = null;
-  let destroyed = false;
+interface ProcessingPipeline {
+  readonly processedTrack: MediaStreamTrack;
+  destroy(): void;
+}
+
+/** AudioWorklet-based pipeline (preferred, runs on audio thread). */
+async function createWorkletPipeline(
+  inputTrack: MediaStreamTrack,
+  audioContext: AudioContext,
+): Promise<ProcessingPipeline> {
+  await audioContext.audioWorklet.addModule("/rnnoise-worklet.js");
+  const wasmResponse = await fetch("/rnnoise.wasm");
+  const wasmBytes = await wasmResponse.arrayBuffer();
+
+  const source = audioContext.createMediaStreamSource(new MediaStream([inputTrack]));
+  const dest = audioContext.createMediaStreamDestination();
+  const workletNode = new AudioWorkletNode(audioContext, "rnnoise-processor", {
+    numberOfInputs: 1,
+    numberOfOutputs: 1,
+    outputChannelCount: [1],
+  });
+
+  const initPromise = new Promise<void>((resolve, reject) => {
+    workletNode.port.onmessage = (event: MessageEvent) => {
+      if (event.data.type === "ready") resolve();
+      else if (event.data.type === "error") reject(new Error(event.data.message));
+    };
+  });
+  workletNode.port.postMessage({ type: "init", wasmBytes }, [wasmBytes]);
+  await initPromise;
+
+  source.connect(workletNode);
+  workletNode.connect(dest);
+
+  log.info("RNNoise AudioWorklet processing active");
 
   return {
-    async process(input: MediaStream): Promise<MediaStream> {
-      if (destroyed) throw new Error("NoiseSuppressor destroyed");
-
-      audioContext = new AudioContext({ sampleRate: 48000 });
-
-      // Load the worklet processor module
-      await audioContext.audioWorklet.addModule("/rnnoise-worklet.js");
-
-      // Fetch WASM bytes to send to the worklet thread
-      const wasmResponse = await fetch("/rnnoise.wasm");
-      const wasmBytes = await wasmResponse.arrayBuffer();
-
-      sourceNode = audioContext.createMediaStreamSource(input);
-      destNode = audioContext.createMediaStreamDestination();
-
-      workletNode = new AudioWorkletNode(audioContext, "rnnoise-processor", {
-        numberOfInputs: 1,
-        numberOfOutputs: 1,
-        outputChannelCount: [1],
-      });
-
-      // Wait for WASM init in the worklet
-      const initPromise = new Promise<void>((resolve, reject) => {
-        if (workletNode === null) { reject(new Error("No worklet")); return; }
-        workletNode.port.onmessage = (event: MessageEvent) => {
-          if (event.data.type === "ready") {
-            resolve();
-          } else if (event.data.type === "error") {
-            reject(new Error(event.data.message));
-          }
-        };
-      });
-
-      // Send WASM bytes to the worklet for initialization
-      workletNode.port.postMessage({ type: "init", wasmBytes }, [wasmBytes]);
-      await initPromise;
-
-      sourceNode.connect(workletNode);
-      workletNode.connect(destNode);
-
-      log.info("RNNoise AudioWorklet processing active");
-      return destNode.stream;
-    },
-
-    destroy(): void {
-      if (destroyed) return;
-      destroyed = true;
-
-      if (workletNode !== null) {
-        workletNode.port.postMessage({ type: "destroy" });
-        workletNode.disconnect();
-        workletNode = null;
-      }
-      if (sourceNode !== null) {
-        sourceNode.disconnect();
-        sourceNode = null;
-      }
-      if (destNode !== null) {
-        destNode.disconnect();
-        destNode = null;
-      }
-      if (audioContext !== null) {
-        void audioContext.close();
-        audioContext = null;
-      }
-      log.info("RNNoise AudioWorklet destroyed");
+    processedTrack: dest.stream.getAudioTracks()[0]!,
+    destroy() {
+      workletNode.port.postMessage({ type: "destroy" });
+      workletNode.disconnect();
+      source.disconnect();
+      dest.disconnect();
+      log.info("RNNoise AudioWorklet pipeline destroyed");
     },
   };
 }
 
-// ---------------------------------------------------------------------------
-// ScriptProcessorNode fallback (deprecated but universal)
-// ---------------------------------------------------------------------------
-
-function createScriptProcessorSuppressor(): NoiseSuppressor {
-  let audioContext: AudioContext | null = null;
-  let sourceNode: MediaStreamAudioSourceNode | null = null;
-  let destNode: MediaStreamAudioDestinationNode | null = null;
-  let processorNode: ScriptProcessorNode | null = null;
-  let rnnoiseState: number = 0;
-  let inputPtr: number = 0;
-  let outputPtr: number = 0;
-  let wasmModule: RNNoiseModule | null = null;
-  let destroyed = false;
+/** ScriptProcessorNode-based pipeline (fallback). */
+async function createScriptProcessorPipeline(
+  inputTrack: MediaStreamTrack,
+  audioContext: AudioContext,
+): Promise<ProcessingPipeline> {
+  const wasmModule = await loadRNNoise();
+  const rnnoiseState = wasmModule._rnnoise_create();
+  const inputPtr = wasmModule._malloc(RNNOISE_FRAME_SIZE * 4);
+  const outputPtr = wasmModule._malloc(RNNOISE_FRAME_SIZE * 4);
 
   const inputRing = new Float32Array(RNNOISE_FRAME_SIZE);
   let inputRingOffset = 0;
@@ -160,7 +134,6 @@ function createScriptProcessorSuppressor(): NoiseSuppressor {
   let outSampleOffset = 0;
 
   function processFrame(): void {
-    if (wasmModule === null) return;
     const inOff = inputPtr / 4;
     for (let i = 0; i < RNNOISE_FRAME_SIZE; i++) {
       wasmModule.HEAPF32[inOff + i] = (inputRing[i] ?? 0) * 32768;
@@ -181,140 +154,121 @@ function createScriptProcessorSuppressor(): NoiseSuppressor {
     outCount++;
   }
 
+  const source = audioContext.createMediaStreamSource(new MediaStream([inputTrack]));
+  const dest = audioContext.createMediaStreamDestination();
+  const processorNode = audioContext.createScriptProcessor(SCRIPT_PROCESSOR_BUFFER, 1, 1);
+
+  processorNode.onaudioprocess = (event: AudioProcessingEvent) => {
+    const inData = event.inputBuffer.getChannelData(0);
+    const outData = event.outputBuffer.getChannelData(0);
+
+    let inIdx = 0;
+    while (inIdx < inData.length) {
+      const needed = RNNOISE_FRAME_SIZE - inputRingOffset;
+      const toCopy = Math.min(needed, inData.length - inIdx);
+      inputRing.set(inData.subarray(inIdx, inIdx + toCopy), inputRingOffset);
+      inputRingOffset += toCopy;
+      inIdx += toCopy;
+      if (inputRingOffset >= RNNOISE_FRAME_SIZE) {
+        processFrame();
+        inputRingOffset = 0;
+      }
+    }
+
+    let outIdx = 0;
+    while (outIdx < outData.length && outCount > 0) {
+      const chunk = outRing[outReadIdx]!;
+      const available = chunk.length - outSampleOffset;
+      const toWrite = Math.min(available, outData.length - outIdx);
+      outData.set(chunk.subarray(outSampleOffset, outSampleOffset + toWrite), outIdx);
+      outIdx += toWrite;
+      outSampleOffset += toWrite;
+      if (outSampleOffset >= chunk.length) {
+        outReadIdx = (outReadIdx + 1) % OUT_RING_CAPACITY;
+        outCount--;
+        outSampleOffset = 0;
+      }
+    }
+    if (outIdx < outData.length) {
+      outData.fill(0, outIdx);
+    }
+  };
+
+  source.connect(processorNode);
+  processorNode.connect(dest);
+
+  log.info("RNNoise ScriptProcessor processing active (fallback)");
+
   return {
-    async process(input: MediaStream): Promise<MediaStream> {
-      if (destroyed) throw new Error("NoiseSuppressor destroyed");
-
-      wasmModule = await loadRNNoise();
-      rnnoiseState = wasmModule._rnnoise_create();
-      inputPtr = wasmModule._malloc(RNNOISE_FRAME_SIZE * 4);
-      outputPtr = wasmModule._malloc(RNNOISE_FRAME_SIZE * 4);
-
-      audioContext = new AudioContext({ sampleRate: 48000 });
-      sourceNode = audioContext.createMediaStreamSource(input);
-      destNode = audioContext.createMediaStreamDestination();
-      processorNode = audioContext.createScriptProcessor(SCRIPT_PROCESSOR_BUFFER, 1, 1);
-
-      processorNode.onaudioprocess = (event: AudioProcessingEvent) => {
-        const inData = event.inputBuffer.getChannelData(0);
-        const outData = event.outputBuffer.getChannelData(0);
-
-        let inIdx = 0;
-        while (inIdx < inData.length) {
-          const needed = RNNOISE_FRAME_SIZE - inputRingOffset;
-          const toCopy = Math.min(needed, inData.length - inIdx);
-          inputRing.set(inData.subarray(inIdx, inIdx + toCopy), inputRingOffset);
-          inputRingOffset += toCopy;
-          inIdx += toCopy;
-          if (inputRingOffset >= RNNOISE_FRAME_SIZE) {
-            processFrame();
-            inputRingOffset = 0;
-          }
-        }
-
-        let outIdx = 0;
-        while (outIdx < outData.length && outCount > 0) {
-          const chunk = outRing[outReadIdx]!;
-          const available = chunk.length - outSampleOffset;
-          const toWrite = Math.min(available, outData.length - outIdx);
-          outData.set(chunk.subarray(outSampleOffset, outSampleOffset + toWrite), outIdx);
-          outIdx += toWrite;
-          outSampleOffset += toWrite;
-          if (outSampleOffset >= chunk.length) {
-            outReadIdx = (outReadIdx + 1) % OUT_RING_CAPACITY;
-            outCount--;
-            outSampleOffset = 0;
-          }
-        }
-        if (outIdx < outData.length) {
-          outData.fill(0, outIdx);
-        }
-      };
-
-      sourceNode.connect(processorNode);
-      processorNode.connect(destNode);
-
-      log.info("RNNoise ScriptProcessor processing active (fallback)");
-      return destNode.stream;
-    },
-
-    destroy(): void {
-      if (destroyed) return;
-      destroyed = true;
-
-      if (processorNode !== null) {
-        processorNode.onaudioprocess = null;
-        processorNode.disconnect();
-        processorNode = null;
-      }
-      if (sourceNode !== null) {
-        sourceNode.disconnect();
-        sourceNode = null;
-      }
-      if (destNode !== null) {
-        destNode.disconnect();
-        destNode = null;
-      }
-      if (audioContext !== null) {
-        void audioContext.close();
-        audioContext = null;
-      }
-      if (wasmModule !== null && rnnoiseState !== 0) {
-        wasmModule._rnnoise_destroy(rnnoiseState);
-        wasmModule._free(inputPtr);
-        wasmModule._free(outputPtr);
-        rnnoiseState = 0;
-      }
-      outWriteIdx = 0;
-      outReadIdx = 0;
-      outCount = 0;
-      outSampleOffset = 0;
-      log.info("RNNoise ScriptProcessor destroyed");
+    processedTrack: dest.stream.getAudioTracks()[0]!,
+    destroy() {
+      processorNode.onaudioprocess = null;
+      processorNode.disconnect();
+      source.disconnect();
+      dest.disconnect();
+      wasmModule._rnnoise_destroy(rnnoiseState);
+      wasmModule._free(inputPtr);
+      wasmModule._free(outputPtr);
+      log.info("RNNoise ScriptProcessor pipeline destroyed");
     },
   };
 }
 
 // ---------------------------------------------------------------------------
-// Factory — tries AudioWorklet first, falls back to ScriptProcessorNode
+// LiveKit TrackProcessor implementation
 // ---------------------------------------------------------------------------
 
-/** Check if AudioWorklet is available in this browser context. */
-function supportsAudioWorklet(): boolean {
-  try {
-    return typeof AudioWorkletNode !== "undefined"
-      && typeof AudioContext !== "undefined"
-      && "audioWorklet" in AudioContext.prototype;
-  } catch {
-    return false;
-  }
-}
+/**
+ * Creates an RNNoise TrackProcessor compatible with LiveKit's
+ * LocalAudioTrack.setProcessor() API.
+ *
+ * Usage:
+ *   const processor = createRNNoiseProcessor();
+ *   await localAudioTrack.setProcessor(processor);
+ *   // Later:
+ *   await localAudioTrack.stopProcessor();
+ */
+export function createRNNoiseProcessor(): TrackProcessor<Track.Kind.Audio, AudioProcessorOptions> {
+  let pipeline: ProcessingPipeline | null = null;
 
-export function createNoiseSuppressor(): NoiseSuppressor {
-  log.debug("Creating noise suppressor", { audioWorkletSupported: supportsAudioWorklet() });
-  if (supportsAudioWorklet()) {
-    // Wrap in a facade that falls back to ScriptProcessor on failure
-    const worklet = createWorkletSuppressor();
-    let fallback: NoiseSuppressor | null = null;
-    let activeSuppressor: NoiseSuppressor = worklet;
+  return {
+    name: "rnnoise",
 
-    return {
-      async process(input: MediaStream): Promise<MediaStream> {
+    async init(opts: AudioProcessorOptions): Promise<void> {
+      log.debug("RNNoise processor init", { audioWorkletSupported: supportsAudioWorklet() });
+      const ctx = opts.audioContext;
+
+      if (supportsAudioWorklet()) {
         try {
-          return await worklet.process(input);
+          pipeline = await createWorkletPipeline(opts.track, ctx);
+          return;
         } catch (err) {
           log.warn("AudioWorklet failed, falling back to ScriptProcessorNode", err);
-          worklet.destroy();
-          fallback = createScriptProcessorSuppressor();
-          activeSuppressor = fallback;
-          return fallback.process(input);
         }
-      },
-      destroy(): void {
-        activeSuppressor.destroy();
-      },
-    };
-  }
+      }
 
-  log.info("AudioWorklet not supported, using ScriptProcessorNode");
-  return createScriptProcessorSuppressor();
+      pipeline = await createScriptProcessorPipeline(opts.track, ctx);
+    },
+
+    async restart(opts: AudioProcessorOptions): Promise<void> {
+      log.debug("RNNoise processor restart");
+      if (pipeline !== null) {
+        pipeline.destroy();
+        pipeline = null;
+      }
+      await this.init(opts);
+    },
+
+    async destroy(): Promise<void> {
+      if (pipeline !== null) {
+        pipeline.destroy();
+        pipeline = null;
+      }
+      log.info("RNNoise processor destroyed");
+    },
+
+    get processedTrack(): MediaStreamTrack | undefined {
+      return pipeline?.processedTrack;
+    },
+  };
 }
diff --git a/Client/tauri-client/src/lib/types.ts b/Client/tauri-client/src/lib/types.ts
index b3916a2b..fa4be75c 100644
--- a/Client/tauri-client/src/lib/types.ts
+++ b/Client/tauri-client/src/lib/types.ts
@@ -28,7 +28,6 @@ export type WsErrorCode =
   | "INVALID_INPUT"
   | "SERVER_ERROR"
   | "CHANNEL_FULL"
-  | "INVALID_SDP"
   | "VOICE_ERROR"
   | "VIDEO_LIMIT";
 
diff --git a/README.md b/README.md
index 44c2176f..286abfc0 100644
--- a/README.md
+++ b/README.md
@@ -1,8 +1,11 @@
 # OwnCord
 
+*The gaming chat platform you actually own.*
+
 A self-hosted Windows chat platform with real-time messaging,
 voice/video, file sharing, and a web admin panel. Run your own
-server and keep everything under your control.
+server and keep everything under your control — zero cloud
+dependencies, works fully on LAN.
 
 ## Features
 
@@ -21,17 +24,14 @@ server and keep everything under your control.
 
 ### Voice & Video
 
-- Voice channels with WebRTC (Pion SFU)
+- Voice channels powered by LiveKit SFU
 - Webcam video chat with responsive grid layout
 - Mute, deafen, camera, and screenshare controls
 - Push-to-talk with global hotkey (non-consuming, works while unfocused)
 - Per-user volume control (right-click user in voice channel)
-- NAT traversal via Google STUN + configurable external IP
-- RNNoise ML noise suppression (AudioWorklet + fallback)
-- Voice activity detection with configurable sensitivity
-- Silence suppression to save bandwidth
-- Configurable audio quality (low/medium/high)
-- Server-enforced max video streams per room
+- RNNoise ML noise suppression
+- Voice activity detection with speaker indicators
+- LiveKit server runs as a companion process alongside `chatserver.exe`
 
 ### Channels & Organization
 
@@ -111,8 +111,8 @@ Two components: a **Go server** and a **Tauri v2 client**
 |  +---------------+  |  HTTPS  |  +---------------+  |
 |  |  REST Client  |--+------->|  |  REST API     |  |
 |  +---------------+  |         |  +---------------+  |
-|  +---------------+  |  WebRTC |  +---------------+  |
-|  |  Voice/Video  |--+------->|  |  SFU (Pion)   |  |
+|  +---------------+  | LiveKit |  +---------------+  |
+|  |  Voice/Video  |--+------->|  |  LiveKit SFU  |  |
 |  +---------------+  |         |  +---------------+  |
 +---------------------+         |  +---------------+  |
                                 |  |  SQLite DB    |  |
@@ -122,7 +122,7 @@ Two components: a **Go server** and a **Tauri v2 client**
 
 - **WebSocket** — chat messages, typing, presence, voice signaling
 - **REST API** — message history, file uploads, channel management, auth
-- **WebRTC** — voice and video via Pion SFU with Google STUN for NAT traversal
+- **LiveKit** — voice and video via LiveKit SFU (companion process)
 
 ## Project Structure
 
@@ -141,9 +141,9 @@ OwnCord/
 │   └── tauri-client/        # Tauri v2 desktop client
 │       ├── src-tauri/       #   Rust backend (plugins, commands)
 │       ├── src/             #   TypeScript frontend
-│       │   ├── lib/         #     Core services (API, WS, WebRTC, updater)
+│       │   ├── lib/         #     Core services (API, WS, LiveKit, updater)
 │       │   ├── stores/      #     Reactive state (auth, channels, messages, voice)
-│       │   ├── components/  #     UI components (36 modules)
+│       │   ├── components/  #     UI components (28 modules)
 │       │   ├── pages/       #     Page layouts
 │       │   └── styles/      #     CSS
 │       └── tests/           #   Unit, integration, and E2E tests
@@ -200,9 +200,10 @@ The server generates a `config.yaml` on first run. Key settings:
 | `server.name` | `OwnCord Server` | Display name |
 | `tls.mode` | `selfsigned` | TLS mode (see docs) |
 | `upload.max_size_mb` | `10` | Max upload size |
-| `voice.quality` | `medium` | `low`, `medium`, `high` |
-| `voice.external_ip` | — | Public IP for NAT traversal |
-| `voice.turn_enabled` | `true` | Enable TURN relay (requires coturn) |
+| `voice.livekit_url` | `ws://localhost:7880` | LiveKit server WebSocket URL |
+| `voice.livekit_api_key` | `devkey` | LiveKit API key |
+| `voice.livekit_api_secret` | — | LiveKit API secret (min 32 chars) |
+| `voice.livekit_binary` | — | Path to `livekit-server` binary (auto-start) |
 | `server.admin_allowed_cidrs` | private nets | CIDRs allowed to access `/admin` |
 | `github.token` | — | Token for update checks |
 
@@ -237,10 +238,10 @@ Detailed docs live in the `docs/brain/` Obsidian vault:
 
 | Component | Technology |
 | --------- | --------- |
-| Server | Go, chi router, Pion WebRTC |
+| Server | Go, chi router, LiveKit server SDK |
 | Database | SQLite (pure Go, embedded) |
 | Client | Tauri v2 (Rust + TypeScript) |
-| Voice/Video | WebRTC with Pion SFU, Google STUN |
+| Voice/Video | LiveKit SFU (companion process) |
 | Build | NSIS installer, GitHub Actions CI |
 
 ## License
diff --git a/docs/brain/06-Specs/CHATSERVER.md b/docs/brain/06-Specs/CHATSERVER.md
index e166653f..84cbac30 100644
--- a/docs/brain/06-Specs/CHATSERVER.md
+++ b/docs/brain/06-Specs/CHATSERVER.md
@@ -1,5 +1,9 @@
 # ChatServer — Self-Hosted Windows Chat Platform
 
+> **Note:** Most Phase 1-6 tasks are complete as of v1.2.0. See
+> [[02-Tasks/Done|Done]] for the detailed completion list and
+> [[00-Overview/Changelog|Changelog]] for version history.
+
 Native Windows desktop client + self-hosted server.
 Two executables: `chatserver.exe` (server) and
 `OwnCord.exe` (Tauri v2 client). Server operator runs
@@ -19,7 +23,7 @@ the server, friends install the client.
 ### Client (`OwnCord.exe`)
 
 **Tauri v2** (Rust backend + TypeScript/HTML/CSS frontend).
-See LANGUAGE-REVIEW.md for the evaluation that led to this
+See [[07-Archive/LANGUAGE-REVIEW|LANGUAGE-REVIEW]] for the evaluation that led to this
 choice, and CLIENT-ARCHITECTURE.md for the full design.
 
 - Tauri v2 desktop app using system WebView2 (NOT Electron)
@@ -66,108 +70,106 @@ CLIENT (OwnCord.exe) — installed by each friend
 
 ## Phase 1: Protocol & Server Core (2–3 weeks)
 
-- [ ] Define client-server protocol over WebSocket
+- [x] Define client-server protocol over WebSocket
   (JSON messages with type/payload structure)
-- [ ] Message types: auth, chat, typing, presence,
+- [x] Message types: auth, chat, typing, presence,
   channel_update, voice_signal, file_transfer
-- [ ] Server: Go project with `go embed` for admin
+- [x] Server: Go project with `go embed` for admin
   panel static files only
-- [ ] SQLite setup with migrations on startup (users,
+- [x] SQLite setup with migrations on startup (users,
   channels, messages, sessions, roles, invites)
-- [ ] config.yaml generation on first run (port, name,
+- [x] config.yaml generation on first run (port, name,
   max upload size, voice quality, TLS mode)
-- [ ] Server systray icon (getlantern/systray) — minimize to tray, status
-  indicator, open admin panel, quit
 - [ ] Windows Firewall handling on first launch
 - [ ] Optional: register as Windows Service for headless operation
 
 ## Phase 2: Auth & Security (2–3 weeks)
 
-- [ ] Invite-only registration — server generates
+- [x] Invite-only registration — server generates
   invite codes, client has "Redeem Invite" flow
-- [ ] bcrypt (cost 12+) passwords, server-side session tokens (256-bit random)
-- [ ] Client stores auth token securely via Windows Credential Manager / DPAPI
-- [ ] Login rate limiting: 5 attempts/min/IP, lockout after 10 failures
-- [ ] Optional TOTP 2FA (`pquerna/otp`) — QR code
-  during setup, prompts on login
-- [ ] Roles: Owner, Admin, Moderator, Member + custom roles with bitfield permissions
-- [ ] Per-channel permission overrides, enforced server-side on every action
-- [ ] TLS modes: self-signed (default), Let's Encrypt,
+- [x] bcrypt (cost 12+) passwords, server-side session tokens (256-bit random)
+- [x] Client stores auth token securely via Windows Credential Manager / DPAPI
+- [x] Login rate limiting: 5 attempts/min/IP, lockout after 10 failures
+- [ ] Optional TOTP 2FA — planned, not yet implemented (T-023 in backlog).
+  DB schema has `totp_secret` column ready.
+- [x] Roles: Owner, Admin, Moderator, Member + custom roles with bitfield permissions
+- [x] Per-channel permission overrides, enforced server-side on every action
+- [x] TLS modes: self-signed (default), Let's Encrypt,
   manual cert, off (Tailscale)
-- [ ] Client: certificate pinning or trust-on-first-use (TOFU) for self-signed certs
+- [x] Client: certificate pinning or trust-on-first-use (TOFU) for self-signed certs
 
 ## Phase 3: Client App — Core UI (3–4 weeks)
 
-- [ ] Connection dialog: server address, port, login/register, invite code entry
-- [ ] Save server profiles (connect to multiple
+- [x] Connection dialog: server address, port, login/register, invite code entry
+- [x] Save server profiles (connect to multiple
   servers like TeamSpeak)
-- [ ] Main window layout: server list → channel list → message area → member list
-- [ ] Channel tree view with categories, text channels, voice channels
-- [ ] Message rendering: markdown, code blocks, timestamps, avatars, replies, reactions
-- [ ] Message input: multi-line, markdown preview, emoji picker, file drag-and-drop
-- [ ] Unread indicators, @mention badges per channel
-- [ ] System tray: minimize to tray, notification popups, badge count
-- [ ] Keyboard shortcuts: Ctrl+K quick switcher,
+- [x] Main window layout: server list → channel list → message area → member list
+- [x] Channel tree view with categories, text channels, voice channels
+- [x] Message rendering: markdown, code blocks, timestamps, avatars, replies, reactions
+- [x] Message input: multi-line, markdown preview, emoji picker, file drag-and-drop
+- [x] Unread indicators, @mention badges per channel
+- [x] System tray: minimize to tray, notification popups, badge count
+- [x] Keyboard shortcuts: Ctrl+K quick switcher,
   Escape to close panels, customizable PTT key
-- [ ] Settings: account, appearance (light/dark),
+- [x] Settings: account, appearance (light/dark),
   notifications, audio devices, keybinds
 
 ## Phase 4: Real-Time Chat Features (2–3 weeks)
 
-- [ ] WebSocket client with auto-reconnect, exponential
+- [x] WebSocket client with auto-reconnect, exponential
   backoff, message replay on reconnect
-- [ ] Send/receive messages in real-time, append to scrollback
-- [ ] Message history: paginated from server on channel switch, scroll-to-load-more
-- [ ] Threads, replies (inline preview), reactions (emoji), edit, delete
-- [ ] Typing indicators ("X is typing..." below input)
-- [ ] Online/offline/idle/DnD presence with status icons in member list
-- [ ] File uploads: drag-and-drop or clipboard paste,
+- [x] Send/receive messages in real-time, append to scrollback
+- [x] Message history: paginated from server on channel switch, scroll-to-load-more
+- [x] Threads, replies (inline preview), reactions (emoji), edit, delete
+- [x] Typing indicators ("X is typing..." below input)
+- [x] Online/offline/idle/DnD presence with status icons in member list
+- [x] File uploads: drag-and-drop or clipboard paste,
   progress bar, inline image previews
-- [ ] Client-side file validation before upload (size check, warn on large files)
-- [ ] Search: query server FTS5 endpoint, display results with jump-to-message
-- [ ] Windows toast notifications with action buttons (reply, mark read)
-- [ ] Notification sounds (configurable, per-channel mute/override)
+- [x] Client-side file validation before upload (size check, warn on large files)
+- [x] Search: query server FTS5 endpoint, display results with jump-to-message
+- [x] Windows toast notifications with action buttons (reply, mark read)
+- [x] Notification sounds (configurable, per-channel mute/override)
 
 ## Phase 5: Voice & Video (3–5 weeks)
 
-- [ ] LiveKit integration in native client for voice/video
+- [x] LiveKit integration in native client for voice/video
   (client connects to LiveKit directly using token from server)
-- [ ] Audio device selection: input/output dropdowns in settings, live preview
-- [ ] Voice channels: click to join/leave, show connected users with speaking indicators
-- [ ] Voice controls: mute (button + keybind), deafen, per-user volume sliders
-- [ ] Push-to-talk: configurable global hotkey that works in fullscreen games
-- [ ] Voice activity detection with configurable sensitivity
-- [ ] Noise suppression (RNNoise or equivalent, bundled with client)
-- [ ] Server-side: LiveKit companion process with token-based auth
+- [x] Audio device selection: input/output dropdowns in settings, live preview
+- [x] Voice channels: click to join/leave, show connected users with speaking indicators
+- [x] Voice controls: mute (button + keybind), deafen, per-user volume sliders
+- [x] Push-to-talk: configurable global hotkey that works in fullscreen games
+- [x] Voice activity detection with configurable sensitivity
+- [x] Noise suppression (RNNoise or equivalent, bundled with client)
+- [x] Server-side: LiveKit companion process with token-based auth
   and webhook sync for voice state updates
-- [ ] Voice quality: low (32kbps) / medium (64kbps) / high (128kbps Opus)
+- [x] Voice quality: low (32kbps) / medium (64kbps) / high (128kbps Opus)
 - [ ] Screen sharing via LiveKit screen share track
-- [ ] Video calls: camera capture, displayed in voice channel panel
+- [x] Video calls: camera capture, displayed in voice channel panel
 
 ## Phase 6: Admin Panel — Web-Based (1–2 weeks)
 
-- [ ] Served by server at `/admin`, browser-only access
-- [ ] Auth: admin credentials, session-based
-- [ ] Dashboard: connected users, message count, disk usage, CPU/RAM, uptime
-- [ ] User management: list all, edit roles, ban/unban, reset password, force disconnect
-- [ ] Channel management: create, rename, reorder, set permissions, archive
+- [x] Served by server at `/admin`, browser-only access
+- [x] Auth: admin credentials, session-based
+- [x] Dashboard: connected users, message count, disk usage, CPU/RAM, uptime
+- [x] User management: list all, edit roles, ban/unban, reset password, force disconnect
+- [x] Channel management: create, rename, reorder, set permissions, archive
 - [ ] Invite management: generate, view active, set expiry/use limit, revoke
-- [ ] Server settings: name, icon, MOTD, max upload size, voice quality, TLS config
-- [ ] Moderation: kick, ban, temp ban, slow mode, mute, word filter, audit log
-- [ ] Backup: trigger manual backup, configure
+- [x] Server settings: name, icon, MOTD, max upload size, voice quality, TLS config
+- [x] Moderation: kick, ban, temp ban, slow mode, mute, word filter, audit log
+- [x] Backup: trigger manual backup, configure
   schedule, view/restore from admin panel
-- [ ] Built with simple HTML/CSS/JS embedded in the server binary
+- [x] Built with simple HTML/CSS/JS embedded in the server binary
 
 ## Phase 7: Distribution & Updates (1–2 weeks)
 
-- [ ] **Server:** GitHub Actions builds
+- [x] **Server:** GitHub Actions builds
   `chatserver.exe` (amd64), SHA256, GitHub Release
-- [ ] **Client:** Tauri bundler (NSIS) installer —
+- [x] **Client:** Tauri bundler (NSIS) installer —
   Program Files, Start Menu, auto-start, protocol
   handler for `chatserver://` invite links
 - [ ] Client auto-update: check GitHub releases on
   launch, prompt to download + install
-- [ ] Server update: admin panel shows available update, one-click download + restart
+- [x] Server update: admin panel shows available update, one-click download + restart
 - [ ] Docs: Quick Start, Port Forwarding, Tailscale,
   Client install guide
 - [ ] Security hardening checklist for server operators
@@ -222,9 +224,15 @@ update integrity (SHA256).
 | LiveKit | `livekit/server-sdk-go` (token generation, webhook validation) |
 | SQLite | `modernc.org/sqlite` (pure Go) |
 | Auth | `golang.org/x/crypto/bcrypt` |
-| TOTP | `pquerna/otp` |
 | Sanitization | `bluemonday` |
 | TLS | `golang.org/x/crypto/acme/autocert` |
-| Systray | `getlantern/systray` |
 | Config | `koanf` |
 | Logging | `log/slog` |
+| Versioning | `golang.org/x/mod/semver` |
+| UUID | `google/uuid` |
+
+> **Removed references:**
+> - ~~`getlantern/systray`~~ — The server has no system tray.
+>   System tray is in the Tauri client (Rust-side).
+> - ~~`pquerna/otp`~~ — TOTP 2FA is not yet implemented
+>   (T-023 in backlog). Not in `go.mod`.
diff --git a/docs/brain/06-Specs/PROTOCOL.md b/docs/brain/06-Specs/PROTOCOL.md
index ab38c344..4d729893 100644
--- a/docs/brain/06-Specs/PROTOCOL.md
+++ b/docs/brain/06-Specs/PROTOCOL.md
@@ -327,9 +327,7 @@ and server URL for the client to connect directly to LiveKit.
 ```
 
 Client uses `bitrate` to configure the Opus encoder. Other fields are
-informational for UI. (`threshold_mode` and `top_speakers` fields have
-been removed — LiveKit handles audio mixing and speaker selection
-internally.)
+informational for UI.
 
 ### Voice Control (Client → Server)
 
@@ -348,9 +346,13 @@ internally.)
 Requires `USE_VIDEO` (bit 11) or `SHARE_SCREEN` (bit 12) permission.
 Rate limit: 2/sec per user.
 
-**Note:** Active speaker detection (`voice_speakers`) is no longer a
-server→client WebSocket message. Speaker detection is handled client-side
-via LiveKit SDK events (`ParticipantEvent.IsSpeakingChanged`).
+### Migration Notes (LiveKit transition)
+
+- `threshold_mode` and `top_speakers` fields have been removed from `voice_config`.
+  LiveKit handles audio mixing and speaker selection internally.
+- Active speaker detection (`voice_speakers`) is no longer a server→client
+  WebSocket message. Speaker detection is handled client-side via LiveKit SDK
+  events (`ParticipantEvent.IsSpeakingChanged`).
 
 ---
 
@@ -444,7 +446,12 @@ then auto-reconnect after the delay expires.
         "color": "#F39C12",
         "permissions": 1073741823
       },
-      { "id": 3, "name": "Member", "color": null, "permissions": 1049601 }
+      {
+        "id": 3, "name": "Moderator",
+        "color": "#3498DB",
+        "permissions": 1048575
+      },
+      { "id": 4, "name": "Member", "color": null, "permissions": 7779 }
     ]
   }
 }
@@ -457,7 +464,7 @@ then auto-reconnect after the delay expires.
 Fetched via REST API, not WebSocket, to keep the WS connection lean.
 
 ```text
-GET /api/channels/{id}/messages?before={msg_id}&limit=50
+GET /api/v1/channels/{id}/messages?before={msg_id}&limit=50
 ```
 
 ---
@@ -491,4 +498,8 @@ Error codes: `FORBIDDEN`, `NOT_FOUND`, `RATE_LIMITED`, `INVALID_INPUT`,
 - Voice signaling: 20/sec per user
 - Voice camera/screenshare: 2/sec per user
 
-Server sends `rate_limited` error with `retry_after` in seconds.
+Server sends a standard `error` message with code `RATE_LIMITED` and `retry_after` in seconds:
+
+```json
+{"type": "error", "payload": {"code": "RATE_LIMITED", "message": "...", "retry_after": 5}}
+```