Root cause: degradationPreference 'maintain-resolution' forced FPS drops under CPU
pressure during fast-motion gaming. Changed to 'maintain-framerate' so the encoder
drops resolution temporarily instead of stuttering. Added contentHint='motion' to
optimize for temporal prediction (more P-frames, fewer I-frames).
Screen sharing was publishing at h360 (640x360) and never ramping to target
resolution due to stale closure in setTimeout, wrong initial quality anchor,
and 5 competing code paths with inconsistent bitrates.
- Create utils/screenShare.ts as single source of truth for all screen share ops
- Publish at target resolution from the start (not h360 → ramp)
- Read store at call time in timers (eliminates stale closure bug)
- Use maintain-resolution for screen content, maintain-framerate for camera
- Fix OS-level "Stop sharing" not resetting store or restoring AEC
- Enable dynacast for SFU quality signaling
- Reconcile QUALITY_MAP to canonical bitrates across all 7 files
LiveKit's server-side VAD has a high, non-configurable threshold that misses
conversational speech. Replace it with a SpeakingDetector singleton that uses
Web Audio AnalyserNodes to read actual RMS levels per participant (50ms poll,
0.008 threshold, 250ms hysteresis hold).
Eliminate double-state architecture (useState → useEffect bridge → store)
that lost speaking events due to React 18 batching. ActiveSpeakersChanged
now writes speakingParticipantIds directly to voiceStore; 200ms poll
catches missed SDK events. Each VoiceUser subscribes to its own identity
via fine-grained selector for minimal re-renders. Also adds connection
quality indicator and ConnectionInfoPopover.
tsc was emitting compiled .js files directly into packages/web/src/
alongside the .tsx source files because noEmit was not set. These
artifacts were never used — Vite compiles from .tsx source directly.
- Add noEmit: true to packages/web/tsconfig.json (tsc = type-check only)
- Delete all 62 orphaned .js files from src/ (-6,129 lines)
- Add packages/web/src/**/*.js to .gitignore as safeguard
- RNNoise (AI Noise Suppression) now enabled by default for all users
- Remove redundant browser Noise Suppression toggle from VoiceControls
and UserSettings — AudioManager handles it automatically as fallback
- Add AI Noise Suppression toggle to UserSettings panel
- Rename toggleRnnoise → setRnnoiseEnabled for clearer API
- Store migration v3→v4: enable RNNoise for existing users
- Keep Echo Cancellation and Auto Gain Control (orthogonal features)
RNNoise worklet outputs mono (1ch) which played only in the left ear.
Replace unreliable automatic up-mixing (channelCount/channelCountMode on
inputGain) with a ChannelMergerNode that duplicates the mono signal to
both L and R channels — guaranteed stereo by the Web Audio spec.
Inject @sapphi-red/web-noise-suppressor into AudioManager's input pipeline
as a toggleable AudioWorkletNode. The worklet is loaded lazily on first
enable, then kept alive — toggling bypasses by rewiring the graph without
destroying the WASM instance. Browser NS is forced off when RNNoise is
active to avoid double-processing. InputGain forced to stereo up-mix to
prevent mono-left-only output from the worklet. Also wires Phase 2 output
device routing (setSinkId) into AppLayout/ChannelSidebar.
- syncMic now uses setMicrophoneEnabled(false/true) to mute/unmute the mic
track in-place instead of tearing down and re-publishing via unpublishTrack().
This eliminates WebRTC renegotiation on every mute cycle and preserves the
Web Audio pipeline for future AudioWorklet injection (RNNoise WASM).
- Unify DmCallView mute/deafen with the primary syncMic path — removed direct
setMicrophoneEnabled calls, added missing auto-mute/unmute on deafen toggle.
- Strip redundant applyConstraints from VoiceControls noise suppression toggle
that was hardcoding echoCancellation:true/autoGainControl:true and fighting
AudioManager's constraint pipeline.
- Redesign DM close as a visibility flag (closed column) instead of row deletion,
so message broadcasts still reach users who closed a DM conversation
- Add broadcastDmMessage() helper that auto-resurfaces closed DMs when new messages arrive
- Wire Remove Friend button in DM welcome header with onClick handler and friend check
- Fix deploy.sh to cd to its own directory so rsync always runs from project root
- Add WS events: dm_channel_created, dm_channel_closed, friend_removed,
channel_created/updated/deleted, server_updated
- Fix first-ever DM: broadcast dm_channel_created to recipient
- Add DELETE /api/dm/:id for closing DMs with re-open support
- Wire Close DM button in sidebar
- Fix dm_message_created for unknown channels (safety net)
- Broadcast friend_removed on friend deletion
- Sound system: track realtimeMessageEvents separately from API loads,
play notification for messages in all channels, not just current
- Optimistic updates: send/edit/delete messages appear instantly with
rollback on failure, temp message deduplication on WS echo
- Channel CRUD broadcasts to all server members
- Server update broadcast on PATCH
- Server join registers user in connectionManager immediately
- Reconnect: only reload current channel, preserve other channel caches
- Activity panel, right panel, member list toggle components
Chrome's AEC uses getDisplayMedia audio as a reference signal and
aggressively ducks the microphone even when headphones are used.
This adds a screenShareActive flag to AudioManager that forces
echoCancellation off during screen share, plus Chromium-specific
goog* constraints as belt-and-suspenders.
Disable AGC by default to prevent Chrome from crushing mic sensitivity
when stream audio is playing. Add user-facing toggles for echo
cancellation, noise suppression, and auto gain control. Decouple voice
and stream audio by routing through ctx.destination instead of shared
compressor. Track mic stream generation to re-publish when settings change.
Add MutationObserver to neutralize LiveKit's re-attached <audio> elements,
mark our keep-alive elements with data-opencord, detach tracks on unsubscribe,
remove dangerous blanket .play(), and soften compressor to transparent limiter.
Show local user's own screen share in their stream tile by bypassing
the isSubscribed check for local participants and auto-watching/unwatching
local streams on publish/unpublish. Eliminate audio double-playback by
detaching LiveKit's auto-attached audio elements so GlobalAudioRenderer
is the sole playback path, making attenuation and volume controls effective.
Fix GainNode leak in useAudioTrackPlayer cleanup paths.
Streams no longer auto-focus into large view when they start — tiles
stay in the equal-size grid until manually clicked. Audio playback is
moved out of VoiceUser/StreamTile into a new GlobalAudioRenderer
component rendered in AppLayout so it survives channel navigation.
Streams now appear as separate tiles in the voice grid alongside the
user's camera/avatar tile, matching Discord's model. Each stream tile
has independent volume, mute, watch/unwatch controls, quality badges,
and stream attenuation that ducks audio when someone speaks.
Pass audio: true to setScreenShareEnabled so the browser offers the
"Share audio" checkbox. Track ScreenShareAudio from remote participants
and play it through a dedicated audio element with the same volume
pipeline (including boost >100%).
- Add WebSocket voice_status/voice_status_update events so mute/deafen
icons are visible in the sidebar without joining the voice channel
- Server tracks voiceUserStates and includes them in the ready payload
- Re-register voice channel on WebSocket reconnect to prevent sidebar
users from disappearing after idle timeout
- Re-broadcast deafen state to late joiners via LiveKit data channel
- Fix black grid tile when video stops (enabled-flag guards)
- Remove duplicate mute/deafen from VoiceControls (replaced with
Video Quality + Noise Suppression)
- Fix missing users in sidebar voice list (identity matching + fallback)
Restructure VoiceControls and UserAreaPanel into a single fixed-positioned
floating card that spans both server and channel sidebars. Panel expands
when voice is connected and contracts to just user area when not.
Shift entire color palette from Discord's Ash theme to the darker Dark theme.
Redesign VoiceControlBar as floating auto-show pill, add gradient Join Voice
screen, move invite icon to server header, add voice channel user status
badges, update auth pages with subtle gradient backgrounds, and sweep all
hardcoded hex colors across 14 files to match the new darker palette.
- New PictureInPicture component: draggable floating window that shows
active stream when navigating away from voice channel, with LIVE badge,
snap-to-edge, click-to-navigate-back
- Fix VoiceUser to check actual track presence instead of boolean flags
for video rendering
- Add LocalTrackPublished/LocalTrackUnpublished events to useLiveKit
so local screen share and camera show immediately in VoiceGrid
- Add pipCollapsed state to uiStore
Restores the working smooth resolution ramp-up pipeline that starts at 360p
to clear the WebRTC handshake, then upgrades to target quality after 2 seconds.
- Implemented 'Overdrive' logic to force high bitrates on Chrome
- Fixed 'Auto' preset to default to stable 720p60
- Added persistent 'Triple-Kick' hammer to prevent bitrate throttling
- Fixed sidebar connection status sync
- Added comprehensive diagnostic logger
- Fix stale message cache: add force param to loadMessages, clearAllMessages action
- Fix reload race condition: URL-based isDmChannel fallback before WS ready
- Add read_states DB table for persistent unread tracking
- Add channel_ack WS event (client→server→echo) with BigInt comparison
- Wire up unread state in chatStore (readStates, unreadChannels, ackChannel)
- Auto-ack channels on MessageList view (200ms debounced)
- Unread pill indicators on server icons in ServerSidebar
- Bold text + white dot on unread channels/DMs in ChannelSidebar
- Replace all showDms reads with data-driven isDmChannel() across 8 files
- Design system, UI polish, and component fixes from previous sessions
- Guard Disconnected/ConnectionStateChanged event handlers against stale rooms:
old room events no longer nuke the new room's state (root cause of buttons
failing, mute getting stuck, DUPLICATE_IDENTITY cascades)
- Reset media state (isMuted/isCameraOn/isScreenSharing) on connect to prevent
desync after reconnects
- Add voiceStates to WS ready payload so users see who's in voice on page load
- Wire VoiceControls buttons to check getActiveRoom() before SDK calls
- Guard ChannelSidebar against re-joining the same voice channel
- Switch LIVEKIT_URL to wss://nova.ddns.net/livekit for HTTPS secure context
(required for getUserMedia in Safari)
Phase 1 - Feature Repair:
- Fix member kick/leave: add missing db.delete() call in servers.ts
- Stabilize invite codes: return existing code instead of regenerating
- Fix user search: use LIKE instead of exact match in social.ts
- Wire DM button on FriendsPage to create/navigate to DM channels
- Add cancel outgoing friend request (DELETE endpoint + frontend)
- Add accept/decline friend request actions with WS real-time events
- Fix replyToId persistence in message creation
- Hydrate reactions and replyTo in message queries
- Add joinByCode to API client and serverStore
- Add friend_request_received/accepted WebSocket events
Phase 2 - Discord UI Overhaul:
- Remove stray borders between layout columns
- Replace shadow-sm with shadow-header on content headers
- Replace all bg-gray-*/text-gray-* with Discord color tokens
- Ensure flat color contrast (#1E1F22, #2B2D31, #313338)
Testing:
- Set up vitest + @testing-library/react + jsdom
- Add 17 tests across InviteModal, JoinServer, FriendsPage (all passing)
- Fix vite resolve.extensions to prefer .tsx over stale .js files
Jannis Braun