molluskandClaude Opus 4.8
1bf14ba08f
The multitrack recorder wrote every per-stem WAV frame (and the potentially
large late-joiner back-pad) inline on the caller thread while holding the
recorder mutex, so a slow/contended disk stalled the playout mixer (local
underruns) and the events loop. This is the multitrack counterpart to A17
(e0325d4), which moved the single-file recorder's writes off the mixer path.
Design: the front (MultitrackRecorder) now keeps only cheap in-memory state
(known-peer set, mic FIFO, a pending-cycle builder) and on each end_cycle
assembles ONE whole-cycle batch (new peers + mic frame + optional mix frame +
the map of peer frames written this cycle) and try_sends it over a bounded
sync_channel(256) to a dedicated writer thread. The writer thread owns every
WavWriter, is authoritative for its own cycle count, back-pads a brand-new
peer by cycles_written*frame_samples, fills absent peer/mix frames with
silence, latches the first write/create error then drains, and finalizes all
headers on channel close.
The unit of hand-off is a whole cycle, not a track: the writer appends exactly
frame_samples to every existing track per applied batch, and a full queue
DROPS the entire batch (counted + logged at 1 and every 256). So a dropped
cycle omits the same 20ms from every stem at once and all tracks stay
equal-length and sample-aligned by construction even under disk back-pressure.
On drop the batch's new-peer announcements are rolled back out of the known set
so they re-announce (and correctly re-back-pad) on the next applied cycle.
Public method signatures are unchanged -> zero core/mod.rs edits. The
WAV/file format is unchanged (no wire/on-disk change), no new deps
(std::sync::mpsc + std::thread, as A17). Writer logic is factored behind a
generic SampleWriter seam so the apply-batch alignment invariant is unit-tested
without spawning the thread; new tests cover the back-pad-on-apply invariant,
the dropped-cycle equal-length property, and async create-error surfacing at
finalize. The three existing end-to-end tests pass unchanged (now exercising
the threaded path). 496 lib tests, clippy --all-targets clean, release builds.
Codex-implemented (gpt-5.5 xhigh), senior-reviewed.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
PeerSpeak
Decentralized, peer-to-peer voice chat — full-mesh, NAT-traversing, with no central server. Built in Rust on iroh (QUIC), PipeWire audio, the Opus codec, and an iced GUI.
Create a room, share the join ticket, and talk. Everyone connects directly to everyone else; relays are only used to punch through NATs when a direct path isn't available.
Screenshots
| Launch screen | In a room |
|---|---|
 |
 |
| Settings |
|---|
 |
Features
Rooms & sessions
- Create a room → shareable join ticket; join by pasting a ticket.
- Full-mesh multi-peer rooms with live presence.
- Recent-rooms list to hop back into a room someone's still in.
- Remembered nickname and in-call duration timer.
Audio
- PipeWire capture/playback, selectable input and output devices, per-app gain.
- Opus codec (48 kHz mono, 20 ms frames) with an adaptive jitter buffer + packet-loss concealment.
- Noise gate with a draggable threshold on a live mic meter (test your mic off-call too).
- Mix-bus soft limiter and opt-in echo cancellation (PipeWire WebRTC AEC + noise suppression).
Voice controls
- Self-mute, deafen, and rebindable push-to-talk.
- Per-peer volume, local mute, and speaking indicators.
Text chat
- In-room text chat over the gossip plane, with clickable links and inline image/audio attachments.
- Drag-selectable, copyable messages; right-click context menu on all text fields.
Shared music listening
- Build a personal playlist of local audio files with a full transport (play/pause, seek, reorder).
- Let others tune in: peers stream your current track, timeline-synced and gapless, sitting under voice at their own volume.
Screen share (via pixelpass)
- Share your screen; peers click 👁 Watch to open the stream in mpv (vlc fallback).
- Live badges on sharing peers; per-app audio capture.
Recording & notifications
- Local call recording (mic + incoming mix → WAV in
~/peerspeak-recordings/). - Desktop notifications and event chimes with per-event custom sound overrides.
UI & networking
- Selectable room layouts (3-Column, Bottom Dock, Drawer) with draggable, persisted dividers.
- 10 built-in themes (Catppuccin, Dracula, Nord, Tokyo Night, Gruvbox, Solarized…), all WCAG-AA checked.
- Network mode picker (relay-no-discovery default, full n0, or direct-only); retained-address reconnect.
- Config, window size/position, and all preferences persisted to
~/.config/peerspeak/.
See docs/FEATURES.md for the full inventory and field-test status, and docs/ARCHITECTURE.md for internals.
Roadmap
- Contacts & invites — friends list with invite-notification one-click join (design in
docs/contacts-plan.md). - Spatial audio & per-peer EQ.
- Soundboard — play short clips into the call mix.
- Room persistence / invite links beyond the raw ticket.
- Windows support — cross-compiles and launches under Wine today; needs a real WASAPI audio pass (see
docs/WINDOWS.md).
Building
PeerSpeak builds with a stable Rust toolchain (edition 2024). Install the system dependencies below, then:
cargo build --release
./target/release/peerspeak
System dependencies
Arch Linux
sudo pacman -S --needed rust pipewire opus pkgconf git
Debian / Ubuntu
sudo apt install build-essential pkg-config clang libclang-dev \
libpipewire-0.3-dev libopus-dev libasound2-dev libxcb1-dev
Plus a Rust toolchain via rustup. clang/libclang are needed for the PipeWire bindings (bindgen).
At runtime you need a running PipeWire server. Screen sharing additionally requires pixelpass on your PATH, and mpv (or vlc) to watch a peer's share.
Packaging
- Arch:
cd packaging && makepkg -si(usespackaging/PKGBUILD). - Debian/Ubuntu:
.debis built withcargo-debfrom the[package.metadata.deb]block inCargo.toml. Build inside a Debian/Ubuntu environment so the binary links that distro's glibc. - Windows: see
docs/WINDOWS.md.
License
PeerSpeak is licensed under the MIT License, © 2026 mollusk.
Third-party components bundled with PeerSpeak (the Rust dependency tree, the
statically bundled Opus codec on some builds, and embedded fonts) are all under
permissive licenses; their texts and a full dependency manifest are collected in
THIRD_PARTY_LICENSES.