Talon
487a561963
Three reported bugs traced to one root cause plus two missing designed features:
1. Stale users + eternal PLC hiss (root cause): ConnSession::close() silently
erased dropped users without broadcasting UserEvent::LEFT, so peers never
learned the user left and their audio engines never called remove_stream —
Opus PLC synthesized comfort noise forever. Fix: broadcast_left() helper
+ close() broadcasts LEFT before erasing.
2. PLC cap (defense-in-depth): on_playback now caps pure PLC at ~2s, then
emits digital silence so a stale stream can never hiss forever even if
remove_stream is skipped. Resets automatically on fresh packets.
3. No timeout / no ping: client never sent Ping, server had no last_seen /
reaper, so half-open connections (NAT timeout, wifi loss, sleep) left
ghost users forever. Fix: client Ping every 15s with RTT measurement,
ConnSession::last_seen bumped on every inbound TCP/UDP frame, steady_timer
reaper sweeps every 15s and drops sessions older than 45s (configurable
via server::Config).
4. UDP KEEPALIVE: client sends plaintext kFrameKeepalive every 5s; server
bumps last_seen + echoes back. Keeps NAT bindings alive and lets media
activity defer the reaper independently of TCP.
5. Graceful client disconnect: vc_disconnect() sends Disconnect{code=0} via
a flag-based io-thread exit (no double-close race); server handles
client-sent Disconnect with immediate close() + LEFT broadcast.
3 new tests: disconnect_left, plc_cap, reaper_timeout. 21/21 ctest green.
Docs: protocol.md §6/§7, voice.md §6, architecture.md §5, PROGRESS.md.
VoiceCat
Self-hosted, native voice & text chat in the spirit of classic TeamSpeak / Mumble —
channel-based voice, channel + private text, one server you run yourself. Plain TCP
(control) and UDP (media), no WebRTC. Encrypted by default. A shared C++ core
(libvoicecat) drives native clients (Swift on macOS/iOS, C# on Windows) and the server.
Status: pre-implementation. The design is complete in
docs/. The code is an M0 skeleton — it compiles and links, but every subsystem is a stub. SeeAGENTS.mdto start building, anddocs/roadmap.mdfor the milestones.
Read the design first
The docs/ folder is the source of truth. Start at docs/README.md,
then architecture → protocol → voice → security → tech-stack → deployment →
roadmap.
Build the skeleton (no dependencies needed yet)
The M0 skeleton builds with just a C++20 compiler + CMake + Ninja — no vcpkg, no third-party libraries, because every subsystem is currently a stub.
cmake --preset dev
cmake --build --preset dev
ctest --preset dev # runs the smoke test (links the core, calls the C ABI)
Artifacts land in build/dev/bin/ (voicecat-server, vccli).
When you start implementing a subsystem that needs real libraries, build with vcpkg deps:
# one-time: git clone https://github.com/microsoft/vcpkg && ./vcpkg/bootstrap-vcpkg.sh
export VCPKG_ROOT=/path/to/vcpkg # set VCPKG_ROOT (works on Linux/macOS/Windows)
cmake --preset server-release # auto-installs deps from vcpkg.json
cmake --build --preset server-release
Layout
docs/ design spec (read this)
core/ libvoicecat — the shared C++ core
include/ voicecat.h (the C ABI all clients call)
proto/ voicecat.proto (control-plane wire format, source of truth)
src/ net/ crypto/ codec/ protocol/ session/ audio/ (stubs today)
server/ voicecat-server (headless; links the core)
tools/vccli/ headless test client — drives the protocol from M1 on
clients/ apple/ (Swift, M4) windows/ (C#, M4) — placeholders for now
tests/ CTest targets
License
Permissive-only dependencies (no GPL/LGPL) so the project can be redistributed freely,
including closed-source. Project license: TBD (see docs/tech-stack.md §5).