WeaveSignal/weavesignal
Watch
2
0
Fork
You've already forked weavesignal
 0
Code Issues 17 Pull requests Projects Releases 1 Packages Wiki Activity Actions
1 release 1 tag
  • v1.0.0 466b32ab6a

    WeaveSignal 1.0.0 — release one Stable

    weave released this 2026-11-04 12:15:42 +00:00 | 5 commits to main since this release

    1.0.0 — release one (2026-11-04)

    The first public release. What each part does is on the documentation site, https://weavesignal.top;
    what is built and what is not is on its page What is built, and what is planned.

    In this release

    • The live room and the desk. Presenters, invited guests and callers in one LiveKit room; the
      desk mixes the programme and a mix-minus for each person in one pass per 20 ms block, with a gate
      per role, talkback, cue, a limiter on every mix and the profanity delay with Dump.
    • Call-in and screening. The player page's call-in button, consent to record before any token,
      a proof-of-work challenge instead of a CAPTCHA, a booth room for each caller, the screener's lanes,
      Banned before and Desk busy, TURN over TLS for networks that block UDP.
    • The studio app for Linux (AppImage, .deb, .rpm) and Windows (an unsigned installer), with desk B,
      the screener, the cart wall, the schedule, settings, staff accounts and signed updates.
    • Playout through Liquidsoap 2.5: automation from clocks with rotation and separation, the US
      statutory rule, a dead-air fallback, a default loop, the per-play log, listener figures, the legal
      logger, now playing, a loudness control towards −16 LUFS, and an MP3 compatibility stream.
    • Video: cameras, a cut-only vision mixer, the slate with a still per show, lower thirds, captions,
      AV1 into CMAF HLS behind the delay, and the audio-only toggle.
    • The live chat with light accounts, moderation with statements of reasons, retention and a minimum
      age by jurisdiction.
    • Shows, cards and feeds: show cards in two layouts, seasons that fill in their own episodes,
      podcast feeds with an Opus alternate.
    • Delivery at scale: content steering, an HLS edge and an audio relay image, Bunny in the guide.
    • The public API: scoped tokens, open read routes, signed webhooks and an OpenAPI description.
    • Packages: the Compose bundle with weavesignal.sh (and --public with a certificate),
      Quadlets, and a Cloudron app.

    Known limits

    • Run a station on a server of its own. On a heavily shared server the listener feed's cushion
      filled from 3 s to its 10 s cap; on an idle 4-vCPU server 0.5 s held for 13 hours unattended.
    • WebRTC from the desk has no congestion control: the desk's image lacks GStreamer's rtpgccbwe.
    • Not run end to end yet: a 24-hour unattended run from a seeded week; two real people over two real
      networks (one person on two networks was run); an off-site edge in front of a station; Safari's Opus
      in fMP4 on a device; the GPU in the Compose bundle; phones on GrapheneOS over a real network.
    • Firefox for Android can call in only once its microphone is allowed in Android's settings.
    • LiveKit's TURN refuses clients that reach it over IPv6: give the TURN name an A record only.
    • podman-compose is refused by weavesignal.sh: it ignores the bundle's settings files. Use Docker
      Compose, or Podman with the docker-compose provider.
    • Not in release one: the presenter kit, the Windows server, WHIP for remote DJs and the DJ overlap
      handover.
    Downloads