Rill

Self-hosted 1:1 video for your product. Two browsers join a call; Rill copies audio, video, and opaque labeled data between them. Your app owns login, chat schema, and hang-up.

Use it when the call is inside your product (support, consult, interview) and you want to run the SFU yourself. Not a meeting app: two people, then call_full.

docker compose up

Open http://127.0.0.1:3000. Create a call, share the link, enable microphone and camera in two browsers. Chrome and desktop Safari. The SFU on port 7880 is API and signaling, not a homepage (/ is 404).

What you get

• Packet copy. Live media is RTP forwarded (SSRC/PT cache). No transcode on the hot path.
• Two people. Opaque sub + JWT publish: { microphone, camera, screen, data }. No host/guest roles.
• Labeled data. call.send(label, data) / on("data"). Max 128 KiB per message; the app chunks larger payloads. Rill does not parse chat or rrweb.
• Headless SFU. /health, /ready, /v1/*, /v1/signal. No dashboard, no /admin.
• Optional recording. rill-record sidecar writes per-participant MP4 + JSONL. If it is down, the call continues.

Browser A  ←RTP + data→  Rill SFU  ←RTP + data→  Browser B
                              ⋮ lossy fork
                         rill-record → S3

In-memory calls: a restart forgets every call. Create again and mint new tokens.

Run · Your app · Recording · HTTP