Self-hosting

Requirements

To run the image	To build from source
Docker or any OCI runtime	Rust 1.79+
No external services required	No CGO boundary — pure Rust in the media path

Run with Docker

docker run -p 3000:3000 \
  ghcr.io/lumyx-dev/sfu-core:latest

# signaling  ws://127.0.0.1:3000/ws
# dashboard  http://127.0.0.1:3000

Scale it out when you actually need to — a single process handles a room until concurrent peers or egress say otherwise.

Build from source

cargo add lumyx-sfu

# or run the workspace binary
cargo run -p lumyx-sfu --release

Rust 1.79+. The media path has no GC and no CGO boundary, so a release build behaves the same on your hardware as it does in the published image.

Configuration

Thresholds, alert destinations and remediation live in lumyx.toml next to the binary. Nothing in it is required to start — every metric below already has a sane default.

# Thresholds are yours. These are the defaults.

[alerts.packet_loss]
above = 0.02
sustained_for = "30s"
severity = "breached"
notify = ["slack:#oncall", "pagerduty"]

[remediation]
auto_audio_only = true

Full field reference: Metrics reference.

Production checklist

TLS termination

Put a reverse proxy in front for wss:// and https:// — the binary speaks plain ws/http.

Persistent storage

In-memory retention only survives the process. Add Postgres or object storage to keep history past that.

Prometheus scrape

Point your existing Prometheus at :3000/metrics — see the Prometheus endpoint reference.

Alert destinations

Wire Slack, PagerDuty or a webhook before traffic arrives, not after the first threshold breach.