
Receives a browser's camera and microphone capture over WebRTC and emits a VideoFrame stream and an AudioFrame stream, each gated on independently, so the graph can take live media from a remote browser tab.
Browser webcam object detection streamed back to the browser.
Voice assistant driven by the browser microphone.
with_video and with_audio gate the two outputs independently, and any enabled output that is left unconsumed back-pressures the component, so wire every output whose gate is on.fps and frame_size are negotiation hints the browser may silently downgrade to what the device supports, and permission denial yields no frames without an error.audio_frame_ms at the sentinel forwards each underlying WebRTC audio callback as its own frame, while a positive value accumulates audio into fixed-size chunks; drop_frames discards frames under back-pressure instead of blocking, which matters under network jitter.webrtc_port_range pins the local UDP ports used for connection setup; leaving it unset lets the system choose, which can fail behind restrictive firewalls.drop_frames takes effect at runtime; the other knobs apply on the next browser session.Code-quality + perf hardening: SDP reparse null-guard, ICE null-guard, APM, read_maybe abstraction