+-----------------------------------------------------------------------------------------+
|                  WebRTC Topologies: Mesh vs MCU vs LiveKit SFU                          |
+-----------------------------------------------------------------------------------------+

PEER-TO-PEER MESH (Crushes Mobile CPU & Bandwidth):
[Client A] <====(Uploads 5 streams!)====> [Client B, C, D, E, F]
* Network Upload: N-1 Streams! Battery dies in 15 minutes!

MULTIPOINT CONTROL UNIT (MCU - Heavy Server Transcoding):
[Clients] ---> [CENTRAL SERVER: Decodes, Mixes & Re-Encodes Video into 1 Grid] ---> [Clients]
* Server CPU: Enormous! High cloud infrastructure cost and +150ms transcoding lag!

SELECTIVE FORWARDING UNIT (LiveKit SFU - 2026 Industry Standard):
[Clients Upload ONLY 1 Stream] ---> [LIVEKIT SFU (Go Media Router)] ---> [Forwarded Selectively!]
* Client Upload: 1 Single Stream! Server CPU: Zero Transcoding! Sub-50ms Routing Latency!

+-----------------------------------------------------------------------------------------+
|                  Adaptive Bitrate Simulcast Video Publishing Flow                       |
+-----------------------------------------------------------------------------------------+

[MOBILE BROADCASTER (1080p Camera Feed)]
                 |
                 +---> [Layer 1 (High): 1080p @ 2.5 Mbps (Sent to Desktop Fullscreen Viewers)]
                 +---> [Layer 2 (Med):   720p @ 1.0 Mbps (Sent to iPad / Tablet Viewers)]
                 +---> [Layer 3 (Low):   360p @ 300 Kbps (Sent to Weak 3G Mobile Viewers)]
                                   |
                                   v
[LIVEKIT SFU MEDIA ROUTER: Inspects each subscriber's downlink network bandwidth]
                                   |
        +--------------------------+--------------------------+
        | (Viewer on Gigabit Wi-Fi)|                          | (Viewer on Subway 3G Cell)
        v                          v                          v
[Delivers 1080p Stream]   [Delivers 720p Stream]     [Delivers 360p (Zero Buffering!)]

// lib/screens/video_room_screen.dart
import 'package:flutter/material.dart';
import 'package:livekit_client/livekit_client.dart';

class VideoRoomScreen extends StatefulWidget {
  final String roomUrl;
  final String token;

  const VideoRoomScreen({super.key, required this.roomUrl, required this.token});

  @override
  State<VideoRoomScreen> createState() => _VideoRoomScreenState();
}

class _VideoRoomScreenState extends State<VideoRoomScreen> {
  Room? _room;
  EventsListener<RoomEvent>? _listener;

  @override
  void initState() {
    super.initState();
    _connectToRoom();
  }

  Future<void> _connectToRoom() async {
    // 1. Configure Hardware-Accelerated Audio Options with Noise Cancellation
    final roomOptions = RoomOptions(
      adaptiveStream: true, // Enables adaptive client-side resolution throttling!
      dynacast: true,       // Pauses unused simulcast layers to save publisher upload!
      defaultAudioCaptureOptions: const AudioCaptureOptions(
        noiseSuppression: true,
        echoCancellation: true,
        autoGainControl: true,
      ),
      defaultVideoCaptureOptions: const VideoCaptureOptions(
        params: VideoParametersPresets.h720_169,
      ),
    );

    final room = Room(roomOptions: roomOptions);
    _listener = room.createListener();

    // 2. Connect to LiveKit SFU Server
    await room.connect(widget.roomUrl, widget.token);

    // 3. Publish Local Camera and Microphone Tracks
    await room.localParticipant?.setCameraEnabled(true);
    await room.localParticipant?.setMicrophoneEnabled(true);

    _listener?.on<RoomEvent>((event) {
      if (mounted) setState(() {});
    });

    setState(() {
      _room = room;
    });
  }

  @override
  void dispose() {
    _listener?.dispose();
    _room?.disconnect();
    _room?.dispose();
    super.dispose();
  }

  @override
  Widget build(BuildContext context) {
    if (_room == null || _room!.connectionState != ConnectionState.connected) {
      return const Scaffold(body: Center(child: CircularProgressIndicator()));
    }

    final remoteParticipants = _room!.remoteParticipants.values.toList();

    return Scaffold(
      backgroundColor: Colors.black,
      body: SafeArea(
        child: GridView.builder(
          gridDelegate: const SliverGridDelegateWithFixedCrossAxisCount(crossAxisCount: 2),
          itemCount: remoteParticipants.length,
          itemBuilder: (context, index) {
            final participant = remoteParticipants[index];
            final videoTrack = participant.videoTrackPublications.firstOrNull?.track;

            if (videoTrack != null && videoTrack is VideoTrack) {
              return VideoTrackRenderer(videoTrack);
            }
            return Center(child: Text(participant.identity, style: const TextStyle(color: Colors.white)));
          },
        ),
      ),
    );
  }
}

+-----------------------------------------------------------------------------------------+
|                  Real-Time Voice AI Agent Streaming Architecture                        |
+-----------------------------------------------------------------------------------------+

[MOBILE USER (Speaks into Microphone)]
                 |
                 v (Sub-50ms Opus Audio Stream over WebRTC)
[LIVEKIT SFU]
                 |
                 v (Piped to Python / Node.js LiveKit AI Worker)
+-----------------------------------------------------------------+
| LIVEKIT VOICE AGENT PIPELINE:                                   |
| 1. Krisp / Silero VAD (Voice Activity Detection in 15ms).       |
| 2. Deepgram Nova-3 (Streaming Speech-to-Text in 120ms).         |
| 3. GPT-4o / Claude 3.7 (First Token Output in 180ms).           |
| 4. Cartesia / ElevenLabs (Streaming Text-to-Speech in 100ms).   |
+--------------------------------+--------------------------------+
                                 |
                                 v (Synthesized Voice Audio Streamed back to WebRTC Room!)
[User hears intelligent AI voice response in < 450ms! Natural conversational turn-taking!]

       +-------------------------------------------------------------+
       |             End-to-End Glass-to-Glass Latency (ms)          |
       +-------------------------------------------------------------+
 Standard HLS / Low-Latency HLS Video | ==================================== [2,800.0 ms]
 LiveKit SFU WebRTC (Cross-Continent) | = [74.0 ms] (37x Faster Interactive Speed!)
                                      +-------------------------------------+
                                      0ms     700ms   1400ms  2100ms  2800ms