|
| 1 | +import Collections |
| 2 | +import Foundation |
| 3 | +import LiveKit |
| 4 | + |
| 5 | +@MainActor |
| 6 | +final class AgentSession: ObservableObject { |
| 7 | + // MARK: - State |
| 8 | + |
| 9 | + @Published private(set) var error: Error? |
| 10 | + |
| 11 | + @Published private(set) var agent: Agent? |
| 12 | + |
| 13 | + @Published private(set) var connectionState: ConnectionState = .disconnected |
| 14 | + @Published private(set) var isListening = false |
| 15 | + var isAvailable: Bool { |
| 16 | + switch connectionState { |
| 17 | + case .disconnected where isListening, |
| 18 | + .connecting where isListening, |
| 19 | + .connected, |
| 20 | + .reconnecting: |
| 21 | + true |
| 22 | + default: |
| 23 | + false |
| 24 | + } |
| 25 | + } |
| 26 | + |
| 27 | + @Published private(set) var localAudioTrack: (any AudioTrack)? |
| 28 | + @Published private(set) var localCameraTrack: (any VideoTrack)? |
| 29 | + @Published private(set) var localScreenShareTrack: (any VideoTrack)? |
| 30 | + |
| 31 | + // TODO: Move camera switching here (vs Devices)? |
| 32 | + |
| 33 | + var isMicrophoneEnabled: Bool { localAudioTrack != nil } |
| 34 | + var isCameraEnabled: Bool { localCameraTrack != nil } |
| 35 | + var isScreenShareEnabled: Bool { localScreenShareTrack != nil } |
| 36 | + |
| 37 | + @Published private(set) var agentAudioTrack: (any AudioTrack)? |
| 38 | + @Published private(set) var avatarCameraTrack: (any VideoTrack)? |
| 39 | + |
| 40 | + @Published private(set) var messages: OrderedDictionary<ReceivedMessage.ID, ReceivedMessage> = [:] |
| 41 | + |
| 42 | + var supportedFeatures: Features { features } |
| 43 | + |
| 44 | + // MARK: - Dependencies |
| 45 | + |
| 46 | + private let environment: Environment |
| 47 | + private let room: Room |
| 48 | + private let features: Features |
| 49 | + private let senders: [any MessageSender] |
| 50 | + private let receivers: [any MessageReceiver] |
| 51 | + |
| 52 | + // MARK: - Internal state |
| 53 | + |
| 54 | + private var waitForAgentTask: Task<Void, Swift.Error>? |
| 55 | + |
| 56 | + // MARK: - Init |
| 57 | + |
| 58 | + init(environment: Environment, context: Context = .init(), senders: [any MessageSender]? = nil, receivers: [any MessageReceiver]? = nil) { |
| 59 | + self.environment = environment |
| 60 | + room = context.room |
| 61 | + features = context.features |
| 62 | + |
| 63 | + let textMessageSender = TextMessageSender(room: room) |
| 64 | + self.senders = senders ?? [textMessageSender] |
| 65 | + self.receivers = receivers ?? [textMessageSender, TranscriptionStreamReceiver(room: room)] |
| 66 | + |
| 67 | + observeRoom() |
| 68 | + observeReceivers() |
| 69 | + } |
| 70 | + |
| 71 | + private func observeRoom() { |
| 72 | + Task { [weak self] in |
| 73 | + guard let changes = self?.room.changes else { return } |
| 74 | + for await _ in changes { |
| 75 | + guard let self else { return } |
| 76 | + |
| 77 | + connectionState = room.connectionState |
| 78 | + agent = room.agentParticipant |
| 79 | + |
| 80 | + localAudioTrack = room.localParticipant.firstAudioTrack |
| 81 | + localCameraTrack = room.localParticipant.firstCameraVideoTrack |
| 82 | + localScreenShareTrack = room.localParticipant.firstScreenShareVideoTrack |
| 83 | + |
| 84 | + agentAudioTrack = room.agentParticipant?.audioTracks.first(where: { $0.source == .microphone })?.track as? AudioTrack // remove bg audio tracks |
| 85 | + avatarCameraTrack = room.agentParticipant?.avatarWorker?.firstCameraVideoTrack |
| 86 | + } |
| 87 | + } |
| 88 | + } |
| 89 | + |
| 90 | + private func observeReceivers() { |
| 91 | + for receiver in receivers { |
| 92 | + Task { [weak self] in |
| 93 | + for await message in try await receiver.messages() { |
| 94 | + guard let self else { return } |
| 95 | + messages.updateValue(message, forKey: message.id) |
| 96 | + } |
| 97 | + } |
| 98 | + } |
| 99 | + } |
| 100 | + |
| 101 | + // MARK: - Public |
| 102 | + |
| 103 | + func connect(options: ConnectOptions? = nil, roomOptions: RoomOptions? = nil, preConnectAudio: Bool = true, waitForAgent: TimeInterval = 20) async { |
| 104 | + error = nil |
| 105 | + waitForAgentTask?.cancel() |
| 106 | + |
| 107 | + defer { |
| 108 | + waitForAgentTask = Task { |
| 109 | + try await Task.sleep(for: .seconds(waitForAgent)) |
| 110 | + try Task.checkCancellation() |
| 111 | + if connectionState == .connected, agent == nil { |
| 112 | + await disconnect() |
| 113 | + self.error = .agentNotConnected |
| 114 | + } |
| 115 | + } |
| 116 | + } |
| 117 | + |
| 118 | + let connection = { @Sendable in |
| 119 | + let (server, token) = try! await self.credentials() |
| 120 | + try await self.room.connect(url: server, token: token, connectOptions: options, roomOptions: roomOptions) |
| 121 | + } |
| 122 | + |
| 123 | + do { |
| 124 | + if preConnectAudio { |
| 125 | + try await room.withPreConnectAudio { try await connection() } |
| 126 | + } else { |
| 127 | + try await connection() |
| 128 | + } |
| 129 | + } catch { |
| 130 | + self.error = .failedToConnect(error) |
| 131 | + } |
| 132 | + } |
| 133 | + |
| 134 | + func disconnect() async { |
| 135 | + await room.disconnect() |
| 136 | + } |
| 137 | + |
| 138 | + func resetError() { |
| 139 | + error = nil |
| 140 | + } |
| 141 | + |
| 142 | + func send(message: SentMessage) async { |
| 143 | + do { |
| 144 | + for sender in senders { |
| 145 | + try await sender.send(message) |
| 146 | + } |
| 147 | + } catch { |
| 148 | + self.error = .failedToSend(error) |
| 149 | + } |
| 150 | + } |
| 151 | + |
| 152 | + func getMessageHistory() -> [ReceivedMessage] { |
| 153 | + messages.values.elements |
| 154 | + } |
| 155 | + |
| 156 | + func restoreMessageHistory(_ messages: [ReceivedMessage]) { |
| 157 | + self.messages = .init(uniqueKeysWithValues: messages.sorted(by: { $0.timestamp < $1.timestamp }).map { ($0.id, $0) }) |
| 158 | + } |
| 159 | + |
| 160 | + func toggleMicrophone() async { |
| 161 | + do { |
| 162 | + try await room.localParticipant.setMicrophone(enabled: !isMicrophoneEnabled) |
| 163 | + } catch { |
| 164 | + self.error = .mediaDevice(error) |
| 165 | + } |
| 166 | + } |
| 167 | + |
| 168 | + func toggleCamera() async { |
| 169 | + let enable = !isCameraEnabled |
| 170 | + do { |
| 171 | + // One video track at a time |
| 172 | + if enable, isScreenShareEnabled { |
| 173 | + try await room.localParticipant.setScreenShare(enabled: false) |
| 174 | + } |
| 175 | + |
| 176 | + // Hm??? |
| 177 | +// let device = try await CameraCapturer.captureDevices().first(where: { $0.uniqueID == selectedVideoDeviceID }) |
| 178 | + try await room.localParticipant.setCamera(enabled: enable) // captureOptions: CameraCaptureOptions(device: device)) |
| 179 | + } catch { |
| 180 | + self.error = .mediaDevice(error) |
| 181 | + } |
| 182 | + } |
| 183 | + |
| 184 | + func toggleScreenShare() async { |
| 185 | + let enable = !isScreenShareEnabled |
| 186 | + do { |
| 187 | + // One video track at a time |
| 188 | + if enable, isCameraEnabled { |
| 189 | + try await room.localParticipant.setCamera(enabled: false) |
| 190 | + } |
| 191 | + try await room.localParticipant.setScreenShare(enabled: enable) |
| 192 | + } catch { |
| 193 | + self.error = .mediaDevice(error) |
| 194 | + } |
| 195 | + } |
| 196 | + |
| 197 | + // MARK: - Private |
| 198 | + |
| 199 | + private func credentials() async throws -> (server: String, token: String) { |
| 200 | + switch environment { |
| 201 | + case let .sandbox(id, room, participant): |
| 202 | + let sandboxConnection = try await Sandbox.getConnection(id: id, roomName: room, participantName: participant) |
| 203 | + return (sandboxConnection.serverUrl, sandboxConnection.participantToken) |
| 204 | + case let .cloud(server, token): |
| 205 | + return (server, token) |
| 206 | + } |
| 207 | + } |
| 208 | +} |
0 commit comments