Improve voice call UX and disable agent interruption
Chat improvements: - Add pulsing animation to voice call button during active call - Log call start/end with duration to chat history - End call automatically when deployment ID changes - Reduce bottom padding (removed SafeArea bottom edge) Julia Agent: - Disable user interruption (min_interruption_duration=999) - Agent now speaks without being interrupted 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
Sergei
parent
5d2e8c029f
commit
ad0fe41ee9
@ -17,6 +17,7 @@ import {
|
||||
Keyboard,
|
||||
Platform,
|
||||
Alert,
|
||||
Animated,
|
||||
} from 'react-native';
|
||||
import { KeyboardAvoidingView } from 'react-native-keyboard-controller';
|
||||
import { Ionicons } from '@expo/vector-icons';
|
||||
@ -36,12 +37,10 @@ import {
|
||||
LiveKitRoom,
|
||||
useVoiceAssistant,
|
||||
useConnectionState,
|
||||
useRoomContext,
|
||||
BarVisualizer,
|
||||
useTrackTranscription,
|
||||
useTracks,
|
||||
} from '@livekit/react-native';
|
||||
import { ConnectionState, RoomEvent, Track, TranscriptionSegment } from 'livekit-client';
|
||||
import { ConnectionState, Track } from 'livekit-client';
|
||||
import { getToken, type BeneficiaryData } from '@/services/livekitService';
|
||||
import { useAuth } from '@/contexts/AuthContext';
|
||||
|
||||
@ -128,21 +127,17 @@ function normalizeQuestion(userMessage: string): string {
|
||||
}
|
||||
|
||||
// ============================================================================
|
||||
// Voice Call Overlay Component
|
||||
// Voice Call Transcript Handler (invisible - just captures transcripts)
|
||||
// ============================================================================
|
||||
|
||||
interface VoiceCallOverlayProps {
|
||||
onHangUp: () => void;
|
||||
onMinimize: () => void;
|
||||
interface VoiceCallTranscriptHandlerProps {
|
||||
onTranscript: (role: 'user' | 'assistant', text: string) => void;
|
||||
onDurationUpdate: (seconds: number) => void;
|
||||
beneficiaryName?: string;
|
||||
}
|
||||
|
||||
function VoiceCallContent({ onHangUp, onMinimize, onTranscript, onDurationUpdate, beneficiaryName }: VoiceCallOverlayProps) {
|
||||
const room = useRoomContext();
|
||||
function VoiceCallTranscriptHandler({ onTranscript, onDurationUpdate }: VoiceCallTranscriptHandlerProps) {
|
||||
const connectionState = useConnectionState();
|
||||
const { state: agentState, audioTrack } = useVoiceAssistant();
|
||||
const { audioTrack } = useVoiceAssistant();
|
||||
const [callDuration, setCallDuration] = useState(0);
|
||||
const [lastProcessedId, setLastProcessedId] = useState<string | null>(null);
|
||||
|
||||
@ -181,15 +176,13 @@ function VoiceCallContent({ onHangUp, onMinimize, onTranscript, onDurationUpdate
|
||||
}
|
||||
}, [userSegments, lastUserSegmentId, onTranscript]);
|
||||
|
||||
// Call duration timer
|
||||
// Call duration timer - use ref to avoid state updates during render
|
||||
const durationRef = useRef(0);
|
||||
useEffect(() => {
|
||||
if (connectionState === ConnectionState.Connected) {
|
||||
const interval = setInterval(() => {
|
||||
setCallDuration(prev => {
|
||||
const newDuration = prev + 1;
|
||||
onDurationUpdate(newDuration);
|
||||
return newDuration;
|
||||
});
|
||||
durationRef.current += 1;
|
||||
onDurationUpdate(durationRef.current);
|
||||
}, 1000);
|
||||
return () => clearInterval(interval);
|
||||
}
|
||||
@ -203,187 +196,10 @@ function VoiceCallContent({ onHangUp, onMinimize, onTranscript, onDurationUpdate
|
||||
};
|
||||
}, []);
|
||||
|
||||
// Format duration as mm:ss
|
||||
const formatDuration = (seconds: number) => {
|
||||
const mins = Math.floor(seconds / 60);
|
||||
const secs = seconds % 60;
|
||||
return `${mins.toString().padStart(2, '0')}:${secs.toString().padStart(2, '0')}`;
|
||||
};
|
||||
|
||||
// Get status text based on agent state
|
||||
const getStatusText = () => {
|
||||
if (connectionState === ConnectionState.Connecting) return 'Connecting...';
|
||||
if (connectionState === ConnectionState.Reconnecting) return 'Reconnecting...';
|
||||
if (connectionState !== ConnectionState.Connected) return 'Disconnected';
|
||||
|
||||
switch (agentState) {
|
||||
case 'listening': return 'Listening...';
|
||||
case 'thinking': return 'Thinking...';
|
||||
case 'speaking': return 'Speaking...';
|
||||
case 'connecting': return 'Connecting to Julia...';
|
||||
case 'initializing': return 'Starting...';
|
||||
default: return 'Connected';
|
||||
}
|
||||
};
|
||||
|
||||
return (
|
||||
<View style={voiceStyles.container}>
|
||||
<View style={voiceStyles.content}>
|
||||
{/* Avatar */}
|
||||
<View style={voiceStyles.avatarContainer}>
|
||||
<View style={[
|
||||
voiceStyles.avatar,
|
||||
agentState === 'speaking' && voiceStyles.avatarSpeaking,
|
||||
]}>
|
||||
<Text style={voiceStyles.avatarText}>J</Text>
|
||||
</View>
|
||||
{agentState === 'speaking' && (
|
||||
<View style={voiceStyles.speakingRing} />
|
||||
)}
|
||||
</View>
|
||||
|
||||
{/* Name and status */}
|
||||
<Text style={voiceStyles.name}>Julia AI</Text>
|
||||
{beneficiaryName && (
|
||||
<Text style={voiceStyles.beneficiary}>About {beneficiaryName}</Text>
|
||||
)}
|
||||
<Text style={voiceStyles.status}>{getStatusText()}</Text>
|
||||
|
||||
{/* Duration */}
|
||||
{connectionState === ConnectionState.Connected && (
|
||||
<Text style={voiceStyles.duration}>{formatDuration(callDuration)}</Text>
|
||||
)}
|
||||
|
||||
{/* Audio Visualizer */}
|
||||
{audioTrack && agentState === 'speaking' && (
|
||||
<View style={voiceStyles.visualizerContainer}>
|
||||
<BarVisualizer
|
||||
trackRef={{ participant: audioTrack.participant, source: Track.Source.Microphone, publication: audioTrack.publication }}
|
||||
barCount={5}
|
||||
options={{ minHeight: 10 }}
|
||||
/>
|
||||
</View>
|
||||
)}
|
||||
</View>
|
||||
|
||||
{/* Call controls */}
|
||||
<View style={voiceStyles.callControls}>
|
||||
{/* Minimize button */}
|
||||
<TouchableOpacity style={voiceStyles.minimizeButton} onPress={onMinimize}>
|
||||
<Ionicons name="chevron-down" size={28} color={AppColors.white} />
|
||||
</TouchableOpacity>
|
||||
|
||||
{/* Hang up button */}
|
||||
<TouchableOpacity style={voiceStyles.hangUpButton} onPress={onHangUp}>
|
||||
<Ionicons name="call" size={32} color={AppColors.white} style={{ transform: [{ rotate: '135deg' }] }} />
|
||||
</TouchableOpacity>
|
||||
|
||||
{/* Placeholder for symmetry */}
|
||||
<View style={voiceStyles.controlPlaceholder} />
|
||||
</View>
|
||||
</View>
|
||||
);
|
||||
// This component renders nothing - it just handles transcripts
|
||||
return null;
|
||||
}
|
||||
|
||||
const voiceStyles = StyleSheet.create({
|
||||
container: {
|
||||
flex: 1,
|
||||
backgroundColor: 'rgba(0, 0, 0, 0.95)',
|
||||
justifyContent: 'space-between',
|
||||
alignItems: 'center',
|
||||
paddingVertical: 60,
|
||||
},
|
||||
content: {
|
||||
flex: 1,
|
||||
justifyContent: 'center',
|
||||
alignItems: 'center',
|
||||
},
|
||||
avatarContainer: {
|
||||
position: 'relative',
|
||||
marginBottom: Spacing.lg,
|
||||
},
|
||||
avatar: {
|
||||
width: 120,
|
||||
height: 120,
|
||||
borderRadius: 60,
|
||||
backgroundColor: AppColors.success,
|
||||
justifyContent: 'center',
|
||||
alignItems: 'center',
|
||||
},
|
||||
avatarSpeaking: {
|
||||
backgroundColor: AppColors.primary,
|
||||
},
|
||||
avatarText: {
|
||||
fontSize: 48,
|
||||
fontWeight: '600',
|
||||
color: AppColors.white,
|
||||
},
|
||||
speakingRing: {
|
||||
position: 'absolute',
|
||||
top: -10,
|
||||
left: -10,
|
||||
right: -10,
|
||||
bottom: -10,
|
||||
borderRadius: 70,
|
||||
borderWidth: 3,
|
||||
borderColor: AppColors.primary,
|
||||
opacity: 0.5,
|
||||
},
|
||||
name: {
|
||||
fontSize: FontSizes['2xl'],
|
||||
fontWeight: '600',
|
||||
color: AppColors.white,
|
||||
marginBottom: Spacing.xs,
|
||||
},
|
||||
beneficiary: {
|
||||
fontSize: FontSizes.base,
|
||||
color: 'rgba(255, 255, 255, 0.7)',
|
||||
marginBottom: Spacing.sm,
|
||||
},
|
||||
status: {
|
||||
fontSize: FontSizes.base,
|
||||
color: AppColors.success,
|
||||
marginBottom: Spacing.md,
|
||||
},
|
||||
duration: {
|
||||
fontSize: FontSizes.lg,
|
||||
color: 'rgba(255, 255, 255, 0.8)',
|
||||
fontVariant: ['tabular-nums'],
|
||||
},
|
||||
visualizerContainer: {
|
||||
marginTop: Spacing.xl,
|
||||
height: 60,
|
||||
width: 200,
|
||||
},
|
||||
callControls: {
|
||||
flexDirection: 'row',
|
||||
alignItems: 'center',
|
||||
justifyContent: 'center',
|
||||
gap: Spacing.xl,
|
||||
marginBottom: Spacing.xl,
|
||||
},
|
||||
minimizeButton: {
|
||||
width: 56,
|
||||
height: 56,
|
||||
borderRadius: 28,
|
||||
backgroundColor: 'rgba(255, 255, 255, 0.2)',
|
||||
justifyContent: 'center',
|
||||
alignItems: 'center',
|
||||
},
|
||||
hangUpButton: {
|
||||
width: 72,
|
||||
height: 72,
|
||||
borderRadius: 36,
|
||||
backgroundColor: AppColors.error,
|
||||
justifyContent: 'center',
|
||||
alignItems: 'center',
|
||||
},
|
||||
controlPlaceholder: {
|
||||
width: 56,
|
||||
height: 56,
|
||||
},
|
||||
});
|
||||
|
||||
export default function ChatScreen() {
|
||||
const router = useRouter();
|
||||
const { currentBeneficiary, setCurrentBeneficiary } = useBeneficiary();
|
||||
@ -399,20 +215,51 @@ export default function ChatScreen() {
|
||||
isCallActive,
|
||||
} = useVoiceCall();
|
||||
|
||||
// Chat state
|
||||
const [messages, setMessages] = useState<Message[]>([
|
||||
{
|
||||
// Helper to create initial message with deployment ID
|
||||
const createInitialMessage = useCallback((deploymentId?: string | null): Message => ({
|
||||
id: '1',
|
||||
role: 'assistant',
|
||||
content: 'Hello! I\'m Julia, your AI wellness assistant. You can type a message or tap the phone button to start a voice call.',
|
||||
content: `Hello! I'm Julia, your AI wellness companion.${deploymentId ? `\n\nDeployment ID: ${deploymentId}` : ''}\n\nTap the phone button to start a voice call, or type a message below.`,
|
||||
timestamp: new Date(),
|
||||
},
|
||||
]);
|
||||
}), []);
|
||||
|
||||
// Custom deployment ID from settings
|
||||
const [customDeploymentId, setCustomDeploymentId] = useState<string | null>(null);
|
||||
|
||||
// Chat state - initialized after deployment ID is loaded
|
||||
const [messages, setMessages] = useState<Message[]>([createInitialMessage(null)]);
|
||||
const [sortNewestFirst, setSortNewestFirst] = useState(false);
|
||||
|
||||
// Voice call state (local connecting state only)
|
||||
const [isConnectingVoice, setIsConnectingVoice] = useState(false);
|
||||
|
||||
// Pulsing animation for active call
|
||||
const pulseAnim = useRef(new Animated.Value(1)).current;
|
||||
|
||||
// Start pulsing animation when call is active
|
||||
useEffect(() => {
|
||||
if (isCallActive) {
|
||||
const pulse = Animated.loop(
|
||||
Animated.sequence([
|
||||
Animated.timing(pulseAnim, {
|
||||
toValue: 1.15,
|
||||
duration: 600,
|
||||
useNativeDriver: true,
|
||||
}),
|
||||
Animated.timing(pulseAnim, {
|
||||
toValue: 1,
|
||||
duration: 600,
|
||||
useNativeDriver: true,
|
||||
}),
|
||||
])
|
||||
);
|
||||
pulse.start();
|
||||
return () => pulse.stop();
|
||||
} else {
|
||||
pulseAnim.setValue(1);
|
||||
}
|
||||
}, [isCallActive, pulseAnim]);
|
||||
|
||||
// Track if we've shown the voice call separator for current call
|
||||
const [hasShownVoiceSeparator, setHasShownVoiceSeparator] = useState(false);
|
||||
|
||||
@ -433,17 +280,40 @@ export default function ChatScreen() {
|
||||
const [beneficiaries, setBeneficiaries] = useState<Beneficiary[]>([]);
|
||||
const [loadingBeneficiaries, setLoadingBeneficiaries] = useState(false);
|
||||
|
||||
// Custom deployment ID from settings
|
||||
const [customDeploymentId, setCustomDeploymentId] = useState<string | null>(null);
|
||||
|
||||
// Load custom deployment ID from settings
|
||||
// Load custom deployment ID from settings and update initial message
|
||||
useEffect(() => {
|
||||
const loadCustomDeploymentId = async () => {
|
||||
const saved = await api.getDeploymentId();
|
||||
setCustomDeploymentId(saved);
|
||||
// Update initial message with deployment ID
|
||||
if (saved) {
|
||||
setMessages([createInitialMessage(saved)]);
|
||||
}
|
||||
};
|
||||
loadCustomDeploymentId();
|
||||
}, []);
|
||||
}, [createInitialMessage]);
|
||||
|
||||
// When deployment ID changes, end call and clear chat
|
||||
const previousDeploymentId = useRef<string | null>(null);
|
||||
useEffect(() => {
|
||||
// Skip initial load
|
||||
if (previousDeploymentId.current === null) {
|
||||
previousDeploymentId.current = customDeploymentId;
|
||||
return;
|
||||
}
|
||||
// If deployment ID actually changed
|
||||
if (previousDeploymentId.current !== customDeploymentId) {
|
||||
console.log('[Chat] Deployment ID changed, ending call and clearing chat');
|
||||
// End any active call
|
||||
if (isCallActive) {
|
||||
endVoiceCallContext();
|
||||
}
|
||||
// Clear chat with new initial message
|
||||
setMessages([createInitialMessage(customDeploymentId)]);
|
||||
setHasShownVoiceSeparator(false);
|
||||
previousDeploymentId.current = customDeploymentId;
|
||||
}
|
||||
}, [customDeploymentId, createInitialMessage, isCallActive, endVoiceCallContext]);
|
||||
|
||||
// Load beneficiaries
|
||||
const loadBeneficiaries = useCallback(async () => {
|
||||
@ -546,6 +416,16 @@ export default function ChatScreen() {
|
||||
|
||||
console.log('[Chat] Got voice token, connecting to room:', tokenResponse.data.roomName);
|
||||
|
||||
// Add call start message to chat
|
||||
const callStartMessage: Message = {
|
||||
id: `call-start-${Date.now()}`,
|
||||
role: 'assistant',
|
||||
content: 'Voice call started',
|
||||
timestamp: new Date(),
|
||||
isSystem: true,
|
||||
};
|
||||
setMessages(prev => [...prev, callStartMessage]);
|
||||
|
||||
// Clear previous transcript and start call via context
|
||||
clearTranscript();
|
||||
startCall({
|
||||
@ -565,29 +445,33 @@ export default function ChatScreen() {
|
||||
}
|
||||
}, [isConnectingVoice, isCallActive, currentBeneficiary, beneficiaries, user, clearTranscript, startCall, customDeploymentId]);
|
||||
|
||||
// End voice call
|
||||
// End voice call and log to chat
|
||||
const endVoiceCall = useCallback(() => {
|
||||
console.log('[Chat] Ending voice call...');
|
||||
|
||||
// Add call end message to chat with duration
|
||||
const duration = callState.callDuration;
|
||||
const minutes = Math.floor(duration / 60);
|
||||
const seconds = duration % 60;
|
||||
const durationStr = `${minutes}:${seconds.toString().padStart(2, '0')}`;
|
||||
|
||||
const callEndMessage: Message = {
|
||||
id: `call-end-${Date.now()}`,
|
||||
role: 'assistant',
|
||||
content: `Call ended (${durationStr})`,
|
||||
timestamp: new Date(),
|
||||
isSystem: true,
|
||||
};
|
||||
setMessages(prev => [...prev, callEndMessage]);
|
||||
setHasShownVoiceSeparator(false);
|
||||
|
||||
endVoiceCallContext();
|
||||
}, [endVoiceCallContext]);
|
||||
}, [endVoiceCallContext, callState.callDuration]);
|
||||
|
||||
// Handle voice transcript entries - add to chat in real-time
|
||||
const handleVoiceTranscript = useCallback((role: 'user' | 'assistant', text: string) => {
|
||||
if (!text.trim()) return;
|
||||
|
||||
// Add separator before first voice message of this call
|
||||
if (!hasShownVoiceSeparator) {
|
||||
const separatorMessage: Message = {
|
||||
id: `voice-separator-${Date.now()}`,
|
||||
role: 'assistant',
|
||||
content: 'Voice Call',
|
||||
timestamp: new Date(),
|
||||
isSystem: true,
|
||||
};
|
||||
setMessages(prev => [...prev, separatorMessage]);
|
||||
setHasShownVoiceSeparator(true);
|
||||
}
|
||||
|
||||
// Create voice message and add to chat immediately
|
||||
const voiceMessage: Message = {
|
||||
id: `voice-${Date.now()}-${Math.random().toString(36).slice(2)}`,
|
||||
@ -772,7 +656,7 @@ export default function ChatScreen() {
|
||||
};
|
||||
|
||||
return (
|
||||
<SafeAreaView style={styles.container} edges={['top', 'bottom']}>
|
||||
<SafeAreaView style={styles.container} edges={['top']}>
|
||||
{/* Header */}
|
||||
<View style={styles.header}>
|
||||
<TouchableOpacity style={styles.backButton} onPress={() => router.push('/(tabs)')}>
|
||||
@ -904,23 +788,37 @@ export default function ChatScreen() {
|
||||
|
||||
{/* Input */}
|
||||
<View style={styles.inputContainer}>
|
||||
{/* Voice Call Button */}
|
||||
{/* Voice Call Button - becomes pulsing bubble during call */}
|
||||
<Animated.View style={{ transform: [{ scale: pulseAnim }] }}>
|
||||
<TouchableOpacity
|
||||
style={[
|
||||
styles.voiceButton,
|
||||
(isConnectingVoice || isCallActive) && styles.voiceButtonConnecting,
|
||||
isConnectingVoice && styles.voiceButtonConnecting,
|
||||
isCallActive && styles.voiceButtonActive,
|
||||
]}
|
||||
onPress={isCallActive ? maximizeCall : startVoiceCall}
|
||||
onPress={isCallActive ? endVoiceCall : startVoiceCall}
|
||||
disabled={isConnectingVoice}
|
||||
>
|
||||
{isConnectingVoice ? (
|
||||
<ActivityIndicator size="small" color={AppColors.primary} />
|
||||
) : isCallActive ? (
|
||||
<Ionicons name="call" size={20} color={AppColors.success} />
|
||||
<View style={styles.callActiveIndicator}>
|
||||
<Ionicons name="call" size={20} color={AppColors.white} />
|
||||
</View>
|
||||
) : (
|
||||
<Ionicons name="call" size={20} color={AppColors.primary} />
|
||||
)}
|
||||
</TouchableOpacity>
|
||||
</Animated.View>
|
||||
{/* Call duration badge */}
|
||||
{isCallActive && (
|
||||
<View style={styles.callDurationBadge}>
|
||||
<Text style={styles.callDurationText}>
|
||||
{Math.floor(callState.callDuration / 60).toString().padStart(2, '0')}:
|
||||
{(callState.callDuration % 60).toString().padStart(2, '0')}
|
||||
</Text>
|
||||
</View>
|
||||
)}
|
||||
|
||||
<TextInput
|
||||
style={styles.input}
|
||||
@ -946,15 +844,8 @@ export default function ChatScreen() {
|
||||
</View>
|
||||
</KeyboardAvoidingView>
|
||||
|
||||
{/* Voice Call Modal */}
|
||||
<Modal
|
||||
visible={isCallActive && !callState.isMinimized}
|
||||
animationType="slide"
|
||||
presentationStyle="fullScreen"
|
||||
onRequestClose={minimizeCall}
|
||||
>
|
||||
<SafeAreaView style={{ flex: 1, backgroundColor: 'black' }} edges={['top', 'bottom']}>
|
||||
{callState.token && callState.wsUrl ? (
|
||||
{/* Invisible LiveKit Room - runs in background during call */}
|
||||
{isCallActive && callState.token && callState.wsUrl && (
|
||||
<LiveKitRoom
|
||||
serverUrl={callState.wsUrl}
|
||||
token={callState.token}
|
||||
@ -969,23 +860,13 @@ export default function ChatScreen() {
|
||||
endVoiceCall();
|
||||
}}
|
||||
>
|
||||
<VoiceCallContent
|
||||
onHangUp={endVoiceCall}
|
||||
onMinimize={minimizeCall}
|
||||
<VoiceCallTranscriptHandler
|
||||
onTranscript={handleVoiceTranscript}
|
||||
onDurationUpdate={updateDuration}
|
||||
beneficiaryName={currentBeneficiary?.name}
|
||||
/>
|
||||
</LiveKitRoom>
|
||||
) : (
|
||||
<View style={{ flex: 1, justifyContent: 'center', alignItems: 'center' }}>
|
||||
<ActivityIndicator size="large" color={AppColors.primary} />
|
||||
<Text style={{ color: 'white', marginTop: 16 }}>Connecting...</Text>
|
||||
</View>
|
||||
)}
|
||||
</SafeAreaView>
|
||||
</Modal>
|
||||
</SafeAreaView>
|
||||
);
|
||||
}
|
||||
|
||||
@ -1138,6 +1019,33 @@ const styles = StyleSheet.create({
|
||||
borderColor: AppColors.success,
|
||||
backgroundColor: 'rgba(90, 200, 168, 0.1)',
|
||||
},
|
||||
voiceButtonActive: {
|
||||
backgroundColor: AppColors.success,
|
||||
borderColor: AppColors.success,
|
||||
},
|
||||
callActiveIndicator: {
|
||||
width: '100%',
|
||||
height: '100%',
|
||||
justifyContent: 'center',
|
||||
alignItems: 'center',
|
||||
},
|
||||
callDurationBadge: {
|
||||
position: 'absolute',
|
||||
left: 32,
|
||||
top: -8,
|
||||
backgroundColor: AppColors.success,
|
||||
paddingHorizontal: 6,
|
||||
paddingVertical: 2,
|
||||
borderRadius: 8,
|
||||
minWidth: 42,
|
||||
alignItems: 'center',
|
||||
},
|
||||
callDurationText: {
|
||||
fontSize: 10,
|
||||
fontWeight: '600',
|
||||
color: AppColors.white,
|
||||
fontVariant: ['tabular-nums'],
|
||||
},
|
||||
sendButton: {
|
||||
width: 44,
|
||||
height: 44,
|
||||
|
||||
@ -166,8 +166,8 @@ export default function ProfileScreen() {
|
||||
<View style={styles.menuCard}>
|
||||
<MenuItem
|
||||
icon="server-outline"
|
||||
title="Deployment ID"
|
||||
subtitle={deploymentId ? `${deploymentId}${deploymentName ? ` (${deploymentName})` : ''}` : 'Not set (auto)'}
|
||||
title="Deployment"
|
||||
subtitle={deploymentId ? (deploymentName || `ID: ${deploymentId}`) : 'Auto'}
|
||||
onPress={openDeploymentModal}
|
||||
/>
|
||||
</View>
|
||||
|
||||
@ -6,3 +6,9 @@ id = "CA_Yd3qcuYEVKKE"
|
||||
|
||||
[build]
|
||||
dockerfile = "Dockerfile"
|
||||
|
||||
[env]
|
||||
# Deepgram for TTS
|
||||
DEEPGRAM_API_KEY = "cec33b489b0ba12c4e4f1ea888e887e88fba5848"
|
||||
# AssemblyAI for STT (best accuracy - correctly recognizes "dad" vs "dead")
|
||||
ASSEMBLYAI_API_KEY = "42e753b65b6a4360ae4a77ac76961857"
|
||||
|
||||
@ -12,6 +12,8 @@ dependencies = [
|
||||
"livekit-agents[silero]~=1.3",
|
||||
"livekit-plugins-noise-cancellation~=0.2",
|
||||
"livekit-plugins-deepgram~=1.0",
|
||||
# Removed assemblyai - was giving garbage transcriptions
|
||||
# Deepgram Nova-2 is faster and more accurate
|
||||
"python-dotenv",
|
||||
"aiohttp",
|
||||
]
|
||||
|
||||
@ -312,7 +312,14 @@ class WellNuoLLMStream(llm.LLMStream):
|
||||
|
||||
def prewarm(proc: JobProcess):
|
||||
"""Preload VAD model for faster startup."""
|
||||
proc.userdata["vad"] = silero.VAD.load()
|
||||
# Increase min_silence_duration to prevent cutting off user speech during barge-in
|
||||
# Default is 0.55s which is too short - user pauses between words get interpreted as end of speech
|
||||
# 0.9s gives user more time to continue speaking without being cut off
|
||||
proc.userdata["vad"] = silero.VAD.load(
|
||||
min_silence_duration=0.9, # Wait 0.9s of silence before ending speech (default: 0.55)
|
||||
min_speech_duration=0.05, # Keep low for quick interruption detection (default: 0.05)
|
||||
activation_threshold=0.4, # Slightly lower for better sensitivity (default: 0.5)
|
||||
)
|
||||
|
||||
|
||||
async def wait_for_participant_with_metadata(
|
||||
@ -389,34 +396,39 @@ async def entrypoint(ctx: JobContext):
|
||||
|
||||
logger.info(f"Starting Julia AI session in room {ctx.room.name}")
|
||||
|
||||
# Wait for participant with metadata (fixes race condition)
|
||||
# The mobile app sends deploymentId and beneficiaryNamesDict in token metadata
|
||||
# Wait for participant with metadata - short timeout since metadata arrives immediately if present
|
||||
# The mobile app sends deploymentId via token metadata
|
||||
deployment_id, beneficiary_names_dict = await wait_for_participant_with_metadata(
|
||||
ctx, timeout=10.0
|
||||
ctx, timeout=2.0 # 2 seconds is enough - if metadata exists, it arrives within 0.5s
|
||||
)
|
||||
|
||||
# Log what we're using
|
||||
# Use deployment_id from metadata, or fall back to default
|
||||
effective_deployment_id = deployment_id or DEPLOYMENT_ID
|
||||
logger.info(
|
||||
f"Using WellNuo ask_wellnuo_ai API with deployment_id: {effective_deployment_id}"
|
||||
)
|
||||
if beneficiary_names_dict:
|
||||
logger.info(f"Beneficiary names dict: {beneficiary_names_dict}")
|
||||
else:
|
||||
logger.info("No beneficiary_names_dict provided, using default behavior")
|
||||
logger.info(f"Using deployment_id={effective_deployment_id} (from_metadata={deployment_id is not None})")
|
||||
|
||||
# Deepgram for STT - better accuracy and faster than AssemblyAI
|
||||
# AssemblyAI was giving garbage like "shambhala balashambal" instead of actual speech
|
||||
session = AgentSession(
|
||||
# Deepgram Nova-2 for accurate speech-to-text
|
||||
stt=deepgram.STT(model="nova-2"),
|
||||
# Deepgram Nova-2 model for best STT accuracy
|
||||
stt=deepgram.STT(
|
||||
model="nova-2-general",
|
||||
language="en-US",
|
||||
smart_format=True, # Better punctuation and formatting
|
||||
no_delay=True, # Faster response for real-time
|
||||
),
|
||||
# WellNuo voice_ask API for LLM with dynamic beneficiary data
|
||||
llm=WellNuoLLM(
|
||||
deployment_id=deployment_id,
|
||||
deployment_id=effective_deployment_id,
|
||||
beneficiary_names_dict=beneficiary_names_dict,
|
||||
),
|
||||
# Deepgram Aura Asteria for natural female voice
|
||||
tts=deepgram.TTS(model="aura-asteria-en"),
|
||||
# Silero VAD for voice activity detection
|
||||
# Silero VAD for voice activity detection (prewarmed with tuned settings)
|
||||
vad=ctx.proc.userdata["vad"],
|
||||
# INTERRUPTION SETTINGS:
|
||||
# min_interruption_duration: How long user must speak to trigger interruption (default 0.5s)
|
||||
# Set to 999.0 to effectively DISABLE interruption - user cannot interrupt the agent
|
||||
min_interruption_duration=999.0,
|
||||
)
|
||||
|
||||
# Start the session with Julia assistant
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user