@omote/core 0.5.3 → 0.5.5

package/dist/index.d.mts

@@ -1,5 +1,5 @@
-import { EventEmitter, OmoteEvents, AISessionState, AnimationEvent } from './events/index.mjs';
-export { BackendEvent, EmotionEvent, GazeEvent, STTFinalEvent, STTPartialEvent, SessionStateEvent, TTSEndEvent, TTSMarkEvent, TTSStartEvent, VisemeEvent } from './events/index.mjs';
+import { EventEmitter, OmoteEvents } from './events/index.mjs';
+export { AnimationEvent, BackendEvent, EmotionEvent, GazeEvent, STTFinalEvent, STTPartialEvent, SessionStateEvent, TTSEndEvent, TTSMarkEvent, TTSStartEvent, VisemeEvent } from './events/index.mjs';
 export { D as DEFAULT_LOGGING_CONFIG, I as ILogger, e as LOG_LEVEL_PRIORITY, b as LogEntry, L as LogFormatter, a as LogLevel, c as LogSink, d as LoggingConfig, f as configureLogging, i as createLogger, g as getLoggingConfig, n as noopLogger, r as resetLoggingConfig, s as setLogLevel, h as setLoggingEnabled } from './Logger-I_k4sGhM.mjs';
 export { ARKitToFLAMEMapping, ApiError, AudioChunkEvent, AvatarFormat, Character, CharacterAvatar, CharacterMemory, CharacterPersonality, CharacterSpec, CharacterVoice, CreateCharacterRequest, CreateCharacterResponse, CreateLAMJobRequest, CreateLAMJobResponse, CreateSessionRequest, CreateSessionResponse, GSplatConfig, LAMJob, LAMJobStatus, PROTOCOL_VERSION, PaginatedResponse, PlatformSession, ErrorEvent as ProtocolErrorEvent, ProtocolEvent, ResponseChunkEvent, ResponseEndEvent, ResponseStartEvent, SessionMessage, SessionStatus, isProtocolEvent } from '@omote/types';
 
@@ -584,6 +584,14 @@ interface FullFacePipelineOptions {
     chunkSize?: number;
     /** A2E inference engine */
     lam: A2EBackend;
+    /**
+     * Identity/style index for the A2E model (default: 0).
+     *
+     * The LAM model uses a 12-class one-hot identity vector as style conditioning.
+     * Different indices produce different expression intensity across face regions.
+     * Only affects Wav2Vec2Inference (GPU). Wav2ArkitCpuInference has identity 11 baked in.
+     */
+    identityIndex?: number;
     /** Per-character expression weight scaling */
     profile?: ExpressionProfile;
     /**
@@ -719,6 +727,78 @@ declare class FullFacePipeline extends EventEmitter<FullFacePipelineEvents> {
     dispose(): void;
 }
 
+/**
+ * Interruption Handler
+ *
+ * VAD-based barge-in detection for AI conversations:
+ * - Monitors VAD probability for user speech
+ * - Detects when user interrupts AI response
+ * - Triggers interruption callbacks
+ */
+
+interface InterruptionEvents {
+    [key: string]: unknown;
+    'speech.detected': {
+        rms: number;
+    };
+    'speech.ended': {
+        durationMs: number;
+    };
+    'interruption.triggered': {
+        rms: number;
+        durationMs: number;
+    };
+}
+/**
+ * Interruption handler configuration
+ *
+ * Industry standards applied:
+ * - vadThreshold: 0.5 (Silero VAD default)
+ * - minSpeechDurationMs: 200ms (Google/Amazon barge-in standard)
+ * - silenceTimeoutMs: 500ms (OpenAI Realtime API standard)
+ */
+interface InterruptionConfig {
+    /** VAD probability threshold for speech detection (default: 0.5, Silero standard) */
+    vadThreshold?: number;
+    /** Minimum speech duration to trigger interruption (default: 200ms, Google/Amazon standard) */
+    minSpeechDurationMs?: number;
+    /** Silence duration to end speech (default: 500ms, OpenAI standard) */
+    silenceTimeoutMs?: number;
+    /** Enable interruption detection (default: true) */
+    enabled?: boolean;
+}
+declare class InterruptionHandler extends EventEmitter<InterruptionEvents> {
+    private config;
+    private isSpeaking;
+    private speechStartTime;
+    private lastSpeechTime;
+    private silenceTimer;
+    private aiIsSpeaking;
+    private interruptionTriggeredThisSession;
+    constructor(config?: InterruptionConfig);
+    /**
+     * Process VAD result for interruption detection
+     * @param vadProbability - Speech probability from VAD (0-1)
+     * @param audioEnergy - Optional RMS energy for logging (default: 0)
+     */
+    processVADResult(vadProbability: number, audioEnergy?: number): void;
+    /** Notify that AI started/stopped speaking */
+    setAISpeaking(speaking: boolean): void;
+    /** Enable/disable interruption detection */
+    setEnabled(enabled: boolean): void;
+    /** Update configuration */
+    updateConfig(config: Partial<InterruptionConfig>): void;
+    /** Reset state */
+    reset(): void;
+    /** Get current state */
+    getState(): {
+        isSpeaking: boolean;
+        speechDurationMs: number;
+    };
+    private onSpeechDetected;
+    private onSilenceDetected;
+}
+
 /**
  * Lazy ONNX Runtime loader with conditional WebGPU/WASM bundle loading
  *
@@ -2187,6 +2267,17 @@ interface A2EProcessorConfig {
     sampleRate?: number;
     /** Samples per inference chunk (default: 16000 = 1s) */
     chunkSize?: number;
+    /**
+     * Identity/style index for the A2E model (default: 0).
+     *
+     * The LAM model uses a one-hot identity vector (12 classes, indices 0-11) as
+     * style conditioning alongside audio features. Different indices produce
+     * different expression intensity across face regions (brows, eyes, cheeks).
+     *
+     * Only affects Wav2Vec2Inference (GPU model). Wav2ArkitCpuInference has
+     * identity 11 baked into the model weights.
+     */
+    identityIndex?: number;
     /** Callback fired with each blendshape frame (push mode) */
     onFrame?: (frame: Float32Array) => void;
     /** Error callback */
@@ -2196,6 +2287,7 @@ declare class A2EProcessor {
     private readonly backend;
     private readonly sampleRate;
     private readonly chunkSize;
+    private readonly identityIndex;
     private readonly onFrame?;
     private readonly onError?;
     private bufferCapacity;
@@ -2821,813 +2913,6 @@ declare class EmotionController {
     reset(): void;
 }
 
-/**
- * AI Adapter Interface
- *
- * Common interface for AI backends (AWS AgentCore, OpenAI Realtime).
- * Adapters handle the conversation flow and emit events for animation.
- *
- * @category AI
- */
-
-/**
- * Tenant configuration for multi-tenant isolation
- */
-interface TenantConfig {
-    /** Unique tenant identifier */
-    tenantId: string;
-    /** Customer-specific API credentials */
-    credentials: {
-        apiKey?: string;
-        authToken?: string;
-        refreshToken?: string;
-    };
-    /** Character configuration for this tenant */
-    characterId: string;
-    /** Optional custom endpoint override */
-    endpoint?: string;
-}
-/**
- * Voice configuration for TTS
- */
-interface VoiceConfig {
-    /** TTS provider */
-    provider: 'elevenlabs' | 'openai';
-    /** Voice ID */
-    voiceId: string;
-    /** Stability (0-1, ElevenLabs) */
-    stability?: number;
-    /** Similarity boost (0-1, ElevenLabs) */
-    similarityBoost?: number;
-}
-/**
- * Session configuration
- */
-interface SessionConfig {
-    /** Session ID (generated or provided) */
-    sessionId: string;
-    /** Tenant this session belongs to */
-    tenant: TenantConfig;
-    /** Initial system prompt / personality */
-    systemPrompt?: string;
-    /** Voice configuration for TTS */
-    voice?: VoiceConfig;
-    /** Initial emotion state */
-    emotion?: string;
-    /** Language code */
-    language?: string;
-}
-/**
- * Message role in conversation
- */
-type MessageRole = 'user' | 'assistant' | 'system';
-/**
- * Conversation message in session history
- */
-interface ConversationMessage {
-    /** Message role */
-    role: MessageRole;
-    /** Text content */
-    content: string;
-    /** Timestamp (ms) */
-    timestamp: number;
-    /** Emotion detected/expressed */
-    emotion?: string;
-    /** Audio duration if applicable (ms) */
-    audioDurationMs?: number;
-}
-
-/**
- * Events emitted by AI adapters
- */
-interface AIAdapterEvents {
-    [key: string]: unknown;
-    'state.change': {
-        state: AISessionState;
-        previousState: AISessionState;
-    };
-    'user.speech.start': {
-        timestamp: number;
-    };
-    'user.speech.end': {
-        timestamp: number;
-        durationMs: number;
-    };
-    'user.transcript.partial': {
-        text: string;
-        confidence: number;
-    };
-    'user.transcript.final': {
-        text: string;
-        confidence: number;
-    };
-    'ai.thinking.start': {
-        timestamp: number;
-    };
-    'ai.response.start': {
-        text?: string;
-        emotion?: string;
-    };
-    'ai.response.chunk': {
-        text: string;
-        isLast: boolean;
-    };
-    'ai.response.end': {
-        fullText: string;
-        durationMs: number;
-    };
-    'audio.output.chunk': {
-        audio: ArrayBuffer;
-        sampleRate: number;
-        timestamp: number;
-    };
-    'audio.output.end': {
-        durationMs: number;
-    };
-    'animation': AnimationEvent;
-    'memory.updated': {
-        messageCount: number;
-        tokenCount?: number;
-    };
-    'connection.opened': {
-        sessionId: string;
-        adapter: string;
-    };
-    'connection.closed': {
-        reason: string;
-    };
-    'connection.error': {
-        error: Error;
-        recoverable: boolean;
-    };
-    'interruption.detected': {
-        timestamp: number;
-    };
-    'interruption.handled': {
-        action: 'stop' | 'continue';
-        timestamp: number;
-    };
-}
-/**
- * Base interface for all AI adapters
- */
-interface AIAdapter {
-    /** Adapter name for logging/debugging */
-    readonly name: string;
-    /** Current session state */
-    readonly state: AISessionState;
-    /** Current session ID (null if not connected) */
-    readonly sessionId: string | null;
-    /** Whether the adapter is connected */
-    readonly isConnected: boolean;
-    /**
-     * Initialize and connect the adapter
-     */
-    connect(config: SessionConfig): Promise<void>;
-    /**
-     * Disconnect and cleanup
-     */
-    disconnect(): Promise<void>;
-    /**
-     * Push user audio for processing
-     * @param audio - PCM audio data (16kHz, mono)
-     */
-    pushAudio(audio: Int16Array | Float32Array): void;
-    /**
-     * Send text message directly (bypasses STT)
-     */
-    sendText(text: string): Promise<void>;
-    /**
-     * Handle user interruption
-     * Stops current AI speech and prepares for new input
-     */
-    interrupt(): void;
-    /**
-     * Get conversation history
-     */
-    getHistory(): ConversationMessage[];
-    /**
-     * Clear conversation history
-     */
-    clearHistory(): void;
-    /**
-     * Check if adapter is available/healthy
-     */
-    healthCheck(): Promise<boolean>;
-    on<K extends keyof AIAdapterEvents>(event: K, callback: (data: AIAdapterEvents[K]) => void): () => void;
-    off<K extends keyof AIAdapterEvents>(event: K, callback: (data: AIAdapterEvents[K]) => void): void;
-    once<K extends keyof AIAdapterEvents>(event: K, callback: (data: AIAdapterEvents[K]) => void): () => void;
-}
-
-/**
- * Conversation Session Interface
- *
- * Represents an active conversation with memory and state.
- *
- * @category AI
- */
-
-/**
- * Serializable session snapshot for persistence
- */
-interface SessionSnapshot {
-    /** Session ID */
-    sessionId: string;
-    /** Tenant ID */
-    tenantId: string;
-    /** Character ID */
-    characterId: string;
-    /** Conversation history */
-    history: ConversationMessage[];
-    /** Custom context */
-    context: Record<string, string>;
-    /** Emotion state */
-    emotion: EmotionWeights;
-    /** Creation timestamp */
-    createdAt: number;
-    /** Last activity timestamp */
-    lastActivityAt: number;
-}
-/**
- * Extended session with memory management
- */
-interface ConversationSession {
-    /** Session identifier */
-    readonly sessionId: string;
-    /** Associated AI adapter */
-    readonly adapter: AIAdapter;
-    /** Session configuration */
-    readonly config: SessionConfig;
-    /** Current state */
-    readonly state: AISessionState;
-    /** Conversation history */
-    readonly history: ConversationMessage[];
-    /** Current emotion state */
-    readonly emotion: EmotionWeights;
-    /** Session creation timestamp */
-    readonly createdAt: number;
-    /** Last activity timestamp */
-    readonly lastActivityAt: number;
-    /**
-     * Start the session (connects adapter)
-     */
-    start(): Promise<void>;
-    /**
-     * End the session (disconnects adapter)
-     */
-    end(): Promise<void>;
-    /**
-     * Push audio input
-     */
-    pushAudio(audio: Int16Array | Float32Array): void;
-    /**
-     * Send text input directly
-     */
-    sendText(text: string): Promise<void>;
-    /**
-     * Interrupt current AI response
-     */
-    interrupt(): void;
-    /**
-     * Update emotion state
-     */
-    setEmotion(emotion: EmotionWeights): void;
-    /**
-     * Add a context item (custom memory)
-     */
-    addContext(key: string, value: string): void;
-    /**
-     * Remove a context item
-     */
-    removeContext(key: string): void;
-    /**
-     * Get all context items
-     */
-    getContext(): Record<string, string>;
-    /**
-     * Export session for persistence
-     */
-    export(): SessionSnapshot;
-    /**
-     * Import session from snapshot
-     */
-    import(snapshot: SessionSnapshot): void;
-}
-
-/**
- * AWS AgentCore Adapter
- *
- * Primary AI adapter for the Omote Platform.
- *
- * Pipeline:
- * User Audio -> Whisper ASR (local) -> Text
- * Text -> AgentCore (WebSocket) -> Response Text + Audio chunks (TTS handled backend-side)
- * Audio chunks -> LAM (local) -> Blendshapes -> Render
- *
- * @category AI
- */
-
-/**
- * AgentCore-specific configuration
- */
-interface AgentCoreConfig {
-    /** AgentCore WebSocket endpoint */
-    endpoint: string;
-    /** AWS region */
-    region?: string;
-    /** Model URLs */
-    models?: {
-        lamUrl?: string;
-    };
-    /** Enable observability */
-    observability?: {
-        tracing?: boolean;
-        metrics?: boolean;
-    };
-}
-/**
- * AWS AgentCore Adapter
- */
-declare class AgentCoreAdapter extends EventEmitter<AIAdapterEvents> implements AIAdapter {
-    readonly name = "AgentCore";
-    private _state;
-    private _sessionId;
-    private _isConnected;
-    private asr;
-    private vad;
-    private lam;
-    private emotionController;
-    private pipeline;
-    private ws;
-    private wsReconnectAttempts;
-    private readonly maxReconnectAttempts;
-    private audioBuffer;
-    private history;
-    private currentConfig;
-    private agentCoreConfig;
-    private isSpeaking;
-    private currentTtsAbortController;
-    private tokenCache;
-    constructor(config: AgentCoreConfig);
-    get state(): AISessionState;
-    get sessionId(): string | null;
-    get isConnected(): boolean;
-    /**
-     * Connect to AgentCore with session configuration
-     */
-    connect(config: SessionConfig): Promise<void>;
-    /**
-     * Disconnect and cleanup
-     */
-    disconnect(): Promise<void>;
-    /**
-     * Push user audio for processing
-     */
-    pushAudio(audio: Int16Array | Float32Array): void;
-    /**
-     * Send text directly to AgentCore
-     */
-    sendText(text: string): Promise<void>;
-    /**
-     * Interrupt current AI response
-     */
-    interrupt(): void;
-    getHistory(): ConversationMessage[];
-    clearHistory(): void;
-    healthCheck(): Promise<boolean>;
-    private setState;
-    private getAuthToken;
-    private initASR;
-    private initLAM;
-    private initPipeline;
-    private connectWebSocket;
-    private handleAgentCoreMessage;
-    private scheduleTranscription;
-    /**
-     * Detect voice activity using Silero VAD
-     * Falls back to simple RMS if VAD not available
-     */
-    private detectVoiceActivity;
-    private base64ToArrayBuffer;
-    private addToHistory;
-    private handleDisconnect;
-}
-
-/**
- * Conversation Orchestrator
- *
- * Manages the conversation pipeline with AgentCore:
- * - Handles session lifecycle and tenant isolation
- * - Manages adapter events and state
- *
- * @category AI
- */
-
-/**
- * Orchestrator configuration
- */
-interface OrchestratorConfig {
-    /** AgentCore adapter config */
-    adapter: AgentCoreConfig;
-    /** Connection timeout in ms */
-    connectionTimeoutMs?: number;
-    /** Max retry attempts */
-    maxRetries?: number;
-}
-/**
- * Orchestrator events (extends AI adapter events)
- */
-interface OrchestratorEvents extends AIAdapterEvents {
-    'session.created': {
-        sessionId: string;
-        tenantId: string;
-    };
-    'session.ended': {
-        sessionId: string;
-        reason: string;
-    };
-}
-/**
- * Conversation Orchestrator
- */
-declare class ConversationOrchestrator extends EventEmitter<OrchestratorEvents> {
-    private config;
-    private adapter;
-    private sessions;
-    private tenants;
-    private healthCheckInterval;
-    private readonly HEALTH_CHECK_INTERVAL_MS;
-    constructor(config: OrchestratorConfig);
-    /**
-     * Register a tenant
-     */
-    registerTenant(tenant: TenantConfig): void;
-    /**
-     * Unregister a tenant
-     */
-    unregisterTenant(tenantId: string): void;
-    /**
-     * Get tenant config
-     */
-    getTenant(tenantId: string): TenantConfig | undefined;
-    /**
-     * Create a new conversation session for a tenant
-     */
-    createSession(tenantId: string, options?: Partial<SessionConfig>): Promise<ConversationSession>;
-    /**
-     * End a session
-     */
-    endSession(sessionId: string): Promise<void>;
-    /**
-     * Get session by ID
-     */
-    getSession(sessionId: string): ConversationSession | undefined;
-    /**
-     * Get all sessions for a tenant
-     */
-    getTenantSessions(tenantId: string): ConversationSession[];
-    /**
-     * Start health monitoring
-     */
-    startHealthMonitoring(): void;
-    /**
-     * Stop health monitoring
-     */
-    stopHealthMonitoring(): void;
-    /**
-     * Dispose all resources
-     */
-    dispose(): Promise<void>;
-    private generateSessionId;
-    private forwardAdapterEvents;
-    private performHealthCheck;
-}
-
-/**
- * Tenant Manager
- *
- * Handles multi-tenant isolation for the Omote Platform:
- * - Credential isolation per tenant
- * - Session scoping per tenant
- * - Quota management
- * - Token refresh
- *
- * @category AI
- */
-
-/**
- * Tenant quota configuration
- */
-interface TenantQuota {
-    /** Max concurrent sessions */
-    maxSessions: number;
-    /** Requests per minute */
-    requestsPerMinute: number;
-    /** Max tokens per conversation */
-    maxTokensPerConversation: number;
-    /** Max audio minutes per day */
-    maxAudioMinutesPerDay: number;
-}
-/**
- * Tenant usage tracking
- */
-interface TenantUsage {
-    /** Current active sessions */
-    currentSessions: number;
-    /** Requests in current minute */
-    requestsThisMinute: number;
-    /** Total tokens used */
-    tokensUsed: number;
-    /** Audio minutes used today */
-    audioMinutesToday: number;
-    /** Last reset timestamp */
-    lastMinuteReset: number;
-    /** Last daily reset timestamp */
-    lastDailyReset: number;
-}
-/**
- * Token refresh callback
- */
-type TokenRefreshCallback = () => Promise<string>;
-/**
- * Tenant Manager
- */
-declare class TenantManager {
-    private tenants;
-    private quotas;
-    private usage;
-    private tokenRefreshCallbacks;
-    /**
-     * Default quota for new tenants
-     */
-    static readonly DEFAULT_QUOTA: TenantQuota;
-    /**
-     * Register a tenant with quota
-     */
-    register(tenant: TenantConfig, quota?: TenantQuota, tokenRefreshCallback?: TokenRefreshCallback): void;
-    /**
-     * Unregister a tenant
-     */
-    unregister(tenantId: string): void;
-    /**
-     * Get tenant config
-     */
-    get(tenantId: string): TenantConfig | undefined;
-    /**
-     * Check if tenant exists
-     */
-    has(tenantId: string): boolean;
-    /**
-     * Get all tenant IDs
-     */
-    getTenantIds(): string[];
-    /**
-     * Check if tenant can create new session
-     */
-    canCreateSession(tenantId: string): boolean;
-    /**
-     * Check if tenant can make request
-     */
-    canMakeRequest(tenantId: string): boolean;
-    /**
-     * Check if tenant can use audio
-     */
-    canUseAudio(tenantId: string, minutes: number): boolean;
-    /**
-     * Increment session count
-     */
-    incrementSessions(tenantId: string): void;
-    /**
-     * Decrement session count
-     */
-    decrementSessions(tenantId: string): void;
-    /**
-     * Record a request
-     */
-    recordRequest(tenantId: string): void;
-    /**
-     * Record token usage
-     */
-    recordTokens(tenantId: string, tokens: number): void;
-    /**
-     * Record audio usage
-     */
-    recordAudioMinutes(tenantId: string, minutes: number): void;
-    /**
-     * Get fresh auth token for tenant
-     */
-    getAuthToken(tenantId: string): Promise<string>;
-    /**
-     * Update tenant credentials
-     */
-    updateCredentials(tenantId: string, credentials: Partial<TenantConfig['credentials']>): void;
-    /**
-     * Get usage stats for tenant
-     */
-    getUsage(tenantId: string): TenantUsage | undefined;
-    /**
-     * Get quota for tenant
-     */
-    getQuota(tenantId: string): TenantQuota | undefined;
-    /**
-     * Update quota for tenant
-     */
-    updateQuota(tenantId: string, quota: Partial<TenantQuota>): void;
-    /**
-     * Reset all usage stats for a tenant
-     */
-    resetUsage(tenantId: string): void;
-    private checkMinuteReset;
-    private checkDailyReset;
-}
-
-/**
- * Audio Sync Manager
- *
- * Synchronizes TTS audio playback with lip sync animation:
- * - Buffers audio for inference
- * - Manages playback timing
- * - Handles audio queue for streaming
- *
- * @category AI
- */
-
-/**
- * Audio sync events
- */
-interface AudioSyncEvents {
-    [key: string]: unknown;
-    'buffer.ready': {
-        audio: Float32Array;
-    };
-    'playback.start': Record<string, never>;
-    'playback.end': Record<string, never>;
-    'sync.drift': {
-        driftMs: number;
-    };
-}
-/**
- * Audio sync configuration
- */
-interface AudioSyncConfig {
-    /** Target sample rate (default: 16000) */
-    sampleRate?: number;
-    /** Buffer size for inference (default: 16640) */
-    bufferSize?: number;
-    /** Overlap between buffers (default: 4160) */
-    overlapSize?: number;
-    /** Max drift before correction (default: 100ms) */
-    maxDriftMs?: number;
-}
-/**
- * Audio Sync Manager
- */
-declare class AudioSyncManager extends EventEmitter<AudioSyncEvents> {
-    private config;
-    private audioBuffer;
-    private bufferPosition;
-    private playbackQueue;
-    private isPlaying;
-    private audioContext;
-    private playbackStartTime;
-    private samplesPlayed;
-    constructor(config?: AudioSyncConfig);
-    /**
-     * Initialize audio context
-     */
-    initialize(): Promise<void>;
-    /**
-     * Push audio chunk for processing and playback
-     */
-    pushAudio(audio: Float32Array): void;
-    /**
-     * Buffer audio for inference
-     */
-    private bufferForInference;
-    /**
-     * Start audio playback
-     */
-    private startPlayback;
-    /**
-     * Process playback queue
-     */
-    private processPlaybackQueue;
-    /**
-     * Check for audio/animation drift
-     */
-    private checkDrift;
-    /**
-     * Clear playback queue
-     */
-    clearQueue(): void;
-    /**
-     * Stop playback
-     */
-    stop(): void;
-    /**
-     * Get current playback position in seconds
-     */
-    getPlaybackPosition(): number;
-    /**
-     * Check if currently playing
-     */
-    getIsPlaying(): boolean;
-    /**
-     * Dispose resources
-     */
-    dispose(): void;
-}
-
-/**
- * Interruption Handler
- *
- * VAD-based interruption detection for AI conversations:
- * - Monitors user audio for speech
- * - Detects when user interrupts AI response
- * - Triggers interruption callbacks
- *
- * @category AI
- */
-
-/**
- * Interruption events
- */
-interface InterruptionEvents {
-    [key: string]: unknown;
-    'speech.detected': {
-        rms: number;
-    };
-    'speech.ended': {
-        durationMs: number;
-    };
-    'interruption.triggered': {
-        rms: number;
-        durationMs: number;
-    };
-}
-/**
- * Interruption handler configuration
- *
- * Industry standards applied:
- * - vadThreshold: 0.5 (Silero VAD default)
- * - minSpeechDurationMs: 200ms (Google/Amazon barge-in standard)
- * - silenceTimeoutMs: 500ms (OpenAI Realtime API standard)
- */
-interface InterruptionConfig {
-    /** VAD probability threshold for speech detection (default: 0.5, Silero standard) */
-    vadThreshold?: number;
-    /** Minimum speech duration to trigger interruption (default: 200ms, Google/Amazon standard) */
-    minSpeechDurationMs?: number;
-    /** Silence duration to end speech (default: 500ms, OpenAI standard) */
-    silenceTimeoutMs?: number;
-    /** Enable interruption detection (default: true) */
-    enabled?: boolean;
-}
-/**
- * Interruption Handler
- */
-declare class InterruptionHandler extends EventEmitter<InterruptionEvents> {
-    private config;
-    private isSpeaking;
-    private speechStartTime;
-    private lastSpeechTime;
-    private silenceTimer;
-    private aiIsSpeaking;
-    private interruptionTriggeredThisSession;
-    constructor(config?: InterruptionConfig);
-    /**
-     * Process VAD result for interruption detection
-     * @param vadProbability - Speech probability from VAD (0-1)
-     * @param audioEnergy - Optional RMS energy for logging (default: 0)
-     */
-    processVADResult(vadProbability: number, audioEnergy?: number): void;
-    /**
-     * Notify that AI started speaking
-     */
-    setAISpeaking(speaking: boolean): void;
-    /**
-     * Enable/disable interruption detection
-     */
-    setEnabled(enabled: boolean): void;
-    /**
-     * Update configuration
-     */
-    updateConfig(config: Partial<InterruptionConfig>): void;
-    /**
-     * Reset state
-     */
-    reset(): void;
-    /**
-     * Get current state
-     */
-    getState(): {
-        isSpeaking: boolean;
-        speechDurationMs: number;
-    };
-    private onSpeechDetected;
-    private onSilenceDetected;
-}
-
 /**
  * Model Cache
  *
@@ -4799,4 +4084,4 @@ declare class ProceduralLifeLayer {
     private updateBrowNoise;
 }
 
-export { type A2EBackend, type A2EModelInfo, A2EOrchestrator, type A2EOrchestratorConfig, A2EProcessor, type A2EProcessorConfig, type A2EProgressEvent, type A2EResult, type AIAdapter, type AIAdapterEvents, AISessionState, ARKIT_BLENDSHAPES, type ActiveSpan, AgentCoreAdapter, type AgentCoreConfig, type AnimationClip, AnimationEvent, AnimationGraph, type AnimationGraphConfig, type AnimationGraphEvents, type AnimationLayer, type AnimationOutput, type AnimationState, type AnimationStateName, type AnimationTrigger, AudioChunkCoalescer, type AudioChunkCoalescerOptions, AudioEnergyAnalyzer, AudioScheduler, type AudioSchedulerOptions, type AudioSyncConfig, type AudioSyncEvents, AudioSyncManager, BLENDSHAPE_TO_GROUP, type BackendPreference, type BlendWeight, type BlendshapeGroup, BlendshapeSmoother, type BlendshapeSmootherConfig, CTC_VOCAB, type CacheConfig, type CacheSpanAttributes, ConsoleExporter, type ConversationMessage, ConversationOrchestrator, type ConversationSession, type CreateA2EConfig, type CreateSenseVoiceConfig, DEFAULT_ANIMATION_CONFIG, EMOTION_NAMES, EMOTION_VECTOR_SIZE, type EmotionAnimationMap, EmotionController, type EmotionLabel, type EmotionName, type EmotionPresetName, EmotionPresets, type EmotionWeights, EmphasisDetector, EventEmitter, type ExpressionProfile, type FetchWithCacheOptions, type FullFaceFrame, FullFacePipeline, type FullFacePipelineEvents, type FullFacePipelineOptions, INFERENCE_LATENCY_BUCKETS, type InferenceSpanAttributes, type InterruptionConfig, type InterruptionEvents, InterruptionHandler, LAM_BLENDSHAPES, type LifeLayerConfig, type LifeLayerInput, type LifeLayerOutput, MODEL_LOAD_TIME_BUCKETS, type MessageRole, type MetricData, MetricNames, MicrophoneCapture, type MicrophoneCaptureConfig, ModelCache, type ModelSpanAttributes, OTLPExporter, type OTLPExporterConfig, OmoteEvents, OmoteTelemetry, type OrchestratorConfig, type OrchestratorEvents, ProceduralLifeLayer, type QuotaInfo, RingBuffer, type RuntimeBackend, type SafariSpeechConfig, SafariSpeechRecognition, type SamplingConfig, type SenseVoiceBackend, type SenseVoiceConfig, SenseVoiceInference, type SenseVoiceLanguage, type SenseVoiceModelInfo, type SenseVoiceResult, SenseVoiceUnifiedAdapter, SenseVoiceWorker, type SenseVoiceWorkerConfig, type SessionConfig, type SessionSnapshot, type SileroVADBackend, type SileroVADConfig, type SileroVADFactoryConfig, SileroVADInference, SileroVADUnifiedAdapter, SileroVADWorker, type SpanAttributes, type SpanData, type SpeechErrorCallback, type SpeechRecognitionResult, type SpeechResultCallback, type SpeechSegment, type TelemetryConfig, type TelemetryExporter, type TelemetryExporterInterface, type TenantConfig, TenantManager, type TenantQuota, type TenantUsage, type TokenRefreshCallback, type Transition, UnifiedInferenceWorker, type VADBackend, type VADModelInfo, type VADResult, type VADWorkerConfig, type VADWorkerModelInfo, type ValidationResult, type VoiceConfig, type Wav2ArkitCpuConfig, Wav2ArkitCpuInference, Wav2ArkitCpuUnifiedAdapter, Wav2ArkitCpuWorker, type Wav2ArkitCpuWorkerConfig, Wav2Vec2Inference, type Wav2Vec2InferenceConfig, type Wav2Vec2Result, blendEmotions, calculatePeak, calculateRMS, configureCacheLimit, configureTelemetry, createA2E, createEmotionVector, createSenseVoice, createSileroVAD, fetchWithCache, formatBytes, getCacheConfig, getCacheKey, getEmotionPreset, getModelCache, getOptimalWasmThreads, getRecommendedBackend, getTelemetry, hasWebGPUApi, isAndroid, isIOS, isIOSSafari, isMobile, isSafari, isSpeechRecognitionAvailable, isWebGPUAvailable, lerpBlendshapes, lerpEmotion, preloadModels, resolveBackend, shouldEnableWasmProxy, shouldUseCpuA2E, shouldUseNativeASR, shouldUseServerA2E, supportsVADWorker };
+export { type A2EBackend, type A2EModelInfo, A2EOrchestrator, type A2EOrchestratorConfig, A2EProcessor, type A2EProcessorConfig, type A2EProgressEvent, type A2EResult, ARKIT_BLENDSHAPES, type ActiveSpan, type AnimationClip, AnimationGraph, type AnimationGraphConfig, type AnimationGraphEvents, type AnimationLayer, type AnimationOutput, type AnimationState, type AnimationStateName, type AnimationTrigger, AudioChunkCoalescer, type AudioChunkCoalescerOptions, AudioEnergyAnalyzer, AudioScheduler, type AudioSchedulerOptions, BLENDSHAPE_TO_GROUP, type BackendPreference, type BlendWeight, type BlendshapeGroup, BlendshapeSmoother, type BlendshapeSmootherConfig, CTC_VOCAB, type CacheConfig, type CacheSpanAttributes, ConsoleExporter, type CreateA2EConfig, type CreateSenseVoiceConfig, DEFAULT_ANIMATION_CONFIG, EMOTION_NAMES, EMOTION_VECTOR_SIZE, type EmotionAnimationMap, EmotionController, type EmotionLabel, type EmotionName, type EmotionPresetName, EmotionPresets, type EmotionWeights, EmphasisDetector, EventEmitter, type ExpressionProfile, type FetchWithCacheOptions, type FullFaceFrame, FullFacePipeline, type FullFacePipelineEvents, type FullFacePipelineOptions, INFERENCE_LATENCY_BUCKETS, type InferenceSpanAttributes, type InterruptionConfig, type InterruptionEvents, InterruptionHandler, LAM_BLENDSHAPES, type LifeLayerConfig, type LifeLayerInput, type LifeLayerOutput, MODEL_LOAD_TIME_BUCKETS, type MetricData, MetricNames, MicrophoneCapture, type MicrophoneCaptureConfig, ModelCache, type ModelSpanAttributes, OTLPExporter, type OTLPExporterConfig, OmoteEvents, OmoteTelemetry, ProceduralLifeLayer, type QuotaInfo, RingBuffer, type RuntimeBackend, type SafariSpeechConfig, SafariSpeechRecognition, type SamplingConfig, type SenseVoiceBackend, type SenseVoiceConfig, SenseVoiceInference, type SenseVoiceLanguage, type SenseVoiceModelInfo, type SenseVoiceResult, SenseVoiceUnifiedAdapter, SenseVoiceWorker, type SenseVoiceWorkerConfig, type SileroVADBackend, type SileroVADConfig, type SileroVADFactoryConfig, SileroVADInference, SileroVADUnifiedAdapter, SileroVADWorker, type SpanAttributes, type SpanData, type SpeechErrorCallback, type SpeechRecognitionResult, type SpeechResultCallback, type SpeechSegment, type TelemetryConfig, type TelemetryExporter, type TelemetryExporterInterface, type Transition, UnifiedInferenceWorker, type VADBackend, type VADModelInfo, type VADResult, type VADWorkerConfig, type VADWorkerModelInfo, type ValidationResult, type Wav2ArkitCpuConfig, Wav2ArkitCpuInference, Wav2ArkitCpuUnifiedAdapter, Wav2ArkitCpuWorker, type Wav2ArkitCpuWorkerConfig, Wav2Vec2Inference, type Wav2Vec2InferenceConfig, type Wav2Vec2Result, blendEmotions, calculatePeak, calculateRMS, configureCacheLimit, configureTelemetry, createA2E, createEmotionVector, createSenseVoice, createSileroVAD, fetchWithCache, formatBytes, getCacheConfig, getCacheKey, getEmotionPreset, getModelCache, getOptimalWasmThreads, getRecommendedBackend, getTelemetry, hasWebGPUApi, isAndroid, isIOS, isIOSSafari, isMobile, isSafari, isSpeechRecognitionAvailable, isWebGPUAvailable, lerpBlendshapes, lerpEmotion, preloadModels, resolveBackend, shouldEnableWasmProxy, shouldUseCpuA2E, shouldUseNativeASR, shouldUseServerA2E, supportsVADWorker };