@speechos/core 0.2.9 → 0.2.11

package/dist/websocket.d.ts

@@ -1,8 +1,8 @@
 /**
  * WebSocket integration for SpeechOS SDK.
  *
- * Provides a direct WebSocket connection to the backend for voice sessions,
- * bypassing LiveKit for lower latency. Uses audio buffering to capture
+ * Provides a direct WebSocket connection to the backend for voice sessions.
+ * Uses audio buffering to capture
  * audio immediately while the connection is being established.
  */
 import type { CommandDefinition, CommandResult, ErrorSource, VoiceSessionOptions } from './types.js';
@@ -92,8 +92,9 @@ declare class WebSocketManager {
     /**
      * Request command matching using the transcript as input.
      * Note: The command definitions were already sent in the auth message via startVoiceSession.
+     * Returns an array of matched commands (empty array if no matches).
      */
-    requestCommand(_commands: CommandDefinition[]): Promise<CommandResult | null>;
+    requestCommand(_commands: CommandDefinition[]): Promise<CommandResult[]>;
     /**
      * Stop audio capture and wait for all data to be sent.
      *
@@ -108,7 +109,7 @@ declare class WebSocketManager {
      * Wait for the WebSocket send buffer to drain.
      *
      * This ensures all audio data has been transmitted before we request
-     * the transcript. Uses the same pattern as LiveKit's ReadableStream approach.
+     * the transcript.
      */
     private waitForBufferDrain;
     /**

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@speechos/core",
-  "version": "0.2.9",
-  "description": "Headless core SDK for SpeechOS - state, events, LiveKit integration",
+  "version": "0.2.11",
+  "description": "Headless core SDK for SpeechOS - state, events, WebSocket integration",
   "type": "module",
   "main": "./dist/index.cjs",
   "module": "./dist/index.js",
@@ -52,8 +52,5 @@
     "tslib": "^2.8.1",
     "typescript": "^5.7.3",
     "vitest": "^4.0.16"
-  },
-  "dependencies": {
-    "livekit-client": "^2.16.1"
   }
 }

package/dist/livekit.d.cts

@@ -1,199 +0,0 @@
-/**
- * LiveKit integration for SpeechOS SDK
- * Handles room connections, audio streaming, and transcription requests
- */
-import { Room } from "livekit-client";
-import type { LiveKitTokenResponse, ErrorSource, CommandDefinition, CommandResult, VoiceSessionOptions } from "./types.js";
-/**
- * A deferred promise with timeout support.
- * Encapsulates resolve/reject/timeout in a single object for cleaner async handling.
- */
-export declare class Deferred<T> {
-    readonly promise: Promise<T>;
-    private _resolve;
-    private _reject;
-    private _timeoutId;
-    private _settled;
-    constructor();
-    /**
-     * Set a timeout that will reject the promise with the given error
-     */
-    setTimeout(ms: number, errorMessage: string, errorCode: string, errorSource: ErrorSource): void;
-    resolve(value: T): void;
-    reject(error: Error): void;
-    private clearTimeout;
-    get isSettled(): boolean;
-}
-/**
- * LiveKit connection manager
- */
-declare class LiveKitManager {
-    private room;
-    private tokenData;
-    private micTrack;
-    private cachedTokenData;
-    private tokenCacheTimestamp;
-    private tokenPrefetchPromise;
-    private tokenRefreshTimer;
-    private autoRefreshEnabled;
-    private pendingTranscript;
-    private pendingEditText;
-    private pendingCommand;
-    private pendingTrackSubscribed;
-    private editOriginalText;
-    private sessionSettings;
-    /**
-     * Check if the cached token is still valid (within TTL)
-     */
-    private isCachedTokenValid;
-    /**
-     * Pre-fetch a LiveKit token for later use
-     * Call this early (e.g., when widget expands) to reduce latency when starting a voice session.
-     * If a prefetch is already in progress, returns the existing promise.
-     * If a valid cached token exists, returns it immediately.
-     */
-    prefetchToken(): Promise<LiveKitTokenResponse>;
-    /**
-     * Fetch a LiveKit token from the backend
-     * Uses cached token if valid, otherwise fetches a fresh one.
-     * Includes language settings and user vocabulary which are stored in the VoiceSession.
-     */
-    fetchToken(): Promise<LiveKitTokenResponse>;
-    /**
-     * Internal method to fetch a fresh token from the server
-     */
-    private fetchTokenFromServer;
-    /**
-     * Connect to a LiveKit room (fresh connection each time)
-     */
-    connect(): Promise<Room>;
-    /**
-     * Wait until the agent is ready to receive audio
-     * Resolves when LocalTrackSubscribed event is received
-     */
-    waitUntilReady(): Promise<void>;
-    /**
-     * Set up LiveKit room event listeners
-     */
-    private setupRoomEvents;
-    /**
-     * Handle incoming data messages from the agent
-     */
-    private handleDataMessage;
-    /**
-     * Publish microphone audio track
-     * Uses the device ID from session settings if set
-     */
-    enableMicrophone(): Promise<void>;
-    /**
-     * Log information about the current microphone track
-     */
-    private logMicrophoneInfo;
-    /**
-     * Disable microphone audio track
-     */
-    disableMicrophone(): Promise<void>;
-    /**
-     * Send a data message to the room
-     */
-    sendDataMessage(message: object): Promise<void>;
-    /**
-     * Start a voice session with pre-connect audio buffering
-     * Fetches a fresh token, then enables mic with preConnectBuffer to capture audio while connecting.
-     * Agent subscription happens in the background - we don't block on it.
-     *
-     * @param options - Session options including action type and parameters
-     */
-    startVoiceSession(options?: VoiceSessionOptions): Promise<void>;
-    /**
-     * Wait for the agent to subscribe to our audio track in the background
-     * Handles timeout errors without blocking the main flow
-     */
-    private waitForAgentSubscription;
-    /**
-     * Enable microphone with pre-connect buffering
-     * This starts capturing audio locally before the room is connected,
-     * buffering it until the connection is established.
-     */
-    private enableMicrophoneWithPreConnectBuffer;
-    /**
-     * Stop the voice session and request the transcript
-     * Returns a promise that resolves with the transcript text
-     * @throws Error if timeout occurs waiting for transcript
-     */
-    stopVoiceSession(): Promise<string>;
-    /**
-     * Alias for stopVoiceSession - granular API naming
-     */
-    stopAndGetTranscript(): Promise<string>;
-    /**
-     * Request text editing using the transcript as instructions
-     * Sends the original text to the backend, which applies the spoken instructions
-     * Returns a promise that resolves with the edited text
-     * @throws Error if timeout occurs waiting for edited text
-     */
-    requestEditText(originalText: string): Promise<string>;
-    /**
-     * Alias for requestEditText - granular API naming
-     */
-    stopAndEdit(originalText: string): Promise<string>;
-    /**
-     * Request command matching using the transcript as input
-     * Sends command definitions to the backend, which matches the user's speech against them
-     * Returns a promise that resolves with the matched command or null if no match
-     * @throws Error if timeout occurs waiting for command result
-     */
-    requestCommand(commands: CommandDefinition[]): Promise<CommandResult | null>;
-    /**
-     * Alias for requestCommand - granular API naming
-     */
-    stopAndCommand(commands: CommandDefinition[]): Promise<CommandResult | null>;
-    /**
-     * Disconnect from the current room
-     * Clears the token so a fresh one is fetched for the next session
-     */
-    disconnect(): Promise<void>;
-    /**
-     * Invalidate the cached token
-     * Call this when settings change that would affect the token (language, vocabulary)
-     */
-    invalidateTokenCache(): void;
-    /**
-     * Start auto-refreshing the token while the widget is expanded.
-     * Call this after a voice session completes to immediately fetch a fresh token
-     * (since each command requires its own token) and keep it fresh for subsequent commands.
-     */
-    startAutoRefresh(): void;
-    /**
-     * Stop auto-refreshing the token.
-     * Call this when the widget collapses or user navigates away.
-     */
-    stopAutoRefresh(): void;
-    /**
-     * Schedule a token refresh before the current cache expires.
-     * Handles computer sleep by checking elapsed time on each refresh attempt.
-     */
-    private scheduleTokenRefresh;
-    /**
-     * Perform the auto-refresh, handling computer sleep scenarios.
-     */
-    private performAutoRefresh;
-    /**
-     * Get the current room instance
-     */
-    getRoom(): Room | null;
-    /**
-     * Get the current token data
-     */
-    getTokenData(): LiveKitTokenResponse | null;
-    /**
-     * Check if connected to a room
-     */
-    isConnected(): boolean;
-    /**
-     * Check if microphone is enabled
-     */
-    isMicrophoneEnabled(): boolean;
-}
-export declare const livekit: LiveKitManager;
-export {};

package/dist/livekit.d.ts

@@ -1,199 +0,0 @@
-/**
- * LiveKit integration for SpeechOS SDK
- * Handles room connections, audio streaming, and transcription requests
- */
-import { Room } from "livekit-client";
-import type { LiveKitTokenResponse, ErrorSource, CommandDefinition, CommandResult, VoiceSessionOptions } from "./types.js";
-/**
- * A deferred promise with timeout support.
- * Encapsulates resolve/reject/timeout in a single object for cleaner async handling.
- */
-export declare class Deferred<T> {
-    readonly promise: Promise<T>;
-    private _resolve;
-    private _reject;
-    private _timeoutId;
-    private _settled;
-    constructor();
-    /**
-     * Set a timeout that will reject the promise with the given error
-     */
-    setTimeout(ms: number, errorMessage: string, errorCode: string, errorSource: ErrorSource): void;
-    resolve(value: T): void;
-    reject(error: Error): void;
-    private clearTimeout;
-    get isSettled(): boolean;
-}
-/**
- * LiveKit connection manager
- */
-declare class LiveKitManager {
-    private room;
-    private tokenData;
-    private micTrack;
-    private cachedTokenData;
-    private tokenCacheTimestamp;
-    private tokenPrefetchPromise;
-    private tokenRefreshTimer;
-    private autoRefreshEnabled;
-    private pendingTranscript;
-    private pendingEditText;
-    private pendingCommand;
-    private pendingTrackSubscribed;
-    private editOriginalText;
-    private sessionSettings;
-    /**
-     * Check if the cached token is still valid (within TTL)
-     */
-    private isCachedTokenValid;
-    /**
-     * Pre-fetch a LiveKit token for later use
-     * Call this early (e.g., when widget expands) to reduce latency when starting a voice session.
-     * If a prefetch is already in progress, returns the existing promise.
-     * If a valid cached token exists, returns it immediately.
-     */
-    prefetchToken(): Promise<LiveKitTokenResponse>;
-    /**
-     * Fetch a LiveKit token from the backend
-     * Uses cached token if valid, otherwise fetches a fresh one.
-     * Includes language settings and user vocabulary which are stored in the VoiceSession.
-     */
-    fetchToken(): Promise<LiveKitTokenResponse>;
-    /**
-     * Internal method to fetch a fresh token from the server
-     */
-    private fetchTokenFromServer;
-    /**
-     * Connect to a LiveKit room (fresh connection each time)
-     */
-    connect(): Promise<Room>;
-    /**
-     * Wait until the agent is ready to receive audio
-     * Resolves when LocalTrackSubscribed event is received
-     */
-    waitUntilReady(): Promise<void>;
-    /**
-     * Set up LiveKit room event listeners
-     */
-    private setupRoomEvents;
-    /**
-     * Handle incoming data messages from the agent
-     */
-    private handleDataMessage;
-    /**
-     * Publish microphone audio track
-     * Uses the device ID from session settings if set
-     */
-    enableMicrophone(): Promise<void>;
-    /**
-     * Log information about the current microphone track
-     */
-    private logMicrophoneInfo;
-    /**
-     * Disable microphone audio track
-     */
-    disableMicrophone(): Promise<void>;
-    /**
-     * Send a data message to the room
-     */
-    sendDataMessage(message: object): Promise<void>;
-    /**
-     * Start a voice session with pre-connect audio buffering
-     * Fetches a fresh token, then enables mic with preConnectBuffer to capture audio while connecting.
-     * Agent subscription happens in the background - we don't block on it.
-     *
-     * @param options - Session options including action type and parameters
-     */
-    startVoiceSession(options?: VoiceSessionOptions): Promise<void>;
-    /**
-     * Wait for the agent to subscribe to our audio track in the background
-     * Handles timeout errors without blocking the main flow
-     */
-    private waitForAgentSubscription;
-    /**
-     * Enable microphone with pre-connect buffering
-     * This starts capturing audio locally before the room is connected,
-     * buffering it until the connection is established.
-     */
-    private enableMicrophoneWithPreConnectBuffer;
-    /**
-     * Stop the voice session and request the transcript
-     * Returns a promise that resolves with the transcript text
-     * @throws Error if timeout occurs waiting for transcript
-     */
-    stopVoiceSession(): Promise<string>;
-    /**
-     * Alias for stopVoiceSession - granular API naming
-     */
-    stopAndGetTranscript(): Promise<string>;
-    /**
-     * Request text editing using the transcript as instructions
-     * Sends the original text to the backend, which applies the spoken instructions
-     * Returns a promise that resolves with the edited text
-     * @throws Error if timeout occurs waiting for edited text
-     */
-    requestEditText(originalText: string): Promise<string>;
-    /**
-     * Alias for requestEditText - granular API naming
-     */
-    stopAndEdit(originalText: string): Promise<string>;
-    /**
-     * Request command matching using the transcript as input
-     * Sends command definitions to the backend, which matches the user's speech against them
-     * Returns a promise that resolves with the matched command or null if no match
-     * @throws Error if timeout occurs waiting for command result
-     */
-    requestCommand(commands: CommandDefinition[]): Promise<CommandResult | null>;
-    /**
-     * Alias for requestCommand - granular API naming
-     */
-    stopAndCommand(commands: CommandDefinition[]): Promise<CommandResult | null>;
-    /**
-     * Disconnect from the current room
-     * Clears the token so a fresh one is fetched for the next session
-     */
-    disconnect(): Promise<void>;
-    /**
-     * Invalidate the cached token
-     * Call this when settings change that would affect the token (language, vocabulary)
-     */
-    invalidateTokenCache(): void;
-    /**
-     * Start auto-refreshing the token while the widget is expanded.
-     * Call this after a voice session completes to immediately fetch a fresh token
-     * (since each command requires its own token) and keep it fresh for subsequent commands.
-     */
-    startAutoRefresh(): void;
-    /**
-     * Stop auto-refreshing the token.
-     * Call this when the widget collapses or user navigates away.
-     */
-    stopAutoRefresh(): void;
-    /**
-     * Schedule a token refresh before the current cache expires.
-     * Handles computer sleep by checking elapsed time on each refresh attempt.
-     */
-    private scheduleTokenRefresh;
-    /**
-     * Perform the auto-refresh, handling computer sleep scenarios.
-     */
-    private performAutoRefresh;
-    /**
-     * Get the current room instance
-     */
-    getRoom(): Room | null;
-    /**
-     * Get the current token data
-     */
-    getTokenData(): LiveKitTokenResponse | null;
-    /**
-     * Check if connected to a room
-     */
-    isConnected(): boolean;
-    /**
-     * Check if microphone is enabled
-     */
-    isMicrophoneEnabled(): boolean;
-}
-export declare const livekit: LiveKitManager;
-export {};