feat: Introduce new core components for conversation and speech management

- Added ConversationManager for managing conversation history with configurable limits.
- Implemented InputQueue for serial processing of input items.
- Created SpeechManager for handling text-to-speech generation and streaming.
- Developed StreamProcessor for processing LLM streams and forwarding events.
- Added TranscriptionManager for audio transcription using AI SDK.
- Introduced WebSocketManager for managing WebSocket connections and messaging.
- Updated VoiceAgent to support new architecture and improved socket handling.
- Refactored index files to export new core components.

dist/core/InputQueue.d.ts

@@ -0,0 +1,33 @@
+/**
+ * A generic serial input queue that ensures only one processor runs at a time.
+ *
+ * @template T  The shape of each queued item (must include resolve/reject)
+ */
+export interface QueueItem<T = string> {
+    resolve: (v: T) => void;
+    reject: (e: unknown) => void;
+}
+export declare class InputQueue<T extends QueueItem<any>> {
+    private queue;
+    private processing;
+    /** Callback invoked for each item — must return a resolved value */
+    processor: (item: T) => Promise<any>;
+    /**
+     * Enqueue an item for serial processing.
+     */
+    enqueue(item: T): void;
+    /**
+     * Reject all pending items (used on disconnect/destroy).
+     */
+    rejectAll(reason: Error): void;
+    /**
+     * Number of items waiting in the queue.
+     */
+    get length(): number;
+    /**
+     * Whether the queue is currently processing an item.
+     */
+    get isProcessing(): boolean;
+    private drain;
+}
+//# sourceMappingURL=InputQueue.d.ts.map