@nextpa/chat-core 0.1.0

package/dist/services/realtime/types.d.ts

@@ -0,0 +1,78 @@
+import type { FunctionCall, ToolDefinition } from '../../types';
+export type RealtimeProviderType = 'openai' | 'google' | 'vertexai';
+export type ProviderError = {
+    readonly code?: string;
+    readonly message: string;
+};
+export type ProviderUsage = {
+    readonly inputTokens: number;
+    readonly outputTokens: number;
+    readonly totalTokens: number;
+    readonly cachedTokens?: number;
+    readonly textCachedInputTokens?: number;
+    readonly audioCachedInputTokens?: number;
+    readonly audioInputTokens?: number;
+    readonly audioOutputTokens?: number;
+    readonly textInputTokens?: number;
+    readonly textOutputTokens?: number;
+};
+export type ProviderEventMap = {
+    sessionReady: () => void;
+    connectionStateChanged: (state: 'connecting' | 'connected' | 'disconnected') => void;
+    userSpeechStarted: () => void;
+    userSpeechStopped: () => void;
+    userSpeechEnded: () => void;
+    /** Normalised microphone input level (0..1), emitted continuously while voice is connected. */
+    userAudioLevel: (level: number) => void;
+    userTranscriptDelta: (delta: string) => void;
+    userTranscriptCompleted: (transcript: string) => void;
+    assistantTranscriptDelta: (delta: string) => void;
+    assistantTranscriptDone: (transcript: string) => void;
+    audioDone: () => void;
+    functionCall: (call: FunctionCall) => void;
+    usageUpdated: (usage: ProviderUsage) => void;
+    error: (error: ProviderError) => void;
+    interrupted: () => void;
+    serverDisconnecting: (timeLeft?: string) => void;
+};
+export type TranscriptionConfig = {
+    mode: 'oob' | 'native';
+    model?: string;
+    language: string;
+    prompt?: string;
+};
+export type RealtimeProviderConnectConfig = {
+    apiKey: string;
+    model: string;
+    instructions: string;
+    voice: string;
+    tools: ToolDefinition[];
+    toolChoice: string;
+    speed?: number;
+    transcription?: TranscriptionConfig;
+    proxyUrl?: string;
+    proxyBearerToken?: string;
+};
+export type RealtimeProviderClient = {
+    readonly status: 'disconnected' | 'connecting' | 'connected';
+    connect(config: RealtimeProviderConnectConfig): Promise<void>;
+    close(): void;
+    mute(muted: boolean): void;
+    /** Stop the assistant talking now: cancel the in-flight response and drop any queued audio. */
+    interrupt(): void;
+    requestResponse(): void;
+    sendToolResult(callId: string, name: string, output: string): void;
+    sendText(text: string): void;
+    waitForAudioPlaybackEnd(timeoutMs?: number): Promise<string>;
+    refreshAudioDevice(): Promise<void>;
+    /**
+     * Re-arm audio after the page returns to the foreground. iOS suspends
+     * AudioContexts and pauses the playback element when an installed PWA is
+     * backgrounded; on return they must be explicitly resumed/replayed or the
+     * call goes silent and the mic-level indicator freezes. No-op where the
+     * provider has nothing suspended.
+     */
+    resumeAudio?(): void;
+    on<K extends keyof ProviderEventMap>(event: K, handler: ProviderEventMap[K]): () => void;
+    off<K extends keyof ProviderEventMap>(event: K, handler: ProviderEventMap[K]): void;
+};

package/dist/services/realtime/types.js

@@ -0,0 +1 @@
+export {};

package/dist/services/realtime/vertexai/index.d.ts

@@ -0,0 +1 @@
+export { VertexAiRealtimeProvider } from './vertexai-provider';

package/dist/services/realtime/vertexai/index.js

@@ -0,0 +1 @@
+export { VertexAiRealtimeProvider } from './vertexai-provider';

package/dist/services/realtime/vertexai/vertexai-provider.d.ts

@@ -0,0 +1,53 @@
+import type { RealtimeProviderClient, RealtimeProviderConnectConfig, ProviderEventMap } from '../types';
+export declare class VertexAiRealtimeProvider implements RealtimeProviderClient {
+    private ws;
+    private readonly audioHandler;
+    private readonly eventHandlers;
+    private _status;
+    private readonly pendingFunctionCalls;
+    private awaitingToolResponse;
+    private isSetupComplete;
+    private modelResponding;
+    private userSpeaking;
+    private pendingCommands;
+    private outputTranscriptBuffer;
+    private inputTranscriptBuffer;
+    private lastCompletedToolCalls;
+    private sentFrames;
+    private droppedFrames;
+    constructor();
+    get status(): 'disconnected' | 'connecting' | 'connected';
+    connect(config: RealtimeProviderConnectConfig): Promise<void>;
+    private handleMicFrame;
+    private describeDropReason;
+    close(): void;
+    resumeAudio(): void;
+    mute(muted: boolean): void;
+    interrupt(): void;
+    requestResponse(): void;
+    sendToolResult(callId: string, name: string, output: string): void;
+    sendText(text: string): void;
+    waitForAudioPlaybackEnd(timeoutMs?: number): Promise<string>;
+    refreshAudioDevice(): Promise<void>;
+    on<K extends keyof ProviderEventMap>(event: K, handler: ProviderEventMap[K]): () => void;
+    off<K extends keyof ProviderEventMap>(event: K, handler: ProviderEventMap[K]): void;
+    private sendJson;
+    private sendSetupMessage;
+    private emit;
+    private resetSessionState;
+    private executeRequestResponse;
+    private executeSendToolResult;
+    private finalizeInputTranscript;
+    private handleServerMessage;
+    private extractAndEmitUsage;
+    private flushPendingCommands;
+    private handleVoiceActivity;
+    private handleServerContent;
+    private static readonly THINKING_RENDER_DELAY_MS;
+    private deferAssistantTranscript;
+    private ensureSpeechStopped;
+    private handleModelTurn;
+    private handleTurnComplete;
+    private handleInterrupted;
+    private handleToolCall;
+}