realtime-voice-agents 2.0.0

package/dist/testing.d.cts

@@ -0,0 +1,191 @@
+import { a as GeminiLiveSessionLike, n as GeminiLiveConnector, t as GeminiConnectParams } from "./GeminiLiveProvider-DvkTgzjG.cjs";
+import { i as WebSocketLike } from "./transport-B_PJFIVd.cjs";
+import { IncomingHttpHeaders } from "node:http";
+import { WebSocket } from "ws";
+//#region src/testing/FakeTwilioMediaStream.d.ts
+type MessageListener = (data: string, isBinary?: boolean) => void;
+type CloseListener = (code?: number, reason?: string) => void;
+type ErrorListener = (error: Error) => void;
+interface FakeConnectOptions {
+  callSid?: string;
+  streamSid?: string;
+  accountSid?: string;
+  customParameters?: Record<string, string>;
+  /** Send Twilio's `connected` preamble frame before `start`. Default true. */
+  sendConnectedFrame?: boolean;
+}
+/** Base64 μ-law silence of the given duration (whole milliseconds at 8 kHz). */
+declare function mulawSilenceBase64(ms: number): string;
+declare class FakeTwilioMediaStream implements WebSocketLike {
+  readyState: number;
+  readonly callSid: string;
+  readonly streamSid: string;
+  /** Everything the bridge sent, parsed, in order. */
+  readonly outbound: Array<Record<string, any>>;
+  /** Base64 payloads of outbound media frames, in order. */
+  readonly sentMediaPayloads: string[];
+  /** Names of marks echoed back to the bridge (played or discarded). */
+  readonly echoedMarks: string[];
+  /** Number of `clear` frames received from the bridge. */
+  clearCount: number;
+  private readonly messageListeners;
+  private readonly closeListeners;
+  private readonly errorListeners;
+  private readonly playoutQueue;
+  private playedMsTotal;
+  private callerClockMs;
+  private mediaSequence;
+  private closedByBridge;
+  constructor(options?: {
+    callSid?: string;
+    streamSid?: string;
+  });
+  send(data: string): void;
+  close(code?: number, reason?: string): void;
+  on(event: 'message', listener: MessageListener): this;
+  on(event: 'close', listener: CloseListener): this;
+  on(event: 'error', listener: ErrorListener): this;
+  /** Send the `connected` preamble and `start` frame. */
+  connect(options?: FakeConnectOptions): void;
+  /** Deliver one caller media frame (base64 μ-law) to the bridge. */
+  sendMedia(payload: string): void;
+  /** Deliver `ms` of caller silence as 20 ms μ-law frames. */
+  sendSilence(ms: number): void;
+  sendDtmf(digit: string): void;
+  /** Deliver the `stop` frame (Twilio ended the stream). */
+  stop(): void;
+  /** Simulate the socket dropping from Twilio's side. */
+  disconnect(code?: number, reason?: string): void;
+  /** Total milliseconds of agent audio actually "played" to the caller. */
+  get playedMs(): number;
+  /** Milliseconds of agent audio still buffered. */
+  get queuedMs(): number;
+  get wasClosedByBridge(): boolean;
+  /**
+   * Advance the playout clock: consume buffered media in order and echo each
+   * mark as playout reaches it.
+   */
+  advancePlayback(ms: number): void;
+  /** Play out everything currently buffered. */
+  playAll(): void;
+  private echoMark;
+  private deliver;
+}
+//#endregion
+//#region src/testing/FakeOpenAIServer.d.ts
+interface FakeOpenAIServerOptions {
+  /** Auto-ack session.update with session.updated. Default true. */
+  autoAckSessionUpdate?: boolean;
+}
+interface FakeAudioResponseOptions {
+  responseId?: string;
+  itemId?: string;
+  /** Base64 μ-law chunks to stream as output_audio deltas. */
+  chunks: string[];
+  transcript?: string;
+  usage?: Record<string, unknown>;
+  /** Emit response.done after the deltas. Default true. */
+  complete?: boolean;
+}
+declare class FakeOpenAIConnection {
+  readonly socket: WebSocket;
+  /** Every parsed client → server frame, in order. */
+  readonly received: Array<Record<string, any>>;
+  /** HTTP headers of the client's upgrade request (auth, extensions…). */
+  readonly upgradeHeaders: IncomingHttpHeaders;
+  private readonly server;
+  constructor(server: FakeOpenAIServer, socket: WebSocket, upgradeHeaders?: IncomingHttpHeaders);
+  /** Base64 payloads from input_audio_buffer.append frames. */
+  get appendedAudio(): string[];
+  eventsOfType(type: string): Array<Record<string, any>>;
+  waitForEvent(predicate: string | ((frame: Record<string, any>) => boolean), timeoutMs?: number): Promise<Record<string, any>>;
+  send(frame: Record<string, unknown>): void;
+  /** Script a complete assistant audio response (created → deltas → done). */
+  sendAudioResponse(options: FakeAudioResponseOptions): {
+    responseId: string;
+    itemId: string;
+  };
+  sendToolCall(options: {
+    name: string;
+    argumentsJson?: string;
+    callId?: string;
+    responseId?: string;
+  }): string;
+  sendSpeechStarted(): void;
+  sendSpeechStopped(): void;
+  sendUserTranscript(text: string): void;
+  sendError(error: Record<string, unknown>): void;
+  drop(code?: number, reason?: string): void;
+}
+declare class FakeOpenAIServer {
+  readonly connections: FakeOpenAIConnection[];
+  private readonly wss;
+  private readonly options;
+  readonly url: string;
+  private constructor();
+  static start(options?: FakeOpenAIServerOptions): Promise<FakeOpenAIServer>;
+  get latest(): FakeOpenAIConnection;
+  waitForConnection(timeoutMs?: number): Promise<FakeOpenAIConnection>;
+  close(): Promise<void>;
+}
+//#endregion
+//#region src/testing/FakeGeminiLive.d.ts
+declare class FakeGeminiSession implements GeminiLiveSessionLike {
+  readonly params: GeminiConnectParams;
+  readonly realtimeInputs: Array<{
+    data: string;
+    mimeType: string;
+  }>;
+  readonly clientContents: Array<{
+    turns: Array<Record<string, any>>;
+    turnComplete?: boolean;
+  }>;
+  readonly toolResponses: Array<{
+    functionResponses: Array<Record<string, any>>;
+  }>;
+  closed: boolean;
+  constructor(params: GeminiConnectParams);
+  sendRealtimeInput(input: {
+    audio: {
+      data: string;
+      mimeType: string;
+    };
+  }): void;
+  sendClientContent(content: {
+    turns: Array<Record<string, any>>;
+    turnComplete?: boolean;
+  }): void;
+  sendToolResponse(response: {
+    functionResponses: Array<Record<string, any>>;
+  }): void;
+  close(): void;
+  serverMessage(message: Record<string, unknown>): void;
+  completeSetup(): void;
+  /** Stream a model turn: PCM24k audio parts (+ transcript), then turnComplete. */
+  sendAudioTurn(options: {
+    pcm24kBase64Chunks: string[];
+    transcript?: string;
+  }): void;
+  sendToolCall(options: {
+    name: string;
+    args?: Record<string, unknown>;
+    id?: string;
+  }): string;
+  sendInterrupted(): void;
+  sendInputTranscription(text: string): void;
+  sendResumptionUpdate(handle: string): void;
+  sendGoAway(timeLeft: string): void;
+  sendUsage(usage: Record<string, unknown>): void;
+  /** Drop the connection from the server side. */
+  drop(code?: number, reason?: string): void;
+}
+declare class FakeGeminiLive {
+  readonly sessions: FakeGeminiSession[];
+  /** Auto-send setupComplete on connect. Default true. */
+  autoCompleteSetup: boolean;
+  readonly connector: GeminiLiveConnector;
+  get latest(): FakeGeminiSession;
+  waitForSessions(count: number, timeoutMs?: number): Promise<FakeGeminiSession>;
+}
+//#endregion
+export { type FakeAudioResponseOptions, type FakeConnectOptions, FakeGeminiLive, FakeGeminiSession, FakeOpenAIConnection, FakeOpenAIServer, type FakeOpenAIServerOptions, FakeTwilioMediaStream, mulawSilenceBase64 };

package/dist/testing.d.mts

@@ -0,0 +1,191 @@
+import { a as GeminiLiveSessionLike, n as GeminiLiveConnector, t as GeminiConnectParams } from "./GeminiLiveProvider-x2nyx5aO.mjs";
+import { i as WebSocketLike } from "./transport-CEaLFV4E.mjs";
+import { WebSocket } from "ws";
+import { IncomingHttpHeaders } from "node:http";
+//#region src/testing/FakeTwilioMediaStream.d.ts
+type MessageListener = (data: string, isBinary?: boolean) => void;
+type CloseListener = (code?: number, reason?: string) => void;
+type ErrorListener = (error: Error) => void;
+interface FakeConnectOptions {
+  callSid?: string;
+  streamSid?: string;
+  accountSid?: string;
+  customParameters?: Record<string, string>;
+  /** Send Twilio's `connected` preamble frame before `start`. Default true. */
+  sendConnectedFrame?: boolean;
+}
+/** Base64 μ-law silence of the given duration (whole milliseconds at 8 kHz). */
+declare function mulawSilenceBase64(ms: number): string;
+declare class FakeTwilioMediaStream implements WebSocketLike {
+  readyState: number;
+  readonly callSid: string;
+  readonly streamSid: string;
+  /** Everything the bridge sent, parsed, in order. */
+  readonly outbound: Array<Record<string, any>>;
+  /** Base64 payloads of outbound media frames, in order. */
+  readonly sentMediaPayloads: string[];
+  /** Names of marks echoed back to the bridge (played or discarded). */
+  readonly echoedMarks: string[];
+  /** Number of `clear` frames received from the bridge. */
+  clearCount: number;
+  private readonly messageListeners;
+  private readonly closeListeners;
+  private readonly errorListeners;
+  private readonly playoutQueue;
+  private playedMsTotal;
+  private callerClockMs;
+  private mediaSequence;
+  private closedByBridge;
+  constructor(options?: {
+    callSid?: string;
+    streamSid?: string;
+  });
+  send(data: string): void;
+  close(code?: number, reason?: string): void;
+  on(event: 'message', listener: MessageListener): this;
+  on(event: 'close', listener: CloseListener): this;
+  on(event: 'error', listener: ErrorListener): this;
+  /** Send the `connected` preamble and `start` frame. */
+  connect(options?: FakeConnectOptions): void;
+  /** Deliver one caller media frame (base64 μ-law) to the bridge. */
+  sendMedia(payload: string): void;
+  /** Deliver `ms` of caller silence as 20 ms μ-law frames. */
+  sendSilence(ms: number): void;
+  sendDtmf(digit: string): void;
+  /** Deliver the `stop` frame (Twilio ended the stream). */
+  stop(): void;
+  /** Simulate the socket dropping from Twilio's side. */
+  disconnect(code?: number, reason?: string): void;
+  /** Total milliseconds of agent audio actually "played" to the caller. */
+  get playedMs(): number;
+  /** Milliseconds of agent audio still buffered. */
+  get queuedMs(): number;
+  get wasClosedByBridge(): boolean;
+  /**
+   * Advance the playout clock: consume buffered media in order and echo each
+   * mark as playout reaches it.
+   */
+  advancePlayback(ms: number): void;
+  /** Play out everything currently buffered. */
+  playAll(): void;
+  private echoMark;
+  private deliver;
+}
+//#endregion
+//#region src/testing/FakeOpenAIServer.d.ts
+interface FakeOpenAIServerOptions {
+  /** Auto-ack session.update with session.updated. Default true. */
+  autoAckSessionUpdate?: boolean;
+}
+interface FakeAudioResponseOptions {
+  responseId?: string;
+  itemId?: string;
+  /** Base64 μ-law chunks to stream as output_audio deltas. */
+  chunks: string[];
+  transcript?: string;
+  usage?: Record<string, unknown>;
+  /** Emit response.done after the deltas. Default true. */
+  complete?: boolean;
+}
+declare class FakeOpenAIConnection {
+  readonly socket: WebSocket;
+  /** Every parsed client → server frame, in order. */
+  readonly received: Array<Record<string, any>>;
+  /** HTTP headers of the client's upgrade request (auth, extensions…). */
+  readonly upgradeHeaders: IncomingHttpHeaders;
+  private readonly server;
+  constructor(server: FakeOpenAIServer, socket: WebSocket, upgradeHeaders?: IncomingHttpHeaders);
+  /** Base64 payloads from input_audio_buffer.append frames. */
+  get appendedAudio(): string[];
+  eventsOfType(type: string): Array<Record<string, any>>;
+  waitForEvent(predicate: string | ((frame: Record<string, any>) => boolean), timeoutMs?: number): Promise<Record<string, any>>;
+  send(frame: Record<string, unknown>): void;
+  /** Script a complete assistant audio response (created → deltas → done). */
+  sendAudioResponse(options: FakeAudioResponseOptions): {
+    responseId: string;
+    itemId: string;
+  };
+  sendToolCall(options: {
+    name: string;
+    argumentsJson?: string;
+    callId?: string;
+    responseId?: string;
+  }): string;
+  sendSpeechStarted(): void;
+  sendSpeechStopped(): void;
+  sendUserTranscript(text: string): void;
+  sendError(error: Record<string, unknown>): void;
+  drop(code?: number, reason?: string): void;
+}
+declare class FakeOpenAIServer {
+  readonly connections: FakeOpenAIConnection[];
+  private readonly wss;
+  private readonly options;
+  readonly url: string;
+  private constructor();
+  static start(options?: FakeOpenAIServerOptions): Promise<FakeOpenAIServer>;
+  get latest(): FakeOpenAIConnection;
+  waitForConnection(timeoutMs?: number): Promise<FakeOpenAIConnection>;
+  close(): Promise<void>;
+}
+//#endregion
+//#region src/testing/FakeGeminiLive.d.ts
+declare class FakeGeminiSession implements GeminiLiveSessionLike {
+  readonly params: GeminiConnectParams;
+  readonly realtimeInputs: Array<{
+    data: string;
+    mimeType: string;
+  }>;
+  readonly clientContents: Array<{
+    turns: Array<Record<string, any>>;
+    turnComplete?: boolean;
+  }>;
+  readonly toolResponses: Array<{
+    functionResponses: Array<Record<string, any>>;
+  }>;
+  closed: boolean;
+  constructor(params: GeminiConnectParams);
+  sendRealtimeInput(input: {
+    audio: {
+      data: string;
+      mimeType: string;
+    };
+  }): void;
+  sendClientContent(content: {
+    turns: Array<Record<string, any>>;
+    turnComplete?: boolean;
+  }): void;
+  sendToolResponse(response: {
+    functionResponses: Array<Record<string, any>>;
+  }): void;
+  close(): void;
+  serverMessage(message: Record<string, unknown>): void;
+  completeSetup(): void;
+  /** Stream a model turn: PCM24k audio parts (+ transcript), then turnComplete. */
+  sendAudioTurn(options: {
+    pcm24kBase64Chunks: string[];
+    transcript?: string;
+  }): void;
+  sendToolCall(options: {
+    name: string;
+    args?: Record<string, unknown>;
+    id?: string;
+  }): string;
+  sendInterrupted(): void;
+  sendInputTranscription(text: string): void;
+  sendResumptionUpdate(handle: string): void;
+  sendGoAway(timeLeft: string): void;
+  sendUsage(usage: Record<string, unknown>): void;
+  /** Drop the connection from the server side. */
+  drop(code?: number, reason?: string): void;
+}
+declare class FakeGeminiLive {
+  readonly sessions: FakeGeminiSession[];
+  /** Auto-send setupComplete on connect. Default true. */
+  autoCompleteSetup: boolean;
+  readonly connector: GeminiLiveConnector;
+  get latest(): FakeGeminiSession;
+  waitForSessions(count: number, timeoutMs?: number): Promise<FakeGeminiSession>;
+}
+//#endregion
+export { type FakeAudioResponseOptions, type FakeConnectOptions, FakeGeminiLive, FakeGeminiSession, FakeOpenAIConnection, FakeOpenAIServer, type FakeOpenAIServerOptions, FakeTwilioMediaStream, mulawSilenceBase64 };