fix(twilio): maintain audio format after handoff (#126)

Co-authored-by: Kazuhiro Sera <[email protected]>

Author: dkundel-openai

packages/agents-extensions/src/TwilioRealtimeTransport.ts

@@ -176,6 +176,11 @@ export class TwilioRealtimeTransportLayer extends OpenAIRealtimeWebSocket {
     await super.connect(options);
   }
 
+  updateSessionConfig(config: Partial<RealtimeSessionConfig>): void {
+    const newConfig = this._setInputAndOutputAudioFormat(config);
+    super.updateSessionConfig(newConfig);
+  }
+
   _interrupt(_elapsedTime: number) {
     const elapsedTime = this.#lastPlayedChunkCount + 50; /* 50ms buffer */
     this.#logger.debug(

packages/agents-extensions/test/TwilioRealtimeTransport.test.ts

@@ -26,6 +26,7 @@ vi.mock('@openai/agents/realtime', () => {
   FakeOpenAIRealtimeWebSocket.prototype.sendAudio = vi.fn();
   FakeOpenAIRealtimeWebSocket.prototype.close = vi.fn();
   FakeOpenAIRealtimeWebSocket.prototype._interrupt = vi.fn();
+  FakeOpenAIRealtimeWebSocket.prototype.updateSessionConfig = vi.fn();
   return { OpenAIRealtimeWebSocket: FakeOpenAIRealtimeWebSocket, utils };
 });
 
@@ -146,4 +147,22 @@ describe('TwilioRealtimeTransportLayer', () => {
     expect(marks[2].mark.name).toBe('b:1');
     expect(audioListener).toHaveBeenCalledTimes(3);
   });
+
+  test('updateSessionConfig keeps audio format', async () => {
+    const twilio = new FakeTwilioWebSocket();
+    const transport = new TwilioRealtimeTransportLayer({
+      twilioWebSocket: twilio as any,
+    });
+    await transport.connect({ apiKey: 'ek_test' } as any);
+    const { OpenAIRealtimeWebSocket } = await import('@openai/agents/realtime');
+    const spy = vi.mocked(
+      OpenAIRealtimeWebSocket.prototype.updateSessionConfig,
+    );
+    transport.updateSessionConfig({ instructions: 'hi' });
+    expect(spy).toHaveBeenCalledWith({
+      instructions: 'hi',
+      inputAudioFormat: 'g711_ulaw',
+      outputAudioFormat: 'g711_ulaw',
+    });
+  });
 });