VAPI-3479 Match telephone-event clock rate to the primary audio codec (fixes silent DTMF loss)

src/v1/bandwidthRtc.test.ts

@@ -184,58 +184,81 @@ describe("bandwidthRtcV1 addStreamToPublishingPeerConnection", () => {
     expect((brtc as any).localDtmfSenders.has("stream-1")).toBe(false);
   });
 
-  test("appends telephone-event codec when missing from audio preferences", () => {
+  test("appends the clock-matched telephone-event when missing from audio preferences", () => {
     const brtc = new BandwidthRtc();
     const transceiver = makeTransceiver();
     withPublishingPeerConnection(brtc, transceiver);
 
-    const telephoneEventCodec = { mimeType: "audio/telephone-event", clockRate: 8000 };
-    (global as any).RTCRtpSender = { getCapabilities: jest.fn().mockReturnValue({ codecs: [telephoneEventCodec] }) };
+    const te8000 = { mimeType: "audio/telephone-event", clockRate: 8000 };
+    const te48000 = { mimeType: "audio/telephone-event", clockRate: 48000 };
+    (global as any).RTCRtpSender = { getCapabilities: jest.fn().mockReturnValue({ codecs: [te8000, te48000] }) };
 
     const opusCodec = { mimeType: "audio/opus", clockRate: 48000 };
     (brtc as any).addStreamToPublishingPeerConnection(makeMockStream("stream-1", "audio"), { audio: [opusCodec] });
 
-    expect(transceiver.setCodecPreferences).toHaveBeenCalledWith([opusCodec, telephoneEventCodec]);
+    // Must pair with telephone-event/48000 (Opus clock), not the 8000 variant.
+    expect(transceiver.setCodecPreferences).toHaveBeenCalledWith([opusCodec, te48000]);
   });
 
-  test("does not duplicate telephone-event when already in preferences", () => {
+  test("replaces a clock-mismatched telephone-event already in preferences", () => {
     const brtc = new BandwidthRtc();
     const transceiver = makeTransceiver();
     withPublishingPeerConnection(brtc, transceiver);
 
     const opusCodec = { mimeType: "audio/opus", clockRate: 48000 };
-    const telephoneEventCodec = { mimeType: "audio/telephone-event", clockRate: 8000 };
-    (brtc as any).addStreamToPublishingPeerConnection(makeMockStream("stream-1", "audio"), { audio: [opusCodec, telephoneEventCodec] });
+    const te8000 = { mimeType: "audio/telephone-event", clockRate: 8000 };
+    const te48000 = { mimeType: "audio/telephone-event", clockRate: 48000 };
+    (global as any).RTCRtpSender = { getCapabilities: jest.fn().mockReturnValue({ codecs: [te48000] }) };
 
-    expect(transceiver.setCodecPreferences).toHaveBeenCalledWith([opusCodec, telephoneEventCodec]);
+    (brtc as any).addStreamToPublishingPeerConnection(makeMockStream("stream-1", "audio"), { audio: [opusCodec, te8000] });
+
+    // The mismatched 8000 event is dropped and replaced with the 48000 one from capabilities.
+    expect(transceiver.setCodecPreferences).toHaveBeenCalledWith([opusCodec, te48000]);
+    expect(transceiver.setCodecPreferences).toHaveBeenCalledTimes(1);
+  });
+
+  test("keeps the clock-matched telephone-event already in preferences", () => {
+    const brtc = new BandwidthRtc();
+    const transceiver = makeTransceiver();
+    withPublishingPeerConnection(brtc, transceiver);
+
+    const opusCodec = { mimeType: "audio/opus", clockRate: 48000 };
+    const te48000 = { mimeType: "audio/telephone-event", clockRate: 48000 };
+    (brtc as any).addStreamToPublishingPeerConnection(makeMockStream("stream-1", "audio"), { audio: [opusCodec, te48000] });
+
+    expect(transceiver.setCodecPreferences).toHaveBeenCalledWith([opusCodec, te48000]);
     expect(transceiver.setCodecPreferences).toHaveBeenCalledTimes(1);
   });
 
-  test("falls back to original preferences when telephone-event not found in capabilities", () => {
+  test("falls back to original preferences when no clock-matched telephone-event is available", () => {
     const brtc = new BandwidthRtc();
     const transceiver = makeTransceiver();
     withPublishingPeerConnection(brtc, transceiver);
 
-    (global as any).RTCRtpSender = { getCapabilities: jest.fn().mockReturnValue({ codecs: [] }) };
+    // Only a mismatched (8000) telephone-event exists; better to omit it than offer a dropped codec.
+    const te8000 = { mimeType: "audio/telephone-event", clockRate: 8000 };
+    (global as any).RTCRtpSender = { getCapabilities: jest.fn().mockReturnValue({ codecs: [te8000] }) };
 
     const opusCodec = { mimeType: "audio/opus", clockRate: 48000 };
     (brtc as any).addStreamToPublishingPeerConnection(makeMockStream("stream-1", "audio"), { audio: [opusCodec] });
 
     expect(transceiver.setCodecPreferences).toHaveBeenCalledWith([opusCodec]);
   });
 
-  test("forces telephone-event into codec preferences even without explicit codecPreferences", () => {
+  test("forces the clock-matched telephone-event even without explicit codecPreferences", () => {
     const brtc = new BandwidthRtc();
     const transceiver = makeTransceiver();
     withPublishingPeerConnection(brtc, transceiver);
 
     const opusCodec = { mimeType: "audio/opus", clockRate: 48000 };
-    const telephoneEventCodec = { mimeType: "audio/telephone-event", clockRate: 8000 };
-    (global as any).RTCRtpSender = { getCapabilities: jest.fn().mockReturnValue({ codecs: [opusCodec, telephoneEventCodec] }) };
+    const te8000 = { mimeType: "audio/telephone-event", clockRate: 8000 };
+    const te48000 = { mimeType: "audio/telephone-event", clockRate: 48000 };
+    (global as any).RTCRtpSender = { getCapabilities: jest.fn().mockReturnValue({ codecs: [opusCodec, te8000, te48000] }) };
 
     (brtc as any).addStreamToPublishingPeerConnection(makeMockStream("stream-1", "audio"));
 
-    expect(transceiver.setCodecPreferences).toHaveBeenCalledWith([opusCodec, telephoneEventCodec]);
+    // Full capabilities collapse to the primary codec + its matching-clock telephone-event.
+    expect(transceiver.setCodecPreferences).toHaveBeenCalledWith([opusCodec, te48000]);
   });
 
   test("skips setCodecPreferences when RTCRtpSender is unavailable (e.g. non-browser environment)", () => {

src/v1/bandwidthRtc.ts

@@ -703,16 +703,25 @@ export class BandwidthRtc {
         // dropped from the SDP offer. Apply it unconditionally (not just when the caller
         // passes codecPreferences) so telephone-event is always present and RTCDTMFSender
         // can send RFC 4733 DTMF packets, regardless of the browser's default codec offer.
+        //
+        // RFC 4733: telephone-event shares the primary audio codec's RTP clock, so it MUST
+        // be offered at that codec's clock rate (e.g. 48000 alongside Opus). A mismatched
+        // telephone-event (e.g. 8000 next to Opus/48000) is silently dropped by the peer and
+        // DTMF never reaches the far end — so drop every telephone-event and re-add only the
+        // one whose clock rate matches, sourced from prefs or full capabilities.
         const audioCapabilities = typeof RTCRtpSender !== "undefined" ? RTCRtpSender.getCapabilities(TRACK_KIND_AUDIO) : undefined;
         const audioCodecs = codecPreferences?.audio ?? audioCapabilities?.codecs;
         if (audioCodecs) {
-          const hasTelephoneEvent = audioCodecs.some((c) => c.mimeType.toLowerCase() === TELEPHONE_EVENT_MIME_TYPE);
-          if (!hasTelephoneEvent) {
-            const telephoneEventCodec = audioCapabilities?.codecs.find((c) => c.mimeType.toLowerCase() === TELEPHONE_EVENT_MIME_TYPE);
-            transceiver.setCodecPreferences(telephoneEventCodec ? [...audioCodecs, telephoneEventCodec] : audioCodecs);
-          } else {
-            transceiver.setCodecPreferences(audioCodecs);
-          }
+          // Strip every telephone-event, then re-add only the one whose clock rate matches the
+          // primary (first) media codec, preferring one already in the caller's list and falling
+          // back to full capabilities. If none matches, leave the codecs untouched rather than
+          // offer a telephone-event the peer will drop.
+          const isTelephoneEvent = (c: RTCRtpCodec) => c.mimeType.toLowerCase() === TELEPHONE_EVENT_MIME_TYPE;
+          const mediaCodecs = audioCodecs.filter((c) => !isTelephoneEvent(c));
+          const primaryClockRate = mediaCodecs[0]?.clockRate;
+          const matchesPrimary = (c: RTCRtpCodec) => isTelephoneEvent(c) && c.clockRate === primaryClockRate;
+          const telephoneEventCodec = audioCodecs.find(matchesPrimary) ?? audioCapabilities?.codecs.find(matchesPrimary);
+          transceiver.setCodecPreferences(telephoneEventCodec ? [...mediaCodecs, telephoneEventCodec] : audioCodecs);
         }
       } else if (track.kind === TRACK_KIND_VIDEO && codecPreferences?.video) {
         transceiver.setCodecPreferences(codecPreferences.video);