Why is base64 audio and transcript missing in conversation.item.created event with WebRTC?

I’m using the OpenAI Realtime API with WebRTC, expecting the conversation.item.created event to include base64 audio and a transcript, as mentioned in the documentation (https://platform.openai.com/docs/api-reference/realtime-server-events/conversation/item). However, when I implement WebRTC, neither of these fields appears in the event payload.

Below is the JSON event payload I receive:

JSONCopied
{
    "event_id": "event_1920",
    "type": "conversation.item.created",
    "previous_item_id": "msg_002",
    "item": {
        "id": "msg_003",
        "object": "realtime.item",
        "type": "message",
        "status": "completed",
        "role": "user",
        "content": [
            {
                "type": "input_audio",
                "transcript": "hello how are you", // this item is null
                "audio": "base64encodedaudio==" // this item does not exists
            }
        ]
    }
}

I create the session in my backend (PHP, Laravel) as follows:

PHPCopied
$data = [
            'model' => $model,
            "modalities" => ["audio", "text"],
            "instructions" => $instruction ?? "You are a friendly assistant.",
            "voice"=> $voice,
            "input_audio_transcription" => [
                'model' =>  "whisper-1"
            ],
            "turn_detection" => [
                "type" => "server_vad"
            ]
        ];

$response = Http::withToken($apiKey)
                ->withHeaders(['Content-Type' => 'application/json'])
                ->post($url, $data);

On the front end, I connect to WebRTC like this:

JAVASCRIPTCopied
  this.aiSession = await this.createRTSession();

  // Create a peer connection
  const pc = new RTCPeerConnection();

  // Set up to play remote audio from the model
  const audioEl = document.createElement("audio");
  audioEl.autoplay = true;

  pc.ontrack = e => {
    const remoteAudioStream = new MediaStream();
    remoteAudioStream.addTrack(e.track);
    audioEl.srcObject = remoteAudioStream;
    this.animate(remoteAudioStream);
  };

  // Add local audio track for microphone input in the browser
  const ms = await this.getMicStream();
  this.startRecording(ms);

  pc.addTrack(ms.getTracks()[0]);

  // Set up data channel for sending and receiving events
  const dc = pc.createDataChannel("oai-events");

  dc.addEventListener("message", (e) => {
    let $data = JSON.parse(e.data);
    // Realtime server events appear here!
    console.log($data.type, $data);
  });

  const offer = await pc.createOffer();
  await pc.setLocalDescription(offer);

  const baseUrl = "https://api.openai.com/v1/realtime";
  const model = this.aiSession.model;
  const sdpResponse = await fetch(`${baseUrl}?model=${model}`, {
    method: "POST",
    body: offer.sdp,
    headers: {
      Authorization: `Bearer ${this.ephemeralToken}`,
      "Content-Type": "application/sdp"
    },
  });

  const answer = {
    type: "answer",
    sdp: await sdpResponse.text(),
  };

  await pc.setRemoteDescription(answer);

Any suggestions on why the base64 audio and transcript fields are not present or how I can resolve this issue?