End Dictation with AUDIO_END

Problem: You copied an ambient streaming client and the Dictation stream never finishes. Solution: On /ws/transcribe, send { "type": "AUDIO", "audioData": "<base64 pcm>" } chunks, then { "type": "EVENT", "event": "AUDIO_END" }. Wait for inbound transcript.transcript === "EOF", close the socket, then call End Dictation session. Do not send ambient RU9G.

This cookbook assumes you already have your Partner ID, authentication tokens, and an active Dictation transcription_session_id.

TypeScript
Python

// After Dictation AUDIO chunks that use audioData (not ambient data):
// ws.send(JSON.stringify({ type: "AUDIO", audioData: base64PcmChunk }));

ws.send(JSON.stringify({ type: "EVENT", event: "AUDIO_END" }));

await new Promise<void>((resolve, reject) => {
  ws.onmessage = (event) => {
    try {
      const payload = JSON.parse(String(event.data));
      if (payload?.transcript?.transcript === "EOF") {
        resolve();
      }
    } catch (err) {
      reject(err);
    }
  };
  ws.onerror = () => reject(new Error("WebSocket error while waiting for EOF"));
});

ws.close();

const endRes = await fetch(
  `https://sdp.suki.ai/api/v1/transcription/session/${transcriptionSessionId}/end`,
  {
    method: "POST",
    headers: {
      sdp_suki_token: sdpSukiToken,
      sdp_provider_id: sdpProviderId,
    },
  }
);
if (!endRes.ok) {
  throw new Error(`End Dictation session failed: ${endRes.status}`);
}

import json
import requests

# After Dictation AUDIO chunks that use audioData (not ambient data):
# ws.send(json.dumps({"type": "AUDIO", "audioData": base64_pcm_chunk}))

ws.send(json.dumps({"type": "EVENT", "event": "AUDIO_END"}))

while True:
    raw = ws.recv()
    payload = json.loads(raw)
    transcript = (payload.get("transcript") or {}).get("transcript")
    if transcript == "EOF":
        break

ws.close()

end = requests.post(
    f"https://sdp.suki.ai/api/v1/transcription/session/{transcription_session_id}/end",
    headers={
        "sdp_suki_token": sdp_suki_token,
        "sdp_provider_id": sdp_provider_id,
    },
)
end.raise_for_status()

Common mistakes

Send { "type": "AUDIO", "data": "RU9G" } on Dictation. That marker is ambient-only.
Close the socket or call REST end before you receive inbound EOF.
Open /ws/transcribe again before the session returns to READY or IDLE.

Other cookbooks

AmbientAPI

Authenticate Browser WebSocket Handshake

Auth Dictation WebSocket with protocols.

5 min

DictationSDK

Write Dictation Text to an EHR Field

Write Dictation text with onSubmit.

5 min

​Common mistakes

​Other cookbooks

​Authenticate Browser WebSocket Handshake

​Write Dictation Text to an EHR Field

Common mistakes

Other cookbooks

Authenticate Browser WebSocket Handshake

Write Dictation Text to an EHR Field