Server API

These events are sent by the server to inform the client about the session lifecycle, turn state changes, and transcription results. See Receiving turn events in the High-Level Flow for how they arrive during a session.

session_started (Server to Client message)

Sent once, after a successful start_session, to acknowledge the session.

• type: session_started
• sessionId: ID of the session.

Example:

{
  "type": "session_started",
  "sessionId": "sess_abc123"
}

turn_started (Server to Client message)

This message is sent when a turn begins. From this point on, the speech recognition engine processes all incoming audio chunks.

• type: turn_started
• turnId: ID of the turn being processed.

Example:

{
  "type": "turn_started",
  "turnId": "123"
}

turn_partial (Server to Client message)

This event is sent when new text has been decoded from the audio (the turn partials).

• type: turn_partial
• turnId: ID of the turn being processed.
• transcript: the text understood so far.
• turnAudioDuration: duration of the audio processed to generate this result.

Example:

{
  "type": "turn_partial",
  "turnId": "123",
  "transcript": "oui c'est bien",
  "turnAudioDuration": 2.3
}

turn_ended (Server to Client message)

This message is sent when a turn ends, and contains the final transcript. In manual mode it is emitted after the client sends end_turn: the server finalizes the transcription of the audio received so far and returns the result.

• type: turn_ended
• turnId: ID of the turn being processed.
• kind: The reason the turn ended. Currently always end_of_utterance.
• transcript: The transcription of the turn.
• turnAudioDuration: Duration of the audio processed to generate this result.

Example:

{
  "type": "turn_ended",
  "turnId": "123",
  "kind": "end_of_utterance",
  "transcript": "oui c'est bien ça",
  "turnAudioDuration": 4.7
}