Skip to main content
WSS
/
v1
/
connection
Messages
Input audio
type:object

Client sends captured audio to the server (base64).

Tool call output
type:object

Client sends the result of a tool call requested by the server.

Server message
type:object

Discriminated by type; payload in data. One of: session_started, response_audio, response_text, input_audio_transcript, speech_started, speech_stopped, session_ended, interruption, tool_call.

Session started
type:object

Sent when the call session has started.

Response audio
type:object

Streamed TTS audio chunk for the current turn.

Response text
type:object

Agent text for the current turn; may repeat or refine while is_final is false.

Input audio transcript
type:object

User speech transcript from STT; false is_final for interim hypotheses.

Speech started
type:object

Sent when user speech (VAD) has started.

Speech stopped
type:object

Sent when user speech (VAD) has stopped.

Session ended
type:object

Sent when the call session has ended.

Interruption
type:object

Sent when user has interrupted the agent.

Tool call
type:object

Server requests the client to execute a tool and respond with tool_call_output.