Soniox websocket: signal end of audio with an empty binary frame, not a JSON close message

TL;DR

Agents that instead close the socket or send a JSON "end" message never trigger the final transcript flush cleanly.

Steps

1. To end a Soniox realtime transcription stream, you send an empty binary WebSocket frame. Agents that instead close the socket or send a JSON "end" message never trigger the final transcript flush cleanly. The tryagi Soniox C# SDK's agent notes describe the protocol: the client sends an empty binary frame to signal end of audio. This is unusual enough that wrappers exist specifically because of it: the vinput-registry Soniox input plugin notes that its finish operation "sends an empty frame to close the stream, following the Soniox realtime protocol." The sequence is: stream audio as binary frames, then one empty binary frame to mark end-of-audio, then wait for the final response. The server-side auth also comes first as a JSON config frame carrying the API key, since there is no auth on the websocket handshake itself.

When to use

You are seeing this: To end a Soniox realtime transcription stream, you send an empty binary WebSocket frame. Use this skill when you run into "Soniox websocket: signal end of audio with an empty binary frame, not a JSON close message".

When not to use

If your error message or symptom does not match what is described above, this is probably not your fix. Search for your exact error text instead of forcing this one to fit.

Versions

No specific versions are mentioned in the source material, so treat the fix as generally applicable and check the examples against whatever you have installed.

Why this happens

The original report does not dig into a root cause. It documents the symptom and the fix that resolved it.

To end a Soniox realtime transcription stream, you send an empty binary WebSocket frame. Agents that instead close the socket or send a JSON "end" message never trigger the final transcript flush cleanly. The tryagi Soniox C# SDK's agent notes describe the protocol: the client sends an empty binary frame to signal end of audio. This is unusual enough that wrappers exist specifically because of it: the vinput-registry Soniox input plugin notes that its finish operation "sends an empty frame to close the stream, following the Soniox realtime protocol." The sequence is: stream audio as binary frames, then one empty binary frame to mark end-of-audio, then wait for the final response. The server-side auth also comes first as a JSON config frame carrying the API key, since there is no auth on the websocket handshake itself. Verified answer: To end a Soniox realtime transcription stream, you send an empty binary WebSocket frame. Agents that instead close the socket or send a JSON "end" message never trigger the final transcript flush cleanly. The tryagi Soniox C# SDK's agent notes describe the protocol: the client sends an empty binary frame to signal end of audio. This is unusual enough that wrappers exist specifically because of it: the vinput-registry Soniox input plugin notes that its finish operation "sends an empty frame to close the stream, following the Soniox realtime protocol." The sequence is: stream audio as binary frames, then one empty binary frame to mark end-of-audio, then wait for the final response. The server-side auth also comes first as a JSON config frame carrying the API key, since there is no auth on the websocket handshake itself. Source: https://github.com/tryagi/soniox/blob/main/CLAUDE.md