What is being streamed in the Talkdesk WebSocket messages?
In the article Conversation Orchestrator: Streaming Unidirectional Audio, you can find the “WebSocket Message protocols” section, which provides detailed information on what is being sent in each WebSocket message.
Is the stream unidirectional or bidirectional?
Both unidirectional and bidirectional streams are supported, depending on which Studio components are being used.
The Stream audio unidirectional component supports unidirectional streaming, while the Connect to Autopilot Voice component supports bidirectional streaming.
Can the audio be streamed before the agent answers the call?
To stream the audio before the agent answers the call, you only need to add the Stream audio unidirectional Studio component, before the “Assignment and Dial” component.
To know more about these message protocols, check the Conversation Orchestrator: Streaming Unidirectional Audio article.
Is it possible to stream the audio from the same inbound call to two or more different WSS endpoints?
Yes, but differences apply to whether the stream is being leveraged in a unidirectional or bidirectional way.
What happens when trying to unidirectionally stream the audio from the same inbound call to two or more different WSS endpoints?
When using the Stream audio unidirectional Studio component, it is possible to add two or more components in the same flow, streaming the audio to different endpoints.
What will happen is that while the agent is engaging in a live call with a caller, the audio of this call will be streamed simultaneously to the endpoint(s) defined in all the “Stream audio unidirectional” components placed before the “Assignment and Dial.”
What happens when trying to bidirectionally stream the audio from the same inbound call to two or more different WSS endpoints?
When using the Connect to Autopilot Voice Studio component to leverage Conversation Orchestrator in a bidirectional way, it is possible to add two or more components in the same flow, but the audio will be streamed in a sequential mode.
Because the bidirectional stream leverages a “stream & hold” technology, only when one stream ends, is it possible to start a sequential stream.
Is there an optimal roundtrip time when using a bidirectional stream?
It is advised that the partner receives and processes the information back on a roundtrip of 500ms to 1000ms.
Our recommendation is that the partner should integrate with the AI systems in real-time streaming technology when possible, instead of recordings back and forth.
What is the timeout for Conversation Orchestrator's connectivity when sending the audio packages?
The timeout is set to 2 seconds. This means that if the audio package is not received by the WebSocket endpoint within 2 seconds, the Conversation Orchestrator will log the action as a timeout.
What is the maximum size per message in the WebSocket sent through the integration?
The maximum size per message in the WebSocket is 2,097.152 bytes.
What is the WebSocket message “data”?
The WebSocket “data” message, without any event parameter, similar to the one shown below, should be ignored.
{"data":"OTM1YzEzNjUtZTA2Yy00MTkxLWE4ZDQtYWNiMjZkYjU5NmI0"}