Usage guide
Setup, environment variables, and Python/JavaScript/Go usage examples.
TurnDetector
Managed end-of-turn detection that predicts when the user has finished speaking, using the Echo turn-detection models. Choose it to avoid cutting users off mid-thought without hand-tuning silence timeouts. Runs on ZeroRuntime’s managed inference and is billed to your ZeroRuntime account, so no provider API key is needed.model chooses the backend: echo-large, echo-omni, echo-small.
Fields
str
default:"echo-large"
Turn-detection model: “echo-small” for lowest latency, “echo-large” for higher accuracy, or “echo-omni”. Defaults to ‘echo-large’.
Optional[int]
default:"UNSET"
Number of audio channels the agent receives. Leave unset to detect it automatically. Defaults to None (not set).
Optional[str]
default:"UNSET"
Override the service host to connect to. Leave unset to use the managed service. Defaults to None (not set).
Optional[str]
default:"UNSET"
Language code of the conversation, e.g. “en”. Leave unset to let the model handle it. Defaults to None (not set).
Optional[int]
default:"UNSET"
Sample rate of incoming audio in Hz. Leave unset to detect it automatically. Defaults to None (not set).
float
default:"UNSET"
Seconds to wait for a turn prediction before giving up and continuing without it. Defaults to 2.0.