Configuration reference

Constants

Constant	Value	Meaning
SAMPLE_RATE	16_000	Required low-level sample rate
AUDIO_BLOCK_SAMPLES	160	Samples passed per detector call
DEFAULT_COOLDOWN	1 second	Listener repeat-suppression period

Config

Config::from_file(path) parses standard JSON and exposes:

Field	Type
model_path	PathBuf
wake_word	String
probability_cutoff	f32
sliding_window_size	usize
feature_step_size_ms	u32
metadata	ModelMetadata

Metadata contains optional author and website values, trained languages, and format version.

DetectorBuilder

Start with Detector::builder(model_path). A model-only build requires all three model-specific settings:

# use micro_wakeword::Detector;
# fn run() -> micro_wakeword::Result<()> {
let detector = Detector::builder("model.tflite")
    .wake_word("hello")
    .probability_cutoff(0.5)
    .sliding_window_size(3)
    .feature_step_size_ms(10)
    .build()?;
# Ok(()) }

.feature_step_size_ms(10) is optional because 10 is the only supported value and the default. .runtime(...) is also optional and defaults to Runtime::Auto.

ListenerBuilder

There are two entry points:

// Parse model values from JSON; builder methods can override them.
let builder = micro_wakeword::Listener::config_builder("wake-word.json")?;

// Standalone model; wake word, cutoff, and window become required.
let builder = micro_wakeword::Listener::builder("model.tflite");
# Ok::<(), micro_wakeword::Error>(())

Methods:

Method	Purpose
.wake_word(...)	Set or override the reported label
.probability_cutoff(...)	Set or override detection threshold
.sliding_window_size(...)	Set or override score smoothing
.runtime(...)	Choose TensorFlow Lite loading policy
.device(...)	Select an input name or index
.cooldown(Duration)	Set repeat suppression; default 1 second

For exact signatures and trait details, use the generated API documentation.