bossiara13/J.A.R.V.I.S-py
1
1
Fork 0
Code Issues Pull requests Projects Releases 6 Packages Activity Actions

vosk usage optimization

Browse source
This commit is contained in:
Priler 2026-01-05 04:20:43 +05:00
parent 75edd47d18
commit bd8e2c989d
6 changed files with 305 additions and 108 deletions
Download patch file Download diff file Expand all files Collapse all files

174
crates/jarvis-core/src/stt/vosk.rs
View file

@ -9,80 +9,154 @@ use crate::stt::vosk_models;
use crate::DB;
static MODEL: OnceCell<Model> = OnceCell::new();
static RECOGNIZER: OnceCell<Mutex<Recognizer>> = OnceCell::new();
static WAKE_RECOGNIZER: OnceCell<Mutex<Recognizer>> = OnceCell::new();
static SPEECH_RECOGNIZER: OnceCell<Mutex<Recognizer>> = OnceCell::new();
pub fn init_vosk() -> Result<(), String> {
if RECOGNIZER.get().is_some() {
if MODEL.get().is_some() {
return Ok(());
} // already initialized
let model_path = get_configured_model_path()?;
info!("Loading Vosk model from: {}", model_path.display());
let model = Model::new(model_path.to_str().unwrap())
.ok_or_else(|| format!("Failed to load Vosk model from: {}", model_path.display()))?;
let mut recognizer = Recognizer::new(&model, 16000.0)
.ok_or("Failed to create Vosk recognizer")?;
//let mut recognizer = Recognizer::new(&model, 16000.0)
// .ok_or("Failed to create Vosk recognizer")?;
let wake_phrases: &[&str] = &[
config::VOSK_FETCH_PHRASE,
"[unk]",
"джон",
"джони",
"джей",
"джонстон",
"привет",
"давай",
];
let mut wake_recognizer = Recognizer::new_with_grammar(&model, 16000.0, wake_phrases)
.ok_or("Failed to create wake word recognizer")?;
recognizer.set_max_alternatives(10);
recognizer.set_words(true);
recognizer.set_partial_words(true);
wake_recognizer.set_max_alternatives(1); // required for confidence check later on
MODEL.set(model);
RECOGNIZER.set(Mutex::new(recognizer));
let mut speech_recognizer = Recognizer::new(&model, 16000.0)
.ok_or("Failed to create speech recognizer")?;
speech_recognizer.set_max_alternatives(config::VOSK_SPEECH_RECOGNIZER_MAX_ALTERNATIVES);
speech_recognizer.set_words(config::VOSK_SPEECH_RECOGNIZER_WORDS);
speech_recognizer.set_partial_words(config::VOSK_SPEECH_PARTIAL_WORDS);
MODEL.set(model).map_err(|_| "Model already set")?;
WAKE_RECOGNIZER.set(Mutex::new(wake_recognizer)).map_err(|_| "Wake recognizer already set")?;
SPEECH_RECOGNIZER.set(Mutex::new(speech_recognizer)).map_err(|_| "Speech recognizer already set")?;
Ok(())
}
pub fn recognize(data: &[i16], include_partial: bool) -> Option<String> {
let state = RECOGNIZER
.get()
.unwrap()
.lock()
.unwrap()
.accept_waveform(data);
match state {
Ok(ds) => {
match ds {
DecodingState::Running => {
if include_partial {
Some(
RECOGNIZER
.get()
.unwrap()
.lock()
.unwrap()
.partial_result()
.partial
.into(),
)
} else {
None
pub fn recognize_wake_word(data: &[i16]) -> Option<(String, f32)> {
let mut recognizer = WAKE_RECOGNIZER.get()?.lock().unwrap();
match recognizer.accept_waveform(data) {
Ok(DecodingState::Running) => {
// partials don't have confidence, skip them
None
}
Ok(DecodingState::Finalized) => {
let result = recognizer.result();
// compensate confidence issues
if let Some(alternatives) = result.multiple() {
if let Some(best) = alternatives.alternatives.first() {
if !best.text.is_empty() {
return Some((best.text.to_string(), best.confidence));
}
}
DecodingState::Finalized => {
// Result will always be multiple because we called set_max_alternatives
RECOGNIZER
.get()
.unwrap()
.lock()
.unwrap()
.result()
.multiple()
.and_then(|m| m.alternatives.first().map(|a| a.text.to_string()))
}
DecodingState::Failed => None,
}
},
Err(err) => {
error!("Vosk accept waveform error.\nError details: {}", err);
None
None
}
_ => None,
}
}
pub fn recognize_speech(data: &[i16]) -> Option<String> {
let mut recognizer = SPEECH_RECOGNIZER.get()?.lock().unwrap();
match recognizer.accept_waveform(data) {
Ok(DecodingState::Finalized) => {
recognizer.result()
.multiple()
.and_then(|m| m.alternatives.first().map(|a| a.text.to_string()))
}
_ => None,
}
}
pub fn reset_speech_recognizer() {
if let Some(recognizer) = SPEECH_RECOGNIZER.get() {
recognizer.lock().unwrap().reset();
}
}
pub fn reset_wake_recognizer() {
if let Some(recognizer) = WAKE_RECOGNIZER.get() {
recognizer.lock().unwrap().reset();
}
}
// pub fn recognize(data: &[i16], include_partial: bool) -> Option<String> {
// let state = RECOGNIZER
// .get()
// .unwrap()
// .lock()
// .unwrap()
// .accept_waveform(data);
// match state {
// Ok(ds) => {
// match ds {
// DecodingState::Running => {
// if include_partial {
// Some(
// RECOGNIZER
// .get()
// .unwrap()
// .lock()
// .unwrap()
// .partial_result()
// .partial
// .into(),
// )
// } else {
// None
// }
// }
// DecodingState::Finalized => {
// // Result will always be multiple because we called set_max_alternatives
// RECOGNIZER
// .get()
// .unwrap()
// .lock()
// .unwrap()
// .result()
// .multiple()
// .and_then(|m| m.alternatives.first().map(|a| a.text.to_string()))
// }
// DecodingState::Failed => None,
// }
// },
// Err(err) => {
// error!("Vosk accept waveform error.\nError details: {}", err);
// None
// }
// }
// }
fn get_configured_model_path() -> Result<std::path::PathBuf, String> {
// try to get from settings
if let Some(db) = DB.get() {