2023-06-11 19:17:50 +05:00
|
|
|
use once_cell::sync::OnceCell;
|
2026-02-18 21:08:48 +05:00
|
|
|
use vosk::{DecodingState, Recognizer};
|
|
|
|
|
use std::sync::Arc;
|
|
|
|
|
use parking_lot::Mutex;
|
2023-04-27 00:28:36 +05:00
|
|
|
|
2026-02-18 21:08:48 +05:00
|
|
|
use crate::{vosk_models, i18n, config, models};
|
|
|
|
|
use crate::models::vosk::VoskModel;
|
2026-01-05 03:38:04 +05:00
|
|
|
use crate::DB;
|
2023-04-27 00:28:36 +05:00
|
|
|
|
2026-02-18 21:08:48 +05:00
|
|
|
// the model Arc keeps the vosk::Model alive for the recognizers
|
|
|
|
|
static VOSK_MODEL: OnceCell<Arc<VoskModel>> = OnceCell::new();
|
2026-01-05 04:20:43 +05:00
|
|
|
static WAKE_RECOGNIZER: OnceCell<Mutex<Recognizer>> = OnceCell::new();
|
|
|
|
|
static SPEECH_RECOGNIZER: OnceCell<Mutex<Recognizer>> = OnceCell::new();
|
2023-04-27 00:28:36 +05:00
|
|
|
|
2026-01-05 03:38:04 +05:00
|
|
|
pub fn init_vosk() -> Result<(), String> {
|
2026-02-18 21:08:48 +05:00
|
|
|
if VOSK_MODEL.get().is_some() {
|
2026-01-05 03:38:04 +05:00
|
|
|
return Ok(());
|
2026-02-18 21:08:48 +05:00
|
|
|
}
|
2023-06-11 19:17:50 +05:00
|
|
|
|
2026-01-05 03:38:04 +05:00
|
|
|
let model_path = get_configured_model_path()?;
|
2026-02-18 21:08:48 +05:00
|
|
|
let model_id = format!("vosk:{}", model_path.display());
|
2026-01-05 03:38:04 +05:00
|
|
|
|
2026-02-18 21:08:48 +05:00
|
|
|
// load through registry (shared if anything else needs the same model)
|
|
|
|
|
let vosk = models::vosk::load(
|
|
|
|
|
models::registry(),
|
|
|
|
|
&model_id,
|
|
|
|
|
model_path.to_str().unwrap(),
|
|
|
|
|
)?;
|
2026-01-05 03:38:04 +05:00
|
|
|
|
2026-01-13 02:21:59 +05:00
|
|
|
// language-specific wake grammar
|
|
|
|
|
let lang = i18n::get_language();
|
|
|
|
|
let wake_grammar = config::get_wake_grammar(&lang);
|
|
|
|
|
info!("Wake grammar for '{}': {:?}", lang, wake_grammar);
|
|
|
|
|
|
2026-02-18 21:08:48 +05:00
|
|
|
let mut wake_recognizer = Recognizer::new_with_grammar(&vosk.model, 16000.0, wake_grammar)
|
2026-01-05 04:20:43 +05:00
|
|
|
.ok_or("Failed to create wake word recognizer")?;
|
2023-06-11 19:17:50 +05:00
|
|
|
|
2026-02-18 21:08:48 +05:00
|
|
|
wake_recognizer.set_max_alternatives(1);
|
2023-06-11 19:17:50 +05:00
|
|
|
|
2026-02-18 21:08:48 +05:00
|
|
|
let mut speech_recognizer = Recognizer::new(&vosk.model, 16000.0)
|
2026-01-05 04:20:43 +05:00
|
|
|
.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);
|
|
|
|
|
|
2026-02-18 21:08:48 +05:00
|
|
|
VOSK_MODEL.set(vosk).map_err(|_| "Model already set")?;
|
2026-01-05 04:20:43 +05:00
|
|
|
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")?;
|
2026-01-05 03:38:04 +05:00
|
|
|
|
|
|
|
|
Ok(())
|
2023-04-27 00:28:36 +05:00
|
|
|
}
|
|
|
|
|
|
2026-01-05 04:20:43 +05:00
|
|
|
|
|
|
|
|
pub fn recognize_wake_word(data: &[i16]) -> Option<(String, f32)> {
|
2026-02-18 21:08:48 +05:00
|
|
|
let mut recognizer = WAKE_RECOGNIZER.get()?.lock();
|
2026-01-05 04:20:43 +05:00
|
|
|
|
|
|
|
|
match recognizer.accept_waveform(data) {
|
|
|
|
|
Ok(DecodingState::Running) => {
|
|
|
|
|
None
|
|
|
|
|
}
|
|
|
|
|
Ok(DecodingState::Finalized) => {
|
|
|
|
|
let result = recognizer.result();
|
|
|
|
|
|
|
|
|
|
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));
|
2025-12-12 02:05:37 +05:00
|
|
|
}
|
|
|
|
|
}
|
2023-04-29 16:00:49 +05:00
|
|
|
}
|
2026-01-05 04:20:43 +05:00
|
|
|
|
|
|
|
|
None
|
|
|
|
|
}
|
|
|
|
|
_ => None,
|
|
|
|
|
}
|
|
|
|
|
}
|
2025-12-12 02:05:37 +05:00
|
|
|
|
2026-01-05 04:20:43 +05:00
|
|
|
|
|
|
|
|
pub fn recognize_speech(data: &[i16]) -> Option<String> {
|
2026-02-18 21:08:48 +05:00
|
|
|
let mut recognizer = SPEECH_RECOGNIZER.get()?.lock();
|
2026-01-05 04:20:43 +05:00
|
|
|
|
|
|
|
|
match recognizer.accept_waveform(data) {
|
|
|
|
|
Ok(DecodingState::Finalized) => {
|
|
|
|
|
recognizer.result()
|
|
|
|
|
.multiple()
|
|
|
|
|
.and_then(|m| m.alternatives.first().map(|a| a.text.to_string()))
|
2023-04-27 00:28:36 +05:00
|
|
|
}
|
2026-01-05 04:20:43 +05:00
|
|
|
_ => None,
|
2023-04-27 00:28:36 +05:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2026-01-05 04:20:43 +05:00
|
|
|
|
|
|
|
|
pub fn reset_speech_recognizer() {
|
|
|
|
|
if let Some(recognizer) = SPEECH_RECOGNIZER.get() {
|
2026-02-18 21:08:48 +05:00
|
|
|
recognizer.lock().reset();
|
2026-01-05 04:20:43 +05:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2026-04-23 11:07:35 +03:00
|
|
|
pub fn finalize_speech() -> Option<String> {
|
|
|
|
|
let mut recognizer = SPEECH_RECOGNIZER.get()?.lock();
|
|
|
|
|
let result = recognizer.final_result()
|
|
|
|
|
.multiple()
|
|
|
|
|
.and_then(|m| m.alternatives.first().map(|a| a.text.to_string()))
|
|
|
|
|
.filter(|s| !s.trim().is_empty());
|
|
|
|
|
recognizer.reset();
|
|
|
|
|
result
|
|
|
|
|
}
|
|
|
|
|
|
2026-01-05 04:20:43 +05:00
|
|
|
pub fn reset_wake_recognizer() {
|
|
|
|
|
if let Some(recognizer) = WAKE_RECOGNIZER.get() {
|
2026-02-18 21:08:48 +05:00
|
|
|
recognizer.lock().reset();
|
2026-01-05 04:20:43 +05:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2026-01-05 03:38:04 +05:00
|
|
|
fn get_configured_model_path() -> Result<std::path::PathBuf, String> {
|
|
|
|
|
// try to get from settings
|
|
|
|
|
if let Some(db) = DB.get() {
|
|
|
|
|
let settings = db.read();
|
|
|
|
|
if !settings.vosk_model.is_empty() {
|
|
|
|
|
if let Some(path) = vosk_models::get_model_path(&settings.vosk_model) {
|
|
|
|
|
return Ok(path);
|
|
|
|
|
}
|
|
|
|
|
warn!("Configured Vosk model '{}' not found, falling back to auto-detect", settings.vosk_model);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2026-02-11 07:21:50 +05:00
|
|
|
// auto-detect: prefer model matching current language
|
2026-01-05 03:38:04 +05:00
|
|
|
let available = vosk_models::scan_vosk_models();
|
2026-02-11 07:21:50 +05:00
|
|
|
let language = i18n::get_language();
|
|
|
|
|
|
|
|
|
|
let lang_code = match language.as_str() {
|
|
|
|
|
"ru" => "ru",
|
2026-02-18 21:08:48 +05:00
|
|
|
"en" => "us",
|
2026-02-11 07:21:50 +05:00
|
|
|
other => other,
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
if let Some(matched) = available.iter().find(|m| m.language == lang_code) {
|
|
|
|
|
info!("Auto-detected Vosk model for '{}': {}", language, matched.name);
|
|
|
|
|
return Ok(matched.path.clone());
|
|
|
|
|
}
|
|
|
|
|
|
2026-01-05 03:38:04 +05:00
|
|
|
if let Some(first) = available.first() {
|
2026-02-11 07:21:50 +05:00
|
|
|
info!("Auto-detected Vosk model (no language match): {}", first.name);
|
2026-01-05 03:38:04 +05:00
|
|
|
return Ok(first.path.clone());
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// fallback to legacy path
|
|
|
|
|
let legacy_path = std::path::Path::new(config::VOSK_MODEL_PATH);
|
|
|
|
|
if legacy_path.exists() {
|
|
|
|
|
return Ok(legacy_path.to_path_buf());
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
Err("No Vosk models found".into())
|
|
|
|
|
}
|