Skip to content

armgit / Handy

Public Imported from GitHub
0

A free, open source, and extensible speech-to-text application that works completely offline.

Parakeet V3 (#111)

CJ Pais45a473dc9fa8
diff --git a/src-tauri/Cargo.lock b/src-tauri/Cargo.lock
index 064bf937..73cc2ab8 100644
--- a/src-tauri/Cargo.lock
+++ b/src-tauri/Cargo.lock
@@ -597,7 +597,7 @@ dependencies = [
"semver",
"serde",
"serde_json",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
]
[[package]]
@@ -1400,6 +1400,19 @@ dependencies = [
"regex",
]
+[[package]]
+name = "env_logger"
+version = "0.10.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4cd405aab171cb85d6735e5c8d9db038c17d3ca007a4d2c25f337935c3d90580"
+dependencies = [
+ "humantime",
+ "is-terminal",
+ "log",
+ "regex",
+ "termcolor",
+]
+
[[package]]
name = "env_logger"
version = "0.11.6"
@@ -2026,7 +2039,7 @@ dependencies = [
"objc2-app-kit 0.3.1",
"once_cell",
"serde",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"windows-sys 0.59.0",
"x11-dl",
]
@@ -2121,7 +2134,8 @@ dependencies = [
"chrono",
"cpal",
"enigo",
- "env_logger",
+ "env_logger 0.11.6",
+ "flate2",
"futures-util",
"hound",
"log",
@@ -2136,6 +2150,7 @@ dependencies = [
"serde",
"serde_json",
"strsim",
+ "tar",
"tauri",
"tauri-build",
"tauri-plugin-autostart",
@@ -2151,8 +2166,8 @@ dependencies = [
"tauri-plugin-store",
"tauri-plugin-updater",
"tokio",
+ "transcribe-rs",
"vad-rs",
- "whisper-rs",
]
[[package]]
@@ -2217,6 +2232,12 @@ version = "0.4.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "fbf6a919d6cf397374f7dfeeea91d974c7c0a7221d0d0f4f20d859d329e53fcc"
+[[package]]
+name = "hermit-abi"
+version = "0.5.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fc0fef456e4baa96da950455cd02c081ca953b141298e41db3fc7e36b1da849c"
+
[[package]]
name = "hex"
version = "0.4.3"
@@ -2704,6 +2725,17 @@ dependencies = [
"once_cell",
]
+[[package]]
+name = "is-terminal"
+version = "0.4.16"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e04d7f318608d35d4b61ddd75cbdaee86b023ebe2bd5a66ee0915f0bf93095a9"
+dependencies = [
+ "hermit-abi 0.5.2",
+ "libc",
+ "windows-sys 0.59.0",
+]
+
[[package]]
name = "is-wsl"
version = "0.4.0"
@@ -2973,9 +3005,9 @@ dependencies = [
[[package]]
name = "log"
-version = "0.4.25"
+version = "0.4.28"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "04cbf5b083de1c7e0222a7a51dbfdba1cbe1c6ab0b15e29fff3f6c077fd9cd9f"
+checksum = "34080505efa8e45a4b816c349525ebe327ceaa8559756f0356cba97ef3bf7432"
[[package]]
name = "mac"
@@ -3154,7 +3186,7 @@ dependencies = [
"once_cell",
"png",
"serde",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"windows-sys 0.59.0",
]
@@ -3711,9 +3743,9 @@ dependencies = [
[[package]]
name = "once_cell"
-version = "1.20.2"
+version = "1.21.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1261fe7e33c73b354eab43b1273a57c8f967d0391e80353e51f764ac02cf6775"
+checksum = "42f5e15c9953c5e4ccceeb2e7382a716482c34515315f7b03532b8b4e8393d2d"
[[package]]
name = "open"
@@ -3834,7 +3866,7 @@ dependencies = [
"objc2-osa-kit",
"serde",
"serde_json",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
]
[[package]]
@@ -4130,7 +4162,7 @@ checksum = "a604568c3202727d1507653cb121dbd627a58684eb09a820fd746bee38b4442f"
dependencies = [
"cfg-if",
"concurrent-queue",
- "hermit-abi",
+ "hermit-abi 0.4.0",
"pin-project-lite",
"rustix 0.38.44",
"tracing",
@@ -4281,7 +4313,7 @@ dependencies = [
"rustc-hash 2.1.0",
"rustls",
"socket2",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"tokio",
"tracing",
]
@@ -4300,7 +4332,7 @@ dependencies = [
"rustls",
"rustls-pki-types",
"slab",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"tinyvec",
"tracing",
"web-time",
@@ -4483,14 +4515,14 @@ checksum = "dd6f9d3d47bdd2ad6945c5015a226ec6155d0bcdfd8f7cd29f86b71f8de99d2b"
dependencies = [
"getrandom 0.2.15",
"libredox",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
]
[[package]]
name = "regex"
-version = "1.11.1"
+version = "1.11.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b544ef1b4eac5dc2db33ea63606ae9ffcfac26c1416a2806ae0bf5f56b201191"
+checksum = "23d7fd106d8c02486a8d64e778353d1cffe08ce79ac2e82f540c86d0facf6912"
dependencies = [
"aho-corasick",
"memchr",
@@ -5296,7 +5328,7 @@ dependencies = [
"serde_json",
"sha2",
"smallvec 1.13.2",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"time",
"tokio",
"tokio-stream",
@@ -5379,7 +5411,7 @@ dependencies = [
"smallvec 1.13.2",
"sqlx-core",
"stringprep",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"time",
"tracing",
"whoami",
@@ -5417,7 +5449,7 @@ dependencies = [
"smallvec 1.13.2",
"sqlx-core",
"stringprep",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"time",
"tracing",
"whoami",
@@ -5442,7 +5474,7 @@ dependencies = [
"serde",
"serde_urlencoded",
"sqlx-core",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"time",
"tracing",
"url",
@@ -5891,7 +5923,7 @@ dependencies = [
"tauri-runtime",
"tauri-runtime-wry",
"tauri-utils",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"tokio",
"tray-icon",
"url",
@@ -5944,7 +5976,7 @@ dependencies = [
"sha2",
"syn 2.0.96",
"tauri-utils",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"time",
"url",
"uuid",
@@ -5993,7 +6025,7 @@ dependencies = [
"serde_json",
"tauri",
"tauri-plugin",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
]
[[package]]
@@ -6008,7 +6040,7 @@ dependencies = [
"serde_json",
"tauri",
"tauri-plugin",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
]
[[package]]
@@ -6028,7 +6060,7 @@ dependencies = [
"tauri",
"tauri-plugin",
"tauri-utils",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"toml",
"url",
]
@@ -6045,7 +6077,7 @@ dependencies = [
"serde_json",
"tauri",
"tauri-plugin",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
]
[[package]]
@@ -6058,7 +6090,7 @@ dependencies = [
"serde",
"tauri",
"tauri-plugin",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
]
[[package]]
@@ -6077,7 +6109,7 @@ dependencies = [
"serde_json",
"tauri",
"tauri-plugin",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"url",
"windows 0.58.0",
"zbus",
@@ -6098,7 +6130,7 @@ dependencies = [
"sys-locale",
"tauri",
"tauri-plugin",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
]
[[package]]
@@ -6120,7 +6152,7 @@ dependencies = [
"serde",
"serde_json",
"tauri",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"tracing",
"windows-sys 0.60.2",
"zbus",
@@ -6140,7 +6172,7 @@ dependencies = [
"sqlx",
"tauri",
"tauri-plugin",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"time",
"tokio",
]
@@ -6156,7 +6188,7 @@ dependencies = [
"serde_json",
"tauri",
"tauri-plugin",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"tokio",
"tracing",
]
@@ -6185,7 +6217,7 @@ dependencies = [
"tauri",
"tauri-plugin",
"tempfile",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"time",
"tokio",
"url",
@@ -6210,7 +6242,7 @@ dependencies = [
"serde",
"serde_json",
"tauri-utils",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"url",
"windows 0.61.3",
]
@@ -6272,7 +6304,7 @@ dependencies = [
"serde_json",
"serde_with",
"swift-rs",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"toml",
"url",
"urlpattern",
@@ -6316,6 +6348,15 @@ dependencies = [
"utf-8",
]
+[[package]]
+name = "termcolor"
+version = "1.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "06794f8f6c5c898b3275aebefa6b8a1cb24cd2c6c79397ab15774837a0bc5755"
+dependencies = [
+ "winapi-util",
+]
+
[[package]]
name = "thiserror"
version = "1.0.69"
@@ -6327,11 +6368,11 @@ dependencies = [
[[package]]
name = "thiserror"
-version = "2.0.11"
+version = "2.0.16"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d452f284b73e6d76dd36758a0c8684b1d5be31f92b89d07fd5822175732206fc"
+checksum = "3467d614147380f2e4e374161426ff399c91084acd2363eaf549172b3d5e60c0"
dependencies = [
- "thiserror-impl 2.0.11",
+ "thiserror-impl 2.0.16",
]
[[package]]
@@ -6347,9 +6388,9 @@ dependencies = [
[[package]]
name = "thiserror-impl"
-version = "2.0.11"
+version = "2.0.16"
source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "26afc1baea8a989337eeb52b6e72a039780ce45c3edfcc9c5b9d112feeb173c2"
+checksum = "6c5e1be1c48b9172ee610da68fd9cd2770e7a4056cb3fc98710ee6906f0c7960"
dependencies = [
"proc-macro2",
"quote",
@@ -6609,6 +6650,23 @@ dependencies = [
"once_cell",
]
+[[package]]
+name = "transcribe-rs"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "eaeba5648789d05eedce5478069195639a0770dcc0338146037d4b85163df4bd"
+dependencies = [
+ "env_logger 0.10.2",
+ "hound",
+ "log",
+ "ndarray",
+ "once_cell",
+ "ort",
+ "regex",
+ "thiserror 2.0.16",
+ "whisper-rs",
+]
+
[[package]]
name = "transpose"
version = "0.2.3"
@@ -6637,7 +6695,7 @@ dependencies = [
"once_cell",
"png",
"serde",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"windows-sys 0.59.0",
]
@@ -7133,7 +7191,7 @@ version = "0.38.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "36695906a1b53a3bf5c4289621efedac12b73eeb0b89e7e1a89b517302d5d75c"
dependencies = [
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"windows 0.61.3",
"windows-core 0.61.2",
]
@@ -7861,7 +7919,7 @@ dependencies = [
"sha2",
"soup3",
"tao-macros",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
"url",
"webkit2gtk",
"webkit2gtk-sys",
@@ -8124,7 +8182,7 @@ dependencies = [
"displaydoc",
"indexmap 2.7.1",
"memchr",
- "thiserror 2.0.11",
+ "thiserror 2.0.16",
]
[[package]]
diff --git a/src-tauri/Cargo.toml b/src-tauri/Cargo.toml
index ec7d9d66..38302375 100644
--- a/src-tauri/Cargo.toml
+++ b/src-tauri/Cargo.toml
@@ -62,15 +62,9 @@ natural = "0.5.0"
chrono = "0.4"
rusqlite = { version = "0.32", features = ["bundled"] }
tauri-plugin-sql = { version = "2", features = ["sqlite"] }
-
-[target.'cfg(target_os = "macos")'.dependencies]
-whisper-rs = { version = "0.13.2", features = ["metal"] }
-
-[target.'cfg(target_os = "windows")'.dependencies]
-whisper-rs = { version = "0.13.2", features = ["vulkan"] }
-
-[target.'cfg(target_os = "linux")'.dependencies]
-whisper-rs = { version = "0.13.2", features = ["vulkan"] }
+tar = "0.4.44"
+flate2 = "1.0"
+transcribe-rs = "0.1.0"
[target.'cfg(not(any(target_os = "android", target_os = "ios")))'.dependencies]
tauri-plugin-autostart = "2"
diff --git a/src-tauri/src/audio_toolkit/mod.rs b/src-tauri/src/audio_toolkit/mod.rs
index 0f88bdc6..54c648cf 100644
--- a/src-tauri/src/audio_toolkit/mod.rs
+++ b/src-tauri/src/audio_toolkit/mod.rs
@@ -1,8 +1,10 @@
pub mod audio;
pub mod constants;
+pub mod text;
pub mod vad;
pub use audio::{
list_input_devices, list_output_devices, save_wav_file, AudioRecorder, CpalDeviceInfo,
};
+pub use text::apply_custom_words;
pub use vad::{SileroVad, VoiceActivityDetector};
diff --git a/src-tauri/src/audio_toolkit/text.rs b/src-tauri/src/audio_toolkit/text.rs
new file mode 100644
index 00000000..6f30ae95
--- /dev/null
+++ b/src-tauri/src/audio_toolkit/text.rs
@@ -0,0 +1,176 @@
+use natural::phonetics::soundex;
+use strsim::levenshtein;
+
+/// Applies custom word corrections to transcribed text using fuzzy matching
+///
+/// This function corrects words in the input text by finding the best matches
+/// from a list of custom words using a combination of:
+/// - Levenshtein distance for string similarity
+/// - Soundex phonetic matching for pronunciation similarity
+///
+/// # Arguments
+/// * `text` - The input text to correct
+/// * `custom_words` - List of custom words to match against
+/// * `threshold` - Maximum similarity score to accept (0.0 = exact match, 1.0 = any match)
+///
+/// # Returns
+/// The corrected text with custom words applied
+pub fn apply_custom_words(text: &str, custom_words: &[String], threshold: f64) -> String {
+ if custom_words.is_empty() {
+ return text.to_string();
+ }
+
+ // Pre-compute lowercase versions to avoid repeated allocations
+ let custom_words_lower: Vec<String> = custom_words.iter().map(|w| w.to_lowercase()).collect();
+
+ let words: Vec<&str> = text.split_whitespace().collect();
+ let mut corrected_words = Vec::new();
+
+ for word in words {
+ let cleaned_word = word
+ .trim_matches(|c: char| !c.is_alphabetic())
+ .to_lowercase();
+
+ if cleaned_word.is_empty() {
+ corrected_words.push(word.to_string());
+ continue;
+ }
+
+ // Skip extremely long words to avoid performance issues
+ if cleaned_word.len() > 50 {
+ corrected_words.push(word.to_string());
+ continue;
+ }
+
+ let mut best_match: Option<&String> = None;
+ let mut best_score = f64::MAX;
+
+ for (i, custom_word_lower) in custom_words_lower.iter().enumerate() {
+ // Skip if lengths are too different (optimization)
+ let len_diff = (cleaned_word.len() as i32 - custom_word_lower.len() as i32).abs();
+ if len_diff > 5 {
+ continue;
+ }
+
+ // Calculate Levenshtein distance (normalized by length)
+ let levenshtein_dist = levenshtein(&cleaned_word, custom_word_lower);
+ let max_len = cleaned_word.len().max(custom_word_lower.len()) as f64;
+ let levenshtein_score = if max_len > 0.0 {
+ levenshtein_dist as f64 / max_len
+ } else {
+ 1.0
+ };
+
+ // Calculate phonetic similarity using Soundex
+ let phonetic_match = soundex(&cleaned_word, custom_word_lower);
+
+ // Combine scores: favor phonetic matches, but also consider string similarity
+ let combined_score = if phonetic_match {
+ levenshtein_score * 0.3 // Give significant boost to phonetic matches
+ } else {
+ levenshtein_score
+ };
+
+ // Accept if the score is good enough (configurable threshold)
+ if combined_score < threshold && combined_score < best_score {
+ best_match = Some(&custom_words[i]);
+ best_score = combined_score;
+ }
+ }
+
+ if let Some(replacement) = best_match {
+ // Preserve the original case pattern as much as possible
+ let corrected = preserve_case_pattern(word, replacement);
+
+ // Preserve punctuation from original word
+ let (prefix, suffix) = extract_punctuation(word);
+ corrected_words.push(format!("{}{}{}", prefix, corrected, suffix));
+ } else {
+ corrected_words.push(word.to_string());
+ }
+ }
+
+ corrected_words.join(" ")
+}
+
+/// Preserves the case pattern of the original word when applying a replacement
+fn preserve_case_pattern(original: &str, replacement: &str) -> String {
+ if original.chars().all(|c| c.is_uppercase()) {
+ replacement.to_uppercase()
+ } else if original.chars().next().map_or(false, |c| c.is_uppercase()) {
+ let mut chars: Vec<char> = replacement.chars().collect();
+ if let Some(first_char) = chars.get_mut(0) {
+ *first_char = first_char.to_uppercase().next().unwrap_or(*first_char);
+ }
+ chars.into_iter().collect()
+ } else {
+ replacement.to_string()
+ }
+}
+
+/// Extracts punctuation prefix and suffix from a word
+fn extract_punctuation(word: &str) -> (&str, &str) {
+ let prefix_end = word.chars().take_while(|c| !c.is_alphabetic()).count();
+ let suffix_start = word
+ .char_indices()
+ .rev()
+ .take_while(|(_, c)| !c.is_alphabetic())
+ .count();
+
+ let prefix = if prefix_end > 0 {
+ &word[..prefix_end]
+ } else {
+ ""
+ };
+
+ let suffix = if suffix_start > 0 {
+ &word[word.len() - suffix_start..]
+ } else {
+ ""
+ };
+
+ (prefix, suffix)
+}
+
+#[cfg(test)]
+mod tests {
+ use super::*;
+
+ #[test]
+ fn test_apply_custom_words_exact_match() {
+ let text = "hello world";
+ let custom_words = vec!["Hello".to_string(), "World".to_string()];
+ let result = apply_custom_words(text, &custom_words, 0.5);
+ assert_eq!(result, "Hello World");
+ }
+
+ #[test]
+ fn test_apply_custom_words_fuzzy_match() {
+ let text = "helo wrold";
+ let custom_words = vec!["hello".to_string(), "world".to_string()];
+ let result = apply_custom_words(text, &custom_words, 0.5);
+ assert_eq!(result, "hello world");
+ }
+
+ #[test]
+ fn test_preserve_case_pattern() {
+ assert_eq!(preserve_case_pattern("HELLO", "world"), "WORLD");
+ assert_eq!(preserve_case_pattern("Hello", "world"), "World");
+ assert_eq!(preserve_case_pattern("hello", "WORLD"), "WORLD");
+ }
+
+ #[test]
+ fn test_extract_punctuation() {
+ assert_eq!(extract_punctuation("hello"), ("", ""));
+ assert_eq!(extract_punctuation("!hello?"), ("!", "?"));
+ assert_eq!(extract_punctuation("...hello..."), ("...", "..."));
+ }
+
+ #[test]
+ fn test_empty_custom_words() {
+ let text = "hello world";
+ let custom_words = vec![];
+ let result = apply_custom_words(text, &custom_words, 0.5);
+ assert_eq!(result, "hello world");
+ }
+}
diff --git a/src-tauri/src/managers/model.rs b/src-tauri/src/managers/model.rs
index 96a988c5..6607a813 100644
--- a/src-tauri/src/managers/model.rs
+++ b/src-tauri/src/managers/model.rs
@@ -8,6 +8,15 @@ use std::io::Write;
use std::path::PathBuf;
use std::sync::Mutex;
use tauri::{App, AppHandle, Emitter, Manager};
+use tar::Archive;
+use std::fs::File;
+use flate2::read::GzDecoder;
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub enum EngineType {
+ Whisper,
+ Parakeet,
+}
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ModelInfo {
@@ -20,6 +29,8 @@ pub struct ModelInfo {
pub is_downloaded: bool,
pub is_downloading: bool,
pub partial_size: u64,
+ pub is_directory: bool,
+ pub engine_type: EngineType,
}
#[derive(Debug, Clone, Serialize, Deserialize)]
@@ -65,6 +76,8 @@ impl ModelManager {
is_downloaded: false,
is_downloading: false,
partial_size: 0,
+ is_directory: false,
+ engine_type: EngineType::Whisper,
},
);
@@ -81,6 +94,8 @@ impl ModelManager {
is_downloaded: false,
is_downloading: false,
partial_size: 0,
+ is_directory: false,
+ engine_type: EngineType::Whisper,
},
);
@@ -96,6 +111,8 @@ impl ModelManager {
is_downloaded: false,
is_downloading: false,
partial_size: 0,
+ is_directory: false,
+ engine_type: EngineType::Whisper,
},
);
@@ -111,6 +128,26 @@ impl ModelManager {
is_downloaded: false,
is_downloading: false,
partial_size: 0,
+ is_directory: false,
+ engine_type: EngineType::Whisper,
+ },
+ );
+
+ // Add NVIDIA Parakeet model (directory-based)
+ available_models.insert(
+ "parakeet-tdt-0.6b-v3".to_string(),
+ ModelInfo {
+ id: "parakeet-tdt-0.6b-v3".to_string(),
+ name: "Parakeet V3".to_string(),
+ description: "The fastest and most accurate model".to_string(),
+ filename: "parakeet-tdt-0.6b-v3-int8".to_string(), // Directory name
+ url: Some("https://blob.handy.computer/parakeet-v3-int8.tar.gz".to_string()),
+ size_mb: 850, // Approximate size for int8 quantized model
+ is_downloaded: false,
+ is_downloading: false,
+ partial_size: 0,
+ is_directory: true,
+ engine_type: EngineType::Parakeet,
},
);
@@ -173,17 +210,41 @@ impl ModelManager {
let mut models = self.available_models.lock().unwrap();
for model in models.values_mut() {
- let model_path = self.models_dir.join(&model.filename);
- let partial_path = self.models_dir.join(format!("{}.partial", &model.filename));
+ if model.is_directory {
+ // For directory-based models, check if the directory exists
+ let model_path = self.models_dir.join(&model.filename);
+ let partial_path = self.models_dir.join(format!("{}.partial", &model.filename));
+ let extracting_path = self.models_dir.join(format!("{}.extracting", &model.filename));
+
+ // Clean up any leftover .extracting directories from interrupted extractions
+ if extracting_path.exists() {
+ println!("Cleaning up interrupted extraction for model: {}", model.id);
+ let _ = fs::remove_dir_all(&extracting_path);
+ }
- model.is_downloaded = model_path.exists();
- model.is_downloading = partial_path.exists();
+ model.is_downloaded = model_path.exists() && model_path.is_dir();
+ model.is_downloading = partial_path.exists();
- // Get partial file size if it exists
- if partial_path.exists() {
- model.partial_size = partial_path.metadata().map(|m| m.len()).unwrap_or(0);
+ // Get partial file size if it exists (for the .tar.gz being downloaded)
+ if partial_path.exists() {
+ model.partial_size = partial_path.metadata().map(|m| m.len()).unwrap_or(0);
+ } else {
+ model.partial_size = 0;
+ }
} else {
- model.partial_size = 0;
+ // For file-based models (existing logic)
+ let model_path = self.models_dir.join(&model.filename);
+ let partial_path = self.models_dir.join(format!("{}.partial", &model.filename));
+
+ model.is_downloaded = model_path.exists();
+ model.is_downloading = partial_path.exists();
+
+ // Get partial file size if it exists
+ if partial_path.exists() {
+ model.partial_size = partial_path.metadata().map(|m| m.len()).unwrap_or(0);
+ } else {
+ model.partial_size = 0;
+ }
}
}
@@ -359,8 +420,74 @@ impl ModelManager {
file.flush()?;
drop(file); // Ensure file is closed before moving
- // Move partial file to final location
- fs::rename(&partial_path, &model_path)?;
+ // Handle directory-based models (extract tar.gz) vs file-based models
+ if model_info.is_directory {
+ // Emit extraction started event
+ let _ = self.app_handle.emit("model-extraction-started", model_id);
+ println!("Extracting archive for directory-based model: {}", model_id);
+
+ // Use a temporary extraction directory to ensure atomic operations
+ let temp_extract_dir = self.models_dir.join(format!("{}.extracting", &model_info.filename));
+ let final_model_dir = self.models_dir.join(&model_info.filename);
+
+ // Clean up any previous incomplete extraction
+ if temp_extract_dir.exists() {
+ let _ = fs::remove_dir_all(&temp_extract_dir);
+ }
+
+ // Create temporary extraction directory
+ fs::create_dir_all(&temp_extract_dir)?;
+
+ // Open the downloaded tar.gz file
+ let tar_gz = File::open(&partial_path)?;
+ let tar = GzDecoder::new(tar_gz);
+ let mut archive = Archive::new(tar);
+
+ // Extract to the temporary directory first
+ archive.unpack(&temp_extract_dir).map_err(|e| {
+ let error_msg = format!("Failed to extract archive: {}", e);
+ // Clean up failed extraction
+ let _ = fs::remove_dir_all(&temp_extract_dir);
+ let _ = self.app_handle.emit("model-extraction-failed", &serde_json::json!({
+ "model_id": model_id,
+ "error": error_msg
+ }));
+ anyhow::anyhow!(error_msg)
+ })?;
+
+ // Find the actual extracted directory (archive might have a nested structure)
+ let extracted_dirs: Vec<_> = fs::read_dir(&temp_extract_dir)?
+ .filter_map(|entry| entry.ok())
+ .filter(|entry| entry.file_type().map(|ft| ft.is_dir()).unwrap_or(false))
+ .collect();
+
+ if extracted_dirs.len() == 1 {
+ // Single directory extracted, move it to the final location
+ let source_dir = extracted_dirs[0].path();
+ if final_model_dir.exists() {
+ fs::remove_dir_all(&final_model_dir)?;
+ }
+ fs::rename(&source_dir, &final_model_dir)?;
+ // Clean up temp directory
+ let _ = fs::remove_dir_all(&temp_extract_dir);
+ } else {
+ // Multiple items or no directories, rename the temp directory itself
+ if final_model_dir.exists() {
+ fs::remove_dir_all(&final_model_dir)?;
+ }
+ fs::rename(&temp_extract_dir, &final_model_dir)?;
+ }
+
+ println!("Successfully extracted archive for model: {}", model_id);
+ // Emit extraction completed event
+ let _ = self.app_handle.emit("model-extraction-completed", model_id);
+
+ // Remove the downloaded tar.gz file
+ let _ = fs::remove_file(&partial_path);
+ } else {
+ // Move partial file to final location for file-based models
+ fs::rename(&partial_path, &model_path)?;
+ }
// Update download status
{
@@ -405,15 +532,28 @@ impl ModelManager {
let mut deleted_something = false;
- // Delete complete model file if it exists
- if model_path.exists() {
- println!("ModelManager: Deleting model file at: {:?}", model_path);
- fs::remove_file(&model_path)?;
- println!("ModelManager: Model file deleted successfully");
- deleted_something = true;
+ if model_info.is_directory {
+ // Delete complete model directory if it exists
+ if model_path.exists() && model_path.is_dir() {
+ println!(
+ "ModelManager: Deleting model directory at: {:?}",
+ model_path
+ );
+ fs::remove_dir_all(&model_path)?;
+ println!("ModelManager: Model directory deleted successfully");
+ deleted_something = true;
+ }
+ } else {
+ // Delete complete model file if it exists
+ if model_path.exists() {
+ println!("ModelManager: Deleting model file at: {:?}", model_path);
+ fs::remove_file(&model_path)?;
+ println!("ModelManager: Model file deleted successfully");
+ deleted_something = true;
+ }
}
- // Delete partial file if it exists
+ // Delete partial file if it exists (same for both types)
if partial_path.exists() {
println!("ModelManager: Deleting partial file at: {:?}", partial_path);
fs::remove_file(&partial_path)?;
@@ -441,7 +581,7 @@ impl ModelManager {
return Err(anyhow::anyhow!("Model not available: {}", model_id));
}
- // Ensure we don't return partial files
+ // Ensure we don't return partial files/directories
if model_info.is_downloading {
return Err(anyhow::anyhow!(
"Model is currently downloading: {}",
@@ -454,14 +594,26 @@ impl ModelManager {
.models_dir
.join(format!("{}.partial", &model_info.filename));
- // Ensure we only return complete model files, not partial ones
- if model_path.exists() && !partial_path.exists() {
- Ok(model_path)
+ if model_info.is_directory {
+ // For directory-based models, ensure the directory exists and is complete
+ if model_path.exists() && model_path.is_dir() && !partial_path.exists() {
+ Ok(model_path)
+ } else {
+ Err(anyhow::anyhow!(
+ "Complete model directory not found: {}",
+ model_id
+ ))
+ }
} else {
- Err(anyhow::anyhow!(
- "Complete model file not found: {}",
- model_id
- ))
+ // For file-based models (existing logic)
+ if model_path.exists() && !partial_path.exists() {
+ Ok(model_path)
+ } else {
+ Err(anyhow::anyhow!(
+ "Complete model file not found: {}",
+ model_id
+ ))
+ }
}
}
diff --git a/src-tauri/src/managers/transcription.rs b/src-tauri/src/managers/transcription.rs
index 93ebf2af..2d753b87 100644
--- a/src-tauri/src/managers/transcription.rs
+++ b/src-tauri/src/managers/transcription.rs
@@ -1,17 +1,22 @@
-use crate::managers::model::ModelManager;
+use crate::audio_toolkit::apply_custom_words;
+use crate::managers::model::{EngineType, ModelManager};
use crate::settings::{get_settings, ModelUnloadTimeout};
use anyhow::Result;
use log::debug;
-use natural::phonetics::soundex;
use serde::Serialize;
use std::sync::atomic::{AtomicBool, AtomicU64, Ordering};
use std::sync::{Arc, Mutex};
use std::thread;
use std::time::{Duration, SystemTime};
-use strsim::levenshtein;
use tauri::{App, AppHandle, Emitter, Manager};
-use whisper_rs::{
- FullParams, SamplingStrategy, WhisperContext, WhisperContextParameters, WhisperState,
+use transcribe_rs::{
+ engines::{
+ parakeet::{
+ ParakeetEngine, ParakeetInferenceParams, ParakeetModelParams, TimestampGranularity,
+ },
+ whisper::{WhisperEngine, WhisperInferenceParams},
+ },
+ TranscriptionEngine,
};
#[derive(Clone, Debug, Serialize)]
@@ -22,10 +27,14 @@ pub struct ModelStateEvent {
pub error: Option<String>,
}
+enum LoadedEngine {
+ Whisper(WhisperEngine),
+ Parakeet(ParakeetEngine),
+}
+
#[derive(Clone)]
pub struct TranscriptionManager {
- state: Arc<Mutex<Option<WhisperState>>>,
- context: Arc<Mutex<Option<WhisperContext>>>,
+ engine: Arc<Mutex<Option<LoadedEngine>>>,
model_manager: Arc<ModelManager>,
app_handle: AppHandle,
current_model_id: Arc<Mutex<Option<String>>>,
@@ -34,121 +43,12 @@ pub struct TranscriptionManager {
watcher_handle: Arc<Mutex<Option<thread::JoinHandle<()>>>>,
}
-fn apply_custom_words(text: &str, custom_words: &[String], threshold: f64) -> String {
- if custom_words.is_empty() {
- return text.to_string();
- }
-
- // Pre-compute lowercase versions to avoid repeated allocations
- let custom_words_lower: Vec<String> = custom_words.iter().map(|w| w.to_lowercase()).collect();
-
- let words: Vec<&str> = text.split_whitespace().collect();
- let mut corrected_words = Vec::new();
-
- for word in words {
- let cleaned_word = word
- .trim_matches(|c: char| !c.is_alphabetic())
- .to_lowercase();
-
- if cleaned_word.is_empty() {
- corrected_words.push(word.to_string());
- continue;
- }
-
- // Skip extremely long words to avoid performance issues
- if cleaned_word.len() > 50 {
- corrected_words.push(word.to_string());
- continue;
- }
-
- let mut best_match: Option<&String> = None;
- let mut best_score = f64::MAX;
-
- for (i, custom_word_lower) in custom_words_lower.iter().enumerate() {
- // Skip if lengths are too different (optimization)
- let len_diff = (cleaned_word.len() as i32 - custom_word_lower.len() as i32).abs();
- if len_diff > 5 {
- continue;
- }
-
- // Calculate Levenshtein distance (normalized by length)
- let levenshtein_dist = levenshtein(&cleaned_word, custom_word_lower);
- let max_len = cleaned_word.len().max(custom_word_lower.len()) as f64;
- let levenshtein_score = if max_len > 0.0 {
- levenshtein_dist as f64 / max_len
- } else {
- 1.0
- };
-
- // Calculate phonetic similarity using Soundex
- let phonetic_match = soundex(&cleaned_word, custom_word_lower);
-
- // Combine scores: favor phonetic matches, but also consider string similarity
- let combined_score = if phonetic_match {
- levenshtein_score * 0.3 // Give significant boost to phonetic matches
- } else {
- levenshtein_score
- };
-
- // Accept if the score is good enough (configurable threshold)
- if combined_score < threshold && combined_score < best_score {
- best_match = Some(&custom_words[i]);
- best_score = combined_score;
- }
- }
-
- if let Some(replacement) = best_match {
- // Preserve the original case pattern as much as possible
- let corrected = if word.chars().all(|c| c.is_uppercase()) {
- replacement.to_uppercase()
- } else if word.chars().next().map_or(false, |c| c.is_uppercase()) {
- let mut chars: Vec<char> = replacement.chars().collect();
- if let Some(first_char) = chars.get_mut(0) {
- *first_char = first_char.to_uppercase().next().unwrap_or(*first_char);
- }
- chars.into_iter().collect()
- } else {
- replacement.clone()
- };
-
- // Preserve punctuation from original word - optimized version
- let prefix_end = word.chars().take_while(|c| !c.is_alphabetic()).count();
- let suffix_start = word
- .char_indices()
- .rev()
- .take_while(|(_, c)| !c.is_alphabetic())
- .count();
-
- let original_prefix = if prefix_end > 0 {
- &word[..prefix_end]
- } else {
- ""
- };
- let original_suffix = if suffix_start > 0 {
- &word[word.len() - suffix_start..]
- } else {
- ""
- };
-
- corrected_words.push(format!(
- "{}{}{}",
- original_prefix, corrected, original_suffix
- ));
- } else {
- corrected_words.push(word.to_string());
- }
- }
-
- corrected_words.join(" ")
-}
-
impl TranscriptionManager {
pub fn new(app: &App, model_manager: Arc<ModelManager>) -> Result<Self> {
let app_handle = app.app_handle().clone();
let manager = Self {
- state: Arc::new(Mutex::new(None)),
- context: Arc::new(Mutex::new(None)),
+ engine: Arc::new(Mutex::new(None)),
model_manager,
app_handle: app_handle.clone(),
current_model_id: Arc::new(Mutex::new(None)),
@@ -230,8 +130,8 @@ impl TranscriptionManager {
}
pub fn is_model_loaded(&self) -> bool {
- let state = self.state.lock().unwrap();
- state.is_some()
+ let engine = self.engine.lock().unwrap();
+ engine.is_some()
}
pub fn unload_model(&self) -> Result<()> {
@@ -239,12 +139,14 @@ impl TranscriptionManager {
debug!("Starting to unload model");
{
- let mut state = self.state.lock().unwrap();
- *state = None; // Dropping state frees GPU/CPU memory
- }
- {
- let mut context = self.context.lock().unwrap();
- *context = None; // Dropping context frees additional memory
+ let mut engine = self.engine.lock().unwrap();
+ if let Some(ref mut loaded_engine) = *engine {
+ match loaded_engine {
+ LoadedEngine::Whisper(ref mut whisper) => whisper.unload_model(),
+ LoadedEngine::Parakeet(ref mut parakeet) => parakeet.unload_model(),
+ }
+ }
+ *engine = None; // Drop the engine to free memory
}
{
let mut current_model = self.current_model_id.lock().unwrap();
@@ -306,19 +208,16 @@ impl TranscriptionManager {
let model_path = self.model_manager.get_model_path(model_id)?;
- let path_str = model_path
- .to_str()
- .ok_or_else(|| anyhow::anyhow!("Invalid path for model: {}", model_id))?;
-
println!(
- "Loading transcription model {} from: {}",
- model_id, path_str
+ "Loading transcription model {} from: {:?}",
+ model_id, model_path
);
- // Create new context
- let context =
- WhisperContext::new_with_params(path_str, WhisperContextParameters::default())
- .map_err(|e| {
+ // Create appropriate engine based on model type
+ let loaded_engine = match model_info.engine_type {
+ EngineType::Whisper => {
+ let mut engine = WhisperEngine::new();
+ engine.load_model(&model_path).map_err(|e| {
let error_msg = format!("Failed to load whisper model {}: {}", model_id, e);
let _ = self.app_handle.emit(
"model-state-changed",
@@ -331,30 +230,34 @@ impl TranscriptionManager {
);
anyhow::anyhow!(error_msg)
})?;
+ LoadedEngine::Whisper(engine)
+ }
+ EngineType::Parakeet => {
+ let mut engine = ParakeetEngine::new();
+ engine
+ .load_model_with_params(&model_path, ParakeetModelParams::int8())
+ .map_err(|e| {
+ let error_msg =
+ format!("Failed to load parakeet model {}: {}", model_id, e);
+ let _ = self.app_handle.emit(
+ "model-state-changed",
+ ModelStateEvent {
+ event_type: "loading_failed".to_string(),
+ model_id: Some(model_id.to_string()),
+ model_name: Some(model_info.name.clone()),
+ error: Some(error_msg.clone()),
+ },
+ );
+ anyhow::anyhow!(error_msg)
+ })?;
+ LoadedEngine::Parakeet(engine)
+ }
+ };
- // Create new state
- let state = context.create_state().map_err(|e| {
- let error_msg = format!("Failed to create state for model {}: {}", model_id, e);
- let _ = self.app_handle.emit(
- "model-state-changed",
- ModelStateEvent {
- event_type: "loading_failed".to_string(),
- model_id: Some(model_id.to_string()),
- model_name: Some(model_info.name.clone()),
- error: Some(error_msg.clone()),
- },
- );
- anyhow::anyhow!(error_msg)
- })?;
-
- // Update the current context and state
- {
- let mut current_context = self.context.lock().unwrap();
- *current_context = Some(context);
- }
+ // Update the current engine and model ID
{
- let mut current_state = self.state.lock().unwrap();
- *current_state = Some(state);
+ let mut engine = self.engine.lock().unwrap();
+ *engine = Some(loaded_engine);
}
{
let mut current_model = self.current_model_id.lock().unwrap();
@@ -398,18 +301,17 @@ impl TranscriptionManager {
let st = std::time::Instant::now();
- let mut result = String::new();
println!("Audio vector length: {}", audio.len());
if audio.len() == 0 {
println!("Empty audio vector");
- return Ok(result);
+ return Ok(String::new());
}
// Check if model is loaded, if not try to load it
{
- let state_guard = self.state.lock().unwrap();
- if state_guard.is_none() {
+ let engine_guard = self.engine.lock().unwrap();
+ if engine_guard.is_none() {
// Model not loaded, try to load the selected model from settings
let settings = get_settings(&self.app_handle);
println!(
@@ -418,7 +320,7 @@ impl TranscriptionManager {
);
// Drop the guard before calling load_model to avoid deadlock
- drop(state_guard);
+ drop(engine_guard);
// Try to load the model
if let Err(e) = self.load_model(&settings.selected_model) {
@@ -430,60 +332,55 @@ impl TranscriptionManager {
}
}
- let mut state_guard = self.state.lock().unwrap();
- let state = state_guard.as_mut().ok_or_else(|| {
- anyhow::anyhow!(
- "Model failed to load after auto-load attempt. Please check your model settings."
- )
- })?;
-
- // Get current settings to check translation preference
+ // Get current settings for configuration
let settings = get_settings(&self.app_handle);
- // Initialize parameters
- let mut params = FullParams::new(SamplingStrategy::BeamSearch {
- beam_size: 3,
- patience: -1.0,
- });
- let language = Some(settings.selected_language.as_str());
- params.set_language(language);
- params.set_print_special(false);
- params.set_print_progress(false);
- params.set_print_realtime(false);
- params.set_print_timestamps(false);
- params.set_suppress_blank(true);
- params.set_suppress_non_speech_tokens(true);
- params.set_no_speech_thold(0.2);
-
- // Enable translation to English if requested
- if settings.translate_to_english {
- params.set_translate(true);
- }
-
- state
- .full(params, &audio)
- .expect("failed to convert samples");
-
- let num_segments = state
- .full_n_segments()
- .expect("failed to get number of segments");
+ // Perform transcription with the appropriate engine
+ let result = {
+ let mut engine_guard = self.engine.lock().unwrap();
+ let engine = engine_guard.as_mut().ok_or_else(|| {
+ anyhow::anyhow!(
+ "Model failed to load after auto-load attempt. Please check your model settings."
+ )
+ })?;
+
+ match engine {
+ LoadedEngine::Whisper(whisper_engine) => {
+ let params = WhisperInferenceParams {
+ language: if settings.selected_language == "auto" {
+ None
+ } else {
+ Some(settings.selected_language.clone())
+ },
+ ..Default::default()
+ };
- for i in 0..num_segments {
- let segment = state
- .full_get_segment_text(i)
- .expect("failed to get segment");
- result.push_str(&segment);
- }
+ whisper_engine
+ .transcribe_samples(audio, Some(params))
+ .map_err(|e| anyhow::anyhow!("Whisper transcription failed: {}", e))?
+ }
+ LoadedEngine::Parakeet(parakeet_engine) => {
+ let params = ParakeetInferenceParams {
+ timestamp_granularity: TimestampGranularity::Segment,
+ ..Default::default()
+ };
+
+ parakeet_engine
+ .transcribe_samples(audio, Some(params))
+ .map_err(|e| anyhow::anyhow!("Parakeet transcription failed: {}", e))?
+ }
+ }
+ };
// Apply word correction if custom words are configured
let corrected_result = if !settings.custom_words.is_empty() {
apply_custom_words(
- &result,
+ &result.text,
&settings.custom_words,
settings.word_correction_threshold,
)
} else {
- result
+ result.text
};
let et = std::time::Instant::now();
@@ -497,8 +394,6 @@ impl TranscriptionManager {
// Check if we should immediately unload the model after transcription
if settings.model_unload_timeout == ModelUnloadTimeout::Immediately {
println!("⚡ Immediately unloading model after transcription");
- // Drop the state guard first to avoid deadlock
- drop(state_guard);
if let Err(e) = self.unload_model() {
eprintln!("Failed to immediately unload model: {}", e);
}
diff --git a/src/components/model-selector/ModelSelector.tsx b/src/components/model-selector/ModelSelector.tsx
index 08809233..88a7b71f 100644
--- a/src/components/model-selector/ModelSelector.tsx
+++ b/src/components/model-selector/ModelSelector.tsx
@@ -20,7 +20,7 @@ interface DownloadProgress {
percentage: number;
}
-type ModelStatus = "ready" | "loading" | "downloading" | "error" | "none";
+type ModelStatus = "ready" | "loading" | "downloading" | "extracting" | "error" | "none";
interface DownloadStats {
startTime: number;
@@ -45,6 +45,9 @@ const ModelSelector: React.FC<ModelSelectorProps> = ({ onError }) => {
const [downloadStats, setDownloadStats] = useState<
Map<string, DownloadStats>
>(new Map());
+ const [extractingModels, setExtractingModels] = useState<Set<string>>(
+ new Set(),
+ );
const dropdownRef = useRef<HTMLDivElement>(null);
@@ -156,6 +159,49 @@ const ModelSelector: React.FC<ModelSelectorProps> = ({ onError }) => {
},
);
+ // Listen for extraction events
+ const extractionStartedUnlisten = listen<string>(
+ "model-extraction-started",
+ (event) => {
+ const modelId = event.payload;
+ setExtractingModels((prev) => new Set(prev.add(modelId)));
+ setModelStatus("extracting");
+ },
+ );
+
+ const extractionCompletedUnlisten = listen<string>(
+ "model-extraction-completed",
+ (event) => {
+ const modelId = event.payload;
+ setExtractingModels((prev) => {
+ const next = new Set(prev);
+ next.delete(modelId);
+ return next;
+ });
+ loadModels(); // Refresh models list
+
+ // Auto-select the newly extracted model
+ setTimeout(() => {
+ loadCurrentModel();
+ handleModelSelect(modelId);
+ }, 500);
+ },
+ );
+
+ const extractionFailedUnlisten = listen<{model_id: string, error: string}>(
+ "model-extraction-failed",
+ (event) => {
+ const modelId = event.payload.model_id;
+ setExtractingModels((prev) => {
+ const next = new Set(prev);
+ next.delete(modelId);
+ return next;
+ });
+ setModelError(`Failed to extract model: ${event.payload.error}`);
+ setModelStatus("error");
+ },
+ );
+
// Click outside to close dropdown
const handleClickOutside = (event: MouseEvent) => {
if (
@@ -173,6 +219,9 @@ const ModelSelector: React.FC<ModelSelectorProps> = ({ onError }) => {
modelStateUnlisten.then((fn) => fn());
downloadProgressUnlisten.then((fn) => fn());
downloadCompleteUnlisten.then((fn) => fn());
+ extractionStartedUnlisten.then((fn) => fn());
+ extractionCompletedUnlisten.then((fn) => fn());
+ extractionFailedUnlisten.then((fn) => fn());
};
}, []);
@@ -241,6 +290,16 @@ const ModelSelector: React.FC<ModelSelectorProps> = ({ onError }) => {
};
const getModelDisplayText = (): string => {
+ if (extractingModels.size > 0) {
+ if (extractingModels.size === 1) {
+ const [modelId] = Array.from(extractingModels);
+ const model = models.find(m => m.id === modelId);
+ return `Extracting ${model?.name || 'Model'}...`;
+ } else {
+ return `Extracting ${extractingModels.size} models...`;
+ }
+ }
+
if (modelDownloadProgress.size > 0) {
if (modelDownloadProgress.size === 1) {
const [progress] = Array.from(modelDownloadProgress.values());
@@ -261,6 +320,8 @@ const ModelSelector: React.FC<ModelSelectorProps> = ({ onError }) => {
return currentModel?.name || "Model Ready";
case "loading":
return currentModel ? `Loading ${currentModel.name}...` : "Loading...";
+ case "extracting":
+ return currentModel ? `Extracting ${currentModel.name}...` : "Extracting...";
case "error":
return modelError || "Model Error";
case "none":
diff --git a/src/components/settings/LanguageSelector.tsx b/src/components/settings/LanguageSelector.tsx
index c0ec7051..0cd6c01b 100644
--- a/src/components/settings/LanguageSelector.tsx
+++ b/src/components/settings/LanguageSelector.tsx
@@ -1,7 +1,9 @@
import React, { useState, useRef, useEffect, useMemo } from "react";
+import { listen } from "@tauri-apps/api/event";
import { SettingContainer } from "../ui/SettingContainer";
import { ResetButton } from "../ui/ResetButton";
import { useSettings } from "../../hooks/useSettings";
+import { useModels } from "../../hooks/useModels";
import { LANGUAGES } from "../../lib/constants/languages";
interface LanguageSelectorProps {
@@ -14,12 +16,15 @@ export const LanguageSelector: React.FC<LanguageSelectorProps> = ({
grouped = false,
}) => {
const { getSetting, updateSetting, resetSetting, isUpdating } = useSettings();
+ const { currentModel, loadCurrentModel } = useModels();
const [isOpen, setIsOpen] = useState(false);
const [searchQuery, setSearchQuery] = useState("");
const dropdownRef = useRef<HTMLDivElement>(null);
const searchInputRef = useRef<HTMLInputElement>(null);
const selectedLanguage = getSetting("selected_language") || "auto";
+ const isParakeetModel = currentModel === "parakeet-tdt-0.6b-v3";
+ const isLanguageSelectionDisabled = isParakeetModel;
useEffect(() => {
const handleClickOutside = (event: MouseEvent) => {
@@ -38,6 +43,17 @@ export const LanguageSelector: React.FC<LanguageSelectorProps> = ({
};
}, []);
+ // Listen for model state changes to update UI reactively
+ useEffect(() => {
+ const modelStateUnlisten = listen("model-state-changed", () => {
+ loadCurrentModel();
+ });
+
+ return () => {
+ modelStateUnlisten.then((fn) => fn());
+ };
+ }, [loadCurrentModel]);
+
useEffect(() => {
if (isOpen && searchInputRef.current) {
searchInputRef.current.focus();
@@ -51,8 +67,9 @@ export const LanguageSelector: React.FC<LanguageSelectorProps> = ({
[searchQuery]
);
- const selectedLanguageName =
- LANGUAGES.find((lang) => lang.value === selectedLanguage)?.label || "Auto";
+ const selectedLanguageName = isParakeetModel
+ ? "Auto"
+ : LANGUAGES.find((lang) => lang.value === selectedLanguage)?.label || "Auto";
const handleLanguageSelect = async (languageCode: string) => {
await updateSetting("selected_language", languageCode);
@@ -65,7 +82,7 @@ export const LanguageSelector: React.FC<LanguageSelectorProps> = ({
};
const handleToggle = () => {
- if (isUpdating("selected_language")) return;
+ if (isUpdating("selected_language") || isLanguageSelectionDisabled) return;
setIsOpen(!isOpen);
};
@@ -86,7 +103,10 @@ export const LanguageSelector: React.FC<LanguageSelectorProps> = ({
return (
<SettingContainer
title="Language"
- description="Select the language for speech recognition. Auto will automatically determine the language, while selecting a specific language can improve accuracy for that language."
+ description={isParakeetModel
+ ? "Parakeet model automatically detects the language. No manual selection is needed."
+ : "Select the language for speech recognition. Auto will automatically determine the language, while selecting a specific language can improve accuracy for that language."
+ }
descriptionMode={descriptionMode}
grouped={grouped}
>
@@ -95,12 +115,12 @@ export const LanguageSelector: React.FC<LanguageSelectorProps> = ({
<button
type="button"
className={`px-2 py-1 text-sm font-semibold bg-mid-gray/10 border border-mid-gray/80 rounded min-w-[200px] text-left flex items-center justify-between transition-all duration-150 ${
- isUpdating("selected_language")
+ isUpdating("selected_language") || isLanguageSelectionDisabled
? "opacity-50 cursor-not-allowed"
: "hover:bg-logo-primary/10 cursor-pointer hover:border-logo-primary"
}`}
onClick={handleToggle}
- disabled={isUpdating("selected_language")}
+ disabled={isUpdating("selected_language") || isLanguageSelectionDisabled}
>
<span className="truncate">{selectedLanguageName}</span>
<svg
@@ -120,7 +140,7 @@ export const LanguageSelector: React.FC<LanguageSelectorProps> = ({
</svg>
</button>
- {isOpen && !isUpdating("selected_language") && (
+ {isOpen && !isUpdating("selected_language") && !isLanguageSelectionDisabled && (
<div className="absolute top-full left-0 right-0 mt-1 bg-background border border-mid-gray/80 rounded shadow-lg z-50 max-h-60 overflow-hidden">
{/* Search input */}
<div className="p-2 border-b border-mid-gray/80">
@@ -164,7 +184,7 @@ export const LanguageSelector: React.FC<LanguageSelectorProps> = ({
</div>
<ResetButton
onClick={handleReset}
- disabled={isUpdating("selected_language")}
+ disabled={isUpdating("selected_language") || isLanguageSelectionDisabled}
/>
</div>
{isUpdating("selected_language") && (
diff --git a/src/hooks/useModels.ts b/src/hooks/useModels.ts
index a799c2a3..7559b237 100644
--- a/src/hooks/useModels.ts
+++ b/src/hooks/useModels.ts
@@ -5,11 +5,14 @@ import { listen } from "@tauri-apps/api/event";
interface ModelInfo {
id: string;
name: string;
+ description: string;
filename: string;
url?: string;
size_mb: number;
is_downloaded: boolean;
- is_bundled: boolean;
+ is_downloading: boolean;
+ partial_size: number;
+ is_directory: boolean;
}
interface DownloadProgress {
@@ -25,6 +28,9 @@ export const useModels = () => {
const [downloadingModels, setDownloadingModels] = useState<Set<string>>(
new Set(),
);
+ const [extractingModels, setExtractingModels] = useState<Set<string>>(
+ new Set(),
+ );
const [downloadProgress, setDownloadProgress] = useState<
Map<string, DownloadProgress>
>(new Map());
@@ -117,6 +123,10 @@ export const useModels = () => {
return downloadingModels.has(modelId);
};
+ const isModelExtracting = (modelId: string): boolean => {
+ return extractingModels.has(modelId);
+ };
+
const getDownloadProgress = (
modelId: string,
): DownloadProgress | undefined => {
@@ -158,9 +168,48 @@ export const useModels = () => {
},
);
+ // Listen for extraction events
+ const extractionStartedUnlisten = listen<string>(
+ "model-extraction-started",
+ (event) => {
+ const modelId = event.payload;
+ setExtractingModels((prev) => new Set(prev.add(modelId)));
+ },
+ );
+
+ const extractionCompletedUnlisten = listen<string>(
+ "model-extraction-completed",
+ (event) => {
+ const modelId = event.payload;
+ setExtractingModels((prev) => {
+ const next = new Set(prev);
+ next.delete(modelId);
+ return next;
+ });
+ // Refresh models list to update download status
+ loadModels();
+ },
+ );
+
+ const extractionFailedUnlisten = listen<{model_id: string, error: string}>(
+ "model-extraction-failed",
+ (event) => {
+ const modelId = event.payload.model_id;
+ setExtractingModels((prev) => {
+ const next = new Set(prev);
+ next.delete(modelId);
+ return next;
+ });
+ setError(`Failed to extract model: ${event.payload.error}`);
+ },
+ );
+
return () => {
progressUnlisten.then((fn) => fn());
completeUnlisten.then((fn) => fn());
+ extractionStartedUnlisten.then((fn) => fn());
+ extractionCompletedUnlisten.then((fn) => fn());
+ extractionFailedUnlisten.then((fn) => fn());
};
}, []);
@@ -170,6 +219,7 @@ export const useModels = () => {
loading,
error,
downloadingModels,
+ extractingModels,
downloadProgress,
hasAnyModels,
isFirstRun,
@@ -181,6 +231,7 @@ export const useModels = () => {
deleteModel,
getModelInfo,
isModelDownloading,
+ isModelExtracting,
getDownloadProgress,
};
};
diff --git a/src/lib/types.ts b/src/lib/types.ts
index 3f99dcfa..68000ddc 100644
--- a/src/lib/types.ts
+++ b/src/lib/types.ts
@@ -68,10 +68,13 @@ export const ModelInfoSchema = z.object({
id: z.string(),
name: z.string(),
description: z.string(),
+ filename: z.string(),
+ url: z.string().optional(),
size_mb: z.number(),
is_downloaded: z.boolean(),
is_downloading: z.boolean(),
partial_size: z.number(),
+ is_directory: z.boolean(),
});
export type ModelInfo = z.infer<typeof ModelInfoSchema>;