Merge pull request #185 from solaoi/main

v0.9.24

package.json

@@ -1,7 +1,7 @@
 {
   "name": "lycoris",
   "private": true,
-  "version": "0.9.23",
+  "version": "0.9.24",
   "type": "module",
   "license": "MIT",
   "engines": {

src-tauri/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lycoris"
-version = "0.9.23"
+version = "0.9.24"
 description = "Lycoris is an offline voice memo"
 authors = ["solaoi"]
 license = "MIT"

src-tauri/migrations/001.sql

@@ -13,6 +13,10 @@ CREATE TABLE speeches (
     created_at_unixtime INTEGER DEFAULT (CAST(strftime('%s', 'now') AS INTEGER)),
     content TEXT,
     content_2 TEXT,
+    is_done_with_hybrid_reazonspeech INTEGER DEFAULT 0,
+    is_done_with_hybrid_whisper INTEGER DEFAULT 0,
+    hybrid_reazonspeech_content TEXT,
+    hybrid_whisper_content TEXT,
     wav TEXT,
     model TEXT,
     -- manual|vosk|whisper
@@ -178,4 +182,8 @@ VALUES("jvnv-F2-jp", "style-bert-vits2-voice");
 INSERT INTO models(model_name, model_type)
 VALUES("jvnv-M1-jp", "style-bert-vits2-voice");
 INSERT INTO models(model_name, model_type)
-VALUES("jvnv-M2-jp", "style-bert-vits2-voice");
+VALUES("jvnv-M2-jp", "style-bert-vits2-voice");
+CREATE INDEX idx_hybrid_status ON speeches(
+    is_done_with_hybrid_reazonspeech,
+    is_done_with_hybrid_whisper
+);

src-tauri/src/main.rs

@@ -4,7 +4,8 @@
 )]
 
 use tauri::{
-    http::{HttpRange, ResponseBuilder}, AppHandle, Manager, PathResolver, State, Window
+    http::{HttpRange, ResponseBuilder},
+    AppHandle, Manager, PathResolver, State, Window,
 };
 use tauri_plugin_sql::{Migration, MigrationKind};
 
@@ -41,6 +42,7 @@ use module::{
     synthesizer::{self, Synthesizer},
     transcription::{TraceCompletion, Transcription},
     transcription_amivoice::TranscriptionAmivoice,
+    transcription_hybrid::TranscriptionHybrid,
     transcription_ja::TranscriptionJa,
     transcription_online::TranscriptionOnline,
     translation_en::TranslationEn,
@@ -325,6 +327,9 @@ fn start_trace_command(
         } else if transcription_accuracy.starts_with("reazonspeech") {
             let mut transcription_ja = TranscriptionJa::new(app_handle, note_id);
             transcription_ja.start(stop_convert_rx, true);
+        } else if transcription_accuracy.starts_with("hybrid-transcript") {
+            let mut transcription_hybrid = TranscriptionHybrid::new(app_handle, note_id);
+            transcription_hybrid.start(stop_convert_rx, true);
         } else {
             let mut transcription = Transcription::new(
                 app_handle,

src-tauri/src/module/action.rs

@@ -342,7 +342,7 @@ impl Action {
     }
 
     pub fn execute(&mut self) {
-        if self.token == "" {
+        if self.token.is_empty() {
             println!("whisper token is empty, so skipping...");
             return;
         }

src-tauri/src/module/chat_online.rs

@@ -36,9 +36,9 @@ impl ChatOnline {
         }
     }
 
-    pub fn start(&mut self, stop_convert_rx: Receiver<()>, is_continuous: bool) {
+    pub fn start(&mut self, stop_convert_rx: Receiver<()>, use_no_vosk_queue_terminate_mode: bool) {
         while Self::convert(self).is_ok() {
-            if is_continuous {
+            if use_no_vosk_queue_terminate_mode {
                 let vosk_speech = self.sqlite.select_vosk(self.note_id);
                 if vosk_speech.is_err() {
                     self.app_handle
@@ -176,10 +176,10 @@ impl ChatOnline {
         );
         headers.insert(CONTENT_TYPE, HeaderValue::from_static("application/json"));
 
-        let post_body = if template != "" {
-            if functions != "" {
+        let post_body = if !template.is_empty() {
+            if !functions.is_empty() {
                 let func: Value = serde_json::from_str(functions).unwrap();
-                if function_call != "" {
+                if !function_call.is_empty() {
                     json!({
                     "model": model,
                     "temperature": temperature,
@@ -204,9 +204,9 @@ impl ChatOnline {
                 })
             }
         } else {
-            if functions != "" {
+            if !functions.is_empty() {
                 let func: Value = serde_json::from_str(functions).unwrap();
-                if function_call != "" {
+                if !function_call.is_empty() {
                     json!({
                       "model": model,
                       "temperature": temperature,
@@ -243,7 +243,7 @@ impl ChatOnline {
         let json_response: Value = response.json().await?;
 
         let response_text = if status == 200 {
-            if functions != "" {
+            if !functions.is_empty() {
                 let name = serde_json::to_string(
                     &json_response["choices"][0]["message"]["function_call"]["name"],
                 )
@@ -286,7 +286,7 @@ impl ChatOnline {
                 let result = self.sqlite.select_ai_resource();
                 let resource = if result.is_ok() {
                     let command = result.unwrap().replace("{{question}}", &question);
-                    if command == "" {
+                    if command.is_empty() {
                         "".to_string()
                     } else {
                         let result = std::process::Command::new("sh")
@@ -305,7 +305,7 @@ impl ChatOnline {
                 };
                 let result = self.sqlite.select_ai_template();
                 let template = if result.is_ok() {
-                    if resource == "" {
+                    if resource.is_empty() {
                         result.unwrap()
                     } else {
                         result
@@ -352,7 +352,7 @@ impl ChatOnline {
                     } else {
                         "".to_string()
                     };
-                    let output = if hook != "" {
+                    let output = if !hook.is_empty() {
                         let command = hook
                             .replace("{{resource}}", &resource)
                             .replace("{{answer}}", &answer)
@@ -370,7 +370,7 @@ impl ChatOnline {
                     } else {
                         "".to_string()
                     };
-                    let message = if output != "" {
+                    let message = if !output.is_empty() {
                         format!("Q. {}\nA. {}\nCLI. {}", question, answer, output)
                     } else {
                         format!("Q. {}\nA. {}", question, answer)

src-tauri/src/module/deleter.rs

@@ -21,7 +21,7 @@ impl NoteDeleter {
         let sqlite = Sqlite::new();
         let speeches = sqlite.select_all_speeches_by(note_id).unwrap();
         let _ = speeches.iter().for_each(|speech| {
-            if speech.wav != "" {
+            if !speech.wav.is_empty() {
                 if let Err(e) = remove_file(&speech.wav) {
                     eprintln!("Failed to delete file {}: {}", &speech.wav, e);
                 }

src-tauri/src/module/mod.rs

@@ -15,8 +15,12 @@ mod sqlite;
 pub mod synthesizer;
 mod transcriber;
 pub mod transcription;
-pub mod transcription_ja;
 pub mod transcription_amivoice;
+pub mod transcription_hybrid;
+mod transcription_hybrid_online;
+mod transcription_hybrid_reazonspeech;
+mod transcription_hybrid_whisper;
+pub mod transcription_ja;
 pub mod transcription_online;
 pub mod translation_en;
 pub mod translation_ja;

src-tauri/src/module/record.rs

@@ -24,8 +24,8 @@ use tauri::{api::path::data_dir, AppHandle, Manager};
 
 use super::{
     chat_online, recognizer::MyRecognizer, sqlite::Sqlite, transcription, transcription_amivoice,
-    transcription_ja, transcription_online, translation_en, translation_ja, translation_ja_high,
-    writer::Writer,
+    transcription_hybrid, transcription_ja, transcription_online, translation_en, translation_ja,
+    translation_ja_high, writer::Writer,
 };
 
 pub struct Record {
@@ -258,6 +258,17 @@ impl Record {
                                 if let Some(singleton) = lock.as_mut() {
                                     singleton.start(stop_convert_rx_clone, false);
                                 }
+                            } else if transcription_accuracy_clone.starts_with("hybrid-transcript")
+                            {
+                                transcription_hybrid::initialize_transcription_hybrid(
+                                    app_handle_clone,
+                                    note_id,
+                                );
+                                let mut lock =
+                                    transcription_hybrid::SINGLETON_INSTANCE.lock().unwrap();
+                                if let Some(singleton) = lock.as_mut() {
+                                    singleton.start(stop_convert_rx_clone, false);
+                                }
                             } else {
                                 transcription::initialize_transcription(
                                     app_handle_clone,
@@ -300,6 +311,7 @@ impl Record {
             translation_ja_high::drop_translation_ja_high();
             transcription_online::drop_transcription_online();
             transcription_amivoice::drop_transcription_amivoice();
+            transcription_hybrid::drop_transcription_hybrid();
             chat_online::drop_chat_online();
         } else {
             drop(stop_convert_tx)

src-tauri/src/module/record_desktop.rs

@@ -38,7 +38,7 @@ use screencapturekit::{
 use vosk::Recognizer;
 
 use super::{
-    chat_online, recognizer::MyRecognizer, sqlite::Sqlite, transcription, transcription_amivoice, transcription_ja, transcription_online, translation_en, translation_ja, translation_ja_high, writer::Writer
+    chat_online, recognizer::MyRecognizer, sqlite::Sqlite, transcription, transcription_amivoice, transcription_hybrid, transcription_ja, transcription_online, translation_en, translation_ja, translation_ja_high, writer::Writer
 };
 
 pub struct RecordDesktop {
@@ -288,6 +288,15 @@ impl RecordDesktop {
                                 if let Some(singleton) = lock.as_mut() {
                                     singleton.start(stop_convert_rx_clone, false);
                                 }
+                            } else if transcription_accuracy_clone.starts_with("hybrid-transcript") {
+                                transcription_hybrid::initialize_transcription_hybrid(
+                                    app_handle_clone,
+                                    note_id,
+                                );
+                                let mut lock = transcription_hybrid::SINGLETON_INSTANCE.lock().unwrap();
+                                if let Some(singleton) = lock.as_mut() {
+                                    singleton.start(stop_convert_rx_clone, false);
+                                }
                             } else {
                                 transcription::initialize_transcription(
                                     app_handle_clone,
@@ -334,6 +343,7 @@ impl RecordDesktop {
             translation_ja_high::drop_translation_ja_high();
             transcription_online::drop_transcription_online();
             transcription_amivoice::drop_transcription_amivoice();
+            transcription_hybrid::drop_transcription_hybrid();
             chat_online::drop_chat_online();
         } else {
             drop(stop_convert_tx)