Merge pull request #60 from solaoi/main

add online support

src-tauri/Cargo.toml

@@ -28,8 +28,8 @@ hound = "3.4.0"
 whisper-rs = "0.4.0"
 rusqlite = "*"
 samplerate-rs = "0.1.0"
-# download model
-reqwest = { version = "0.11", features = ["stream"] }
+# download model and openai request
+reqwest = { version = "0.11", features = ["stream", "multipart", "json"] }
 tokio = { version = "1", features = ["full"] }
 futures-util = "0.3.26"
 # serve audio

src-tauri/migrations/001.sql

@@ -22,6 +22,8 @@ CREATE TABLE settings (
 );
 INSERT INTO settings(setting_name, setting_status) VALUES("speakerLanguage", NULL);
 INSERT INTO settings(setting_name, setting_status) VALUES("transcriptionAccuracy", "off");
+INSERT INTO settings(setting_name, setting_status) VALUES("settingKey", "");
+INSERT INTO settings(setting_name, setting_status) VALUES("settingLanguage", "日本語");
 CREATE TABLE models (
     id INTEGER PRIMARY KEY AUTOINCREMENT,
     model_name TEXT,

src-tauri/src/main.rs

@@ -73,7 +73,6 @@ fn start_command(
     let mut lock = state.0.lock().unwrap();
     let (stop_record_tx, stop_record_rx) = unbounded();
     *lock = Some(stop_record_tx);
-
     std::thread::spawn(move || {
         let record = module::record::Record::new(window.app_handle().clone());
         record.start(
@@ -107,13 +106,23 @@ fn start_trace_command(
     *lock = Some(stop_convert_tx);
 
     std::thread::spawn(move || {
-        let mut transcription = module::transcription::Transcription::new(
-            window.app_handle(),
-            transcription_accuracy,
-            speaker_language,
-            note_id,
-        );
-        transcription.start(stop_convert_rx, true);
+        if transcription_accuracy.starts_with("online") {
+            let mut transcription_online = module::transcription_online::TranscriptionOnline::new(
+                window.app_handle(),
+                transcription_accuracy,
+                speaker_language,
+                note_id,
+            );
+            transcription_online.start(stop_convert_rx, true);
+        } else {
+            let mut transcription = module::transcription::Transcription::new(
+                window.app_handle(),
+                transcription_accuracy,
+                speaker_language,
+                note_id,
+            );
+            transcription.start(stop_convert_rx, true);
+        }
     });
 }
 
@@ -124,10 +133,7 @@ fn stop_trace_command(state: State<'_, RecordState>, window: tauri::Window) {
         stop_convert_tx.send(()).unwrap_or_else(|_| {
             window
                 .app_handle()
-                .emit_all(
-                    "traceCompletion",
-                    TraceCompletion {},
-                )
+                .emit_all("traceCompletion", TraceCompletion {})
                 .unwrap();
         })
     }

src-tauri/src/module/mod.rs

@@ -8,4 +8,5 @@ pub mod record;
 mod sqlite;
 mod transcriber;
 pub mod transcription;
+pub mod transcription_online;
 mod writer;

src-tauri/src/module/record.rs

@@ -23,7 +23,8 @@ use crossbeam_channel::{unbounded, Receiver};
 use tauri::{api::path::data_dir, AppHandle, Manager};
 
 use super::{
-    recognizer::MyRecognizer, sqlite::Sqlite, transcription::Transcription, writer::Writer,
+    recognizer::MyRecognizer, sqlite::Sqlite, transcription::Transcription,
+    transcription_online::TranscriptionOnline, writer::Writer,
 };
 
 pub struct Record {
@@ -140,7 +141,6 @@ impl Record {
         let is_converting = Arc::new(Mutex::new(false));
         let (stop_convert_tx, stop_convert_rx) = unbounded();
         let is_no_transcription = transcription_accuracy == "off";
-
         let app_handle = self.app_handle.clone();
         thread::spawn(move || loop {
             match notify_decoding_state_is_finalized_rx.try_recv() {
@@ -177,7 +177,6 @@ impl Record {
                         .lock()
                         .unwrap()
                         .replace(Writer::build(&audio_path.to_str().expect("error"), spec));
-
                     if !is_no_transcription && !*is_converting.lock().unwrap() {
                         let is_converting_clone = Arc::clone(&is_converting);
                         let app_handle_clone = app_handle.clone();
@@ -188,13 +187,24 @@ impl Record {
                             let mut lock = is_converting_clone.lock().unwrap();
                             *lock = true;
                             drop(lock);
-                            let mut transcription = Transcription::new(
-                                app_handle_clone,
-                                transcription_accuracy_clone,
-                                speaker_language_clone,
-                                note_id,
-                            );
-                            transcription.start(stop_convert_rx_clone, false);
+                            if transcription_accuracy_clone.starts_with("online") {
+                                let mut transcription_online = TranscriptionOnline::new(
+                                    app_handle_clone,
+                                    transcription_accuracy_clone,
+                                    speaker_language_clone,
+                                    note_id,
+                                );
+                                transcription_online.start(stop_convert_rx_clone, false);
+                            } else {
+                                let mut transcription = Transcription::new(
+                                    app_handle_clone,
+                                    transcription_accuracy_clone,
+                                    speaker_language_clone,
+                                    note_id,
+                                );
+                                transcription.start(stop_convert_rx_clone, false);
+                            }
+
                             let mut lock = is_converting_clone.lock().unwrap();
                             *lock = false;
                             drop(lock);

src-tauri/src/module/sqlite.rs

@@ -73,6 +73,14 @@ impl Sqlite {
             });
     }
 
+    pub fn select_whisper_token(&self) -> Result<String, rusqlite::Error> {
+        return self.conn.query_row(
+            "SELECT setting_status FROM settings WHERE setting_name = \"settingKey\"",
+            params![],
+            |row| Ok(row.get_unwrap(0)),
+        );
+    }
+
     pub fn update_model_vosk_to_whisper(
         &self,
         id: u16,

src-tauri/src/module/transcriber.rs

@@ -29,44 +29,47 @@ impl Transcriber {
         speaker_language: String,
         transcription_accuracy: String,
     ) -> FullParams<'static, 'static> {
-        let mut language = "ja";
-        if speaker_language.starts_with("en-us") || speaker_language.starts_with("small-en-us") {
-            language = "en";
+        let language = if speaker_language.starts_with("en-us")
+            || speaker_language.starts_with("small-en-us")
+        {
+            "en"
         } else if speaker_language.starts_with("cn") || speaker_language.starts_with("small-cn") {
-            language = "zh";
+            "zh"
         } else if speaker_language.starts_with("small-ko") {
-            language = "ko";
+            "ko"
         } else if speaker_language.starts_with("fr") || speaker_language.starts_with("small-fr") {
-            language = "fr";
+            "fr"
         } else if speaker_language.starts_with("de") || speaker_language.starts_with("small-de") {
-            language = "de";
+            "de"
         } else if speaker_language.starts_with("ru") || speaker_language.starts_with("small-ru") {
-            language = "ru";
+            "ru"
         } else if speaker_language.starts_with("es") || speaker_language.starts_with("small-es") {
-            language = "es";
+            "es"
         } else if speaker_language.starts_with("small-pt") {
-            language = "pt";
+            "pt"
         } else if speaker_language.starts_with("small-tr") {
-            language = "tr";
+            "tr"
         } else if speaker_language.starts_with("vn") || speaker_language.starts_with("small-vn") {
-            language = "vi";
+            "vi"
         } else if speaker_language.starts_with("it") || speaker_language.starts_with("small-it") {
-            language = "it";
+            "it"
         } else if speaker_language.starts_with("small-nl") {
-            language = "nl";
+            "nl"
         } else if speaker_language.starts_with("small-ca") {
-            language = "ca";
+            "ca"
         } else if speaker_language.starts_with("uk") || speaker_language.starts_with("small-uk") {
-            language = "uk";
+            "uk"
         } else if speaker_language.starts_with("small-sv") {
-            language = "sv";
+            "sv"
         } else if speaker_language.starts_with("hi") || speaker_language.starts_with("small-hi") {
-            language = "hi";
+            "hi"
         } else if speaker_language.starts_with("small-cs") {
-            language = "cs";
+            "cs"
         } else if speaker_language.starts_with("small-pl") {
-            language = "pl";
-        }
+            "pl"
+        } else {
+            "ja"
+        };
         let mut params = FullParams::new(SamplingStrategy::BeamSearch {
             beam_size: 5,
             patience: 1.0,