Wire up browser source

Browser source can be started and stopped via the UI. It still serves a
hello world json blob.

Observing occasional crashes when stopping the C++ transcription engine.

6 files changed, 67 insertions, 29 deletions

diff --git a/GUI/GUI/GUI/BrowserSource.cpp b/GUI/GUI/GUI/BrowserSource.cpp
index 030b982..0b4ad31 100644
--- a/GUI/GUI/GUI/BrowserSource.cpp
+++ b/GUI/GUI/GUI/BrowserSource.cpp
@@ -1,15 +1,21 @@
 #include "BrowserSource.h"

 #include "Logging.h"

+#include "ScopeGuard.h"

 

 #include "oatpp/network/Server.hpp"

 

+using ::Logging::Log;

+

 BrowserSource::BrowserSource(uint16_t port, wxTextCtrl *out)

 	: port_(port), out_(out)

 {}

 

 void BrowserSource::Run(volatile bool* run)

 {

-	AppComponent components;

+	oatpp::base::Environment::init();

+	ScopeGuard oatpp_env_cleanup([]() { oatpp::base::Environment::destroy(); });

+

+	AppComponent components(port_);

 

 	OATPP_COMPONENT(std::shared_ptr<oatpp::web::server::HttpRouter>, router);

 	router->addController(std::make_shared<AppController>());

@@ -20,8 +26,8 @@ void BrowserSource::Run(volatile bool* run)
 

 	oatpp::network::Server server(connectionProvider, connectionHandler);

 

-	OATPP_LOGI("BrowserSource", "Server running on port %s",

-		connectionProvider->getProperty("port").getData());

+	Log(out_, "Server running on port {}\n",

+		static_cast<const char*>(connectionProvider->getProperty("port").getData()));

 

 	server.run(std::function<bool()>([run]() { return *run == true; }));

 }

diff --git a/GUI/GUI/GUI/BrowserSource.h b/GUI/GUI/GUI/BrowserSource.h
index 5b75c21..37d45a7 100644
--- a/GUI/GUI/GUI/BrowserSource.h
+++ b/GUI/GUI/GUI/BrowserSource.h
@@ -52,9 +52,14 @@ public:
 class AppComponent

 {

 public:

-    // TODO(yum) parameterize port

+    AppComponent(uint16_t port) : port_(port) {}

+

+    // TODO(yum) make port configurable. oatpp examples show how to use a port

+    // that's available at boot time. Plumbing this with port_ or port causes

+    // oatpp to use a different port every time, which I think is caused by port_

+    // being uninitialized.

 	OATPP_CREATE_COMPONENT(std::shared_ptr<oatpp::network::ServerConnectionProvider>, serverConnectionProvider)([] {

-        return oatpp::network::tcp::server::ConnectionProvider::createShared({ "0.0.0.0", 8000, oatpp::network::Address::IP_4 });

+        return oatpp::network::tcp::server::ConnectionProvider::createShared({ "0.0.0.0", 9517, oatpp::network::Address::IP_4 });

         }());

 

     OATPP_CREATE_COMPONENT(std::shared_ptr<oatpp::web::server::HttpRouter>, httpRouter)([] {

@@ -69,6 +74,9 @@ public:
     OATPP_CREATE_COMPONENT(std::shared_ptr<oatpp::data::mapping::ObjectMapper>, apiObjectMapper)([] {

         return oatpp::parser::json::mapping::ObjectMapper::createShared();

         }());

+

+private:

+    const uint16_t port_;

 };

 

 class BrowserSource

diff --git a/GUI/GUI/GUI/Config.cpp b/GUI/GUI/GUI/Config.cpp
index bcc7375..3e09a2e 100644
--- a/GUI/GUI/GUI/Config.cpp
+++ b/GUI/GUI/GUI/Config.cpp
@@ -92,7 +92,7 @@ AppConfig::AppConfig()
 	whisper_model("ggml-base.en.bin"),

 	whisper_mic(0),

 

-	browser_src_port(9010),

+	browser_src_port(9517),

 	whisper_enable_builtin(false),

 	whisper_enable_custom(true),

 	whisper_enable_browser_src(false)

diff --git a/GUI/GUI/GUI/Frame.cpp b/GUI/GUI/GUI/Frame.cpp
index ce3d832..ec7de2e 100644
--- a/GUI/GUI/GUI/Frame.cpp
+++ b/GUI/GUI/GUI/Frame.cpp
@@ -952,10 +952,12 @@ Frame::Frame()
                         "but are far more accurate.");

                     whisper_window_duration_ = whisper_window_duration;

 

+                    // TODO(yum) make this mutable once we figure out how to

+                    // get oatpp to accept a runtime src port.

                     auto* whisper_browser_src_port = new wxTextCtrl(

                         whisper_config_panel_pairs, ID_WHISPER_BROWSER_SRC_PORT,

                         std::to_string(app_c_.browser_src_port), wxDefaultPosition,

-                        wxDefaultSize, /*style=*/0);

+                        wxDefaultSize, wxTE_READONLY);

                     whisper_browser_src_port->SetToolTip(

                         "This is the port that the browser source is hosted "

                         "on. If you aren't using TaSTT to stream, you can "

@@ -1200,7 +1202,7 @@ Frame::Frame()
 		sizer->Add(whisper_panel, /*proportion=*/1, /*flags=*/wxEXPAND);

     }

 

-	Bind(wxEVT_MENU, &Frame::OnExit, this, wxID_EXIT);

+	Bind(wxEVT_CLOSE_WINDOW, &Frame::OnExit, this, wxID_EXIT);

 	Bind(wxEVT_BUTTON, &Frame::OnNavbarTranscribe, this,

         ID_NAVBAR_BUTTON_TRANSCRIBE);

 	Bind(wxEVT_BUTTON, &Frame::OnNavbarUnity, this, ID_NAVBAR_BUTTON_UNITY);

@@ -1350,10 +1352,10 @@ void Frame::PopulateDynamicInputFields()
     }

 }

 

-void Frame::OnExit(wxCommandEvent& event)

+void Frame::OnExit(wxCloseEvent& event)

 {

-    OnAppStop(event);

-    Close(true);

+    OnAppStop();

+    OnWhisperStop();

 }

 

 void Frame::OnNavbarTranscribe(wxCommandEvent& event)

@@ -1887,7 +1889,7 @@ void Frame::OnAppStart(wxCommandEvent& event) {
     py_app_ = p;

 }

 

-void Frame::OnAppStop(wxCommandEvent& event) {

+void Frame::OnAppStop() {

     if (py_app_) {

         const long pid = py_app_->GetPid();

 

@@ -1927,6 +1929,10 @@ void Frame::OnAppStop(wxCommandEvent& event) {
     }

 }

 

+void Frame::OnAppStop(wxCommandEvent& event) {

+    OnAppStop();

+}

+

 void Frame::OnWhisperStart(wxCommandEvent& event) {

 	Log(whisper_out_, "Launching transcription engine\n");

 

@@ -1956,7 +1962,6 @@ void Frame::OnWhisperStart(wxCommandEvent& event) {
     }

     const bool enable_local_beep = whisper_enable_local_beep_->GetValue();

     const bool use_cpu = whisper_use_cpu_->GetValue();

-    const bool use_builtin = whisper_enable_builtin_->GetValue();

     std::string rows_str = whisper_rows_->GetValue().ToStdString();

     std::string cols_str = whisper_cols_->GetValue().ToStdString();

     std::string chars_per_sync_str =

@@ -1965,13 +1970,16 @@ void Frame::OnWhisperStart(wxCommandEvent& event) {
         kBytesPerChar[bytes_per_char_idx].ToStdString();

     std::string window_duration_str =

         whisper_window_duration_->GetValue().ToStdString();

-    int rows, cols, chars_per_sync, bytes_per_char, window_duration;

+    std::string browser_src_port_str =

+        whisper_browser_src_port_->GetValue().ToStdString();

+    int rows, cols, chars_per_sync, bytes_per_char, window_duration, browser_src_port;

     try {

         rows = std::stoi(rows_str);

         cols = std::stoi(cols_str);

         chars_per_sync = std::stoi(chars_per_sync_str);

         bytes_per_char = std::stoi(bytes_per_char_str);

         window_duration = std::stoi(window_duration_str);

+        browser_src_port = std::stoi(browser_src_port_str);

     }

     catch (const std::invalid_argument&) {

 		Log(whisper_out_, "Could not parse rows \"{}\", cols \"{}\", chars "

@@ -2003,6 +2011,14 @@ void Frame::OnWhisperStart(wxCommandEvent& event) {
         return;

     }

 

+    const int min_port = 1024;

+    const int max_port = 65535;

+    if (browser_src_port < min_port || browser_src_port > max_port) {

+        Log(whisper_out_, "Browser source port not on [{},{}]\n",

+            min_port, max_port);

+        return;

+    }

+

     app_c_.whisper_mic = which_mic;

     app_c_.language = kLangChoices[which_lang].ToStdString();

     app_c_.whisper_model = kWhisperModelChoices[which_model].ToStdString();

@@ -2014,15 +2030,28 @@ void Frame::OnWhisperStart(wxCommandEvent& event) {
     app_c_.window_duration = std::to_string(window_duration);

     app_c_.enable_local_beep = enable_local_beep;

     app_c_.use_cpu = use_cpu;

-    app_c_.use_builtin = use_builtin;

+    app_c_.browser_src_port = browser_src_port;

+    app_c_.whisper_enable_browser_src = whisper_enable_browser_src_->GetValue();

+    app_c_.whisper_enable_builtin = whisper_enable_builtin_->GetValue();

+    app_c_.whisper_enable_custom = whisper_enable_custom_->GetValue();

     app_c_.Serialize(AppConfig::kConfigPath);

 

     whisper_->Start(app_c_);

-    Log(whisper_out_, "Control flow exit start button\n");

+    if (whisper_enable_browser_src_->GetValue()) {

+        Log(whisper_out_, "Frame launching browser src\n");

+        whisper_->StartBrowserSource(app_c_);

+    }

 }

 

-void Frame::OnWhisperStop(wxCommandEvent& event) {

+void Frame::OnWhisperStop() {

     whisper_->Stop();

+    if (whisper_enable_browser_src_->GetValue()) {

+        whisper_->StopBrowserSource();

+    }

+}

+

+void Frame::OnWhisperStop(wxCommandEvent& event) {

+    OnWhisperStop();

 }

 

 void Frame::OnAppDrain(wxTimerEvent& event) {

diff --git a/GUI/GUI/GUI/Frame.h b/GUI/GUI/GUI/Frame.h
index d40ba24..7da65a3 100644
--- a/GUI/GUI/GUI/Frame.h
+++ b/GUI/GUI/GUI/Frame.h
@@ -93,7 +93,7 @@ private:
     // Populate dynamically-generated input fields, such as microphone lists.

     void PopulateDynamicInputFields();

 

-    void OnExit(wxCommandEvent& event);

+    void OnExit(wxCloseEvent& event);

     void OnNavbarTranscribe(wxCommandEvent& event);

     void OnNavbarUnity(wxCommandEvent& event);

     void OnNavbarDebug(wxCommandEvent& event);

@@ -101,8 +101,10 @@ private:
     void OnSetupPython(wxCommandEvent& event);

     void OnDumpMics(wxCommandEvent& event);

     void OnAppStart(wxCommandEvent& event);

+    void OnAppStop();

     void OnAppStop(wxCommandEvent& event);

     void OnWhisperStart(wxCommandEvent& event);

+    void OnWhisperStop();

     void OnWhisperStop(wxCommandEvent& event);

     void OnAppDrain(wxTimerEvent& event);

     void OnGenerateFX(wxCommandEvent& event);

diff --git a/GUI/GUI/GUI/WhisperCPP.cpp b/GUI/GUI/GUI/WhisperCPP.cpp
index 3007834..b610941 100644
--- a/GUI/GUI/GUI/WhisperCPP.cpp
+++ b/GUI/GUI/GUI/WhisperCPP.cpp
@@ -240,7 +240,6 @@ void WhisperCPP::Start(const AppConfig& c) {
 	}

 

 	transcription_thd_ = std::async(std::launch::async, [&]() -> void {

-		Log(out_, "Transcription thread top\n");

 		run_transcription_ = true;

 

 		Whisper::iAudioCapture* mic_stream;

@@ -361,12 +360,6 @@ void WhisperCPP::Start(const AppConfig& c) {
 				Log(app->out_, "Exit transcription loop\n");

 				return S_FALSE;

 			}

-			// Sleeping here prevents the GUI from hanging.

-			// For some reason, printing is also required to prevent hanging.

-			static int i = 0;

-			if (++i % 20 == 0) {

-				Log(app->out_, "Spin {}\n", i);

-			}

 			return S_OK;

 		};

 		callbacks.pv = this;

@@ -394,20 +387,20 @@ void WhisperCPP::Stop() {
 

 void WhisperCPP::StartBrowserSource(const AppConfig& c) {

 	if (!browser_src_thd_.valid()) {

-		Log(out_, "Transcription engine already running\n");

+		Log(out_, "Browser source already running\n");

 		return;

 	}

 

 	browser_src_thd_ = std::async(std::launch::async, [&]() -> void {

-#if 0

+		run_browser_src_ = true;

 		BrowserSource src(c.browser_src_port, out_);

 		src.Run(&run_browser_src_);

-#endif

+		Log(out_, "Browser source thread exit\n");

 	});

 }

 

 void WhisperCPP::StopBrowserSource() {

-	Log(out_, "Stopping browser source engine...\n");

+	Log(out_, "Stopping browser source...\n");

 	run_browser_src_ = false;

 	browser_src_thd_.wait();

 	Log(out_, "Done!\n");