feat/106: converting playht speed option to float instead of integer (#107)

* feat/106: converting playht speed option to float instead of integer

* feat/106: Using strtod

* feat/106: Using strtod

Signed-off-by: Rammohan Yadavalli <rammohan.yadavalli@kore.com>

---------

Signed-off-by: Rammohan Yadavalli <rammohan.yadavalli@kore.com>
Signed-off-by: Rammohan Yadavalli <rammohan.yadavalli@gmail.com>
Co-authored-by: Rammohan Yadavalli <rammohan.yadavalli@kore.com>

mod_deepgram_transcribe/dg_transcribe_glue.cpp

@@ -172,7 +172,7 @@ namespace {
     else if (customModel) oss << "&model=" << customModel;
 
     if (var = switch_channel_get_variable(channel, "DEEPGRAM_SPEECH_MODEL_VERSION")) {
-     oss <<  "&version";
+     oss <<  "&version=";
      oss <<  var;
     }
     oss <<  "&language=";

mod_deepgram_tts/deepgram_glue.cpp

@@ -742,7 +742,7 @@ extern "C" {
     std::string url;
     std::ostringstream url_stream;
     // always use sample_rate=8000 for support jambonz caching system.
-    url_stream << "https://api.deepgram.com/v1/speak?model=" << d->voice_name << "&encoding=linear16&sample_rate=8000";
+    url_stream << (d->endpoint != nullptr ? d->endpoint : "https://api.deepgram.com") << "/v1/speak?model=" << d->voice_name << "&encoding=linear16&sample_rate=8000";
     url = url_stream.str();
 
     /* create the JSON body */

mod_deepgram_tts/mod_deepgram_tts.c

@@ -19,10 +19,12 @@ static void cleardeepgram(deepgram_t* d, int freeAll) {
   if (d->connect_time_ms) free(d->connect_time_ms);
   if (d->final_response_time_ms) free(d->final_response_time_ms);
   if (d->cache_filename) free(d->cache_filename);
+  if (d->endpoint) free(d->endpoint);
   
 
   d->api_key = NULL;
   d->request_id = NULL;
+  d->endpoint = NULL;
 
   d->reported_model_name = NULL;
   d->reported_model_uuid = NULL;
@@ -121,6 +123,9 @@ static void d_text_param_tts(switch_speech_handle_t *sh, char *param, const char
   if (0 == strcmp(param, "api_key")) {
     if (d->api_key) free(d->api_key);
     d->api_key = strdup(val);
+  } else if (0 == strcmp(param, "endpoint")) {
+    if (d->endpoint) free(d->endpoint);
+    d->endpoint = strdup(val);
   } else if (0 == strcmp(param, "voice")) {
     if (d->voice_name) free(d->voice_name);
     d->voice_name = strdup(val);

mod_deepgram_tts/mod_deepgram_tts.h

@@ -7,6 +7,7 @@
 typedef struct deepgram_data {
   char *voice_name;
   char *api_key;
+  char *endpoint;
 
   /* result data */
   long response_code;

mod_dub/mod_dub.c

@@ -271,7 +271,7 @@ SWITCH_STANDARD_API(dub_function)
         }
         else {
           char* text = argv[3];
-          int gain = argc > 4 ? atoi(argv[4]) : 0;
+          int gain = argc > 5 ? atoi(argv[5]) : 0;
           switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "sayOnTrack %s gain %d\n", text, gain);
           status = dub_say_on_track(session, track, text, gain);
         }

mod_dub/tts_vendor_parser.cpp

@@ -378,7 +378,10 @@ switch_status_t playht_parse_text(const std::map<std::string, std::string>& para
     cJSON_AddStringToObject(jResult, "quality", quality.c_str());
   }
   if (!speed.empty()) {
-    cJSON_AddNumberToObject(jResult, "speed", atoi(speed.c_str()));
+      double val = strtod(speed.c_str(), NULL);
+      if (val != 0.0) {
+        cJSON_AddNumberToObject(jResult, "speed", val);
+      }
   }
   if (!seed.empty()) {
     cJSON_AddNumberToObject(jResult, "seed", atoi(seed.c_str()));

mod_jambonz_transcribe/audio_pipe.cpp

@@ -229,16 +229,6 @@ int AudioPipe::lws_callback(struct lws *wsi,
 }
 
 
-// static members
-static const lws_retry_bo_t retry = {
-    nullptr,   // retry_ms_table
-    0,         // retry_ms_table_count
-    0,         // conceal_count
-    UINT16_MAX,         // secs_since_valid_ping
-    UINT16_MAX,        // secs_since_valid_hangup
-    0          // jitter_percent
-};
-
 struct lws_context *AudioPipe::context = nullptr;
 std::thread AudioPipe::serviceThread;
 std::mutex AudioPipe::mutex_connects;
@@ -376,6 +366,28 @@ bool AudioPipe::lws_service_thread() {
     { NULL, NULL, 0, 0 }
   };
 
+  uint16_t secs_sinceq_valid_ping = UINT16_MAX;
+  uint16_t secs_since_valid_hangup = UINT16_MAX;
+
+  char* wsVar = std::getenv("WS_PING_INTERVAL");
+  if (wsVar != nullptr) {
+    secs_sinceq_valid_ping = std::atoi(wsVar);
+  }
+
+  wsVar = std::getenv("WS_NO_PONG_HANGUP_INTERVAL");
+  if (wsVar != nullptr) {
+    secs_since_valid_hangup = std::atoi(wsVar);
+  }
+
+  const lws_retry_bo_t retry = {
+    nullptr,                  // retry_ms_table
+    0,                        // retry_ms_table_count
+    0,                        // conceal_count
+    secs_sinceq_valid_ping,   // secs_sinceq_valid_ping
+    secs_since_valid_hangup,  // secs_since_valid_hangup
+    0                         // jitter_percent
+  };
+
   memset(&info, 0, sizeof info); 
   info.port = CONTEXT_PORT_NO_LISTEN; 
   info.options = LWS_SERVER_OPTION_DO_SSL_GLOBAL_INIT;

mod_jambonz_transcribe/jb_transcribe_glue.cpp

@@ -145,7 +145,7 @@ namespace {
     cJSON_AddStringToObject(json, "format", "raw");
     cJSON_AddStringToObject(json, "encoding", "LINEAR16");
     cJSON_AddBoolToObject(json, "interimResults", tech_pvt->interim);
-    cJSON_AddNumberToObject(json, "sampleRateHz", 8000);
+    cJSON_AddNumberToObject(json, "sampleRateHz", tech_pvt->sampling);
     if (var = switch_channel_get_variable(channel, "JAMBONZ_STT_OPTIONS")) {
       cJSON* jOptions = cJSON_Parse(var);
       if (jOptions)  {
@@ -353,7 +353,7 @@ extern "C" {
   }
 	
   switch_status_t jb_transcribe_session_init(switch_core_session_t *session, 
-    responseHandler_t responseHandler, uint32_t samples_per_second, uint32_t channels, 
+    responseHandler_t responseHandler, uint32_t samples_per_second, int desiredSampling, uint32_t channels, 
     char* lang, int interim, char* bugname, void **ppUserData)
   {    	
     int err;
@@ -365,7 +365,7 @@ extern "C" {
       return SWITCH_STATUS_FALSE;
     }
 
-    if (SWITCH_STATUS_SUCCESS != fork_data_init(tech_pvt, session, samples_per_second, 8000, channels, lang, interim, bugname, responseHandler)) {
+    if (SWITCH_STATUS_SUCCESS != fork_data_init(tech_pvt, session, samples_per_second, desiredSampling, channels, lang, interim, bugname, responseHandler)) {
       destroy_tech_pvt(tech_pvt);
       return SWITCH_STATUS_FALSE;
     }

mod_jambonz_transcribe/jb_transcribe_glue.h

@@ -6,7 +6,7 @@ int parse_ws_uri(switch_channel_t *channel, const char* szServerUri, char* host,
 switch_status_t jb_transcribe_init();
 switch_status_t jb_transcribe_cleanup();
 switch_status_t jb_transcribe_session_init(switch_core_session_t *session, responseHandler_t responseHandler, 
-		uint32_t samples_per_second, uint32_t channels, char* lang, int interim, char* bugname, void **ppUserData);
+		uint32_t samples_per_second, int desiredSampling, uint32_t channels, char* lang, int interim, char* bugname, void **ppUserData);
 switch_status_t jb_transcribe_session_stop(switch_core_session_t *session, int channelIsClosing, char* bugname);
 switch_bool_t jb_transcribe_frame(switch_core_session_t *session, switch_media_bug_t *bug);
 

mod_jambonz_transcribe/mod_jambonz_transcribe.c

@@ -73,6 +73,8 @@ static switch_status_t start_capture(switch_core_session_t *session, switch_medi
 	switch_codec_implementation_t read_impl = { 0 };
 	void *pUserData;
 	uint32_t samples_per_second;
+	uint32_t desiredSampling = 8000;
+	const char* var;
 
   if (!switch_channel_get_variable(channel, "JAMBONZ_STT_URL")) {
     switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Missing JAMBONZ_STT_URL channel var\n");
@@ -91,8 +93,11 @@ static switch_status_t start_capture(switch_core_session_t *session, switch_medi
 	}
 
 	samples_per_second = !strcasecmp(read_impl.iananame, "g722") ? read_impl.actual_samples_per_second : read_impl.samples_per_second;
-
+	var = switch_channel_get_variable(channel, "JAMBONZ_STT_SAMPLING");
-	if (SWITCH_STATUS_FALSE == jb_transcribe_session_init(session, responseHandler, samples_per_second, flags & SMBF_STEREO ? 2 : 1, lang, interim, bugname, &pUserData)) {
+	if (var != NULL) {
+		desiredSampling = atoi(var);
+	}
+	if (SWITCH_STATUS_FALSE == jb_transcribe_session_init(session, responseHandler, samples_per_second, desiredSampling, flags & SMBF_STEREO ? 2 : 1, lang, interim, bugname, &pUserData)) {
 		switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error initializing jb speech session.\n");
 		return SWITCH_STATUS_FALSE;
 	}

mod_playht_tts/playht_glue.cpp

@@ -774,7 +774,10 @@ extern "C" {
       cJSON_AddStringToObject(jResult, "quality", p->quality);
     }
     if (p->speed) {
-      cJSON_AddNumberToObject(jResult, "speed", atoi(p->speed));
+      double val = strtod(p->speed, NULL);
+      if (val != 0.0) {
+        cJSON_AddNumberToObject(jResult, "speed", val);
+      }
     }
     if (p->seed) {
       cJSON_AddNumberToObject(jResult, "seed", atoi(p->seed));