revert change that added support for aws language model, as this does not work on our earlier version of aws-sdk-cpp and recent versions have hugh performance issue on debian 12

Signed-off-by: Dave Horton <daveh@beachdognet.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_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;
 	}