Fix playht timeout while downloading audio

Signed-off-by: Quan HL <quan.luuhoang8@gmail.com>

mod_aws_lex/aws_lex_glue.cpp

@@ -173,13 +173,12 @@ public:
 		char *metadata,
 		const char* awsAccessKeyId, 
 		const char* awsSecretAccessKey,
+		const char* awsSessionToken,
 		responseHandler_t responseHandler,
 		errorHandler_t  errorHandler) : 
 	m_bot(bot), m_alias(alias), m_region(region), m_sessionId(sessionId), m_finished(false), m_finishing(false), m_packets(0),
 	m_pStream(nullptr), m_bPlayDone(false), m_bDiscardAudio(false)
 	{
-		Aws::String key(awsAccessKeyId);
-		Aws::String secret(awsSecretAccessKey);
 		Aws::String awsLocale(locale);
 		Aws::Client::ClientConfiguration config;
 		config.region = region;
@@ -191,7 +190,10 @@ public:
 		keySnippet[19] = '\0';
 
 		switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "GStreamer %p ACCESS_KEY_ID %s\n", this, keySnippet);
-		if (*awsAccessKeyId && *awsSecretAccessKey) {
+		if (*awsAccessKeyId && *awsSecretAccessKey && *awsSessionToken) {
+			switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "using AWS creds %s %s %s\n", awsAccessKeyId, awsSecretAccessKey, awsSessionToken);	
+			m_client = Aws::MakeUnique<LexRuntimeV2Client>(ALLOC_TAG, AWSCredentials(awsAccessKeyId, awsSecretAccessKey, awsSessionToken), config);
+		} else if (*awsAccessKeyId && *awsSecretAccessKey) {
 			switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "using AWS creds %s %s\n", awsAccessKeyId, awsSecretAccessKey);	
 			m_client = Aws::MakeUnique<LexRuntimeV2Client>(ALLOC_TAG, AWSCredentials(awsAccessKeyId, awsSecretAccessKey), config);
 		}
@@ -540,7 +542,7 @@ static void *SWITCH_THREAD_FUNC lex_thread(switch_thread_t *thread, void *obj) {
 	bool ok = true;
 	switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "lex_thread: starting cb %p\n", (void *) cb);
 	GStreamer* pStreamer = new GStreamer(cb->sessionId, cb->bot, cb->alias, cb->region, cb->locale, 
-		cb->intent, cb->metadata, cb->awsAccessKeyId, cb->awsSecretAccessKey, 
+		cb->intent, cb->metadata, cb->awsAccessKeyId, cb->awsSecretAccessKey, cb->awsSessionToken,
 		cb->responseHandler, cb->errorHandler);
 	if (!pStreamer) {
 		switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "lex_thread: Error allocating streamer\n");
@@ -641,6 +643,7 @@ extern "C" {
 		memset(cb, sizeof(cb), 0);
 		const char* awsAccessKeyId = switch_channel_get_variable(channel, "AWS_ACCESS_KEY_ID");
 		const char* awsSecretAccessKey = switch_channel_get_variable(channel, "AWS_SECRET_ACCESS_KEY");
+		const char* awsSessionToken = switch_channel_get_variable(channel, "AWS_SESSION_TOKEN");
 
 		if (!hasDefaultCredentials && (!awsAccessKeyId || !awsSecretAccessKey)) {
 			switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_ERROR, 
@@ -654,6 +657,7 @@ extern "C" {
 		if (awsAccessKeyId && awsSecretAccessKey) {
 			strncpy(cb->awsAccessKeyId, awsAccessKeyId, 128);
 			strncpy(cb->awsSecretAccessKey, awsSecretAccessKey, 128);
+			if (awsSessionToken) strncpy(cb->awsSessionToken, awsSessionToken, 1024);
 		}
 		else {
 			strncpy(cb->awsAccessKeyId, std::getenv("AWS_ACCESS_KEY_ID"), 128);

mod_aws_lex/mod_aws_lex.h

@@ -30,6 +30,7 @@ struct cap_cb {
 	char sessionId[256];
   char awsAccessKeyId[128];
   char awsSecretAccessKey[128];
+	char awsSessionToken[1024];
   SpeexResamplerState *resampler;
 	void* streamer;
 	responseHandler_t responseHandler;

mod_aws_transcribe/aws_transcribe_glue.cpp

@@ -49,12 +49,11 @@ public:
 		const char* region, 
 		const char* awsAccessKeyId, 
 		const char* awsSecretAccessKey,
+		const char* awsSessionToken,
 		responseHandler_t responseHandler
   ) : m_sessionId(sessionId), m_bugname(bugname), m_finished(false), m_interim(interim), m_finishing(false), m_connected(false), m_connecting(false),
 	 		m_packets(0), m_responseHandler(responseHandler), m_pStream(nullptr), 
 			m_audioBuffer(320 * (samples_per_second == 8000 ? 1 : 2), 15) {
-		Aws::String key(awsAccessKeyId);
-		Aws::String secret(awsSecretAccessKey);
 		Aws::Client::ClientConfiguration config;
 		if (region != nullptr && strlen(region) > 0) config.region = region;
 		char keySnippet[20];
@@ -64,7 +63,9 @@ public:
 		keySnippet[19] = '\0';
 
 		switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "GStreamer %p ACCESS_KEY_ID %s, region %s\n", this, keySnippet, region);
-		if (*awsAccessKeyId && *awsSecretAccessKey) {
+		if (*awsAccessKeyId && *awsSecretAccessKey && *awsSessionToken) {
+			m_client = Aws::MakeUnique<TranscribeStreamingServiceClient>(ALLOC_TAG, AWSCredentials(awsAccessKeyId, awsSecretAccessKey, awsSessionToken), config);
+		} else if (*awsAccessKeyId && *awsSecretAccessKey) {
 			m_client = Aws::MakeUnique<TranscribeStreamingServiceClient>(ALLOC_TAG, AWSCredentials(awsAccessKeyId, awsSecretAccessKey), config);
 		}
 		else {
@@ -320,8 +321,8 @@ static void *SWITCH_THREAD_FUNC aws_transcribe_thread(switch_thread_t *thread, v
 	struct cap_cb *cb = (struct cap_cb *) obj;
 	bool ok = true;
 	switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "transcribe_thread: starting cb %p\n", (void *) cb);
-	GStreamer* pStreamer = new GStreamer(cb->sessionId, cb->bugname, cb->channels, cb->lang, cb->interim, cb->samples_per_second, cb->region, cb->awsAccessKeyId, cb->awsSecretAccessKey, 
+	GStreamer* pStreamer = new GStreamer(cb->sessionId, cb->bugname, cb->channels, cb->lang, cb->interim, cb->samples_per_second,
-		cb->responseHandler);
+		cb->region, cb->awsAccessKeyId, cb->awsSecretAccessKey, cb->awsSessionToken, cb->responseHandler);
 	if (!pStreamer) {
 		switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "transcribe_thread: Error allocating streamer\n");
 		return nullptr;
@@ -408,6 +409,7 @@ extern "C" {
 		memset(cb, sizeof(cb), 0);
 		const char* awsAccessKeyId = switch_channel_get_variable(channel, "AWS_ACCESS_KEY_ID");
 		const char* awsSecretAccessKey = switch_channel_get_variable(channel, "AWS_SECRET_ACCESS_KEY");
+		const char* awsSessionToken = switch_channel_get_variable(channel, "AWS_SESSION_TOKEN");
 		const char* awsRegion = switch_channel_get_variable(channel, "AWS_REGION");
 		cb->channels = channels;
 		LanguageCode code = LanguageCodeMapper::GetLanguageCodeForName(lang);
@@ -419,12 +421,12 @@ extern "C" {
 		strncpy(cb->sessionId, switch_core_session_get_uuid(session), MAX_SESSION_ID);
 		strncpy(cb->bugname, bugname, MAX_BUG_LEN);
 
-		if (awsAccessKeyId && awsSecretAccessKey && awsRegion) {
+    if (awsRegion) strncpy(cb->region, awsRegion, MAX_REGION);
+		if (awsAccessKeyId && awsSecretAccessKey) {
 			switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_DEBUG, "Using channel vars for aws authentication\n");
 			strncpy(cb->awsAccessKeyId, awsAccessKeyId, 128);
 			strncpy(cb->awsSecretAccessKey, awsSecretAccessKey, 128);
-			strncpy(cb->region, awsRegion, MAX_REGION);
+			if (awsSessionToken) strncpy(cb->awsSessionToken, awsSessionToken, 1024);
-
 		}
 		else if (std::getenv("AWS_ACCESS_KEY_ID") &&
 			std::getenv("AWS_SECRET_ACCESS_KEY") &&

mod_aws_transcribe/mod_aws_transcribe.h

@@ -28,6 +28,7 @@ struct cap_cb {
 	char sessionId[MAX_SESSION_ID+1];
   char awsAccessKeyId[128];
   char awsSecretAccessKey[128];
+	char awsSessionToken[1024];
 	uint32_t channels;
   SpeexResamplerState *resampler;
 	void* streamer;

mod_azure_tts/mod_azure_tts.c

@@ -82,6 +82,8 @@ static switch_status_t a_speech_feed_tts(switch_speech_handle_t *sh, char *text,
   a->draining = 0;
   a->reads = 0;
   a->flushed = 0;
+  a->response_code = 0;
+  a->err_msg = NULL;
 
   return azure_speech_feed_tts(a, text, flags);
 }

mod_deepgram_tts/mod_deepgram_tts.c

@@ -84,6 +84,8 @@ static switch_status_t d_speech_feed_tts(switch_speech_handle_t *sh, char *text,
   deepgram_t *d = createOrRetrievePrivateData(sh);
   d->draining = 0;
   d->reads = 0;
+  d->response_code = 0;
+  d->err_msg = NULL;
 
   switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "d_speech_feed_tts\n");
 

mod_elevenlabs_tts/mod_elevenlabs_tts.c

@@ -106,6 +106,8 @@ static switch_status_t ell_speech_feed_tts(switch_speech_handle_t *sh, char *tex
   elevenlabs_t *el = createOrRetrievePrivateData(sh);
   el->draining = 0;
   el->reads = 0;
+  el->response_code = 0;
+  el->err_msg = NULL;
   
   switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "ell_speech_feed_tts\n");
   

mod_jambonz_transcribe/jb_transcribe_glue.cpp

@@ -404,12 +404,14 @@ extern "C" {
         jambonz::AudioPipe *pAudioPipe = static_cast<jambonz::AudioPipe *>(tech_pvt->pAudioPipe);
         if (pAudioPipe) reaper(tech_pvt);
         destroy_tech_pvt(tech_pvt);
+        switch_mutex_unlock(tech_pvt->mutex);
         switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_DEBUG, "(%u) jb_transcribe_session_stop, bug removed\n", id);
         return SWITCH_STATUS_SUCCESS;
       }
       else {
         switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_DEBUG, "jb_transcribe_session_stop: race condition, previous close completed\n");
       }
+      switch_mutex_unlock(tech_pvt->mutex);
     }
     return SWITCH_STATUS_FALSE;
   }

mod_jambonz_transcribe/mod_jambonz_transcribe.c

@@ -100,7 +100,7 @@ static switch_status_t start_capture(switch_core_session_t *session, switch_medi
 		return status;
 	}
   switch_channel_set_private(channel, bugname, bug);
-	switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "added media bug for jb transcribe\n");
+	switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "added media bug for jb transcribe: %s\n", bugname);
 
 	return SWITCH_STATUS_SUCCESS;
 }
@@ -146,7 +146,7 @@ SWITCH_STANDARD_API(jb_transcribe_function)
 		if ((lsession = switch_core_session_locate(argv[0]))) {
 			if (!strcasecmp(argv[1], "stop")) {
 				char *bugname = argc > 2 ? argv[2] : MY_BUG_NAME;
-    		switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "stop transcribing\n");
+        switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "stop transcribing %s\n", bugname);
 				status = do_stop(lsession, bugname);
 			} else if (!strcasecmp(argv[1], "start")) {
         char* lang = argv[2];

mod_playht_tts/mod_playht_tts.c

@@ -99,6 +99,8 @@ static switch_status_t p_speech_feed_tts(switch_speech_handle_t *sh, char *text,
   playht_t *p = createOrRetrievePrivateData(sh);
   p->draining = 0;
   p->reads = 0;
+  p->response_code = 0;
+  p->err_msg = NULL;
 
   switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "p_speech_feed_tts\n");
 

mod_playht_tts/playht_glue.cpp

@@ -76,7 +76,8 @@ static CURL* createEasyHandle(void) {
 
   // set connect timeout to 3 seconds and total timeout to 109 seconds
   curl_easy_setopt(easy, CURLOPT_CONNECTTIMEOUT_MS, 3000L);
-  curl_easy_setopt(easy, CURLOPT_TIMEOUT, 10L);
+  //For long text, PlayHT took more than 20 seconds to complete the download.
+  curl_easy_setopt(easy, CURLOPT_TIMEOUT, 60L);
 
   return easy ;    
 }

mod_rimelabs_tts/mod_rimelabs_tts.c

@@ -82,6 +82,8 @@ static switch_status_t d_speech_feed_tts(switch_speech_handle_t *sh, char *text,
   rimelabs_t *d = createOrRetrievePrivateData(sh);
   d->draining = 0;
   d->reads = 0;
+  d->response_code = 0;
+  d->err_msg = NULL;
 
   switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "d_speech_feed_tts\n");
 

mod_whisper_tts/mod_whisper_tts.c

@@ -90,6 +90,8 @@ static switch_status_t w_speech_feed_tts(switch_speech_handle_t *sh, char *text,
   whisper_t *w = createOrRetrievePrivateData(sh);
   w->draining = 0;
   w->reads = 0;
+  w->response_code = 0;
+  w->err_msg = NULL;
 
   switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "w_speech_feed_tts\n");