change default base dir for streaming tts cache files to /tmp/

* fixes for unlocking mutex and minimizing time under lock

* call .get() on future returned from speechSynthesizer->SpeakTextAsync

* mod_azure_tts: various fixes, including dangling session lock and reducing latency

* mod_deepgram_transcribe: add support for transcribing filler words

mod_azure_tts/Makefile.am

@@ -7,4 +7,4 @@ mod_azure_tts_la_CFLAGS   = $(AM_CFLAGS)
 mod_azure_tts_la_CXXFLAGS = $(AM_CXXFLAGS) -std=c++14 -I/usr/local/include/MicrosoftSpeechSDK/cxx_api -I/usr/local/include/MicrosoftSpeechSDK/c_api
 
 mod_azure_tts_la_LIBADD   = $(switch_builddir)/libfreeswitch.la
-mod_azure_tts_la_LDFLAGS  = -avoid-version -module -no-undefined -L/usr/local/lib/MicrosoftSpeechSDK/x64 -lMicrosoft.CognitiveServices.Speech.core -shared `pkg-config --libs boost` -lstdc++
+mod_azure_tts_la_LDFLAGS  = -avoid-version -module -no-undefined -L/usr/local/lib/MicrosoftSpeechSDK/x64 -lMicrosoft.CognitiveServices.Speech.core -shared -lstdc++ -lboost_system -lboost_thread

mod_azure_tts/azure_glue.cpp

@@ -7,6 +7,7 @@
 #include <cstdlib>
 #include <string>
 #include <chrono>
+#include <thread>
 
 #define BUFFER_SIZE 8129
 
@@ -16,6 +17,34 @@ using namespace Microsoft::CognitiveServices::Speech;
 
 static std::string fullDirPath;
 
+static void start_synthesis(std::shared_ptr<SpeechSynthesizer> speechSynthesizer, const char* text, azure_t* a) {
+    try {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "start_synthesis calling \n");
+      auto result = std::strncmp(text, "<speak", 6) == 0 ?
+        speechSynthesizer->SpeakSsmlAsync(text).get() :
+        speechSynthesizer->SpeakTextAsync(text).get();
+
+      if (result->Reason == ResultReason::SynthesizingAudioCompleted) {
+        a->response_code = 200;
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "start_synthesis completed id %s, audio data - bytes: %ld, duration: %ldms\n", 
+          result->ResultId.c_str(), result->GetAudioLength(), result->AudioDuration.count());
+      } else if (result->Reason == ResultReason::Canceled) {
+        auto cancellation = SpeechSynthesisCancellationDetails::FromResult(result);
+        a->response_code = static_cast<long int>(cancellation->ErrorCode);
+        a->err_msg = strdup(cancellation->ErrorDetails.c_str());
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error synthesizing text %d with error string: %s.\n",
+          static_cast<int>(cancellation->ErrorCode), cancellation->ErrorDetails.c_str());
+      } else {
+        a->response_code = 500;
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error synthsize text %s (%d).\n", text, static_cast<int>(result->Reason));
+      }
+    } catch (const std::exception& e) {
+        a->response_code = 500;
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "mod_azure_tts: Exception in start_synthesis %s\n",  e.what());
+    }
+    a->draining = 1;
+}
+
 extern "C" {
   switch_status_t azure_speech_load() {
     switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "azure_speech_loading..\n");
@@ -23,7 +52,7 @@ extern "C" {
     /* create temp folder for cache files */
     const char* baseDir = std::getenv("JAMBONZ_TMP_CACHE_FOLDER");
     if (!baseDir) {
-      baseDir = "/var/";
+      baseDir = "/tmp/";
     }
     if (strcmp(baseDir, "/") == 0) {
       switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "failed to create folder %s\n", baseDir);
@@ -109,8 +138,12 @@ extern "C" {
     if (a->session_id) {
       int err;
       switch_codec_implementation_t read_impl;
+      
+      /* lock and unlock session */
       switch_core_session_t *psession = switch_core_session_locate(a->session_id);
       switch_core_session_get_read_impl(psession, &read_impl);
+      switch_core_session_rwunlock(psession);
+
       uint32_t samples_per_second = !strcasecmp(read_impl.iananame, "g722") ? read_impl.actual_samples_per_second : read_impl.samples_per_second;
       a->samples_rate = samples_per_second;
       if (samples_per_second != 8000 /*Hz*/) {
@@ -150,43 +183,32 @@ extern "C" {
 
     speechSynthesizer->SynthesisStarted += [a](const SpeechSynthesisEventArgs& e) {
         switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "azure_speech_feed_tts SynthesisStarted\n");
-        a->response_code = 200;
     };
 
     speechSynthesizer->Synthesizing += [a](const SpeechSynthesisEventArgs& e) {
+      if (a->flushed) return;
       bool fireEvent = false;
       CircularBuffer_t *cBuffer = (CircularBuffer_t *) a->circularBuffer;
-      std::vector<uint16_t> pcm_data;
 
-      //switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "Synthesizing: received data\n");
-
-      if (a->flushed) {
-        return;
+      auto audioData = e.Result->GetAudioData();
+      if (a->file) {
+        fwrite(audioData->data(), 1, audioData->size(), a->file);
       }
-      {
-        switch_mutex_lock(a->mutex);
-        auto audioData = e.Result->GetAudioData();
-        for (size_t i = 0; i < audioData->size(); i += sizeof(int16_t)) {
-            int16_t value = static_cast<int16_t>((*audioData)[i]) | (static_cast<int16_t>((*audioData)[i + 1]) << 8);
-            pcm_data.push_back(value);
-        }
 
-        /* and write to the file */
-        size_t bytesResampled = pcm_data.size() * sizeof(uint16_t);
-        if (a->file) fwrite(pcm_data.data(), sizeof(uint16_t), pcm_data.size(), a->file);
+      /**
+       * this sort of reinterpretation can be dangerous as a general rule, but in this case we know that the data
+       * is 16-bit PCM, so it's safe to do this and its much faster than copying the data byte by byte
+       */
+      const uint16_t* begin = reinterpret_cast<const uint16_t*>(audioData->data());
+      const uint16_t* end = reinterpret_cast<const uint16_t*>(audioData->data() + audioData->size());
 
-        // Resize the buffer if necessary
-        if (cBuffer->capacity() - cBuffer->size() < (bytesResampled / sizeof(uint16_t))) {
-
-          //TODO: if buffer exceeds some max size, return CURL_WRITEFUNC_ERROR to abort the transfer
-          cBuffer->set_capacity(cBuffer->size() + std::max((bytesResampled / sizeof(uint16_t)), (size_t)BUFFER_SIZE));
-        }
-
-        /* Push the data into the buffer */
-        cBuffer->insert(cBuffer->end(), pcm_data.data(), pcm_data.data() + pcm_data.size());
-
-        switch_mutex_unlock(a->mutex);
+      /* lock as briefly as possible */
+      switch_mutex_lock(a->mutex);
+      if (cBuffer->capacity() - cBuffer->size() < audioData->size()) {
+        cBuffer->set_capacity(cBuffer->size() + std::max( audioData->size(), (size_t)BUFFER_SIZE));
       }
+      cBuffer->insert(cBuffer->end(), begin, end);
+      switch_mutex_unlock(a->mutex);
 
       if (0 == a->reads++) {
         fireEvent = true;
@@ -200,6 +222,7 @@ extern "C" {
         switch_core_session_t* session = switch_core_session_locate(a->session_id);
         if (session) {
           switch_channel_t *channel = switch_core_session_get_channel(session);
+          switch_core_session_rwunlock(session);
           if (channel) {
             switch_event_t *event;
             if (switch_event_create(&event, SWITCH_EVENT_PLAYBACK_START) == SWITCH_STATUS_SUCCESS) {
@@ -216,31 +239,12 @@ extern "C" {
           }else {
             switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "speechSynthesizer->Synthesizing: channel not found\n");
           }
-          switch_core_session_rwunlock(session);
         }
       }
     };
 
-    speechSynthesizer->SynthesisCompleted += [a](const SpeechSynthesisEventArgs& e) {
-       switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "azure_speech_feed_tts SynthesisCompleted\n");
-       a->draining = 1;
-    };
-
-    speechSynthesizer->SynthesisCanceled += [a](const SpeechSynthesisEventArgs& e) {
-      if (e.Result->Reason == ResultReason::Canceled) {
-        auto cancellation = SpeechSynthesisCancellationDetails::FromResult(e.Result);
-        a->response_code = static_cast<long int>(cancellation->ErrorCode);
-        a->err_msg = strdup(cancellation->ErrorDetails.c_str());
-        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error synthsize tex %d with error string: %s.\n", static_cast<int>(cancellation->ErrorCode), cancellation->ErrorDetails.c_str());
-      }
-    };
-    // switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "azure_speech_feed_tts before sending synthesize request\n");
-    if (std::strncmp(text, "<speak", 6) == 0) {
-      speechSynthesizer->SpeakSsmlAsync(text);
-    } else {
-      speechSynthesizer->SpeakTextAsync(text);
-    }
-    // switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "azure_speech_feed_tts sent synthesize request\n");
+    std::thread(start_synthesis, speechSynthesizer, text, a).detach();
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "azure_speech_feed_tts sent synthesize request\n");
     return SWITCH_STATUS_SUCCESS;
   }
 
@@ -248,33 +252,31 @@ extern "C" {
     CircularBuffer_t *cBuffer = (CircularBuffer_t *) a->circularBuffer;
     std::vector<uint16_t> pcm_data;
 
-    {
-      switch_mutex_lock(a->mutex);
-      if (a->response_code > 0 && a->response_code != 200) {
-        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "azure_speech_read_tts, returning failure\n") ;
-        return SWITCH_STATUS_FALSE;
-      }
-      if (a->flushed) {
+    if (a->response_code > 0 && a->response_code != 200) {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "azure_speech_read_tts, returning failure\n") ;
+      return SWITCH_STATUS_FALSE;
+    }
+    if (a->flushed) {
+      return SWITCH_STATUS_BREAK;
+    }
+    switch_mutex_lock(a->mutex);
+    size_t bufSize = cBuffer->size();
+    if (cBuffer->empty()) {
+      switch_mutex_unlock(a->mutex);
+      if (a->draining) {
         return SWITCH_STATUS_BREAK;
       }
-      if (cBuffer->empty()) {
-        if (a->draining) {
-          switch_mutex_unlock(a->mutex);
-          return SWITCH_STATUS_BREAK;
-        }
-        /* no audio available yet so send silence */
-        memset(data, 255, *datalen);
-        switch_mutex_unlock(a->mutex);
-        return SWITCH_STATUS_SUCCESS;
-      }
-      // azure returned 8000hz 16 bit data, we have to take enough data based on call sample rate.
-      size_t size = a->samples_rate ?
-        std::min((*datalen/(2 * a->samples_rate / 8000)), cBuffer->size()) :
-        std::min((*datalen/2), cBuffer->size());
-      pcm_data.insert(pcm_data.end(), cBuffer->begin(), cBuffer->begin() + size);
-      cBuffer->erase(cBuffer->begin(), cBuffer->begin() + size);
-      switch_mutex_unlock(a->mutex);
+      /* no audio available yet so send silence */
+      memset(data, 255, *datalen);
+      return SWITCH_STATUS_SUCCESS;
     }
+    // azure returned 8000hz 16 bit data, we have to take enough data based on call sample rate.
+    size_t size = a->samples_rate ?
+      std::min((*datalen/(2 * a->samples_rate / 8000)), bufSize) :
+      std::min((*datalen/2), bufSize);
+    pcm_data.insert(pcm_data.end(), cBuffer->begin(), cBuffer->begin() + size);
+    cBuffer->erase(cBuffer->begin(), cBuffer->begin() + size);
+    switch_mutex_unlock(a->mutex);
 
     size_t data_size = pcm_data.size();
 
@@ -336,6 +338,9 @@ extern "C" {
       switch_core_session_t* session = switch_core_session_locate(a->session_id);
       if (session) {
         switch_channel_t *channel = switch_core_session_get_channel(session);
+
+        /* unlock as quickly as possible */
+        switch_core_session_rwunlock(session);
         if (channel) {
           switch_event_t *event;
           if (switch_event_create(&event, SWITCH_EVENT_PLAYBACK_STOP) == SWITCH_STATUS_SUCCESS) {
@@ -357,7 +362,6 @@ extern "C" {
         else {
           switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: channel not found\n");
         }
-        switch_core_session_rwunlock(session);
       }
     }
 

mod_deepgram_transcribe/README.md

@@ -25,10 +25,11 @@ Stop transcription on the channel.
 | variable | Description |
 | --- | ----------- |
 | DEEPGRAM_API_KEY | Deepgram API key used to authenticate |
-| DEEPGRAM_SPEECH_TIER | https://developers.deepgram.com/documentation/features/tier/ |
 | DEEPGRAM_SPEECH_CUSTOM_MODEL | custom model id |
 | DEEPGRAM_SPEECH_MODEL | https://developers.deepgram.com/documentation/features/model/ |
 | DEEPGRAM_SPEECH_MODEL_VERSION | https://developers.deepgram.com/documentation/features/version/ |
+| DEEPGRAM_SPEECH_ENABLE_SMART_FORMAT | https://developers.deepgram.com/docs/smart-format |
+| DEEPGRAM_SPEECH_ENABLE_FILLER_WORDS | https://developers.deepgram.com/docs/filler-words |
 | DEEPGRAM_SPEECH_ENABLE_AUTOMATIC_PUNCTUATION | https://developers.deepgram.com/documentation/features/punctuate/ |
 | DEEPGRAM_SPEECH_PROFANITY_FILTER | https://developers.deepgram.com/documentation/features/profanity-filter/ |
 | DEEPGRAM_SPEECH_REDACT | https://developers.deepgram.com/documentation/features/redact/ |
@@ -42,6 +43,7 @@ Stop transcription on the channel.
 | DEEPGRAM_SPEECH_REPLACE | https://developers.deepgram.com/documentation/features/replace/  |
 | DEEPGRAM_SPEECH_TAG | https://developers.deepgram.com/documentation/features/tag/ |
 | DEEPGRAM_SPEECH_ENDPOINTING  | https://developers.deepgram.com/documentation/features/endpointing/ |
+| DEEPGRAM_SPEECH_UTTERANCE_END_MS | https://developers.deepgram.com/docs/utterance-end |
 | DEEPGRAM_SPEECH_VAD_TURNOFF | https://developers.deepgram.com/documentation/features/voice-activity-detection/ |
 
 

mod_deepgram_transcribe/dg_transcribe_glue.cpp

@@ -184,6 +184,9 @@ namespace {
       * 
       */
     }
+    if (var = switch_channel_get_variable(channel, "DEEPGRAM_SPEECH_ENABLE_FILLER_WORDS")) {
+     oss <<  "&filler_words=true";
+    }
     if (var = switch_channel_get_variable(channel, "DEEPGRAM_SPEECH_ENABLE_AUTOMATIC_PUNCTUATION")) {
      oss <<  "&punctuate=true";
     }

mod_deepgram_tts/Makefile.am

@@ -5,4 +5,4 @@ mod_LTLIBRARIES = mod_deepgram_tts.la
 mod_deepgram_tts_la_SOURCES  = mod_deepgram_tts.c deepgram_glue.cpp
 mod_deepgram_tts_la_CFLAGS   = $(AM_CFLAGS)
 mod_deepgram_tts_la_LIBADD   = $(switch_builddir)/libfreeswitch.la
-mod_deepgram_tts_la_LDFLAGS  = -avoid-version -module -no-undefined -shared `pkg-config --libs boost` -lstdc++
+mod_deepgram_tts_la_LDFLAGS  = -avoid-version -module -no-undefined -shared -lstdc++ -lboost_system -lboost_thread

mod_deepgram_tts/deepgram_glue.cpp

@@ -461,6 +461,7 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
     switch_core_session_t* session = switch_core_session_locate(d->session_id);
     if (session) {
       switch_channel_t *channel = switch_core_session_get_channel(session);
+      switch_core_session_rwunlock(session);
       if (channel) {
         switch_event_t *event;
         if (switch_event_create(&event, SWITCH_EVENT_PLAYBACK_START) == SWITCH_STATUS_SUCCESS) {
@@ -501,7 +502,6 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
       else {
         switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: channel not found\n");
       }
-      switch_core_session_rwunlock(session);
     }
     else {
       switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: session %s not found\n", d->session_id);
@@ -607,7 +607,7 @@ extern "C" {
     /* create temp folder for cache files */
     const char* baseDir = std::getenv("JAMBONZ_TMP_CACHE_FOLDER");
     if (!baseDir) {
-      baseDir = "/var/";
+      baseDir = "/tmp/";
     }
     if (strcmp(baseDir, "/") == 0) {
       switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "failed to create folder %s\n", baseDir);
@@ -888,6 +888,7 @@ extern "C" {
       switch_core_session_t* session = switch_core_session_locate(d->session_id);
       if (session) {
         switch_channel_t *channel = switch_core_session_get_channel(session);
+        switch_core_session_rwunlock(session);
         if (channel) {
           switch_event_t *event;
           if (switch_event_create(&event, SWITCH_EVENT_PLAYBACK_STOP) == SWITCH_STATUS_SUCCESS) {
@@ -909,7 +910,6 @@ extern "C" {
         else {
           switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: channel not found\n");
         }
-        switch_core_session_rwunlock(session);
       }
     }
     return SWITCH_STATUS_SUCCESS;

mod_elevenlabs_tts/elevenlabs_glue.cpp

@@ -662,7 +662,7 @@ extern "C" {
     /* create temp folder for cache files */
     const char* baseDir = std::getenv("JAMBONZ_TMP_CACHE_FOLDER");
     if (!baseDir) {
-      baseDir = "/var/";
+      baseDir = "/tmp/";
     }
     if (strcmp(baseDir, "/") == 0) {
       switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "failed to create folder %s\n", baseDir);

mod_whisper_tts/Makefile.am

@@ -5,4 +5,4 @@ mod_LTLIBRARIES = mod_whisper_tts.la
 mod_whisper_tts_la_SOURCES  = mod_whisper_tts.c whisper_glue.cpp
 mod_whisper_tts_la_CFLAGS   = $(AM_CFLAGS)
 mod_whisper_tts_la_LIBADD   = $(switch_builddir)/libfreeswitch.la
-mod_whisper_tts_la_LDFLAGS  = -avoid-version -module -no-undefined -shared `pkg-config --libs boost` -lstdc++ -lmpg123
+mod_whisper_tts_la_LDFLAGS  = -avoid-version -module -no-undefined -shared -lstdc++ -lboost_system -lboost_thread -lmpg123

mod_whisper_tts/mod_whisper_tts.c

@@ -6,7 +6,7 @@ SWITCH_MODULE_SHUTDOWN_FUNCTION(mod_whisper_tts_shutdown);
 SWITCH_MODULE_DEFINITION(mod_whisper_tts, mod_whisper_tts_load, mod_whisper_tts_shutdown, NULL);
 
 static void clearWhisper(whisper_t* w, int freeAll) {
-  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "clearWhisper\n");
+  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "clearWhisper\n");
   if (w->api_key) free(w->api_key);
   if (w->model_id) free(w->model_id);
   if (w->speed) free(w->speed);
@@ -65,7 +65,7 @@ static switch_status_t w_speech_close(switch_speech_handle_t *sh, switch_speech_
 {
   switch_status_t rc;
   whisper_t *w = createOrRetrievePrivateData(sh);
-  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "w_speech_close\n");
+  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "w_speech_close\n");
 
   switch_mutex_destroy(w->mutex);
 
@@ -83,7 +83,7 @@ static switch_status_t w_speech_feed_tts(switch_speech_handle_t *sh, char *text,
   w->draining = 0;
   w->reads = 0;
 
-  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "w_speech_feed_tts\n");
+  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "w_speech_feed_tts\n");
 
   return whisper_speech_feed_tts(w, text, flags);
 }
@@ -94,7 +94,6 @@ static switch_status_t w_speech_feed_tts(switch_speech_handle_t *sh, char *text,
 static switch_status_t w_speech_read_tts(switch_speech_handle_t *sh, void *data, size_t *datalen, switch_speech_flag_t *flags)
 {
   whisper_t *w = createOrRetrievePrivateData(sh);
-  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "w_speech_read_tts\n");
   return whisper_speech_read_tts(w, data, datalen, flags);
 }
 
@@ -104,7 +103,7 @@ static switch_status_t w_speech_read_tts(switch_speech_handle_t *sh, void *data,
 static void w_speech_flush_tts(switch_speech_handle_t *sh)
 {
   whisper_t *w = createOrRetrievePrivateData(sh);
-  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "w_speech_flush_tts\n");
+  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "w_speech_flush_tts\n");
   whisper_speech_flush_tts(w);
 
   clearWhisper(w, 0);
@@ -113,7 +112,7 @@ static void w_speech_flush_tts(switch_speech_handle_t *sh)
 static void w_text_param_tts(switch_speech_handle_t *sh, char *param, const char *val)
 {
   whisper_t *w = createOrRetrievePrivateData(sh);
-  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "w_text_param_tts: %s=%s\n", param, val);
+  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "w_text_param_tts: %s=%s\n", param, val);
   if (0 == strcmp(param, "api_key")) {
     if (w->api_key) free(w->api_key);
     w->api_key = strdup(val);

mod_whisper_tts/whisper_glue.cpp

@@ -442,8 +442,6 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
   }
   {
     switch_mutex_lock(w->mutex);
-    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "write_cb: received data, response %ld\n", 
-      w->response_code);
 
     if (w->response_code > 0 && w->response_code != 200) {
       std::string body((char *) ptr, bytes_received);
@@ -482,6 +480,7 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
     switch_core_session_t* session = switch_core_session_locate(w->session_id);
     if (session) {
       switch_channel_t *channel = switch_core_session_get_channel(session);
+      switch_core_session_rwunlock(session);
       if (channel) {
         switch_event_t *event;
         if (switch_event_create(&event, SWITCH_EVENT_PLAYBACK_START) == SWITCH_STATUS_SUCCESS) {
@@ -525,7 +524,6 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
       else {
         switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: channel not found\n");
       }
-      switch_core_session_rwunlock(session);
     }
     else {
       switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: session %s not found\n", w->session_id);
@@ -632,7 +630,7 @@ extern "C" {
     /* create temp folder for cache files */
     const char* baseDir = std::getenv("JAMBONZ_TMP_CACHE_FOLDER");
     if (!baseDir) {
-      baseDir = "/var/";
+      baseDir = "/tmp/";
     }
     if (strcmp(baseDir, "/") == 0) {
       switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "failed to create folder %s\n", baseDir);
@@ -810,6 +808,7 @@ extern "C" {
       switch_codec_implementation_t read_impl;
       switch_core_session_t *psession = switch_core_session_locate(w->session_id);
       switch_core_session_get_read_impl(psession, &read_impl);
+      switch_core_session_rwunlock(psession);
       uint32_t samples_per_second = !strcasecmp(read_impl.iananame, "g722") ? read_impl.actual_samples_per_second : read_impl.samples_per_second;
       if (mpg123_param(mh, MPG123_FORCE_RATE, samples_per_second /*Hz*/, 0) != MPG123_OK) {
         switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error mpg123_param!\n");
@@ -897,8 +896,7 @@ extern "C" {
 
   switch_status_t whisper_speech_flush_tts(whisper_t* w) {
     bool download_complete = w->response_code == 200;
-    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "whisper_speech_flush_tts, download complete? %s\n", download_complete ? "yes" : "no") ;  
-
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "whisper_speech_flush_tts, download complete? %s\n", download_complete ? "yes" : "no") ;  
     ConnInfo_t *conn = (ConnInfo_t *) w->conn;
     CircularBuffer_t *cBuffer = (CircularBuffer_t *) w->circularBuffer;
     delete cBuffer;
@@ -930,6 +928,7 @@ extern "C" {
       switch_core_session_t* session = switch_core_session_locate(w->session_id);
       if (session) {
         switch_channel_t *channel = switch_core_session_get_channel(session);
+        switch_core_session_rwunlock(session);
         if (channel) {
           switch_event_t *event;
           if (switch_event_create(&event, SWITCH_EVENT_PLAYBACK_STOP) == SWITCH_STATUS_SUCCESS) {
@@ -951,14 +950,13 @@ extern "C" {
         else {
           switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: channel not found\n");
         }
-        switch_core_session_rwunlock(session);
       }
     }
     return SWITCH_STATUS_SUCCESS;
   }
 
 	switch_status_t whisper_speech_close(whisper_t* w) {
-    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "whisper_speech_close\n") ;
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "whisper_speech_close\n") ;
 		return SWITCH_STATUS_SUCCESS;
 	}
 }