wip

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

mod_audio_fork/audio_pipe.cpp

@@ -480,7 +480,7 @@ AudioPipe::AudioPipe(const char* uuid, const char* host, unsigned int port, cons
 }
 AudioPipe::~AudioPipe() {
   if (m_audio_buffer) delete [] m_audio_buffer;
-  if (m_recv_buf) delete [] m_recv_buf;
+  if (m_recv_buf) free(m_recv_buf);
 }
 
 void AudioPipe::connect(void) {

mod_azure_tts/azure_glue.cpp

@@ -59,7 +59,7 @@ extern "C" {
       return SWITCH_STATUS_FALSE;
     }
 
-    fullDirPath = std::string(baseDir) + "jambonz-tts-cache-files";
+    fullDirPath = std::string(baseDir) + "tts-cache-files";
 
     // Create the directory with read, write, and execute permissions for everyone
     mode_t oldMask = umask(0);
@@ -135,23 +135,12 @@ extern "C" {
       return SWITCH_STATUS_FALSE;
     }
 
-    if (a->session_id) {
+    if (a->rate != 8000 /*Hz*/) {
       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*/) {
-        a->resampler = speex_resampler_init(1, 8000, samples_per_second, SWITCH_RESAMPLE_QUALITY, &err);
-        if (0 != err) {
-          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error initializing resampler: %s.\n", speex_resampler_strerror(err));
-          return SWITCH_STATUS_FALSE;
-        }
+      a->resampler = speex_resampler_init(1, 8000, a->rate, SWITCH_RESAMPLE_QUALITY, &err);
+      if (0 != err) {
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error initializing resampler: %s.\n", speex_resampler_strerror(err));
+        return SWITCH_STATUS_FALSE;
       }
     }
 
@@ -179,72 +168,77 @@ extern "C" {
 			speechConfig->SetEndpointId(a->endpointId);
 		}
 
-    auto speechSynthesizer = SpeechSynthesizer::FromConfig(speechConfig);
+    try {
+      auto speechSynthesizer = SpeechSynthesizer::FromConfig(speechConfig);
 
-    speechSynthesizer->SynthesisStarted += [a](const SpeechSynthesisEventArgs& e) {
-        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "azure_speech_feed_tts SynthesisStarted\n");
-    };
+      speechSynthesizer->SynthesisStarted += [a](const SpeechSynthesisEventArgs& e) {
+          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "azure_speech_feed_tts SynthesisStarted\n");
+      };
 
-    speechSynthesizer->Synthesizing += [a](const SpeechSynthesisEventArgs& e) {
-      if (a->flushed) return;
-      bool fireEvent = false;
-      CircularBuffer_t *cBuffer = (CircularBuffer_t *) a->circularBuffer;
+      speechSynthesizer->Synthesizing += [a](const SpeechSynthesisEventArgs& e) {
+        if (a->flushed) return;
+        bool fireEvent = false;
+        CircularBuffer_t *cBuffer = (CircularBuffer_t *) a->circularBuffer;
 
-      auto audioData = e.Result->GetAudioData();
-      if (a->file) {
-        fwrite(audioData->data(), 1, audioData->size(), a->file);
-      }
+        auto audioData = e.Result->GetAudioData();
+        if (a->file) {
+          fwrite(audioData->data(), 1, audioData->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());
+        /**
+         * 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());
 
-      /* 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);
+        /* 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;
-      }
+        if (0 == a->reads++) {
+          fireEvent = true;
+        }
 
-      if (fireEvent && a->session_id) {
-        auto endTime = std::chrono::high_resolution_clock::now();
-        auto startTime = *static_cast<std::chrono::time_point<std::chrono::high_resolution_clock>*>(a->startTime);
-        auto duration = std::chrono::duration_cast<std::chrono::milliseconds>(endTime - startTime);
-        auto time_to_first_byte_ms = std::to_string(duration.count());
-        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) {
-              switch_channel_event_set_data(channel, event);
-              switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "Playback-File-Type", "tts_stream");
-              switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_time_to_first_byte_ms", time_to_first_byte_ms.c_str());
-              if (a->cache_filename) {
-                switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_cache_filename", a->cache_filename);
+        if (fireEvent && a->session_id) {
+          auto endTime = std::chrono::high_resolution_clock::now();
+          auto startTime = *static_cast<std::chrono::time_point<std::chrono::high_resolution_clock>*>(a->startTime);
+          auto duration = std::chrono::duration_cast<std::chrono::milliseconds>(endTime - startTime);
+          auto time_to_first_byte_ms = std::to_string(duration.count());
+          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) {
+                switch_channel_event_set_data(channel, event);
+                switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "Playback-File-Type", "tts_stream");
+                switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_time_to_first_byte_ms", time_to_first_byte_ms.c_str());
+                if (a->cache_filename) {
+                  switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_cache_filename", a->cache_filename);
+                }
+                switch_event_fire(&event);
+              } else {
+                switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "speechSynthesizer->Synthesizing: failed to create event\n");
               }
-              switch_event_fire(&event);
-            } else {
-              switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "speechSynthesizer->Synthesizing: failed to create event\n");
+            }else {
+              switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "speechSynthesizer->Synthesizing: channel not found\n");
             }
-          }else {
-            switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "speechSynthesizer->Synthesizing: channel not found\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");
+      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");
+    } catch (const std::exception& e) {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "mod_azure_tts: Exception: %s\n", e.what());
+      return SWITCH_STATUS_FALSE;
+    }
     return SWITCH_STATUS_SUCCESS;
   }
 
@@ -271,9 +265,7 @@ extern "C" {
       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);
+    size_t size = std::min((*datalen/(2 * a->rate / 8000)), bufSize);
     pcm_data.insert(pcm_data.end(), cBuffer->begin(), cBuffer->begin() + size);
     cBuffer->erase(cBuffer->begin(), cBuffer->begin() + size);
     switch_mutex_unlock(a->mutex);

mod_azure_tts/mod_azure_tts.c

@@ -82,7 +82,6 @@ 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->samples_rate = 0;
 
   return azure_speech_feed_tts(a, text, flags);
 }

mod_azure_tts/mod_azure_tts.h

@@ -25,7 +25,6 @@ typedef struct azure_data {
   int reads;
   int cache_audio;
   int flushed;
-  uint32_t samples_rate;
 
   void *startTime;
 

mod_custom_tts/Makefile.am

@@ -0,0 +1,8 @@
+include $(top_srcdir)/build/modmake.rulesam
+MODNAME=mod_custom_tts
+
+mod_LTLIBRARIES = mod_custom_tts.la
+mod_custom_tts_la_SOURCES  = mod_custom_tts.c custom_glue.cpp
+mod_custom_tts_la_CFLAGS   = $(AM_CFLAGS)
+mod_custom_tts_la_LIBADD   = $(switch_builddir)/libfreeswitch.la
+mod_custom_tts_la_LDFLAGS  = -avoid-version -module -no-undefined -shared -lstdc++ -lboost_system -lboost_thread -lmpg123

mod_custom_tts/README.md

@@ -1,3 +1,3 @@
-# mod_google_tts
+# mod_custom_tts
 
-A Freeswitch module that allows speak text to speech audio from Google stream.
+A Freeswitch module that allows speak text to speech audio from custom vendor stream.

mod_custom_tts/custom_glue.cpp

@@ -0,0 +1,960 @@
+#include "mod_custom_tts.h"
+#include <switch.h>
+#include <switch_json.h>
+#include <curl/curl.h>
+#include <cstdlib>
+
+#include <boost/circular_buffer.hpp>
+#include <boost/thread.hpp>
+#include <boost/asio.hpp>
+#include <boost/asio/ssl.hpp>
+#include <boost/pool/object_pool.hpp>
+#include <boost/bind/bind.hpp>
+#include <boost/tokenizer.hpp>
+#include <boost/foreach.hpp>
+#include <boost/asio.hpp>
+#include <boost/assign/list_of.hpp>
+#include <boost/algorithm/string.hpp>
+
+#include "mpg123.h"
+
+#define BUFFER_GROW_SIZE (8192)
+#define MP3_DCACHE 8192 * 2
+
+typedef boost::circular_buffer<uint16_t> CircularBuffer_t;
+/* Global information, common to all connections */
+typedef struct
+{
+  CURLM *multi;
+  int still_running;
+} GlobalInfo_t;
+static GlobalInfo_t global;
+
+/* Information associated with a specific easy handle */
+typedef struct
+{
+  CURL *easy;
+  custom_t* custom;
+  char* body;
+  struct curl_slist *hdr_list;
+  GlobalInfo_t *global;
+  mpg123_handle *mh;
+  char error[CURL_ERROR_SIZE];
+  FILE* file;
+  std::chrono::time_point<std::chrono::high_resolution_clock> startTime;
+  bool flushed;
+} ConnInfo_t;
+
+
+static boost::object_pool<ConnInfo_t> pool ;
+static std::map<curl_socket_t, boost::asio::ip::tcp::socket *> socket_map;
+static boost::asio::io_service io_service;
+static boost::asio::deadline_timer timer(io_service);
+static std::string fullDirPath;
+static std::thread worker_thread;
+
+std::string secondsToMillisecondsString(double seconds) {
+    // Convert to milliseconds
+    double milliseconds = seconds * 1000.0;
+
+    // Truncate to remove fractional part
+    long milliseconds_long = static_cast<long>(milliseconds);
+
+    // Convert to string
+    return std::to_string(milliseconds_long);
+}
+
+static CURL* createEasyHandle(void) {
+  CURL* easy = curl_easy_init();
+  if(!easy) {
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "curl_easy_init() failed!\n");
+    return nullptr ;
+  }  
+
+  curl_easy_setopt(easy, CURLOPT_FOLLOWLOCATION, 1L);
+  curl_easy_setopt(easy, CURLOPT_USERAGENT, "jambonz/0.8.5");
+
+  // 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);
+
+  return easy ;    
+}
+
+static void cleanupConn(ConnInfo_t *conn) {
+  auto c = conn->custom;
+
+  if (conn->mh) {
+    mpg123_close(conn->mh);
+    mpg123_delete(conn->mh);
+  }
+
+  if( conn->hdr_list ) {
+    curl_slist_free_all(conn->hdr_list);
+    conn->hdr_list = nullptr ;
+  }
+  curl_easy_cleanup(conn->easy);
+
+  if (conn->file) {
+    if (fclose(conn->file) != 0) {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "cleanupConn: error closing audio cache file\n");
+    }
+    conn->file = nullptr ;
+  }
+
+  c->conn = nullptr ;
+  c->draining = 1;
+
+  memset(conn, 0, sizeof(ConnInfo_t));
+  pool.destroy(conn) ;
+}
+
+/* Check for completed transfers, and remove their easy handles */
+void check_multi_info(GlobalInfo_t *g) {
+  CURLMsg *msg;
+  int msgs_left;
+  ConnInfo_t *conn;
+  CURL *easy;
+  CURLcode res;
+  
+  while((msg = curl_multi_info_read(g->multi, &msgs_left))) {
+    if(msg->msg == CURLMSG_DONE) {
+      long response_code;
+      double namelookup=0, connect=0, total=0 ;
+      char *ct = NULL ;
+
+      easy = msg->easy_handle;
+      res = msg->data.result;
+      curl_easy_getinfo(easy, CURLINFO_PRIVATE, &conn);
+      curl_easy_getinfo(easy, CURLINFO_RESPONSE_CODE, &response_code);
+      curl_easy_getinfo(easy, CURLINFO_CONTENT_TYPE, &ct);
+
+      curl_easy_getinfo(easy, CURLINFO_NAMELOOKUP_TIME, &namelookup);
+      curl_easy_getinfo(easy, CURLINFO_CONNECT_TIME, &connect);
+      curl_easy_getinfo(easy, CURLINFO_TOTAL_TIME, &total);
+
+      auto c = conn->custom;
+      c->response_code = response_code;
+      if (ct) c->ct = strdup(ct);
+
+      std::string name_lookup_ms = secondsToMillisecondsString(namelookup);
+      std::string connect_ms = secondsToMillisecondsString(connect);
+      std::string final_response_time_ms = secondsToMillisecondsString(total);
+
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, 
+        "mod_custom_tts: response: %ld, content-type %s,"
+        "dns(ms): %"  CURL_FORMAT_CURL_OFF_T ".%06ld, "
+        "connect(ms): %"  CURL_FORMAT_CURL_OFF_T ".%06ld, "
+        "total(ms): %"  CURL_FORMAT_CURL_OFF_T ".%06ld\n",
+        response_code, ct,
+        (long)(namelookup), (long)(fmod(namelookup, 1.0) * 1000000),
+        (long)(connect), (long)(fmod(connect, 1.0) * 1000000),
+        (long)(total), (long)(fmod(total, 1.0) * 1000000));
+
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "name lookup time: %s\n", name_lookup_ms.c_str());
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "connect time: %s\n", connect_ms.c_str());
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "final response time: %s\n", final_response_time_ms.c_str());
+
+      c->name_lookup_time_ms = strdup(name_lookup_ms.c_str());
+      c->connect_time_ms = strdup(connect_ms.c_str());
+      c->final_response_time_ms = strdup(final_response_time_ms.c_str());
+
+      curl_multi_remove_handle(g->multi, easy);
+      cleanupConn(conn);
+    }
+  }
+}
+
+int mcode_test(const char *where, CURLMcode code) {
+  if(CURLM_OK != code) {
+    const char *s;
+    switch(code) {
+    case CURLM_CALL_MULTI_PERFORM:
+      s = "CURLM_CALL_MULTI_PERFORM";
+      break;
+    case CURLM_BAD_HANDLE:
+      s = "CURLM_BAD_HANDLE";
+      break;
+    case CURLM_BAD_EASY_HANDLE:
+      s = "CURLM_BAD_EASY_HANDLE";
+      break;
+    case CURLM_OUT_OF_MEMORY:
+      s = "CURLM_OUT_OF_MEMORY";
+      break;
+    case CURLM_INTERNAL_ERROR:
+      s = "CURLM_INTERNAL_ERROR";
+      break;
+    case CURLM_UNKNOWN_OPTION:
+      s = "CURLM_UNKNOWN_OPTION";
+      break;
+    case CURLM_LAST:
+      s = "CURLM_LAST";
+      break;
+    default:
+      s = "CURLM_unknown";
+      break;
+    case CURLM_BAD_SOCKET:
+      s = "CURLM_BAD_SOCKET";
+      break;
+    }
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "mcode_test ERROR: %s returns %s:%d\n", where, s, code);
+
+    return -1;
+  }
+  return 0 ;
+}
+
+static void remsock(int *f, GlobalInfo_t *g) {
+  if(f) {
+    free(f);
+    f = NULL;
+  }
+}
+
+/* Called by asio when there is an action on a socket */
+static void event_cb(GlobalInfo_t *g, curl_socket_t s, int action, const boost::system::error_code & error, int *fdp) {
+  int f = *fdp;
+
+  //switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "event_cb socket %#X has action %d\n", s, action) ; 
+
+  // Socket already POOL REMOVED.
+  if (f == CURL_POLL_REMOVE) {
+    //switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "event_cb socket %#X removed\n", s); 
+    remsock(fdp, g);
+    return;
+  }
+
+  if(socket_map.find(s) == socket_map.end()) {
+    //switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "event_cb: socket  %#X already closed\n, s");
+    return;
+  }
+
+  /* make sure the event matches what are wanted */
+  if(f == action || f == CURL_POLL_INOUT) {
+    if(error) {
+      action = CURL_CSELECT_ERR;
+    }
+    CURLMcode rc = curl_multi_socket_action(g->multi, s, action, &g->still_running);
+
+    mcode_test("event_cb: curl_multi_socket_action", rc);
+    check_multi_info(g);
+
+    if(g->still_running <= 0) {
+      timer.cancel();
+    }
+
+    /* keep on watching.
+      * the socket may have been closed and/or fdp may have been changed
+      * in curl_multi_socket_action(), so check them both */
+    if(!error && socket_map.find(s) != socket_map.end() &&
+        (f == action || f == CURL_POLL_INOUT)) {
+      boost::asio::ip::tcp::socket *tcp_socket = socket_map.find(s)->second;
+
+      if(action == CURL_POLL_IN) {
+        tcp_socket->async_read_some(boost::asio::null_buffers(),
+                                    boost::bind(&event_cb, g, s,
+                                                action, boost::placeholders::_1, fdp));
+      }
+      if(action == CURL_POLL_OUT) {
+        tcp_socket->async_write_some(boost::asio::null_buffers(),
+                                      boost::bind(&event_cb, g, s,
+                                                  action, boost::placeholders::_1, fdp));
+      } 
+    }
+  }
+}
+
+/* socket functions */
+static void setsock(int *fdp, curl_socket_t s, CURL *e, int act, int oldact, GlobalInfo_t *g) {
+  std::map<curl_socket_t, boost::asio::ip::tcp::socket *>::iterator it = socket_map.find(s);
+
+  if(it == socket_map.end()) {
+    //switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "setsock: socket  %#X not found\n, s");
+    return;
+  }
+
+  boost::asio::ip::tcp::socket * tcp_socket = it->second;
+
+  *fdp = act;
+
+  if(act == CURL_POLL_IN) {
+    if(oldact != CURL_POLL_IN && oldact != CURL_POLL_INOUT) {
+      tcp_socket->async_read_some(boost::asio::null_buffers(),
+                                  boost::bind(&event_cb, g, s,
+                                  CURL_POLL_IN, boost::placeholders::_1, fdp));
+    }
+  }
+  else if(act == CURL_POLL_OUT) {
+    if(oldact != CURL_POLL_OUT && oldact != CURL_POLL_INOUT) {
+      tcp_socket->async_write_some(boost::asio::null_buffers(),
+                                    boost::bind(&event_cb, g, s,
+                                    CURL_POLL_OUT, boost::placeholders::_1, fdp));
+    }
+  }
+  else if(act == CURL_POLL_INOUT) {
+    if(oldact != CURL_POLL_IN && oldact != CURL_POLL_INOUT) {
+      tcp_socket->async_read_some(boost::asio::null_buffers(),
+                                  boost::bind(&event_cb, g, s,
+                                  CURL_POLL_IN, boost::placeholders::_1, fdp));
+    }
+    if(oldact != CURL_POLL_OUT && oldact != CURL_POLL_INOUT) {
+      tcp_socket->async_write_some(boost::asio::null_buffers(),
+                                    boost::bind(&event_cb, g, s,
+                                    CURL_POLL_OUT, boost::placeholders::_1, fdp));
+    }
+  }
+}
+
+static void addsock(curl_socket_t s, CURL *easy, int action, GlobalInfo_t *g) {
+  /* fdp is used to store current action */
+  int *fdp = (int *) calloc(sizeof(int), 1);
+
+  setsock(fdp, s, easy, action, 0, g);
+  curl_multi_assign(g->multi, s, fdp);
+}
+
+static int sock_cb(CURL *e, curl_socket_t s, int what, void *cbp, void *sockp) {
+  GlobalInfo_t *g = &global;
+
+  int *actionp = (int *) sockp;
+  static const char *whatstr[] = { "none", "IN", "OUT", "INOUT", "REMOVE"};
+
+  if(what == CURL_POLL_REMOVE) {
+    *actionp = what;
+  }
+  else {
+    if(!actionp) {
+      addsock(s, e, what, g);
+    }
+    else {
+      setsock(actionp, s, e, what, *actionp, g);
+    }
+  }
+  return 0;  
+}
+
+static void threadFunc() {      
+  /* to make sure the event loop doesn't terminate when there is no work to do */
+  io_service.reset() ;
+  boost::asio::io_service::work work(io_service);
+  
+  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "mod_custom_tts threadFunc - starting\n");
+
+  for(;;) {
+      
+    try {
+      io_service.run() ;
+      break ;
+    }
+    catch( std::exception& e) {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "mod_custom_tts threadFunc - Error: %s\n", e.what());
+    }
+  }
+  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "mod_custom_tts threadFunc - ending\n");
+}
+
+
+/* Called by asio when our timeout expires */
+static void timer_cb(const boost::system::error_code & error, GlobalInfo_t *g)
+{
+  //switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "timer_cb\n");
+
+  if(!error) {
+    CURLMcode rc = curl_multi_socket_action(g->multi, CURL_SOCKET_TIMEOUT, 0, &g->still_running);
+    mcode_test("timer_cb: curl_multi_socket_action", rc);
+    check_multi_info(g);
+  }
+}
+
+int multi_timer_cb(CURLM *multi, long timeout_ms, GlobalInfo_t *g) {
+
+  /* cancel running timer */
+  timer.cancel();
+
+  if(timeout_ms >= 0) {
+    // from libcurl 7.88.1-10+deb12u4 does not allow call curl_multi_socket_action or curl_multi_perform in curl_multi callback directly
+    timer.expires_from_now(boost::posix_time::millisec(timeout_ms ? timeout_ms : 1));
+    timer.async_wait(boost::bind(&timer_cb, boost::placeholders::_1, g));
+  }
+
+  return 0;
+}
+
+static std::vector<uint16_t> convert_mp3_to_linear(ConnInfo_t *conn, uint8_t *data, size_t len) {
+  std::vector<uint16_t> linear_data;
+  int eof = 0;
+  int mp3err = 0;
+  unsigned char decode_buf[MP3_DCACHE];
+
+  if(mpg123_feed(conn->mh, data, len) == MPG123_OK) {
+    while(!eof) {
+      size_t usedlen = 0;
+      off_t frame_offset;
+      unsigned char* audio;
+
+      int decode_status = mpg123_decode_frame(conn->mh, &frame_offset, &audio, &usedlen);
+
+      switch(decode_status) {
+        case MPG123_NEW_FORMAT:
+          continue;
+
+        case MPG123_OK:
+          for(size_t i = 0; i < usedlen; i += 2) {
+            uint16_t value = reinterpret_cast<uint16_t*>(audio)[i / 2];
+            linear_data.push_back(value);
+          }
+          break;
+
+        case MPG123_DONE:
+        case MPG123_NEED_MORE:
+          eof = 1;
+          break;
+
+        case MPG123_ERR:
+        default:
+          if(++mp3err >= 5) {
+            switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Decoder Error!\n");
+            eof = 1;
+          }
+      }
+
+      if (eof)
+        break;
+
+      mp3err = 0;
+    }
+  }
+
+  return linear_data;
+}
+/* CURLOPT_WRITEFUNCTION */
+static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
+  bool fireEvent = false;
+  uint8_t *data = (uint8_t *) ptr;
+  size_t bytes_received = size * nmemb;
+  auto c = conn->custom;
+  CircularBuffer_t *cBuffer = (CircularBuffer_t *) c->circularBuffer;
+  std::vector<uint16_t> pcm_data;
+  
+  if (conn->flushed || cBuffer == nullptr) {
+    /* this will abort the transfer */
+    return 0;
+  }
+  {
+    switch_mutex_lock(c->mutex);
+
+    if (c->response_code > 0 && c->response_code != 200) {
+      std::string body((char *) ptr, bytes_received);
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: received body %s\n", body.c_str());
+      c->err_msg = strdup(body.c_str());
+      switch_mutex_unlock(c->mutex);
+      return 0;
+    }
+
+    /* cache file will stay in the mp3 format for size (smaller) and simplicity */
+    if (conn->file) fwrite(data, sizeof(uint8_t), bytes_received, conn->file);
+
+    pcm_data = convert_mp3_to_linear(conn, data, bytes_received);
+    size_t bytesResampled = pcm_data.size() * sizeof(uint16_t);
+
+    // Resize the buffer if necessary
+    if (cBuffer->capacity() - cBuffer->size() < (bytesResampled / sizeof(uint16_t))) {
+      //switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "write_cb growing buffer\n"); 
+
+      //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_GROW_SIZE));
+    }
+    
+    /* Push the data into the buffer */
+    cBuffer->insert(cBuffer->end(), pcm_data.data(), pcm_data.data() + pcm_data.size());
+
+    if (0 == c->reads++) {
+      fireEvent = true;
+    }
+    switch_mutex_unlock(c->mutex);
+  }
+  if (fireEvent && c->session_id) {
+    auto endTime = std::chrono::high_resolution_clock::now();
+    auto duration = std::chrono::duration_cast<std::chrono::milliseconds>(endTime - conn->startTime);
+    auto time_to_first_byte_ms = std::to_string(duration.count());
+    switch_core_session_t* session = switch_core_session_locate(c->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) {
+          switch_channel_event_set_data(channel, event);
+
+          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "write_cb: firing playback-started\n");
+
+          switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "Playback-File-Type", "tts_stream");
+          if (c->name_lookup_time_ms) {
+            switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_custom_name_lookup_time_ms", c->name_lookup_time_ms);
+          }
+          if (c->connect_time_ms) {
+            switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_custom_connect_time_ms", c->connect_time_ms);
+          }
+          if (c->final_response_time_ms) {
+            switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_custom_final_response_time_ms", c->final_response_time_ms);
+          }
+          if (c->voice_name) {
+            switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_custom_voice_name", c->voice_name);
+          }
+          if (c->cache_filename) {
+            switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_cache_filename", c->cache_filename);
+          }
+
+          switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_time_to_first_byte_ms", time_to_first_byte_ms.c_str());
+          switch_event_fire(&event);
+        }
+        else {
+          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: failed to create event\n");
+        }
+      }
+      else {
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: channel not found\n");
+      }
+    }
+    else {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: session %s not found\n", c->session_id);
+    }
+  }
+  return bytes_received;
+}
+
+static bool parseHeader(const std::string& str, std::string& header, std::string& value) {
+    std::vector<std::string> parts;
+    boost::split(parts, str, boost::is_any_of(":"), boost::token_compress_on);
+
+    if (parts.size() != 2)
+        return false;
+
+    header = boost::trim_copy(parts[0]);
+    value = boost::trim_copy(parts[1]);
+    return true;
+}
+
+static int extract_response_code(const std::string& input) {
+  std::size_t space_pos = input.find(' ');
+  if (space_pos == std::string::npos) {
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Invalid HTTP response format %s\n", input.c_str());
+    return 0;
+  }
+
+  std::size_t code_start_pos = space_pos + 1;
+  std::size_t code_end_pos = input.find(' ', code_start_pos);
+  if (code_end_pos == std::string::npos) {
+    code_end_pos = input.length();
+  }
+
+  std::string code_str = input.substr(code_start_pos, code_end_pos - code_start_pos);
+  int response_code = std::stoi(code_str);
+  return response_code;
+}
+
+static size_t header_callback(char *buffer, size_t size, size_t nitems, ConnInfo_t *conn) {
+  size_t bytes_received = size * nitems;
+  const std::string prefix = "HTTP/";
+  custom_t* c = conn->custom;
+  std::string header, value;
+  std::string input(buffer, bytes_received);
+  if (parseHeader(input, header, value)) {
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "recv header: %s with value %s\n", header.c_str(), value.c_str());
+  }
+  else {
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "header_callback: %s\n", input.c_str());
+    if (input.rfind(prefix, 0) == 0) {
+      try {
+        c->response_code = extract_response_code(input);
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "header_callback: parsed response code: %ld\n", c->response_code);
+      } catch (const std::invalid_argument& e) {
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "header_callback: invalid response code %s\n", input.substr(prefix.length()).c_str());
+      }
+    }
+  }
+  return bytes_received;
+}
+
+/* CURLOPT_OPENSOCKETFUNCTION */
+static curl_socket_t opensocket(void *clientp, curlsocktype purpose, struct curl_sockaddr *address) {
+  curl_socket_t sockfd = CURL_SOCKET_BAD;
+
+  //switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "opensocket: %d\n", purpose);
+  /* restrict to IPv4 */
+  if(purpose == CURLSOCKTYPE_IPCXN && address->family == AF_INET) {
+    /* create a tcp socket object */
+    boost::asio::ip::tcp::socket *tcp_socket = new boost::asio::ip::tcp::socket(io_service);
+
+    /* open it and get the native handle*/
+    boost::system::error_code ec;
+    tcp_socket->open(boost::asio::ip::tcp::v4(), ec);
+
+    if(ec) {
+      /* An error occurred */
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Couldn't open socket [%ld][%s]\n", ec, ec.message().c_str());
+    }
+    else {
+      sockfd = tcp_socket->native_handle();
+
+      /* save it for monitoring */
+      socket_map.insert(std::pair<curl_socket_t, boost::asio::ip::tcp::socket *>(sockfd, tcp_socket));
+    }
+  }
+  return sockfd;
+}
+
+/* CURLOPT_CLOSESOCKETFUNCTION */
+static int close_socket(void *clientp, curl_socket_t item) {
+  //switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "close_socket : %#X\n", item);
+
+  std::map<curl_socket_t, boost::asio::ip::tcp::socket *>::iterator it = socket_map.find(item);
+  if(it != socket_map.end()) {
+    delete it->second;
+    socket_map.erase(it);
+  }
+  return 0;
+}
+
+
+extern "C" {
+  switch_status_t custom_speech_load() {
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "custom_speech_loading..\n");
+    memset(&global, 0, sizeof(GlobalInfo_t));
+    global.multi = curl_multi_init();
+
+     if (!global.multi) {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "custom_speech_load curl_multi_init() failed, exiting!\n");
+      return SWITCH_STATUS_FALSE;
+    }
+
+    curl_multi_setopt(global.multi, CURLMOPT_SOCKETFUNCTION, sock_cb);
+    curl_multi_setopt(global.multi, CURLMOPT_SOCKETDATA, &global);
+    curl_multi_setopt(global.multi, CURLMOPT_TIMERFUNCTION, multi_timer_cb);
+    curl_multi_setopt(global.multi, CURLMOPT_TIMERDATA, &global);
+    curl_multi_setopt(global.multi, CURLMOPT_PIPELINING, CURLPIPE_MULTIPLEX);
+
+    /* create temp folder for cache files */
+    const char* baseDir = std::getenv("JAMBONZ_TMP_CACHE_FOLDER");
+    if (!baseDir) {
+      baseDir = "/tmp/";
+    }
+    if (strcmp(baseDir, "/") == 0) {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "failed to create folder %s\n", baseDir);
+      return SWITCH_STATUS_FALSE;
+    }
+
+    fullDirPath = std::string(baseDir) + "tts-cache-files";
+
+    // Create the directory with read, write, and execute permissions for everyone
+    mode_t oldMask = umask(0);
+    int result = mkdir(fullDirPath.c_str(), S_IRWXU | S_IRWXG | S_IRWXO);
+    umask(oldMask);
+    if (result != 0) {
+      if (errno != EEXIST) {
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "failed to create folder %s\n", fullDirPath.c_str());
+        fullDirPath = "";
+      }
+      else switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "folder %s already exists\n", fullDirPath.c_str());
+    }
+    else {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "created folder %s\n", fullDirPath.c_str());
+    }
+    // init mgp123
+    if (mpg123_init() != MPG123_OK) {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "failed to initiate MPG123");
+      return SWITCH_STATUS_FALSE;
+    }
+
+    /* start worker thread that handles transfers*/
+    std::thread t(threadFunc) ;
+    worker_thread.swap( t ) ;
+
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "custom_speech_loaded..\n");
+
+
+    return SWITCH_STATUS_SUCCESS;
+  }
+
+  switch_status_t custom_speech_unload() {
+    /* stop the ASIO IO service */
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "custom_speech_unload: stopping io service\n");
+    io_service.stop();
+
+    /* Join the worker thread */
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "custom_speech_unload: wait for worker thread to complete\n");
+    if (worker_thread.joinable()) {
+        worker_thread.join();
+    }
+
+    /* cleanup curl multi handle*/
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "custom_speech_unload: release curl multi\n");
+    curl_multi_cleanup(global.multi);
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "custom_speech_unload: completed\n");
+
+    mpg123_exit();
+
+		return SWITCH_STATUS_SUCCESS;
+  }
+
+  switch_status_t custom_speech_open(custom_t* custom) {
+    return SWITCH_STATUS_SUCCESS;
+  }
+
+  switch_status_t custom_speech_feed_tts(custom_t* c, char* text, switch_speech_flag_t *flags) {
+    CURLMcode rc;
+
+    const int MAX_CHARS = 20;
+    char tempText[MAX_CHARS + 4]; // +4 for the ellipsis and null terminator
+
+    if (strlen(text) > MAX_CHARS) {
+        strncpy(tempText, text, MAX_CHARS);
+        strcpy(tempText + MAX_CHARS, "...");
+    } else {
+        strcpy(tempText, text);
+    }
+
+    /* open cache file */
+    if (c->cache_audio && fullDirPath.length() > 0) {
+      switch_uuid_t uuid;
+      char uuid_str[SWITCH_UUID_FORMATTED_LENGTH + 1];
+      char outfile[512] = "";
+      int fd;
+
+      switch_uuid_get(&uuid);
+      switch_uuid_format(uuid_str, &uuid);
+
+      switch_snprintf(outfile, sizeof(outfile), "%s%s%s.mp3", fullDirPath.c_str(), SWITCH_PATH_SEPARATOR, uuid_str);
+      c->cache_filename = strdup(outfile);
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "writing audio cache file to %s\n", c->cache_filename);
+
+      mode_t oldMask = umask(0);
+      fd = open(outfile, O_WRONLY | O_CREAT | O_TRUNC, S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH | S_IWOTH);
+      umask(oldMask);
+      if (fd == -1 ) {
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error opening cache file %s: %s\n", outfile, strerror(errno));
+      }
+      else {
+        c->file = fdopen(fd, "wb");
+        if (!c->file) {
+          close(fd);
+          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error opening cache file %s: %s\n", outfile, strerror(errno));
+        }
+      }
+    }
+
+    if (!c->custom_tts_url) {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "custom_speech_feed_tts: no custom_tts_url provided\n");
+      return SWITCH_STATUS_FALSE;
+    }
+
+    /* format url*/
+    std::string url = c->custom_tts_url;
+
+    /* create the JSON body */
+    cJSON * jResult = cJSON_CreateObject();
+    cJSON_AddStringToObject(jResult, "text", text);
+    cJSON_AddStringToObject(jResult, "type", std::strncmp(text, "<speak", 6) == 0 ? "ssml" : "text");
+    if (c->language) {
+      cJSON_AddStringToObject(jResult, "language", c->language);
+    }
+    if (c->voice_name) {
+      cJSON_AddStringToObject(jResult, "voice", c->voice_name);
+    }
+    char *json = cJSON_PrintUnformatted(jResult);
+
+    cJSON_Delete(jResult);
+
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "custom_speech_feed_tts: [%s] [%s]\n", url.c_str(), tempText);
+
+    ConnInfo_t *conn = pool.malloc() ;
+
+    // COnfigure MPG123
+    int mhError = 0;
+    mpg123_handle *mh = mpg123_new("auto", &mhError);
+    if (!mh) {
+      const char *mhErr = mpg123_plain_strerror(mhError);
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error allocating mpg123 handle! %s\n", switch_str_nil(mhErr));
+      return SWITCH_STATUS_FALSE;
+    }
+
+    if (mpg123_open_feed(mh) != MPG123_OK) {
+			switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error mpg123_open_feed!\n");
+      return SWITCH_STATUS_FALSE;
+		}
+
+    if (mpg123_format_all(mh) != MPG123_OK) {
+			switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error mpg123_format_all!\n");
+      return SWITCH_STATUS_FALSE;
+		}
+
+    if (mpg123_param(mh, MPG123_FLAGS, MPG123_MONO_MIX, 0) != MPG123_OK) {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error forcing single channel!\n");
+      return SWITCH_STATUS_FALSE;
+    }
+
+    CURL* easy = createEasyHandle();
+    c->conn = (void *) conn ;
+    conn->custom = c;
+    conn->easy = easy;
+    conn->mh = mh;
+    conn->global = &global;
+    conn->hdr_list = NULL ;
+    conn->file = c->file;
+    conn->body = json;
+    conn->flushed = false;
+    
+
+    c->circularBuffer = (void *) new CircularBuffer_t(8192);
+
+    if (mpg123_param(mh, MPG123_FORCE_RATE, c->rate /*Hz*/, 0) != MPG123_OK) {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error mpg123_param!\n");
+      return SWITCH_STATUS_FALSE;
+    }
+
+    curl_easy_setopt(easy, CURLOPT_URL, url.c_str());
+    curl_easy_setopt(easy, CURLOPT_WRITEFUNCTION, write_cb);
+    curl_easy_setopt(easy, CURLOPT_WRITEDATA, conn);
+    curl_easy_setopt(easy, CURLOPT_ERRORBUFFER, conn->error);
+    curl_easy_setopt(easy, CURLOPT_PRIVATE, conn);
+    curl_easy_setopt(easy, CURLOPT_VERBOSE, 0L);
+    curl_easy_setopt(easy, CURLOPT_NOPROGRESS, 1L);
+    curl_easy_setopt(easy, CURLOPT_HEADERFUNCTION, header_callback);
+    curl_easy_setopt(easy, CURLOPT_HEADERDATA, conn);
+    
+    /* call this function to get a socket */
+    curl_easy_setopt(easy, CURLOPT_OPENSOCKETFUNCTION, opensocket);
+
+    /* call this function to close a socket */
+    curl_easy_setopt(easy, CURLOPT_CLOSESOCKETFUNCTION, close_socket);
+
+    if (c->auth_token) {
+      std::ostringstream api_key_stream;
+      api_key_stream << "Authorization: Bearer " << c->auth_token;
+      conn->hdr_list = curl_slist_append(conn->hdr_list, api_key_stream.str().c_str());
+    }
+    conn->hdr_list = curl_slist_append(conn->hdr_list, "Accept: audio/mpeg");
+    conn->hdr_list = curl_slist_append(conn->hdr_list, "Content-Type: application/json");
+    curl_easy_setopt(easy, CURLOPT_HTTPHEADER, conn->hdr_list);
+
+    curl_easy_setopt(easy, CURLOPT_POSTFIELDS, conn->body);
+    //curl_easy_setopt(easy, CURLOPT_POSTFIELDSIZE, body.length());
+
+    curl_easy_setopt(easy, CURLOPT_HTTP_VERSION, CURL_HTTP_VERSION_2_0);
+
+    rc = curl_multi_add_handle(global.multi, conn->easy);
+    mcode_test("new_conn: curl_multi_add_handle", rc);
+
+    /* start a timer to measure the duration until we receive first byte of audio */
+    conn->startTime = std::chrono::high_resolution_clock::now();
+
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "custom_speech_feed_tts: called curl_multi_add_handle\n");
+
+
+    return SWITCH_STATUS_SUCCESS;
+  }
+
+  switch_status_t custom_speech_read_tts(custom_t* c, void *data, size_t *datalen, switch_speech_flag_t *flags) {
+    CircularBuffer_t *cBuffer = (CircularBuffer_t *) c->circularBuffer;
+    std::vector<uint16_t> pcm_data;
+
+    {
+      switch_mutex_lock(c->mutex);
+      ConnInfo_t *conn = (ConnInfo_t *) c->conn;
+      if (c->response_code > 0 && c->response_code != 200) {
+        switch_mutex_unlock(c->mutex);
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "custom_speech_read_tts, returning failure\n") ;  
+        return SWITCH_STATUS_FALSE;
+      }
+      if (conn && conn->flushed) {
+        switch_mutex_unlock(c->mutex);
+        return SWITCH_STATUS_BREAK;
+      }
+      if (cBuffer->empty()) {
+        if (c->draining) {
+          switch_mutex_unlock(c->mutex);
+          return SWITCH_STATUS_BREAK;
+        }
+        /* no audio available yet so send silence */
+        memset(data, 255, *datalen);
+        switch_mutex_unlock(c->mutex);
+        return SWITCH_STATUS_SUCCESS;
+      }
+      size_t 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(c->mutex);
+    }
+
+    memcpy(data, pcm_data.data(), pcm_data.size() * sizeof(uint16_t));
+    *datalen = pcm_data.size() * sizeof(uint16_t);
+
+    return SWITCH_STATUS_SUCCESS;
+  }
+
+  switch_status_t custom_speech_flush_tts(custom_t* c) {
+    bool download_complete = c->response_code == 200;
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "custom_speech_flush_tts, download complete? %s\n", download_complete ? "yes" : "no") ;  
+    ConnInfo_t *conn = (ConnInfo_t *) c->conn;
+    CircularBuffer_t *cBuffer = (CircularBuffer_t *) c->circularBuffer;
+    delete cBuffer;
+    c->circularBuffer = nullptr ;
+
+    if (conn) {
+      conn->flushed = true;
+      if (!download_complete) {
+        if (conn->file) {
+          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "closing audio cache file %s because download was interrupted\n", c->cache_filename);
+          if (fclose(conn->file) != 0) {
+            switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "error closing audio cache file\n");
+          }
+          conn->file = nullptr ;
+        }
+
+        if (c->cache_filename) {
+          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "removing audio cache file %s because download was interrupted\n", c->cache_filename);
+          if (unlink(c->cache_filename) != 0) {
+            switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "cleanupConn: error removing audio cache file %s: %d:%s\n", 
+              c->cache_filename, errno, strerror(errno));
+          }
+          free(c->cache_filename);
+          c->cache_filename = nullptr ;
+        }
+      }
+    }
+    if (c->session_id) {
+      switch_core_session_t* session = switch_core_session_locate(c->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) {
+            switch_channel_event_set_data(channel, event);
+            switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "Playback-File-Type", "tts_stream");
+            switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_custom_response_code", std::to_string(c->response_code).c_str());
+            if (c->cache_filename && c->response_code == 200) {
+              switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_cache_filename", c->cache_filename);
+            }
+            if (c->response_code != 200 && c->err_msg) {
+              switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_error", c->err_msg);
+            }
+            switch_event_fire(&event);
+          }
+          else {
+            switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: failed to create event\n");
+          }
+        }
+        else {
+          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: channel not found\n");
+        }
+      }
+    }
+    return SWITCH_STATUS_SUCCESS;
+  }
+
+	switch_status_t custom_speech_close(custom_t* c) {
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "custom_speech_close\n") ;
+		return SWITCH_STATUS_SUCCESS;
+	}
+}

mod_custom_tts/custom_glue.h

@@ -0,0 +1,12 @@
+#ifndef __CUSTOM_GLUE_H__
+#define __CUSTOM_GLUE_H__
+
+switch_status_t custom_speech_load();
+switch_status_t custom_speech_open(custom_t* custom);
+switch_status_t custom_speech_feed_tts(custom_t* custom, char* text, switch_speech_flag_t *flags);
+switch_status_t custom_speech_read_tts(custom_t* custom, void *data, size_t *datalen, switch_speech_flag_t *flags);
+switch_status_t custom_speech_flush_tts(custom_t* custom);
+switch_status_t custom_speech_close(custom_t* custom);
+switch_status_t custom_speech_unload();
+
+#endif

mod_custom_tts/mod_custom_tts.c

@@ -0,0 +1,158 @@
+#include "mod_custom_tts.h"
+#include "custom_glue.h"
+
+SWITCH_MODULE_LOAD_FUNCTION(mod_custom_tts_load);
+SWITCH_MODULE_SHUTDOWN_FUNCTION(mod_custom_tts_shutdown);
+SWITCH_MODULE_DEFINITION(mod_custom_tts, mod_custom_tts_load, mod_custom_tts_shutdown, NULL);
+
+static void clearCustomVendor(custom_t* c, int freeAll) {
+  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "clearCustomVendor\n");
+  if (c->auth_token) free(c->auth_token);
+  if (c->custom_tts_url) free(c->custom_tts_url);
+  if (c->language) free(c->language);
+  if (c->ct) free(c->ct);
+  if (c->err_msg) free(c->err_msg);
+  if (c->name_lookup_time_ms) free(c->name_lookup_time_ms);
+  if (c->connect_time_ms) free(c->connect_time_ms);
+  if (c->final_response_time_ms) free(c->final_response_time_ms);
+  if (c->cache_filename) free(c->cache_filename);
+  
+
+  c->auth_token = NULL;
+  c->custom_tts_url = NULL;
+  c->language = NULL;
+  c->ct = NULL;
+  c->err_msg = NULL;
+  c->name_lookup_time_ms = NULL;
+  c->connect_time_ms = NULL;
+  c->final_response_time_ms = NULL;
+  c->cache_filename = NULL;
+
+  if (freeAll) {
+    if (c->voice_name) free(c->voice_name);
+    if (c->session_id) free(c->session_id);
+    c->voice_name = NULL;
+    c->session_id = NULL;
+  }
+}
+
+static custom_t * createOrRetrievePrivateData(switch_speech_handle_t *sh) {
+  custom_t *c = (custom_t *) sh->private_info;  
+  if (!c) {
+    c = switch_core_alloc(sh->memory_pool, sizeof(*c));
+  	sh->private_info = c;
+    memset(c, 0, sizeof(*c));
+    switch_mutex_init(&c->mutex, SWITCH_MUTEX_NESTED, sh->memory_pool);
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "allocated custom_t\n");
+  }
+  return c;
+}
+
+switch_status_t w_speech_open(switch_speech_handle_t *sh, const char *voice_name, int rate, int channels, switch_speech_flag_t *flags)
+{
+  custom_t *c = createOrRetrievePrivateData(sh);
+  c->voice_name = strdup(voice_name);
+  c->rate = rate;
+  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "w_speech_open voice: %s, rate %d, channels %d\n", voice_name, rate, channels);
+  return custom_speech_open(c);
+}
+
+static switch_status_t w_speech_close(switch_speech_handle_t *sh, switch_speech_flag_t *flags)
+{
+  switch_status_t rc;
+  custom_t *c = createOrRetrievePrivateData(sh);
+  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "w_speech_close\n");
+
+  switch_mutex_destroy(c->mutex);
+
+  rc = custom_speech_close(c);
+  clearCustomVendor(c, 1);
+  return rc;
+}
+
+/**
+ * Freeswitch will call this function to feed us text to speak
+ */
+static switch_status_t w_speech_feed_tts(switch_speech_handle_t *sh, char *text, switch_speech_flag_t *flags)
+{
+  custom_t *c = createOrRetrievePrivateData(sh);
+  c->draining = 0;
+  c->reads = 0;
+  c->response_code = 0;
+  c->err_msg = NULL;
+
+  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "w_speech_feed_tts\n");
+
+  return custom_speech_feed_tts(c, text, flags);
+}
+
+/**
+ * Freeswitch calls periodically to get some rendered audio in L16 format. We can provide up to 8k of audio at a time.
+ */
+static switch_status_t w_speech_read_tts(switch_speech_handle_t *sh, void *data, size_t *datalen, switch_speech_flag_t *flags)
+{
+  custom_t *c = createOrRetrievePrivateData(sh);
+  return custom_speech_read_tts(c, data, datalen, flags);
+}
+
+/**
+ * This is called at the end, not sure exactly what we need to do here..
+ */
+static void w_speech_flush_tts(switch_speech_handle_t *sh)
+{
+  custom_t *c = createOrRetrievePrivateData(sh);
+  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "w_speech_flush_tts\n");
+  custom_speech_flush_tts(c);
+
+  clearCustomVendor(c, 0);
+}
+
+static void w_text_param_tts(switch_speech_handle_t *sh, char *param, const char *val)
+{
+  custom_t *c = createOrRetrievePrivateData(sh);
+  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "w_text_param_tts: %s=%s\n", param, val);
+  if (0 == strcmp(param, "auth_token")) {
+    if (c->auth_token) free(c->auth_token);
+    c->auth_token = strdup(val);
+  } else if (0 == strcmp(param, "custom_tts_url")) {
+    if (c->custom_tts_url) free(c->custom_tts_url);
+    c->custom_tts_url = strdup(val);
+  } else if (0 == strcmp(param, "language")) {
+    if (c->language) free(c->language);
+    c->language = strdup(val);
+  } else if (0 == strcmp(param, "session-uuid")) {
+    if (c->session_id) free(c->session_id);
+    c->session_id = strdup(val);
+  } else if (0 == strcmp(param, "write_cache_file") && switch_true(val)) {
+    c->cache_audio = 1;
+  }
+}
+static void w_numeric_param_tts(switch_speech_handle_t *sh, char *param, int val)
+{
+}
+static void w_float_param_tts(switch_speech_handle_t *sh, char *param, double val)
+{
+}
+
+SWITCH_MODULE_LOAD_FUNCTION(mod_custom_tts_load)
+{
+  switch_speech_interface_t *speech_interface;
+
+  *module_interface = switch_loadable_module_create_module_interface(pool, modname);
+  speech_interface = switch_loadable_module_create_interface(*module_interface, SWITCH_SPEECH_INTERFACE);
+  speech_interface->interface_name = "custom";
+  speech_interface->speech_open = w_speech_open;
+  speech_interface->speech_close = w_speech_close;
+  speech_interface->speech_feed_tts = w_speech_feed_tts;
+  speech_interface->speech_read_tts = w_speech_read_tts;
+	speech_interface->speech_flush_tts = w_speech_flush_tts;
+	speech_interface->speech_text_param_tts = w_text_param_tts;
+	speech_interface->speech_numeric_param_tts = w_numeric_param_tts;
+	speech_interface->speech_float_param_tts = w_float_param_tts;
+  return custom_speech_load();
+}
+
+SWITCH_MODULE_SHUTDOWN_FUNCTION(mod_custom_tts_shutdown)
+{
+  return custom_speech_unload();
+}

mod_custom_tts/mod_custom_tts.h

@@ -1,34 +1,31 @@
-#ifndef __MOD_GOOGLE_TTS_H__
-#define __MOD_GOOGLE_TTS_H__
+#ifndef __MOD_CUSTOM_TTS_H__
+#define __MOD_CUSTOM_TTS_H__
 
 #include <switch.h>
-#include <speex/speex_resampler.h>
-
-typedef struct google_data {
+typedef struct custom_data {
   char *voice_name;
-  char *model;
-  char *reported_usage;
+  char *auth_token;
+  char *custom_tts_url;
   char *language;
-  char *gender;
 
   /* result data */
   long response_code;
-  char *session_id;
-  char *cache_filename;
+  char *ct;
+  char *name_lookup_time_ms;
+  char *connect_time_ms;
+  char *final_response_time_ms;
   char *err_msg;
+  char *cache_filename;
+  char *session_id;
 
   int rate;
   int draining;
   int reads;
   int cache_audio;
-  int flushed;
 
-  void *startTime;
-
-  FILE *file;
-  SpeexResamplerState *resampler;
+	void *conn;
   void *circularBuffer;
   switch_mutex_t *mutex;
-} google_t;
-
+  FILE *file;
+} custom_t;
 #endif

mod_deepgram_tts/deepgram_glue.cpp

@@ -385,7 +385,7 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
   auto d = conn->deepgram;
   CircularBuffer_t *cBuffer = (CircularBuffer_t *) d->circularBuffer;
   
-  if (conn->flushed) {
+  if (conn->flushed || cBuffer == nullptr) {
     /* this will abort the transfer */
     return 0;
   }
@@ -641,7 +641,7 @@ extern "C" {
       return SWITCH_STATUS_FALSE;
     }
 
-    fullDirPath = std::string(baseDir) + "jambonz-tts-cache-files";
+    fullDirPath = std::string(baseDir) + "tts-cache-files";
 
     // Create the directory with read, write, and execute permissions for everyone
     mode_t oldMask = umask(0);
@@ -764,7 +764,9 @@ extern "C" {
     conn->file = d->file;
     conn->body = json;
     conn->flushed = false;
-    
+    conn->has_last_byte = false;
+    conn->last_byte = 0;
+
 
     d->circularBuffer = (void *) new CircularBuffer_t(BUFFER_GROW_SIZE);
     // Always use deepgram at rate 8000 for helping cache audio from jambonz.

mod_dub/tts_vendor_parser.cpp

@@ -4,6 +4,60 @@
 #include <switch_json.h>
 #include <map>
 
+
+switch_status_t custom_vendor_parse_text(const std::map<std::string, std::string>& params, const std::string& text, 
+  std::string& url, std::string& body, std::vector<std::string>& headers) {
+  std::string auth_token;
+  std::string voice_name;
+  std::string custom_tts_url;
+  std::string language;
+
+  for (const auto& pair : params) {
+    if (pair.first == "auth_token") {
+      auth_token = pair.second;
+    } else if (pair.first == "voice") {
+      voice_name = pair.second;
+    } else if (pair.first == "custom_tts_url") {
+      custom_tts_url = pair.second;
+    } else if (pair.first == "language") {
+      language = pair.second;
+    }
+  }
+
+  if (custom_tts_url.empty()) {
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "custom_vendor_parse_text: no custom_tts_url provided\n");
+    return SWITCH_STATUS_FALSE;
+  }
+
+  url = custom_tts_url;
+
+  /* create the JSON body */
+  cJSON * jResult = cJSON_CreateObject();
+  cJSON_AddStringToObject(jResult, "text", text.c_str());
+  cJSON_AddStringToObject(jResult, "type", text.substr(0, 6) == "<speak" ? "ssml" : "text");
+  cJSON_AddNumberToObject(jResult, "samplingRate", 8000);
+  if (!voice_name.empty()) {
+    cJSON_AddStringToObject(jResult, "voice", voice_name.c_str());
+  }
+  if (!language.empty()) {
+    cJSON_AddStringToObject(jResult, "language", language.c_str());
+  }
+  char* _body = cJSON_PrintUnformatted(jResult);
+  body = _body;
+
+  cJSON_Delete(jResult);
+  free(_body);
+
+  // Create headers
+  if (!auth_token.empty()) {
+    headers.push_back("Authorization: Bearer " + auth_token);
+  }
+  headers.push_back("Accept: audio/mp3");
+  headers.push_back("Content-Type: application/json");
+
+  return SWITCH_STATUS_SUCCESS;
+}
+
 switch_status_t rimelabs_parse_text(const std::map<std::string, std::string>& params, const std::string& text, 
   std::string& url, std::string& body, std::vector<std::string>& headers) {
   std::string api_key;
@@ -476,7 +530,9 @@ switch_status_t tts_vendor_parse_text(const std::string& say, std::string& url,
     return playht_parse_text(params, text, url, body, headers);
   } else if (params["vendor"] == "rimelabs") {
     return rimelabs_parse_text(params, text, url, body, headers);
-  } else {
+  } else if (params["vendor"] == "custom") {
+    return custom_vendor_parse_text(params, text, url, body, headers);
+  }  else {
     switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "tts_vendor_parse_text: There is no available parser for vendor %s\n", params["vendor"]);
     return SWITCH_STATUS_FALSE;
   }

mod_elevenlabs_tts/elevenlabs_glue.cpp

@@ -445,7 +445,7 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
   CircularBuffer_t *cBuffer = (CircularBuffer_t *) el->circularBuffer;
   std::vector<uint16_t> pcm_data;
   
-  if (conn->flushed) {
+  if (conn->flushed || cBuffer == nullptr) {
     /* this will abort the transfer */
     return 0;
   }
@@ -688,7 +688,7 @@ extern "C" {
       return SWITCH_STATUS_FALSE;
     }
 
-    fullDirPath = std::string(baseDir) + "jambonz-tts-cache-files";
+    fullDirPath = std::string(baseDir) + "tts-cache-files";
 
     // Create the directory with read, write, and execute permissions for everyone
     mode_t oldMask = umask(0);
@@ -841,7 +841,6 @@ extern "C" {
     CURL* easy = createEasyHandle();
 
     el->conn = (void *) conn ;
-    el->sample_rate = 0;
     conn->elevenlabs = el;
     conn->easy = easy;
     conn->global = &global;
@@ -852,20 +851,12 @@ extern "C" {
 
     el->circularBuffer = (void *) new CircularBuffer_t(8192);
 
-    if (el->session_id) {
+    if (el->rate != 8000 /*Hz*/) {
       int err;
-      switch_codec_implementation_t read_impl;
-      switch_core_session_t *psession = switch_core_session_locate(el->session_id);
-      switch_core_session_get_read_impl(psession, &read_impl);
-      uint32_t samples_per_second = !strcasecmp(read_impl.iananame, "g722") ? read_impl.actual_samples_per_second : read_impl.samples_per_second;
-      el->sample_rate = samples_per_second;
-      // elevenlabs output is PCMU 8000
-      if (samples_per_second != 8000 /*Hz*/) {
-        el->resampler = speex_resampler_init(1, 8000, samples_per_second, SWITCH_RESAMPLE_QUALITY, &err);
-        if (0 != err) {
-          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error initializing resampler: %s.\n", speex_resampler_strerror(err));
-          return SWITCH_STATUS_FALSE;
-        }
+      el->resampler = speex_resampler_init(1, 8000, el->rate, SWITCH_RESAMPLE_QUALITY, &err);
+      if (0 != err) {
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error initializing resampler: %s.\n", speex_resampler_strerror(err));
+        return SWITCH_STATUS_FALSE;
       }
     }
 
@@ -939,9 +930,7 @@ extern "C" {
         switch_mutex_unlock(el->mutex);
         return SWITCH_STATUS_SUCCESS;
       }
-      size_t size = el->sample_rate ?
-        std::min((*datalen/(2 * el->sample_rate / 8000)), cBuffer->size()) :
-        std::min((*datalen/2), cBuffer->size());
+      size_t size = std::min((*datalen/(2 * el->rate / 8000)), cBuffer->size());
       pcm_data.insert(pcm_data.end(), cBuffer->begin(), cBuffer->begin() + size);
       cBuffer->erase(cBuffer->begin(), cBuffer->begin() + size);
       switch_mutex_unlock(el->mutex);

mod_elevenlabs_tts/mod_elevenlabs_tts.h

@@ -31,7 +31,6 @@ struct elevenlabs_data {
   char *cache_filename;
 
 	int rate;
-  uint32_t sample_rate;
 
   void *conn;
 	FILE *file;

mod_google_transcribe/google_glue_v1.cpp

@@ -236,6 +236,7 @@ static void *SWITCH_THREAD_FUNC grpc_read_thread(switch_thread_t *thread, void *
       switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "grpc_read_thread: error %s (%d)\n", status.message().c_str(), status.code()) ;
       cJSON* json = cJSON_CreateObject();
       cJSON_AddStringToObject(json, "type", "error");
+      cJSON_AddStringToObject(json, "error_cause", "stream_response");
       cJSON_AddStringToObject(json, "error", status.message().c_str());
       char* jsonString = cJSON_PrintUnformatted(json);
       cb->responseHandler(session, jsonString, cb->bugname);
@@ -339,9 +340,10 @@ static void *SWITCH_THREAD_FUNC grpc_read_thread(switch_thread_t *thread, void *
           cb->responseHandler(session, "no_audio", cb->bugname);
         }
       }
-      else {
+      else if (status.error_code() != 0) {
         cJSON* json = cJSON_CreateObject();
         cJSON_AddStringToObject(json, "type", "error");
+        cJSON_AddStringToObject(json, "error_cause", "stream_close");
         cJSON_AddItemToObject(json, "error_code", cJSON_CreateNumber(status.error_code()));
         cJSON_AddStringToObject(json, "error_message", status.error_message().c_str());
         char* jsonString = cJSON_PrintUnformatted(json);

mod_google_transcribe/google_glue_v2.cpp

@@ -349,6 +349,7 @@ static void *SWITCH_THREAD_FUNC grpc_read_thread(switch_thread_t *thread, void *
       else if (status.error_code() != 0) {
         cJSON* json = cJSON_CreateObject();
         cJSON_AddStringToObject(json, "type", "error");
+        cJSON_AddStringToObject(json, "error_cause", "stream_close");
         cJSON_AddItemToObject(json, "error_code", cJSON_CreateNumber(status.error_code()));
         cJSON_AddStringToObject(json, "error_message", status.error_message().c_str());
         char* jsonString = cJSON_PrintUnformatted(json);

mod_google_tts/Makefile.am

@@ -1,9 +0,0 @@
-include $(top_srcdir)/build/modmake.rulesam
-MODNAME=mod_google_tts
-
-mod_LTLIBRARIES = mod_google_tts.la
-mod_google_tts_la_SOURCES  = mod_google_tts.c google_glue.cpp
-mod_google_tts_la_CFLAGS   = $(AM_CFLAGS)
-mod_google_tts_la_CXXFLAGS = -I $(top_srcdir)/libs/googleapis/gens $(AM_CXXFLAGS) -std=c++17
-mod_google_tts_la_LIBADD   = $(switch_builddir)/libfreeswitch.la
-mod_google_tts_la_LDFLAGS  = -avoid-version -module -no-undefined -shared `pkg-config --libs grpc++ grpc` 

mod_google_tts/google_glue.cpp

@@ -1,405 +0,0 @@
-#include "mod_google_tts.h"
-#include <switch.h>
-
-#include <boost/circular_buffer.hpp>
-
-#include <cstdlib>
-#include <string>
-#include <chrono>
-#include <thread>
-
-#include "google/cloud/texttospeech/v1/cloud_tts.grpc.pb.h"
-#include <grpc++/grpc++.h>
-
-using google::cloud::texttospeech::v1::SynthesizeSpeechRequest;
-using google::cloud::texttospeech::v1::SynthesizeSpeechResponse;
-using google::cloud::texttospeech::v1::TextToSpeech;
-using google::cloud::texttospeech::v1::Voice;
-using google::cloud::texttospeech::v1::SsmlVoiceGender;
-using google::cloud::texttospeech::v1::SsmlVoiceGender_Name;
-using google::cloud::texttospeech::v1::SynthesisInput;
-using google::cloud::texttospeech::v1::AudioEncoding;
-
-#define BUFFER_SIZE 8129
-
-typedef boost::circular_buffer<uint16_t> CircularBuffer_t;
-
-static std::string fullDirPath;
-
-static std::shared_ptr<grpc::Channel> create_grpc_channel(switch_channel_t *channel) {
-  const char* google_uri = "texttospeech.googleapis.com";
-  const char* var;
-  if (var = switch_channel_get_variable(channel, "GOOGLE_TTS_APPLICATION_CREDENTIALS")) {
-    auto channelCreds = grpc::SslCredentials(grpc::SslCredentialsOptions());
-    auto callCreds = grpc::ServiceAccountJWTAccessCredentials(var);
-    auto creds = grpc::CompositeChannelCredentials(channelCreds, callCreds);
-    return grpc::CreateChannel(google_uri, creds);
-  }
-  else {
-    auto creds = grpc::GoogleDefaultCredentials();
-    return grpc::CreateChannel(google_uri, creds);
-  }
-}
-
-static void start_synthesis(const char* text, google_t* g) {
-    try {
-      SynthesizeSpeechRequest request;
-      SynthesizeSpeechResponse response;
-      grpc::ClientContext context;
-      auto input = request.mutable_input();
-      auto voice = request.mutable_voice();
-      auto custom_voice = voice->mutable_custom_voice();
-      auto audio_config = request.mutable_audio_config();
-      /* lock and unlock session */
-      switch_core_session_t *psession = switch_core_session_locate(g->session_id);
-      switch_channel_t *swChannel = switch_core_session_get_channel(psession);
-      switch_core_session_rwunlock(psession);
-      auto channel = create_grpc_channel(swChannel);
-      auto stub = TextToSpeech::NewStub(channel);
-
-      if (strstr(text, "<speak") == text) {
-        input->set_ssml(text);
-      }
-      else {
-        input->set_text(text);
-      }
-      if (g->gender) {
-        if (strcmp(g->gender, "MALE") == 0) {
-          voice->set_ssml_gender(google::cloud::texttospeech::v1::SsmlVoiceGender::MALE);
-        } else if (strcmp(g->gender, "FEMALE") == 0) {
-          voice->set_ssml_gender(google::cloud::texttospeech::v1::SsmlVoiceGender::FEMALE);
-        } else if (strcmp(g->gender, "NEUTRAL") == 0) {
-          voice->set_ssml_gender(google::cloud::texttospeech::v1::SsmlVoiceGender::NEUTRAL);
-        } else {
-          voice->set_ssml_gender(google::cloud::texttospeech::v1::SsmlVoiceGender::SSML_VOICE_GENDER_UNSPECIFIED);
-        }
-      }
-
-      if (g->model) {
-        custom_voice->set_model(g->model);
-        if (strcmp(g->reported_usage, "OFFLINE") == 0) {
-          custom_voice->set_reported_usage(google::cloud::texttospeech::v1::CustomVoiceParams_ReportedUsage::CustomVoiceParams_ReportedUsage_OFFLINE);
-        } else if (strcmp(g->reported_usage, "REALTIME") == 0) {
-          custom_voice->set_reported_usage(google::cloud::texttospeech::v1::CustomVoiceParams_ReportedUsage::CustomVoiceParams_ReportedUsage_REALTIME);
-        } else {
-          custom_voice->set_reported_usage(google::cloud::texttospeech::v1::CustomVoiceParams_ReportedUsage::CustomVoiceParams_ReportedUsage_REPORTED_USAGE_UNSPECIFIED);
-        }
-      } else {
-        voice->set_name(g->voice_name);
-      }
-      voice->set_language_code(g->language);
-      audio_config->set_audio_encoding(AudioEncoding::LINEAR16);
-      audio_config->set_sample_rate_hertz(8000);
-      grpc::Status status = stub->SynthesizeSpeech(&context, request, &response);
-      if (!status.ok()) {
-        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, 
-          "start_synthesis: error synthesizing speech: %s: details: %s\n", 
-          status.error_message().c_str(), status.error_details().c_str()); 
-        return;
-      }
-      g->response_code = 200;
-      auto audioData = response.audio_content();
-      if (g->flushed) return;
-      bool fireEvent = false;
-      CircularBuffer_t *cBuffer = (CircularBuffer_t *) g->circularBuffer;
-      if (g->file) {
-        // google return linear16 wav data that contains 44 bytes wav header, let remove them
-        fwrite(audioData.data() + 44, 1, audioData.size() - 44, g->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
-       */
-      // google return linear16 wav data that contains 44 bytes wav header, let remove them
-      const uint16_t* begin = reinterpret_cast<const uint16_t*>(audioData.data() + 44);
-      const uint16_t* end = reinterpret_cast<const uint16_t*>(audioData.data() + audioData.size());
-
-      /* lock as briefly as possible */
-      switch_mutex_lock(g->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(g->mutex);
-
-      if (0 == g->reads++) {
-        fireEvent = true;
-      }
-
-      if (fireEvent && g->session_id) {
-        auto endTime = std::chrono::high_resolution_clock::now();
-        auto startTime = *static_cast<std::chrono::time_point<std::chrono::high_resolution_clock>*>(g->startTime);
-        auto duration = std::chrono::duration_cast<std::chrono::milliseconds>(endTime - startTime);
-        auto time_to_first_byte_ms = std::to_string(duration.count());
-        switch_core_session_t* session = switch_core_session_locate(g->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) {
-              switch_channel_event_set_data(channel, event);
-              switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "Playback-File-Type", "tts_stream");
-              switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_time_to_first_byte_ms", time_to_first_byte_ms.c_str());
-              if (g->cache_filename) {
-                switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_cache_filename", g->cache_filename);
-              }
-              switch_event_fire(&event);
-            } else {
-              switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "mod_google_tts start_synthesis: failed to create event\n");
-            }
-          }else {
-            switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "mod_google_tts start_synthesis: channel not found\n");
-          }
-        }
-      }
-    } catch (const std::exception& e) {
-        g->response_code = 500;
-        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "mod_google_tts: Exception in start_synthesis %s\n",  e.what());
-    }
-    g->draining = 1;
-}
-
-extern "C" {
-  switch_status_t google_speech_load() {
-    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "google_speech_loading..\n");
-
-    /* create temp folder for cache files */
-    const char* baseDir = std::getenv("JAMBONZ_TMP_CACHE_FOLDER");
-    if (!baseDir) {
-      baseDir = "/tmp/";
-    }
-    if (strcmp(baseDir, "/") == 0) {
-      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "failed to create folder %s\n", baseDir);
-      return SWITCH_STATUS_FALSE;
-    }
-
-    fullDirPath = std::string(baseDir) + "jambonz-tts-cache-files";
-
-    // Create the directory with read, write, and execute permissions for everyone
-    mode_t oldMask = umask(0);
-    int result = mkdir(fullDirPath.c_str(), S_IRWXU | S_IRWXG | S_IRWXO);
-    umask(oldMask);
-    if (result != 0) {
-      if (errno != EEXIST) {
-        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "failed to create folder %s\n", fullDirPath.c_str());
-        fullDirPath = "";
-      }
-      else switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "folder %s already exists\n", fullDirPath.c_str());
-    }
-    else {
-      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "created folder %s\n", fullDirPath.c_str());
-    }
-
-    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "google_speech_loaded..\n");
-
-    return SWITCH_STATUS_SUCCESS;
-  }
-
-    switch_status_t google_speech_open(google_t* google) {
-    return SWITCH_STATUS_SUCCESS;
-  }
-
-  switch_status_t google_speech_feed_tts(google_t* g, char* text, switch_speech_flag_t *flags) {
-    const int MAX_CHARS = 20;
-    char tempText[MAX_CHARS + 4]; // +4 for the ellipsis and null terminator
-
-    if (strlen(text) > MAX_CHARS) {
-        strncpy(tempText, text, MAX_CHARS);
-        strcpy(tempText + MAX_CHARS, "...");
-    } else {
-        strcpy(tempText, text);
-    }
-
-    /* open cache file */
-    if (g->cache_audio && fullDirPath.length() > 0) {
-      switch_uuid_t uuid;
-      char uuid_str[SWITCH_UUID_FORMATTED_LENGTH + 1];
-      char outfile[512] = "";
-      int fd;
-
-      switch_uuid_get(&uuid);
-      switch_uuid_format(uuid_str, &uuid);
-
-      switch_snprintf(outfile, sizeof(outfile), "%s%s%s.r8", fullDirPath.c_str(), SWITCH_PATH_SEPARATOR, uuid_str);
-      g->cache_filename = strdup(outfile);
-      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "writing audio cache file to %s\n", g->cache_filename);
-
-      mode_t oldMask = umask(0);
-      fd = open(outfile, O_WRONLY | O_CREAT | O_TRUNC, S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH | S_IWOTH);
-      umask(oldMask);
-      if (fd == -1 ) {
-        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error opening cache file %s: %s\n", outfile, strerror(errno));
-      }
-      else {
-        g->file = fdopen(fd, "wb");
-        if (!g->file) {
-          close(fd);
-          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error opening cache file %s: %s\n", outfile, strerror(errno));
-        }
-      }
-    }
-
-    if (!g->language) {
-      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "google_speech_feed_tts: no language provided\n");
-      return SWITCH_STATUS_FALSE;
-    }
-
-    if (!g->voice_name && !g->model) {
-      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "google_speech_feed_tts: no voice_name or model provided\n");
-      return SWITCH_STATUS_FALSE;
-    }
-
-    if (g->rate != 8000 /*Hz*/) {
-      int err;
-      g->resampler = speex_resampler_init(1, 8000, g->rate, SWITCH_RESAMPLE_QUALITY, &err);
-      if (0 != err) {
-        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error initializing resampler: %s.\n", speex_resampler_strerror(err));
-        return SWITCH_STATUS_FALSE;
-      }
-    }
-
-    std::chrono::time_point<std::chrono::high_resolution_clock>* ptr = new std::chrono::time_point<std::chrono::high_resolution_clock>(std::chrono::high_resolution_clock::now());
-    g->startTime = ptr;
-
-    g->circularBuffer = (void *) new CircularBuffer_t(BUFFER_SIZE);
-
-    std::thread(start_synthesis, text, g).detach();
-    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "google_speech_feed_tts sent synthesize request\n");
-    return SWITCH_STATUS_SUCCESS;
-  }
-
-  switch_status_t google_speech_read_tts(google_t* g, void *data, size_t *datalen, switch_speech_flag_t *flags) {
-    CircularBuffer_t *cBuffer = (CircularBuffer_t *) g->circularBuffer;
-    std::vector<uint16_t> pcm_data;
-
-    if (g->response_code > 0 && g->response_code != 200) {
-      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "google_speech_read_tts, returning failure\n") ;
-      return SWITCH_STATUS_FALSE;
-    }
-    if (g->flushed) {
-      return SWITCH_STATUS_BREAK;
-    }
-    switch_mutex_lock(g->mutex);
-    size_t bufSize = cBuffer->size();
-    if (cBuffer->empty()) {
-      switch_mutex_unlock(g->mutex);
-      if (g->draining) {
-        return SWITCH_STATUS_BREAK;
-      }
-      /* no audio available yet so send silence */
-      memset(data, 255, *datalen);
-      return SWITCH_STATUS_SUCCESS;
-    }
-    // google returned 8000hz 16 bit data, we have to take enough data based on call sample rate.
-    size_t size = std::min((*datalen/(2 * g->rate / 8000)), bufSize);
-    pcm_data.insert(pcm_data.end(), cBuffer->begin(), cBuffer->begin() + size);
-    cBuffer->erase(cBuffer->begin(), cBuffer->begin() + size);
-    switch_mutex_unlock(g->mutex);
-
-    size_t data_size = pcm_data.size();
-
-    if (g->resampler) {
-      std::vector<int16_t> in(pcm_data.begin(), pcm_data.end());
-
-      std::vector<int16_t> out((*datalen));
-      spx_uint32_t in_len = data_size;
-      spx_uint32_t out_len = out.size();
-
-      speex_resampler_process_interleaved_int(g->resampler, in.data(), &in_len, out.data(), &out_len);
-
-      if (out_len > out.size()) {
-        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_CRIT, "Resampler output exceeded maximum buffer size!\n");
-        return SWITCH_STATUS_FALSE;
-      }
-
-      memcpy(data, out.data(), out_len * sizeof(int16_t));
-      *datalen = out_len * sizeof(int16_t);
-    } else {
-      memcpy(data, pcm_data.data(), data_size * sizeof(int16_t));
-      *datalen = data_size * sizeof(int16_t);
-    }
-
-    return SWITCH_STATUS_SUCCESS;
-  }
-
-  switch_status_t google_speech_flush_tts(google_t* g) {
-    bool download_complete = g->response_code == 200;
-    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "google_speech_flush_tts, download complete? %s\n", download_complete ? "yes" : "no") ;
-
-    CircularBuffer_t *cBuffer = (CircularBuffer_t *) g->circularBuffer;
-    delete cBuffer;
-    g->circularBuffer = nullptr ;
-    delete static_cast<std::chrono::time_point<std::chrono::high_resolution_clock>*>(g->startTime);
-    g->startTime = nullptr;
-
-    g->flushed = 1;
-    if (!download_complete) {
-      if (g->file) {
-        //switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "closing audio cache file %s because download was interrupted\n", g->cache_filename);
-        if (fclose(g->file) != 0) {
-          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "error closing audio cache file\n");
-        }
-        g->file = nullptr ;
-      }
-
-      if (g->cache_filename) {
-        //switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "removing audio cache file %s because download was interrupted\n", g->cache_filename);
-        if (unlink(g->cache_filename) != 0) {
-          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "cleanupConn: error removing audio cache file %s: %d:%s\n", 
-            g->cache_filename, errno, strerror(errno));
-        }
-        free(g->cache_filename);
-        g->cache_filename = nullptr ;
-      }
-    }
-    if (g->session_id) {
-      switch_core_session_t* session = switch_core_session_locate(g->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) {
-            switch_channel_event_set_data(channel, event);
-            switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "Playback-File-Type", "tts_stream");
-            switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_google_response_code", std::to_string(g->response_code).c_str());
-            if (g->cache_filename && download_complete) {
-              switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_cache_filename", g->cache_filename);
-            }
-            if (!download_complete && g->err_msg) {
-              switch_event_add_header_string(event, SWITCH_STACK_BOTTOM, "variable_tts_error", g->err_msg);
-            }
-            switch_event_fire(&event);
-          }
-          else {
-            switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: failed to create event\n");
-          }
-        }
-        else {
-          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "write_cb: channel not found\n");
-        }
-      }
-    }
-
-    return SWITCH_STATUS_SUCCESS;
-  }
-
-  switch_status_t google_speech_close(google_t* g) {
-    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "google_speech_close\n") ;
-    if (g->resampler) {
-      speex_resampler_destroy(g->resampler);
-    }
-
-    g->resampler = NULL;
-    return SWITCH_STATUS_SUCCESS;
-  }
-
-  switch_status_t google_speech_unload() {
-    return SWITCH_STATUS_SUCCESS;
-  }
-
-} 

mod_google_tts/google_glue.h

@@ -1,12 +0,0 @@
-#ifndef __GOOGLE_TTS_GLUE_H__
-#define __GOOGLE_TTS_GLUE_H__
-
-switch_status_t google_speech_load();
-switch_status_t google_speech_open(google_t* google);
-switch_status_t google_speech_feed_tts(google_t* google, char* text, switch_speech_flag_t *flags);
-switch_status_t google_speech_read_tts(google_t* google, void *data, size_t *datalen, switch_speech_flag_t *flags);
-switch_status_t google_speech_flush_tts(google_t* google);
-switch_status_t google_speech_close(google_t* google);
-switch_status_t google_speech_unload();
-
-#endif

mod_google_tts/mod_google_tts.c

@@ -1,154 +0,0 @@
-#include "mod_google_tts.h"
-#include "google_glue.h"
-
-SWITCH_MODULE_LOAD_FUNCTION(mod_google_tts_load);
-SWITCH_MODULE_SHUTDOWN_FUNCTION(mod_google_tts_shutdown);
-SWITCH_MODULE_DEFINITION(mod_google_tts, mod_google_tts_load, mod_google_tts_shutdown, NULL);
-
-static void clearGoogle(google_t* g, int freeAll) {
-  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "clearGoogle\n");
-
-  if (g->cache_filename) free(g->cache_filename);
-  if (g->model) free(g->model);
-  if (g->reported_usage) free(g->reported_usage);
-  if (g->gender) free(g->gender);
-  if (g->language) free(g->language);
-  if (g->err_msg) free(g->err_msg);
-  
-  g->cache_filename = NULL;
-  g->model = NULL;
-  g->reported_usage = NULL;
-  g->gender = NULL;
-  g->language = NULL;
-  g->err_msg = NULL;
-
-
-  if (freeAll) {
-    if (g->voice_name) free(g->voice_name);
-    if (g->session_id) free(g->session_id);
-    g->voice_name = NULL;
-    g->session_id = NULL;
-  }
-
-}
-
-static google_t * createOrRetrievePrivateData(switch_speech_handle_t *sh) {
-  google_t *g = (google_t *) sh->private_info;  
-  if (!g) {
-    g = switch_core_alloc(sh->memory_pool, sizeof(*g));
-  	sh->private_info = g;
-    memset(g, 0, sizeof(*g));
-    switch_mutex_init(&g->mutex, SWITCH_MUTEX_NESTED, sh->memory_pool);
-    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "allocated google_t\n");
-  }
-  return g;
-}
-
-switch_status_t g_speech_open(switch_speech_handle_t *sh, const char *voice_name, int rate, int channels, switch_speech_flag_t *flags)
-{
-  google_t *g = createOrRetrievePrivateData(sh);
-  g->voice_name = strdup(voice_name);
-  g->rate = rate;
-  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "g_speech_open voice: %s, rate %d, channels %d\n", voice_name, rate, channels);
-  return google_speech_open(g);
-}
-
-static switch_status_t g_speech_close(switch_speech_handle_t *sh, switch_speech_flag_t *flags)
-{
-  switch_status_t rc;
-  google_t *g = createOrRetrievePrivateData(sh);
-  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "g_speech_close\n");
-
-  switch_mutex_destroy(g->mutex);
-
-  rc = google_speech_close(g);
-  clearGoogle(g, 1);
-  return rc;
-}
-
-/**
- * Freeswitch will call this function to feed us text to speak
- */
-static switch_status_t g_speech_feed_tts(switch_speech_handle_t *sh, char *text, switch_speech_flag_t *flags)
-{
-  google_t *g = createOrRetrievePrivateData(sh);
-  g->draining = 0;
-  g->reads = 0;
-  g->flushed = 0;
-
-  return google_speech_feed_tts(g, text, flags);
-}
-
-/**
- * Freeswitch calls periodically to get some rendered audio in L16 format. We can provide up to 8k of audio at a time.
- */
-static switch_status_t g_speech_read_tts(switch_speech_handle_t *sh, void *data, size_t *datalen, switch_speech_flag_t *flags)
-{
-  google_t *g = createOrRetrievePrivateData(sh);
-  return google_speech_read_tts(g, data, datalen, flags);
-}
-
-/**
- * This is called at the end, not sure exactly what we need to do here..
- */
-static void g_speech_flush_tts(switch_speech_handle_t *sh)
-{
-  google_t *g = createOrRetrievePrivateData(sh);
-  google_speech_flush_tts(g);
-
-  clearGoogle(g, 0);
-}
-
-static void g_text_param_tts(switch_speech_handle_t *sh, char *param, const char *val)
-{
-  google_t *g = createOrRetrievePrivateData(sh);
-  switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "g_text_param_tts: %s=%s\n", param, val);
-  if (0 == strcmp(param, "voice")) {
-    if (g->voice_name) free(g->voice_name);
-    g->voice_name = strdup(val);
-  } else if (0 == strcmp(param, "model")) {
-    if (g->model) free(g->model);
-    g->model = strdup(val);
-  } else if (0 == strcmp(param, "reported_usage")) {
-    if (g->reported_usage) free(g->reported_usage);
-    g->reported_usage = strdup(val);
-  } else if (0 == strcmp(param, "language")) {
-    if (g->language) free(g->language);
-    g->language = strdup(val);
-  } else if (0 == strcmp(param, "session-uuid")) {
-    if (g->session_id) free(g->session_id);
-    g->session_id = strdup(val);
-  } else if (0 == strcmp(param, "write_cache_file") && switch_true(val)) {
-    g->cache_audio = 1;
-  }
-}
-
-static void g_numeric_param_tts(switch_speech_handle_t *sh, char *param, int val)
-{
-}
-static void g_float_param_tts(switch_speech_handle_t *sh, char *param, double val)
-{
-}
-
-SWITCH_MODULE_LOAD_FUNCTION(mod_google_tts_load)
-{
-  switch_speech_interface_t *speech_interface;
-
-  *module_interface = switch_loadable_module_create_module_interface(pool, modname);
-  speech_interface = switch_loadable_module_create_interface(*module_interface, SWITCH_SPEECH_INTERFACE);
-  speech_interface->interface_name = "google";
-  speech_interface->speech_open = g_speech_open;
-  speech_interface->speech_close = g_speech_close;
-  speech_interface->speech_feed_tts = g_speech_feed_tts;
-  speech_interface->speech_read_tts = g_speech_read_tts;
-	speech_interface->speech_flush_tts = g_speech_flush_tts;
-	speech_interface->speech_text_param_tts = g_text_param_tts;
-	speech_interface->speech_numeric_param_tts = g_numeric_param_tts;
-	speech_interface->speech_float_param_tts = g_float_param_tts;
-  return google_speech_load();
-}
-
-SWITCH_MODULE_SHUTDOWN_FUNCTION(mod_google_tts_shutdown)
-{
-  return google_speech_unload();
-}

mod_playht_tts/playht_glue.cpp

@@ -436,7 +436,7 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
   CircularBuffer_t *cBuffer = (CircularBuffer_t *) p->circularBuffer;
   std::vector<uint16_t> pcm_data;
   
-  if (conn->flushed) {
+  if (conn->flushed || cBuffer == nullptr) {
     /* this will abort the transfer */
     return 0;
   }
@@ -648,7 +648,7 @@ extern "C" {
       return SWITCH_STATUS_FALSE;
     }
 
-    fullDirPath = std::string(baseDir) + "jambonz-tts-cache-files";
+    fullDirPath = std::string(baseDir) + "tts-cache-files";
 
     // Create the directory with read, write, and execute permissions for everyone
     mode_t oldMask = umask(0);
@@ -838,17 +838,9 @@ extern "C" {
 
     p->circularBuffer = (void *) new CircularBuffer_t(8192);
 
-    if (p->session_id) {
-      int err;
-      switch_codec_implementation_t read_impl;
-      switch_core_session_t *psession = switch_core_session_locate(p->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");
-        return SWITCH_STATUS_FALSE;
-      }
+    if (mpg123_param(mh, MPG123_FORCE_RATE, p->rate /*Hz*/, 0) != MPG123_OK) {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error mpg123_param!\n");
+      return SWITCH_STATUS_FALSE;
     }
 
     std::ostringstream api_key_stream;

mod_rimelabs_tts/rimelabs_glue.cpp

@@ -386,7 +386,7 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
   auto d = conn->rimelabs;
   CircularBuffer_t *cBuffer = (CircularBuffer_t *) d->circularBuffer;
   
-  if (conn->flushed) {
+  if (conn->flushed || cBuffer == nullptr) {
     /* this will abort the transfer */
     return 0;
   }
@@ -394,6 +394,8 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
   std::unique_ptr<uint8_t[]> combinedData;
 
   if (conn->has_last_byte) {
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "rimelabs write_cb:  processing last byte from previous read\n");
+
     conn->has_last_byte = false;  // We'll handle the last_byte now, so toggle the flag off
 
     // Allocate memory for the new data array
@@ -415,6 +417,8 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
 
   // If we now have an odd total, save the last byte for next time
   if ((total_bytes_to_process % sizeof(int16_t)) != 0) {
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "rimelabs write_cb: got odd number of bytes %d\n", total_bytes_to_process);
+
     conn->last_byte = data[total_bytes_to_process - 1];
     conn->has_last_byte = true;
     total_bytes_to_process--;
@@ -626,7 +630,7 @@ extern "C" {
       return SWITCH_STATUS_FALSE;
     }
 
-    fullDirPath = std::string(baseDir) + "jambonz-tts-cache-files";
+    fullDirPath = std::string(baseDir) + "tts-cache-files";
 
     // Create the directory with read, write, and execute permissions for everyone
     mode_t oldMask = umask(0);
@@ -758,6 +762,8 @@ extern "C" {
     conn->file = d->file;
     conn->body = json;
     conn->flushed = false;
+    conn->has_last_byte = false;
+    conn->last_byte = 0;
     
 
     d->circularBuffer = (void *) new CircularBuffer_t(BUFFER_GROW_SIZE);

mod_whisper_tts/whisper_glue.cpp

@@ -436,7 +436,7 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
   CircularBuffer_t *cBuffer = (CircularBuffer_t *) w->circularBuffer;
   std::vector<uint16_t> pcm_data;
   
-  if (conn->flushed) {
+  if (conn->flushed || cBuffer == nullptr) {
     /* this will abort the transfer */
     return 0;
   }
@@ -671,7 +671,7 @@ extern "C" {
       return SWITCH_STATUS_FALSE;
     }
 
-    fullDirPath = std::string(baseDir) + "jambonz-tts-cache-files";
+    fullDirPath = std::string(baseDir) + "tts-cache-files";
 
     // Create the directory with read, write, and execute permissions for everyone
     mode_t oldMask = umask(0);