Tts/elevenlabs streaming (#629)

* update to fsmrf with fix

* changes to support elevenlabs tts streaming

* say: add vendor data to span

* bug: tts spans must include cached property

* add env for JAMBONES_USE_FREESWITCH_TIMER_FD

* fix bug in prev commit

* wip

* linting

* wip - caching files generating by streaming tts

* wip caching

* cleanup some logs

* handle tts streaming failure, write alert

* update node version dependency

* set timerfd on outbound call scenarios

* default model to nova-2-phonecall when using deepgram

---------

Co-authored-by: Dave Horton <daveh@beachdognet.com>

lib/config.js

@@ -132,6 +132,8 @@ const JAMBONES_DISABLE_DIRECT_P2P_CALL = process.env.JAMBONES_DISABLE_DIRECT_P2P
 
 const JAMBONES_EAGERLY_PRE_CACHE_AUDIO = process.env.JAMBONES_EAGERLY_PRE_CACHE_AUDIO;
 
+const JAMBONES_USE_FREESWITCH_TIMER_FD = process.env.JAMBONES_USE_FREESWITCH_TIMER_FD;
+
 module.exports = {
   JAMBONES_MYSQL_HOST,
   JAMBONES_MYSQL_USER,
@@ -213,5 +215,6 @@ module.exports = {
   JAMBONZ_RECORD_WS_USERNAME,
   JAMBONZ_RECORD_WS_PASSWORD,
   JAMBONZ_DISABLE_DIAL_PAI_HEADER,
-  JAMBONES_DISABLE_DIRECT_P2P_CALL
+  JAMBONES_DISABLE_DIRECT_P2P_CALL,
+  JAMBONES_USE_FREESWITCH_TIMER_FD
 };

lib/session/call-session.js

@@ -21,6 +21,7 @@ const {
   JAMBONES_INJECT_CONTENT,
   JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
   AWS_REGION,
+  JAMBONES_USE_FREESWITCH_TIMER_FD
 } = require('../config');
 const BackgroundTaskManager = require('../utils/background-task-manager');
 const BADPRECONDITIONS = 'preconditions not met';
@@ -689,7 +690,7 @@ class CallSession extends Emitter {
         (type === 'stt' && credential.use_for_stt)
       )) {
         this.logger.info(
-          `Speech vendor: ${credential.vendor} ${credential.label ? `, label: ${credential.label}` : ''} selected`);
+          `${type}: ${credential.vendor} ${credential.label ? `, label: ${credential.label}` : ''}  `);
         if ('google' === vendor) {
           if (type === 'tts' && !credential.tts_tested_ok ||
             type === 'stt' && !credential.stt_tested_ok) {
@@ -2033,8 +2034,12 @@ Duration=${duration} `
   }
 
   _configMsEndpoint() {
-    if (this.onHoldMusic) {
-      this.ep.set({hold_music: `shout://${this.onHoldMusic.replace(/^https?:\/\//, '')}`});
+    const opts = {
+      ...(this.onHoldMusic && {holdMusic: `shout://${this.onHoldMusic.replace(/^https?:\/\//, '')}`}),
+      ...(JAMBONES_USE_FREESWITCH_TIMER_FD && {timer_name: 'timerfd'})
+    };
+    if (Object.keys(opts).length > 0) {
+      this.ep.set(opts);
     }
   }
 

lib/tasks/listen.js

@@ -123,8 +123,6 @@ class TaskListen extends Task {
       ci,
       this.metadata);
     if (this.hook.auth) {
-      this.logger.debug({username: this.hook.auth.username, password: this.hook.auth.password},
-        'TaskListen:_startListening basic auth');
       await this.ep.set({
         'MOD_AUDIO_BASIC_AUTH_USERNAME': this.hook.auth.username,
         'MOD_AUDIO_BASIC_AUTH_PASSWORD': this.hook.auth.password

lib/tasks/say.js

@@ -23,6 +23,12 @@ const breakLengthyTextIfNeeded = (logger,  text) => {
   }
 };
 
+const parseTextFromSayString = (text) => {
+  const closingBraceIndex = text.indexOf('}');
+  if (closingBraceIndex === -1) return text;
+  return text.slice(closingBraceIndex + 1);
+};
+
 class TaskSay extends Task {
   constructor(logger, opts, parentTask) {
     super(logger, opts);
@@ -60,7 +66,7 @@ class TaskSay extends Task {
   }
 
   async _synthesizeWithSpecificVendor(cs, ep, {vendor, language, voice, label, preCache = false}) {
-    const {srf} = cs;
+    const {srf, accountSid:account_sid} = cs;
     const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, srf);
     const {writeAlerts, AlertType, stats} = srf.locals;
     const {synthAudio} = srf.locals.dbHelpers;
@@ -97,11 +103,17 @@ class TaskSay extends Task {
       voice = this.options.voice_id || voice;
     }
 
+    this.ep.set({
+      tts_engine: vendor,
+      tts_voice: voice,
+      cache_speech_handles: 1,
+    }).catch((err) => this.logger.info({err}, 'Error setting tts_engine on endpoint'));
+
     if (!preCache) this.logger.info({vendor, language, voice, model}, 'TaskSay:exec');
     try {
       if (!credentials) {
         writeAlerts({
-          account_sid: cs.accountSid,
+          account_sid,
           alert_type: AlertType.TTS_NOT_PROVISIONED,
           vendor
         }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
@@ -120,18 +132,17 @@ class TaskSay extends Task {
         if (text.startsWith('silence_stream://')) return text;
 
         /* otel: trace time for tts */
-        let otelSpan;
         if (!preCache)  {
           const {span} = this.startChildSpan('tts-generation', {
             'tts.vendor': vendor,
             'tts.language': language,
             'tts.voice': voice
           });
-          otelSpan = span;
+          this.otelSpan = span;
         }
         try {
           const {filePath, servedFromCache, rtt} = await synthAudio(stats, {
-            account_sid: cs.accountSid,
+            account_sid,
             text,
             vendor,
             language,
@@ -141,17 +152,21 @@ class TaskSay extends Task {
             salt,
             credentials,
             options: this.options,
-            disableTtsCache : this.disableTtsCache
+            disableTtsCache : this.disableTtsCache,
+            preCache
           });
+          if (!filePath.startsWith('say:')) {
             this.logger.debug(`file ${filePath}, served from cache ${servedFromCache}`);
             if (filePath) cs.trackTmpFile(filePath);
+            if (this.otelSpan) {
+              this.otelSpan.setAttributes({'tts.cached': servedFromCache});
+              this.otelSpan.end();
+              this.otelSpan = null;
+            }
             if (!servedFromCache && !lastUpdated) {
               lastUpdated = true;
-            updateSpeechCredentialLastUsed(credentials.speech_credential_sid)
-              .catch(() => {/*already logged error */});
+              updateSpeechCredentialLastUsed(credentials.speech_credential_sid).catch(() => {/* logged error */});
             }
-          if (otelSpan) otelSpan.setAttributes({'tts.cached': servedFromCache});
-          if (otelSpan) otelSpan.end();
             if (!servedFromCache && rtt && !preCache) {
               this.notifyStatus({
                 event: 'synthesized-audio',
@@ -161,10 +176,16 @@ class TaskSay extends Task {
                 elapsedTime: rtt
               });
             }
+          }
+          else {
+            this.logger.debug('a streaming tts api will be used');
+            const modifiedPath = filePath.replace('say:{', `say:{session-uuid=${this.ep.uuid},`);
+            return modifiedPath;
+          }
           return filePath;
         } catch (err) {
           this.logger.info({err}, 'Error synthesizing tts');
-          if (otelSpan) otelSpan.end();
+          if (this.otelSpan) this.otelSpan.end();
           writeAlerts({
             account_sid: cs.accountSid,
             alert_type: AlertType.TTS_FAILURE,
@@ -186,6 +207,11 @@ class TaskSay extends Task {
   }
 
   async exec(cs, {ep}) {
+    const {srf, accountSid:account_sid} = cs;
+    const {writeAlerts, AlertType} = srf.locals;
+    const {addFileToCache} = srf.locals.dbHelpers;
+    const engine = this.synthesizer.engine || 'standard';
+
     await super.exec(cs);
     this.ep = ep;
 
@@ -243,7 +269,39 @@ class TaskSay extends Task {
           await this.playToConfMember(this.ep, memberId, confName, confUuid, filepath[segment]);
         }
         else {
-          this.logger.debug(`Say:exec sending command to play file ${filepath[segment]}`);
+          this.logger.debug(`Say:exec sending  ${filepath[segment].substring(0, 64)}`);
+          this.ep.once('playback-start', (evt) => {
+            this.logger.debug({evt}, 'got playback-start');
+            if (this.otelSpan) {
+              this.logger.debug({evt}, 'got playback-start');
+              this._addStreamingTtsAttributes(this.otelSpan, evt);
+              this.otelSpan.end();
+              this.otelSpan = null;
+              if (evt.variable_tts_cache_filename) cs.trackTmpFile(evt.variable_tts_cache_filename);
+            }
+          });
+          this.ep.once('playback-stop', (evt) => {
+            this.logger.debug({evt}, 'got playback-stop');
+            if (evt.variable_tts_error) {
+              writeAlerts({
+                account_sid,
+                alert_type: AlertType.TTS_FAILURE,
+                vendor,
+                detail: evt.variable_tts_error
+              }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
+            }
+            if (evt.variable_tts_cache_filename) {
+              const text = parseTextFromSayString(this.text[segment]);
+              addFileToCache(evt.variable_tts_cache_filename, {
+                account_sid,
+                vendor,
+                language,
+                voice,
+                engine,
+                text
+              }).catch((err) => this.logger.info({err}, 'Error adding file to cache'));
+            }
+          });
           await ep.play(filepath[segment]);
           this.logger.debug(`Say:exec completed play file ${filepath[segment]}`);
         }
@@ -265,8 +323,30 @@ class TaskSay extends Task {
         this.notifyStatus({event: 'kill-playback'});
         this.ep.api('uuid_break', this.ep.uuid);
       }
-    }
+      this.ep.removeAllListeners('playback-start');
+      this.ep.removeAllListeners('playback-stop');
     }
   }
 
+  _addStreamingTtsAttributes(span, evt) {
+    const attrs = {'tts.cached': false};
+    for (const [key, value] of Object.entries(evt)) {
+      if (key.startsWith('variable_tts_')) {
+        let newKey = key.substring('variable_tts_'.length)
+          .replace('elevenlabs_', 'elevenlabs.');
+        if (spanMapping[newKey]) newKey = spanMapping[newKey];
+        attrs[newKey] = value;
+      }
+    }
+    span.setAttributes(attrs);
+  }
+}
+
+const spanMapping = {
+  'elevenlabs.reported_latency_ms': 'elevenlabs.latency_ms',
+  'elevenlabs.request_id': 'elevenlabs.req_id',
+  'elevenlabs.history_item_id': 'elevenlabs.item_id',
+  'elevenlabs.optimize_streaming_latency': 'elevenlabs.optimization',
+};
+
 module.exports = TaskSay;

lib/tasks/task.js

@@ -177,7 +177,6 @@ class Task extends Emitter {
           const makeTask = require('./make_task');
           const tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
           if (tasks && tasks.length > 0) {
-            this.logger.info({tasks: tasks}, `${this.name} replacing application with ${tasks.length} tasks`);
             this.callSession.replaceApplication(tasks);
           }
         }

lib/utils/install-srf-locals.js

@@ -176,6 +176,7 @@ function installSrfLocals(srf, logger) {
   const registrar = new Registrar(logger, client);
   const {
     synthAudio,
+    addFileToCache,
     getNuanceAccessToken,
     getIbmAccessToken,
   } = require('@jambonz/speech-utils')({}, logger);
@@ -215,6 +216,7 @@ function installSrfLocals(srf, logger) {
       listCalls,
       deleteCall,
       synthAudio,
+      addFileToCache,
       createHash,
       retrieveHash,
       deleteKey,

lib/utils/place-outdial.js

@@ -16,6 +16,9 @@ const uuidv4 = require('uuid-random');
 const HttpRequestor = require('./http-requestor');
 const WsRequestor = require('./ws-requestor');
 const {makeOpusFirst} = require('./sdp-utils');
+const {
+  JAMBONES_USE_FREESWITCH_TIMER_FD
+} = require('../config');
 
 class SingleDialer extends Emitter {
   constructor({logger, sbcAddress, target, opts, application, callInfo, accountInfo, rootSpan, startSpan, dialTask,
@@ -324,8 +327,12 @@ class SingleDialer extends Emitter {
   }
 
   _configMsEndpoint() {
-    if (this.onHoldMusic) {
-      this.ep.set({hold_music: `shout://${this.onHoldMusic.replace(/^https?:\/\//, '')}`});
+    const opts = {
+      ...(this.onHoldMusic && {holdMusic: `shout://${this.onHoldMusic.replace(/^https?:\/\//, '')}`}),
+      ...(JAMBONES_USE_FREESWITCH_TIMER_FD && {timer_name: 'timerfd'})
+    };
+    if (Object.keys(opts).length > 0) {
+      this.ep.set(opts);
     }
   }
 

lib/utils/transcription-utils.js

@@ -107,30 +107,30 @@ const optimalDeepramModels = {
   'zh-CN':['base', 'base'],
   'zh-TW': ['base', 'base'],
   da: ['enhanced', 'enhanced'],
-  en: ['nova-2-conversationalai', 'nova-2'],
-  'en-US': ['nova-2-conversationalai', 'nova-2'],
-  'en-AU': ['nova-2-conversationalai', 'nova-2'],
-  'en-GB': ['nova-2-conversationalai', 'nova-2'],
-  'en-IN': ['nova-2-conversationalai', 'nova-2'],
-  'en-NZ': ['nova-2-conversationalai', 'nova-2'],
-  nl: ['nova-2-conversationalai', 'nova-2'],
-  fr: ['nova-2-conversationalai', 'nova-2'],
-  'fr-CA': ['nova-2-conversationalai', 'nova-2'],
-  de: ['nova-2-conversationalai', 'nova-2'],
-  hi: ['nova-2-conversationalai', 'nova-2'],
-  'hi-Latn': ['nova-2-conversationalai', 'nova-2'],
+  en: ['nova-2-phonecall', 'nova-2'],
+  'en-US': ['nova-2-phonecall', 'nova-2'],
+  'en-AU': ['nova-2-phonecall', 'nova-2'],
+  'en-GB': ['nova-2-phonecall', 'nova-2'],
+  'en-IN': ['nova-2-phonecall', 'nova-2'],
+  'en-NZ': ['nova-2-phonecall', 'nova-2'],
+  nl: ['nova-2-phonecall', 'nova-2'],
+  fr: ['nova-2-phonecall', 'nova-2'],
+  'fr-CA': ['nova-2-phonecall', 'nova-2'],
+  de: ['nova-2-phonecall', 'nova-2'],
+  hi: ['nova-2-phonecall', 'nova-2'],
+  'hi-Latn': ['nova-2-phonecall', 'nova-2'],
   id: ['base', 'base'],
   it: ['enhanced', 'enhanced'],
   ja: ['enhanced', 'enhanced'],
   ko: ['enhanced', 'enhanced'],
   no: ['enhanced', 'enhanced'],
   pl: ['enhanced', 'enhanced'],
-  pt: ['nova-2-conversationalai', 'nova-2'],
-  'pt-BR': ['nova-2-conversationalai', 'nova-2'],
+  pt: ['nova-2-phonecall', 'nova-2'],
+  'pt-BR': ['nova-2-phonecall', 'nova-2'],
   'pt-PT': ['base', 'base'],
   ru: ['base', 'base'],
-  es: ['nova-2-conversationalai', 'nova-2'],
-  'es-419': ['nova-2-conversationalai', 'nova-2'],
+  es: ['nova-2-phonecall', 'nova-2'],
+  'es-419': ['nova-2-phonecall', 'nova-2'],
   'es-LATAM': ['enhanced', 'enhanced'],
   sv: ['enhanced', 'enhanced'],
   ta: ['enhanced', 'enhanced'],

package.json

@@ -3,7 +3,7 @@
   "version": "0.8.5",
   "main": "app.js",
   "engines": {
-    "node": ">= 10.16.0"
+    "node": ">= 18.x"
   },
   "keywords": [
     "sip",
@@ -31,7 +31,7 @@
     "@jambonz/http-health-check": "^0.0.1",
     "@jambonz/mw-registrar": "^0.2.4",
     "@jambonz/realtimedb-helpers": "^0.8.7",
-    "@jambonz/speech-utils": "^0.0.33",
+    "@jambonz/speech-utils": "^0.0.38",
     "@jambonz/stats-collector": "^0.1.9",
     "@jambonz/time-series": "^0.2.8",
     "@jambonz/verb-specifications": "^0.0.50",
@@ -47,7 +47,7 @@
     "bent": "^7.3.12",
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
-    "drachtio-fsmrf": "^3.0.33",
+    "drachtio-fsmrf": "^3.0.37",
     "drachtio-srf": "^4.5.31",
     "express": "^4.18.2",
     "express-validator": "^7.0.1",