initial changes for major tts revamp

lib/session/call-session.js

@@ -1,4 +1,5 @@
 const Emitter = require('events');
+const fs = require('fs');
 const {CallDirection, TaskPreconditions, CallStatus, TaskName} = require('../utils/constants');
 const moment = require('moment');
 const assert = require('assert');
@@ -40,6 +41,8 @@ class CallSession extends Emitter {
     this.stackIdx = 0;
     this.callGone = false;
 
+    this.tmpFiles = new Set();
+
     sessionTracker.add(this.callSid, this);
   }
 
@@ -99,6 +102,12 @@ class CallSession extends Emitter {
   get speechSynthesisVoice() {
     return this.application.speech_synthesis_voice;
   }
+  /**
+   * default language to use for speech synthesis if not provided in the app
+   */
+  get speechSynthesisLanguage() {
+    return this.application.speech_synthesis_language;
+  }
 
   /**
    * default vendor to use for speech recognition if not provided in the app
@@ -172,6 +181,10 @@ class CallSession extends Emitter {
     sessionTracker.remove(this.callSid);
   }
 
+  trackTmpFile(path) {
+    this.tmpFiles.add(path);
+  }
+
   normalizeUrl(url, method, auth) {
     const hook = {
       url,
@@ -492,6 +505,17 @@ class CallSession extends Emitter {
         this.logger.error(err, 'CallSession:_clearResources error');
       }
     }
+
+    // remove any temporary tts files that were created (audio is still cached in redis)
+    for (const path of this.tmpFiles) {
+      fs.unlink(path, (err) => {
+        if (err) {
+          return this.logger.error(err, `CallSession:_clearResources Error deleting tmp file ${path}`);
+        }
+        this.logger.debug(`CallSession:_clearResources successfully deleted ${path}`);
+      });
+    }
+    this.tmpFiles.clear();
   }
 
   /**

lib/tasks/make_task.js

@@ -25,6 +25,10 @@ function makeTask(logger, obj, parent) {
       const TaskHangup = require('./hangup');
       return new TaskHangup(logger, data, parent);
     case TaskName.Say:
+      if (data.vendor === 'google' && !data.language) {
+        const TaskSayLegacy = require('./say-legacy');
+        return new TaskSayLegacy(logger, data, parent);
+      }
       const TaskSay = require('./say');
       return new TaskSay(logger, data, parent);
     case TaskName.Play:

lib/tasks/say-legacy.js

@@ -0,0 +1,53 @@
+const Task = require('./task');
+const {TaskName, TaskPreconditions} = require('../utils/constants');
+
+class TaskSayLegacy extends Task {
+  constructor(logger, opts, parentTask) {
+    super(logger, opts);
+    this.preconditions = TaskPreconditions.Endpoint;
+
+    this.text = this.data.text;
+    this.loop = this.data.loop || 1;
+    this.earlyMedia = this.data.earlyMedia === true || (parentTask && parentTask.earlyMedia);
+    if (this.data.synthesizer) {
+      this.voice = this.data.synthesizer.voice;
+      switch (this.data.synthesizer.vendor) {
+        case 'google':
+          this.ttsEngine = 'google_tts';
+          break;
+        default:
+          throw new Error(`unsupported tts vendor ${this.data.synthesizer.vendor}`);
+      }
+    }
+  }
+
+  get name() { return TaskName.SayLegacy; }
+
+  async exec(cs, ep) {
+    super.exec(cs);
+    this.ep = ep;
+    try {
+      while (!this.killed && this.loop--) {
+        this.logger.debug(`TaskSayLegacy: remaining loops ${this.loop}`);
+        await ep.speak({
+          ttsEngine: 'google_tts',
+          voice: this.voice || this.callSession.speechSynthesisVoice,
+          text: this.text
+        });
+      }
+    } catch (err) {
+      this.logger.info(err, 'TaskSayLegacy:exec error');
+    }
+    this.emit('playDone');
+  }
+
+  async kill() {
+    super.kill();
+    if (this.ep.connected) {
+      this.logger.debug('TaskSayLegacy:kill - killing audio');
+      await this.ep.api('uuid_break', this.ep.uuid).catch((err) => this.logger.info(err, 'Error killing audio'));
+    }
+  }
+}
+
+module.exports = TaskSayLegacy;

lib/tasks/say.js

@@ -9,31 +9,32 @@ class TaskSay extends Task {
     this.text = this.data.text;
     this.loop = this.data.loop || 1;
     this.earlyMedia = this.data.earlyMedia === true || (parentTask && parentTask.earlyMedia);
-    if (this.data.synthesizer) {
-      this.voice = this.data.synthesizer.voice;
-      switch (this.data.synthesizer.vendor) {
-        case 'google':
-          this.ttsEngine = 'google_tts';
-          break;
-        default:
-          throw new Error(`unsupported tts vendor ${this.data.synthesizer.vendor}`);
-      }
-    }
+    this.synthesizer = this.data.synthesizer || {};
   }
 
   get name() { return TaskName.Say; }
 
   async exec(cs, ep) {
+    const {srf} = cs;
+    const {synthAudio} = srf.locals.dbHelpers;
     super.exec(cs);
     this.ep = ep;
     try {
+      let filepath;
+      const opts = Object.assign({
+        text: this.text,
+        vendor: cs.speechSynthesisVendor,
+        language: cs.speechSynthesisLanguage,
+        voice: cs.speechSynthesisVoice
+      }, this.synthesizer);
+
       while (!this.killed && this.loop--) {
         this.logger.debug(`TaskSay: remaining loops ${this.loop}`);
-        await ep.speak({
-          ttsEngine: 'google_tts',
-          voice: this.voice || this.callSession.speechSynthesisVoice,
-          text: this.text
-        });
+        if (!filepath) {
+          filepath = await synthAudio(opts);
+          cs.trackTmpFile(filepath);
+        }
+        await ep.play(filepath);
       }
     } catch (err) {
       this.logger.info(err, 'TaskSay:exec error');

lib/tasks/specs.json

@@ -187,12 +187,14 @@
     "properties": {
       "vendor": {
         "type": "string",
-        "enum": ["google"]
+        "enum": ["google", "aws", "polly"]
       },
+      "language": "string",
       "voice": "string"
     },
     "required": [
-      "vendor"
+      "vendor",
+      "language"
     ]
   },
   "recognizer": {

lib/utils/constants.json

@@ -12,6 +12,7 @@
     "SipNotify": "sip:notify",
     "SipRedirect": "sip:redirect",
     "Say": "say",
+    "SayLegacy": "say:legacy",
     "Tag": "tag",
     "Transcribe": "transcribe"
   },

lib/utils/install-srf-locals.js

@@ -105,7 +105,8 @@ function installSrfLocals(srf, logger) {
     updateCallStatus,
     retrieveCall,
     listCalls,
-    deleteCall
+    deleteCall,
+    synthAudio
   } = require('jambonz-realtimedb-helpers')({
     host: process.env.JAMBONES_REDIS_HOST,
     port: process.env.JAMBONES_REDIS_PORT || 6379
@@ -119,7 +120,8 @@ function installSrfLocals(srf, logger) {
       updateCallStatus,
       retrieveCall,
       listCalls,
-      deleteCall
+      deleteCall,
+      synthAudio
     },
     parentLogger: logger,
     ipv4: localIp,