0.0.20

support self hosted microsoft

config/test.json

@@ -8,6 +8,8 @@
   },
   "redis-auth": {
     "host": "127.0.0.1",
-    "port": 3380
+    "port": 3380,
+    "username": "daveh",
+    "password": "foobarbazzle"
   }
 }

index.js

@@ -1,15 +1,23 @@
 const {noopLogger} = require('./lib/utils');
-const promisify = require('@jambonz/promisify-redis');
-const redis = promisify(require('redis'));
+const Redis = require('ioredis');
 
 module.exports = (opts, logger) => {
-  const {host = '127.0.0.1', port = 6379, tls = false} = opts;
   logger = logger || noopLogger;
+  const connectionOpts = {...opts};
+  // Support legacy app
+  if (process.env.JAMBONES_REDIS_USERNAME && process.env.JAMBONES_REDIS_PASSWORD) {
+    if (Array.isArray(connectionOpts)) {
+      for (const o of opts) {
+        o.username = process.env.JAMBONES_REDIS_USERNAME;
+        o.password = process.env.JAMBONES_REDIS_PASSWORD;
+      }
+    } else {
+      connectionOpts.username = process.env.JAMBONES_REDIS_USERNAME;
+      connectionOpts.password = process.env.JAMBONES_REDIS_PASSWORD;
+    }
+  }
 
-  const url = process.env.JAMBONES_REDIS_USERNAME && process.env.JAMBONES_REDIS_PASSWORD ?
-    `${process.env.JAMBONES_REDIS_USERNAME}:${process.env.JAMBONES_REDIS_PASSWORD}@${host}:${port}` :
-    `${host}:${port}`;
-  const client = redis.createClient(tls ? `rediss://${url}` : `redis://${url}`);
+  const client = new Redis(connectionOpts);
   ['ready', 'connect', 'reconnecting', 'error', 'end', 'warning']
     .forEach((event) => {
       client.on(event, (...args) => {
@@ -23,6 +31,7 @@ module.exports = (opts, logger) => {
 
   return {
     client,
+    getTtsSize: require('./lib/get-tts-size').bind(null, client, logger),
     purgeTtsCache: require('./lib/purge-tts-cache').bind(null, client, logger),
     synthAudio: require('./lib/synth-audio').bind(null, client, logger),
     getNuanceAccessToken: require('./lib/get-nuance-access-token').bind(null, client, logger),

lib/get-ibm-access-token.js

@@ -9,7 +9,7 @@ async function getIbmAccessToken(client, logger, apiKey) {
   logger = logger || noopLogger;
   try {
     const key = makeIbmKey(apiKey);
-    const access_token = await client.getAsync(key);
+    const access_token = await client.get(key);
     if (access_token) return {access_token, servedFromCache: true};
 
     /* access token not found in cache, so fetch it from Ibm */

lib/get-nuance-access-token.js

@@ -9,7 +9,7 @@ async function getNuanceAccessToken(client, logger, clientId, secret, scope) {
   logger = logger || noopLogger;
   try {
     const key = makeNuanceKey(clientId, secret, scope);
-    const access_token = await client.getAsync(key);
+    const access_token = await client.get(key);
     if (access_token) return {access_token, servedFromCache: true};
 
     /* access token not found in cache, so fetch it from Nuance */

lib/get-tts-size.js

@@ -0,0 +1,11 @@
+async function getTtsSize(client, logger, pattern = null) {
+  let keys;
+  if (pattern) {
+    keys = await client.keys(pattern);
+  } else {
+    keys = await client.keys('tts:*');
+  }
+  return keys.length;
+}
+
+module.exports = getTtsSize;

lib/purge-tts-cache.js

@@ -19,12 +19,12 @@ async function purgeTtsCache(client, logger, {all, account_sid, vendor,
 
   try {
     if (all) {
-      const keys = await client.keysAsync('tts:*');
-      purgedCount = await client.delAsync(keys);
+      const keys = await client.keys('tts:*');
+      purgedCount = await client.del(keys);
 
     } else if (account_sid && !vendor && !language && !voice && !engine && !text) {
-      const keys = await client.keysAsync(`tts:${account_sid}:*`);
-      purgedCount = await client.delAsync(keys);
+      const keys = await client.keys(`tts:${account_sid}:*`);
+      purgedCount = await client.del(keys);
     }
     else {
       const key = makeSynthKey({
@@ -35,7 +35,7 @@ async function purgeTtsCache(client, logger, {all, account_sid, vendor,
         engine,
         text,
       });
-      purgedCount = await client.delAsync(key);
+      purgedCount = await client.del(key);
       if (purgedCount === 0) error = 'Specified item not found';
     }
 

lib/synth-audio.js

@@ -14,7 +14,13 @@ const {
   CancellationDetails,
   SpeechSynthesisOutputFormat
 } = sdk;
-const {makeSynthKey, createNuanceClient, createKryptonClient, createRivaClient, noopLogger} = require('./utils');
+const {
+  makeSynthKey,
+  createNuanceClient,
+  createKryptonClient,
+  createRivaClient,
+  noopLogger
+} = require('./utils');
 const getNuanceAccessToken = require('./get-nuance-access-token');
 const {
   SynthesisRequest,
@@ -30,9 +36,27 @@ const {
 const {SynthesizeSpeechRequest} = require('../stubs/riva/proto/riva_tts_pb');
 const {AudioEncoding} = require('../stubs/riva/proto/riva_audio_pb');
 const debug = require('debug')('jambonz:realtimedb-helpers');
-const EXPIRES = (process.env.JAMBONES_TTS_CACHE_DURATION_MINS || 24 * 60) * 60; // cache tts for 24 hours
+const EXPIRES = (process.env.JAMBONES_TTS_CACHE_DURATION_MINS || 4 * 60) * 60; // cache tts for 4 hours
 const TMP_FOLDER = '/tmp';
 
+
+const trimTrailingSilence = (buffer) => {
+  assert.ok(buffer instanceof Buffer, 'trimTrailingSilence - argument is not a Buffer');
+
+  let offset = buffer.length;
+  while (offset > 0) {
+    // Get 16-bit value from the buffer (read in reverse)
+    const value = buffer.readUInt16BE(offset - 2);
+    if (value !== 0) {
+      break;
+    }
+    offset -= 2;
+  }
+
+  // Trim the silence from the end
+  return offset === buffer.length ? buffer : buffer.subarray(0, offset);
+};
+
 /**
  * Synthesize speech to an mp3 file, and also cache the generated speech
  * in redis (base64 format) for 24 hours so as to avoid unnecessarily paying
@@ -81,7 +105,7 @@ async function synthAudio(client, logger, stats, { account_sid,
   else if ('nvidia' === vendor) {
     assert.ok(voice, 'synthAudio requires voice when nvidia is used');
     assert.ok(language, 'synthAudio requires language when nvidia is used');
-    assert.ok(credentials.riva_uri, 'synthAudio requires riva_uri in credentials when nuance is used');
+    assert.ok(credentials.riva_server_uri, 'synthAudio requires riva_server_uri in credentials when nvidia is used');
   }
   else if ('ibm' === vendor) {
     assert.ok(voice, 'synthAudio requires voice when ibm is used');
@@ -104,14 +128,19 @@ async function synthAudio(client, logger, stats, { account_sid,
     text
   });
   let filePath;
-  if (['nuance', 'nvidia'].includes(vendor)) {
+  if (['nuance', 'nvidia'].includes(vendor) ||
+    (
+      process.env.JAMBONES_TTS_TRIM_SILENCE &&
+      ['microsoft', 'azure'].includes(vendor)
+    )
+  ) {
     filePath = `${TMP_FOLDER}/${key.replace('tts:', `tts-${salt || ''}`)}.r8`;
   }
   else filePath = `${TMP_FOLDER}/${key.replace('tts:', `tts-${salt || ''}`)}.mp3`;
   debug(`synth key is ${key}`);
   let cached;
   if (!disableTtsCache) {
-    cached = await client.getAsync(key);
+    cached = await client.get(key);
   }
   if (cached) {
     // found in cache - extend the expiry and use it
@@ -119,7 +148,7 @@ async function synthAudio(client, logger, stats, { account_sid,
     servedFromCache = true;
     stats.increment('tts.cache.requests', ['found:yes']);
     audioBuffer = Buffer.from(cached, 'base64');
-    client.expireAsync(key, EXPIRES).catch((err) => logger.info(err, 'Error setting expires'));
+    client.expire(key, EXPIRES).catch((err) => logger.info(err, 'Error setting expires'));
   }
   if (!cached) {
     // not found in cache - go get it from speech vendor and add to cache
@@ -168,7 +197,7 @@ async function synthAudio(client, logger, stats, { account_sid,
     debug(`tts rtt time for ${text.length} chars on ${vendorLabel}: ${rtt}`);
     logger.info(`tts rtt time for ${text.length} chars on ${vendorLabel}: ${rtt}`);
 
-    client.setexAsync(key, EXPIRES, audioBuffer.toString('base64'))
+    client.setex(key, EXPIRES, audioBuffer.toString('base64'))
       .catch((err) => logger.error(err, `error calling setex on key ${key}`));
   }
 
@@ -274,6 +303,47 @@ const synthIbm = async(logger, {credentials, stats, voice, text}) => {
   }
 };
 
+async function _synthOnPremMicrosoft(logger, {
+  credentials,
+  stats,
+  language,
+  voice,
+  text,
+  filePath
+}) {
+  const {use_custom_tts, custom_tts_endpoint_url} = credentials;
+  let content = text;
+
+  if (use_custom_tts && !content.startsWith('<speak')) {
+    /**
+     * Note: it seems that to use custom voice ssml is required with the voice attribute
+     * Otherwise sending plain text we get "Voice does not match"
+     */
+    content = `<speak>${text}</speak>`;
+  }
+
+  if (content.startsWith('<speak>')) {
+    /* microsoft enforces some properties and uses voice xml element so if the user did not supply do it for them */
+    const words = content.slice(7, -8).trim().replace(/(\r\n|\n|\r)/gm, ' ');
+    // eslint-disable-next-line max-len
+    content = `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts" xml:lang="${language}"><voice name="${voice}">${words}</voice></speak>`;
+    logger.info({content}, 'synthMicrosoft');
+  }
+
+  try {
+    const post = bent('POST', 'buffer', {
+      'X-Microsoft-OutputFormat': 'audio-16khz-128kbitrate-mono-mp3',
+      'Content-Type': 'application/ssml+xml',
+      'User-Agent': 'Jambonz'
+    });
+    const mp3 = await post(custom_tts_endpoint_url, content);
+    return mp3;
+  } catch (err) {
+    logger.info({err}, '_synthMicrosoftByHttp returned error');
+    throw err;
+  }
+}
+
 const synthMicrosoft = async(logger, {
   credentials,
   stats,
@@ -283,21 +353,35 @@ const synthMicrosoft = async(logger, {
   filePath
 }) => {
   try {
-    const {api_key: apiKey, region, use_custom_tts, custom_tts_endpoint} = credentials;
+    const {api_key: apiKey, region, use_custom_tts, custom_tts_endpoint, custom_tts_endpoint_url} = credentials;
+    if (use_custom_tts && custom_tts_endpoint_url) {
+      return await _synthOnPremMicrosoft(logger, {
+        credentials,
+        stats,
+        language,
+        voice,
+        text,
+        filePath
+      });
+    }
+    const trimSilence = filePath.endsWith('.r8');
     let content = text;
     const speechConfig = SpeechConfig.fromSubscription(apiKey, region);
     speechConfig.speechSynthesisLanguage = language;
     speechConfig.speechSynthesisVoiceName = voice;
     if (use_custom_tts && custom_tts_endpoint) {
       speechConfig.endpointId = custom_tts_endpoint;
-
+    }
+    if (use_custom_tts && !content.startsWith('<speak')) {
       /**
        * Note: it seems that to use custom voice ssml is required with the voice attribute
        * Otherwise sending plain text we get "Voice does not match"
        */
-      if (!content.startsWith('<speak')) content = `<speak>${text}</speak>`;
+      content = `<speak>${text}</speak>`;
     }
-    speechConfig.speechSynthesisOutputFormat = SpeechSynthesisOutputFormat.Audio16Khz32KBitRateMonoMp3;
+    speechConfig.speechSynthesisOutputFormat = trimSilence ?
+      SpeechSynthesisOutputFormat.Raw8Khz16BitMonoPcm :
+      SpeechSynthesisOutputFormat.Audio16Khz32KBitRateMonoMp3;
     const synthesizer = new SpeechSynthesizer(speechConfig);
 
     if (content.startsWith('<speak>')) {
@@ -323,7 +407,9 @@ const synthMicrosoft = async(logger, {
               reject(cancellation.errorDetails);
               break;
             case ResultReason.SynthesizingAudioCompleted:
-              resolve(Buffer.from(result.audioData));
+              let buffer = Buffer.from(result.audioData);
+              if (trimSilence) buffer = trimTrailingSilence(buffer);
+              resolve(buffer);
               synthesizer.close();
               stats.increment('tts.count', ['vendor:microsoft', 'accepted:yes']);
               break;
@@ -425,20 +511,25 @@ const synthNuance = async(client, logger, {credentials, stats, voice, model, tex
 };
 
 const synthNvidia = async(client, logger, {credentials, stats, language,  voice, model, text}) => {
-  const {riva_uri} = credentials;
-  const rivaClient = await createRivaClient(riva_uri);
-
-  const request = new SynthesizeSpeechRequest();
-  request.setVoiceName(voice);
-  request.setLanguageCode(language);
-  request.setSampleRateHz(8000);
-  request.setEncoding(AudioEncoding.LINEAR_PCM);
-  request.setText(text);
+  const {riva_server_uri} = credentials;
+  let rivaClient, request;
+  try {
+    rivaClient = await createRivaClient(riva_server_uri);
+    request = new SynthesizeSpeechRequest();
+    request.setVoiceName(voice);
+    request.setLanguageCode(language);
+    request.setSampleRateHz(8000);
+    request.setEncoding(AudioEncoding.LINEAR_PCM);
+    request.setText(text);
+  } catch (err) {
+    logger.info({err}, 'error creating riva client');
+    return Promise.reject(err);
+  }
 
   return new Promise((resolve, reject) => {
     rivaClient.synthesize(request, (err, response) => {
       if (err) {
-        console.error(err);
+        logger.info({err, voice, language}, 'error synthesizing speech using Nvidia');
         return reject(err);
       }
       resolve(Buffer.from(response.getAudio()));

lib/utils.js

@@ -106,7 +106,6 @@ const createRivaClient = async(rivaUri) => {
   return client;
 };
 
-
 module.exports = {
   makeSynthKey,
   makeNuanceKey,

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/speech-utils",
-  "version": "0.0.12",
+  "version": "0.0.20",
   "description": "TTS-related speech utilities for jambonz",
   "main": "index.js",
   "author": "Dave Horton",
@@ -9,7 +9,7 @@
     "test": "test"
   },
   "scripts": {
-    "test": "NODE_ENV=test JAMBONES_REDIS_USERNAME=daveh JAMBONES_REDIS_PASSWORD=foobarbazzle node test/ ",
+    "test": "NODE_ENV=test node test/ ",
     "coverage": "nyc --reporter html --report-dir ./coverage npm run test",
     "jslint": "eslint index.js lib",
     "build": "./build_stubs.sh"
@@ -24,17 +24,16 @@
   },
   "homepage": "https://github.com/jambonz/speech-utils#readme",
   "dependencies": {
-    "@aws-sdk/client-polly": "^3.303.0",
+    "@aws-sdk/client-polly": "^3.359.0",
     "@google-cloud/text-to-speech": "^4.2.1",
     "@grpc/grpc-js": "^1.8.13",
-    "@jambonz/promisify-redis": "^0.0.6",
     "bent": "^7.3.12",
     "debug": "^4.3.4",
     "form-urlencoded": "^6.1.0",
     "google-protobuf": "^3.21.2",
     "ibm-watson": "^8.0.0",
-    "microsoft-cognitiveservices-speech-sdk": "^1.26.0",
-    "redis": "^3.1.2",
+    "microsoft-cognitiveservices-speech-sdk": "^1.31.0",
+    "ioredis": "^5.3.2",
     "undici": "^5.21.0"
   },
   "devDependencies": {

test/ibm.js

@@ -31,7 +31,7 @@ test('IBM - create access key', async(t) => {
     //console.log({obj}, 'received access token from IBM - second request');
     t.ok(obj.access_token && obj.servedFromCache, 'successfully received access token from cache');
  
-    await client.flushallAsync();
+    await client.flushall();
     t.end();
   }
   catch (err) {
@@ -65,7 +65,7 @@ test('IBM - retrieve tts voices test', async(t) => {
     t.ok(voices.length > 0 && voices[0].language, 
       `GetVoices: successfully retrieved ${voices.length} voices from IBM`);
  
-    await client.flushallAsync();
+    await client.flushall();
 
     t.end();
 

test/list-voices.js

@@ -31,7 +31,7 @@ test('IBM - create access key', async(t) => {
     //console.log({obj}, 'received access token from IBM - second request');
     t.ok(obj.access_token && obj.servedFromCache, 'successfully received access token from cache');
  
-    await client.flushallAsync();
+    await client.flushall();
     t.end();
   }
   catch (err) {
@@ -65,7 +65,7 @@ test('IBM - retrieve tts voices test', async(t) => {
     t.ok(voices.length > 0 && voices[0].language, 
       `GetVoices: successfully retrieved ${voices.length} voices from IBM`);
  
-    await client.flushallAsync();
+    await client.flushall();
 
     t.end();
 
@@ -99,7 +99,7 @@ test('Nuance hosted tests', async(t) => {
     t.ok(voices.length > 0 && voices[0].language, 
       `GetVoices: successfully retrieved ${voices.length} voices from Nuance`);
 
-    await client.flushallAsync();
+    await client.flushall();
 
     t.end();
 
@@ -132,7 +132,7 @@ test('Nuance on-prem tests', async(t) => {
     t.ok(voices.length > 0 && voices[0].language, 
       `GetVoices: successfully retrieved ${voices.length} voices from Nuance`);
 
-    await client.flushallAsync();
+    await client.flushall();
 
     t.end();
 
@@ -162,7 +162,7 @@ test('Google tests', async(t) => {
     let result = await getTtsVoices(opts);
     t.ok(result[0].voices.length > 0, `GetVoices: successfully retrieved ${result[0].voices.length} voices from Google`);
 
-    await client.flushallAsync();
+    await client.flushall();
 
     t.end();
   }
@@ -193,7 +193,7 @@ test('AWS tests', async(t) => {
     let result = await getTtsVoices(opts);
     t.ok(result?.Voices?.length > 0, `GetVoices: successfully retrieved ${result.Voices.length} voices from AWS`);
 
-    await client.flushallAsync();
+    await client.flushall();
 
     t.end();
   }

test/nuance.js

@@ -34,7 +34,7 @@ test('Nuance hosted tests', async(t) => {
     t.ok(voices.length > 0 && voices[0].language, 
       `GetVoices: successfully retrieved ${voices.length} voices from Nuance`);
 
-    await client.flushallAsync();
+    await client.flushall();
 
     t.end();
 
@@ -67,7 +67,7 @@ test('Nuance on-prem tests', async(t) => {
     t.ok(voices.length > 0 && voices[0].language, 
       `GetVoices: successfully retrieved ${voices.length} voices from Nuance`);
 
-    await client.flushallAsync();
+    await client.flushall();
 
     t.end();
 

test/synth.js

@@ -300,7 +300,7 @@ test('Nvidia speech synth tests', async(t) => {
     let opts = await synthAudio(stats, {
       vendor: 'nvidia',
       credentials: {
-        riva_uri: process.env.RIVA_URI,
+        riva_server_uri: process.env.RIVA_URI,
       },
       language: 'en-US',
       voice: 'English-US.Female-1',
@@ -311,7 +311,7 @@ test('Nvidia speech synth tests', async(t) => {
     opts = await synthAudio(stats, {
       vendor: 'nvidia',
       credentials: {
-        riva_uri: process.env.RIVA_URI,
+        riva_server_uri: process.env.RIVA_URI,
       },
       language: 'en-US',
       voice: 'English-US.Female-1',
@@ -413,18 +413,21 @@ test('Custom Vendor speech synth tests', async(t) => {
 
 test('TTS Cache tests', async(t) => {
   const fn = require('..');
-  const {purgeTtsCache, client} = fn(opts, logger);
+  const {purgeTtsCache, getTtsSize, client} = fn(opts, logger);
 
   try {
     // save some random tts keys to cache
     const minRecords = 8;
     for (const i in Array(minRecords).fill(0)) {
-      await client.setAsync(makeSynthKey({vendor: i, language: i, voice: i, engine: i, text: i}), i);
+      await client.set(makeSynthKey({vendor: i, language: i, voice: i, engine: i, text: i}), i);
     }
+    const count = await getTtsSize();
+    t.ok(count >= minRecords, 'getTtsSize worked.');
+    
     const {purgedCount} = await purgeTtsCache();
     t.ok(purgedCount >= minRecords, `successfully purged at least ${minRecords} tts records from cache`);
 
-    const cached = (await client.keysAsync('tts:*')).length;
+    const cached = (await client.keys('tts:*')).length;
     t.equal(cached, 0, `successfully purged all tts records from cache`);
 
   } catch (err) {
@@ -435,11 +438,11 @@ test('TTS Cache tests', async(t) => {
   try {
     // save some random tts keys to cache
     for (const i in Array(10).fill(0)) {
-      await client.setAsync(makeSynthKey({vendor: i, language: i, voice: i, engine: i, text: i}), i);
+      await client.set(makeSynthKey({vendor: i, language: i, voice: i, engine: i, text: i}), i);
     }
     // save a specific key to tts cache
     const opts = {vendor: 'aws', language: 'en-US', voice: 'MALE', engine: 'Engine', text: 'Hello World!'};
-    await client.setAsync(makeSynthKey(opts), opts.text);
+    await client.set(makeSynthKey(opts), opts.text);
 
     const {purgedCount} = await purgeTtsCache({all: false, ...opts});
     t.ok(purgedCount === 1, `successfully purged one specific tts record from cache`);
@@ -455,7 +458,7 @@ test('TTS Cache tests', async(t) => {
     t.ok(error, `error returned when specified key was not found`);
 
     // make sure other tts keys are still there
-    const cached = (await client.keysAsync('tts:*')).length;
+    const cached = (await client.keys('tts:*')).length;
     t.ok(cached >= 1, `successfully kept all non-specified tts records in cache`);
 
   } catch (err) {
@@ -471,21 +474,21 @@ test('TTS Cache tests', async(t) => {
     const account_sid = "12412512_cabc_5aff"
     const account_sid2 = "22412512_cabc_5aff"
     for (const i in Array(minRecords).fill(0)) {
-      await client.setAsync(makeSynthKey({account_sid, vendor: i, language: i, voice: i, engine: i, text: i}), i);
+      await client.set(makeSynthKey({account_sid, vendor: i, language: i, voice: i, engine: i, text: i}), i);
     }
     for (const i in Array(minRecords).fill(0)) {
-      await client.setAsync(makeSynthKey({account_sid: account_sid2, vendor: i, language: i, voice: i, engine: i, text: i}), i);
+      await client.set(makeSynthKey({account_sid: account_sid2, vendor: i, language: i, voice: i, engine: i, text: i}), i);
     }
     const {purgedCount} = await purgeTtsCache({account_sid});
     t.equal(purgedCount, minRecords, `successfully purged at least ${minRecords} tts records from cache for account_sid:${account_sid}`);
 
-    let cached = (await client.keysAsync('tts:*')).length;
+    let cached = (await client.keys('tts:*')).length;
     t.equal(cached, minRecords, `successfully purged all tts records from cache for account_sid:${account_sid}`);
 
     const {purgedCount: purgedCount2} = await purgeTtsCache({account_sid: account_sid2});
     t.equal(purgedCount2, minRecords, `successfully purged at least ${minRecords} tts records from cache for account_sid:${account_sid2}`);
 
-    cached = (await client.keysAsync('tts:*')).length;
+    cached = (await client.keys('tts:*')).length;
     t.equal(cached, 0, `successfully purged all tts records from cache`);
 
   } catch (err) {