@jambonz/speech-utils 1.0.17 → 1.0.19

package/.github/workflows/ci.yml

@@ -7,11 +7,18 @@ on:
 jobs:
   build:
     runs-on: ubuntu-latest
+    permissions:
+      id-token: write   # required to request the GitHub OIDC token for AWS
+      contents: read
     steps:
       - uses: actions/checkout@v4
       - uses: actions/setup-node@v4
         with:
           node-version: '20'
+      - uses: aws-actions/configure-aws-credentials@v4
+        with:
+          role-to-assume: ${{ secrets.AWS_ROLE_ARN }}
+          aws-region: us-east-1
       - run: npm install
       - run: npm run jslint
       - run: sudo apt update && sudo apt install -y squid
@@ -19,10 +26,12 @@ jobs:
       - run: sudo systemctl start squid
       - run: npm test
         env:
-          AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
-          AWS_REGION: ${{ secrets.AWS_REGION }}
-          AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
+          # AWS_* are exported by configure-aws-credentials above as short-lived
+          # OIDC credentials; no AWS keys are stored as repository secrets.
           GCP_JSON_KEY: ${{ secrets.GCP_JSON_KEY }}
+          # Enables the "AWS speech synth tests by RoleArn" test, which exercises the
+          # AssumeRole credential path used in production.
+          AWS_ROLE_ARN: ${{ secrets.AWS_ROLE_ARN }}
 
           MICROSOFT_API_KEY: ${{ secrets.MICROSOFT_API_KEY }}
           MICROSOFT_REGION: ${{ secrets.MICROSOFT_REGION }}

package/lib/synth-audio.js

@@ -80,8 +80,8 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
   logger = logger || noopLogger;
 
   assert.ok(['google', 'aws', 'polly', 'microsoft', 'wellsaid', 'nvidia', 'elevenlabs',
-    'whisper', 'deepgram', 'deepgramflux', 'rimelabs', 'cartesia', 'gradium', 'nineninesix', 'inworld', 'resemble',
-    'murf', 'xai']
+    'whisper', 'deepgram', 'deepgramflux', 'rimelabs', 'cartesia', 'gradium', 'kugelaudio', 'nineninesix', 'inworld',
+    'resemble', 'murf', 'xai', 'fishaudio']
     .includes(vendor) ||
   vendor.startsWith('custom'),
   `synthAudio supported vendors are google, aws, microsoft, nvidia and wellsaid ..etc, not ${vendor}`);
@@ -139,6 +139,9 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
   } else if ('gradium' === vendor) {
     assert.ok(voice, 'synthAudio requires voice when gradium is used');
     assert.ok(credentials.api_key, 'synthAudio requires api_key when gradium is used');
+  } else if ('kugelaudio' === vendor) {
+    assert.ok(voice, 'synthAudio requires voice when kugelaudio is used');
+    assert.ok(credentials.api_key, 'synthAudio requires api_key when kugelaudio is used');
   } else if ('nineninesix' === vendor) {
     assert.ok(voice, 'synthAudio requires voice when nineninesix is used');
     assert.ok(credentials.api_key, 'synthAudio requires api_key when nineninesix is used');
@@ -149,6 +152,10 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
   } else if (vendor === 'resemble') {
     assert.ok(voice, 'synthAudio requires voice when resemble is used');
     assert.ok(credentials.api_key, 'synthAudio requires api_key when resemble is used');
+  } else if ('fishaudio' === vendor) {
+    /* no voice assert: fish synthesizes with its own default voice when
+       reference_id is omitted, which is what the 'default' selection means */
+    assert.ok(credentials.api_key, 'synthAudio requires api_key when fishaudio is used');
   }
 
   const key = makeSynthKey({
@@ -220,11 +227,21 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
           credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming,
           disableTtsCache});
         break;
+      case 'fishaudio':
+        audioData = await synthFishaudio(logger, {
+          credentials, options, stats, voice, key, text, renderForCaching, disableTtsStreaming,
+          disableTtsCache});
+        break;
       case 'gradium':
         audioData = await synthGradium(logger, {
           credentials, options, stats, voice, key, text, renderForCaching, disableTtsStreaming,
           disableTtsCache});
         break;
+      case 'kugelaudio':
+        audioData = await synthKugelaudio(logger, {
+          credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming,
+          disableTtsCache});
+        break;
       case 'nineninesix':
         audioData = await synthNineninesix(logger, {
           credentials, stats, language, voice, key, text, renderForCaching, disableTtsStreaming,
@@ -1503,8 +1520,165 @@ const synthGradium = async(logger, {
   }
 };
 
+/* kugelaudio — json websocket (/ws/tts/stream) for streaming, and POST /v1/tts/generate
+   for the cache render. the POST streams back bare little-endian 16-bit samples at the
+   requested sample_rate, which is exactly the r8 container at 8000.
+
+   voices are numeric ids (or public handles). language is an ISO 639-1 code that drives
+   text normalization; jambonz carries BCP-47, so only the primary subtag is sent. the
+   api rejects codes outside its list, so an unsupported or unset language is omitted
+   and the voice's own language applies. options.api_uri pins a region
+   (e.g. api.eu.kugelaudio.com).
+*/
+const KUGELAUDIO_LANGUAGES = ['ar', 'bg', 'bn', 'cs', 'da', 'de', 'el', 'en', 'es', 'fa', 'fi', 'fr', 'he', 'hi',
+  'hr', 'hu', 'id', 'it', 'ja', 'ko', 'ms', 'nl', 'no', 'pl', 'pt', 'ro', 'ru', 'sk', 'sl', 'sr', 'sv', 'ta', 'th',
+  'tr', 'uk', 'ur', 'vi', 'yue', 'zh'];
+const synthKugelaudio = async(logger, {
+  credentials, options, stats, language, voice, key, text, renderForCaching, disableTtsStreaming,
+  disableTtsCache
+}) => {
+  const {api_key, model_id} = credentials;
+  const {api_uri, speed, cfg_scale, temperature, normalize, project_id, dictionary_ids} = options || {};
+  const isSet = (v) => v !== null && v !== undefined;
+
+  /* default to using the streaming interface, unless disabled by env var OR we want just a cache file */
+  if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
+    let params = '{';
+    params += `api_key=${api_key}`;
+    params += `,playback_id=${key}`;
+    params += ',vendor=kugelaudio';
+    params += `,voice=${voice}`;
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
+    params += `,model_id=${model_id || 'kugel-3'}`;
+    if (language) params += `,language=${language}`;
+    if (api_uri) params += `,api_uri=${api_uri}`;
+    if (isSet(speed)) params += `,speed=${speed}`;
+    if (isSet(cfg_scale)) params += `,cfg_scale=${cfg_scale}`;
+    if (isSet(temperature)) params += `,temperature=${temperature}`;
+    if (isSet(normalize)) params += `,normalize=${normalize}`;
+    if (isSet(project_id)) params += `,project_id=${project_id}`;
+    /* the say: param parser is bracket-aware, so the json array survives intact */
+    if (Array.isArray(dictionary_ids)) params += `,dictionary_ids=${JSON.stringify(dictionary_ids)}`;
+    params += '}';
+
+    return {
+      filePath: `say:${params}${text.replace(/\n/g, ' ').replace(/\r/g, ' ')}`,
+      servedFromCache: false,
+      rtt: 0
+    };
+  }
+
+  try {
+    const sampleRate = 8000;
+    const host = (api_uri || 'api.kugelaudio.com').replace(/^[a-z]+:\/\//, '').replace(/\/$/, '');
+    const post = bent(`https://${host}`, 'POST', 'buffer', {
+      'Authorization': `Bearer ${api_key}`,
+      'Content-Type': 'application/json; charset=utf-8'
+    });
+    const voiceId = /^\d+$/.test(`${voice}`) ? Number(voice) : voice;
+    const lang = language && language.split('-')[0].toLowerCase();
+    const audioContent = await post('/v1/tts/generate', {
+      text,
+      voice_id: voiceId,
+      model_id: model_id || 'kugel-3',
+      sample_rate: sampleRate,
+      ...(KUGELAUDIO_LANGUAGES.includes(lang) && {language: lang}),
+      ...(isSet(speed) && {speed: Number(speed)}),
+      ...(isSet(cfg_scale) && {cfg_scale: Number(cfg_scale)}),
+      ...(isSet(temperature) && {temperature: Number(temperature)}),
+      ...(isSet(normalize) && {normalize: normalize === true || normalize === 'true'}),
+      ...(isSet(project_id) && {project_id: Number(project_id)}),
+      ...(Array.isArray(dictionary_ids) && {dictionary_ids})
+    });
+    return {
+      audioContent,
+      extension: 'r8',
+      sampleRate
+    };
+  } catch (err) {
+    logger.info({err}, 'synth kugelaudio returned error');
+    stats.increment('tts.count', ['vendor:kugelaudio', 'accepted:no']);
+    throw err;
+  }
+};
+
 /* nineninesix.ai — a Cartesia-compatible API, but only raw/wav come back
    (mp3 is rejected), so the cache render asks for wav rather than mp3. */
+/* fish.audio — msgpack websocket for streaming, and a POST endpoint for the cache
+   render. format:pcm + sample_rate:8000 returns bare little-endian 16-bit samples,
+   which is exactly the r8 container. we avoid fish's wav output because its RIFF
+   header carries a placeholder size (0xffffff24) — length is unknown up front, as
+   with gradium.
+
+   fish is a voice-cloning vendor: the "voice" is a reference_id returned by
+   POST /model, and omitting it entirely synthesizes with fish's default voice.
+   the sentinel value 'default' (the bundled fallback entry in the portal) means
+   exactly that — send no reference_id.
+*/
+const synthFishaudio = async(logger, {
+  credentials, options, stats, voice, key, text, renderForCaching, disableTtsStreaming, disableTtsCache
+}) => {
+  const {api_key, model_id, fishaudio_tts_uri} = credentials;
+  const {reference_id, latency, chunk_length, speed, volume} = options || {};
+
+  /* free-text reference_id in the vendor options wins over the voice selector */
+  const refId = reference_id || (voice && voice !== 'default' ? voice : null);
+
+  /* default to using the streaming interface, unless disabled by env var OR we want just a cache file */
+  if (!JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
+    let params = '{';
+    params += `api_key=${api_key}`;
+    params += `,playback_id=${key}`;
+    params += ',vendor=fishaudio';
+    params += `,voice=${refId || 'default'}`;
+    params += `,write_cache_file=${disableTtsCache ? 0 : 1}`;
+    if (model_id) params += `,model_id=${model_id}`;
+    if (latency) params += `,latency=${latency}`;
+    if (chunk_length) params += `,chunk_length=${chunk_length}`;
+    if (speed) params += `,speed=${speed}`;
+    if (volume) params += `,volume=${volume}`;
+    if (fishaudio_tts_uri) params += `,endpoint=${fishaudio_tts_uri}`;
+    params += '}';
+
+    return {
+      filePath: `say:${params}${text.replace(/\n/g, ' ').replace(/\r/g, ' ')}`,
+      servedFromCache: false,
+      rtt: 0
+    };
+  }
+
+  try {
+    const sampleRate = 8000;
+    const post = bent(fishaudio_tts_uri || 'https://api.fish.audio', 'POST', 'buffer', {
+      'Authorization': `Bearer ${api_key}`,
+      'Content-Type': 'application/json',
+      /* the model is selected by header, not in the body */
+      'model': model_id || 's2.1-pro'
+    });
+    const audioContent = await post('/v1/tts', {
+      text,
+      format: 'pcm',
+      sample_rate: sampleRate,
+      ...(refId && {reference_id: refId}),
+      ...(latency && {latency}),
+      ...(chunk_length && {chunk_length: parseInt(chunk_length, 10)}),
+      ...((speed || volume) && {prosody: {
+        ...(speed && {speed: parseFloat(speed)}),
+        ...(volume && {volume: parseFloat(volume)})
+      }})
+    });
+    return {
+      audioContent,
+      extension: 'r8',
+      sampleRate
+    };
+  } catch (err) {
+    logger.info({err}, 'synth fishaudio returned error');
+    stats.increment('tts.count', ['vendor:fishaudio', 'accepted:no']);
+    throw err;
+  }
+};
+
 const synthNineninesix = async(logger, {
   credentials, stats, voice, language, key, text, renderForCaching, disableTtsStreaming, disableTtsCache
 }) => {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/speech-utils",
-  "version": "1.0.17",
+  "version": "1.0.19",
   "description": "TTS-related speech utilities for jambonz",
   "main": "index.js",
   "author": "Dave Horton",

package/test/aws-credentials.js

@@ -0,0 +1,23 @@
+/**
+ * Resolve AWS credentials for the test suite.
+ *
+ * Returns null when AWS is not configured, so the caller skips.
+ *
+ * When AWS_SESSION_TOKEN is set the credentials are temporary -- GitHub OIDC in CI, or
+ * `aws sso login` locally -- and the key pair must not be passed through. Doing so sends
+ * lib/get-aws-sts-token.js down its accessKeyId branch, which calls GetSessionToken, and
+ * AWS rejects GetSessionToken when it is called with session credentials. Returning the
+ * region alone routes to the SDK's default credential provider chain instead, which
+ * handles temporary credentials correctly.
+ */
+module.exports = () => {
+  const region = process.env.AWS_REGION;
+  if (!region) return null;
+
+  const accessKeyId = process.env.AWS_ACCESS_KEY_ID;
+  const secretAccessKey = process.env.AWS_SECRET_ACCESS_KEY;
+
+  if (process.env.AWS_SESSION_TOKEN) return {region};
+  if (accessKeyId && secretAccessKey) return {accessKeyId, secretAccessKey, region};
+  return {region};
+};

package/test/aws.js

@@ -6,33 +6,33 @@ process.on('unhandledRejection', (reason, p) => {
   console.log('Unhandled Rejection at: Promise', p, 'reason:', reason);
 });
 
+const awsCredentials = require('./aws-credentials');
+
 const sleep = (ms) => new Promise((resolve) => setTimeout(resolve, ms));
 
 test('AWS - create and cache auth token', async(t) => {
   const fn = require('..');
   const {client, getAwsAuthToken} = fn(opts, logger);
 
-  if (!process.env.AWS_ACCESS_KEY_ID || !process.env.AWS_SECRET_ACCESS_KEY || !process.env.AWS_REGION) {
+  const credentials = awsCredentials();
+  if (!credentials) {
       t.pass('skipping AWS auth token tests since no AWS credentials provided');
       t.end();
       client.quit();
       return;
   }
+  // getAwsAuthToken derives its cache key from roleArn || accessKeyId || speech_credential_sid.
+  // With temporary credentials none of the first two are passed, so supply a stable sid --
+  // which is what production does for instance-profile credentials.
+  const args = {...credentials, speech_credential_sid: 'test-aws-speech-credential'};
+
   try {
-    let obj = await getAwsAuthToken({
-      accessKeyId: process.env.AWS_ACCESS_KEY_ID, 
-      secretAccessKey: process.env.AWS_SECRET_ACCESS_KEY,
-      region: process.env.AWS_REGION
-    });
+    let obj = await getAwsAuthToken(args);
     //console.log({obj}, 'received auth token from AWS');
     t.ok(obj.securityToken && !obj.servedFromCache, 'successfullY generated auth token from AWS');
 
     await sleep(250);
-    obj = await getAwsAuthToken({
-      accessKeyId: process.env.AWS_ACCESS_KEY_ID, 
-      secretAccessKey: process.env.AWS_SECRET_ACCESS_KEY,
-      region: process.env.AWS_REGION
-    });
+    obj = await getAwsAuthToken(args);
     //console.log({obj}, 'received auth token from AWS - second request');
     t.ok(obj.securityToken && obj.servedFromCache, 'successfully received access token from cache');
  

package/test/list-voices.js

@@ -1,4 +1,5 @@
 const test = require('tape').test ;
+const awsCredentials = require('./aws-credentials');
 const config = require('config');
 const opts = config.get('redis');
 const fs = require('fs');
@@ -45,18 +46,15 @@ test('AWS tests', async(t) => {
   const fn = require('..');
   const {client, getTtsVoices} = fn(opts, logger);
 
-  if (!process.env.AWS_ACCESS_KEY_ID || !process.env.AWS_SECRET_ACCESS_KEY || !process.env.AWS_REGION) {
-    t.pass('skipping AWS speech synth tests since AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, or AWS_REGION not provided');
+  const credentials = awsCredentials();
+  if (!credentials) {
+    t.pass('skipping AWS speech synth tests since AWS_REGION not provided');
     return t.end();
   }
   try {
     const opts = {
       vendor: 'aws',
-      credentials: {
-        accessKeyId: process.env.AWS_ACCESS_KEY_ID,
-        secretAccessKey: process.env.AWS_SECRET_ACCESS_KEY,
-        region: process.env.AWS_REGION,
-      }
+      credentials
     };
     let result = await getTtsVoices(opts);
     t.ok(result?.Voices?.length > 0, `GetVoices: successfully retrieved ${result.Voices.length} voices from AWS`);

package/test/synth.js

@@ -1,4 +1,5 @@
 const test = require('tape').test;
+const awsCredentials = require('./aws-credentials');
 const config = require('config');
 const opts = config.get('redis');
 const fs = require('fs');
@@ -668,18 +669,15 @@ test('AWS speech synth tests', async(t) => {
   const fn = require('..');
   const {synthAudio, client} = fn(opts, logger);
 
-  if (!process.env.AWS_ACCESS_KEY_ID || !process.env.AWS_SECRET_ACCESS_KEY || !process.env.AWS_REGION) {
-    t.pass('skipping AWS speech synth tests since AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, or AWS_REGION not provided');
+  const credentials = awsCredentials();
+  if (!credentials) {
+    t.pass('skipping AWS speech synth tests since AWS_REGION not provided');
     return t.end();
   }
   try {
     let opts = await synthAudio(stats, {
       vendor: 'aws',
-      credentials: {
-        accessKeyId: process.env.AWS_ACCESS_KEY_ID,
-        secretAccessKey: process.env.AWS_SECRET_ACCESS_KEY,
-        region: process.env.AWS_REGION,
-      },
+      credentials,
       language: 'en-US',
       voice: 'Joey',
       text: 'This is a test.  This is only a test',
@@ -689,11 +687,7 @@ test('AWS speech synth tests', async(t) => {
 
     opts = await synthAudio(stats, {
       vendor: 'aws',
-      credentials: {
-        accessKeyId: process.env.AWS_ACCESS_KEY_ID,
-        secretAccessKey: process.env.AWS_SECRET_ACCESS_KEY,
-        region: process.env.AWS_REGION,
-      },
+      credentials,
       language: 'en-US',
       voice: 'Joey',
       text: 'This is a test.  This is only a test',
@@ -724,9 +718,18 @@ test('AWS speech synth tests by RoleArn', async(t) => {
       },
       language: 'en-US',
       voice: 'Joey',
-      text: 'This is a test.  This is only a test',
+      // Distinct text on purpose: the 'AWS speech synth tests' above cache audio for
+      // the same vendor/voice/language, and identical text would hit that cache entry,
+      // making servedFromCache true and this assertion fail.
+      text: 'This is a roleArn test.  This is only a roleArn test',
+      // Without this, synthPolly returns the mediajam streaming params string, which
+      // embeds accessKeyId/secretAccessKey/sessionToken (see lib/synth-audio.js).
+      // Every other synth test in this file renders for caching; this one did not,
+      // so it printed live credentials into a public CI log.
+      renderForCaching: true,
     });
-    t.ok(!opts.servedFromCache, `successfully synthesized aws by roleArn audio to ${opts.filePath}`);
+    // Deliberately does not interpolate opts.filePath -- it can carry credentials.
+    t.ok(!opts.servedFromCache, 'successfully synthesized aws by roleArn audio');
   } catch (err) {
     console.error(err);
     t.end(err);
@@ -1087,6 +1090,77 @@ test('gradium speech synth tests', async(t) => {
   client.quit();
 });
 
+test('kugelaudio speech synth tests', async(t) => {
+  const fn = require('..');
+  const {synthAudio, client} = fn(opts, logger);
+
+  if (!process.env.KUGELAUDIO_API_KEY) {
+    t.pass('skipping kugelaudio speech synth tests since KUGELAUDIO_API_KEY is not provided');
+    return t.end();
+  }
+  const text = 'Guten Tag und willkommen bei jambonz! Ihre Bestellung kostet 12,99 Euro. ' + Date.now();
+  try {
+    const opts = await synthAudio(stats, {
+      vendor: 'kugelaudio',
+      credentials: {
+        api_key: process.env.KUGELAUDIO_API_KEY,
+        model_id: 'kugel-3'
+      },
+      language: 'de-DE',
+      voice: '1930',
+      text,
+      renderForCaching: true
+    });
+    t.ok(!opts.servedFromCache, `successfully synthed kugelaudio audio to ${opts.filePath}`);
+
+  } catch (err) {
+    console.error(JSON.stringify(err));
+    t.end(err);
+  }
+  client.quit();
+});
+
+test('fishaudio speech synth tests', async(t) => {
+  const fn = require('..');
+  const {synthAudio, client} = fn(opts, logger);
+
+  if (!process.env.FISHAUDIO_API_KEY) {
+    t.pass('skipping fishaudio speech synth tests since FISHAUDIO_API_KEY is not provided');
+    return t.end();
+  }
+  const text = 'Hi there and welcome to jambones! ' + Date.now();
+  try {
+    /* voice 'default' means "send no reference_id" — fish's own default voice */
+    const o = await synthAudio(stats, {
+      vendor: 'fishaudio',
+      credentials: {
+        api_key: process.env.FISHAUDIO_API_KEY,
+        model_id: 's2.1-pro'
+      },
+      voice: 'default',
+      text,
+      renderForCaching: true
+    });
+    t.ok(!o.servedFromCache, `successfully synthed fishaudio audio to ${o.filePath}`);
+
+    /* the cache render must be raw 8k pcm (r8): fish's wav header carries a
+       placeholder RIFF size, so we never ask for wav */
+    const o2 = await synthAudio(stats, {
+      vendor: 'fishaudio',
+      credentials: {api_key: process.env.FISHAUDIO_API_KEY},
+      voice: 'default',
+      text: text + ' two',
+      renderForCaching: true,
+      disableTtsCache: true
+    });
+    t.ok(!o2.servedFromCache, 'fishaudio synthed a second uncached render');
+  } catch (err) {
+    console.error(JSON.stringify(err));
+    t.end(err);
+  }
+  client.quit();
+});
+
 test('nineninesix speech synth tests', async(t) => {
   const fn = require('..');
   const {synthAudio, client} = fn(opts, logger);
@@ -1104,7 +1178,7 @@ test('nineninesix speech synth tests', async(t) => {
         model_id: 'gepard-1.0'
       },
       language: 'en',
-      voice: '3ad7a827-7fd1-4954-bf35-47d4cc33d9ed',
+      voice: '775e4dfd-819c-4325-92ec-250f487ef7e3',
       text,
       renderForCaching: true
     });