mirror of
https://github.com/jambonz/speech-utils.git
synced 2026-10-03 23:33:59 +00:00
Compare commits
7
Commits
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
b442b050e7 | ||
|
|
d96ab5cdf3 | ||
|
|
3bf74671c6 | ||
|
|
a47ef6d7c4 | ||
|
|
84089fa528 | ||
|
|
72be44eea2 | ||
|
|
05d6c4b32d |
@@ -7,22 +7,23 @@ const CACHE_EXPIRY = process.env.AWS_STS_SESSION_RESET_EXPIRY || (EXPIRY - 600);
|
||||
|
||||
async function getAwsAuthToken(
|
||||
logger, createHash, retrieveHash,
|
||||
{accessKeyId, secretAccessKey, region, roleArn}) {
|
||||
{speech_credential_sid, accessKeyId, secretAccessKey, region, roleArn}) {
|
||||
logger = logger || noopLogger;
|
||||
try {
|
||||
const key = makeAwsKey(roleArn || accessKeyId);
|
||||
// if incase instance profile is used, speech_credential_sid will be used as key to lookup cache
|
||||
const key = makeAwsKey(roleArn || accessKeyId || speech_credential_sid);
|
||||
const obj = await retrieveHash(key);
|
||||
if (obj) return {...obj, servedFromCache: true};
|
||||
|
||||
/* access token not found in cache, so generate it using STS */
|
||||
let data;
|
||||
let expiry = CACHE_EXPIRY;
|
||||
if (roleArn) {
|
||||
const stsClient = new STSClient({ region });
|
||||
const roleToAssume = { RoleArn: roleArn, RoleSessionName: 'Jambonz_Speech', DurationSeconds: EXPIRY};
|
||||
const command = new AssumeRoleCommand(roleToAssume);
|
||||
|
||||
data = await stsClient.send(command);
|
||||
} else {
|
||||
/* access token not found in cache, so generate it using STS */
|
||||
} else if (accessKeyId) {
|
||||
const stsClient = new STSClient({
|
||||
region,
|
||||
credentials: {
|
||||
@@ -32,6 +33,26 @@ async function getAwsAuthToken(
|
||||
});
|
||||
const command = new GetSessionTokenCommand({DurationSeconds: EXPIRY});
|
||||
data = await stsClient.send(command);
|
||||
} else {
|
||||
// instance profile is used.
|
||||
const stsClient = new STSClient({ region });
|
||||
const cred = await stsClient.config.credentials();
|
||||
// method in the AWS SDK automatically fetches credentials using the default credential
|
||||
// provider chain. If the credentials come from an instance profile or an environment
|
||||
// variable, their expiration is controlled by AWS and not explicitly by our code.
|
||||
if (cred && cred.expiration) {
|
||||
const currentTime = new Date();
|
||||
const expiryTime = new Date(cred.expiration);
|
||||
const remainingTimeInSeconds = Math.round((expiryTime - currentTime) / 1000);
|
||||
expiry = remainingTimeInSeconds;
|
||||
}
|
||||
data = {
|
||||
Credentials: {
|
||||
AccessKeyId: cred.accessKeyId,
|
||||
SecretAccessKey: cred.secretAccessKey,
|
||||
SessionToken: cred.sessionToken
|
||||
}
|
||||
};
|
||||
}
|
||||
|
||||
const credentials = {
|
||||
@@ -40,9 +61,11 @@ async function getAwsAuthToken(
|
||||
sessionToken: data.Credentials.SessionToken,
|
||||
securityToken: data.Credentials.SessionToken
|
||||
};
|
||||
|
||||
createHash(key, credentials, CACHE_EXPIRY)
|
||||
.catch((err) => logger.error(err, `Error saving hash for key ${key}`));
|
||||
// Only cache if expiry is good
|
||||
if (expiry > 0) {
|
||||
createHash(key, credentials, expiry)
|
||||
.catch((err) => logger.error(err, `Error saving hash for key ${key}`));
|
||||
}
|
||||
|
||||
return {...credentials, servedFromCache: false};
|
||||
} catch (err) {
|
||||
|
||||
+25
-4
@@ -170,6 +170,8 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
|
||||
renderForCaching
|
||||
});
|
||||
let filePath;
|
||||
// used only for custom vendor
|
||||
let fileExtension;
|
||||
filePath = makeFilePath({vendor, voice, key, salt, renderForCaching});
|
||||
debug(`synth key is ${key}`);
|
||||
let cached;
|
||||
@@ -201,7 +203,17 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
|
||||
debug('result WAS found in cache');
|
||||
servedFromCache = true;
|
||||
stats.increment('tts.cache.requests', ['found:yes']);
|
||||
audioBuffer = Buffer.from(cached, 'base64');
|
||||
if (vendor.startsWith('custom')) {
|
||||
// custom vendors support multiple mime types such as: mp3, wav, r8, r16 ...etc,
|
||||
// mime type/file extension is available when http response has header Content-type.
|
||||
// In cache, file extension is store together with audiBuffer in a json.
|
||||
// Normal cache audio will be base64 string
|
||||
const payload = JSON.parse(cached);
|
||||
filePath = filePath.replace(/\.[^\.]*$/g, payload.fileExtension);
|
||||
audioBuffer = Buffer.from(payload.audioBuffer, 'base64');
|
||||
} else {
|
||||
audioBuffer = Buffer.from(cached, 'base64');
|
||||
}
|
||||
client.expire(key, EXPIRES).catch((err) => logger.info(err, 'Error setting expires'));
|
||||
}
|
||||
if (!cached) {
|
||||
@@ -268,7 +280,7 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
|
||||
renderForCaching, disableTtsStreaming});
|
||||
break;
|
||||
case vendor.startsWith('custom') ? vendor : 'cant_match_value':
|
||||
({ audioBuffer, filePath } = await synthCustomVendor(logger,
|
||||
({ audioBuffer, filePath, fileExtension } = await synthCustomVendor(logger,
|
||||
{credentials, stats, language, voice, text, filePath}));
|
||||
break;
|
||||
default:
|
||||
@@ -282,7 +294,14 @@ async function synthAudio(client, createHash, retrieveHash, logger, stats, { acc
|
||||
debug(`tts rtt time for ${text.length} chars on ${vendorLabel}: ${rtt}`);
|
||||
logger.info(`tts rtt time for ${text.length} chars on ${vendorLabel}: ${rtt}`);
|
||||
|
||||
client.setex(key, EXPIRES, audioBuffer.toString('base64'))
|
||||
const base64Audio = audioBuffer.toString('base64');
|
||||
const cacheContent = vendor.startsWith('custom') ?
|
||||
JSON.stringify({
|
||||
audioBuffer: base64Audio,
|
||||
fileExtension
|
||||
}) : base64Audio;
|
||||
|
||||
client.setex(key, EXPIRES, cacheContent)
|
||||
.catch((err) => logger.error(err, `error calling setex on key ${key}`));
|
||||
}
|
||||
|
||||
@@ -729,9 +748,11 @@ const synthCustomVendor = async(logger, {credentials, stats, language, voice, te
|
||||
const regex = /\.[^\.]*$/g;
|
||||
const mime = response.headers['content-type'];
|
||||
const buffer = await response.arrayBuffer();
|
||||
const fileExtension = getFileExtFromMime(mime);
|
||||
return {
|
||||
audioBuffer: buffer,
|
||||
filePath: filePath.replace(regex, getFileExtFromMime(mime))
|
||||
filePath: filePath.replace(regex, fileExtension),
|
||||
fileExtension
|
||||
};
|
||||
} catch (err) {
|
||||
logger.info({err}, `Vendor ${vendor} returned error`);
|
||||
|
||||
Generated
+214
-593
File diff suppressed because it is too large
Load Diff
+2
-2
@@ -1,6 +1,6 @@
|
||||
{
|
||||
"name": "@jambonz/speech-utils",
|
||||
"version": "0.1.21",
|
||||
"version": "0.1.22",
|
||||
"description": "TTS-related speech utilities for jambonz",
|
||||
"main": "index.js",
|
||||
"author": "Dave Horton",
|
||||
@@ -35,7 +35,7 @@
|
||||
"debug": "^4.3.4",
|
||||
"form-urlencoded": "^6.1.4",
|
||||
"google-protobuf": "^3.21.2",
|
||||
"ibm-watson": "^8.0.0",
|
||||
"ibm-watson": "^10.0.0",
|
||||
"microsoft-cognitiveservices-speech-sdk": "1.38.0",
|
||||
"openai": "^4.25.0",
|
||||
"undici": "^6.4.0"
|
||||
|
||||
@@ -554,6 +554,7 @@ test('Custom Vendor speech synth tests', async(t) => {
|
||||
text: 'This is a test. This is only a test',
|
||||
});
|
||||
t.ok(!opts.servedFromCache, `successfully synthesized custom vendor audio to ${opts.filePath}`);
|
||||
t.ok(opts.filePath.endsWith('wav'), 'audio is cached as wav file');
|
||||
let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/somethingnew`);
|
||||
t.ok(obj.headers.Authorization == 'Bearer some_jwt_token', 'Custom Vendor Authentication Header is correct');
|
||||
t.ok(obj.body.language == 'en-US', 'Custom Vendor Language is correct');
|
||||
@@ -561,6 +562,21 @@ test('Custom Vendor speech synth tests', async(t) => {
|
||||
t.ok(obj.body.type == 'text', 'Custom Vendor type is correct');
|
||||
t.ok(obj.body.text == 'This is a test. This is only a test', 'Custom Vendor text is correct');
|
||||
|
||||
// Checking if cache is stored with wav format
|
||||
opts = await synthAudio(stats, {
|
||||
vendor: 'custom:somethingnew',
|
||||
credentials: {
|
||||
use_for_tts: 1,
|
||||
custom_tts_url: "http://127.0.0.1:3100/somethingnew",
|
||||
auth_token: 'some_jwt_token'
|
||||
},
|
||||
language: 'en-US',
|
||||
voice: 'English-US.Female-1',
|
||||
text: 'This is a test. This is only a test',
|
||||
});
|
||||
t.ok(opts.servedFromCache, `successfully get custom vendor cached audio to ${opts.filePath}`);
|
||||
t.ok(opts.filePath.endsWith('wav'), 'audio is cached as wav file');
|
||||
|
||||
opts = await synthAudio(stats, {
|
||||
vendor: 'custom:somethingnew2',
|
||||
credentials: {
|
||||
|
||||
Reference in New Issue
Block a user