ben.carrasco/jambonz-api-server
1
0
Fork 0
mirror of https://github.com/jambonz/jambonz-api-server.git synced 2026-10-09 21:14:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

TTS/STT languages and voices for each provider (#281)

Browse Source 
* wip

* wip

* wip

* wip

* wip

* wip

* add testcase
This commit is contained in:
Hoan Luu Huu
2024-01-08 18:15:40 -05:00
committed by GitHub
parent 997ff05f3c
commit 7e046ac7f3
27 changed files with 11536 additions and 82 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/routes/api/speech-credentials.js
+22 -81
View File
@@ -5,8 +5,9 @@ const SpeechCredential = require('../../models/speech-credential');
const sysError = require('../error');
const {decrypt, encrypt} = require('../../utils/encrypt-decrypt');
const {parseAccountSid, parseServiceProviderSid, parseSpeechCredentialSid} = require('./utils');
const {decryptCredential, testWhisper, testDeepgramTTS} = require('../../utils/speech-utils');
const {DbErrorUnprocessableRequest, DbErrorForbidden} = require('../../utils/errors');
const {decryptCredential, testWhisper, testDeepgramTTS,
getLanguagesAndVoicesForVendor} = require('../../utils/speech-utils');
const {DbErrorUnprocessableRequest, DbErrorForbidden, DbErrorBadRequest} = require('../../utils/errors');
const {
testGoogleTts,
testGoogleStt,
@@ -24,7 +25,6 @@ const {
testElevenlabs,
testAssemblyStt
} = require('../../utils/speech-utils');
const bent = require('bent');
const {promisePool} = require('../../db');
const validateAdd = async(req) => {
@@ -751,89 +751,30 @@ router.get('/:sid/test', async(req, res) => {
* Fetch speech voices and languages
*/
router.post('/voices', async(req, res) => {
router.get('/speech/supportedLanguagesAndVoices', async(req, res) => {
const logger = req.app.locals.logger;
const {vendor, label} = req.body;
const account_sid = req.user.account_sid || req.body.account_sid;
const service_provider_sid = req.user.service_provider_sid ||
req.body.service_provider_sid || parseServiceProviderSid(req);
try {
res.status(200).json(await getTtsVoices(vendor, label, service_provider_sid, account_sid));
const {vendor, label} = req.query;
if (!vendor) {
throw new DbErrorBadRequest('vendor is required');
}
const account_sid = req.user.account_sid || req.body.account_sid;
const service_provider_sid = req.user.service_provider_sid ||
req.body.service_provider_sid || parseServiceProviderSid(req);
const credentials = await SpeechCredential.getSpeechCredentialsByVendorAndLabel(
service_provider_sid, account_sid, vendor, label);
const tmp = credentials && credentials.length > 0 ? credentials[0] : null;
const cred = tmp ? JSON.parse(decrypt(tmp.credential)) : null;
try {
const data = await getLanguagesAndVoicesForVendor(logger, vendor, cred);
res.status(200).json(data);
} catch (err) {
throw new DbErrorUnprocessableRequest(err.message);
}
} catch (err) {
sysError(logger, res, err);
}
});
router.post('/languages', async(req, res) => {
const logger = req.app.locals.logger;
const {vendor, label} = req.body;
const account_sid = req.user.account_sid || req.body.account_sid;
const service_provider_sid = req.user.service_provider_sid ||
req.body.service_provider_sid || parseServiceProviderSid(req);
try {
res.status(200).json(await getTtsLanguages(vendor, label, service_provider_sid, account_sid));
} catch (err) {
sysError(logger, res, err);
}
});
const getTtsVoices = async(vendor, label, service_provider_sid, account_sid) => {
const credentials = await SpeechCredential.getSpeechCredentialsByVendorAndLabel(
service_provider_sid, account_sid, vendor, label);
const tmp = credentials && credentials.length > 0 ? credentials[0] : null;
const cred = tmp ? JSON.parse(decrypt(tmp.credential)) : null;
if (vendor === 'elevenlabs') {
const get = bent('https://api.elevenlabs.io', 'GET', 'json', {
...(cred && {
'xi-api-key' : cred.api_key
})
});
const resp = await get('/v1/voices');
return resp ? resp.voices.map((v) => {
let name = `${v.name}${v.category !== 'premade' ? ` (${v.category})` : ''} -
${v.labels.accent ? ` ${v.labels.accent},` : ''}
${v.labels.description ? ` ${v.labels.description},` : ''}
${v.labels.age ? ` ${v.labels.age},` : ''}
${v.labels.gender ? ` ${v.labels.gender},` : ''}
${v.labels['use case'] ? ` ${v.labels['use case']},` : ''}
`;
const lastIndex = name.lastIndexOf(',');
if (lastIndex !== -1) {
name = name.substring(0, lastIndex);
}
return {
value: v.voice_id,
name
};
}).sort((a, b) => a.name.localeCompare(b.name)) : [];
}
return [];
};
const getTtsLanguages = async(vendor, label, service_provider_sid, account_sid) => {
const credentials = await SpeechCredential.getSpeechCredentialsByVendorAndLabel(
service_provider_sid, account_sid, vendor, label);
const tmp = credentials && credentials.length > 0 ? credentials[0] : null;
const cred = tmp ? JSON.parse(decrypt(tmp.credential)) : null;
if (vendor === 'elevenlabs') {
if (!cred) {
return [];
}
const get = bent('https://api.elevenlabs.io', 'GET', 'json', {
'xi-api-key' : cred.api_key
});
const resp = await get('/v1/models');
if (!resp || resp.length === 0) {
return [];
}
const model = resp.find((m) => m.model_id === cred.model_id);
return model ? model.languages.map((l) => {
return {
value: l.language_id,
name: l.name
};
}).sort((a, b) => a.name.localeCompare(b.name)) : [];
}
};
module.exports = router;