ben.carrasco/jambonz-api-server
1
0
Fork 0
mirror of https://github.com/jambonz/jambonz-api-server.git synced 2026-07-24 21:12:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

support mod cartesia transcribe (#463)

Browse Source
This commit is contained in:
Hoan Luu Huu
2025-06-18 01:53:22 +07:00
committed by GitHub
parent fcff3d4b32
commit e2fc0216e1
3 changed files with 54 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/routes/api/speech-credentials.js
+32 -5
View File
@@ -135,6 +135,8 @@ const encryptCredential = (obj) => {
custom_tts_endpoint,
custom_tts_endpoint_url,
use_custom_stt,
use_for_stt,
use_for_tts,
custom_stt_endpoint,
custom_stt_endpoint_url,
tts_api_key,
@@ -148,7 +150,10 @@ const encryptCredential = (obj) => {
custom_tts_streaming_url,
auth_token = '',
cobalt_server_uri,
// For most vendors, model_id is being used for both TTS and STT, or one of them.
// for Cartesia, model_id is used for TTS only. introduce stt_model_id for STT
model_id,
stt_model_id,
user_id,
voice_engine,
engine_version,
@@ -259,8 +264,17 @@ const encryptCredential = (obj) => {
case 'cartesia':
assert(api_key, 'invalid cartesia speech credential: api_key is required');
assert(model_id, 'invalid cartesia speech credential: model_id is required');
const cartesiaData = JSON.stringify({api_key, model_id, options});
if (use_for_tts) {
assert(model_id, 'invalid cartesia speech credential: model_id is required');
}
if (use_for_stt) {
assert(stt_model_id, 'invalid cartesia speech credential: stt_model_id is required');
}
const cartesiaData = JSON.stringify({
api_key,
...(model_id && {model_id}),
...(stt_model_id && {stt_model_id}),
options});
return encrypt(cartesiaData);
case 'rimelabs':
@@ -487,6 +501,7 @@ router.put('/:sid', async(req, res) => {
custom_tts_streaming_url,
cobalt_server_uri,
model_id,
stt_model_id,
voice_engine,
options,
deepgram_stt_uri,
@@ -518,6 +533,7 @@ router.put('/:sid', async(req, res) => {
custom_tts_streaming_url,
cobalt_server_uri,
model_id,
stt_model_id,
voice_engine,
options,
deepgram_stt_uri,
@@ -833,17 +849,28 @@ router.get('/:sid/test', async(req, res) => {
}
}
} else if (cred.vendor === 'cartesia') {
if (cred.use_for_tts) {
if (cred.use_for_tts || cred.use_for_stt) {
try {
// Cartesia does not have API for testing STT, same key is used for both TTS and STT
await testCartesia(logger, synthAudio, credential);
results.tts.status = 'ok';
if (cred.use_for_tts) {
results.tts.status = 'ok';
}
if (cred.use_for_stt) {
results.stt.status = 'ok';
}
SpeechCredential.ttsTestResult(sid, true);
} catch (err) {
let reason = err.message;
try {
reason = await err.text();
} catch {}
results.tts = {status: 'fail', reason};
if (cred.use_for_tts) {
results.tts = {status: 'fail', reason};
}
if (cred.use_for_stt) {
results.stt = {status: 'fail', reason};
}
SpeechCredential.ttsTestResult(sid, false);
}
}