ben.carrasco/jambonz-feature-server
1
0
Fork 0
mirror of https://github.com/jambonz/jambonz-feature-server.git synced 2026-02-15 10:49:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

Speech vendor/cobalt (#463)

Browse Source 
* initial changes for cobalt speech

* wip

* wip

* update to drachtio-fsmrf that supports cobalt

* update to verb-specifications with cobalt speech support

* more wip

* lint

* use node 18 with gh actions

* support for compiling cobalt hints

* fix bug in uuid_cobalt_compile_context

* update verb-specifications

* remove repeated code

* cobalt support for transcribe

* update to verb specs
This commit is contained in:
Dave Horton
2023-09-13 09:47:30 -04:00
committed by GitHub
parent d220733dea
commit a1793ac359
11 changed files with 204 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
lib/tasks/transcribe.js
View File

@@ -7,6 +7,7 @@ const {
AzureTranscriptionEvents,
DeepgramTranscriptionEvents,
SonioxTranscriptionEvents,
CobaltTranscriptionEvents,
IbmTranscriptionEvents,
NvidiaTranscriptionEvents,
JambonzTranscriptionEvents,
@@ -103,6 +104,12 @@ class TaskTranscribe extends SttTask {
}
}
/* when using cobalt model is required */
if (this.vendor === 'cobalt' && !this.data.recognizer.model) {
this.notifyError({ msg: 'ASR error', details:'Cobalt requires a model to be specified'});
throw new Error('Cobalt requires a model to be specified');
}
try {
await this._startTranscribing(cs, ep, 1);
if (this.separateRecognitionPerChannel && ep2) {
@@ -163,7 +170,9 @@ class TaskTranscribe extends SttTask {
}
}
async _startTranscribing(cs, ep, channel) {
async _setSpeechHandlers(cs, ep, channel) {
if (this._speechHandlersSet) return;
this._speechHandlersSet = true;
const opts = this.setChannelVarsForStt(this, this.sttCredentials, this.data.recognizer);
switch (this.vendor) {
case 'google':
@@ -216,6 +225,29 @@ class TaskTranscribe extends SttTask {
ep.addCustomEventListener(SonioxTranscriptionEvents.Transcription,
this._onTranscription.bind(this, cs, ep, channel));
break;
case 'cobalt':
this.bugname = 'cobalt_transcribe';
ep.addCustomEventListener(CobaltTranscriptionEvents.Transcription,
this._onTranscription.bind(this, cs, ep, channel));
/* special case: if using hints with cobalt we need to compile them */
if (this.vendor === 'cobalt' && opts.COBALT_SPEECH_HINTS) {
try {
const context = await this.compileHintsForCobalt(
ep,
opts.COBALT_SERVER_URI,
this.data.recognizer.model,
opts.COBALT_CONTEXT_TOKEN,
opts.COBALT_SPEECH_HINTS
);
if (context) opts.COBALT_COMPILED_CONTEXT_DATA = context;
delete opts.COBALT_SPEECH_HINTS;
} catch (err) {
this.logger.error({err}, 'Error compiling hints for cobalt');
}
}
break;
case 'ibm':
this.bugname = 'ibm_transcribe';
ep.addCustomEventListener(IbmTranscriptionEvents.Transcription,
@@ -237,6 +269,7 @@ class TaskTranscribe extends SttTask {
ep.addCustomEventListener(NvidiaTranscriptionEvents.VadDetected,
this._onVadDetected.bind(this, cs, ep));
break;
default:
if (this.vendor.startsWith('custom:')) {
this.bugname = `${this.vendor}_transcribe`;
@@ -258,7 +291,10 @@ class TaskTranscribe extends SttTask {
ep.addCustomEventListener(JambonzTranscriptionEvents.Error, this._onJambonzError.bind(this, cs, ep));
await ep.set(opts)
.catch((err) => this.logger.info(err, 'Error setting channel variables'));
}
async _startTranscribing(cs, ep, channel) {
await this._setSpeechHandlers(cs, ep, channel);
await this._transcribe(ep);
/* start child span for this channel */