ben.carrasco/jambonz-feature-server
1
0
Fork 0
mirror of https://github.com/jambonz/jambonz-feature-server.git synced 2025-12-20 08:40:38 +00:00
Code Issues Packages Projects Releases Wiki Activity

Feature/ibm watson (#193)

Browse Source 
* initial changes to support ibm watson

* update specs.json for ibm

* update to drachtio-fsmrf with support for ibm

* bugfix: set access token for ibm stt, not api_key

* fix name of api_key

* normalize ibm transcription results

* rework ibm credentials

* bugfix setting runtime speech creds

* bugfix: ibm region

* typo

* changes to transcribe for ibm watson

* implement connect handler

* bugfix: bind error

* proper use of result_index

* ibm error handling
This commit is contained in:
Dave Horton
2022-11-21 22:09:37 -05:00
committed by GitHub
parent 8686348454
commit 71a2435c63
9 changed files with 217 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

72
lib/utils/transcription-utils.js
View File

@@ -27,6 +27,23 @@ const normalizeDeepgram = (evt, channel, language) => {
};
};
const normalizeIbm = (evt, channel, language) => {
const copy = JSON.parse(JSON.stringify(evt));
//const idx = evt.result_index;
const result = evt.results[0];
return {
language_code: language,
channel_tag: channel,
is_final: result.final,
alternatives: result.alternatives,
vendor: {
name: 'ibm',
evt: copy
}
};
};
const normalizeGoogle = (evt, channel, language) => {
const copy = JSON.parse(JSON.stringify(evt));
return {
@@ -113,6 +130,8 @@ module.exports = (logger) => {
return normalizeAws(evt, channel, language);
case 'nuance':
return normalizeNuance(evt, channel, language);
case 'ibm':
return normalizeIbm(evt, channel, language);
default:
logger.error(`Unknown vendor ${vendor}`);
return evt;
@@ -311,6 +330,31 @@ module.exports = (logger) => {
{DEEPGRAM_SPEECH_VAD_TURNOFF: deepgramOptions.tag}
};
}
else if ('ibm' === rOpts.vendor) {
const {ibmOptions = {}} = rOpts;
opts = {
...opts,
...(sttCredentials.access_token) &&
{IBM_ACCESS_TOKEN: sttCredentials.access_token},
...(sttCredentials.stt_region) &&
{IBM_SPEECH_REGION: sttCredentials.stt_region},
...(sttCredentials.instance_id) &&
{IBM_SPEECH_INSTANCE_ID: sttCredentials.instance_id},
...(ibmOptions.model) &&
{IBM_SPEECH_MODEL: ibmOptions.model},
...(ibmOptions.language_customization_id) &&
{IBM_SPEECH_LANGUAGE_CUSTOMIZATION_ID: ibmOptions.language_customization_id},
...(ibmOptions.acoustic_customization_id) &&
{IBM_SPEECH_ACOUSTIC_CUSTOMIZATION_ID: ibmOptions.acoustic_customization_id},
...(ibmOptions.baseModelVersion) &&
{IBM_SPEECH_BASE_MODEL_VERSION: ibmOptions.baseModelVersion},
...(ibmOptions.watsonMetadata) &&
{IBM_SPEECH_WATSON_METADATA: ibmOptions.watsonMetadata},
...(ibmOptions.watsonLearningOptOut) &&
{IBM_SPEECH_WATSON_LEARNING_OPT_OUT: ibmOptions.watsonLearningOptOut}
};
}
logger.debug({opts}, 'recognizer channel vars');
return opts;
};
@@ -336,12 +380,34 @@ module.exports = (logger) => {
ep.removeCustomEventListener(DeepgramTranscriptionEvents.Transcription);
ep.removeCustomEventListener(DeepgramTranscriptionEvents.Connect);
ep.removeCustomEventListener(DeepgramTranscriptionEvents.ConnectFailure);
};
const setSpeechCredentialsAtRuntime = (recognizer) => {
if (!recognizer) return;
if (recognizer.vendor === 'nuance') {
const {clientId, secret} = recognizer.nuanceOptions;
if (clientId && secret) return {client_id: clientId, secret};
}
else if (recognizer.vendor === 'deepgram') {
const {apiKey} = recognizer.deepgramOptions;
if (apiKey) return {api_key: apiKey};
}
else if (recognizer.vendor === 'ibm') {
const {ttsApiKey, ttsRegion, sttApiKey, sttRegion, instanceId} = recognizer.ibmOptions;
if (ttsApiKey || sttApiKey) return {
tts_api_key: ttsApiKey,
tts_region: ttsRegion,
stt_api_key: sttApiKey,
stt_region: sttRegion,
instance_id: instanceId
};
}
};
return {
normalizeTranscription,
setChannelVarsForStt,
removeSpeechListeners
removeSpeechListeners,
setSpeechCredentialsAtRuntime
};
};