support fetch tts/stt deepgram models from rest api (#457)

lib/routes/api/speech-credentials.js

@@ -215,7 +215,8 @@ const encryptCredential = (obj) => {
       if (!deepgram_stt_uri || !deepgram_tts_uri) {
         assert(api_key, 'invalid deepgram speech credential: api_key is required');
       }
-      const deepgramData = JSON.stringify({api_key, deepgram_stt_uri, deepgram_stt_use_tls, deepgram_tts_uri});
+      const deepgramData = JSON.stringify({api_key, deepgram_stt_uri,
+        deepgram_stt_use_tls, deepgram_tts_uri, model_id});
       return encrypt(deepgramData);
 
     case 'ibm':

lib/utils/speech-data/stt-model-deepgram.js

@@ -0,0 +1,52 @@
+module.exports = [
+  // Nova-3
+  { name: 'Nova-3', value: 'nova-3' },
+  { name: 'Nova-3 General', value: 'nova-3-general' },
+  { name: 'Nova-3 Medical', value: 'nova-3-medical' },
+
+  // Nova-2
+  { name: 'Nova-2', value: 'nova-2' },
+  { name: 'Nova-2 General', value: 'nova-2-general' },
+  { name: 'Nova-2 Meeting', value: 'nova-2-meeting' },
+  { name: 'Nova-2 Phonecall', value: 'nova-2-phonecall' },
+  { name: 'Nova-2 Finance', value: 'nova-2-finance' },
+  { name: 'Nova-2 Conversational AI', value: 'nova-2-conversationalai' },
+  { name: 'Nova-2 Voicemail', value: 'nova-2-voicemail' },
+  { name: 'Nova-2 Video', value: 'nova-2-video' },
+  { name: 'Nova-2 Medical', value: 'nova-2-medical' },
+  { name: 'Nova-2 Drivethru', value: 'nova-2-drivethru' },
+  { name: 'Nova-2 Automotive', value: 'nova-2-automotive' },
+  { name: 'Nova-2 ATC', value: 'nova-2-atc' },
+
+  // Nova (legacy)
+  { name: 'Nova', value: 'nova' },
+  { name: 'Nova General', value: 'nova-general' },
+  { name: 'Nova Phonecall', value: 'nova-phonecall' },
+  { name: 'Nova Medical', value: 'nova-medical' },
+
+  // Enhanced (legacy)
+  { name: 'Enhanced', value: 'enhanced' },
+  { name: 'Enhanced General', value: 'enhanced-general' },
+  { name: 'Enhanced Meeting', value: 'enhanced-meeting' },
+  { name: 'Enhanced Phonecall', value: 'enhanced-phonecall' },
+  { name: 'Enhanced Finance', value: 'enhanced-finance' },
+
+  // Base (legacy)
+  { name: 'Base', value: 'base' },
+  { name: 'Base General', value: 'base-general' },
+  { name: 'Base Meeting', value: 'base-meeting' },
+  { name: 'Base Phonecall', value: 'base-phonecall' },
+  { name: 'Base Finance', value: 'base-finance' },
+  { name: 'Base Conversational AI', value: 'base-conversationalai' },
+  { name: 'Base Voicemail', value: 'base-voicemail' },
+  { name: 'Base Video', value: 'base-video' },
+
+  // Whisper
+  { name: 'Whisper Tiny', value: 'whisper-tiny' },
+  { name: 'Whisper Base', value: 'whisper-base' },
+  { name: 'Whisper Small', value: 'whisper-small' },
+  { name: 'Whisper Medium', value: 'whisper-medium' },
+  { name: 'Whisper Large', value: 'whisper-large' },
+  { name: 'Whisper', value: 'whisper' },
+];
+

lib/utils/speech-utils.js

@@ -48,6 +48,12 @@ const SttOpenaiLanguagesVoices = require('./speech-data/stt-openai');
 
 
 const SttModelOpenai = require('./speech-data/stt-model-openai');
+const sttModelDeepgram = require('./speech-data/stt-model-deepgram');
+
+function capitalizeFirst(str) {
+  if (!str) return str;
+  return str.charAt(0).toUpperCase() + str.slice(1);
+}
 
 
 const testSonioxStt = async(logger, credentials) => {
@@ -636,6 +642,7 @@ function decryptCredential(obj, credential, logger, isObscureKey = true) {
     obj.deepgram_stt_uri = o.deepgram_stt_uri;
     obj.deepgram_stt_use_tls = o.deepgram_stt_use_tls;
     obj.deepgram_tts_uri = o.deepgram_tts_uri;
+    obj.model_id = o.model_id;
   }
   else if ('ibm' === obj.vendor) {
     const o = JSON.parse(decrypt(credential));
@@ -851,8 +858,47 @@ async function getLanguagesVoicesForNuane(credential, getTtsVoices, logger) {
   return tranform(TtsNuanceLanguagesVoices, SttNuanceLanguagesVoices);
 }
 
-async function getLanguagesVoicesForDeepgram(credential) {
-  return tranform(TtsLanguagesDeepgram, SttDeepgramLanguagesVoices, TtsModelDeepgram);
+async function getLanguagesVoicesForDeepgram(credential, getTtsVoices, logger) {
+  if (credential) {
+    const {model_id, api_key, deepgram_stt_uri, deepgram_tts_uri} = credential;
+    // currently just fetching STT and TTS models from Deepgram cloud
+    if (!deepgram_stt_uri && !deepgram_tts_uri) {
+      const response = await fetch('https://api.deepgram.com/v1/models', {
+        headers: {
+          'Authorization': `Token ${api_key}`
+        }
+      });
+      if (!response.ok) {
+        logger.error({response}, 'Error fetching Deepgram voices');
+        throw new Error('failed to list voices');
+      }
+      const {stt, tts} = await response.json();
+      let sttLangs = SttDeepgramLanguagesVoices;
+      const sttModels = Array.from(
+        new Map(
+          stt.map((m) => [m.canonical_name, { name: capitalizeFirst(m.canonical_name), value: m.canonical_name }])
+        ).values()
+      ).sort((a, b) => a.name.localeCompare(b.name));
+      const ttsModels = Array.from(
+        new Map(
+          tts.map((m) => [m.canonical_name, { name: capitalizeFirst(m.canonical_name), value: m.canonical_name }])
+        ).values()
+      ).sort((a, b) => a.name.localeCompare(b.name));
+      // if model_id is not provided, return all models, all voices, all languages
+      if (!model_id) {
+        return tranform(TtsLanguagesDeepgram, sttLangs, ttsModels, sttModels);
+      }
+
+      const selectedSttModel = stt.find((m) => m.canonical_name === model_id);
+      const selectedSttLangs = selectedSttModel ? selectedSttModel.languages : [];
+      sttLangs = SttDeepgramLanguagesVoices.filter((l) => {
+        return selectedSttLangs.includes(l.value);
+      });
+      return tranform(TtsLanguagesDeepgram, sttLangs, ttsModels, sttModels);
+    }
+  }
+  return tranform(TtsLanguagesDeepgram, SttDeepgramLanguagesVoices,
+    TtsModelDeepgram, sttModelDeepgram.sort((a, b) => a.name.localeCompare(b.name)));
 }
 
 async function getLanguagesVoicesForIbm(credential, getTtsVoices, logger) {
@@ -1072,9 +1118,9 @@ async function getLanguagesVoicesForRimelabs(credential) {
     Object.keys(voices).length > 0 ? voices[Object.keys(voices)[0]] : [];
   const ttsVoices = Object.entries(modelVoices).map(([key, voices]) => ({
     value: key,
-    name: key.charAt(0).toUpperCase() + key.slice(1),
+    name: capitalizeFirst(key),
     voices: voices.map((v) => ({
-      name: v.charAt(0).toUpperCase() + v.slice(1),
+      name: capitalizeFirst(v),
       value: v
     }))
   }));