support deepgram stt model (#528)

* support deepgram stt model

* wip

* wip

src/containers/internal/views/applications/form.tsx

@@ -922,6 +922,9 @@ export const ApplicationForm = ({ application }: ApplicationFormProps) => {
           serviceProviderSid={
             currentServiceProvider?.service_provider_sid || ""
           }
+          application_speech_synthesis_voice={
+            application?.data?.speech_synthesis_voice
+          }
           accountSid={accountSid}
           credentials={credentials}
           ttsVendor={[synthVendor, setSynthVendor]}
@@ -952,6 +955,9 @@ export const ApplicationForm = ({ application }: ApplicationFormProps) => {
                 currentServiceProvider?.service_provider_sid || ""
               }
               accountSid={accountSid}
+              application_speech_synthesis_voice={
+                application?.data?.fallback_speech_synthesis_voice
+              }
               credentials={credentials}
               ttsVendor={[
                 fallbackSpeechSynthsisVendor,

src/containers/internal/views/applications/speech-selection.tsx

@@ -44,6 +44,7 @@ import {
 type SpeechProviderSelectionProbs = {
   accountSid: string;
   serviceProviderSid: string;
+  application_speech_synthesis_voice: string | null | undefined;
   credentials: SpeechCredential[] | undefined;
   ttsVendor: [
     keyof SynthesisVendors,
@@ -67,6 +68,7 @@ type SpeechProviderSelectionProbs = {
 export const SpeechProviderSelection = ({
   accountSid,
   serviceProviderSid,
+  application_speech_synthesis_voice,
   credentials,
   ttsVendor: [synthVendor, setSynthVendor],
   ttsVendorOptions,
@@ -242,10 +244,6 @@ export const SpeechProviderSelection = ({
         // Extract model
         if (json.models && json.models.length) {
           setSynthesisModelOptions(json.models);
-          if (synthVendor === VENDOR_DEEPGRAM) {
-            setSynthVoice(json.models[0].value);
-            return;
-          }
         }
 
         if (json.tts && json.tts.length) {
@@ -337,6 +335,7 @@ export const SpeechProviderSelection = ({
 
   const updateTtsVoice = (language: string, voice: string) => {
     if (shouldUpdateTtsVoice.current) {
+      console.log("xhoaluu");
       setSynthLang(language);
       setSynthVoice(voice);
       shouldUpdateTtsVoice.current = false;
@@ -387,6 +386,20 @@ export const SpeechProviderSelection = ({
         toastError(error.msg);
       });
   };
+
+  useEffect(() => {
+    if (
+      synthVendor === VENDOR_DEEPGRAM &&
+      synthesisModelOptions.length > 0 &&
+      !synthesisModelOptions.some(
+        (m) => m.value === application_speech_synthesis_voice,
+      )
+    ) {
+      setSynthVoice(synthesisModelOptions[0].value);
+    } else {
+      setSynthVoice(application_speech_synthesis_voice || "");
+    }
+  }, [synthesisModelOptions, application_speech_synthesis_voice]);
   return (
     <>
       <fieldset>

src/containers/internal/views/speech-services/form.tsx

@@ -259,6 +259,7 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
       case VENDOR_PLAYHT:
         return "Voice Engine";
       case VENDOR_CARTESIA:
+      case VENDOR_DEEPGRAM:
         return "Model ID";
       default:
         return "Model";
@@ -436,6 +437,7 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
           deepgram_stt_uri: deepgramSttUri || null,
           deepgram_tts_uri: deepgramTtsUri || null,
           deepgram_stt_use_tls: deepgramSttUseTls ? 1 : 0,
+          model_id: sttModelId || null,
         }),
         ...(vendor === VENDOR_SPEECHMATICS && {
           speechmatics_stt_uri: speechmaticsEndpoint || null,
@@ -562,7 +564,8 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
       vendor === VENDOR_PLAYHT ||
       vendor === VENDOR_RIMELABS ||
       vendor === VENDOR_CARTESIA ||
-      vendor === VENDOR_OPENAI
+      vendor === VENDOR_OPENAI ||
+      vendor === VENDOR_DEEPGRAM
     ) {
       getSpeechSupportedLanguagesAndVoices(
         currentServiceProvider?.service_provider_sid,
@@ -572,21 +575,9 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
       ).then(({ json }) => {
         if (json.models) {
           setTtsModels(json.models);
-          if (
-            json.models.length > 0 &&
-            !json.models.find((m) => m.value === ttsModelId)
-          ) {
-            setTtsModelId(json.models[0].value);
-          }
         }
         if (json.sttModels) {
           setSttModels(json.sttModels);
-          if (
-            json.sttModels.length > 0 &&
-            !json.sttModels.some((m) => m.value === sttModelId)
-          ) {
-            setSttModelId(json.sttModels[0].value);
-          }
         }
       });
     } else {
@@ -594,6 +585,24 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
     }
   }, [vendor]);
 
+  useEffect(() => {
+    const modelId = credential?.data?.model_id || "";
+    if (sttModels.length > 0 && !sttModels.some((m) => m.value === modelId)) {
+      setSttModelId(sttModels[0].value);
+    } else {
+      setSttModelId(modelId);
+    }
+  }, [credential, sttModels]);
+
+  useEffect(() => {
+    const modelId = credential?.data?.model_id || "";
+    if (ttsModels.length > 0 && !ttsModels.some((m) => m.value === modelId)) {
+      setTtsModelId(sttModels[0].value);
+    } else {
+      setTtsModelId(modelId);
+    }
+  }, [credential, ttsModels]);
+
   useEffect(() => {
     setLocation();
     if (credential && credential.data) {
@@ -742,7 +751,10 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
       if (credential.data.model_id) {
         setTtsModelId(credential.data.model_id);
       }
-      if (credential.data.model_id && vendor === VENDOR_OPENAI) {
+      if (
+        credential.data.model_id &&
+        (vendor === VENDOR_OPENAI || vendor === VENDOR_DEEPGRAM)
+      ) {
         setSttModelId(credential.data.model_id);
       }
       if (credential?.data?.playht_tts_uri) {
@@ -1717,22 +1729,23 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
               />
             </fieldset>
           )}
-        {vendor == VENDOR_OPENAI && sttModels.length > 0 && (
+        {(vendor == VENDOR_OPENAI || vendor === VENDOR_DEEPGRAM) &&
-          <fieldset>
+          sttModels.length > 0 && (
-            <label htmlFor={`${vendor}_stt_model_id`}>
+            <fieldset>
-              {getModelLabelByVendor(vendor)}
+              <label htmlFor={`${vendor}_stt_model_id`}>
-            </label>
+                {getModelLabelByVendor(vendor)}
-            <Selector
+              </label>
-              id={"stt_model_id"}
+              <Selector
-              name={"stt_model_id"}
+                id={"stt_model_id"}
-              value={sttModelId}
+                name={"stt_model_id"}
-              options={sttModels}
+                value={sttModelId}
-              onChange={(e) => {
+                options={sttModels}
-                setSttModelId(e.target.value);
+                onChange={(e) => {
-              }}
+                  setSttModelId(e.target.value);
-            />
+                }}
-          </fieldset>
+              />
-        )}
+            </fieldset>
+          )}
         {(vendor === VENDOR_ELEVENLABS ||
           vendor === VENDOR_PLAYHT ||
           vendor === VENDOR_CARTESIA ||