Add deepgram as STT (#151)

* add deepgram as stt

* add all languages

* remove unused types

* temporary remove Flemish

* apply review comments

Co-authored-by: eglehelms <e.helms@cognigy.com>

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

@@ -19,6 +19,7 @@ import {
   VENDOR_AWS,
   VENDOR_WELLSAID,
   useSpeechVendors,
+  VENDOR_DEEPGRAM,
 } from "src/vendor";
 import {
   postApplication,
@@ -368,7 +369,9 @@ export const ApplicationForm = ({ application }: ApplicationFormProps) => {
               id="synthesis_vendor"
               name="synthesis_vendor"
               value={synthVendor}
-              options={vendors}
+              options={vendors.filter(
+                (vendor) => vendor.value != VENDOR_DEEPGRAM
+              )}
               onChange={(e) => {
                 const vendor = e.target.value as keyof SynthesisVendors;
                 setSynthVendor(vendor);

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

@@ -24,6 +24,7 @@ import {
   VENDOR_MICROSOFT,
   VENDOR_NUANCE,
   VENDOR_WELLSAID,
+  VENDOR_DEEPGRAM,
 } from "src/vendor";
 import { MSG_REQUIRED_FIELDS } from "src/constants";
 import { getObscuredSecret } from "src/utils";
@@ -137,7 +138,9 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
           access_key_id: vendor === VENDOR_AWS ? accessKeyId : null,
           secret_access_key: vendor === VENDOR_AWS ? secretAccessKey : null,
           api_key:
-            vendor === VENDOR_MICROSOFT || vendor === VENDOR_WELLSAID
+            vendor === VENDOR_MICROSOFT ||
+            vendor === VENDOR_WELLSAID ||
+            vendor === VENDOR_DEEPGRAM
               ? apiKey
               : null,
           client_id: vendor === VENDOR_NUANCE ? clientId : null,
@@ -264,26 +267,30 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
         </fieldset>
         {vendor && (
           <fieldset>
-            <label htmlFor="use_for_tts" className="chk">
-              <input
-                id="use_for_tts"
-                name="use_for_tts"
-                type="checkbox"
-                onChange={(e) => setTtsCheck(e.target.checked)}
-                defaultChecked={ttsCheck}
-              />
-              <div>Use for text-to-speech</div>
-            </label>
-            <label htmlFor="use_for_stt" className="chk">
-              <input
-                id="use_for_stt"
-                name="use_for_stt"
-                type="checkbox"
-                onChange={(e) => setSttCheck(e.target.checked)}
-                defaultChecked={sttCheck}
-              />
-              <div>Use for speech-to-text</div>
-            </label>
+            {vendor !== VENDOR_DEEPGRAM && (
+              <label htmlFor="use_for_tts" className="chk">
+                <input
+                  id="use_for_tts"
+                  name="use_for_tts"
+                  type="checkbox"
+                  onChange={(e) => setTtsCheck(e.target.checked)}
+                  defaultChecked={ttsCheck}
+                />
+                <div>Use for text-to-speech</div>
+              </label>
+            )}
+            {vendor !== VENDOR_WELLSAID && (
+              <label htmlFor="use_for_stt" className="chk">
+                <input
+                  id="use_for_stt"
+                  name="use_for_stt"
+                  type="checkbox"
+                  onChange={(e) => setSttCheck(e.target.checked)}
+                  defaultChecked={sttCheck}
+                />
+                <div>Use for speech-to-text</div>
+              </label>
+            )}
           </fieldset>
         )}
         {vendor === VENDOR_GOOGLE && (
@@ -379,8 +386,9 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
             />
           </fieldset>
         )}
-        {/* Single API key state var is used for both Microsoft and WellSaid */}
-        {(vendor === VENDOR_MICROSOFT || vendor === VENDOR_WELLSAID) && (
+        {(vendor === VENDOR_MICROSOFT ||
+          vendor === VENDOR_WELLSAID ||
+          vendor === VENDOR_DEEPGRAM) && (
           <fieldset>
             <label htmlFor={`${vendor}_apikey`}>
               API Key<span>*</span>

src/vendor/index.tsx

@@ -14,6 +14,7 @@ export const VENDOR_AWS = "aws";
 export const VENDOR_MICROSOFT = "microsoft";
 export const VENDOR_WELLSAID = "wellsaid";
 export const VENDOR_NUANCE = "nuance";
+export const VENDOR_DEEPGRAM = "deepgram";
 
 export const vendors: VendorOptions[] = [
   {
@@ -36,6 +37,10 @@ export const vendors: VendorOptions[] = [
     name: "Nuance",
     value: VENDOR_NUANCE,
   },
+  {
+    name: "deepgram",
+    value: VENDOR_DEEPGRAM,
+  },
 ];
 
 export const useRegionVendors = () => {
@@ -78,6 +83,7 @@ export const useSpeechVendors = () => {
       import("./speech-recognizer/google-speech-recognizer-lang"),
       import("./speech-recognizer/ms-speech-recognizer-lang"),
       import("./speech-recognizer/nuance-speech-recognizer-lang"),
+      import("./speech-recognizer/deepgram-speech-recognizer-lang"),
       import("./speech-synthesis/aws-speech-synthesis-lang"),
       import("./speech-synthesis/google-speech-synthesis-lang"),
       import("./speech-synthesis/ms-speech-synthesis-lang"),
@@ -89,6 +95,7 @@ export const useSpeechVendors = () => {
         { default: googleRecognizer },
         { default: msRecognizer },
         { default: nuanceRecognizer },
+        { default: deepgramRecognizer },
         { default: awsSynthesis },
         { default: googleSynthesis },
         { default: msSynthesis },
@@ -109,6 +116,7 @@ export const useSpeechVendors = () => {
               google: googleRecognizer,
               microsoft: msRecognizer,
               nuance: nuanceRecognizer,
+              deepgram: deepgramRecognizer,
             },
           });
         }

src/vendor/speech-recognizer/deepgram-speech-recognizer-lang.ts

@@ -0,0 +1,126 @@
+import type { Language } from "../types";
+
+export const languages: Language[] = [
+  {
+    name: "Chinese - general",
+    code: "zh",
+  },
+  {
+    name: "Chinese (China)",
+    code: "zh-CN",
+  },
+  {
+    name: "Chinese (Taiwan)",
+    code: "zh-TW",
+  },
+  {
+    name: "Dutch - general",
+    code: "nl",
+  },
+  {
+    name: "English - general",
+    code: "en",
+  },
+  {
+    name: "English (Australia)",
+    code: "en-AU",
+  },
+  {
+    name: "English (United Kingdom)",
+    code: "en-GB",
+  },
+  {
+    name: "English (India)",
+    code: "en-IN",
+  },
+  {
+    name: "English (New Zealand)",
+    code: "en-NZ",
+  },
+  {
+    name: "English (United States)",
+    code: "en-US",
+  },
+  {
+    name: "French - general",
+    code: "fr",
+  },
+  {
+    name: "French (Canada)",
+    code: "fr-CA",
+  },
+  {
+    name: "German - general",
+    code: "de",
+  },
+  {
+    name: "Hindi - general",
+    code: "hi",
+  },
+  {
+    name: "Hindi (Roman Script)",
+    code: "hi-Latin",
+  },
+  {
+    name: "Indonesian - general",
+    code: "in",
+  },
+  {
+    name: "Italian - general",
+    code: "it",
+  },
+  {
+    name: "Japanese - general",
+    code: "ja",
+  },
+  {
+    name: "Korean - general",
+    code: "ko",
+  },
+  {
+    name: "Norwegian - general",
+    code: "no",
+  },
+  {
+    name: "Polish - general",
+    code: "pl",
+  },
+  {
+    name: "Portuguese - general",
+    code: "pt",
+  },
+  {
+    name: "Portuguese (Brazil)",
+    code: "pt-BR",
+  },
+  {
+    name: "Portuguese (Portugal)",
+    code: "pt-PT",
+  },
+  {
+    name: "Russian - general",
+    code: "ru",
+  },
+  {
+    name: "Spanish - general",
+    code: "es",
+  },
+  {
+    name: "Spanish (Latin America)",
+    code: "es-419",
+  },
+  {
+    name: "Swedish - general",
+    code: "sv",
+  },
+  {
+    name: "Turkish - general",
+    code: "tr",
+  },
+  {
+    name: "Ukrainian - general",
+    code: "uk",
+  },
+];
+
+export default languages;

src/vendor/types.ts

@@ -1,4 +1,10 @@
-export type Vendor = "Google" | "AWS" | "Microsoft" | "WellSaid" | "Nuance";
+export type Vendor =
+  | "Google"
+  | "AWS"
+  | "Microsoft"
+  | "WellSaid"
+  | "Nuance"
+  | "deepgram";
 
 export interface VendorOptions {
   name: Vendor;
@@ -50,6 +56,7 @@ export interface RecognizerVendors {
   google: Language[];
   microsoft: Language[];
   nuance: Language[];
+  deepgram: Language[];
 }
 
 export interface SynthesisVendors {