From f1f8a7d808d7b3739becf64b50b26738847e46a5 Mon Sep 17 00:00:00 2001
From: Hoan Luu Huu <110280845+xquanluu@users.noreply.github.com>
Date: Wed, 13 Aug 2025 19:17:18 +0700
Subject: [PATCH] support resemble tts (#559)

* support resemble tts

* wip
---
 src/api/types.ts                              |  2 +
 .../views/applications/speech-selection.tsx   |  2 +
 .../internal/views/speech-services/form.tsx   | 77 +++++++++++++++++++
 src/vendor/index.tsx                          |  5 ++
 src/vendor/types.ts                           |  3 +-
 5 files changed, 88 insertions(+), 1 deletion(-)

diff --git a/src/api/types.ts b/src/api/types.ts
index 6089461..2d02cf1 100644
--- a/src/api/types.ts
+++ b/src/api/types.ts
@@ -440,6 +440,8 @@ export interface SpeechCredential {
   deepgram_stt_use_tls: number;
   speechmatics_stt_uri: null | string;
   playht_tts_uri: null | string;
+  resemble_tts_uri: null | string;
+  resemble_tts_use_tls: number;
 }
 
 export interface Alert {
diff --git a/src/containers/internal/views/applications/speech-selection.tsx b/src/containers/internal/views/applications/speech-selection.tsx
index f7c1aa3..fd103c6 100644
--- a/src/containers/internal/views/applications/speech-selection.tsx
+++ b/src/containers/internal/views/applications/speech-selection.tsx
@@ -37,6 +37,7 @@ import {
   VENDOR_OPENAI,
   VENDOR_INWORLD,
   VENDOR_DEEPGRAM_RIVER,
+  VENDOR_RESEMBLE,
 } from "src/vendor";
 import {
   LabelOptions,
@@ -587,6 +588,7 @@ export const SpeechProviderSelection = ({
               vendor.value != VENDOR_WELLSAID &&
               vendor.value != VENDOR_ELEVENLABS &&
               vendor.value != VENDOR_WHISPER &&
+              vendor.value !== VENDOR_RESEMBLE &&
               vendor.value !== VENDOR_CUSTOM,
           )}
           onChange={(e) => {
diff --git a/src/containers/internal/views/speech-services/form.tsx b/src/containers/internal/views/speech-services/form.tsx
index 733fba9..10164d0 100644
--- a/src/containers/internal/views/speech-services/form.tsx
+++ b/src/containers/internal/views/speech-services/form.tsx
@@ -54,6 +54,7 @@ import {
   VENDOR_OPENAI,
   VENDOR_INWORLD,
   VENDOR_DEEPGRAM_RIVER,
+  VENDOR_RESEMBLE,
 } from "src/vendor";
 import { MSG_REQUIRED_FIELDS } from "src/constants";
 import {
@@ -207,6 +208,12 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
   const [tmpPlayhtTtsUri, setTmpPlayhtTtsUri] = useState("");
   const [initialPlayhtOnpremCheck, setInitialPlayhtOnpremCheck] =
     useState(false);
+  const [resembleTtsUri, setResembleTtsUri] = useState("");
+  const [tmpResembleTtsUri, setTmpResembleTtsUri] = useState("");
+  const [initialResembleOnpremCheck, setInitialResembleOnpremCheck] =
+    useState(false);
+  const [resembleTtsUseTls, setResembleTtsUseTls] = useState(false);
+  const [tmpResembleTtsUseTls, setTmpResembleTtsUseTls] = useState(false);
   const handleFile = (file: File) => {
     const handleError = () => {
       setGoogleServiceKey(null);
@@ -481,6 +488,10 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
         ...(vendor === VENDOR_PLAYHT && {
           playht_tts_uri: playhtTtsUri || null,
         }),
+        ...(vendor === VENDOR_RESEMBLE && {
+          resemble_tts_uri: resembleTtsUri || null,
+          resemble_tts_use_tls: resembleTtsUseTls ? 1 : 0,
+        }),
       };
 
       if (credential && credential.data) {
@@ -531,6 +542,7 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
               vendor === VENDOR_WHISPER ||
               vendor === VENDOR_CARTESIA ||
               vendor === VENDOR_OPENAI ||
+              vendor === VENDOR_RESEMBLE ||
               vendor === VENDOR_DEEPGRAM_RIVER
                 ? apiKey
                 : null,
@@ -798,6 +810,9 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
       if (credential?.data?.playht_tts_uri) {
         setPlayhtTtsUri(credential.data.playht_tts_uri);
       }
+      if (credential?.data?.resemble_tts_uri) {
+        setResembleTtsUri(credential.data.resemble_tts_uri);
+      }
     }
     if (credential?.data?.options) {
       setOptions(credential.data.options);
@@ -867,6 +882,15 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
       setSpeechmaticsEndpoint(credential.data.speechmatics_stt_uri);
     }
     setInitialPlayhtOnpremCheck(hasValue(credential?.data?.playht_tts_uri));
+    setInitialResembleOnpremCheck(hasValue(credential?.data?.resemble_tts_uri));
+    if (credential?.data?.resemble_tts_use_tls) {
+      setResembleTtsUseTls(
+        credential?.data?.resemble_tts_use_tls > 0 ? true : false,
+      );
+      setTmpResembleTtsUseTls(
+        credential?.data?.resemble_tts_use_tls > 0 ? true : false,
+      );
+    }
   }, [credential]);
 
   const updateCustomVoices = (
@@ -1005,6 +1029,7 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
               vendor !== VENDOR_PLAYHT &&
               vendor !== VENDOR_RIMELABS &&
               vendor !== VENDOR_INWORLD &&
+              vendor !== VENDOR_RESEMBLE &&
               vendor !== VENDOR_ELEVENLABS && (
                 <label htmlFor="use_for_stt" className="chk">
                   <input
@@ -1744,6 +1769,57 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
           </fieldset>
         )}
 
+        {vendor === VENDOR_RESEMBLE && (
+          <fieldset>
+            <Checkzone
+              disabled={hasValue(credential)}
+              hidden
+              name="use_on-prem_resemble_container"
+              label="Use on-prem Resemble container"
+              initialCheck={initialResembleOnpremCheck}
+              handleChecked={(e) => {
+                setInitialResembleOnpremCheck(e.target.checked);
+                if (e.target.checked) {
+                  if (tmpResembleTtsUri) {
+                    setResembleTtsUri(tmpResembleTtsUri);
+                  }
+                  if (tmpResembleTtsUseTls) {
+                    setResembleTtsUseTls(tmpResembleTtsUseTls);
+                  }
+                } else {
+                  setTmpResembleTtsUri(resembleTtsUri);
+                  setResembleTtsUri("");
+                  setTmpResembleTtsUseTls(resembleTtsUseTls);
+                  setResembleTtsUseTls(false);
+                }
+              }}
+            >
+              <label htmlFor="resemble_uri_for_tts">
+                TTS Container URI<span>*</span>
+              </label>
+              <input
+                id="resemble_uri_for_tts"
+                required
+                type="text"
+                name="resemble_uri_for_tts"
+                placeholder=""
+                value={resembleTtsUri}
+                onChange={(e) => setResembleTtsUri(e.target.value)}
+              />
+              <label htmlFor="resemble_stt_use_tls" className="chk">
+                <input
+                  id="resemble_stt_use_tls"
+                  name="resemble_stt_use_tls"
+                  type="checkbox"
+                  onChange={(e) => setResembleTtsUseTls(e.target.checked)}
+                  defaultChecked={resembleTtsUseTls}
+                />
+                <div>Use TLS</div>
+              </label>
+            </Checkzone>
+          </fieldset>
+        )}
+
         {(vendor === VENDOR_WELLSAID ||
           vendor === VENDOR_ASSEMBLYAI ||
           vendor === VENDOR_VOXIST ||
@@ -1755,6 +1831,7 @@ export const SpeechServiceForm = ({ credential }: SpeechServiceFormProps) => {
           vendor === VENDOR_CARTESIA ||
           vendor === VENDOR_OPENAI ||
           vendor === VENDOR_DEEPGRAM_RIVER ||
+          vendor === VENDOR_RESEMBLE ||
           vendor === VENDOR_SPEECHMATICS) && (
           <fieldset>
             <label htmlFor={`${vendor}_apikey`}>
diff --git a/src/vendor/index.tsx b/src/vendor/index.tsx
index 94c69c8..1785abf 100644
--- a/src/vendor/index.tsx
+++ b/src/vendor/index.tsx
@@ -29,6 +29,7 @@ export const VENDOR_INWORLD = "inworld";
 export const VENDOR_VERBIO = "verbio";
 export const VENDOR_CARTESIA = "cartesia";
 export const VENDOR_OPENAI = "openai";
+export const VENDOR_RESEMBLE = "resemble";
 
 export const vendors: VendorOptions[] = [
   {
@@ -123,6 +124,10 @@ export const vendors: VendorOptions[] = [
     name: "OpenAI",
     value: VENDOR_OPENAI,
   },
+  {
+    name: "Resemble",
+    value: VENDOR_RESEMBLE,
+  },
 ].sort((a, b) => a.name.localeCompare(b.name)) as VendorOptions[];
 
 export const AWS_CREDENTIAL_ACCESS_KEY = "access_key";
diff --git a/src/vendor/types.ts b/src/vendor/types.ts
index 3d86dc9..d2b3685 100644
--- a/src/vendor/types.ts
+++ b/src/vendor/types.ts
@@ -21,7 +21,8 @@ export type Vendor =
   | "inworld"
   | "verbio"
   | "openai"
-  | "Cartesia";
+  | "Cartesia"
+  | "Resemble";
 
 export interface VendorOptions {
   name: Vendor;