may config tts baseURL

enjoy/src/i18n/en.json

@@ -84,6 +84,7 @@
       "ttsEngine": "TTS engine",
       "ttsModel": "TTS model",
       "ttsVoice": "TTS voice",
+      "ttsBaseUrl": "TTS base URL",
       "notFound": "Conversation not found",
       "contentRequired": "Content required"
     },

enjoy/src/i18n/zh-CN.json

@@ -84,6 +84,7 @@
       "ttsEngine": "TTS 引擎",
       "ttsModel": "TTS 模型",
       "ttsVoice": "TTS 声音",
+      "ttsBaseUrl": "TTS 请求地址",
       "notFound": "未找到对话",
       "contentRequired": "对话内容不能为空"
     },

enjoy/src/main/db/models/speech.ts

@@ -161,6 +161,7 @@ export class Speech extends Model<Speech> {
       engine = "openai",
       model = "tts-1",
       voice = "alloy",
+      baseUrl,
     } = configuration || {};
 
     logger.debug("Generating speech", { engine, model, voice });
@@ -176,7 +177,9 @@ export class Speech extends Model<Speech> {
       }
       const openai = new OpenAI({
         apiKey: key,
+        baseURL: baseUrl,
       });
+      logger.debug("baseURL", openai.baseURL);
 
       const file = await openai.audio.speech.create({
         input: text,

enjoy/src/renderer/components/conversations/conversation-form.tsx

@@ -53,6 +53,7 @@ const conversationFormSchema = z.object({
           engine: z.enum(["openai"]).default("openai"),
           model: z.string().default("tts-1"),
           voice: z.string().optional(),
+          baseUrl: z.string().optional(),
         })
         .optional(),
     })
@@ -102,6 +103,9 @@ export const ConversationForm = (props: {
           engine: conversation.engine,
           configuration: {
             ...conversation.configuration,
+            tts: {
+              ...conversation.configuration?.tts,
+            },
           },
         }
       : {
@@ -128,6 +132,10 @@ export const ConversationForm = (props: {
       configuration.baseUrl = LLM_PROVIDERS[engine]?.baseUrl;
     }
 
+    if (!configuration.tts.baseUrl) {
+      configuration.tts.baseUrl = LLM_PROVIDERS[engine]?.baseUrl;
+    }
+
     if (conversation?.id) {
       EnjoyApp.conversations
         .update(conversation.id, {
@@ -547,6 +555,21 @@ export const ConversationForm = (props: {
                 </FormItem>
               )}
             />
+
+            <FormField
+              control={form.control}
+              name="configuration.tts.baseUrl"
+              render={({ field }) => (
+                <FormItem>
+                  <FormLabel>{t("models.conversation.ttsBaseUrl")}</FormLabel>
+                  <Input {...field} />
+                  <FormDescription>
+                    {t("models.conversation.baseUrl")}
+                  </FormDescription>
+                  <FormMessage />
+                </FormItem>
+              )}
+            />
           </div>
         </ScrollArea>
 
@@ -584,7 +607,9 @@ export const ConversationForm = (props: {
           )}
 
           <Button
-            disabled={submitting || (conversation.id && !form.formState.isDirty)}
+            disabled={
+              submitting || (conversation.id && !form.formState.isDirty)
+            }
             className="w-full h-12"
             size="lg"
             type="submit"
@@ -688,6 +713,7 @@ const conversationDefaultConfiguration = {
     frequencyPenalty: 0,
     historyBufferSize: 0,
     tts: {
+      baseUrl: "",
       engine: "openai",
       model: "tts-1",
       voice: "alloy",

enjoy/src/renderer/components/messages/assistant-message.tsx

@@ -54,6 +54,7 @@ export const AssistantMessageComponent = (props: {
         engine: configuration?.tts?.engine,
         model: configuration?.tts?.model,
         voice: configuration?.tts?.voice,
+        baseUrl: configuration?.tts?.baseUrl,
       })
       .then((speech) => {
         setSpeech(speech);