🐛 fix: Fix client fetch

api/open-stt.ts

@@ -15,5 +15,9 @@ export default async (req: Request) => {
   if (!OPENAI_API_KEY) return new Response('OPENAI_API_KEY is not set', { status: 500 });
   const openai = new OpenAI({ apiKey: OPENAI_API_KEY, baseURL: OPENAI_PROXY_URL });
   const res = await createOpenaiAudioTranscriptionsCompletion({ openai, payload });
-  return res;
+  return new Response(JSON.stringify(res), {
+    headers: {
+      'content-type': 'application/json;charset=UTF-8',
+    },
+  });
 };

src/const/api.ts

@@ -1,3 +1,5 @@
+import urlJoin from 'url-join';
+
 export const MICROSOFT_SPEECH_URL =
   'https://southeastasia.api.speech.microsoft.com/accfreetrial/texttospeech/acc/v3.0-beta1/vcg/speak';
 export const EDGE_SPEECH_URL =
@@ -19,3 +21,7 @@ export const OPENAI_PROXY_URL =
   process.env.OPENAI_PROXY_URL ||
   process.env.NEXT_PUBLIC_OPENAI_PROXY_URL ||
   'https://api.openai.com/v1';
+
+export const OPENAI_TTS_URL = (api?: string) => urlJoin(api || OPENAI_PROXY_URL, 'audio/speech');
+export const OPENAI_STT_URL = (api?: string) =>
+  urlJoin(api || OPENAI_PROXY_URL, 'audio/transcriptions');

src/server/createOpenaiAudioTranscriptionsCompletion.ts

@@ -25,5 +25,5 @@ export const createOpenaiAudioTranscriptionsCompletion = async ({
     { headers: { Accept: '*/*' } },
   );
 
-  return response.text;
+  return response;
 };

src/services/fetchEdgeSpeech.ts

@@ -19,6 +19,10 @@ export const fetchEdgeSpeech = async (
     ? fetch(api.url, { body: JSON.stringify(payload), method: 'POST' })
     : await createEdgeSpeechComletion({ payload }));
 
+  if (!response.ok) {
+    throw new Error('Network response was not ok');
+  }
+
   const arrayBuffer = await response.arrayBuffer();
   const audioBuffer = await arrayBufferConvert(arrayBuffer);
   return audioBuffer;

src/services/fetchOpenaiSTT.ts

@@ -1,10 +1,18 @@
-import OpenAI from 'openai';
-
-import { OPENAI_API_KEY, OPENAI_PROXY_URL } from '@/const/api';
-import { createOpenaiAudioTranscriptionsCompletion } from '@/server/createOpenaiAudioTranscriptionsCompletion';
+import { OPENAI_API_KEY, OPENAI_PROXY_URL, OPENAI_STT_URL } from '@/const/api';
 import { OpenAISTTPayload } from '@/server/types';
 import { RecordMineType, getRecordMineType } from '@/utils/getRecordMineType';
 
+const genSTTBody = ({ blob, options }: OpenAISTTPayload) => {
+  const filename = `${Date.now()}.${options.mineType.extension}`;
+  const file = new File([blob], filename, {
+    type: options.mineType.mineType,
+  });
+
+  const body = new FormData();
+  body.append('file', file);
+  body.append('model', options.model);
+  return body;
+};
 export interface OpenaiSttOptions {
   api?: {
     key?: string;
@@ -14,8 +22,6 @@ export interface OpenaiSttOptions {
   mineType?: RecordMineType;
   model?: 'whisper-1';
 }
-
-// 纯文本生成语音
 export const fetchOpenaiSTT = async (
   speech: Blob,
   { api = {}, model = 'whisper-1', mineType }: OpenaiSttOptions,
@@ -31,12 +37,21 @@ export const fetchOpenaiSTT = async (
     },
   };
 
-  const response = (await (api?.url
+  const response = await (api?.url
     ? fetch(api.url, { body: JSON.stringify(payload), method: 'POST' })
-    : createOpenaiAudioTranscriptionsCompletion({
-        openai: new OpenAI({ apiKey: key, baseURL: url }),
-        payload,
-      }))) as string;
+    : fetch(OPENAI_STT_URL(url), {
+        body: genSTTBody(payload),
+        headers: new Headers({
+          Authorization: `Bearer ${key}`,
+        }),
+        method: 'POST',
+      }));
+
+  if (!response.ok) {
+    throw new Error('Network response was not ok');
+  }
+
+  const json = await response.json();
 
-  return response;
+  return json.text;
 };

src/services/fetchOpenaiTTS.ts

@@ -1,7 +1,4 @@
-import OpenAI from 'openai';
-
-import { OPENAI_API_KEY, OPENAI_PROXY_URL } from '@/const/api';
-import { createOpenaiAudioSpeechCompletion } from '@/server/createOpenaiAudioSpeechCompletion';
+import { OPENAI_API_KEY, OPENAI_PROXY_URL, OPENAI_TTS_URL } from '@/const/api';
 import { OpenAITTSPayload } from '@/server/types';
 import { arrayBufferConvert } from '@/utils/arrayBufferConvert';
 import { type SsmlOptions } from '@/utils/genSSML';
@@ -34,9 +31,17 @@ export const fetchOpenaiTTS = async (
 
   const response = await (api?.url
     ? fetch(api.url, { body: JSON.stringify(payload), method: 'POST' })
-    : await createOpenaiAudioSpeechCompletion({
-        openai: new OpenAI({ apiKey: key, baseURL: url }),
-        payload,
+    : fetch(OPENAI_TTS_URL(url), {
+        body: JSON.stringify({
+          input,
+          model,
+          voice,
+        }),
+        headers: new Headers({
+          'Authorization': `Bearer ${key}`,
+          'Content-Type': 'application/json',
+        }),
+        method: 'POST',
       }));
 
   if (!response.ok) {