♻️ refactor: Refactor api

README.md

@@ -86,17 +86,17 @@ Click button below to deploy your private plugins' gateway.
 
 This project provides some additional configuration items set with environment variables:
 
-| Environment Variable         | Description                                                                                                                                             | Default                     |
-| ---------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------- |
-| `ALLOW_ORIGINS`              | Allow origins , string or string array                                                                                                                  |                             |
-| `OPENAI_API_KEY`             | This is the API key you apply on the OpenAI account page                                                                                                | `sk-xxxxxx...xxxxxx`        |
-| `OPENAI_PROXY_URL`           | If you manually configure the OpenAI interface proxy, you can use this configuration item to override the default OpenAI API request base URL           | `https://api.openai.com/v1` |
-| `AZURE_SPEECH_KEY`           | This is the API key of Azure Speech Service                                                                                                             |                             |
-| `AZURE_SPEECH_REGION`        | This is the region of Azure Speech Service                                                                                                              |                             |
-| `AZURE_SPEECH_PROXY_URL`     | If you manually configure the AZURE Speech interface proxy, you can use this configuration item to override the default Speech API request base URL     | `/api/azure-speech`         |
-| `MICROSOFT_SPEECH_PROXY_URL` | If you manually configure the Microsoft Speech interface proxy, you can use this configuration item to override the default Speech API request base URL | `/api/microsoft-speech`     |
-| `EDDGE_API_TOKEN`            | This is the API key of Edge Speech Service                                                                                                              |                             |
-| `EDDGE_PROXY_URL`            | If you manually configure the Edge interface proxy, you can use this configuration item to override the default Edge wss request base URL               |                             |
+| Environment Variable       | Description                                                                                                                                             | Default                     |
+| -------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------- |
+| `ALLOW_ORIGINS`            | Allow origins , string or string array                                                                                                                  |                             |
+| `OPENAI_API_KEY`           | This is the API key you apply on the OpenAI account page                                                                                                | `sk-xxxxxx...xxxxxx`        |
+| `OPENAI_PROXY_URL`         | If you manually configure the OpenAI interface proxy, you can use this configuration item to override the default OpenAI API request base URL           | `https://api.openai.com/v1` |
+| `AZURE_SPEECH_KEY`         | This is the API key of Azure Speech Service                                                                                                             |                             |
+| `AZURE_SPEECH_REGION`      | This is the region of Azure Speech Service                                                                                                              |                             |
+| `AZURE_SPEECH_PROXY_URL`   | If you manually configure the AZURE Speech interface proxy, you can use this configuration item to override the default Speech API request base URL     | `/api/azure-speech`         |
+| `MICROSOFT_SPEECH_API_URL` | If you manually configure the Microsoft Speech interface proxy, you can use this configuration item to override the default Speech API request base URL | `/api/microsoft-speech`     |
+| `EDGE_API_TOKEN`           | This is the API key of Edge Speech Service                                                                                                              |                             |
+| `EDGE_SPEECH_API_URL`      | If you manually configure the Edge interface proxy, you can use this configuration item to override the default Edge wss request base URL               |                             |
 
 <div align="right">
 

api/edge-speech.ts

@@ -0,0 +1,13 @@
+import { createEdgeSpeechComletion } from '../src/server/createEdgeSpeechComletion';
+import { EdgeSpeechPayload } from '../src/server/types';
+
+export const config = {
+  runtime: 'edge',
+};
+
+export default async (req: Request) => {
+  if (req.method !== 'POST') return new Response('Method Not Allowed', { status: 405 });
+  const payload = (await req.json()) as EdgeSpeechPayload;
+  const res = await createEdgeSpeechComletion({ payload });
+  return res;
+};

api/microsoft-speech.ts

@@ -1,15 +1,13 @@
-import cors from '../src/server/cors';
-import { getAllowOrigins } from '../src/server/getAllowOrigins';
-import { handleMicrosoftSpeechRequest } from '../src/server/handleMicrosoftSpeechRequest';
+import { createMicrosoftSpeechComletion } from '../src/server/createMicrosoftSpeechComletion';
+import { MicrosoftSpeechPayload } from '../src/server/types';
 
 export const config = {
   runtime: 'edge',
 };
 
 export default async (req: Request) => {
   if (req.method !== 'POST') return new Response('Method Not Allowed', { status: 405 });
-  const origin = getAllowOrigins(req);
-  if (!origin) return new Response('Origin Not Allowed', { status: 403 });
-  const res = await handleMicrosoftSpeechRequest(req);
-  return cors(req, new Response(res.body, res), { methods: ['POST'], origin });
+  const payload = (await req.json()) as MicrosoftSpeechPayload;
+  const res = await createMicrosoftSpeechComletion({ payload });
+  return res;
 };

api/open-stt.ts

@@ -0,0 +1,19 @@
+import OpenAI from 'openai';
+
+import { OPENAI_API_KEY, OPENAI_PROXY_URL } from '@/const/api';
+
+import { createOpenaiAudioTranscriptionsCompletion } from '../src/server/createOpenaiAudioTranscriptionsCompletion';
+import { OpenAISTTPayload } from '../src/server/types';
+
+export const config = {
+  runtime: 'edge',
+};
+
+export default async (req: Request) => {
+  if (req.method !== 'POST') return new Response('Method Not Allowed', { status: 405 });
+  const payload = (await req.json()) as OpenAISTTPayload;
+  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;
+};

api/openai-tts.ts

@@ -0,0 +1,19 @@
+import OpenAI from 'openai';
+
+import { OPENAI_API_KEY, OPENAI_PROXY_URL } from '@/const/api';
+
+import { createOpenaiAudioSpeechCompletion } from '../src/server/createOpenaiAudioSpeechCompletion';
+import { OpenAITTSPayload } from '../src/server/types';
+
+export const config = {
+  runtime: 'edge',
+};
+
+export default async (req: Request) => {
+  if (req.method !== 'POST') return new Response('Method Not Allowed', { status: 405 });
+  const payload = (await req.json()) as OpenAITTSPayload;
+  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 createOpenaiAudioSpeechCompletion({ openai, payload });
+  return res;
+};

package.json

@@ -66,11 +66,10 @@
     "antd-style": "^3",
     "lodash-es": "^4",
     "lucide-react": "latest",
-    "microsoft-cognitiveservices-speech-sdk": "^1",
+    "openai": "^4.17.3",
     "query-string": "^8",
     "react-error-boundary": "^4.0.11",
     "react-layout-kit": "^1",
-    "ssml-document": "^1",
     "swr": "^2",
     "url-join": "^5",
     "uuid": "^9"

src/const/api.ts

@@ -1,15 +1,14 @@
-import urlJoin from 'url-join';
-
-export const MICROSOFT_SPPECH_URL =
+export const MICROSOFT_SPEECH_URL =
   'https://southeastasia.api.speech.microsoft.com/accfreetrial/texttospeech/acc/v3.0-beta1/vcg/speak';
-export const MICROSOFT_SPEECH_PROXY_URL =
-  process.env.MICROSOFT_SPEECH_PROXY_URL ||
-  process.env.NEXT_PUBLIC_MICROSOFT_SPEECH_PROXY_URL ||
-  '/api/microsoft-speech';
-export const AZURE_SPEECH_PROXY_URL =
-  process.env.AZURE_SPEECH_PROXY_URL ||
-  process.env.NEXT_PUBLIC_AZURE_SPEECH_PROXY_URL ||
-  '/api/azure-speech';
+export const EDGE_SPEECH_URL =
+  'wss://speech.platform.bing.com/consumer/speech/synthesize/readaloud/edge/v1';
+export const EDGE_API_TOKEN = '6A5AA1D4EAFF4E9FB37E23D68491D6F4';
+
+export const MICROSOFT_SPEECH_API_URL = '/api/microsoft-speech';
+export const EDGE_SPEECH_API_URL = '/api/edge-speech';
+export const OPENAI_TTS_API_URL = '/api/openai-tts';
+export const OPENAI_STT_API_URL = '/api/openai-stt';
+
 export const AZURE_SPEECH_KEY =
   process.env.AZURE_SPEECH_KEY || process.env.NEXT_PUBLIC_AZURE_SPEECH_KEY || '';
 export const AZURE_SPEECH_REGION =
@@ -20,14 +19,3 @@ 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');
-export const EDDGE_PROXY_URL =
-  process.env.EDDGE_PROXY_URL ||
-  process.env.NEXT_PUBLIC_EDDGE_PROXY_UR ||
-  'wss://speech.platform.bing.com/consumer/speech/synthesize/readaloud/edge/v1';
-export const EDDGE_API_TOKEN =
-  process.env.EDDGE_API_TOKEN ||
-  process.env.NEXT_PUBLIC_EDDGE_API_TOKEN ||
-  '6A5AA1D4EAFF4E9FB37E23D68491D6F4';

src/index.ts

@@ -3,19 +3,17 @@ export { default as AudioVisualizer, type AudioVisualizerProps } from './AudioVi
 export { default as azureVoiceList } from './data/azureVoiceList';
 export { default as edgeVoiceList } from './data/edgeVoiceList';
 export { default as voiceLocale } from './data/locales';
-export { default as nameList } from './data/nameList';
 export { default as openaiVoiceList } from './data/openaiVoiceList';
+export { default as voiceList } from './data/voiceList';
 export { useAudioPlayer } from './hooks/useAudioPlayer';
 export { useAudioVisualizer } from './hooks/useAudioVisualizer';
 export { useBlobUrl } from './hooks/useBlobUrl';
 export { useStreamAudioPlayer } from './hooks/useStreamAudioPlayer';
-export { type AzureSpeechOptions, fetchAzureSpeech } from './services/fetchAzureSpeech';
 export { type EdgeSpeechOptions, fetchEdgeSpeech } from './services/fetchEdgeSpeech';
 export { fetchMicrosoftSpeech, type MicrosoftSpeechOptions } from './services/fetchMicrosoftSpeech';
 export { fetchOpenaiSTT, type OpenaiSttOptions } from './services/fetchOpenaiSTT';
 export { fetchOpenaiTTS, type OpenaiTtsOptions } from './services/fetchOpenaiTTS';
 export { useAudioRecorder } from './useAudioRecorder';
-export { useAzureSpeech } from './useAzureSpeech';
 export { useEdgeSpeech } from './useEdgeSpeech';
 export { useMicrosoftSpeech } from './useMicrosoftSpeech';
 export {
@@ -42,3 +40,9 @@ export {
   getSpeechSynthesVoiceOptions,
   getVoiceLocaleOptions,
 } from './utils/getVoiceList';
+export {
+  EDGE_SPEECH_API_URL,
+  MICROSOFT_SPEECH_API_URL,
+  OPENAI_STT_API_URL,
+  OPENAI_TTS_API_URL,
+} from '@/const/api';

src/server.ts

@@ -1,2 +1,5 @@
-export { handleAzureSpeechRequest } from './server/handleAzureSpeechRequest';
-export { handleMicrosoftSpeechRequest } from './server/handleMicrosoftSpeechRequest';
+export { createEdgeSpeechComletion } from '@/server/createEdgeSpeechComletion';
+export { createMicrosoftSpeechComletion } from '@/server/createMicrosoftSpeechComletion';
+export { createOpenaiAudioSpeechCompletion } from '@/server/createOpenaiAudioSpeechCompletion';
+export { createOpenaiAudioTranscriptionsCompletion } from '@/server/createOpenaiAudioTranscriptionsCompletion';
+export * from '@/server/types';