feat: 更新语音转文字功能，支持从音频链接获取音频数据并转换为base64；添加错误处理和权限验证

src/aura/asr/index.ts

@@ -1,20 +1,109 @@
-import { app } from '@/app.ts'
+import { app, oss } from '@/app.ts'
 import { asr } from './modules/index.ts'
+import z from 'zod'
+import { baseURL } from '@/modules/domain.ts'
+import { getObjectByPathname } from '@/modules/fm-manager/index.ts'
+export const createAsr = async (opts: { base64Data: string }) => {
+  const { base64Data } = opts
+  const result = await asr.getText({
+    audio: {
+      data: base64Data,
+      format: 'wav' as any,
+      rate: 16000,
+      channel: 1
+    },
+    request: {
+      enable_words: true,
+      enable_sentence_info: true,
+      enable_utterance_info: true,
+      enable_punctuation_prediction: true,
+      enable_inverse_text_normalization: true
+    }
+  })
+  return {
+    text: result.result?.text || '',
+    result
+  };
+}
 app.route({
   path: 'asr',
   key: 'text',
   middleware: ['auth'],
-  description: '语音转文字，将base64的音频数据转换为文字, 参数: base64Audio， 为base64编码的音频数据',
-}).define(async (ctx) => {
-  const base64Audio = ctx.query.base64Audio as string
-  if (!base64Audio) {
-    ctx.throw('Missing base64Audio parameter')
-  }
-  const result = await asr.getText({
-    audio: {
-      data: base64Audio
+  description: '语音转文字，将base64的音频数据转换为文字, 参数: base64Data 为base64编码的音频数据',
+  metadata: {
+    args: {
+      base64Data: z.string().describe('base64编码的音频数据').nonempty('base64Data参数不能为空'),
     }
-  })
-  ctx.body = result
+  }
+}).define(async (ctx) => {
+  const base64Data = ctx.query.base64Data as string
+  if (!base64Data) {
+    ctx.throw(400, 'base64Data参数不能为空')
+  }
+  const result = await createAsr({ base64Data })
+  ctx.body = {
+    text: result.text
+  }
 })
   .addTo(app)
+
+app.route({
+  path: 'asr',
+  key: 'link',
+  middleware: ['auth'],
+  description: '语音转文字，将音频链接的音频数据转换为文字, 参数: url 为音频链接',
+  metadata: {
+    args: {
+      url: z.string().describe('音频链接').nonempty('url参数不能为空'),
+    }
+  }
+}).define(async (ctx) => {
+  const tokenUser = ctx.state.tokenUser;
+  const url = ctx.query.url as string
+  if (!url) {
+    ctx.throw(400, 'url参数不能为空')
+  }
+  let base64Data: string = null;
+  if (url.startsWith(baseURL) || url.startsWith('/')) {
+    const pathname = new URL(url, baseURL).pathname;
+    const [username] = pathname.split('/').filter(Boolean)
+    if (username !== tokenUser.username) {
+      ctx.throw(403, '没有权限访问该音频链接')
+    }
+    let data: Awaited<ReturnType<typeof oss.getObject>>;
+    try {
+      console.log('fetch audio from minio with objectName', pathname.slice(1))
+      const objectName = getObjectByPathname({ pathname })
+      data = await oss.getObject(objectName.objectName)
+    } catch (e: any) {
+      if (e?.name === 'NoSuchKey' || e?.$metadata?.httpStatusCode === 404) {
+        ctx.throw(404, '音频文件不存在')
+      }
+    }
+    if (!data.Body) {
+      ctx.throw(404, '音频文件内容为空')
+    }
+    const bytes = await data.Body.transformToByteArray()
+    base64Data = Buffer.from(bytes).toString('base64')
+  } else if (url.startsWith('http')) {
+    base64Data = await fetchAudioAsBase64(url)
+  } else {
+    ctx.throw(400, 'url参数必须是有效的链接')
+  }
+  // 这里需要将音频链接转换为base64数据，可以使用fetch获取音频数据并转换为base64
+  const result = await createAsr({ base64Data })
+  ctx.body = {
+    text: result.text
+  }
+})
+  .addTo(app)
+
+const fetchAudioAsBase64 = async (url: string): Promise<string> => {
+  const response = await fetch(url);
+  if (!response.ok) {
+    throw new Error(`Failed to fetch audio from URL: ${response.statusText}`);
+  }
+  const arrayBuffer = await response.arrayBuffer();
+  const base64String = btoa(String.fromCharCode(...new Uint8Array(arrayBuffer)));
+  return base64String;
+};

src/aura/asr/modules/index.ts

@@ -4,4 +4,5 @@ import { auraConfig } from '../../config.ts'
 export const asr = new Asr({
   appid: auraConfig.VOLCENGINE_AUC_APPID,
   token: auraConfig.VOLCENGINE_AUC_TOKEN,
+  type: 'flash'
 })