update

src/asr/provider/volcengine/asr-ws-big-model-client.ts

@@ -1,5 +1,5 @@
-import * as zlib from 'zlib';
-import { promisify } from 'util';
+import * as zlib from 'node:zlib';
+import { promisify } from 'node:util';
 import { nanoid } from 'nanoid';
 import { VolcEngineBase, uuid } from './base.ts';
 
@@ -61,6 +61,39 @@ function generateBeforePayload(sequence: number): Buffer {
   return beforePayload;
 }
 
+export type ParsedMessage = {
+  isLastPackage: boolean;
+  payloadSequence?: number;
+  payloadMsg?: {
+    audio_info?: {
+      duration: number;
+    };
+    result?: {
+      additions?: {
+        log_id?: string;
+      };
+      text?: string;
+      utterances?: Array<{
+        additions?: {
+          fixed_prefix_result?: string;
+        };
+        definite?: boolean;
+        end_time?: number;
+        start_time?: number;
+        text?: string;
+        words?: Array<{
+          end_time: number;
+          start_time: number;
+          text: string;
+        }>;
+      }>;
+    };
+    error?: any;
+  };
+  payloadSize?: number;
+  code?: number;
+  seq?: number;
+};
 /**
  * Parse response from the WebSocket server
  */
@@ -393,10 +426,11 @@ export class AsrWsClient extends VolcEngineBase {
     // Wait for response
     await sendVoice(audioData, segmentSize);
   }
+
   async onMessage(event: MessageEvent) {
     try {
       const parsed = parseResponse(Buffer.from(event.data as ArrayBuffer));
-      console.log(`Seq ${parsed.payloadSequence} response:`, parsed);
+      // console.log(`Seq ${parsed.payloadSequence} response:`, parsed);
       if (typeof event.data === 'string') {
         throw new Error('event.data is string: ' + event.data);
       }
@@ -405,10 +439,9 @@ export class AsrWsClient extends VolcEngineBase {
         this.emitter.emit('error', parsed);
         this.isError = true;
       }
+      this.emitter.emit('message', parsed);
       if (parsed.isLastPackage) {
         this.emitter.emit('end', parsed);
-      } else {
-        this.emitter.emit('message', parsed);
       }
     } catch (error) {
       console.error('Error processing response:', error);
@@ -440,6 +473,14 @@ export class AsrWsClient extends VolcEngineBase {
       throw error;
     }
   }
+  async setIsEnd(isEnd: boolean) {
+      super.setIsEnd(isEnd);
+      if (isEnd) {
+        // 发送空白包
+        const emptyBuffer = Buffer.alloc(10000);
+        this.sendVoiceStream(emptyBuffer);
+      }
+  }
   /**
    * 发送语音流, 最小10000
    * @param data

src/asr/provider/volcengine/asr-ws-client.ts

@@ -238,7 +238,7 @@ interface AudioItem {
   id: string | number;
   path: string;
 }
-
+// 流式语音识别
 export class AsrWsClient extends VolcEngineBase {
   private audioPath: string;
   private cluster: string;

src/asr/provider/volcengine/auc.ts

@@ -1,11 +1,11 @@
-import crypto from "node:crypto"
+import { nanoid } from "nanoid"
 
 export const FlashURL = "https://openspeech.bytedance.com/api/v3/auc/bigmodel/recognize/flash"
 export const AsrBaseURL = 'https://openspeech.bytedance.com/api/v3/auc/bigmodel/submit'
 export const AsrBase = 'volc.bigasr.auc'
 export const AsrTurbo = 'volc.bigasr.auc_turbo'
 
-const uuid = () => crypto.randomUUID()
+const uuid = () => nanoid()
 
 type AsrOptions = {
   url?: string

src/asr/provider/volcengine/base.ts

@@ -1,4 +1,3 @@
-import { initWs } from '../../../ws-adapter/index.ts';
 import { WSServer } from '../../provider/ws-server.ts';
 import { nanoid } from 'nanoid';
 

src/asr/provider/volcengine/test/asr-bigmodel.ts

@@ -7,15 +7,22 @@ import fs from 'fs';
 const main = async () => {
   const audioId = '123';
   const asrClient = new AsrWsClient({
-    appid: config.APP_ID,
-    token: config.TOKEN,
+    appid: config.VOLCENGINE_ASR_MODEL_APPID,
+    token: config.VOLCENGINE_ASR_MODEL_TOKEN,
   });
+  asrClient.emitter.on('message', (result) => {
+    console.log('识别结果', JSON.stringify(result, null, 2));
+  })
+  asrClient.emitter.on('end', (result) => {
+    console.log('识别结束', JSON.stringify(result, null, 2));
+  })
   await new Promise((resolve) => setTimeout(resolve, 2000));
   const data = fs.readFileSync(audioPath);
   await asrClient.sendVoiceFile(data);
-  await asrClient.sendVoiceFile(fs.readFileSync(blankAudioPath));
+  // await asrClient.sendVoiceFile(fs.readFileSync(blankAudioPath));
   asrClient.setIsEnd(true);
-  await asrClient.sendVoiceFile(fs.readFileSync(audioPath2));
+  // await asrClient.sendVoiceFile(fs.readFileSync(audioPath2));
+
 };
 
 main();

src/ws-adapter/index.ts

@@ -1,4 +1,6 @@
-const isBrowser = process?.env?.BROWSER === 'true' || (typeof window !== 'undefined' && typeof window.document !== 'undefined');
+const isBrowser = (typeof process === 'undefined') || 
+  (typeof window !== 'undefined' && typeof window.document !== 'undefined') || 
+  (typeof process !== 'undefined' && process?.env?.BROWSER === 'true');
 const chantHttpToWs = (url: string) => {
   if (url.startsWith('http://')) {
     return url.replace('http://', 'ws://');