feat(mistral): implement Mistral TTS plugin and STT configs

Author: CarltonBags

plugins/mistral/src/tts.ts

@@ -10,6 +10,7 @@ import {
   tts,
 } from '@livekit/agents';
 import { Mistral } from '@mistralai/mistralai';
+import * as crypto from 'node:crypto';
 import type { MistralTTSModels } from './models.js';
 
 // Confirmed from WAV header: Mistral TTS PCM output is 24000 Hz, mono, 16-bit signed
@@ -135,7 +136,8 @@ export class ChunkedStream extends tts.ChunkedStream {
         stream: true,
       });
 
-      const requestId = this.#text.slice(0, 8);
+      const requestId = crypto.randomUUID();
+      const segmentId = crypto.randomUUID();
       const audioByteStream = new AudioByteStream(MISTRAL_TTS_SAMPLE_RATE, MISTRAL_TTS_CHANNELS);
 
       let lastFrame: import('@livekit/rtc-node').AudioFrame | undefined;
@@ -152,7 +154,7 @@ export class ChunkedStream extends tts.ChunkedStream {
           const pcmBytes = Buffer.from(event.data.audioData, 'base64');
           const frames = audioByteStream.write(pcmBytes);
           for (const frame of frames) {
-            sendLastFrame(requestId, false);
+            sendLastFrame(segmentId, false);
             lastFrame = frame;
           }
         } else if (event.data.type === 'speech.audio.done') {
@@ -163,11 +165,11 @@ export class ChunkedStream extends tts.ChunkedStream {
       // Flush any remaining buffered audio
       const flushFrames = audioByteStream.flush();
       for (const frame of flushFrames) {
-        sendLastFrame(requestId, false);
+        sendLastFrame(segmentId, false);
         lastFrame = frame;
       }
 
-      sendLastFrame(requestId, true);
+      sendLastFrame(segmentId, true);
       this.queue.close();
     } catch (error: unknown) {
       if (this.abortController?.signal.aborted) return;