@tuturuuu/ai 0.8.1 → 0.8.2

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@tuturuuu/ai",
   "license": "MIT",
-  "version": "0.8.1",
+  "version": "0.8.2",
   "repository": {
     "type": "git",
     "url": "https://github.com/tutur3u/platform",
@@ -76,9 +76,9 @@
     "@streamdown/math": "^1.0.2",
     "@streamdown/mermaid": "^1.0.2",
     "@tuturuuu/google": "0.1.0",
-    "@tuturuuu/internal-api": "0.29.0",
+    "@tuturuuu/internal-api": "0.29.1",
     "@tuturuuu/supabase": "0.5.0",
-    "@tuturuuu/utils": "0.23.0",
+    "@tuturuuu/utils": "0.24.2",
     "@vercel/sandbox": "^2.9.2",
     "@zernio/chat-sdk-adapter": "^0.5.0",
     "ai": "^7.0.56",

package/src/meetings/transcription/input.ts

@@ -0,0 +1,60 @@
+export const MAX_INLINE_TRANSCRIPTION_AUDIO_BYTES = 18 * 1024 * 1024;
+export const TRANSCRIPTION_MULTIPART_HEADROOM_BYTES = 1024 * 1024;
+export const MAX_TRANSCRIPTION_MULTIPART_REQUEST_BYTES =
+  MAX_INLINE_TRANSCRIPTION_AUDIO_BYTES + TRANSCRIPTION_MULTIPART_HEADROOM_BYTES;
+
+const TRANSCRIPTION_AUDIO_MEDIA_TYPES = [
+  'audio/webm',
+  'audio/mp4',
+  'audio/ogg',
+  'audio/mpeg',
+] as const;
+
+export type TranscriptionAudioMediaType =
+  (typeof TRANSCRIPTION_AUDIO_MEDIA_TYPES)[number];
+
+type TranscriptionAudioValidation =
+  | { mediaType: TranscriptionAudioMediaType; ok: true }
+  | { message: string; ok: false; status: 400 | 413 | 415 };
+
+export function validateTranscriptionAudioFile(
+  audio: Pick<File, 'size' | 'type'>
+): TranscriptionAudioValidation {
+  if (audio.size <= 0) {
+    return { message: 'The recording is empty.', ok: false, status: 400 };
+  }
+
+  if (audio.size > MAX_INLINE_TRANSCRIPTION_AUDIO_BYTES) {
+    return {
+      message: 'The recording exceeds the inline transcription limit.',
+      ok: false,
+      status: 413,
+    };
+  }
+
+  const baseMediaType = audio.type.split(';', 1)[0]?.trim().toLowerCase();
+  const mediaType = TRANSCRIPTION_AUDIO_MEDIA_TYPES.find(
+    (allowedMediaType) => allowedMediaType === baseMediaType
+  );
+  if (!mediaType) {
+    return {
+      message:
+        'The recording format is not supported for inline transcription.',
+      ok: false,
+      status: 415,
+    };
+  }
+
+  return { mediaType, ok: true };
+}
+
+export function hasOversizedDeclaredContentLength(req: Request) {
+  const contentLength = req.headers.get('content-length')?.trim();
+  if (!contentLength || !/^\d+$/u.test(contentLength)) {
+    return false;
+  }
+
+  return (
+    BigInt(contentLength) > BigInt(MAX_TRANSCRIPTION_MULTIPART_REQUEST_BYTES)
+  );
+}

package/src/meetings/transcription/route.ts

@@ -3,6 +3,10 @@ import { createClient } from '@tuturuuu/supabase/next/server';
 import { generateObject } from 'ai';
 import { z } from 'zod';
 import { withAiMemory } from '../../memory';
+import {
+  hasOversizedDeclaredContentLength,
+  validateTranscriptionAudioFile,
+} from './input';
 
 const DEFAULT_MODEL_NAME = 'gemini-3.1-flash-lite';
 
@@ -42,14 +46,23 @@ export function createPOST() {
         return new Response('Unauthorized', { status: 401 });
       }
 
+      if (hasOversizedDeclaredContentLength(req)) {
+        return new Response('Audio file too large', { status: 413 });
+      }
+
       const formData = await req.formData();
-      const audioFile = formData.get('audio') as File;
+      const audioFile = formData.get('audio');
       const wsId = formData.get('wsId') as string | null;
 
-      if (!audioFile) {
+      if (!(audioFile instanceof File)) {
         return new Response('No audio file provided', { status: 400 });
       }
 
+      const validation = validateTranscriptionAudioFile(audioFile);
+      if (!validation.ok) {
+        return new Response(validation.message, { status: validation.status });
+      }
+
       // Convert file to buffer
       const audioBuffer = await audioFile.arrayBuffer();
       const audioUint8Array = new Uint8Array(audioBuffer);
@@ -73,7 +86,7 @@ export function createPOST() {
             content: [
               {
                 type: 'file',
-                mediaType: 'audio/mpeg',
+                mediaType: validation.mediaType,
                 data: audioUint8Array,
               },
             ],