@ai-sdk/provider-utils 5.0.11 → 5.0.12

package/CHANGELOG.md

@@ -1,5 +1,15 @@
 # @ai-sdk/provider-utils
 
+## 5.0.12
+
+### Patch Changes
+
+- 02ffdcb: fix(provider-utils): bound media-type sniffing decode for ID3-prefixed input
+
+  Media-type detection stripped ID3 tags before the ~18-byte prefix cap, decoding the entire base64 attachment (plus a full-size copy) whenever the data began with `ID3`/`SUQz`. This turned the intended O(1) sniff into an O(N) decode of the whole attachment. Detection now decodes at most a bounded prefix and skips the ID3 tag within that bound, keeping cost O(1) in input size on all paths (image, audio, and combined).
+
+- 76cb673: fix: detect MP4 audio from its ftyp box during transcription
+
 ## 5.0.11
 
 ### Patch Changes

package/dist/index.js

@@ -427,7 +427,7 @@ var documentMediaTypeSignatures = [
     // %PDF
   }
 ];
-var audioMediaTypeSignatures = [
+var audioMediaTypeSignaturesWithoutMp4 = [
   {
     mediaType: "audio/mpeg",
     bytesPrefix: [255, 251]
@@ -489,15 +489,28 @@ var audioMediaTypeSignatures = [
     mediaType: "audio/aac",
     bytesPrefix: [64, 21, 0, 0]
   },
-  {
-    mediaType: "audio/mp4",
-    bytesPrefix: [102, 116, 121, 112]
-  },
   {
     mediaType: "audio/webm",
     bytesPrefix: [26, 69, 223, 163]
   }
 ];
+var audioMediaTypeSignatures = [
+  ...audioMediaTypeSignaturesWithoutMp4,
+  {
+    mediaType: "audio/mp4",
+    bytesPrefix: [
+      0,
+      0,
+      0,
+      null,
+      102,
+      116,
+      121,
+      112
+      // ftyp
+    ]
+  }
+];
 var videoMediaTypeSignatures = [
   {
     mediaType: "video/mp4",
@@ -540,25 +553,37 @@ var videoMediaTypeSignatures = [
     // RIFF (AVI)
   }
 ];
-var stripID3 = (data) => {
-  const bytes = typeof data === "string" ? convertBase64ToUint8Array(data) : data;
+var DEFAULT_SNIFF_BYTES = 18;
+var MAX_SIGNATURE_BYTES = 12;
+var MAX_ID3_TAG_BYTES = 128 * 1024;
+var ID3_SCAN_BYTES = MAX_ID3_TAG_BYTES + MAX_SIGNATURE_BYTES;
+function decodePrefix(data, maxBytes) {
+  if (typeof data !== "string") {
+    return data.length > maxBytes ? data.subarray(0, maxBytes) : data;
+  }
+  const maxChars = Math.ceil(maxBytes / 3) * 4;
+  const bytes = convertBase64ToUint8Array(
+    data.substring(0, Math.min(data.length, maxChars))
+  );
+  return bytes.length > maxBytes ? bytes.subarray(0, maxBytes) : bytes;
+}
+function hasID3(bytes) {
+  return bytes.length > 10 && bytes[0] === 73 && // 'I'
+  bytes[1] === 68 && // 'D'
+  bytes[2] === 51;
+}
+var stripID3 = (bytes) => {
   const id3Size = (bytes[6] & 127) << 21 | (bytes[7] & 127) << 14 | (bytes[8] & 127) << 7 | bytes[9] & 127;
-  return bytes.slice(id3Size + 10);
+  return bytes.subarray(id3Size + 10);
 };
-function stripID3TagsIfPresent(data) {
-  const hasId3 = typeof data === "string" && data.startsWith("SUQz") || typeof data !== "string" && data.length > 10 && data[0] === 73 && // 'I'
-  data[1] === 68 && // 'D'
-  data[2] === 51;
-  return hasId3 ? stripID3(data) : data;
-}
 function detectMediaTypeBySignatures({
   data,
   signatures
 }) {
-  const processedData = stripID3TagsIfPresent(data);
-  const bytes = typeof processedData === "string" ? convertBase64ToUint8Array(
-    processedData.substring(0, Math.min(processedData.length, 24))
-  ) : processedData;
+  let bytes = decodePrefix(data, DEFAULT_SNIFF_BYTES);
+  if (hasID3(bytes)) {
+    bytes = stripID3(decodePrefix(data, ID3_SCAN_BYTES));
+  }
   for (const signature of signatures) {
     if (bytes.length >= signature.bytesPrefix.length && signature.bytesPrefix.every(
       (byte, index) => byte === null || bytes[index] === byte
@@ -584,7 +609,9 @@ function detectMediaType({
       signatures: [
         ...imageMediaTypeSignatures,
         ...documentMediaTypeSignatures,
-        ...audioMediaTypeSignatures,
+        // MP4 containers cannot be distinguished as audio or video by ftyp alone.
+        // Preserve the generic detection result as video/mp4.
+        ...audioMediaTypeSignaturesWithoutMp4,
         ...videoMediaTypeSignatures
       ]
     });
@@ -1153,7 +1180,7 @@ function withUserAgentSuffix(headers, ...userAgentSuffixParts) {
 }
 
 // src/version.ts
-var VERSION = true ? "5.0.11" : "0.0.0-test";
+var VERSION = true ? "5.0.12" : "0.0.0-test";
 
 // src/get-from-api.ts
 var getOriginalFetch = () => globalThis.fetch;