Fix base64 audio handling for automatic-speech-recognition

Author: alvarobartt

src/huggingface_inference_toolkit/serialization/json_utils.py

@@ -14,34 +14,49 @@ def default(obj):
     raise TypeError
 
 
+def _is_base64_audio(data):
+    """Check if data appears to be base64 encoded audio."""
+    if not isinstance(data, str):
+        return False
+
+    try:
+        decoded = base64.b64decode(data)
+        # Check for common audio file headers
+        # WAV: starts with 'RIFF' and contains 'WAVE'
+        if decoded.startswith(b"RIFF") and b"WAVE" in decoded[:20]:
+            return True
+        # MP3: starts with 'ID3' or has MP3 frame sync
+        if decoded.startswith(b"ID3") or (len(decoded) > 2 and decoded[0] == 0xFF and (decoded[1] & 0xE0) == 0xE0):
+            return True
+        # FLAC: starts with 'fLaC'
+        if decoded.startswith(b"fLaC"):
+            return True
+        # OGG: starts with 'OggS'
+        if decoded.startswith(b"OggS"):
+            return True
+        # M4A/AAC: starts with specific atoms
+        if decoded.startswith((b"ftypM4A", b"ftypisom", b"ftypmp42")):
+            return True
+    except Exception:
+        return False
+
+    return False
+
+
 class Jsoner:
     @staticmethod
     def deserialize(body):
-        return orjson.loads(body)
+        data = orjson.loads(body)
+
+        if isinstance(data, dict) and "inputs" in data:
+            if _is_base64_audio(data["inputs"]):
+                try:
+                    data["inputs"] = base64.b64decode(data["inputs"])
+                except Exception:
+                    pass
+
+        return data
 
     @staticmethod
     def serialize(body, accept=None):
         return orjson.dumps(body, option=orjson.OPT_SERIALIZE_NUMPY, default=default)
-
-
-# class _JSONEncoder(json.JSONEncoder):
-#     """
-#     custom `JSONEncoder` to make sure float and int64 ar converted
-#     """
-
-#     def default(self, obj):
-#         if isinstance(obj, np.integer):
-#             return int(obj)
-#         elif isinstance(obj, np.floating):
-#             return float(obj)
-#         elif isinstance(obj, np.ndarray):
-#             return obj.tolist()
-#         elif isinstance(obj, datetime.datetime):
-#             return obj.__str__()
-#         elif isinstance(obj, Image.Image):
-#             with BytesIO() as out:
-#                 obj.save(out, format="PNG")
-#                 png_string = out.getvalue()
-#                 return base64.b64encode(png_string).decode("utf-8")
-#         else:
-#             return super(_JSONEncoder, self).default(obj)