fix: preserve ByteStream subclasses during deserialization

haystack/dataclasses/byte_stream.py

@@ -119,6 +119,6 @@ def from_dict(cls, data: dict[str, Any]) -> "ByteStream":
 
         :param data: A dictionary with keys 'data', 'meta', and 'mime_type'.
 
-        :returns: A ByteStream instance.
+        :returns: An instance of this class.
         """
-        return ByteStream(data=bytes(data["data"]), meta=data.get("meta", {}), mime_type=data.get("mime_type"))
+        return cls(data=bytes(data["data"]), meta=data.get("meta", {}), mime_type=data.get("mime_type"))

releasenotes/notes/preserve-bytestream-subclass-8f2c408199657a06.yaml

@@ -0,0 +1,4 @@
+fixes:
+  - |
+    ``ByteStream.from_dict()`` now preserves subclasses when deserializing a
+    dictionary, instead of always returning a ``ByteStream`` instance.

test/dataclasses/test_byte_stream.py

@@ -159,6 +159,17 @@ def test_from_dict():
     assert str(b2) == str(b)
 
 
+def test_from_dict_preserves_subclass():
+    class CustomByteStream(ByteStream):
+        pass
+
+    original = CustomByteStream.from_string("Hello, world!", mime_type="text/plain", meta={"foo": "bar"})
+    restored = CustomByteStream.from_dict(original.to_dict())
+
+    assert isinstance(restored, CustomByteStream)
+    assert restored.to_dict() == original.to_dict()
+
+
 def test_no_warning_on_init():
     with warnings.catch_warnings():
         warnings.simplefilter("error", Warning)