diff --git a/openhexa/sdk/datasets/dataset.py b/openhexa/sdk/datasets/dataset.py
index f22b87c..6fddbdb 100644
--- a/openhexa/sdk/datasets/dataset.py
+++ b/openhexa/sdk/datasets/dataset.py
@@ -203,14 +203,14 @@ def get_file(self, filename: str) -> DatasetFile:
 
     def add_file(
         self,
-        source: typing.Union[str, PathLike[str], typing.IO],
+        source: typing.Union[str, PathLike[str], typing.IO, bytes],
         filename: typing.Optional[str] = None,
     ) -> DatasetFile:
         """Create a new dataset file and add it to the dataset version."""
         mime_type = None
         if isinstance(source, (str, PathLike)):
             path = Path(source)
-            filename = path.name
+            filename = path.name if filename is None else filename
             mime_type, _ = mimetypes.guess_type(path)
         else:
             if filename is None:
@@ -249,19 +249,18 @@ def add_file(
                 raise ValueError("A file with this name already exists in this dataset version")
             else:
                 raise Exception(errors)
-
-        upload_url = data["createDatasetVersionFile"]["uploadUrl"]
-        content = read_content(source)
-
-        response = requests.put(upload_url, data=content, headers={"Content-Type": mime_type})
+        result = data["createDatasetVersionFile"]
+        upload_url = result["uploadUrl"]
+        with read_content(source) as content:
+            response = requests.put(upload_url, data=content, headers={"Content-Type": mime_type})
         response.raise_for_status()
         return DatasetFile(
             version=self,
-            id=data["createDatasetVersionFile"]["file"]["id"],
-            filename=data["createDatasetVersionFile"]["file"]["filename"],
-            content_type=data["createDatasetVersionFile"]["file"]["contentType"],
-            uri=data["createDatasetVersionFile"]["file"]["uri"],
-            created_at=data["createDatasetVersionFile"]["file"]["createdAt"],
+            id=result["file"]["id"],
+            filename=result["file"]["filename"],
+            content_type=result["file"]["contentType"],
+            uri=result["file"]["uri"],
+            created_at=result["file"]["createdAt"],
         )
 
 
diff --git a/openhexa/sdk/utils.py b/openhexa/sdk/utils.py
index b569397..e651607 100644
--- a/openhexa/sdk/utils.py
+++ b/openhexa/sdk/utils.py
@@ -1,6 +1,7 @@
 """Miscellaneous utility functions."""
 
 import abc
+import contextlib
 import enum
 import os
 import typing
@@ -173,20 +174,21 @@ def __next__(self):
         return result
 
 
-def read_content(source: typing.Union[str, os.PathLike[str], typing.IO], encoding: str = "utf-8") -> bytes:
+@contextlib.contextmanager
+def read_content(source: typing.Union[str, os.PathLike[str], typing.IO, bytes]):
     """Read file content and return it as bytes."""
-    # If source is a string or PathLike object
-    if isinstance(source, (str, os.PathLike)):
-        with open(os.fspath(source), "rb") as f:
-            return f.read()
-
-    # If source is a buffer
-    elif hasattr(source, "read"):
-        content = source.read()
+    try:
+        if isinstance(source, bytes):
+            yield source
+        elif hasattr(source, "read"):
+            yield source
+        # If source is a string or PathLike object
+        elif isinstance(source, (str, os.PathLike)):
+            with open(os.fspath(source), "rb") as f:
+                yield f
 
-        if not isinstance(content, bytes):
-            return content.encode(encoding)
         else:
-            return content
-
-    raise ValueError("Unsupported type for source")
+            raise ValueError("Unsupported type for source")
+    finally:
+        if hasattr(source, "close"):
+            source.close()