diff --git a/src/skore/project.py b/src/skore/project.py
index d752d387c..404fba1ee 100644
--- a/src/skore/project.py
+++ b/src/skore/project.py
@@ -6,6 +6,8 @@
 from pathlib import Path
 from typing import Any, List
 
+import pandas
+
 
 def get_filepath_from_project_name(project_name) -> Path:
     return None
@@ -33,61 +35,24 @@ class JSONItem(Item):
     """An item that is a JSON serializable Python object."""
 
     def __init__(self, value):
-        self.raw = value
-
-    # to do cache
-    @property
-    def raw(self):
-        return self.raw
+        self.__raw = None
+        self.__transformed = json.dumps(value)
 
-    # to do cache
     @property
-    def transformed(self):
-        return json.dumps(self.raw)
-
+    def value(self):
+        return self.__raw
 
 
 class DataFrameItem(Item):
     """An item that is a pandas or polars data frame."""
 
     def __init__(self, value):
+        self.__raw = None
         self.__transformed = value.to_json(orient="split")
 
     @property
-    def raw(self):...
-        # return self.__raw
-
-    @property
-    def transformed(self):
-        return self.__transformed
-
-    def from_json(string): ...
-
-    def to_json(string): ...
-
-
-class NumpyArrayItem(Item):
-    @property
-    def raw(self): ...
-
-    @property
-    def transformed(self): ...
-
-    def to_json(self): ...
-
-    def from_json(self): ...
-
-
-class MediaItem(Item):
-    """An item that is a media object."""
-
-
-class ScikitLearnModelItem(Item):
-    """An item that is a Scikit-Learn model."""
-
-    def __init__(self, value):
-        self.value = value
-
+    def value(self):
+        return pandas.read_json(self.__transformed, orient="split")
 
 class Transformer:
     @staticmethod
@@ -122,8 +87,8 @@ def transform(value) -> Item:
             return item_class_map[value_class](value)
 
         # For scikit-learn models, we need to check if it's a BaseEstimator
-        if hasattr(value, "fit") and hasattr(value, "predict"):
-            return ScikitLearnModelItem(value)
+        # if hasattr(value, "fit") and hasattr(value, "predict"):
+        #     return ScikitLearnModelItem(value)
 
         # If we don't have a specific Item subclass for this type,
         # we'll use the base Item class
@@ -139,9 +104,11 @@ def __init__(self, transformer, storage):
 
     def put(self, key, value):
         """Put a value into the project."""
+        i = self.transformer.transform(value)
         self.storage.save(
             key,
-            self.transformer.transform(value),
+            i.value,
+            i.metadata,
         )
 
     def put_item(self, key, item):
diff --git a/tests/unit/test_project.py b/tests/unit/test_project.py
new file mode 100644
index 000000000..93c5b04ae
--- /dev/null
+++ b/tests/unit/test_project.py
@@ -0,0 +1,16 @@
+from pandas import DataFrame
+from skore.project import Project
+
+
+def test_json_item():
+    project = Project()
+    d = {"a": 1, "b": 2}
+    project.put("test", d)
+    assert project.get("test") == d
+
+
+def test_dataframe_item():
+    project = Project()
+    df = DataFrame({"a": [1, 2, 3], "b": [4, 5, 6]})
+    project.put("pandas", df)
+    assert project.get("pandas") == df