enhance: use bytes for dataset element contents

gptscript/datasets.py

@@ -1,5 +1,7 @@
+import base64
 from typing import Dict
-from pydantic import BaseModel
+from pydantic import BaseModel, field_serializer, field_validator, BeforeValidator
+
 
 class DatasetElementMeta(BaseModel):
     name: str
@@ -9,7 +11,17 @@ class DatasetElementMeta(BaseModel):
 class DatasetElement(BaseModel):
     name: str
     description: str
-    contents: str
+    contents: bytes
+
+    @field_serializer("contents")
+    def serialize_contents(self, value: bytes) -> str:
+        return base64.b64encode(value).decode("utf-8")
+
+    @field_validator("contents", mode="before")
+    def deserialize_contents(cls, value) -> bytes:
+        if isinstance(value, str):
+            return base64.b64decode(value)
+        return value
 
 
 class DatasetMeta(BaseModel):

gptscript/gptscript.py

@@ -242,7 +242,7 @@ async def create_dataset(self, workspace_id: str, name: str, description: str =
         )
         return Dataset.model_validate_json(res)
 
-    async def add_dataset_element(self, workspace_id: str, datasetID: str, elementName: str, elementContent: str,
+    async def add_dataset_element(self, workspace_id: str, datasetID: str, elementName: str, elementContent: bytes,
                                   elementDescription: str = "") -> DatasetElementMeta:
         if workspace_id == "":
             workspace_id = os.environ["GPTSCRIPT_WORKSPACE_ID"]
@@ -251,7 +251,7 @@ async def add_dataset_element(self, workspace_id: str, datasetID: str, elementNa
             raise ValueError("datasetID cannot be empty")
         elif elementName == "":
             raise ValueError("elementName cannot be empty")
-        elif elementContent == "":
+        elif not elementContent:
             raise ValueError("elementContent cannot be empty")
 
         res = await self._run_basic_command(
@@ -260,7 +260,7 @@ async def add_dataset_element(self, workspace_id: str, datasetID: str, elementNa
                 "input": json.dumps({
                     "datasetID": datasetID,
                     "elementName": elementName,
-                    "elementContent": elementContent,
+                    "elementContent": base64.b64encode(elementContent).decode("utf-8"),
                     "elementDescription": elementDescription,
                 }),
                 "workspaceID": workspace_id,

tests/test_gptscript.py

@@ -771,29 +771,29 @@ async def test_datasets(gptscript):
     assert len(dataset.elements) == 0, "Expected dataset elements to be empty"
 
     # Add an element
-    element_meta = await gptscript.add_dataset_element(workspace_id, dataset.id, "element1", "element1 contents",
+    element_meta = await gptscript.add_dataset_element(workspace_id, dataset.id, "element1", b"element1 contents",
                                                        "element1 description")
     assert element_meta.name == "element1", "Expected element name to match"
     assert element_meta.description == "element1 description", "Expected element description to match"
 
     # Add two more elements
     await gptscript.add_dataset_elements(workspace_id, dataset.id, [
-        DatasetElement(name="element2", contents="element2 contents", description="element2 description"),
-        DatasetElement(name="element3", contents="element3 contents", description="element3 description"),
+        DatasetElement(name="element2", contents=b"element2 contents", description="element2 description"),
+        DatasetElement(name="element3", contents=b"element3 contents", description="element3 description"),
     ])
 
     # Get the elements
     e1 = await gptscript.get_dataset_element(workspace_id, dataset.id, "element1")
     assert e1.name == "element1", "Expected element name to match"
-    assert e1.contents == "element1 contents", "Expected element contents to match"
+    assert e1.contents == b"element1 contents", "Expected element contents to match"
     assert e1.description == "element1 description", "Expected element description to match"
     e2 = await gptscript.get_dataset_element(workspace_id, dataset.id, "element2")
     assert e2.name == "element2", "Expected element name to match"
-    assert e2.contents == "element2 contents", "Expected element contents to match"
+    assert e2.contents == b"element2 contents", "Expected element contents to match"
     assert e2.description == "element2 description", "Expected element description to match"
     e3 = await gptscript.get_dataset_element(workspace_id, dataset.id, "element3")
     assert e3.name == "element3", "Expected element name to match"
-    assert e3.contents == "element3 contents", "Expected element contents to match"
+    assert e3.contents == b"element3 contents", "Expected element contents to match"
     assert e3.description == "element3 description", "Expected element description to match"
 
     # List elements in the dataset