orca-sdk 0.1.6__py3-none-any.whl → 0.1.7__py3-none-any.whl

orca_sdk/memoryset.py

@@ -27,13 +27,12 @@ from .async_client import OrcaAsyncClient
 from .client import (
     CascadingEditSuggestion,
     CloneMemorysetRequest,
+    CreateMemorysetFromDatasourceRequest,
     CreateMemorysetRequest,
     FilterItem,
 )
 from .client import LabeledMemory as LabeledMemoryResponse
-from .client import (
-    LabeledMemoryInsert,
-)
+from .client import LabeledMemoryInsert
 from .client import LabeledMemoryLookup as LabeledMemoryLookupResponse
 from .client import (
     LabeledMemoryUpdate,
@@ -50,9 +49,7 @@ from .client import (
     PredictionFeedback,
 )
 from .client import ScoredMemory as ScoredMemoryResponse
-from .client import (
-    ScoredMemoryInsert,
-)
+from .client import ScoredMemoryInsert
 from .client import ScoredMemoryLookup as ScoredMemoryLookupResponse
 from .client import (
     ScoredMemoryUpdate,
@@ -937,7 +934,7 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
     length: int
     created_at: datetime
     updated_at: datetime
-    insertion_status: Status
+    insertion_status: Status | None
     embedding_model: EmbeddingModelBase
     index_type: IndexType
     index_params: dict[str, Any]
@@ -959,7 +956,9 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
         self.length = metadata["length"]
         self.created_at = datetime.fromisoformat(metadata["created_at"])
         self.updated_at = datetime.fromisoformat(metadata["updated_at"])
-        self.insertion_status = Status(metadata["insertion_status"])
+        self.insertion_status = (
+            Status(metadata["insertion_status"]) if metadata["insertion_status"] is not None else None
+        )
         self._last_refresh = datetime.now()
         self.index_type = metadata["index_type"]
         self.index_params = metadata["index_params"]
@@ -971,7 +970,7 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
 
     def __repr__(self) -> str:
         return (
-            "Memoryset({\n"
+            f"{self.memory_type.capitalize()}Memoryset(" + "{\n"
             f"    name: '{self.name}',\n"
             f"    length: {self.length},\n"
             f"    embedding_model: {self.embedding_model},\n"
@@ -1022,11 +1021,11 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
         return existing
 
     @classmethod
-    def create(
+    def _create_from_datasource(
         cls,
         name: str,
-        datasource: Datasource,
         *,
+        datasource: Datasource,
         embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
         value_column: str = "value",
         label_column: str | None = None,
@@ -1047,54 +1046,9 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
         memory_type: MemoryType | None = None,
     ) -> Self | Job[Self]:
         """
-        Create a new memoryset in the OrcaCloud
-
-        All columns from the datasource that are not specified in the `value_column`,
-        `label_column`, `source_id_column`, or `partition_id_column` will be stored as metadata in the memoryset.
-
-        Params:
-            name: Name for the new memoryset (must be unique)
-            datasource: Source data to populate the memories in the memoryset
-            embedding_model: Embedding model to use for embedding memory values for semantic search.
-                If not provided, a default embedding model for the memoryset will be used.
-            value_column: Name of the column in the datasource that contains the memory values
-            label_column: Name of the column in the datasource that contains the memory labels.
-                Must contain categorical values as integers or strings. String labels will be
-                converted to integers with the unique strings extracted as `label_names`
-            score_column: Name of the column in the datasource that contains the memory scores
-            source_id_column: Optional name of the column in the datasource that contains the ids in
-                the system of reference
-            partition_id_column: Optional name of the column in the datasource that contains the partition ids
-            description: Optional description for the memoryset, this will be used in agentic flows,
-                so make sure it is concise and describes the contents of your memoryset not the
-                datasource or the embedding model.
-            label_names: List of human-readable names for the labels in the memoryset, must match
-                the number of labels in the `label_column`. Will be automatically inferred if string
-                labels are provided or if a [Dataset][datasets.Dataset] with a
-                [`ClassLabel`][datasets.ClassLabel] feature for labels is used as the datasource
-            max_seq_length_override: Maximum sequence length of values in the memoryset, if the
-                value is longer than this it will be truncated, will default to the model's max
-                sequence length if not provided
-            prompt: Optional prompt to use when embedding documents/memories for storage
-            remove_duplicates: Whether to remove duplicates from the datasource before inserting
-                into the memoryset
-            index_type: Type of vector index to use for the memoryset, defaults to `"FLAT"`. Valid
-                values are `"FLAT"`, `"IVF_FLAT"`, `"IVF_SQ8"`, `"IVF_PQ"`, `"HNSW"`, and `"DISKANN"`.
-            index_params: Parameters for the vector index, defaults to `{}`
-            if_exists: What to do if a memoryset with the same name already exists, defaults to
-                `"error"`. Other option is `"open"` to open the existing memoryset.
-            background: Whether to run the operation none blocking and return a job handle
-            hidden: Whether the memoryset should be hidden
-            subsample: Optional number (int) of rows to insert or fraction (float in (0, 1]) of the
-                datasource to insert. Use to limit the size of the initial memoryset.
-            memory_type: Type of memoryset to create, defaults to `"LABELED"` if `label_column` is provided,
-                and `"SCORED"` if `score_column` is provided, must be specified for other cases.
-        Returns:
-            Handle to the new memoryset in the OrcaCloud
+        Create a memoryset from a datasource by calling the API.
 
-        Raises:
-            ValueError: If the memoryset already exists and if_exists is `"error"` or if it is
-                `"open"` and the params do not match those of the existing memoryset.
+        This is a private method that performs the actual API call to create a memoryset from a datasource.
         """
         if embedding_model is None:
             embedding_model = PretrainedEmbeddingModel.GTE_BASE
@@ -1108,7 +1062,7 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
         if existing is not None:
             return existing
 
-        payload: CreateMemorysetRequest = {
+        payload: CreateMemorysetFromDatasourceRequest = {
             "name": name,
             "description": description,
             "datasource_name_or_id": datasource.id,
@@ -1138,141 +1092,582 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
             raise ValueError("Invalid embedding model")
         client = OrcaClient._resolve_client()
         response = client.POST("/memoryset", json=payload)
+
+        if response["insertion_job_id"] is None:
+            raise ValueError("Create memoryset operation failed to produce an insertion job")
+
         job = Job(response["insertion_job_id"], lambda: cls.open(response["id"]))
         return job if background else job.result()
 
     @overload
     @classmethod
-    def from_hf_dataset(cls, name: str, hf_dataset: Dataset, background: Literal[True], **kwargs: Any) -> Self:
-        pass
-
-    @overload
-    @classmethod
-    def from_hf_dataset(cls, name: str, hf_dataset: Dataset, background: Literal[False] = False, **kwargs: Any) -> Self:
+    def create(
+        cls,
+        name: str,
+        *,
+        datasource: None = None,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        description: str | None = None,
+        label_names: list[str] | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
+        hidden: bool = False,
+        memory_type: MemoryType | None = None,
+    ) -> Self:
         pass
 
-    @classmethod
-    def from_hf_dataset(
-        cls, name: str, hf_dataset: Dataset, background: bool = False, **kwargs: Any
-    ) -> Self | Job[Self]:
-        """
-        Create a new memoryset from a Hugging Face [`Dataset`][datasets.Dataset] in the OrcaCloud
-
-        This will automatically create a [`Datasource`][orca_sdk.Datasource] with the same name
-        appended with `_datasource` and use that as the datasource for the memoryset.
-
-        All features that are not specified to be used as `value_column`, `label_column`, or
-        `source_id_column` will be stored as metadata in the memoryset.
-
-        Params:
-            name: Name for the new memoryset (must be unique)
-            hf_dataset: Hugging Face dataset to create the memoryset from
-            kwargs: Additional parameters for creating the memoryset. See
-                [`create`][orca_sdk.memoryset.MemorysetBase.create] attributes for details.
-
-        Returns:
-            Handle to the new memoryset in the OrcaCloud
-        """
-        if_exists = kwargs.get("if_exists", "error")
-        existing = cls._handle_if_exists(
-            name,
-            if_exists=if_exists,
-            label_names=kwargs.get("label_names"),
-            embedding_model=kwargs.get("embedding_model"),
-        )
-        if existing is not None:
-            return existing
-
-        datasource = Datasource.from_hf_dataset(
-            f"{name}_datasource", hf_dataset, if_exists=kwargs.get("if_exists", "error")
-        )
-        kwargs["background"] = background
-        return cls.create(name, datasource, **kwargs)
-
     @overload
     @classmethod
-    def from_pytorch(
+    def create(
         cls,
         name: str,
-        torch_data: TorchDataLoader | TorchDataset,
         *,
-        column_names: list[str] | None = None,
+        datasource: Datasource,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        value_column: str = "value",
+        label_column: str | None = None,
+        score_column: str | None = None,
+        source_id_column: str | None = None,
+        partition_id_column: str | None = None,
+        description: str | None = None,
+        label_names: list[str] | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        remove_duplicates: bool = True,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
         background: Literal[True],
-        **kwargs: Any,
+        hidden: bool = False,
+        subsample: int | float | None = None,
+        memory_type: MemoryType | None = None,
     ) -> Job[Self]:
         pass
 
     @overload
     @classmethod
-    def from_pytorch(
+    def create(
         cls,
         name: str,
-        torch_data: TorchDataLoader | TorchDataset,
         *,
-        column_names: list[str] | None = None,
+        datasource: Datasource,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        value_column: str = "value",
+        label_column: str | None = None,
+        score_column: str | None = None,
+        source_id_column: str | None = None,
+        partition_id_column: str | None = None,
+        description: str | None = None,
+        label_names: list[str] | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        remove_duplicates: bool = True,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
         background: Literal[False] = False,
-        **kwargs: Any,
+        hidden: bool = False,
+        subsample: int | float | None = None,
+        memory_type: MemoryType | None = None,
     ) -> Self:
         pass
 
     @classmethod
-    def from_pytorch(
+    def create(
         cls,
         name: str,
-        torch_data: TorchDataLoader | TorchDataset,
         *,
-        column_names: list[str] | None = None,
+        datasource: Datasource | None = None,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        value_column: str = "value",
+        label_column: str | None = None,
+        score_column: str | None = None,
+        source_id_column: str | None = None,
+        partition_id_column: str | None = None,
+        description: str | None = None,
+        label_names: list[str] | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        remove_duplicates: bool = True,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
         background: bool = False,
-        **kwargs: Any,
+        hidden: bool = False,
+        subsample: int | float | None = None,
+        memory_type: MemoryType | None = None,
     ) -> Self | Job[Self]:
         """
-        Create a new memoryset from a PyTorch [`DataLoader`][torch.utils.data.DataLoader] or
-        [`Dataset`][torch.utils.data.Dataset] in the OrcaCloud
+        Create a new memoryset in the OrcaCloud
 
-        This will automatically create a [`Datasource`][orca_sdk.Datasource] with the same name
-        appended with `_datasource` and use that as the datasource for the memoryset.
+        If `datasource` is provided, all columns from the datasource that are not specified in the
+        `value_column`, `label_column`, `source_id_column`, or `partition_id_column` will be stored
+        as metadata in the memoryset.
 
-        All properties that are not specified to be used as `value_column`, `label_column`, or
-        `source_id_column`, or `partition_id_column` will be stored as metadata in the memoryset.
+        If `datasource` is omitted (None), an empty memoryset will be created with no initial memories.
+        You can add memories later using the `insert` method.
 
         Params:
             name: Name for the new memoryset (must be unique)
-            torch_data: PyTorch data loader or dataset to create the memoryset from
-            column_names: If the provided dataset or data loader returns unnamed tuples, this
-                argument must be provided to specify the names of the columns.
-            background: Whether to run the operation in the background
-            kwargs: Additional parameters for creating the memoryset. See
-                [`create`][orca_sdk.memoryset.MemorysetBase.create] attributes for details.
-
+            datasource: Optional source data to populate the memories in the memoryset. If omitted,
+                an empty memoryset will be created.
+            embedding_model: Embedding model to use for embedding memory values for semantic search.
+                If not provided, a default embedding model for the memoryset will be used.
+            value_column: Name of the column in the datasource that contains the memory values
+            label_column: Name of the column in the datasource that contains the memory labels.
+                Must contain categorical values as integers or strings. String labels will be
+                converted to integers with the unique strings extracted as `label_names`
+            score_column: Name of the column in the datasource that contains the memory scores
+            source_id_column: Optional name of the column in the datasource that contains the ids in
+                the system of reference
+            partition_id_column: Optional name of the column in the datasource that contains the partition ids
+            description: Optional description for the memoryset, this will be used in agentic flows,
+                so make sure it is concise and describes the contents of your memoryset not the
+                datasource or the embedding model.
+            label_names: List of human-readable names for the labels in the memoryset, must match
+                the number of labels in the `label_column`. Will be automatically inferred if string
+                labels are provided or if a [Dataset][datasets.Dataset] with a
+                [`ClassLabel`][datasets.ClassLabel] feature for labels is used as the datasource
+            max_seq_length_override: Maximum sequence length of values in the memoryset, if the
+                value is longer than this it will be truncated, will default to the model's max
+                sequence length if not provided
+            prompt: Optional prompt to use when embedding documents/memories for storage
+            remove_duplicates: Whether to remove duplicates from the datasource before inserting
+                into the memoryset
+            index_type: Type of vector index to use for the memoryset, defaults to `"FLAT"`. Valid
+                values are `"FLAT"`, `"IVF_FLAT"`, `"IVF_SQ8"`, `"IVF_PQ"`, `"HNSW"`, and `"DISKANN"`.
+            index_params: Parameters for the vector index, defaults to `{}`
+            if_exists: What to do if a memoryset with the same name already exists, defaults to
+                `"error"`. Other option is `"open"` to open the existing memoryset.
+            background: Whether to run the operation none blocking and return a job handle.
+                Note: This parameter is ignored when creating an empty memoryset (when datasource is None).
+            hidden: Whether the memoryset should be hidden
+            subsample: Optional number (int) of rows to insert or fraction (float in (0, 1]) of the
+                datasource to insert. Use to limit the size of the initial memoryset.
+            memory_type: Type of memoryset to create, defaults to `"LABELED"` if `label_column` is provided,
+                and `"SCORED"` if `score_column` is provided, must be specified for other cases.
         Returns:
             Handle to the new memoryset in the OrcaCloud
-        """
-        if_exists = kwargs.get("if_exists", "error")
-        existing = cls._handle_if_exists(
-            name,
-            if_exists=if_exists,
-            label_names=kwargs.get("label_names"),
-            embedding_model=kwargs.get("embedding_model"),
-        )
-        if existing is not None:
-            return existing
 
-        datasource = Datasource.from_pytorch(
-            f"{name}_datasource", torch_data, column_names=column_names, if_exists=kwargs.get("if_exists", "error")
-        )
-        kwargs["background"] = background
-        return cls.create(name, datasource, **kwargs)
+        Raises:
+            ValueError: If the memoryset already exists and if_exists is `"error"` or if it is
+                `"open"` and the params do not match those of the existing memoryset.
+        """
+        if datasource is None:
+            return cls._create_empty(
+                name,
+                embedding_model=embedding_model,
+                description=description,
+                label_names=label_names,
+                max_seq_length_override=max_seq_length_override,
+                prompt=prompt,
+                index_type=index_type,
+                index_params=index_params,
+                if_exists=if_exists,
+                hidden=hidden,
+                memory_type=memory_type,
+            )
+        else:
+            return cls._create_from_datasource(
+                name,
+                datasource=datasource,
+                embedding_model=embedding_model,
+                value_column=value_column,
+                label_column=label_column,
+                score_column=score_column,
+                source_id_column=source_id_column,
+                partition_id_column=partition_id_column,
+                description=description,
+                label_names=label_names,
+                max_seq_length_override=max_seq_length_override,
+                prompt=prompt,
+                remove_duplicates=remove_duplicates,
+                index_type=index_type,
+                index_params=index_params,
+                if_exists=if_exists,
+                background=background,
+                hidden=hidden,
+                subsample=subsample,
+                memory_type=memory_type,
+            )
 
     @overload
     @classmethod
-    def from_list(
+    def from_datasource(
         cls,
         name: str,
-        data: list[dict],
         *,
-        background: Literal[True],
-        **kwargs: Any,
-    ) -> Job[Self]:
+        datasource: Datasource,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        value_column: str = "value",
+        label_column: str | None = None,
+        score_column: str | None = None,
+        source_id_column: str | None = None,
+        partition_id_column: str | None = None,
+        description: str | None = None,
+        label_names: list[str] | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        remove_duplicates: bool = True,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
+        background: Literal[True],
+        hidden: bool = False,
+        subsample: int | float | None = None,
+        memory_type: MemoryType | None = None,
+    ) -> Job[Self]:
+        pass
+
+    @overload
+    @classmethod
+    def from_datasource(
+        cls,
+        name: str,
+        *,
+        datasource: Datasource,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        value_column: str = "value",
+        label_column: str | None = None,
+        score_column: str | None = None,
+        source_id_column: str | None = None,
+        partition_id_column: str | None = None,
+        description: str | None = None,
+        label_names: list[str] | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        remove_duplicates: bool = True,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
+        background: Literal[False] = False,
+        hidden: bool = False,
+        subsample: int | float | None = None,
+        memory_type: MemoryType | None = None,
+    ) -> Self:
+        pass
+
+    @classmethod
+    def from_datasource(
+        cls,
+        name: str,
+        *,
+        datasource: Datasource,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        value_column: str = "value",
+        label_column: str | None = None,
+        score_column: str | None = None,
+        source_id_column: str | None = None,
+        partition_id_column: str | None = None,
+        description: str | None = None,
+        label_names: list[str] | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        remove_duplicates: bool = True,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
+        background: bool = False,
+        hidden: bool = False,
+        subsample: int | float | None = None,
+        memory_type: MemoryType | None = None,
+    ) -> Self | Job[Self]:
+        """
+        Create a new memoryset in the OrcaCloud from a datasource.
+
+        This is a convenience method that is equivalent to calling `create` with a datasource.
+        All columns from the datasource that are not specified in the `value_column`,
+        `label_column`, `source_id_column`, or `partition_id_column` will be stored as metadata
+        in the memoryset.
+
+        Params:
+            name: Name for the new memoryset (must be unique)
+            datasource: Source data to populate the memories in the memoryset.
+            embedding_model: Embedding model to use for embedding memory values for semantic search.
+                If not provided, a default embedding model for the memoryset will be used.
+            value_column: Name of the column in the datasource that contains the memory values
+            label_column: Name of the column in the datasource that contains the memory labels.
+                Must contain categorical values as integers or strings. String labels will be
+                converted to integers with the unique strings extracted as `label_names`
+            score_column: Name of the column in the datasource that contains the memory scores
+            source_id_column: Optional name of the column in the datasource that contains the ids in
+                the system of reference
+            partition_id_column: Optional name of the column in the datasource that contains the partition ids
+            description: Optional description for the memoryset, this will be used in agentic flows,
+                so make sure it is concise and describes the contents of your memoryset not the
+                datasource or the embedding model.
+            label_names: List of human-readable names for the labels in the memoryset, must match
+                the number of labels in the `label_column`. Will be automatically inferred if string
+                labels are provided or if a [Dataset][datasets.Dataset] with a
+                [`ClassLabel`][datasets.ClassLabel] feature for labels is used as the datasource
+            max_seq_length_override: Maximum sequence length of values in the memoryset, if the
+                value is longer than this it will be truncated, will default to the model's max
+                sequence length if not provided
+            prompt: Optional prompt to use when embedding documents/memories for storage
+            remove_duplicates: Whether to remove duplicates from the datasource before inserting
+                into the memoryset
+            index_type: Type of vector index to use for the memoryset, defaults to `"FLAT"`. Valid
+                values are `"FLAT"`, `"IVF_FLAT"`, `"IVF_SQ8"`, `"IVF_PQ"`, `"HNSW"`, and `"DISKANN"`.
+            index_params: Parameters for the vector index, defaults to `{}`
+            if_exists: What to do if a memoryset with the same name already exists, defaults to
+                `"error"`. Other option is `"open"` to open the existing memoryset.
+            background: Whether to run the operation none blocking and return a job handle.
+            hidden: Whether the memoryset should be hidden
+            subsample: Optional number (int) of rows to insert or fraction (float in (0, 1]) of the
+                datasource to insert. Use to limit the size of the initial memoryset.
+            memory_type: Type of memoryset to create, defaults to `"LABELED"` if `label_column` is provided,
+                and `"SCORED"` if `score_column` is provided, must be specified for other cases.
+        Returns:
+            Handle to the new memoryset in the OrcaCloud
+
+        Raises:
+            ValueError: If the memoryset already exists and if_exists is `"error"` or if it is
+                `"open"` and the params do not match those of the existing memoryset.
+        """
+        return cls._create_from_datasource(
+            name,
+            datasource=datasource,
+            embedding_model=embedding_model,
+            value_column=value_column,
+            label_column=label_column,
+            score_column=score_column,
+            source_id_column=source_id_column,
+            partition_id_column=partition_id_column,
+            description=description,
+            label_names=label_names,
+            max_seq_length_override=max_seq_length_override,
+            prompt=prompt,
+            remove_duplicates=remove_duplicates,
+            index_type=index_type,
+            index_params=index_params,
+            if_exists=if_exists,
+            background=background,
+            hidden=hidden,
+            subsample=subsample,
+            memory_type=memory_type,
+        )
+
+    @classmethod
+    def _create_empty(
+        cls,
+        name: str,
+        *,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        description: str | None = None,
+        label_names: list[str] | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
+        hidden: bool = False,
+        memory_type: MemoryType | None = None,
+    ) -> Self:
+        """
+        Create an empty memoryset in the OrcaCloud
+
+        This creates a memoryset with no initial memories. You can add memories later using
+        the `insert` method.
+
+        Params:
+            name: Name for the new memoryset (must be unique)
+            embedding_model: Embedding model to use for embedding memory values for semantic search.
+                If not provided, a default embedding model for the memoryset will be used.
+            description: Optional description for the memoryset, this will be used in agentic flows,
+                so make sure it is concise and describes the contents of your memoryset not the
+                datasource or the embedding model.
+            label_names: List of human-readable names for the labels in the memoryset
+            max_seq_length_override: Maximum sequence length of values in the memoryset, if the
+                value is longer than this it will be truncated, will default to the model's max
+                sequence length if not provided
+            prompt: Optional prompt to use when embedding documents/memories for storage
+            index_type: Type of vector index to use for the memoryset, defaults to `"FLAT"`. Valid
+                values are `"FLAT"`, `"IVF_FLAT"`, `"IVF_SQ8"`, `"IVF_PQ"`, `"HNSW"`, and `"DISKANN"`.
+            index_params: Parameters for the vector index, defaults to `{}`
+            if_exists: What to do if a memoryset with the same name already exists, defaults to
+                `"error"`. Other option is `"open"` to open the existing memoryset.
+            hidden: Whether the memoryset should be hidden
+            memory_type: Type of memoryset to create, defaults to `"LABELED"` if called from
+                `LabeledMemoryset` and `"SCORED"` if called from `ScoredMemoryset`.
+
+        Returns:
+            Handle to the new memoryset in the OrcaCloud
+
+        Raises:
+            ValueError: If the memoryset already exists and if_exists is `"error"` or if it is
+                `"open"` and the params do not match those of the existing memoryset.
+        """
+        if embedding_model is None:
+            embedding_model = PretrainedEmbeddingModel.GTE_BASE
+
+        existing = cls._handle_if_exists(
+            name,
+            if_exists=if_exists,
+            label_names=label_names,
+            embedding_model=embedding_model,
+        )
+        if existing is not None:
+            return existing
+
+        payload: CreateMemorysetRequest = {
+            "name": name,
+            "description": description,
+            "label_names": label_names,
+            "max_seq_length_override": max_seq_length_override,
+            "index_type": index_type,
+            "index_params": index_params,
+            "hidden": hidden,
+        }
+        if memory_type is not None:
+            payload["memory_type"] = memory_type
+        if prompt is not None:
+            payload["prompt"] = prompt
+        if isinstance(embedding_model, PretrainedEmbeddingModel):
+            payload["pretrained_embedding_model_name"] = embedding_model.name
+        elif isinstance(embedding_model, FinetunedEmbeddingModel):
+            payload["finetuned_embedding_model_name_or_id"] = embedding_model.id
+        else:
+            raise ValueError("Invalid embedding model")
+
+        client = OrcaClient._resolve_client()
+        response = client.POST("/memoryset/empty", json=payload)
+        return cls.open(response["id"])
+
+    @overload
+    @classmethod
+    def from_hf_dataset(cls, name: str, hf_dataset: Dataset, background: Literal[True], **kwargs: Any) -> Self:
+        pass
+
+    @overload
+    @classmethod
+    def from_hf_dataset(cls, name: str, hf_dataset: Dataset, background: Literal[False] = False, **kwargs: Any) -> Self:
+        pass
+
+    @classmethod
+    def from_hf_dataset(
+        cls, name: str, hf_dataset: Dataset, background: bool = False, **kwargs: Any
+    ) -> Self | Job[Self]:
+        """
+        Create a new memoryset from a Hugging Face [`Dataset`][datasets.Dataset] in the OrcaCloud
+
+        This will automatically create a [`Datasource`][orca_sdk.Datasource] with the same name
+        appended with `_datasource` and use that as the datasource for the memoryset.
+
+        All features that are not specified to be used as `value_column`, `label_column`, or
+        `source_id_column` will be stored as metadata in the memoryset.
+
+        Params:
+            name: Name for the new memoryset (must be unique)
+            hf_dataset: Hugging Face dataset to create the memoryset from
+            kwargs: Additional parameters for creating the memoryset. See
+                [`create`][orca_sdk.memoryset.MemorysetBase.create] attributes for details.
+
+        Returns:
+            Handle to the new memoryset in the OrcaCloud
+        """
+        if_exists = kwargs.get("if_exists", "error")
+        existing = cls._handle_if_exists(
+            name,
+            if_exists=if_exists,
+            label_names=kwargs.get("label_names"),
+            embedding_model=kwargs.get("embedding_model"),
+        )
+        if existing is not None:
+            return existing
+
+        datasource = Datasource.from_hf_dataset(
+            f"{name}_datasource", hf_dataset, if_exists=kwargs.get("if_exists", "error")
+        )
+        kwargs["background"] = background
+        return cls.create(name, datasource=datasource, **kwargs)
+
+    @overload
+    @classmethod
+    def from_pytorch(
+        cls,
+        name: str,
+        torch_data: TorchDataLoader | TorchDataset,
+        *,
+        column_names: list[str] | None = None,
+        background: Literal[True],
+        **kwargs: Any,
+    ) -> Job[Self]:
+        pass
+
+    @overload
+    @classmethod
+    def from_pytorch(
+        cls,
+        name: str,
+        torch_data: TorchDataLoader | TorchDataset,
+        *,
+        column_names: list[str] | None = None,
+        background: Literal[False] = False,
+        **kwargs: Any,
+    ) -> Self:
+        pass
+
+    @classmethod
+    def from_pytorch(
+        cls,
+        name: str,
+        torch_data: TorchDataLoader | TorchDataset,
+        *,
+        column_names: list[str] | None = None,
+        background: bool = False,
+        **kwargs: Any,
+    ) -> Self | Job[Self]:
+        """
+        Create a new memoryset from a PyTorch [`DataLoader`][torch.utils.data.DataLoader] or
+        [`Dataset`][torch.utils.data.Dataset] in the OrcaCloud
+
+        This will automatically create a [`Datasource`][orca_sdk.Datasource] with the same name
+        appended with `_datasource` and use that as the datasource for the memoryset.
+
+        All properties that are not specified to be used as `value_column`, `label_column`, or
+        `source_id_column`, or `partition_id_column` will be stored as metadata in the memoryset.
+
+        Params:
+            name: Name for the new memoryset (must be unique)
+            torch_data: PyTorch data loader or dataset to create the memoryset from
+            column_names: If the provided dataset or data loader returns unnamed tuples, this
+                argument must be provided to specify the names of the columns.
+            background: Whether to run the operation in the background
+            kwargs: Additional parameters for creating the memoryset. See
+                [`create`][orca_sdk.memoryset.MemorysetBase.create] attributes for details.
+
+        Returns:
+            Handle to the new memoryset in the OrcaCloud
+        """
+        if_exists = kwargs.get("if_exists", "error")
+        existing = cls._handle_if_exists(
+            name,
+            if_exists=if_exists,
+            label_names=kwargs.get("label_names"),
+            embedding_model=kwargs.get("embedding_model"),
+        )
+        if existing is not None:
+            return existing
+
+        datasource = Datasource.from_pytorch(
+            f"{name}_datasource", torch_data, column_names=column_names, if_exists=kwargs.get("if_exists", "error")
+        )
+        kwargs["background"] = background
+        return cls.create(name, datasource=datasource, **kwargs)
+
+    @overload
+    @classmethod
+    def from_list(
+        cls,
+        name: str,
+        data: list[dict],
+        *,
+        background: Literal[True],
+        **kwargs: Any,
+    ) -> Job[Self]:
         pass
 
     @overload
@@ -1333,7 +1728,7 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
 
         datasource = Datasource.from_list(f"{name}_datasource", data, if_exists=kwargs.get("if_exists", "error"))
         kwargs["background"] = background
-        return cls.create(name, datasource, **kwargs)
+        return cls.create(name, datasource=datasource, **kwargs)
 
     @overload
     @classmethod
@@ -1406,7 +1801,7 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
 
         datasource = Datasource.from_dict(f"{name}_datasource", data, if_exists=kwargs.get("if_exists", "error"))
         kwargs["background"] = background
-        return cls.create(name, datasource, **kwargs)
+        return cls.create(name, datasource=datasource, **kwargs)
 
     @overload
     @classmethod
@@ -1472,7 +1867,7 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
 
         datasource = Datasource.from_pandas(f"{name}_datasource", dataframe, if_exists=kwargs.get("if_exists", "error"))
         kwargs["background"] = background
-        return cls.create(name, datasource, **kwargs)
+        return cls.create(name, datasource=datasource, **kwargs)
 
     @overload
     @classmethod
@@ -1540,7 +1935,7 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
             f"{name}_datasource", pyarrow_table, if_exists=kwargs.get("if_exists", "error")
         )
         kwargs["background"] = background
-        return cls.create(name, datasource, **kwargs)
+        return cls.create(name, datasource=datasource, **kwargs)
 
     @overload
     @classmethod
@@ -1613,7 +2008,7 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
 
         datasource = Datasource.from_disk(f"{name}_datasource", file_path, if_exists=kwargs.get("if_exists", "error"))
         kwargs["background"] = background
-        return cls.create(name, datasource, **kwargs)
+        return cls.create(name, datasource=datasource, **kwargs)
 
     @classmethod
     def open(cls, name: str) -> Self:
@@ -1830,6 +2225,10 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
 
         client = OrcaClient._resolve_client()
         metadata = client.POST("/memoryset/{name_or_id}/clone", params={"name_or_id": self.id}, json=payload)
+
+        if metadata["insertion_job_id"] is None:
+            raise ValueError("Create memoryset operation failed to produce an insertion job")
+
         job = Job(
             metadata["insertion_job_id"],
             lambda: self.open(metadata["id"]),
@@ -2482,9 +2881,11 @@ class MemorysetBase(Generic[MemoryT, MemoryLookupT], ABC):
 
                 - **`"duplicate"`**: Find potentially duplicate memories in the memoryset
                 - **`"cluster"`**: Cluster the memories in the memoryset
-                - **`"label"`**: Analyze the labels to find potential mislabelings
-                - **`"distribution"`**: Analyze the embedding distribution to populate
+                - **`"distribution"`**: Analyze the embedding distribution
                 - **`"projection"`**: Create a 2D projection of the embeddings for visualization
+                - **`"label"`**: Analyze the labels to find potential mislabelings (labeled memorysets only)
+                - **`"class_patterns"`**: Analyze class patterns and find representative memories (labeled memorysets only)
+                - **`"concepts"`**: Discover and name conceptual clusters in the memoryset (labeled memorysets only)
 
             lookup_count: Number of memories to lookup for each memory in the memoryset
             clear_metrics: Whether to clear any existing metrics from the memories before running the analysis
@@ -2590,35 +2991,246 @@ class LabeledMemoryset(MemorysetBase[LabeledMemory, LabeledMemoryLookup]):
     """
     A Handle to a collection of memories with labels in the OrcaCloud
 
-    Attributes:
-        id: Unique identifier for the memoryset
-        name: Unique name of the memoryset
-        description: Description of the memoryset
-        label_names: Names for the class labels in the memoryset
-        length: Number of memories in the memoryset
-        embedding_model: Embedding model used to embed the memory values for semantic search
-        created_at: When the memoryset was created, automatically generated on create
-        updated_at: When the memoryset was last updated, automatically updated on updates
-    """
+    Attributes:
+        id: Unique identifier for the memoryset
+        name: Unique name of the memoryset
+        description: Description of the memoryset
+        label_names: Names for the class labels in the memoryset
+        length: Number of memories in the memoryset
+        embedding_model: Embedding model used to embed the memory values for semantic search
+        created_at: When the memoryset was created, automatically generated on create
+        updated_at: When the memoryset was last updated, automatically updated on updates
+    """
+
+    label_names: list[str]
+    memory_type: MemoryType = "LABELED"
+
+    def __init__(self, metadata: MemorysetMetadata):
+        super().__init__(metadata)
+        assert metadata["label_names"] is not None
+        self.label_names = metadata["label_names"]
+
+    def __eq__(self, other) -> bool:
+        return isinstance(other, LabeledMemoryset) and self.id == other.id
+
+    @overload
+    @classmethod
+    def create(
+        cls,
+        name: str,
+        *,
+        datasource: None = None,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        description: str | None = None,
+        label_names: list[str],
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
+        hidden: bool = False,
+    ) -> Self:
+        pass
+
+    @overload
+    @classmethod
+    def create(
+        cls,
+        name: str,
+        *,
+        datasource: Datasource,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        value_column: str = "value",
+        label_column: str | None = "label",
+        source_id_column: str | None = None,
+        partition_id_column: str | None = None,
+        description: str | None = None,
+        label_names: list[str] | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        remove_duplicates: bool = True,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
+        background: Literal[True],
+        hidden: bool = False,
+        subsample: int | float | None = None,
+    ) -> Job[Self]:
+        pass
+
+    @overload
+    @classmethod
+    def create(
+        cls,
+        name: str,
+        *,
+        datasource: Datasource,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        value_column: str = "value",
+        label_column: str | None = "label",
+        source_id_column: str | None = None,
+        partition_id_column: str | None = None,
+        description: str | None = None,
+        label_names: list[str] | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        remove_duplicates: bool = True,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
+        background: Literal[False] = False,
+        hidden: bool = False,
+        subsample: int | float | None = None,
+    ) -> Self:
+        pass
+
+    @classmethod
+    def create(  # type: ignore[override]
+        cls,
+        name: str,
+        *,
+        datasource: Datasource | None = None,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        value_column: str = "value",
+        label_column: str | None = "label",
+        source_id_column: str | None = None,
+        partition_id_column: str | None = None,
+        description: str | None = None,
+        label_names: list[str] | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        remove_duplicates: bool = True,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
+        background: bool = False,
+        hidden: bool = False,
+        subsample: int | float | None = None,
+    ) -> Self | Job[Self]:
+        """
+        Create a new labeled memoryset in the OrcaCloud
+
+        If `datasource` is provided, all columns from the datasource that are not specified in the
+        `value_column`, `label_column`, `source_id_column`, or `partition_id_column` will be stored
+        as metadata in the memoryset.
 
-    label_names: list[str]
-    memory_type: MemoryType = "LABELED"
+        If `datasource` is omitted (None), an empty memoryset will be created with no initial memories.
+        You can add memories later using the `insert` method.
 
-    def __init__(self, metadata: MemorysetMetadata):
-        super().__init__(metadata)
-        assert metadata["label_names"] is not None
-        self.label_names = metadata["label_names"]
+        Params:
+            name: Name for the new memoryset (must be unique)
+            datasource: Optional source data to populate the memories in the memoryset. If omitted,
+                an empty memoryset will be created.
+            embedding_model: Embedding model to use for embedding memory values for semantic search.
+                If not provided, a default embedding model for the memoryset will be used.
+            value_column: Name of the column in the datasource that contains the memory values
+            label_column: Name of the column in the datasource that contains the memory labels.
+                Must contain categorical values as integers or strings. String labels will be
+                converted to integers with the unique strings extracted as `label_names`. To create
+                a memoryset with all none labels, set to `None`.
+            source_id_column: Optional name of the column in the datasource that contains the ids in
+                the system of reference
+            partition_id_column: Optional name of the column in the datasource that contains the partition ids
+            description: Optional description for the memoryset, this will be used in agentic flows,
+                so make sure it is concise and describes the contents of your memoryset not the
+                datasource or the embedding model.
+            label_names: List of human-readable names for the labels in the memoryset, must match
+                the number of labels in the `label_column`. Will be automatically inferred if string
+                labels are provided or if a [Dataset][datasets.Dataset] with a
+                [`ClassLabel`][datasets.ClassLabel] feature for labels is used as the datasource
+            max_seq_length_override: Maximum sequence length of values in the memoryset, if the
+                value is longer than this it will be truncated, will default to the model's max
+                sequence length if not provided
+            prompt: Optional prompt to use when embedding documents/memories for storage
+            remove_duplicates: Whether to remove duplicates from the datasource before inserting
+                into the memoryset
+            index_type: Type of vector index to use for the memoryset, defaults to `"FLAT"`. Valid
+                values are `"FLAT"`, `"IVF_FLAT"`, `"IVF_SQ8"`, `"IVF_PQ"`, `"HNSW"`, and `"DISKANN"`.
+            index_params: Parameters for the vector index, defaults to `{}`
+            if_exists: What to do if a memoryset with the same name already exists, defaults to
+                `"error"`. Other option is `"open"` to open the existing memoryset.
+            background: Whether to run the operation none blocking and return a job handle
+            hidden: Whether the memoryset should be hidden
 
-    def __eq__(self, other) -> bool:
-        return isinstance(other, LabeledMemoryset) and self.id == other.id
+        Returns:
+            Handle to the new memoryset in the OrcaCloud
+
+        Raises:
+            ValueError: If the memoryset already exists and if_exists is `"error"` or if it is
+                `"open"` and the params do not match those of the existing memoryset.
+        """
+        if datasource is None:
+            return super().create(
+                name,
+                datasource=None,
+                embedding_model=embedding_model,
+                description=description,
+                label_names=label_names,
+                max_seq_length_override=max_seq_length_override,
+                prompt=prompt,
+                index_type=index_type,
+                index_params=index_params,
+                if_exists=if_exists,
+                hidden=hidden,
+                memory_type="LABELED",
+            )
+        else:
+            # Type narrowing: datasource is definitely Datasource here
+            assert datasource is not None
+            if background:
+                return super().create(
+                    name,
+                    datasource=datasource,
+                    label_column=label_column,
+                    score_column=None,
+                    embedding_model=embedding_model,
+                    value_column=value_column,
+                    source_id_column=source_id_column,
+                    partition_id_column=partition_id_column,
+                    description=description,
+                    label_names=label_names,
+                    max_seq_length_override=max_seq_length_override,
+                    prompt=prompt,
+                    remove_duplicates=remove_duplicates,
+                    index_type=index_type,
+                    index_params=index_params,
+                    if_exists=if_exists,
+                    background=True,
+                    hidden=hidden,
+                    subsample=subsample,
+                    memory_type="LABELED",
+                )
+            else:
+                return super().create(
+                    name,
+                    datasource=datasource,
+                    label_column=label_column,
+                    score_column=None,
+                    embedding_model=embedding_model,
+                    value_column=value_column,
+                    source_id_column=source_id_column,
+                    partition_id_column=partition_id_column,
+                    description=description,
+                    label_names=label_names,
+                    max_seq_length_override=max_seq_length_override,
+                    prompt=prompt,
+                    remove_duplicates=remove_duplicates,
+                    index_type=index_type,
+                    index_params=index_params,
+                    if_exists=if_exists,
+                    background=False,
+                    hidden=hidden,
+                    subsample=subsample,
+                    memory_type="LABELED",
+                )
 
     @overload
     @classmethod
-    def create(
+    def from_datasource(
         cls,
         name: str,
-        datasource: Datasource,
         *,
+        datasource: Datasource,
         embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
         value_column: str = "value",
         label_column: str | None = "label",
@@ -2640,11 +3252,11 @@ class LabeledMemoryset(MemorysetBase[LabeledMemory, LabeledMemoryLookup]):
 
     @overload
     @classmethod
-    def create(
+    def from_datasource(
         cls,
         name: str,
-        datasource: Datasource,
         *,
+        datasource: Datasource,
         embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
         value_column: str = "value",
         label_column: str | None = "label",
@@ -2665,11 +3277,11 @@ class LabeledMemoryset(MemorysetBase[LabeledMemory, LabeledMemoryLookup]):
         pass
 
     @classmethod
-    def create(  # type: ignore[override]
+    def from_datasource(  # type: ignore[override]
         cls,
         name: str,
-        datasource: Datasource,
         *,
+        datasource: Datasource,
         embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
         value_column: str = "value",
         label_column: str | None = "label",
@@ -2688,14 +3300,16 @@ class LabeledMemoryset(MemorysetBase[LabeledMemory, LabeledMemoryLookup]):
         subsample: int | float | None = None,
     ) -> Self | Job[Self]:
         """
-        Create a new labeled memoryset in the OrcaCloud
+        Create a new labeled memoryset in the OrcaCloud from a datasource.
 
+        This is a convenience method that is equivalent to calling `create` with a datasource.
         All columns from the datasource that are not specified in the `value_column`,
-        `label_column`, `source_id_column`, or `partition_id_column` will be stored as metadata in the memoryset.
+        `label_column`, `source_id_column`, or `partition_id_column` will be stored as metadata
+        in the memoryset.
 
         Params:
             name: Name for the new memoryset (must be unique)
-            datasource: Source data to populate the memories in the memoryset
+            datasource: Source data to populate the memories in the memoryset.
             embedding_model: Embedding model to use for embedding memory values for semantic search.
                 If not provided, a default embedding model for the memoryset will be used.
             value_column: Name of the column in the datasource that contains the memory values
@@ -2724,8 +3338,10 @@ class LabeledMemoryset(MemorysetBase[LabeledMemory, LabeledMemoryLookup]):
             index_params: Parameters for the vector index, defaults to `{}`
             if_exists: What to do if a memoryset with the same name already exists, defaults to
                 `"error"`. Other option is `"open"` to open the existing memoryset.
-            background: Whether to run the operation none blocking and return a job handle
+            background: Whether to run the operation none blocking and return a job handle.
             hidden: Whether the memoryset should be hidden
+            subsample: Optional number (int) of rows to insert or fraction (float in (0, 1]) of the
+                datasource to insert. Use to limit the size of the initial memoryset.
 
         Returns:
             Handle to the new memoryset in the OrcaCloud
@@ -2734,28 +3350,52 @@ class LabeledMemoryset(MemorysetBase[LabeledMemory, LabeledMemoryLookup]):
             ValueError: If the memoryset already exists and if_exists is `"error"` or if it is
                 `"open"` and the params do not match those of the existing memoryset.
         """
-        return super().create(
-            name,
-            datasource,
-            label_column=label_column,
-            score_column=None,
-            embedding_model=embedding_model,
-            value_column=value_column,
-            source_id_column=source_id_column,
-            partition_id_column=partition_id_column,
-            description=description,
-            label_names=label_names,
-            max_seq_length_override=max_seq_length_override,
-            prompt=prompt,
-            remove_duplicates=remove_duplicates,
-            index_type=index_type,
-            index_params=index_params,
-            if_exists=if_exists,
-            background=background,
-            hidden=hidden,
-            subsample=subsample,
-            memory_type="LABELED",
-        )
+        if background:
+            return super().create(
+                name,
+                datasource=datasource,
+                label_column=label_column,
+                score_column=None,
+                embedding_model=embedding_model,
+                value_column=value_column,
+                source_id_column=source_id_column,
+                partition_id_column=partition_id_column,
+                description=description,
+                label_names=label_names,
+                max_seq_length_override=max_seq_length_override,
+                prompt=prompt,
+                remove_duplicates=remove_duplicates,
+                index_type=index_type,
+                index_params=index_params,
+                if_exists=if_exists,
+                background=True,
+                hidden=hidden,
+                subsample=subsample,
+                memory_type="LABELED",
+            )
+        else:
+            return super().create(
+                name,
+                datasource=datasource,
+                label_column=label_column,
+                score_column=None,
+                embedding_model=embedding_model,
+                value_column=value_column,
+                source_id_column=source_id_column,
+                partition_id_column=partition_id_column,
+                description=description,
+                label_names=label_names,
+                max_seq_length_override=max_seq_length_override,
+                prompt=prompt,
+                remove_duplicates=remove_duplicates,
+                index_type=index_type,
+                index_params=index_params,
+                if_exists=if_exists,
+                background=False,
+                hidden=hidden,
+                subsample=subsample,
+                memory_type="LABELED",
+            )
 
     def display_label_analysis(self):
         """
@@ -2793,8 +3433,26 @@ class ScoredMemoryset(MemorysetBase[ScoredMemory, ScoredMemoryLookup]):
     def create(
         cls,
         name: str,
-        datasource: Datasource,
         *,
+        datasource: None = None,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        description: str | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
+        hidden: bool = False,
+    ) -> Self:
+        pass
+
+    @overload
+    @classmethod
+    def create(
+        cls,
+        name: str,
+        *,
+        datasource: Datasource,
         embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
         value_column: str = "value",
         score_column: str | None = "score",
@@ -2818,8 +3476,8 @@ class ScoredMemoryset(MemorysetBase[ScoredMemory, ScoredMemoryLookup]):
     def create(
         cls,
         name: str,
-        datasource: Datasource,
         *,
+        datasource: Datasource,
         embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
         score_column: str | None = "score",
         value_column: str = "value",
@@ -2842,8 +3500,8 @@ class ScoredMemoryset(MemorysetBase[ScoredMemory, ScoredMemoryLookup]):
     def create(  # type: ignore[override]
         cls,
         name: str,
-        datasource: Datasource,
         *,
+        datasource: Datasource | None = None,
         embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
         value_column: str = "value",
         score_column: str | None = "score",
@@ -2863,12 +3521,17 @@ class ScoredMemoryset(MemorysetBase[ScoredMemory, ScoredMemoryLookup]):
         """
         Create a new scored memoryset in the OrcaCloud
 
-        All columns from the datasource that are not specified in the `value_column`,
-        `score_column`, `source_id_column`, or `partition_id_column` will be stored as metadata in the memoryset.
+        If `datasource` is provided, all columns from the datasource that are not specified in the
+        `value_column`, `score_column`, `source_id_column`, or `partition_id_column` will be stored
+        as metadata in the memoryset.
+
+        If `datasource` is omitted (None), an empty memoryset will be created with no initial memories.
+        You can add memories later using the `insert` method.
 
         Params:
             name: Name for the new memoryset (must be unique)
-            datasource: Source data to populate the memories in the memoryset
+            datasource: Optional source data to populate the memories in the memoryset. If omitted,
+                an empty memoryset will be created.
             embedding_model: Embedding model to use for embedding memory values for semantic search.
                 If not provided, a default embedding model for the memoryset will be used.
             value_column: Name of the column in the datasource that contains the memory values
@@ -2901,23 +3564,222 @@ class ScoredMemoryset(MemorysetBase[ScoredMemory, ScoredMemoryLookup]):
             ValueError: If the memoryset already exists and if_exists is `"error"` or if it is
                 `"open"` and the params do not match those of the existing memoryset.
         """
-        return super().create(
-            name,
-            datasource,
-            embedding_model=embedding_model,
-            value_column=value_column,
-            score_column=score_column,
-            source_id_column=source_id_column,
-            partition_id_column=partition_id_column,
-            description=description,
-            max_seq_length_override=max_seq_length_override,
-            prompt=prompt,
-            remove_duplicates=remove_duplicates,
-            index_type=index_type,
-            index_params=index_params,
-            if_exists=if_exists,
-            background=background,
-            hidden=hidden,
-            subsample=subsample,
-            memory_type="SCORED",
-        )
+        if datasource is None:
+            return super().create(
+                name,
+                datasource=None,
+                embedding_model=embedding_model,
+                description=description,
+                max_seq_length_override=max_seq_length_override,
+                prompt=prompt,
+                index_type=index_type,
+                index_params=index_params,
+                if_exists=if_exists,
+                hidden=hidden,
+                memory_type="SCORED",
+            )
+        else:
+            # Type narrowing: datasource is definitely Datasource here
+            assert datasource is not None
+            if background:
+                return super().create(
+                    name,
+                    datasource=datasource,
+                    embedding_model=embedding_model,
+                    value_column=value_column,
+                    score_column=score_column,
+                    source_id_column=source_id_column,
+                    partition_id_column=partition_id_column,
+                    description=description,
+                    max_seq_length_override=max_seq_length_override,
+                    prompt=prompt,
+                    remove_duplicates=remove_duplicates,
+                    index_type=index_type,
+                    index_params=index_params,
+                    if_exists=if_exists,
+                    background=True,
+                    hidden=hidden,
+                    subsample=subsample,
+                    memory_type="SCORED",
+                )
+            else:
+                return super().create(
+                    name,
+                    datasource=datasource,
+                    embedding_model=embedding_model,
+                    value_column=value_column,
+                    score_column=score_column,
+                    source_id_column=source_id_column,
+                    partition_id_column=partition_id_column,
+                    description=description,
+                    max_seq_length_override=max_seq_length_override,
+                    prompt=prompt,
+                    remove_duplicates=remove_duplicates,
+                    index_type=index_type,
+                    index_params=index_params,
+                    if_exists=if_exists,
+                    background=False,
+                    hidden=hidden,
+                    subsample=subsample,
+                    memory_type="SCORED",
+                )
+
+    @overload
+    @classmethod
+    def from_datasource(
+        cls,
+        name: str,
+        *,
+        datasource: Datasource,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        value_column: str = "value",
+        score_column: str | None = "score",
+        source_id_column: str | None = None,
+        partition_id_column: str | None = None,
+        description: str | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        remove_duplicates: bool = True,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
+        background: Literal[True],
+        hidden: bool = False,
+        subsample: int | float | None = None,
+    ) -> Job[Self]:
+        pass
+
+    @overload
+    @classmethod
+    def from_datasource(
+        cls,
+        name: str,
+        *,
+        datasource: Datasource,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        score_column: str | None = "score",
+        value_column: str = "value",
+        source_id_column: str | None = None,
+        partition_id_column: str | None = None,
+        description: str | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        remove_duplicates: bool = True,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
+        background: Literal[False] = False,
+        hidden: bool = False,
+        subsample: int | float | None = None,
+    ) -> Self:
+        pass
+
+    @classmethod
+    def from_datasource(  # type: ignore[override]
+        cls,
+        name: str,
+        *,
+        datasource: Datasource,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        value_column: str = "value",
+        score_column: str | None = "score",
+        source_id_column: str | None = None,
+        partition_id_column: str | None = None,
+        description: str | None = None,
+        max_seq_length_override: int | None = None,
+        prompt: str | None = None,
+        remove_duplicates: bool = True,
+        index_type: IndexType = "FLAT",
+        index_params: dict[str, Any] = {},
+        if_exists: CreateMode = "error",
+        background: bool = False,
+        hidden: bool = False,
+        subsample: int | float | None = None,
+    ) -> Self | Job[Self]:
+        """
+        Create a new scored memoryset in the OrcaCloud from a datasource.
+
+        This is a convenience method that is equivalent to calling `create` with a datasource.
+        All columns from the datasource that are not specified in the `value_column`,
+        `score_column`, `source_id_column`, or `partition_id_column` will be stored as metadata
+        in the memoryset.
+
+        Params:
+            name: Name for the new memoryset (must be unique)
+            datasource: Source data to populate the memories in the memoryset.
+            embedding_model: Embedding model to use for embedding memory values for semantic search.
+                If not provided, a default embedding model for the memoryset will be used.
+            value_column: Name of the column in the datasource that contains the memory values
+            score_column: Name of the column in the datasource that contains the memory scores. Must
+                contain numerical values. To create a memoryset with all none scores, set to `None`.
+            source_id_column: Optional name of the column in the datasource that contains the ids in
+                the system of reference
+            partition_id_column: Optional name of the column in the datasource that contains the partition ids
+            description: Optional description for the memoryset, this will be used in agentic flows,
+                so make sure it is concise and describes the contents of your memoryset not the
+                datasource or the embedding model.
+            max_seq_length_override: Maximum sequence length of values in the memoryset, if the
+                value is longer than this it will be truncated, will default to the model's max
+                sequence length if not provided
+            prompt: Optional prompt to use when embedding documents/memories for storage
+            remove_duplicates: Whether to remove duplicates from the datasource before inserting
+                into the memoryset
+            index_type: Type of vector index to use for the memoryset, defaults to `"FLAT"`. Valid
+                values are `"FLAT"`, `"IVF_FLAT"`, `"IVF_SQ8"`, `"IVF_PQ"`, `"HNSW"`, and `"DISKANN"`.
+            index_params: Parameters for the vector index, defaults to `{}`
+            if_exists: What to do if a memoryset with the same name already exists, defaults to
+                `"error"`. Other option is `"open"` to open the existing memoryset.
+            background: Whether to run the operation none blocking and return a job handle.
+            hidden: Whether the memoryset should be hidden
+            subsample: Optional number (int) of rows to insert or fraction (float in (0, 1]) of the
+                datasource to insert. Use to limit the size of the initial memoryset.
+
+        Returns:
+            Handle to the new memoryset in the OrcaCloud
+
+        Raises:
+            ValueError: If the memoryset already exists and if_exists is `"error"` or if it is
+                `"open"` and the params do not match those of the existing memoryset.
+        """
+        if background:
+            return super().create(
+                name,
+                datasource=datasource,
+                embedding_model=embedding_model,
+                value_column=value_column,
+                score_column=score_column,
+                source_id_column=source_id_column,
+                partition_id_column=partition_id_column,
+                description=description,
+                max_seq_length_override=max_seq_length_override,
+                prompt=prompt,
+                remove_duplicates=remove_duplicates,
+                index_type=index_type,
+                index_params=index_params,
+                if_exists=if_exists,
+                background=True,
+                hidden=hidden,
+                subsample=subsample,
+                memory_type="SCORED",
+            )
+        else:
+            return super().create(
+                name,
+                datasource=datasource,
+                embedding_model=embedding_model,
+                value_column=value_column,
+                score_column=score_column,
+                source_id_column=source_id_column,
+                partition_id_column=partition_id_column,
+                description=description,
+                max_seq_length_override=max_seq_length_override,
+                prompt=prompt,
+                remove_duplicates=remove_duplicates,
+                index_type=index_type,
+                index_params=index_params,
+                if_exists=if_exists,
+                background=False,
+                hidden=hidden,
+                subsample=subsample,
+                memory_type="SCORED",
+            )