orca-sdk 0.0.78__py3-none-any.whl

orca_sdk/memoryset.py

@@ -0,0 +1,1209 @@
+from __future__ import annotations
+
+import logging
+from datetime import datetime, timedelta
+from os import PathLike
+from typing import Any, Iterable, Literal, cast, overload
+
+import pandas as pd
+import pyarrow as pa
+from datasets import Dataset
+from torch.utils.data import DataLoader as TorchDataLoader
+from torch.utils.data import Dataset as TorchDataset
+
+from ._generated_api_client.api import (
+    clone_memoryset,
+    create_analysis,
+    create_embedding_evaluation,
+    create_memoryset,
+    delete_datasource,
+    delete_memories,
+    delete_memoryset,
+    get_analysis,
+    get_datasource,
+    get_embedding_evaluation,
+    get_memories,
+    get_memory,
+    get_memoryset,
+    insert_memories_gpu,
+    list_datasources,
+    list_memorysets,
+    memoryset_lookup_gpu,
+    query_memoryset,
+    update_memories_gpu,
+    update_memory_gpu,
+)
+from ._generated_api_client.models import (
+    AnalyzeNeighborLabelsResult,
+    CloneLabeledMemorysetRequest,
+    ColumnType,
+    CreateLabeledMemorysetRequest,
+    DatasourceMetadata,
+    DeleteMemoriesRequest,
+    EmbeddingEvaluationRequest,
+    EmbeddingEvaluationResponse,
+    FilterItem,
+    FilterItemOp,
+    FindDuplicatesAnalysisResult,
+    GetMemoriesRequest,
+)
+from ._generated_api_client.models import LabeledMemory as LabeledMemoryResponse
+from ._generated_api_client.models import (
+    LabeledMemoryInsert,
+    LabeledMemoryInsertMetadata,
+)
+from ._generated_api_client.models import (
+    LabeledMemoryLookup as LabeledMemoryLookupResponse,
+)
+from ._generated_api_client.models import (
+    LabeledMemoryMetrics,
+    LabeledMemorysetMetadata,
+    LabeledMemoryUpdate,
+    LabeledMemoryUpdateMetadataType0,
+    LabelPredictionMemoryLookup,
+    ListMemoriesRequest,
+    LookupRequest,
+    MemorysetAnalysisRequest,
+    MemorysetAnalysisRequestType,
+    PretrainedEmbeddingModelName,
+    TaskStatus,
+)
+from ._generated_api_client.types import UNSET as CLIENT_UNSET
+from ._utils.common import UNSET, CreateMode, DropMode
+from ._utils.task import wait_for_task
+from .datasource import Datasource
+from .embedding_model import (
+    FinetunedEmbeddingModel,
+    PretrainedEmbeddingModel,
+    _EmbeddingModel,
+)
+
+FilterOperation = Literal["==", "!=", ">", ">=", "<", "<=", "in", "not in", "like"]
+"""
+Operations that can be used in a filter expression.
+"""
+
+FilterValue = str | int | float | bool | datetime | None | list[str] | list[int] | list[float] | list[bool]
+"""
+Values that can be used in a filter expression.
+"""
+
+FilterItemTuple = tuple[str, FilterOperation, FilterValue]
+"""
+Filter expression consisting of a field, an operator, and a value:
+
+* **`field`**: The field to filter on.
+* **`operation`**: The operation to apply to the field and value.
+* **`value`**: The value to compare the field against.
+
+Examples:
+    >>> ("label", "==", 0)
+    >>> ("metadata.author", "like", "John")
+    >>> ("source_id", "in", ["123", "456"])
+"""
+
+
+DEFAULT_COLUMN_NAMES = {"value", "label", "source_id"}
+FORBIDDEN_METADATA_COLUMN_NAMES = {"memory_id", "memory_version", "embedding", "created_at", "updated_at", "metrics"}
+
+
+def _parse_filter_item_from_tuple(input: FilterItemTuple) -> FilterItem:
+    field = input[0].split(".")
+    if len(field) == 1 and field[0] not in DEFAULT_COLUMN_NAMES | FORBIDDEN_METADATA_COLUMN_NAMES:
+        field = ["metadata", field[0]]
+    op = FilterItemOp(input[1])
+    value = input[2]
+    return FilterItem(field=field, op=op, value=value)
+
+
+def _parse_memory_insert(memory: dict[str, Any]) -> LabeledMemoryInsert:
+    value = memory.get("value")
+    if not isinstance(value, str):
+        raise ValueError("Memory value must be a string")
+    label = memory.get("label")
+    if not isinstance(label, int):
+        raise ValueError("Memory label must be an integer")
+    source_id = memory.get("source_id")
+    if source_id and not isinstance(source_id, str):
+        raise ValueError("Memory source_id must be a string")
+    metadata = LabeledMemoryInsertMetadata.from_dict({k: v for k, v in memory.items() if k not in DEFAULT_COLUMN_NAMES})
+    if any(k in metadata for k in FORBIDDEN_METADATA_COLUMN_NAMES):
+        raise ValueError(f"The following column names are reserved: {', '.join(FORBIDDEN_METADATA_COLUMN_NAMES)}")
+    return LabeledMemoryInsert(value=value, label=label, source_id=source_id, metadata=metadata)
+
+
+def _parse_memory_update(update: dict[str, Any]) -> LabeledMemoryUpdate:
+    if "memory_id" not in update:
+        raise ValueError("memory_id must be specified in the update dictionary")
+    memory_id = update["memory_id"]
+    if not isinstance(memory_id, str):
+        raise ValueError("memory_id must be a string")
+    value = update.get("value", CLIENT_UNSET)
+    if value is not CLIENT_UNSET and not isinstance(value, str):
+        raise ValueError("value must be a string or unset")
+    label = update.get("label", CLIENT_UNSET)
+    if label is not CLIENT_UNSET and not isinstance(label, int):
+        raise ValueError("label must be an integer or unset")
+    source_id = update.get("source_id", CLIENT_UNSET)
+    if source_id is not CLIENT_UNSET and not isinstance(source_id, str):
+        raise ValueError("source_id must be a string or unset")
+    metadata = LabeledMemoryUpdateMetadataType0.from_dict(
+        {k: v for k, v in update.items() if k not in DEFAULT_COLUMN_NAMES | {"memory_id"}}
+    )
+    if any(k in metadata for k in FORBIDDEN_METADATA_COLUMN_NAMES):
+        raise ValueError(f"Cannot update the following metadata keys: {', '.join(FORBIDDEN_METADATA_COLUMN_NAMES)}")
+    return LabeledMemoryUpdate(memory_id=memory_id, value=value, label=label, source_id=source_id, metadata=metadata)
+
+
+class LabeledMemory:
+    """
+    A row of the [`LabeledMemoryset`][orca_sdk.LabeledMemoryset]
+
+    Attributes:
+        value: Value represented by the row
+        embedding: Embedding of the value of the memory for semantic search, automatically generated
+            with the [`LabeledMemoryset.embedding_model`][orca_sdk.LabeledMemoryset]
+        label: Class label of the memory
+        label_name: Human-readable name of the label, automatically populated from the
+            [`LabeledMemoryset.label_names`][orca_sdk.LabeledMemoryset]
+        source_id: Optional unique identifier of the memory in a system of reference
+        metrics: Metrics about the memory, generated when running an analysis on the
+            [`LabeledMemoryset`][orca_sdk.LabeledMemoryset]
+        metadata: Metadata associated with the memory that is not used in the model. Metadata
+            properties are also accessible as individual attributes on the instance.
+        memory_id: Unique identifier for the memory, automatically generated on insert
+        memory_version: Version of the memory, automatically updated when the label or value changes
+        created_at: When the memory was created, automatically generated on insert
+        updated_at: When the memory was last updated, automatically updated on update
+
+    ## Other Attributes:
+    * **`...`** (<code>[str][str] | [float][float] | [int][int] | [bool][bool] | None</code>): All metadata properties can be accessed as attributes
+    """
+
+    value: str
+    embedding: list[float]
+    label: int
+    label_name: str | None
+    source_id: str | None
+    created_at: datetime
+    updated_at: datetime
+    metadata: dict[str, str | float | int | bool | None]
+    metrics: dict[str, Any]
+    memory_id: str
+    memory_version: int
+
+    def __init__(
+        self,
+        memoryset_id: str,
+        memory: LabeledMemoryResponse | LabeledMemoryLookupResponse | LabelPredictionMemoryLookup,
+    ):
+        # for internal use only, do not document
+        self.memoryset_id = memoryset_id
+        self.memory_id = memory.memory_id
+        self.memory_version = memory.memory_version
+        self.value = memory.value
+        self.embedding = memory.embedding
+        self.label = memory.label
+        self.label_name = memory.label_name
+        self.source_id = memory.source_id
+        self.created_at = memory.created_at
+        self.updated_at = memory.updated_at
+        self.metadata = memory.metadata.to_dict()
+        self.metrics = memory.metrics.to_dict() if memory.metrics else {}
+
+    def __getattr__(self, key: str) -> Any:
+        if key.startswith("__") or key not in self.metadata:
+            raise AttributeError(f"{key} is not a valid attribute")
+        return self.metadata[key]
+
+    def __repr__(self) -> str:
+        return (
+            "LabeledMemory({ "
+            + f"label: {('<' + self.label_name + ': ' + str(self.label) + '>') if self.label_name else str(self.label)}"
+            + f", value: '{self.value[:100] + '...' if len(self.value) > 100 else self.value}'"
+            + (f", source_id: '{self.source_id}'" if self.source_id is not None else "")
+            + " })"
+        )
+
+    def __eq__(self, other: object) -> bool:
+        return isinstance(other, LabeledMemory) and self.memory_id == other.memory_id
+
+    def update(
+        self,
+        *,
+        value: str = UNSET,
+        label: int = UNSET,
+        source_id: str | None = UNSET,
+        **metadata: None | bool | float | int | str,
+    ) -> LabeledMemory:
+        """
+        Update the memory with new values
+
+        Note:
+            If a field is not provided, it will default to [UNSET][orca_sdk.UNSET] and not be updated.
+
+        Params:
+            value: New value of the memory
+            label: New label of the memory
+            source_id: New source ID of the memory
+            **metadata: New values for metadata properties
+
+        Returns:
+            The updated memory
+        """
+        response = update_memory_gpu(
+            self.memoryset_id,
+            body=_parse_memory_update(
+                {"memory_id": self.memory_id}
+                | ({"value": value} if value is not UNSET else {})
+                | ({"label": label} if label is not UNSET else {})
+                | ({"source_id": source_id} if source_id is not UNSET else {})
+                | metadata
+            ),
+        )
+        self.__dict__.update(LabeledMemory(self.memoryset_id, response).__dict__)
+        return self
+
+
+class LabeledMemoryLookup(LabeledMemory):
+    """
+    Lookup result for a memory in a memoryset
+
+    Attributes:
+        lookup_score: Similarity between the memory embedding and search query embedding
+        attention_weight: Weight the model assigned to the memory during prediction if this lookup
+            happened as part of a prediction
+        value: Value represented by the row
+        embedding: Embedding of the value of the memory for semantic search, automatically generated
+            with the [`LabeledMemoryset.embedding_model`][orca_sdk.LabeledMemoryset]
+        label: Class label of the memory
+        label_name: Human-readable name of the label, automatically populated from the
+            [`LabeledMemoryset.label_names`][orca_sdk.LabeledMemoryset]
+        source_id: Optional unique identifier of the memory in a system of reference
+        metrics: Metrics about the memory, generated when running an analysis on the
+            [`LabeledMemoryset`][orca_sdk.LabeledMemoryset]
+        metadata: Metadata associated with the memory that is not used in the model. Metadata
+            properties are also accessible as individual attributes on the instance.
+        memory_id: The unique identifier for the memory, automatically generated on insert
+        memory_version: The version of the memory, automatically updated when the label or value changes
+        created_at: When the memory was created, automatically generated on insert
+        updated_at: When the memory was last updated, automatically updated on update
+
+    ## Other Attributes:
+    * **`...`** (<code>[str][str] | [float][float] | [int][int] | [bool][bool] | None</code>): All metadata properties can be accessed as attributes
+    """
+
+    lookup_score: float
+    attention_weight: float | None
+
+    def __init__(self, memoryset_id: str, memory_lookup: LabeledMemoryLookupResponse | LabelPredictionMemoryLookup):
+        # for internal use only, do not document
+        super().__init__(memoryset_id, memory_lookup)
+        self.lookup_score = memory_lookup.lookup_score
+        self.attention_weight = (
+            memory_lookup.attention_weight if isinstance(memory_lookup, LabelPredictionMemoryLookup) else None
+        )
+
+    def __repr__(self) -> str:
+        return (
+            "LabeledMemoryLookup({ "
+            + f"label: {('<' + self.label_name + ': ' + str(self.label) + '>') if self.label_name else str(self.label)}"
+            + f", lookup_score: {self.lookup_score:.2f}"
+            + (f", attention_weight: {self.attention_weight:.2f}" if self.attention_weight is not None else "")
+            + f", value: '{self.value[:100] + '...' if len(self.value) > 100 else self.value}'"
+            + (f", source_id: '{self.source_id}'" if self.source_id is not None else "")
+            + " })"
+        )
+
+
+class LabeledMemoryset:
+    """
+    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
+        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
+    """
+
+    id: str
+    name: str
+    label_names: list[str]
+    length: int
+    created_at: datetime
+    updated_at: datetime
+    insertion_status: TaskStatus
+    embedding_model: _EmbeddingModel
+
+    def __init__(self, metadata: LabeledMemorysetMetadata):
+        # for internal use only, do not document
+        if metadata.pretrained_embedding_model_name:
+            self.embedding_model = PretrainedEmbeddingModel._get(metadata.pretrained_embedding_model_name)
+        elif metadata.finetuned_embedding_model_id:
+            self.embedding_model = FinetunedEmbeddingModel.open(metadata.finetuned_embedding_model_id)
+        else:
+            raise ValueError("Either pretrained_embedding_model_name or finetuned_embedding_model_id must be provided")
+        self.id = metadata.id
+        self.name = metadata.name
+        self.label_names = metadata.label_names
+        self.length = metadata.length
+        self.created_at = metadata.created_at
+        self.updated_at = metadata.updated_at
+        self.insertion_status = metadata.insertion_status
+        self._last_refresh = datetime.now()
+
+    def __eq__(self, other) -> bool:
+        return isinstance(other, LabeledMemoryset) and self.id == other.id
+
+    def __repr__(self) -> str:
+        return (
+            "LabeledMemoryset({\n"
+            f"    name: '{self.name}',\n"
+            f"    length: {self.length},\n"
+            f"    label_names: {self.label_names},\n"
+            f"    embedding_model: {self.embedding_model},\n"
+            "})"
+        )
+
+    @classmethod
+    def create(
+        cls,
+        name: str,
+        datasource: Datasource,
+        *,
+        embedding_model: FinetunedEmbeddingModel | PretrainedEmbeddingModel | None = None,
+        value_column: str = "value",
+        label_column: str = "label",
+        source_id_column: str | None = None,
+        label_names: list[str] | None = None,
+        max_seq_length_override: int | None = None,
+        if_exists: CreateMode = "error",
+    ) -> LabeledMemoryset:
+        """
+        Create a new memoryset in the OrcaCloud
+
+        All columns from the datasource that are not specified in the `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)
+            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,
+                these must be contiguous integers starting from 0
+            source_id_column: Optional name of the column in the datasource that contains the ids in
+                the system of reference
+            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 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
+            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.
+
+        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.CDE_SMALL
+
+        if cls.exists(name):
+            if if_exists == "error":
+                raise ValueError(f"Memoryset with name {name} already exists")
+            elif if_exists == "open":
+                existing = cls.open(name)
+                for attribute in {"label_names", "embedding_model"}:
+                    if locals()[attribute] is not None and locals()[attribute] != getattr(existing, attribute):
+                        raise ValueError(f"Memoryset with name {name} already exists with a different {attribute}.")
+                return existing
+
+        response = create_memoryset(
+            body=CreateLabeledMemorysetRequest(
+                name=name,
+                datasource_id=datasource.id,
+                datasource_label_column=label_column,
+                datasource_value_column=value_column,
+                datasource_source_id_column=source_id_column,
+                pretrained_embedding_model_name=(
+                    embedding_model._model_name if isinstance(embedding_model, PretrainedEmbeddingModel) else None
+                ),
+                finetuned_embedding_model_id=(
+                    embedding_model.id if isinstance(embedding_model, FinetunedEmbeddingModel) else None
+                ),
+                label_names=label_names or [],
+                max_seq_length_override=max_seq_length_override,
+            ),
+        )
+        wait_for_task(response.insertion_task_id, description="Inserting datasource")
+        return cls.open(response.id)
+
+    @classmethod
+    def from_hf_dataset(cls, name: str, hf_dataset: Dataset, **kwargs: Any) -> LabeledMemoryset:
+        """
+        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.LabeledMemoryset.create] attributes for details.
+
+
+        Returns:
+            Handle to the new memoryset in the OrcaCloud
+        """
+        datasource = Datasource.from_hf_dataset(
+            f"{name}_datasource", hf_dataset, if_exists=kwargs.get("if_exists", "error")
+        )
+        return cls.create(name, datasource, **kwargs)
+
+    @classmethod
+    def from_pytorch(
+        cls,
+        name: str,
+        torch_data: TorchDataLoader | TorchDataset,
+        *,
+        column_names: list[str] | None = None,
+        **kwargs: Any,
+    ) -> LabeledMemoryset:
+        """
+        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` 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.
+            kwargs: Additional parameters for creating the memoryset. See
+                [`create`][orca_sdk.LabeledMemoryset.create] attributes for details.
+
+
+        Returns:
+            Handle to the new memoryset in the OrcaCloud
+        """
+        datasource = Datasource.from_pytorch(
+            f"{name}_datasource", torch_data, column_names=column_names, if_exists=kwargs.get("if_exists", "error")
+        )
+        return cls.create(name, datasource, **kwargs)
+
+    @classmethod
+    def from_list(cls, name: str, data: list[dict], **kwargs: Any) -> LabeledMemoryset:
+        """
+        Create a new memoryset from a list of dictionaries 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` will be stored as metadata in the memoryset.
+
+        Params:
+            name: Name for the new memoryset (must be unique)
+            data: List of dictionaries to create the memoryset from
+            kwargs: Additional parameters for creating the memoryset. See
+                [`create`][orca_sdk.LabeledMemoryset.create] attributes for details.
+
+        Returns:
+            Handle to the new memoryset in the OrcaCloud
+
+        Examples:
+            >>> LabeledMemoryset.from_list("my_memoryset", [
+            ...     {"value": "hello", "label": 0, "tag": "tag1"},
+            ...     {"value": "world", "label": 1, "tag": "tag2"},
+            ... ])
+        """
+        datasource = Datasource.from_list(f"{name}_datasource", data, if_exists=kwargs.get("if_exists", "error"))
+        return cls.create(name, datasource, **kwargs)
+
+    @classmethod
+    def from_dict(cls, name: str, data: dict, **kwargs: Any) -> LabeledMemoryset:
+        """
+        Create a new memoryset from a dictionary of columns 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 columns from the datasource that are not specified in the `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)
+            data: Dictionary of columns to create the memoryset from
+            kwargs: Additional parameters for creating the memoryset. See
+                [`create`][orca_sdk.LabeledMemoryset.create] attributes for details.
+
+        Returns:
+            Handle to the new memoryset in the OrcaCloud
+
+        Examples:
+            >>> LabeledMemoryset.from_dict("my_memoryset", {
+            ...     "value": ["hello", "world"],
+            ...     "label": [0, 1],
+            ...     "tag": ["tag1", "tag2"],
+            ... })
+        """
+        datasource = Datasource.from_dict(f"{name}_datasource", data, if_exists=kwargs.get("if_exists", "error"))
+        return cls.create(name, datasource, **kwargs)
+
+    @classmethod
+    def from_pandas(cls, name: str, dataframe: pd.DataFrame, **kwargs: Any) -> LabeledMemoryset:
+        """
+        Create a new memoryset from a pandas [`DataFrame`][pandas.DataFrame] 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 columns 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)
+            dataframe: Dataframe to create the memoryset from
+            kwargs: Additional parameters for creating the memoryset. See
+                [`create`][orca_sdk.LabeledMemoryset.create] attributes for details.
+
+        Returns:
+            Handle to the new memoryset in the OrcaCloud
+        """
+        datasource = Datasource.from_pandas(f"{name}_datasource", dataframe, if_exists=kwargs.get("if_exists", "error"))
+        return cls.create(name, datasource, **kwargs)
+
+    @classmethod
+    def from_arrow(cls, name: str, pyarrow_table: pa.Table, **kwargs: Any) -> LabeledMemoryset:
+        """
+        Create a new memoryset from a PyArrow [`Table`][pyarrow.Table] 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 columns 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)
+            pyarrow_table: PyArrow table to create the memoryset from
+            kwargs: Additional parameters for creating the memoryset. See
+                [`create`][orca_sdk.LabeledMemoryset.create] attributes for details.
+
+        Returns:
+            Handle to the new memoryset in the OrcaCloud
+        """
+        datasource = Datasource.from_arrow(
+            f"{name}_datasource", pyarrow_table, if_exists=kwargs.get("if_exists", "error")
+        )
+        return cls.create(name, datasource, **kwargs)
+
+    @classmethod
+    def from_disk(cls, name: str, file_path: str | PathLike, **kwargs: Any) -> LabeledMemoryset:
+        """
+        Create a new memoryset from a file on disk 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 columns from the datasource that are not specified in the `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)
+            file_path: Path to the file on disk to create the memoryset from. The file type will
+                be inferred from the file extension. The following file types are supported:
+
+                - .pkl: [`Pickle`][pickle] files containing lists of dictionaries or dictionaries of columns
+                - .json/.jsonl: [`JSON`][json] and [`JSON`] Lines files
+                - .csv: [`CSV`][csv] files
+                - .parquet: [`Parquet`][pyarrow.parquet.ParquetFile] files
+                - dataset directory: Directory containing a saved HuggingFace [`Dataset`][datasets.Dataset]
+            kwargs: Additional parameters for creating the memoryset. See
+                [`create`][orca_sdk.LabeledMemoryset.create] attributes for details.
+
+        Returns:
+            Handle to the new memoryset in the OrcaCloud
+        """
+        datasource = Datasource.from_disk(f"{name}_datasource", file_path, if_exists=kwargs.get("if_exists", "error"))
+        return cls.create(name, datasource, **kwargs)
+
+    @classmethod
+    def open(cls, name: str) -> LabeledMemoryset:
+        """
+        Get a handle to a memoryset in the OrcaCloud
+
+        Params:
+            name: Name or unique identifier of the memoryset
+
+        Returns:
+            Handle to the existing memoryset in the OrcaCloud
+
+        Raises:
+            LookupError: If the memoryset does not exist
+        """
+        metadata = get_memoryset(name)
+        return cls(metadata)
+
+    @classmethod
+    def exists(cls, name_or_id: str) -> bool:
+        """
+        Check if a memoryset exists in the OrcaCloud
+
+        Params:
+            name_or_id: Name or id of the memoryset
+
+        Returns:
+            True if the memoryset exists, False otherwise
+        """
+        try:
+            cls.open(name_or_id)
+            return True
+        except LookupError:
+            return False
+
+    @classmethod
+    def all(cls) -> list[LabeledMemoryset]:
+        """
+        Get a list of handles to all memorysets in the OrcaCloud
+
+        Returns:
+            List of handles to all memorysets in the OrcaCloud
+        """
+        return [cls(metadata) for metadata in list_memorysets()]
+
+    @classmethod
+    def drop(cls, name_or_id: str, if_not_exists: DropMode = "error"):
+        """
+        Delete a memoryset from the OrcaCloud
+
+        Params:
+            name_or_id: Name or id of the memoryset
+            if_not_exists: What to do if the memoryset does not exist, defaults to `"error"`.
+                Other options are `"ignore"` to do nothing if the memoryset does not exist.
+
+        Raises:
+            LookupError: If the memoryset does not exist and if_not_exists is `"error"`
+        """
+        try:
+            delete_memoryset(name_or_id)
+            logging.info(f"Deleted memoryset {name_or_id}")
+        except LookupError:
+            if if_not_exists == "error":
+                raise
+
+    def clone(
+        self,
+        name: str,
+        *,
+        embedding_model: PretrainedEmbeddingModel | FinetunedEmbeddingModel | None = None,
+        max_seq_length_override: int | None = None,
+        if_exists: CreateMode = "error",
+    ) -> LabeledMemoryset:
+        """
+        Create a clone of the memoryset with a new name
+
+        Params:
+            name: Name for the new memoryset (must be unique)
+            embedding_model: Optional new embedding model to use for re-embedding the memory values
+            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
+            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.
+
+        Returns:
+            Handle to the cloned memoryset in the OrcaCloud
+
+        Examples:
+            >>> memoryset = LabeledMemoryset.open("my_memoryset")
+            >>> finetuned_embedding_model = PretrainedEmbeddingModel.GTE_BASE.finetune(
+            ...     "gte_base_finetuned", my_memoryset
+            ... )
+            >>> new_memoryset = memoryset.clone(
+            ...     "my_memoryset_finetuned", embedding_model=finetuned_embedding_model,
+            ... )
+        """
+        if self.exists(name):
+            if if_exists == "error":
+                raise ValueError(f"Memoryset with name {name} already exists")
+            elif if_exists == "open":
+                existing = self.open(name)
+                for attribute in {"embedding_model"}:
+                    if locals()[attribute] is not None and locals()[attribute] != getattr(existing, attribute):
+                        raise ValueError(f"Memoryset with name {name} already exists with a different {attribute}.")
+                return existing
+
+        metadata = clone_memoryset(
+            self.id,
+            body=CloneLabeledMemorysetRequest(
+                name=name,
+                pretrained_embedding_model_name=(
+                    embedding_model._model_name if isinstance(embedding_model, PretrainedEmbeddingModel) else None
+                ),
+                finetuned_embedding_model_id=(
+                    embedding_model.id if isinstance(embedding_model, FinetunedEmbeddingModel) else None
+                ),
+                max_seq_length_override=max_seq_length_override,
+            ),
+        )
+        wait_for_task(metadata.insertion_task_id, description="Cloning memoryset")
+        return LabeledMemoryset.open(metadata.id)
+
+    def refresh(self, throttle: float = 0):
+        """
+        Refresh the information about the memoryset from the OrcaCloud
+
+        Params:
+            throttle: Minimum time in seconds between refreshes
+        """
+        current_time = datetime.now()
+        # Skip refresh if last refresh was too recent
+        if (current_time - self._last_refresh) < timedelta(seconds=throttle):
+            return
+
+        self.__dict__.update(LabeledMemoryset.open(self.id).__dict__)
+        self._last_refresh = current_time
+
+    def __len__(self) -> int:
+        """Get the number of memories in the memoryset"""
+        self.refresh(throttle=5)
+        return self.length
+
+    @overload
+    def __getitem__(self, index: int | str) -> LabeledMemory:
+        pass
+
+    @overload
+    def __getitem__(self, index: slice) -> list[LabeledMemory]:
+        pass
+
+    def __getitem__(self, index: int | slice | str) -> LabeledMemory | list[LabeledMemory]:
+        """
+        Get memories from the memoryset by index or memory id
+
+        Params:
+            index: Index or memory to retrieve or slice of memories to retrieve or unique
+                identifier of the memory to retrieve
+
+        Returns:
+            Memory or memories from the memoryset
+
+        Raises:
+            LookupError: If the id is not found or the index is out of bounds
+
+        Examples:
+            Retrieve the first memory in the memoryset:
+            >>> memoryset[0]
+            LabeledMemory({ label: <positive: 1>, value: 'I am happy' })
+
+            Retrieve the last memory in the memoryset:
+            >>> memoryset[-1]
+            LabeledMemory({ label: <negative: 0>, value: 'I am sad' })
+
+            Retrieve a slice of memories in the memoryset:
+            >>> memoryset[1:3]
+            [
+                LabeledMemory({ label: <positive: 1>, value: 'I am happy' }),
+                LabeledMemory({ label: <negative: 0>, value: 'I am sad' }),
+            ]
+
+            Retrieve a memory by id:
+            >>> memoryset["0195019a-5bc7-7afb-b902-5945ee1fb766"]
+            LabeledMemory({ label: <positive: 1>, value: 'I am happy' })
+        """
+        if isinstance(index, int):
+            return self.query(offset=len(self) + index if index < 0 else index, limit=1)[0]
+        elif isinstance(index, str):
+            return self.get(index)
+        elif isinstance(index, slice):
+            start = 0 if index.start is None else (len(self) + index.start) if index.start < 0 else index.start
+            stop = len(self) if index.stop is None else (len(self) + index.stop) if index.stop < 0 else index.stop
+            return self.query(offset=start, limit=stop - start)
+        else:
+            raise ValueError(f"Invalid index type: {type(index)}")
+
+    @overload
+    def search(self, query: str, *, count: int = 1) -> list[LabeledMemoryLookup]:
+        pass
+
+    @overload
+    def search(self, query: list[str], *, count: int = 1) -> list[list[LabeledMemoryLookup]]:
+        pass
+
+    def search(
+        self, query: str | list[str], *, count: int = 1
+    ) -> list[LabeledMemoryLookup] | list[list[LabeledMemoryLookup]]:
+        """
+        Search for memories that are semantically similar to the query
+
+        Params:
+            query: Query to lookup memories in the memoryset, can be a single query or a list
+            count: Number of memories to return for each query
+
+        Returns:
+            List of memories from the memoryset that match the query. If a single query is provided,
+                the return value is a list containing a single list of memories. If a list of
+                queries is provided, the return value is a list of lists of memories.
+
+        Examples:
+            Search for similar memories:
+            >>> memoryset.search("I am happy", count=2)
+            [
+                LabeledMemoryLookup({ label: <positive: 1>, value: 'I am happy' }),
+                LabeledMemoryLookup({ label: <positive: 1>, value: 'I am content' }),
+            ]
+
+            Search for similar memories for multiple queries:
+            >>> memoryset.search(["I am happy", "I am sad"], count=1)
+            [
+                [
+                    LabeledMemoryLookup({ label: <positive: 1>, value: 'I am happy' }),
+                ],
+                [
+                    LabeledMemoryLookup({ label: <negative: 0>, value: 'I am sad' }),
+                ],
+            ]
+        """
+        response = memoryset_lookup_gpu(
+            name_or_id=self.id,
+            body=LookupRequest(
+                query=query if isinstance(query, list) else [query],
+                count=count,
+            ),
+        )
+        lookups = [[LabeledMemoryLookup(self.id, lookup_response) for lookup_response in batch] for batch in response]
+        return lookups if isinstance(query, list) else lookups[0]
+
+    def query(
+        self,
+        offset: int = 0,
+        limit: int = 100,
+        filters: list[FilterItemTuple] = [],
+    ) -> list[LabeledMemory]:
+        """
+        Query the memoryset for memories that match the filters
+
+        Params:
+            offset: The offset of the first memory to return
+            limit: The maximum number of memories to return
+            filters: List of filters to apply to the query.
+
+        Returns:
+            List of memories from the memoryset that match the filters
+
+        Examples:
+            >>> memoryset.query(filters=[("label", "==", 0)], limit=2)
+            [
+                LabeledMemory({ label: <positive: 1>, value: "I am happy" }),
+                LabeledMemory({ label: <negative: 0>, value: "I am sad" }),
+            ]
+        """
+        return [
+            LabeledMemory(self.id, memory)
+            for memory in query_memoryset(
+                self.id,
+                body=ListMemoriesRequest(
+                    offset=offset,
+                    limit=limit,
+                    filters=[
+                        _parse_filter_item_from_tuple(filter) if isinstance(filter, tuple) else filter
+                        for filter in filters
+                    ],
+                ),
+            )
+        ]
+
+    def insert(self, items: Iterable[dict[str, Any]] | dict[str, Any]) -> None:
+        """
+        Insert memories into the memoryset
+
+        Params:
+            items: List of memories to insert into the memoryset. This should be a list of
+                dictionaries with the following keys:
+
+                - `value`: Value of the memory
+                - `label`: Label of the memory
+                - `source_id`: Optional unique ID of the memory in a system of reference
+                - `...`: Any other metadata to store for the memory
+
+        Examples:
+            >>> memoryset.insert([
+            ...     {"value": "I am happy", "label": 1, "source_id": "user_123", "tag": "happy"},
+            ...     {"value": "I am sad", "label": 0, "source_id": "user_124", "tag": "sad"},
+            ... ])
+        """
+        insert_memories_gpu(
+            self.id,
+            body=(
+                [
+                    _parse_memory_insert(memory)
+                    for memory in (cast(list[dict[str, Any]], [items]) if isinstance(items, dict) else items)
+                ]
+            ),
+        )
+        self.refresh()
+
+    @overload
+    def get(self, memory_id: str) -> LabeledMemory:  # type: ignore -- this takes precedence
+        pass
+
+    @overload
+    def get(self, memory_id: Iterable[str]) -> list[LabeledMemory]:
+        pass
+
+    def get(self, memory_id: str | Iterable[str]) -> LabeledMemory | list[LabeledMemory]:
+        """
+        Fetch a memory or memories from the memoryset
+
+        Params:
+            memory_id: Unique identifier of the memory or memories to fetch
+
+        Returns:
+            Memory or list of memories from the memoryset
+
+        Raises:
+            LookupError: If no memory with the given id is found
+
+        Examples:
+            Fetch a single memory:
+            >>> memoryset.get("0195019a-5bc7-7afb-b902-5945ee1fb766")
+            LabeledMemory({ label: <positive: 1>, value: 'I am happy' })
+
+            Fetch multiple memories:
+            >>> memoryset.get([
+            ...     "0195019a-5bc7-7afb-b902-5945ee1fb766",
+            ...     "019501a1-ea08-76b2-9f62-95e4800b4841",
+            ... ])
+            [
+                LabeledMemory({ label: <positive: 1>, value: 'I am happy' }),
+                LabeledMemory({ label: <negative: 0>, value: 'I am sad' }),
+            ]
+        """
+        if isinstance(memory_id, str):
+            return LabeledMemory(self.id, get_memory(self.id, memory_id))
+        else:
+            return [
+                LabeledMemory(self.id, memory)
+                for memory in get_memories(self.id, body=GetMemoriesRequest(memory_ids=list(memory_id)))
+            ]
+
+    @overload
+    def update(self, updates: dict[str, Any]) -> LabeledMemory:
+        pass
+
+    @overload
+    def update(self, updates: Iterable[dict[str, Any]]) -> list[LabeledMemory]:
+        pass
+
+    def update(self, updates: dict[str, Any] | Iterable[dict[str, Any]]) -> LabeledMemory | list[LabeledMemory]:
+        """
+        Update one or multiple memories in the memoryset
+
+        Params:
+            updates: List of updates to apply to the memories. Each update should be a dictionary
+                with the following keys:
+
+                - `memory_id`: Unique identifier of the memory to update (required)
+                - `value`: Optional new value of the memory
+                - `label`: Optional new label of the memory
+                - `source_id`: Optional new source ID of the memory
+                - `...`: Optional new values for metadata properties
+
+        Returns:
+            Updated memory or list of updated memories
+
+        Examples:
+            Update a single memory:
+            >>> memoryset.update(
+            ...     {
+            ...         "memory_id": "019501a1-ea08-76b2-9f62-95e4800b4841",
+            ...         "tag": "happy",
+            ...     },
+            ... )
+
+            Update multiple memories:
+            >>> memoryset.update(
+            ...     {"memory_id": m.memory_id, "label": 2}
+            ...     for m in memoryset.query(filters=[("tag", "==", "happy")])
+            ... )
+        """
+        response = update_memories_gpu(
+            self.id,
+            body=[
+                _parse_memory_update(update)
+                for update in (cast(list[dict[str, Any]], [updates]) if isinstance(updates, dict) else updates)
+            ],
+        )
+        updated_memories = [LabeledMemory(self.id, memory) for memory in response]
+        return updated_memories[0] if isinstance(updates, dict) else updated_memories
+
+    def delete(self, memory_id: str | Iterable[str]) -> None:
+        """
+        Delete memories from the memoryset
+
+        Params:
+            memory_id: unique identifiers of the memories to delete
+
+        Examples:
+            Delete a single memory:
+            >>> memoryset.delete("0195019a-5bc7-7afb-b902-5945ee1fb766")
+
+            Delete multiple memories:
+            >>> memoryset.delete([
+            ...     "0195019a-5bc7-7afb-b902-5945ee1fb766",
+            ...     "019501a1-ea08-76b2-9f62-95e4800b4841",
+            ... )
+
+        """
+        memory_ids = [memory_id] if isinstance(memory_id, str) else list(memory_id)
+        delete_memories(self.id, body=DeleteMemoriesRequest(memory_ids=memory_ids))
+        logging.info(f"Deleted {len(memory_ids)} memories from memoryset.")
+        self.refresh()
+
+    def find_duplicates(self) -> dict:
+        """
+        Run an analysis to find duplicate memories in the memoryset
+
+        The results of the analysis will be stored in the [`LabeledMemory.metrics`][orca_sdk.LabeledMemory]
+        attribute of each memory in the memoryset.
+
+        Returns:
+            Summary of analysis with number of duplicate memories found
+
+        Examples:
+            >>> memoryset.find_duplicate_memories()
+            { "num_duplicates": 10 }
+            >>> memoryset.delete(
+            ...     m.memory_id
+            ...     for m in memoryset.query(
+            ...         filters=[("metrics.is_duplicate", "==", True)]
+            ...     )
+            ... )
+        """
+        analysis = create_analysis(
+            self.id,
+            body=MemorysetAnalysisRequest(
+                type=MemorysetAnalysisRequestType.ANALYZE_DUPLICATE_MEMORIES,
+            ),
+        )
+        wait_for_task(analysis.task_id, description="Analyzing duplicates")
+        analysis = get_analysis(self.id, analysis.task_id)
+        assert isinstance(analysis.result, FindDuplicatesAnalysisResult)
+        # TODO: return a custom duplicate analysis class instance with helper methods
+        return analysis.result.to_dict()
+
+    def analyze_labels(self, neighbor_count: int = 10) -> dict:
+        """
+        Run an analysis to access if the labels in the memoryset are consistent to detect possibly
+        mislabeled memories.
+
+        The results of the analysis will be stored in the [`LabeledMemory.metrics`][orca_sdk.LabeledMemory]
+        attribute of each memory in the memoryset.
+
+        Returns:
+            Summary of analysis with aggregate metrics for each label class
+
+        Examples:
+            >>> memoryset.analyze_labels()
+            {
+                "label_metrics": [{
+                    "label": 0,
+                    "label_name": "negative",
+                    "average_lookup_score": 0.95,
+                    "memory_count": 100,
+                }, {
+                    "label": 1,
+                    "label_name": "positive",
+                    "average_lookup_score": 0.90,
+                    "memory_count": 100,
+                }]
+            }
+            >>> memoryset.display_label_analysis()
+        """
+        analysis = create_analysis(
+            self.id,
+            body=MemorysetAnalysisRequest(
+                type=MemorysetAnalysisRequestType.ANALYZE_MEMORY_NEIGHBOR_LABELS,
+                neighbor_count=neighbor_count,
+            ),
+        )
+        wait_for_task(analysis.task_id, description="Analyzing labels")
+        analysis = get_analysis(self.id, analysis.task_id)
+        assert isinstance(analysis.result, AnalyzeNeighborLabelsResult)
+        # TODO: return a custom label analysis class instance with helper methods
+        return analysis.result.to_dict()
+
+    def display_label_analysis(self):
+        """Display a UI to review and act upon the label analysis results"""
+        from ._utils.analysis_ui import display_suggested_memory_relabels
+
+        display_suggested_memory_relabels(self)
+
+    @staticmethod
+    def run_embedding_evaluation(
+        datasource: Datasource,
+        value_column: str = "value",
+        label_column: str = "label",
+        source_id_column: str | None = None,
+        neighbor_count: int = 5,
+        embedding_models: list[str] | None = None,
+    ) -> dict:
+        """
+        This function runs an embedding evaluation on the datasource. The embedding evaluation will
+        test the quality of embeddings for the datasource by computing metrics such as prediction accuracy.
+
+        Params:
+            datasource: The datasource to run the embedding evaluation on
+            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,
+                these must be contiguous integers starting from 0
+            source_id_column: Optional name of the column in the datasource that contains the ids in
+                the system of reference
+            neighbor_count: The number of neighbors to select for prediction
+            embedding_models: Optional list of embedding model keys to evaluate, if not provided all
+                available embedding models will be used
+
+        Returns:
+            A dictionary containing the results of the embedding evaluation
+        """
+
+        if embedding_models is not None:
+            embedding_model_enums = [PretrainedEmbeddingModelName(model) for model in embedding_models]
+        else:
+            embedding_model_enums = None
+
+        request = EmbeddingEvaluationRequest(
+            value_column=value_column,
+            label_column=label_column,
+            source_id_column=source_id_column,
+            neighbor_count=neighbor_count,
+            embedding_models=embedding_model_enums,
+        )
+
+        response = create_embedding_evaluation(name_or_id=datasource.id, body=request)
+        wait_for_task(response.task_id, description="Running embedding evaluation")
+
+        response = get_embedding_evaluation(datasource.id, response.task_id)
+        assert response.result is not None
+        return response.result.to_dict()