orca-sdk 0.0.91__py3-none-any.whl → 0.0.93__py3-none-any.whl

orca_sdk/memoryset_test.py

@@ -1,23 +1,45 @@
 import random
+import time
+from typing import Generator
 from uuid import uuid4
 
 import pytest
+from datasets import ClassLabel, Features, Value
 from datasets.arrow_dataset import Dataset
 
+from orca_sdk.conftest import SAMPLE_DATA
+
+from ._generated_api_client.models import CascadingEditSuggestion
 from .classification_model import ClassificationModel
 from .datasource import Datasource
 from .embedding_model import PretrainedEmbeddingModel
 from .memoryset import LabeledMemoryset, TaskStatus
 
+"""
+Test Performance Note:
+
+Creating new `LabeledMemoryset` objects is expensive, so this test file applies the following optimizations:
+
+- Two fixtures are used to manage memorysets:
+    - `readonly_memoryset` is a session-scoped fixture shared across tests that do not modify state.
+      It should only be used in nullipotent tests.
+    - `writable_memoryset` is a function-scoped, regenerating fixture.
+      It can be used in tests that mutate or delete the memoryset, and will be reset before each test.
+
+- To minimize fixture overhead, tests using `writable_memoryset` should combine related behaviors.
+  For example, prefer a single `test_delete` that covers both single and multiple deletion cases,
+  rather than separate `test_delete_single` and `test_delete_multiple` tests.
+"""
 
-def test_create_memoryset(memoryset: LabeledMemoryset, hf_dataset: Dataset, label_names: list[str]):
-    assert memoryset is not None
-    assert memoryset.name == "test_memoryset"
-    assert memoryset.embedding_model == PretrainedEmbeddingModel.GTE_BASE
-    assert memoryset.label_names == label_names
-    assert memoryset.insertion_status == TaskStatus.COMPLETED
-    assert isinstance(memoryset.length, int)
-    assert memoryset.length == len(hf_dataset)
+
+def test_create_memoryset(readonly_memoryset: LabeledMemoryset, hf_dataset: Dataset, label_names: list[str]):
+    assert readonly_memoryset is not None
+    assert readonly_memoryset.name == "test_readonly_memoryset"
+    assert readonly_memoryset.embedding_model == PretrainedEmbeddingModel.GTE_BASE
+    assert readonly_memoryset.label_names == label_names
+    assert readonly_memoryset.insertion_status == TaskStatus.COMPLETED
+    assert isinstance(readonly_memoryset.length, int)
+    assert readonly_memoryset.length == len(hf_dataset)
 
 
 def test_create_memoryset_unauthenticated(unauthenticated, datasource):
@@ -29,61 +51,55 @@ def test_create_memoryset_invalid_input(datasource):
     # invalid name
     with pytest.raises(ValueError, match=r"Invalid input:.*"):
         LabeledMemoryset.create("test memoryset", datasource)
-    # invalid datasource
-    datasource.id = str(uuid4())
-    with pytest.raises(ValueError, match=r"Invalid input:.*"):
-        LabeledMemoryset.create("test_memoryset_invalid_datasource", datasource)
 
 
-def test_create_memoryset_already_exists_error(hf_dataset, label_names, memoryset):
+def test_create_memoryset_already_exists_error(hf_dataset, label_names, readonly_memoryset):
+    memoryset_name = readonly_memoryset.name
     with pytest.raises(ValueError):
-        LabeledMemoryset.from_hf_dataset("test_memoryset", hf_dataset, label_names=label_names, value_column="text")
+        LabeledMemoryset.from_hf_dataset(memoryset_name, hf_dataset, label_names=label_names)
     with pytest.raises(ValueError):
-        LabeledMemoryset.from_hf_dataset(
-            "test_memoryset", hf_dataset, label_names=label_names, value_column="text", if_exists="error"
-        )
+        LabeledMemoryset.from_hf_dataset(memoryset_name, hf_dataset, label_names=label_names, if_exists="error")
 
 
-def test_create_memoryset_already_exists_open(hf_dataset, label_names, memoryset):
+def test_create_memoryset_already_exists_open(hf_dataset, label_names, readonly_memoryset):
     # invalid label names
     with pytest.raises(ValueError):
         LabeledMemoryset.from_hf_dataset(
-            memoryset.name,
+            readonly_memoryset.name,
             hf_dataset,
             label_names=["turtles", "frogs"],
-            value_column="text",
             if_exists="open",
         )
     # different embedding model
     with pytest.raises(ValueError):
         LabeledMemoryset.from_hf_dataset(
-            memoryset.name,
+            readonly_memoryset.name,
             hf_dataset,
             label_names=label_names,
             embedding_model=PretrainedEmbeddingModel.DISTILBERT,
             if_exists="open",
         )
     opened_memoryset = LabeledMemoryset.from_hf_dataset(
-        memoryset.name,
+        readonly_memoryset.name,
         hf_dataset,
         embedding_model=PretrainedEmbeddingModel.GTE_BASE,
         if_exists="open",
     )
     assert opened_memoryset is not None
-    assert opened_memoryset.name == memoryset.name
+    assert opened_memoryset.name == readonly_memoryset.name
     assert opened_memoryset.length == len(hf_dataset)
 
 
-def test_open_memoryset(memoryset, hf_dataset):
-    fetched_memoryset = LabeledMemoryset.open(memoryset.name)
+def test_open_memoryset(readonly_memoryset, hf_dataset):
+    fetched_memoryset = LabeledMemoryset.open(readonly_memoryset.name)
     assert fetched_memoryset is not None
-    assert fetched_memoryset.name == memoryset.name
+    assert fetched_memoryset.name == readonly_memoryset.name
     assert fetched_memoryset.length == len(hf_dataset)
 
 
-def test_open_memoryset_unauthenticated(unauthenticated, memoryset):
+def test_open_memoryset_unauthenticated(unauthenticated, readonly_memoryset):
     with pytest.raises(ValueError, match="Invalid API key"):
-        LabeledMemoryset.open(memoryset.name)
+        LabeledMemoryset.open(readonly_memoryset.name)
 
 
 def test_open_memoryset_not_found():
@@ -96,15 +112,15 @@ def test_open_memoryset_invalid_input():
         LabeledMemoryset.open("not valid id")
 
 
-def test_open_memoryset_unauthorized(unauthorized, memoryset):
+def test_open_memoryset_unauthorized(unauthorized, readonly_memoryset):
     with pytest.raises(LookupError):
-        LabeledMemoryset.open(memoryset.name)
+        LabeledMemoryset.open(readonly_memoryset.name)
 
 
-def test_all_memorysets(memoryset):
+def test_all_memorysets(readonly_memoryset: LabeledMemoryset):
     memorysets = LabeledMemoryset.all()
     assert len(memorysets) > 0
-    assert any(memoryset.name == memoryset.name for memoryset in memorysets)
+    assert any(memoryset.name == readonly_memoryset.name for memoryset in memorysets)
 
 
 def test_all_memorysets_unauthenticated(unauthenticated):
@@ -112,51 +128,39 @@ def test_all_memorysets_unauthenticated(unauthenticated):
         LabeledMemoryset.all()
 
 
-def test_all_memorysets_unauthorized(unauthorized, memoryset):
-    assert memoryset not in LabeledMemoryset.all()
-
-
-def test_drop_memoryset(hf_dataset):
-    memoryset = LabeledMemoryset.from_hf_dataset(
-        "test_memoryset_delete",
-        hf_dataset.select(range(1)),
-        value_column="text",
-    )
-    assert LabeledMemoryset.exists(memoryset.name)
-    LabeledMemoryset.drop(memoryset.name)
-    assert not LabeledMemoryset.exists(memoryset.name)
+def test_all_memorysets_unauthorized(unauthorized, readonly_memoryset):
+    assert readonly_memoryset not in LabeledMemoryset.all()
 
 
-def test_drop_memoryset_unauthenticated(unauthenticated, memoryset):
+def test_drop_memoryset_unauthenticated(unauthenticated, readonly_memoryset):
     with pytest.raises(ValueError, match="Invalid API key"):
-        LabeledMemoryset.drop(memoryset.name)
+        LabeledMemoryset.drop(readonly_memoryset.name)
 
 
-def test_drop_memoryset_not_found(memoryset):
+def test_drop_memoryset_not_found():
     with pytest.raises(LookupError):
         LabeledMemoryset.drop(str(uuid4()))
     # ignores error if specified
     LabeledMemoryset.drop(str(uuid4()), if_not_exists="ignore")
 
 
-def test_drop_memoryset_unauthorized(unauthorized, memoryset):
+def test_drop_memoryset_unauthorized(unauthorized, readonly_memoryset):
     with pytest.raises(LookupError):
-        LabeledMemoryset.drop(memoryset.name)
+        LabeledMemoryset.drop(readonly_memoryset.name)
 
 
-def test_update_memoryset_metadata(memoryset: LabeledMemoryset):
-    memoryset.update_metadata(description="New description")
-    assert memoryset.description == "New description"
+def test_update_memoryset_metadata(writable_memoryset: LabeledMemoryset):
+    # NOTE: We're combining multiple tests into one here to avoid multiple API calls
 
+    writable_memoryset.update_metadata(description="New description")
+    assert writable_memoryset.description == "New description"
 
-def test_update_memoryset_no_description(memoryset: LabeledMemoryset):
-    assert memoryset.description is not None
-    memoryset.update_metadata(description=None)
-    assert memoryset.description is None
+    writable_memoryset.update_metadata(description=None)
+    assert writable_memoryset.description is None
 
 
-def test_search(memoryset: LabeledMemoryset):
-    memory_lookups = memoryset.search(["i love soup", "cats are cute"])
+def test_search(readonly_memoryset: LabeledMemoryset):
+    memory_lookups = readonly_memoryset.search(["i love soup", "cats are cute"])
     assert len(memory_lookups) == 2
     assert len(memory_lookups[0]) == 1
     assert len(memory_lookups[1]) == 1
@@ -164,53 +168,53 @@ def test_search(memoryset: LabeledMemoryset):
     assert memory_lookups[1][0].label == 1
 
 
-def test_search_count(memoryset: LabeledMemoryset):
-    memory_lookups = memoryset.search("i love soup", count=3)
+def test_search_count(readonly_memoryset: LabeledMemoryset):
+    memory_lookups = readonly_memoryset.search("i love soup", count=3)
     assert len(memory_lookups) == 3
     assert memory_lookups[0].label == 0
     assert memory_lookups[1].label == 0
     assert memory_lookups[2].label == 0
 
 
-def test_get_memory_at_index(memoryset: LabeledMemoryset, hf_dataset: Dataset, label_names: list[str]):
-    memory = memoryset[0]
-    assert memory.value == hf_dataset[0]["text"]
+def test_get_memory_at_index(readonly_memoryset: LabeledMemoryset, hf_dataset: Dataset, label_names: list[str]):
+    memory = readonly_memoryset[0]
+    assert memory.value == hf_dataset[0]["value"]
     assert memory.label == hf_dataset[0]["label"]
     assert memory.label_name == label_names[hf_dataset[0]["label"]]
     assert memory.source_id == hf_dataset[0]["source_id"]
     assert memory.score == hf_dataset[0]["score"]
     assert memory.key == hf_dataset[0]["key"]
-    last_memory = memoryset[-1]
-    assert last_memory.value == hf_dataset[-1]["text"]
+    last_memory = readonly_memoryset[-1]
+    assert last_memory.value == hf_dataset[-1]["value"]
     assert last_memory.label == hf_dataset[-1]["label"]
 
 
-def test_get_range_of_memories(memoryset: LabeledMemoryset, hf_dataset: Dataset):
-    memories = memoryset[1:3]
+def test_get_range_of_memories(readonly_memoryset: LabeledMemoryset, hf_dataset: Dataset):
+    memories = readonly_memoryset[1:3]
     assert len(memories) == 2
-    assert memories[0].value == hf_dataset["text"][1]
-    assert memories[1].value == hf_dataset["text"][2]
+    assert memories[0].value == hf_dataset["value"][1]
+    assert memories[1].value == hf_dataset["value"][2]
 
 
-def test_get_memory_by_id(memoryset: LabeledMemoryset, hf_dataset: Dataset):
-    memory = memoryset.get(memoryset[0].memory_id)
-    assert memory.value == hf_dataset[0]["text"]
-    assert memory == memoryset[memory.memory_id]
+def test_get_memory_by_id(readonly_memoryset: LabeledMemoryset, hf_dataset: Dataset):
+    memory = readonly_memoryset.get(readonly_memoryset[0].memory_id)
+    assert memory.value == hf_dataset[0]["value"]
+    assert memory == readonly_memoryset[memory.memory_id]
 
 
-def test_get_memories_by_id(memoryset: LabeledMemoryset, hf_dataset: Dataset):
-    memories = memoryset.get([memoryset[0].memory_id, memoryset[1].memory_id])
+def test_get_memories_by_id(readonly_memoryset: LabeledMemoryset, hf_dataset: Dataset):
+    memories = readonly_memoryset.get([readonly_memoryset[0].memory_id, readonly_memoryset[1].memory_id])
     assert len(memories) == 2
-    assert memories[0].value == hf_dataset[0]["text"]
-    assert memories[1].value == hf_dataset[1]["text"]
+    assert memories[0].value == hf_dataset[0]["value"]
+    assert memories[1].value == hf_dataset[1]["value"]
 
 
-def test_query_memoryset(memoryset: LabeledMemoryset):
-    memories = memoryset.query(filters=[("label", "==", 1)])
+def test_query_memoryset(readonly_memoryset: LabeledMemoryset):
+    memories = readonly_memoryset.query(filters=[("label", "==", 1)])
     assert len(memories) == 8
     assert all(memory.label == 1 for memory in memories)
-    assert len(memoryset.query(limit=2)) == 2
-    assert len(memoryset.query(filters=[("metadata.key", "==", "val1")])) == 1
+    assert len(readonly_memoryset.query(limit=2)) == 2
+    assert len(readonly_memoryset.query(filters=[("metadata.key", "==", "val1")])) == 1
 
 
 def test_query_memoryset_with_feedback_metrics(model: ClassificationModel):
@@ -268,19 +272,19 @@ def test_query_memoryset_with_feedback_metrics_sort(model: ClassificationModel):
     assert memories[-1].feedback_metrics["positive"]["avg"] == -1.0
 
 
-def test_insert_memories(memoryset: LabeledMemoryset):
-    memoryset.refresh()
-    prev_length = memoryset.length
-    memoryset.insert(
+def test_insert_memories(writable_memoryset: LabeledMemoryset):
+    writable_memoryset.refresh()
+    prev_length = writable_memoryset.length
+    writable_memoryset.insert(
         [
             dict(value="tomato soup is my favorite", label=0),
             dict(value="cats are fun to play with", label=1),
         ]
     )
-    assert memoryset.length == prev_length + 2
-    memoryset.insert(dict(value="tomato soup is my favorite", label=0, key="test", source_id="test"))
-    assert memoryset.length == prev_length + 3
-    last_memory = memoryset[-1]
+    assert writable_memoryset.length == prev_length + 2
+    writable_memoryset.insert(dict(value="tomato soup is my favorite", label=0, key="test", source_id="test"))
+    assert writable_memoryset.length == prev_length + 3
+    last_memory = writable_memoryset[-1]
     assert last_memory.value == "tomato soup is my favorite"
     assert last_memory.label == 0
     assert last_memory.metadata
@@ -288,25 +292,26 @@ def test_insert_memories(memoryset: LabeledMemoryset):
     assert last_memory.source_id == "test"
 
 
-def test_update_memory(memoryset: LabeledMemoryset, hf_dataset: Dataset):
-    memory_id = memoryset[0].memory_id
-    updated_memory = memoryset.update(dict(memory_id=memory_id, value="i love soup so much"))
+def test_update_memories(writable_memoryset: LabeledMemoryset, hf_dataset: Dataset):
+    # We've combined the update tests into one to avoid multiple expensive requests for a writable_memoryset
+
+    # test updating a single memory
+    memory_id = writable_memoryset[0].memory_id
+    updated_memory = writable_memoryset.update(dict(memory_id=memory_id, value="i love soup so much"))
     assert updated_memory.value == "i love soup so much"
     assert updated_memory.label == hf_dataset[0]["label"]
-    assert memoryset.get(memory_id).value == "i love soup so much"
-
+    assert writable_memoryset.get(memory_id).value == "i love soup so much"
 
-def test_update_memory_instance(memoryset: LabeledMemoryset, hf_dataset: Dataset):
-    memory = memoryset[0]
+    # test updating a memory instance
+    memory = writable_memoryset[0]
     updated_memory = memory.update(value="i love soup even more")
     assert updated_memory is memory
     assert memory.value == "i love soup even more"
     assert memory.label == hf_dataset[0]["label"]
 
-
-def test_update_memories(memoryset: LabeledMemoryset):
-    memory_ids = [memory.memory_id for memory in memoryset[:2]]
-    updated_memories = memoryset.update(
+    # test updating multiple memories
+    memory_ids = [memory.memory_id for memory in writable_memoryset[:2]]
+    updated_memories = writable_memoryset.update(
         [
             dict(memory_id=memory_ids[0], value="i love soup so much"),
             dict(memory_id=memory_ids[1], value="cats are so cute"),
@@ -316,35 +321,37 @@ def test_update_memories(memoryset: LabeledMemoryset):
     assert updated_memories[1].value == "cats are so cute"
 
 
-def test_delete_memory(memoryset: LabeledMemoryset):
-    memoryset.refresh()
-    prev_length = memoryset.length
-    memory_id = memoryset[0].memory_id
-    memoryset.delete(memory_id)
-    with pytest.raises(LookupError):
-        memoryset.get(memory_id)
-    assert memoryset.length == prev_length - 1
+def test_delete_memories(writable_memoryset: LabeledMemoryset):
+    # We've combined the delete tests into one to avoid multiple expensive requests for a writable_memoryset
 
+    # test deleting a single memory
+    prev_length = writable_memoryset.length
+    memory_id = writable_memoryset[0].memory_id
+    writable_memoryset.delete(memory_id)
+    with pytest.raises(LookupError):
+        writable_memoryset.get(memory_id)
+    assert writable_memoryset.length == prev_length - 1
 
-def test_delete_memories(memoryset: LabeledMemoryset):
-    prev_length = memoryset.length
-    memoryset.delete([memoryset[0].memory_id, memoryset[1].memory_id])
-    assert memoryset.length == prev_length - 2
+    # test deleting multiple memories
+    prev_length = writable_memoryset.length
+    writable_memoryset.delete([writable_memoryset[0].memory_id, writable_memoryset[1].memory_id])
+    assert writable_memoryset.length == prev_length - 2
 
 
-def test_clone_memoryset(memoryset: LabeledMemoryset):
-    cloned_memoryset = memoryset.clone("test_cloned_memoryset", embedding_model=PretrainedEmbeddingModel.DISTILBERT)
+def test_clone_memoryset(readonly_memoryset: LabeledMemoryset):
+    cloned_memoryset = readonly_memoryset.clone(
+        "test_cloned_memoryset", embedding_model=PretrainedEmbeddingModel.DISTILBERT
+    )
     assert cloned_memoryset is not None
     assert cloned_memoryset.name == "test_cloned_memoryset"
-    assert cloned_memoryset.length == memoryset.length
+    assert cloned_memoryset.length == readonly_memoryset.length
     assert cloned_memoryset.embedding_model == PretrainedEmbeddingModel.DISTILBERT
     assert cloned_memoryset.insertion_status == TaskStatus.COMPLETED
 
 
-def test_embedding_evaluation(hf_dataset):
-    datasource = Datasource.from_hf_dataset("eval_datasource", hf_dataset, if_exists="open")
+def test_embedding_evaluation(eval_datasource: Datasource):
     response = LabeledMemoryset.run_embedding_evaluation(
-        datasource, embedding_models=["CDE_SMALL"], neighbor_count=2, value_column="text"
+        eval_datasource, embedding_models=["CDE_SMALL"], neighbor_count=2
     )
     assert response is not None
     assert isinstance(response, dict)
@@ -358,8 +365,8 @@ def test_embedding_evaluation(hf_dataset):
 
 
 @pytest.fixture(scope="function")
-async def test_group_potential_duplicates(memoryset: LabeledMemoryset):
-    memoryset.insert(
+async def test_group_potential_duplicates(writable_memoryset: LabeledMemoryset):
+    writable_memoryset.insert(
         [
             dict(value="raspberry soup Is my favorite", label=0),
             dict(value="Raspberry soup is MY favorite", label=0),
@@ -375,7 +382,74 @@ async def test_group_potential_duplicates(memoryset: LabeledMemoryset):
         ]
     )
 
-    memoryset.analyze({"name": "duplicate", "possible_duplicate_threshold": 0.97})
-    response = memoryset.get_potential_duplicate_groups()
+    writable_memoryset.analyze({"name": "duplicate", "possible_duplicate_threshold": 0.97})
+    response = writable_memoryset.get_potential_duplicate_groups()
     assert isinstance(response, list)
     assert sorted([len(res) for res in response]) == [5, 6]  # 5 favorite, 6 mom
+
+
+def test_get_cascading_edits_suggestions(writable_memoryset: LabeledMemoryset):
+    # Insert a memory to test cascading edits
+    SOUP = 0
+    CATS = 1
+    query_text = "i love soup"  # from SAMPLE_DATA in conftest.py
+    mislabeled_soup_text = "soup is comfort in a bowl"
+    writable_memoryset.insert(
+        [
+            dict(value=mislabeled_soup_text, label=CATS),  # mislabeled soup memory
+        ]
+    )
+
+    # Fetch the memory to update
+    memory = writable_memoryset.query(filters=[("value", "==", query_text)])[0]
+
+    # Update the label and get cascading edit suggestions
+    suggestions = writable_memoryset.get_cascading_edits_suggestions(
+        memory=memory,
+        old_label=CATS,
+        new_label=SOUP,
+        max_neighbors=10,
+        max_validation_neighbors=5,
+    )
+
+    # Validate the suggestions
+    assert len(suggestions) == 1
+    assert suggestions[0].neighbor.value == mislabeled_soup_text
+
+
+def test_analyze_invalid_analysis_name(readonly_memoryset: LabeledMemoryset):
+    """Test that analyze() raises ValueError for invalid analysis names"""
+    memoryset = LabeledMemoryset.open(readonly_memoryset.name)
+
+    # Test with string input
+    with pytest.raises(ValueError) as excinfo:
+        memoryset.analyze("invalid_name")
+    assert "Invalid analysis name: invalid_name" in str(excinfo.value)
+    assert "Valid names are:" in str(excinfo.value)
+
+    # Test with dict input
+    with pytest.raises(ValueError) as excinfo:
+        memoryset.analyze({"name": "invalid_name"})
+    assert "Invalid analysis name: invalid_name" in str(excinfo.value)
+    assert "Valid names are:" in str(excinfo.value)
+
+    # Test with multiple analyses where one is invalid
+    with pytest.raises(ValueError) as excinfo:
+        memoryset.analyze("duplicate", "invalid_name")
+    assert "Invalid analysis name: invalid_name" in str(excinfo.value)
+    assert "Valid names are:" in str(excinfo.value)
+
+    # Test with valid analysis names
+    result = memoryset.analyze("duplicate", "cluster")
+    assert isinstance(result, dict)
+    assert "duplicate" in result
+    assert "cluster" in result
+
+
+def test_drop_memoryset(writable_memoryset: LabeledMemoryset):
+    # NOTE: Keep this test at the end to ensure the memoryset is dropped after all tests.
+    # Otherwise, it would be recreated on the next test run if it were dropped earlier, and
+    # that's expensive.
+    assert LabeledMemoryset.exists(writable_memoryset.name)
+    LabeledMemoryset.drop(writable_memoryset.name)
+    assert not LabeledMemoryset.exists(writable_memoryset.name)

orca_sdk/telemetry.py

@@ -135,6 +135,7 @@ class LabelPrediction:
     anomaly_score: float | None
     memoryset: LabeledMemoryset
     model: ClassificationModel
+    logits: list[float] | None
 
     def __init__(
         self,
@@ -147,6 +148,7 @@ class LabelPrediction:
         memoryset: LabeledMemoryset | str,
         model: ClassificationModel | str,
         telemetry: LabelPredictionWithMemoriesAndFeedback | None = None,
+        logits: list[float] | None = None,
     ):
         # for internal use only, do not document
         from .classification_model import ClassificationModel
@@ -159,6 +161,7 @@ class LabelPrediction:
         self.memoryset = LabeledMemoryset.open(memoryset) if isinstance(memoryset, str) else memoryset
         self.model = ClassificationModel.open(model) if isinstance(model, str) else model
         self.__telemetry = telemetry if telemetry else None
+        self.logits = logits
 
     def __repr__(self):
         return (

{orca_sdk-0.0.91.dist-info → orca_sdk-0.0.93.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: orca_sdk
-Version: 0.0.91
+Version: 0.0.93
 Summary: SDK for interacting with Orca Services
 License: Apache-2.0
 Author: Orca DB Inc.
@@ -20,7 +20,9 @@ Requires-Dist: pandas (>=2.2.3,<3.0.0)
 Requires-Dist: pyarrow (>=18.0.0,<19.0.0)
 Requires-Dist: python-dateutil (>=2.8.0,<3.0.0)
 Requires-Dist: python-dotenv (>=1.1.0,<2.0.0)
+Requires-Dist: scikit-learn (>=1.6.1,<2.0.0)
 Requires-Dist: torch (>=2.5.1,<3.0.0)
+Requires-Dist: transformers (>=4.51.3,<5.0.0)
 Description-Content-Type: text/markdown
 
 <!--