Dataset bugfixes, pytest integration bugfixes, evaluation improvements

.github/workflows/lib-langchain-tests.yml

@@ -33,7 +33,7 @@ jobs:
           python-version: ${{matrix.python_version}}
 
       - name: Install opik
-        run: pip install -e .
+        run: pip install .
 
       - name: Install test tools
         run: |

.github/workflows/lib-openai-tests.yml

@@ -33,7 +33,7 @@ jobs:
           python-version: ${{matrix.python_version}}
 
       - name: Install opik
-        run: pip install -e .
+        run: pip install .
 
       - name: Install test tools
         run: |

.github/workflows/python_sdk_unit_tests.yml

@@ -37,7 +37,7 @@ jobs:
           python-version: ${{ matrix.python_version }}
 
       - name: Install opik
-        run: pip install -e .
+        run: pip install .
 
       - name: Install test requirements
         run: |

sdks/python/src/opik/api_objects/dataset/converters.py

@@ -69,8 +69,3 @@ def from_json(
         result.append(dataset_item.DatasetItem(**item_kwargs))
 
     return result
-
-
-def from_dicts(dict_items: List[Dict[str, Any]]) -> List[dataset_item.DatasetItem]:
-    items = [dataset_item.DatasetItem(**dict_item) for dict_item in dict_items]
-    return items

sdks/python/src/opik/api_objects/dataset/dataset.py

@@ -1,6 +1,6 @@
 import logging
 import json
-from typing import Optional, Any, List, Dict, Union, cast
+from typing import Optional, Any, List, Dict, Union, Sequence
 
 from opik.rest_api import client as rest_api_client
 from opik.rest_api.types import dataset_item as rest_dataset_item
@@ -39,7 +39,7 @@ def description(self) -> Optional[str]:
         return self._description
 
     def insert(
-        self, items: Union[List[dataset_item.DatasetItem], List[Dict[str, Any]]]
+        self, items: Sequence[Union[dataset_item.DatasetItem, Dict[str, Any]]]
     ) -> None:
         """
         Insert new items into the dataset.
@@ -48,10 +48,10 @@ def insert(
             items: List of DatasetItem objects or dicts (which will be converted to DatasetItem objects)
                 to add to the dataset.
         """
-        if len(items) > 0 and isinstance(items[0], dict):
-            items = converters.from_dicts(items)  # type: ignore
-
-        items = cast(List[dataset_item.DatasetItem], items)
+        items: List[dataset_item.DatasetItem] = [  # type: ignore
+            (dataset_item.DatasetItem(**item) if isinstance(item, dict) else item)
+            for item in items
+        ]
 
         rest_items = [
             rest_dataset_item.DatasetItem(

sdks/python/src/opik/api_objects/dataset/dataset_item.py

@@ -1,12 +1,13 @@
 from typing import Optional, Dict, Any
-import dataclasses
+import pydantic
 from .. import constants
 
 
-@dataclasses.dataclass
-class DatasetItem:
+class DatasetItem(pydantic.BaseModel):
     """A DatasetItem object representing an item in a dataset."""
 
+    model_config = pydantic.ConfigDict(strict=True)
+
     input: Dict[str, Any]
     """The input data for the dataset item."""
 

sdks/python/src/opik/config.py

@@ -38,7 +38,7 @@ def settings_customise_sources(
 
     # Below are Opik configurations
 
-    url_override: str = "https://comet.com/opik/api"
+    url_override: str = "https://www.comet.com/opik/api"
     """Opik backend base URL"""
 
     project_name: str = "Default Project"

sdks/python/src/opik/evaluation/evaluator.py

@@ -7,7 +7,7 @@
 from ..api_objects.experiment import experiment_item
 from ..api_objects import opik_client
 
-from . import task_runner, test_result, scoring_runner, scores_logger, report
+from . import tasks_scorer, test_result, scores_logger, report
 
 
 def evaluate(
@@ -47,18 +47,12 @@ def evaluate(
     client = opik_client.get_client_cached()
     start_time = time.time()
 
-    test_cases = task_runner.run(
+    test_results = tasks_scorer.run(
         client=client,
         dataset_=dataset,
         task=task,
-        workers=task_threads,
-        verbose=verbose,
-    )
-
-    test_results = scoring_runner.run(
-        test_cases=test_cases,
         scoring_metrics=scoring_metrics,
-        workers=scoring_threads,
+        workers=task_threads,
         verbose=verbose,
     )
 

sdks/python/src/opik/evaluation/metrics/llm_judges/hallucination/metric.py

@@ -114,7 +114,9 @@ def _parse_model_output(self, content: str) -> score_result.ScoreResult:
             verdict: str = dict_content[template.VERDICT_KEY]
             score = 1.0 if verdict.lower() == template.HALLUCINATION_VERDICT else 0.0
             return score_result.ScoreResult(
-                name=self.name, value=score, reason=dict_content[template.REASON_KEY]
+                name=self.name,
+                value=score,
+                reason=str(dict_content[template.REASON_KEY]),
             )
         except Exception:
             raise exceptions.MetricComputationError(