pydantic-evals 0.4.5__tar.gz → 0.4.6__tar.gz

{pydantic_evals-0.4.5 → pydantic_evals-0.4.6}/.gitignore

@@ -15,7 +15,6 @@ examples/pydantic_ai_examples/.chat_app_messages.sqlite
 .vscode/
 /question_graph_history.json
 /docs-site/.wrangler/
-/CLAUDE.md
 node_modules/
 **.idea/
 .coverage*

{pydantic_evals-0.4.5 → pydantic_evals-0.4.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pydantic-evals
-Version: 0.4.5
+Version: 0.4.6
 Summary: Framework for evaluating stochastic code execution, especially code making use of LLMs
 Project-URL: Homepage, https://ai.pydantic.dev/evals
 Project-URL: Source, https://github.com/pydantic/pydantic-ai
@@ -32,7 +32,7 @@ Requires-Python: >=3.9
 Requires-Dist: anyio>=0
 Requires-Dist: eval-type-backport>=0; python_version < '3.11'
 Requires-Dist: logfire-api>=1.2.0
-Requires-Dist: pydantic-ai-slim==0.4.5
+Requires-Dist: pydantic-ai-slim==0.4.6
 Requires-Dist: pydantic>=2.10
 Requires-Dist: pyyaml>=6.0.2
 Requires-Dist: rich>=13.9.4

{pydantic_evals-0.4.5 → pydantic_evals-0.4.6}/pydantic_evals/dataset.py

@@ -43,7 +43,7 @@ from .evaluators.common import DEFAULT_EVALUATORS
 from .evaluators.context import EvaluatorContext
 from .otel import SpanTree
 from .otel._context_subtree import context_subtree
-from .reporting import EvaluationReport, ReportCase
+from .reporting import EvaluationReport, ReportCase, ReportCaseAggregate
 
 if sys.version_info < (3, 11):
     from exceptiongroup import ExceptionGroup  # pragma: lax no cover
@@ -83,6 +83,10 @@ DEFAULT_SCHEMA_PATH_TEMPLATE = './{stem}_schema.json'
 _YAML_SCHEMA_LINE_PREFIX = '# yaml-language-server: $schema='
 
 
+_REPORT_CASES_ADAPTER = TypeAdapter(list[ReportCase])
+_REPORT_CASE_AGGREGATE_ADAPTER = TypeAdapter(ReportCaseAggregate)
+
+
 class _CaseModel(BaseModel, Generic[InputsT, OutputT, MetadataT], extra='forbid'):
     """Internal model for a case, used for serialization/deserialization."""
 
@@ -303,9 +307,9 @@ class Dataset(BaseModel, Generic[InputsT, OutputT, MetadataT], extra='forbid', a
                 ),
             )
             # TODO(DavidM): This attribute will be too big in general; remove it once we can use child spans in details panel:
-            eval_span.set_attribute('cases', report.cases)
+            eval_span.set_attribute('cases', _REPORT_CASES_ADAPTER.dump_python(report.cases))
             # TODO(DavidM): Remove this 'averages' attribute once we compute it in the details panel
-            eval_span.set_attribute('averages', report.averages())
+            eval_span.set_attribute('averages', _REPORT_CASE_AGGREGATE_ADAPTER.dump_python(report.averages()))
         return report
 
     def evaluate_sync(