pydantic-ai 0.0.52__tar.gz → 0.0.53__tar.gz

{pydantic_ai-0.0.52 → pydantic_ai-0.0.53}/.gitignore

@@ -16,3 +16,4 @@ examples/pydantic_ai_examples/.chat_app_messages.sqlite
 /question_graph_history.json
 /docs-site/.wrangler/
 /CLAUDE.md
+node_modules/

{pydantic_ai-0.0.52 → pydantic_ai-0.0.53}/Makefile

@@ -8,8 +8,12 @@
 .pre-commit: ## Check that pre-commit is installed
 	@pre-commit -V || echo 'Please install pre-commit: https://pre-commit.com/'
 
+.PHONY: .deno
+.deno: ## Check that deno is installed
+	@deno --version > /dev/null 2>&1 || (printf "\033[0;31m✖ Error: deno is not installed, but is needed for mcp-run-python\033[0m\n    Please install deno: https://docs.deno.com/runtime/getting_started/installation/\n" && exit 1)
+
 .PHONY: install
-install: .uv .pre-commit ## Install the package, dependencies, and pre-commit for local development
+install: .uv .pre-commit .deno ## Install the package, dependencies, and pre-commit for local development
 	uv sync --frozen --all-extras --all-packages --group lint --group docs
 	pre-commit install --install-hooks
 

{pydantic_ai-0.0.52 → pydantic_ai-0.0.53}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pydantic-ai
-Version: 0.0.52
+Version: 0.0.53
 Summary: Agent Framework / shim to use Pydantic with LLMs
 Project-URL: Homepage, https://ai.pydantic.dev
 Project-URL: Source, https://github.com/pydantic/pydantic-ai
@@ -28,9 +28,9 @@ Classifier: Topic :: Internet
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Requires-Python: >=3.9
-Requires-Dist: pydantic-ai-slim[anthropic,bedrock,cli,cohere,evals,groq,mcp,mistral,openai,vertexai]==0.0.52
+Requires-Dist: pydantic-ai-slim[anthropic,bedrock,cli,cohere,evals,groq,mcp,mistral,openai,vertexai]==0.0.53
 Provides-Extra: examples
-Requires-Dist: pydantic-ai-examples==0.0.52; extra == 'examples'
+Requires-Dist: pydantic-ai-examples==0.0.53; extra == 'examples'
 Provides-Extra: logfire
 Requires-Dist: logfire>=3.11.0; extra == 'logfire'
 Description-Content-Type: text/markdown

{pydantic_ai-0.0.52 → pydantic_ai-0.0.53}/tests/evals/test_dataset.py

@@ -611,10 +611,10 @@ async def test_from_text_failure():
                 '2 error(s) loading evaluators from registry',
                 [
                     ValueError(
-                        "Evaluator 'NotAnEvaluator' is not in the provided registry. Registered choices: ['Equals', 'EqualsExpected', 'Contains', 'IsInstance', 'MaxDuration', 'LLMJudge', 'HasMatchingSpan']"
+                        "Evaluator 'NotAnEvaluator' is not in the provided `custom_evaluator_types`. Valid choices: ['Equals', 'EqualsExpected', 'Contains', 'IsInstance', 'MaxDuration', 'LLMJudge', 'HasMatchingSpan']. If you are trying to use a custom evaluator, you must include its type in the `custom_evaluator_types` argument."
                     ),
                     ValueError(
-                        "Evaluator 'NotAnEvaluator' is not in the provided registry. Registered choices: ['Equals', 'EqualsExpected', 'Contains', 'IsInstance', 'MaxDuration', 'LLMJudge', 'HasMatchingSpan']"
+                        "Evaluator 'NotAnEvaluator' is not in the provided `custom_evaluator_types`. Valid choices: ['Equals', 'EqualsExpected', 'Contains', 'IsInstance', 'MaxDuration', 'LLMJudge', 'HasMatchingSpan']. If you are trying to use a custom evaluator, you must include its type in the `custom_evaluator_types` argument."
                     ),
                 ],
             )

{pydantic_ai-0.0.52 → pydantic_ai-0.0.53}/tests/evals/test_evaluator_common.py

@@ -222,10 +222,10 @@ async def test_llm_judge_evaluator(mocker: MockerFixture):
     assert result.value is True
     assert result.reason == 'Test passed'
 
-    mock_judge_output.assert_called_once_with('Hello world', 'Content contains a greeting', 'openai:gpt-4o')
+    mock_judge_output.assert_called_once_with('Hello world', 'Content contains a greeting', None)
 
     # Test with input
-    evaluator = LLMJudge(rubric='Output contains input', include_input=True)
+    evaluator = LLMJudge(rubric='Output contains input', include_input=True, model='openai:gpt-4o')
     result = await evaluator.evaluate(ctx)
     assert isinstance(result, EvaluationReason)
     assert result.value is True

{pydantic_ai-0.0.52 → pydantic_ai-0.0.53}/tests/evals/test_evaluators.py

@@ -7,6 +7,11 @@ import pytest
 from inline_snapshot import snapshot
 from pydantic import BaseModel, TypeAdapter
 
+from pydantic_ai.messages import ModelMessage, ModelResponse
+from pydantic_ai.models import Model, ModelRequestParameters
+from pydantic_ai.settings import ModelSettings
+from pydantic_ai.usage import Usage
+
 from ..conftest import try_import
 
 with try_import() as imports_successful:
@@ -108,6 +113,34 @@ async def test_evaluator_spec_serialization():
     assert adapter.dump_python(spec_single_arg, context={'use_short_form': True}) == snapshot({'MyEvaluator': 'value1'})
 
 
+async def test_llm_judge_serialization():
+    # Ensure models are serialized based on their system + name when used with LLMJudge
+
+    class MyModel(Model):
+        async def request(
+            self,
+            messages: list[ModelMessage],
+            model_settings: ModelSettings | None,
+            model_request_parameters: ModelRequestParameters,
+        ) -> tuple[ModelResponse, Usage]:
+            raise NotImplementedError
+
+        @property
+        def model_name(self) -> str:
+            return 'my-model'
+
+        @property
+        def system(self) -> str:
+            return 'my-system'
+
+    adapter = TypeAdapter(Evaluator)
+
+    assert adapter.dump_python(LLMJudge(rubric='my rubric', model=MyModel())) == {
+        'name': 'LLMJudge',
+        'arguments': {'model': 'my-system:my-model', 'rubric': 'my rubric'},
+    }
+
+
 async def test_evaluator_call(test_context: EvaluatorContext[TaskInput, TaskOutput, TaskMetadata]):
     """Test calling an Evaluator."""
 

{pydantic_ai-0.0.52 → pydantic_ai-0.0.53}/tests/models/test_gemini.py

@@ -61,9 +61,10 @@ async def test_model_simple(allow_model_requests: None):
     assert m.model_name == 'gemini-1.5-flash'
     assert 'x-goog-api-key' in m.client.headers
 
-    arc = ModelRequestParameters(function_tools=[], allow_text_result=True, result_tools=[])
-    tools = m._get_tools(arc)
-    tool_config = m._get_tool_config(arc, tools)
+    mrp = ModelRequestParameters(function_tools=[], allow_text_result=True, result_tools=[])
+    mrp = m.customize_request_parameters(mrp)
+    tools = m._get_tools(mrp)
+    tool_config = m._get_tool_config(mrp, tools)
     assert tools is None
     assert tool_config is None
 
@@ -93,9 +94,10 @@ async def test_model_tools(allow_model_requests: None):
         {'type': 'object', 'title': 'Result', 'properties': {'spam': {'type': 'number'}}, 'required': ['spam']},
     )
 
-    arc = ModelRequestParameters(function_tools=tools, allow_text_result=True, result_tools=[result_tool])
-    tools = m._get_tools(arc)
-    tool_config = m._get_tool_config(arc, tools)
+    mrp = ModelRequestParameters(function_tools=tools, allow_text_result=True, result_tools=[result_tool])
+    mrp = m.customize_request_parameters(mrp)
+    tools = m._get_tools(mrp)
+    tool_config = m._get_tool_config(mrp, tools)
     assert tools == snapshot(
         _GeminiTools(
             function_declarations=[
@@ -134,9 +136,10 @@ async def test_require_response_tool(allow_model_requests: None):
         'This is the tool for the final Result',
         {'type': 'object', 'title': 'Result', 'properties': {'spam': {'type': 'number'}}},
     )
-    arc = ModelRequestParameters(function_tools=[], allow_text_result=False, result_tools=[result_tool])
-    tools = m._get_tools(arc)
-    tool_config = m._get_tool_config(arc, tools)
+    mrp = ModelRequestParameters(function_tools=[], allow_text_result=False, result_tools=[result_tool])
+    mrp = m.customize_request_parameters(mrp)
+    tools = m._get_tools(mrp)
+    tool_config = m._get_tool_config(mrp, tools)
     assert tools == snapshot(
         _GeminiTools(
             function_declarations=[
@@ -193,9 +196,9 @@ async def test_json_def_replaced(allow_model_requests: None):
         'This is the tool for the final Result',
         json_schema,
     )
-    assert m._get_tools(
-        ModelRequestParameters(function_tools=[], allow_text_result=True, result_tools=[result_tool])
-    ) == snapshot(
+    mrp = ModelRequestParameters(function_tools=[], allow_text_result=True, result_tools=[result_tool])
+    mrp = m.customize_request_parameters(mrp)
+    assert m._get_tools(mrp) == snapshot(
         _GeminiTools(
             function_declarations=[
                 _GeminiFunction(
@@ -240,9 +243,9 @@ async def test_json_def_replaced_any_of(allow_model_requests: None):
         'This is the tool for the final Result',
         json_schema,
     )
-    assert m._get_tools(
-        ModelRequestParameters(function_tools=[], allow_text_result=True, result_tools=[result_tool])
-    ) == snapshot(
+    mrp = ModelRequestParameters(function_tools=[], allow_text_result=True, result_tools=[result_tool])
+    mrp = m.customize_request_parameters(mrp)
+    assert m._get_tools(mrp) == snapshot(
         _GeminiTools(
             function_declarations=[
                 _GeminiFunction(
@@ -304,7 +307,9 @@ async def test_json_def_recursive(allow_model_requests: None):
         json_schema,
     )
     with pytest.raises(UserError, match=r'Recursive `\$ref`s in JSON Schema are not supported by Gemini'):
-        m._get_tools(ModelRequestParameters(function_tools=[], allow_text_result=True, result_tools=[result_tool]))
+        mrp = ModelRequestParameters(function_tools=[], allow_text_result=True, result_tools=[result_tool])
+        mrp = m.customize_request_parameters(mrp)
+        m._get_tools(mrp)
 
 
 async def test_json_def_date(allow_model_requests: None):
@@ -335,9 +340,9 @@ async def test_json_def_date(allow_model_requests: None):
         'This is the tool for the final Result',
         json_schema,
     )
-    assert m._get_tools(
-        ModelRequestParameters(function_tools=[], allow_text_result=True, result_tools=[result_tool])
-    ) == snapshot(
+    mrp = ModelRequestParameters(function_tools=[], allow_text_result=True, result_tools=[result_tool])
+    mrp = m.customize_request_parameters(mrp)
+    assert m._get_tools(mrp) == snapshot(
         _GeminiTools(
             function_declarations=[
                 _GeminiFunction(