pydantic-evals 0.4.3__py3-none-any.whl → 0.4.5__py3-none-any.whl

pydantic_evals/evaluators/llm_as_a_judge.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+from collections.abc import Sequence
 from textwrap import dedent
 from typing import Any
 
@@ -7,6 +8,7 @@ from pydantic import BaseModel, Field
 from pydantic_core import to_json
 
 from pydantic_ai import Agent, models
+from pydantic_ai.messages import MultiModalContentTypes, UserContent
 from pydantic_ai.settings import ModelSettings
 
 __all__ = (
@@ -62,16 +64,7 @@ async def judge_output(
     If the model is not specified, a default model is used. The default model starts as 'openai:gpt-4o',
     but this can be changed using the `set_default_judge_model` function.
     """
-    user_prompt = dedent(
-        f"""
-        <Output>
-        {_stringify(output)}
-        </Output>
-        <Rubric>
-        {rubric}
-        </Rubric>
-        """
-    )
+    user_prompt = _build_prompt(output=output, rubric=rubric)
     return (
         await _judge_output_agent.run(user_prompt, model=model or _default_model, model_settings=model_settings)
     ).output
@@ -112,19 +105,8 @@ async def judge_input_output(
     If the model is not specified, a default model is used. The default model starts as 'openai:gpt-4o',
     but this can be changed using the `set_default_judge_model` function.
     """
-    user_prompt = dedent(
-        f"""
-        <Input>
-        {_stringify(inputs)}
-        </Input>
-        <Output>
-        {_stringify(output)}
-        </Output>
-        <Rubric>
-        {rubric}
-        </Rubric>
-        """
-    )
+    user_prompt = _build_prompt(inputs=inputs, output=output, rubric=rubric)
+
     return (
         await _judge_input_output_agent.run(user_prompt, model=model or _default_model, model_settings=model_settings)
     ).output
@@ -168,22 +150,7 @@ async def judge_input_output_expected(
     If the model is not specified, a default model is used. The default model starts as 'openai:gpt-4o',
     but this can be changed using the `set_default_judge_model` function.
     """
-    user_prompt = dedent(
-        f"""
-        <Input>
-        {_stringify(inputs)}
-        </Input>
-        <ExpectedOutput>
-        {_stringify(expected_output)}
-        </ExpectedOutput>
-        <Output>
-        {_stringify(output)}
-        </Output>
-        <Rubric>
-        {rubric}
-        </Rubric>
-        """
-    )
+    user_prompt = _build_prompt(inputs=inputs, output=output, rubric=rubric, expected_output=expected_output)
 
     return (
         await _judge_input_output_expected_agent.run(
@@ -227,19 +194,7 @@ async def judge_output_expected(
     If the model is not specified, a default model is used. The default model starts as 'openai:gpt-4o',
     but this can be changed using the `set_default_judge_model` function.
     """
-    user_prompt = dedent(
-        f"""
-        <ExpectedOutput>
-        {_stringify(expected_output)}
-        </ExpectedOutput>
-        <Output>
-        {_stringify(output)}
-        </Output>
-        <Rubric>
-        {rubric}
-        </Rubric>
-        """
-    )
+    user_prompt = _build_prompt(output=output, rubric=rubric, expected_output=expected_output)
     return (
         await _judge_output_expected_agent.run(
             user_prompt, model=model or _default_model, model_settings=model_settings
@@ -265,3 +220,41 @@ def _stringify(value: Any) -> str:
         return to_json(value).decode()
     except Exception:
         return repr(value)
+
+
+def _build_prompt(
+    output: Any,
+    rubric: str,
+    inputs: Any | None = None,
+    expected_output: Any | None = None,
+) -> str | Sequence[str | UserContent]:
+    """Build a prompt that includes input, output, and rubric."""
+    sections: list[str | UserContent] = []
+
+    if inputs is not None:
+        if isinstance(inputs, str):
+            sections.append(f'<Input>\n{inputs}\n</Input>')
+        else:
+            sections.append('<Input>\n')
+            if isinstance(inputs, Sequence):
+                for item in inputs:  # type: ignore
+                    if isinstance(item, (str, MultiModalContentTypes)):
+                        sections.append(item)
+                    else:
+                        sections.append(_stringify(item))
+            elif isinstance(inputs, MultiModalContentTypes):
+                sections.append(inputs)
+            else:
+                sections.append(_stringify(inputs))
+            sections.append('</Input>')
+
+    sections.append(f'<Output>\n{_stringify(output)}\n</Output>')
+    sections.append(f'<Rubric>\n{rubric}\n</Rubric>')
+
+    if expected_output is not None:
+        sections.append(f'<ExpectedOutput>\n{_stringify(expected_output)}\n</ExpectedOutput>')
+
+    if inputs is None or isinstance(inputs, str):
+        return '\n\n'.join(sections)  # type: ignore[arg-type]
+    else:
+        return sections

pydantic_evals/generation.py

@@ -47,7 +47,7 @@ async def generate_dataset(
         path: Optional path to save the generated dataset. If provided, the dataset will be saved to this location.
         dataset_type: The type of dataset to generate, with the desired input, output, and metadata types.
         custom_evaluator_types: Optional sequence of custom evaluator classes to include in the schema.
-        model: The PydanticAI model to use for generation. Defaults to 'gpt-4o'.
+        model: The Pydantic AI model to use for generation. Defaults to 'gpt-4o'.
         n_examples: Number of examples to generate. Defaults to 3.
         extra_instructions: Optional additional instructions to provide to the LLM.
 
@@ -59,7 +59,7 @@ async def generate_dataset(
     """
     output_schema = dataset_type.model_json_schema_with_evaluators(custom_evaluator_types)
 
-    # TODO(DavidM): Update this once we add better response_format and/or ResultTool support to PydanticAI
+    # TODO(DavidM): Update this once we add better response_format and/or ResultTool support to Pydantic AI
     agent = Agent(
         model,
         system_prompt=(

{pydantic_evals-0.4.3.dist-info → pydantic_evals-0.4.5.dist-info}/METADATA

@@ -1,12 +1,12 @@
 Metadata-Version: 2.4
 Name: pydantic-evals
-Version: 0.4.3
+Version: 0.4.5
 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
 Project-URL: Documentation, https://ai.pydantic.dev/evals
 Project-URL: Changelog, https://github.com/pydantic/pydantic-ai/releases
-Author-email: David Montague <david@pydantic.dev>
+Author-email: Samuel Colvin <samuel@pydantic.dev>, Marcelo Trylesinski <marcelotryle@gmail.com>, David Montague <david@pydantic.dev>, Alex Hall <alex@pydantic.dev>, Douwe Maan <douwe@pydantic.dev>
 License-Expression: MIT
 License-File: LICENSE
 Classifier: Development Status :: 4 - Beta
@@ -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==0.4.3
+Requires-Dist: pydantic-ai-slim==0.4.5
 Requires-Dist: pydantic>=2.10
 Requires-Dist: pyyaml>=6.0.2
 Requires-Dist: rich>=13.9.4
@@ -51,18 +51,18 @@ Description-Content-Type: text/markdown
 This is a library for evaluating non-deterministic (or "stochastic") functions in Python. It provides a simple,
 Pythonic interface for defining and running stochastic functions, and analyzing the results of running those functions.
 
-While this library is developed as part of [PydanticAI](https://ai.pydantic.dev), it only uses PydanticAI for a small
+While this library is developed as part of [Pydantic AI](https://ai.pydantic.dev), it only uses Pydantic AI for a small
 subset of generative functionality internally, and it is designed to be used with arbitrary "stochastic function"
-implementations. In particular, it can be used with other (non-PydanticAI) AI libraries, agent frameworks, etc.
+implementations. In particular, it can be used with other (non-Pydantic AI) AI libraries, agent frameworks, etc.
 
-As with PydanticAI, this library prioritizes type safety and use of common Python syntax over esoteric, domain-specific
+As with Pydantic AI, this library prioritizes type safety and use of common Python syntax over esoteric, domain-specific
 use of Python syntax.
 
 Full documentation is available at [ai.pydantic.dev/evals](https://ai.pydantic.dev/evals).
 
 ## Example
 
-While you'd typically use Pydantic Evals with more complex functions (such as PydanticAI agents or graphs), here's a
+While you'd typically use Pydantic Evals with more complex functions (such as Pydantic AI agents or graphs), here's a
 quick example that evaluates a simple function against a test case using both custom and built-in evaluators:
 
 ```python
@@ -110,7 +110,7 @@ report.print(include_input=True, include_output=True)
 """
 ```
 
-Using the library with more complex functions, such as PydanticAI agents, is similar — all you need to do is define a
+Using the library with more complex functions, such as Pydantic AI agents, is similar — all you need to do is define a
 task function wrapping the function you want to evaluate, with a signature that matches the inputs and outputs of your
 test cases.
 

{pydantic_evals-0.4.3.dist-info → pydantic_evals-0.4.5.dist-info}/RECORD

@@ -1,7 +1,7 @@
 pydantic_evals/__init__.py,sha256=OKRbfhdc8UZPzrPJMZUQwvzIxLhXmEZxz1ZuD921fy4,839
 pydantic_evals/_utils.py,sha256=PfhmPbdQp-q90s568LuG45zDDXxgO13BEz8MQJK8qw4,2922
 pydantic_evals/dataset.py,sha256=SY0k2htYG0d0KRRem3pnQdN7rPztJ_TCFnCb0zkXbCk,46477
-pydantic_evals/generation.py,sha256=-w-4-zpJuW8mLj5ed60LUYm--b-2G42p-UDuPhOQgRE,3492
+pydantic_evals/generation.py,sha256=Yd1rfbsDjjBBHDk-1KDu48hlITjM2-74rTnPBD_sqbA,3494
 pydantic_evals/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 pydantic_evals/evaluators/__init__.py,sha256=uj110viFiDvqrIcuDcWexk_uBgJqhKMGPHT2YvDb7LA,624
 pydantic_evals/evaluators/_run_evaluator.py,sha256=Dsnqxno7CrcKWYcnkLuwvPKWQGDRBmbBTwwstcmc0ak,2448
@@ -9,7 +9,7 @@ pydantic_evals/evaluators/_spec.py,sha256=Xi_FHwnmAZ1x2hoJFw4MBZuG0TilNKqMRW3P74
 pydantic_evals/evaluators/common.py,sha256=ZBrNTfPJoOpT4WNXTRGS0UcKhnuhfYJxjNzum-zHFk8,12064
 pydantic_evals/evaluators/context.py,sha256=8osQRCFW8Vekw7JiiOOCHHH3HOGdhDaUlr8i-twSetg,3870
 pydantic_evals/evaluators/evaluator.py,sha256=yOEKLOxElm7_4tLcq6_myXI0e4Ei9svZP9y5DTq4SYI,11147
-pydantic_evals/evaluators/llm_as_a_judge.py,sha256=raty91NWdu_FEt4ze_ugQHCouj1o72gYe3abJBtMqlU,8793
+pydantic_evals/evaluators/llm_as_a_judge.py,sha256=xQjaGuCRXZdlExacFyR4Y4kFmwBh2QxAfEyaed_aqvk,9615
 pydantic_evals/otel/__init__.py,sha256=i2p3vDrOW039N4XM-UkozDhCm0ZmE6ZSs1yV5t03vd0,117
 pydantic_evals/otel/_context_in_memory_span_exporter.py,sha256=vIDF9-6lDuNKZuSM5hN_R8VRK4jzmdfe1DgWdXwxVbc,6758
 pydantic_evals/otel/_context_subtree.py,sha256=Iazp4w3IIBMCrkqWL-hTG-2QG_-2X81p794WG9MAsGk,1175
@@ -17,7 +17,7 @@ pydantic_evals/otel/_errors.py,sha256=aW1414eTofpA7R_DUgOeT-gj7YA6OXmm8Y4oYeFukD
 pydantic_evals/otel/span_tree.py,sha256=LV5Hsyo4riJzevHyBz8wxP82S-ry5zeKYi9bKWjGCS8,23057
 pydantic_evals/reporting/__init__.py,sha256=k_3tteqXGh0yGvgpN68gB0CjG9wzrakzDTve2GHend4,42148
 pydantic_evals/reporting/render_numbers.py,sha256=8SKlK3etbD7HnSWWHCE993ceCNLZCepVQ-SsqUIhyxk,6916
-pydantic_evals-0.4.3.dist-info/METADATA,sha256=Zdwy3MQ-fI3g1GxcKplPCQQZLsR8FdmyYVH2tE1capk,7785
-pydantic_evals-0.4.3.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-pydantic_evals-0.4.3.dist-info/licenses/LICENSE,sha256=vA6Jc482lEyBBuGUfD1pYx-cM7jxvLYOxPidZ30t_PQ,1100
-pydantic_evals-0.4.3.dist-info/RECORD,,
+pydantic_evals-0.4.5.dist-info/METADATA,sha256=6axvUhXnQNjpGjtEIsUv2FHDNQnOQeBMEZmaqzfoo_s,7938
+pydantic_evals-0.4.5.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+pydantic_evals-0.4.5.dist-info/licenses/LICENSE,sha256=vA6Jc482lEyBBuGUfD1pYx-cM7jxvLYOxPidZ30t_PQ,1100
+pydantic_evals-0.4.5.dist-info/RECORD,,