judgeval 0.0.31__py3-none-any.whl → 0.0.33__py3-none-any.whl

judgeval/scorers/judgeval_scorers/__init__.py

@@ -1,176 +0,0 @@
-from typing import Type, Optional, Any
-
-# Import implementations
-from judgeval.scorers.judgeval_scorers.api_scorers import (
-    ExecutionOrderScorer as APIExecutionOrderScorer,
-    JSONCorrectnessScorer as APIJSONCorrectnessScorer,
-    SummarizationScorer as APISummarizationScorer,
-    HallucinationScorer as APIHallucinationScorer,
-    FaithfulnessScorer as APIFaithfulnessScorer,
-    ContextualRelevancyScorer as APIContextualRelevancyScorer,
-    ContextualPrecisionScorer as APIContextualPrecisionScorer,
-    ContextualRecallScorer as APIContextualRecallScorer,
-    AnswerRelevancyScorer as APIAnswerRelevancyScorer,
-    AnswerCorrectnessScorer as APIAnswerCorrectnessScorer,  
-    ComparisonScorer as APIComparisonScorer,
-    InstructionAdherenceScorer as APIInstructionAdherenceScorer,
-    GroundednessScorer as APIGroundednessScorer,
-    DerailmentScorer as APIDerailmentScorer,
-)
-
-from judgeval.scorers.judgeval_scorers.local_implementations import (
-    AnswerRelevancyScorer as LocalAnswerRelevancyScorer,
-    ContextualPrecisionScorer as LocalContextualPrecisionScorer,
-    ContextualRecallScorer as LocalContextualRecallScorer,
-    ContextualRelevancyScorer as LocalContextualRelevancyScorer,
-    FaithfulnessScorer as LocalFaithfulnessScorer,
-    JsonCorrectnessScorer as LocalJsonCorrectnessScorer,
-    ExecutionOrderScorer as LocalExecutionOrderScorer,
-    HallucinationScorer as LocalHallucinationScorer,
-    SummarizationScorer as LocalSummarizationScorer,
-    AnswerCorrectnessScorer as LocalAnswerCorrectnessScorer,
-    ComparisonScorer as LocalComparisonScorer,
-    InstructionAdherenceScorer as LocalInstructionAdherenceScorer,
-)
-
-from judgeval.scorers.judgeval_scorers.classifiers import Text2SQLScorer
-
-
-class ScorerWrapper:
-    """
-    Wrapper class that can dynamically load either API or local implementation of a scorer.
-    """
-    def __init__(self, api_implementation: Type, local_implementation: Optional[Type] = None):
-        self.api_implementation = api_implementation
-        self.local_implementation = local_implementation
-        self._instance = None
-        self._init_args = None
-        self._init_kwargs = None
-        
-    def __call__(self, *args, **kwargs):
-        """Store initialization arguments for later use when implementation is loaded"""
-        self._init_args = args
-        self._init_kwargs = kwargs
-        return self
-    
-    def load_implementation(self, use_judgment: bool = True) -> Any:
-        """
-        Load the appropriate implementation based on the use_judgment flag.
-        
-        Args:
-            use_judgment (bool): If True, use API implementation. If False, use local implementation.
-        
-        Returns:
-            Instance of the appropriate implementation
-        
-        Raises:
-            ValueError: If local implementation is requested but not available
-        """
-        if self._instance is not None:
-            return self._instance
-            
-        if use_judgment:
-            implementation = self.api_implementation
-        else:
-            if self.local_implementation is None:
-                raise ValueError("No local implementation available for this scorer")
-            implementation = self.local_implementation
-            
-        args = self._init_args or ()
-        kwargs = self._init_kwargs or {}
-        self._instance = implementation(*args, **kwargs)
-        return self._instance
-    
-    def __getattr__(self, name):
-        """Defer all attribute access to the loaded implementation"""
-        if self._instance is None:
-            raise RuntimeError("Implementation not loaded. Call load_implementation() first")
-        return getattr(self._instance, name)
-
-# Create wrapped versions of all scorers
-
-AnswerCorrectnessScorer = ScorerWrapper(
-    api_implementation=APIAnswerCorrectnessScorer,
-    local_implementation=LocalAnswerCorrectnessScorer
-)
-
-AnswerRelevancyScorer = ScorerWrapper(
-    api_implementation=APIAnswerRelevancyScorer,
-    local_implementation=LocalAnswerRelevancyScorer
-)
-
-ExecutionOrderScorer = ScorerWrapper(
-    api_implementation=APIExecutionOrderScorer,
-    local_implementation=LocalExecutionOrderScorer
-)
-
-JSONCorrectnessScorer = ScorerWrapper(
-    api_implementation=APIJSONCorrectnessScorer,
-    local_implementation=LocalJsonCorrectnessScorer
-)
-
-SummarizationScorer = ScorerWrapper(
-    api_implementation=APISummarizationScorer,
-    local_implementation=LocalSummarizationScorer
-)
-
-HallucinationScorer = ScorerWrapper(
-    api_implementation=APIHallucinationScorer,
-    local_implementation=LocalHallucinationScorer
-)
-
-FaithfulnessScorer = ScorerWrapper(
-    api_implementation=APIFaithfulnessScorer,
-    local_implementation=LocalFaithfulnessScorer
-)
-
-ContextualRelevancyScorer = ScorerWrapper(
-    api_implementation=APIContextualRelevancyScorer,
-    local_implementation=LocalContextualRelevancyScorer
-)
-
-ContextualPrecisionScorer = ScorerWrapper(
-    api_implementation=APIContextualPrecisionScorer,
-    local_implementation=LocalContextualPrecisionScorer
-)
-
-ContextualRecallScorer = ScorerWrapper(
-    api_implementation=APIContextualRecallScorer,
-    local_implementation=LocalContextualRecallScorer
-)
-
-InstructionAdherenceScorer = ScorerWrapper(
-    api_implementation=APIInstructionAdherenceScorer,
-    local_implementation=LocalInstructionAdherenceScorer
-)
-
-def ComparisonScorer(threshold: float, criteria: str, description: str):
-    return ScorerWrapper(
-        api_implementation=APIComparisonScorer,
-        local_implementation=LocalComparisonScorer
-    )(threshold=threshold, criteria=criteria, description=description)
-
-GroundednessScorer = ScorerWrapper(
-    api_implementation=APIGroundednessScorer,
-)
-
-DerailmentScorer = ScorerWrapper(
-    api_implementation=APIDerailmentScorer,
-    local_implementation=LocalInstructionAdherenceScorer # TODO: add local implementation
-)
-
-__all__ = [
-    "ExecutionOrderScorer",
-    "JSONCorrectnessScorer",
-    "SummarizationScorer",
-    "HallucinationScorer",
-    "FaithfulnessScorer",
-    "ContextualRelevancyScorer",
-    "ContextualPrecisionScorer",
-    "ContextualRecallScorer",
-    "AnswerRelevancyScorer",
-    "Text2SQLScorer",
-    "ComparisonScorer",
-    "GroundednessScorer",
-    "DerailmentScorer",
-]

judgeval/version_check.py

@@ -0,0 +1,22 @@
+import importlib.metadata
+import requests
+import threading
+
+def check_latest_version(package_name: str = "judgeval"):
+    def _check():
+        try:
+            current_version = importlib.metadata.version(package_name)
+            response = requests.get(f"https://pypi.org/pypi/{package_name}/json", timeout=2)
+            latest_version = response.json()["info"]["version"]
+
+            if current_version != latest_version:
+                print(
+                    f"\033[93mUPDATE AVAILABLE:\033[0m You are using '{package_name}=={current_version}', "
+                    f"but the latest version is '{latest_version}'. While this version is still supported, "
+                    f"we recommend upgrading to avoid potential issues or missing features: "
+                    f"`pip install --upgrade {package_name}`"
+                )
+        except Exception:
+            pass
+
+    threading.Thread(target=_check, daemon=True).start()

{judgeval-0.0.31.dist-info → judgeval-0.0.33.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: judgeval
-Version: 0.0.31
+Version: 0.0.33
 Summary: Judgeval Package
 Project-URL: Homepage, https://github.com/JudgmentLabs/judgeval
 Project-URL: Issues, https://github.com/JudgmentLabs/judgeval/issues
@@ -12,12 +12,13 @@ Classifier: Programming Language :: Python :: 3
 Requires-Python: >=3.11
 Requires-Dist: anthropic
 Requires-Dist: fastapi
+Requires-Dist: google-genai
 Requires-Dist: langchain
 Requires-Dist: langchain-anthropic
 Requires-Dist: langchain-core
 Requires-Dist: langchain-huggingface
 Requires-Dist: langchain-openai
-Requires-Dist: litellm
+Requires-Dist: litellm==1.38.12
 Requires-Dist: nest-asyncio
 Requires-Dist: openai
 Requires-Dist: openpyxl
@@ -94,9 +95,21 @@ Create a file named `traces.py` with the following code:
     from judgeval.common.tracer import Tracer, wrap
     from openai import OpenAI
 
+    # Basic initialization
     client = wrap(OpenAI())
     judgment = Tracer(project_name="my_project")
 
+    # Or with S3 storage enabled
+    # NOTE: Make sure AWS creds correspond to an account with write access to the specified S3 bucket
+    judgment = Tracer(
+        project_name="my_project",
+        use_s3=True,
+        s3_bucket_name="my-traces-bucket", # Bucket created automatically if it doesn't exist
+        s3_aws_access_key_id="your-access-key",  # Optional: defaults to AWS_ACCESS_KEY_ID env var
+        s3_aws_secret_access_key="your-secret-key",  # Optional: defaults to AWS_SECRET_ACCESS_KEY env var
+        s3_region_name="us-west-1"  # Optional: defaults to AWS_REGION env var or "us-west-1"
+    )
+
     @judgment.observe(span_type="tool")
     def my_tool():
         return "Hello world!"

judgeval-0.0.33.dist-info/RECORD

@@ -0,0 +1,63 @@
+judgeval/__init__.py,sha256=x9HWt4waJwJMAqTuJSg2MezF9Zg-macEjeU-ajbly-8,330
+judgeval/clients.py,sha256=6VQmEqmfCngUdS2MuPBIpHvtDFqOENm8-_BmMvjLyRQ,944
+judgeval/constants.py,sha256=_XmVAkebMyGrDvvanAVlMgVd4p6MLHdEVsTQFI0kz1k,5411
+judgeval/evaluation_run.py,sha256=WGzx-Ug2qhSmunFo8NrmSstBRsOUc5KpKq0Lc51rqsM,6739
+judgeval/judgment_client.py,sha256=brRYmphZR-2IUre9kdOhfse1mYDilcIqUzzH21ROAdk,22208
+judgeval/rules.py,sha256=jkh1cXXcUf8oRY7xJUZfcQBYWn_rjUW4GvrhRt15PeU,20265
+judgeval/run_evaluation.py,sha256=elMpFHahyeukKKa09fmJM3c_afwJ00mbZRqm18l5f00,28481
+judgeval/version_check.py,sha256=bvJEidB7rAeXozoUbN9Yb97QOR_s2hgvpvj74jJ5HlY,943
+judgeval/common/__init__.py,sha256=7d24BRxtncpMj3AAJCj8RS7TqgjXmW777HVZH6-3sBs,289
+judgeval/common/exceptions.py,sha256=U-TxHLn7oVMezsMuoYouNDb2XuS8RCggfntYf5_6u4E,565
+judgeval/common/logger.py,sha256=KO75wWXCxhUHUMvLaTU31ZzOk6tkZBa7heQ7y0f-zFE,6062
+judgeval/common/s3_storage.py,sha256=W8wq9S7qJZdqdBR4sk3aEZ4K3-pz40DOoolOJrWs9Vo,3768
+judgeval/common/tracer.py,sha256=YsObK8VQXp1DDbU9xncU8NjuY-JUI54BqmG4olezrZc,92507
+judgeval/common/utils.py,sha256=LUQV5JfDr6wj7xHAJoNq-gofNZ6mjXbeKrGKzBME1KM,33533
+judgeval/data/__init__.py,sha256=xuKx_KCVHGp6CXvQuVmKl3v7pJp-qDaz0NccKxwjtO0,481
+judgeval/data/custom_example.py,sha256=QRBqiRiZS8UgVeTRHY0r1Jzm6yAYsyg6zmHxQGxdiQs,739
+judgeval/data/example.py,sha256=cJrmPGLel_P2sy1UaRvuVSAi35EnA9XMR11Lhp4aDLo,5930
+judgeval/data/result.py,sha256=Gb9tiSDsk1amXgh0cFG6JmlW_BMKxS2kuTwNA0rrHjA,3184
+judgeval/data/scorer_data.py,sha256=JVlaTx1EP2jw2gh3Vgx1CSEsvIFABAN26IquKyxwiJQ,3273
+judgeval/data/sequence.py,sha256=FmKVdzQP5VTujRCHDWk097MKRR-rJgbsdrxyCKee6tA,1994
+judgeval/data/sequence_run.py,sha256=RmYjfWKMWg-pcF5PLeiWfrhuDkjDZi5VEmAIEXN3Ib0,2104
+judgeval/data/datasets/__init__.py,sha256=IdNKhQv9yYZ_op0rdBacrFaFVmiiYQ3JTzXzxOTsEVQ,176
+judgeval/data/datasets/dataset.py,sha256=dhLo30hvpmmOK2R6O5wDs_neawUJ4lS8bb4S42SufNQ,13034
+judgeval/data/datasets/eval_dataset_client.py,sha256=xjj66BO9Es9IxXqzQe1RT_e0kpeKlt7OrhRoSuj4KHM,15085
+judgeval/integrations/langgraph.py,sha256=J-cQfFP52TjJewdSTe-fcsUC4HDvjNbXoxmbmF0SgiE,11743
+judgeval/judges/__init__.py,sha256=6X7VSwrwsdxGBNxCyapVRWGghhKOy3MVxFNMQ62kCXM,308
+judgeval/judges/base_judge.py,sha256=ch_S7uBB7lyv44Lf1d7mIGFpveOO58zOkkpImKgd9_4,994
+judgeval/judges/litellm_judge.py,sha256=EIL58Teptv8DzZUO3yP2RDQCDq-aoBB6HPZzPdK6KTg,2424
+judgeval/judges/mixture_of_judges.py,sha256=IJoi4Twk8ze1CJWVEp69k6TSqTCTGrmVYQ0qdffer60,15549
+judgeval/judges/together_judge.py,sha256=l00hhPerAZXg3oYBd8cyMtWsOTNt_0FIqoxhKJKQe3k,2302
+judgeval/judges/utils.py,sha256=9lvUxziGV86ISvVFxYBWc09TWFyAQgUTyPf_a9mD5Rs,2686
+judgeval/scorers/__init__.py,sha256=Mk-mWUt_gNpJqY_WIEuQynD6fxc34fWSRSuobMSrj94,1238
+judgeval/scorers/api_scorer.py,sha256=NQ_CrrUPhSUk1k2Q8rKpCG_TU2FT32sFEqvb-Yi54B0,2688
+judgeval/scorers/exceptions.py,sha256=eGW5CuJgZ5YJBFrE4FHDSF651PO1dKAZ379mJ8gOsfo,178
+judgeval/scorers/judgeval_scorer.py,sha256=79-JJurqHP-qTaWNWInx4SjvQYwXc9lvfPPNgwsh2yA,6773
+judgeval/scorers/prompt_scorer.py,sha256=PaAs2qRolw1P3_I061Xvk9qzvF4O-JR8g_39RqXnHcM,17728
+judgeval/scorers/score.py,sha256=r9QiT4-LIvivcJ6XxByrbswKSO8eQTtAD1UlXT_lcmo,18741
+judgeval/scorers/utils.py,sha256=iHQVTlIANbmCTXz9kTeSdOytgUZ_T74Re61ajqsk_WQ,6827
+judgeval/scorers/judgeval_scorers/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+judgeval/scorers/judgeval_scorers/api_scorers/__init__.py,sha256=_sDUBxSG536KGqXNi6dFpaYKghjEAadxBxaaxV9HuuE,1764
+judgeval/scorers/judgeval_scorers/api_scorers/answer_correctness.py,sha256=Fnd9CVIOZ73sWEWymsU5eBrrZqPFjMZ0BKpeW-PDyTg,711
+judgeval/scorers/judgeval_scorers/api_scorers/answer_relevancy.py,sha256=oETeN9K0HSIRdL2SDqn82Vskpwh5SlKnZvs5VDm2OBU,658
+judgeval/scorers/judgeval_scorers/api_scorers/comparison.py,sha256=kuzf9OWvpY38yYSwlBgneLkUZwJNM4FQqvbS66keA90,1249
+judgeval/scorers/judgeval_scorers/api_scorers/contextual_precision.py,sha256=tpSuzFAaW8X9xqA0aLLKwh7qmBK0Pc_bJZMIe_q412U,770
+judgeval/scorers/judgeval_scorers/api_scorers/contextual_recall.py,sha256=pFVhk4pLtQ-FnNlbI-dFF-SIh69Jza7erHqiPkFWoBo,758
+judgeval/scorers/judgeval_scorers/api_scorers/contextual_relevancy.py,sha256=RQ6DZwEhChfecd89Ey-T7ke--7qTaXZlRsNxwH8gaME,823
+judgeval/scorers/judgeval_scorers/api_scorers/derailment_scorer.py,sha256=V9WPuwNMm097V7IknKs8UkmAk0yjnBXTcJha_BHXxTA,475
+judgeval/scorers/judgeval_scorers/api_scorers/execution_order.py,sha256=Pb3CiNF2Ca826B92wJCVAi_68lJjLhqqCKwQKaflSUg,1294
+judgeval/scorers/judgeval_scorers/api_scorers/faithfulness.py,sha256=-BwOapqjryYNKNydtdkUiKIij76dY0O1jBmdc6dKazQ,692
+judgeval/scorers/judgeval_scorers/api_scorers/groundedness.py,sha256=ntEEeTANEOsGlcbiTAF_3r6BeSJEaVDns8po8T0L6Vg,692
+judgeval/scorers/judgeval_scorers/api_scorers/hallucination.py,sha256=k5gDOki-8KXrZXydvdSqDt3NZqQ28hXoOCHQf6jNxr4,686
+judgeval/scorers/judgeval_scorers/api_scorers/instruction_adherence.py,sha256=XnSGEkQfwVqaqnHEGMCsxNiHVzrsrej48uDbLoWc8CQ,678
+judgeval/scorers/judgeval_scorers/api_scorers/json_correctness.py,sha256=mMKEuR87_yanEuZJ5YSGFMHDD_oLVZ6-rQuciFaDOMA,1095
+judgeval/scorers/judgeval_scorers/api_scorers/summarization.py,sha256=QmWB8bVbDYHY5FcF0rYZE_3c2XXgMLRmR6aXJWfdMC4,655
+judgeval/scorers/judgeval_scorers/classifiers/__init__.py,sha256=Qt81W5ZCwMvBAne0LfQDb8xvg5iOG1vEYP7WizgwAZo,67
+judgeval/scorers/judgeval_scorers/classifiers/text2sql/__init__.py,sha256=8iTzMvou1Dr8pybul6lZHKjc9Ye2-0_racRGYkhEdTY,74
+judgeval/scorers/judgeval_scorers/classifiers/text2sql/text2sql_scorer.py,sha256=ly72Z7s_c8NID6-nQnuW8qEGEW2MqdvpJ-5WfXzbAQg,2579
+judgeval/tracer/__init__.py,sha256=wy3DYpH8U_z0GO_K_gOSkK0tTTD-u5eLDo0T5xIBoAc,147
+judgeval/utils/alerts.py,sha256=O19Xj7DA0YVjl8PWiuH4zfdZeu3yiLVvHfY8ah2wG0g,2759
+judgeval-0.0.33.dist-info/METADATA,sha256=KzTkGTHYE8Uplehvtk_7x30XrV0xe1bpd-tU5lt0mHg,6097
+judgeval-0.0.33.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+judgeval-0.0.33.dist-info/licenses/LICENSE.md,sha256=tKmCg7k5QOmxPK19XMfzim04QiQJPmgIm0pAn55IJwk,11352
+judgeval-0.0.33.dist-info/RECORD,,

judgeval/scorers/base_scorer.py

@@ -1,58 +0,0 @@
-"""
-Judgment Scorer class.
-
-Scores `Example`s using ready-made Judgment evaluators.
-"""
-
-from pydantic import BaseModel, field_validator
-from judgeval.common.logger import debug, info, warning, error
-
-from judgeval.constants import APIScorer, UNBOUNDED_SCORERS
-
-
-class APIJudgmentScorer(BaseModel):
-    """
-    Class for ready-made, "out-of-the-box" scorer that uses Judgment evaluators to score `Example`s.
-
-    Args:
-        score_type (APIScorer): The Judgment metric to use for scoring `Example`s
-        threshold (float): A value between 0 and 1 that determines the scoring threshold
-    """
-    score_type: APIScorer
-    threshold: float
-
-    @field_validator('threshold')
-    def validate_threshold(cls, v, info):
-        """
-        Validates that the threshold is between 0 and 1 inclusive.
-        """
-        score_type = info.data.get('score_type')
-        if score_type in UNBOUNDED_SCORERS:
-            if v < 0:
-                error(f"Threshold for {score_type} must be greater than 0, got: {v}")
-                raise ValueError(f"Threshold for {score_type} must be greater than 0, got: {v}")
-        else:
-            if not 0 <= v <= 1:
-                error(f"Threshold for {score_type} must be between 0 and 1, got: {v}")
-                raise ValueError(f"Threshold for {score_type} must be between 0 and 1, got: {v}")
-        return v
-
-    @field_validator('score_type')
-    def convert_to_enum_value(cls, v):
-        """
-        Validates that the `score_type` is a valid `JudgmentMetric` enum value.
-        Converts string values to `JudgmentMetric` enum values.
-        """
-        debug(f"Attempting to convert score_type value: {v}")
-        if isinstance(v, APIScorer):
-            info(f"Using existing JudgmentMetric: {v.value}")
-            return v.value
-        elif isinstance(v, str):
-            debug(f"Converting string value to JudgmentMetric enum: {v}")
-            return APIScorer[v.upper()].value
-        error(f"Invalid score_type value: {v}")
-        raise ValueError(f"Invalid value for score_type: {v}")
-    
-    def __str__(self):
-        return f"JudgmentScorer(score_type={self.score_type.value}, threshold={self.threshold})"
-    

judgeval/scorers/judgeval_scorers/local_implementations/__init__.py

@@ -1,27 +0,0 @@
-from judgeval.scorers.judgeval_scorers.local_implementations.answer_relevancy.answer_relevancy_scorer import AnswerRelevancyScorer
-from judgeval.scorers.judgeval_scorers.local_implementations.contextual_precision.contextual_precision_scorer import ContextualPrecisionScorer
-from judgeval.scorers.judgeval_scorers.local_implementations.contextual_recall.contextual_recall_scorer import ContextualRecallScorer
-from judgeval.scorers.judgeval_scorers.local_implementations.contextual_relevancy.contextual_relevancy_scorer import ContextualRelevancyScorer
-from judgeval.scorers.judgeval_scorers.local_implementations.faithfulness.faithfulness_scorer import FaithfulnessScorer
-from judgeval.scorers.judgeval_scorers.local_implementations.json_correctness.json_correctness_scorer import JsonCorrectnessScorer
-from judgeval.scorers.judgeval_scorers.local_implementations.execution_order.execution_order import ExecutionOrderScorer 
-from judgeval.scorers.judgeval_scorers.local_implementations.hallucination.hallucination_scorer import HallucinationScorer
-from judgeval.scorers.judgeval_scorers.local_implementations.summarization.summarization_scorer import SummarizationScorer
-from judgeval.scorers.judgeval_scorers.local_implementations.answer_correctness.answer_correctness_scorer import AnswerCorrectnessScorer
-from judgeval.scorers.judgeval_scorers.local_implementations.comparison.comparison_scorer import ComparisonScorer
-from judgeval.scorers.judgeval_scorers.local_implementations.instruction_adherence.instruction_adherence import InstructionAdherenceScorer
-
-__all__ = [
-    "AnswerCorrectnessScorer",
-    "AnswerRelevancyScorer",
-    "ComparisonScorer",
-    "ContextualPrecisionScorer",
-    "ContextualRecallScorer",
-    "ContextualRelevancyScorer",
-    "FaithfulnessScorer",
-    "JsonCorrectnessScorer",
-    "ExecutionOrderScorer",
-    "HallucinationScorer",
-    "SummarizationScorer",
-    "InstructionAdherenceScorer",
-]

judgeval/scorers/judgeval_scorers/local_implementations/answer_correctness/__init__.py

@@ -1,4 +0,0 @@
-from judgeval.scorers.judgeval_scorers.local_implementations.answer_correctness.answer_correctness_scorer import AnswerCorrectnessScorer
-
-
-__all__ = ["AnswerCorrectnessScorer"]

judgeval/scorers/judgeval_scorers/local_implementations/answer_correctness/answer_correctness_scorer.py

@@ -1,276 +0,0 @@
-from typing import Optional, List, Union, Tuple
-
-from judgeval.constants import APIScorer
-from judgeval.judges import JudgevalJudge
-from judgeval.judges.utils import create_judge
-from judgeval.data import Example, ExampleParams
-from judgeval.scorers import JudgevalScorer
-from judgeval.scorers.utils import (
-    get_or_create_event_loop,
-    parse_response_json,
-    scorer_progress_meter,
-    create_verbose_logs,
-    check_example_params,
-)
-from .prompts import (
-    ACVerdict,
-    AnswerCorrectnessTemplate,
-    Statements,
-    Verdicts,
-    Reason,
-)
-
-
-required_params = [
-    ExampleParams.INPUT,
-    ExampleParams.ACTUAL_OUTPUT,
-    ExampleParams.EXPECTED_OUTPUT,
-]
-
-
-class AnswerCorrectnessScorer(JudgevalScorer):
-    def __init__(
-        self,
-        threshold: float = 0.5,
-        model: Optional[Union[str, JudgevalJudge]] = None,
-        include_reason: bool = True,
-        async_mode: bool = True,
-        strict_mode: bool = False,
-        verbose_mode: bool = False
-    ):
-        super().__init__(
-            score_type=APIScorer.ANSWER_CORRECTNESS,
-            threshold=1 if strict_mode else threshold,
-            evaluation_model=None,
-            include_reason=include_reason,
-            async_mode=async_mode,
-            strict_mode=strict_mode,
-            verbose_mode=verbose_mode
-        )
-        self.model, self.using_native_model = create_judge(model)
-        self.evaluation_model = self.model.get_model_name()
-
-    async def _a_get_statements(self, expected_output: str) -> List[str]:
-        prompt = AnswerCorrectnessTemplate.deduce_statements(
-            expected_output=expected_output,
-        )
-        if self.using_native_model:
-            res = await self.model.a_generate(prompt)
-            data = parse_response_json(res, self)
-            return data["statements"]
-        else:
-            try:
-                res: Statements = await self.model.a_generate(
-                    prompt, schema=Statements
-                )
-                return res.statements
-            except TypeError:
-                res = await self.model.a_generate(prompt)
-                data = parse_response_json(res, self)
-                return data["statements"]
-
-    def _get_statements(self, expected_output: str) -> List[str]:
-        prompt = AnswerCorrectnessTemplate.deduce_statements(
-            expected_output=expected_output,
-        )
-        if self.using_native_model:
-            res = self.model.generate(prompt)
-            data = parse_response_json(res, self)
-            return data["statements"]
-        else:
-            try:
-                res: Statements = self.model.generate(
-                    prompt, schema=Statements
-                )
-                return res.statements
-            except TypeError:
-                res = self.model.generate(prompt)
-                data = parse_response_json(res, self)
-                return data["statements"]
-
-    async def _a_get_verdicts(self, actual_output: str) -> List[ACVerdict]:
-        if len(self.statements) == 0:
-            return []
-
-        prompt = AnswerCorrectnessTemplate.generate_verdicts(
-            actual_output=actual_output,
-            statements=self.statements,
-        )
-
-        if self.using_native_model:
-            res = await self.model.a_generate(prompt)
-            data = parse_response_json(res, self)
-            return [ACVerdict(**item) for item in data["verdicts"]]
-        else:
-            try:
-                res: Verdicts = await self.model.a_generate(prompt, schema=Verdicts)
-                return [item for item in res.verdicts]
-            except TypeError:
-                res = await self.model.a_generate(prompt)
-                data = parse_response_json(res, self)
-                return [ACVerdict(**item) for item in data["verdicts"]]
-
-    def _get_verdicts(self, actual_output: str) -> List[ACVerdict]:
-        if len(self.statements) == 0:
-            return []
-
-        prompt = AnswerCorrectnessTemplate.generate_verdicts(
-            actual_output=actual_output,
-            statements=self.statements,
-        )
-
-        if self.using_native_model:
-            res = self.model.generate(prompt)
-            data = parse_response_json(res, self)
-            return [ACVerdict(**item) for item in data["verdicts"]]
-        else:
-            try:
-                res: Verdicts = self.model.generate(prompt, schema=Verdicts)
-                return [item for item in res.verdicts]
-            except TypeError:
-                res = self.model.generate(prompt)
-                data = parse_response_json(res, self)
-                return [ACVerdict(**item) for item in data["verdicts"]]
-
-    async def _a_get_reason(self) -> str:
-        if self.include_reason is False:
-            return None
-
-        incorrect_statements: List[Tuple[str, str]] = []
-        for idx, verdict in enumerate(self.verdicts):
-            if verdict.verdict.strip().lower() == "no":
-                incorrect_statements.append((self.statements[idx], verdict.reason))
-
-        prompt = AnswerCorrectnessTemplate.generate_reason(
-            incorrect_statements=incorrect_statements,
-            score=format(self.score, ".2f"),
-        )
-        if self.using_native_model:
-            res = await self.model.a_generate(prompt)
-            data = parse_response_json(res, self)
-            return data["reason"]
-        else:
-            try:
-                res: Reason = await self.model.a_generate(
-                    prompt=prompt, schema=Reason
-                )
-                return res.reason
-            except TypeError:
-                res = await self.model.a_generate(prompt)
-                data = parse_response_json(res, self)
-                return data["reason"]
-
-    def _get_reason(self) -> str:
-        if self.include_reason is False:
-            return None
-
-        incorrect_statements: List[Tuple[str, str]] = []
-        for idx, verdict in enumerate(self.verdicts):
-            if verdict.verdict.strip().lower() == "no":
-                incorrect_statements.append((self.statements[idx], verdict.reason))
-
-        prompt = AnswerCorrectnessTemplate.generate_reason(
-            incorrect_statements=incorrect_statements,
-            score=format(self.score, ".2f"),
-        )
-        if self.using_native_model:
-            res = self.model.generate(prompt)
-            data = parse_response_json(res, self)
-            return data["reason"]
-        else:
-            try:
-                res: Reason = self.model.generate(
-                    prompt=prompt, schema=Reason
-                )
-                return res.reason
-            except TypeError:
-                res = self.model.generate(prompt)
-                data = parse_response_json(res, self)
-                return data["reason"]
-
-    def _compute_score(self) -> float:
-        number_of_verdicts = len(self.verdicts)
-        if number_of_verdicts == 0:
-            return 1
-
-        correct_count = 0
-        for verdict in self.verdicts:
-            if verdict.verdict.strip().lower() == "yes":
-                correct_count += 1
-
-        score = correct_count / number_of_verdicts
-        return 0 if self.strict_mode and score < self.threshold else score
-
-    def score_example(
-        self,
-        example: Example,
-        _show_indicator: bool = True,
-    ) -> float:
-        check_example_params(example, required_params, self)
-
-        with scorer_progress_meter(self, display_meter=_show_indicator):
-            try:
-                if self.async_mode:
-                    loop = get_or_create_event_loop()
-                    loop.run_until_complete(
-                        self.a_score_example(example, _show_indicator=False)
-                    )
-                else:
-                    self.statements = self._get_statements(example.expected_output)
-                    self.verdicts = self._get_verdicts(example.actual_output)
-                    self.score = self._compute_score()
-                    self.reason = self._get_reason()
-                    self.success = self.score >= self.threshold
-                    self.verbose_logs = create_verbose_logs(
-                        self,
-                        steps=[
-                            f"Statements:\n{self.statements}",
-                            f"Verdicts:\n{[v.model_dump() for v in self.verdicts]}",
-                            f"Score: {self.score}\nReason: {self.reason}",
-                        ],
-                    )
-                return self.score
-            except Exception as e:
-                print(f"Error in score_example for AnswerCorrectnessScorer: {e}")
-                raise
-    
-    async def a_score_example(
-        self, 
-        example: Example,
-        _show_indicator: bool = True,
-    ) -> float:
-        check_example_params(example, required_params, self)
-
-        with scorer_progress_meter(self, async_mode=True, display_meter=_show_indicator):
-            try:
-                self.statements: List[str] = await self._a_get_statements(example.expected_output)
-                self.verdicts: List[ACVerdict] = await self._a_get_verdicts(example.actual_output)
-                self.score = self._compute_score()
-                self.reason = await self._a_get_reason()
-                self.success = self.score >= self.threshold
-                self.verbose_logs = create_verbose_logs(
-                    self,
-                    steps=[
-                        f"Statements:\n{self.statements}",
-                        f"Verdicts:\n{[v.model_dump() for v in self.verdicts]}",
-                        f"Score: {self.score}\nReason: {self.reason}",
-                    ],
-                )
-                return self.score
-            except Exception as e:
-                print(f"Error in a_score_example for AnswerCorrectnessScorer: {e}")
-                raise
-
-    def _success_check(self) -> bool:
-        if self.error is not None:
-            self.success = False
-        else:
-            try:
-                self.success = self.score >= self.threshold
-            except:
-                self.success = False
-        return self.success
-
-    @property
-    def __name__(self):
-        return "Answer Correctness"