Update error message (#9970)

in evals

libs/langchain/langchain/evaluation/loading.py

@@ -108,7 +108,6 @@ def load_evaluator(
     >>> from langchain.evaluation import load_evaluator, EvaluatorType
     >>> evaluator = load_evaluator(EvaluatorType.QA)
     """
-    llm = llm or ChatOpenAI(model="gpt-4", temperature=0)
     if evaluator not in _EVALUATOR_MAP:
         raise ValueError(
             f"Unknown evaluator type: {evaluator}"
@@ -116,6 +115,16 @@ def load_evaluator(
         )
     evaluator_cls = _EVALUATOR_MAP[evaluator]
     if issubclass(evaluator_cls, LLMEvalChain):
+        try:
+            llm = llm or ChatOpenAI(model="gpt-4", temperature=0)
+        except Exception as e:
+            raise ValueError(
+                f"Evaluation with the {evaluator_cls} requires a "
+                "language model to function."
+                " Failed to create the default 'gpt-4' model."
+                " Please manually provide an evaluation LLM"
+                " or check your openai credentials."
+            ) from e
         return evaluator_cls.from_llm(llm=llm, **kwargs)
     else:
         return evaluator_cls(**kwargs)
@@ -154,7 +163,6 @@ def load_evaluators(
     >>> evaluators = [EvaluatorType.QA, EvaluatorType.CRITERIA]
     >>> loaded_evaluators = load_evaluators(evaluators, criteria="helpfulness")
     """
-    llm = llm or ChatOpenAI(model="gpt-4", temperature=0)
     loaded = []
     for evaluator in evaluators:
         _kwargs = config.get(evaluator, {}) if config else {}

libs/langchain/langchain/smith/evaluation/runner_utils.py

@@ -35,7 +35,6 @@ from langchain.callbacks.tracers.base import BaseTracer
 from langchain.callbacks.tracers.evaluation import EvaluatorCallbackHandler
 from langchain.callbacks.tracers.langchain import LangChainTracer
 from langchain.chains.base import Chain
-from langchain.chat_models.openai import ChatOpenAI
 from langchain.evaluation.loading import load_evaluator
 from langchain.evaluation.schema import EvaluatorType, StringEvaluator
 from langchain.schema import ChatResult, LLMResult
@@ -493,7 +492,7 @@ def _determine_reference_key(
 
 def _construct_run_evaluator(
     eval_config: Union[EvaluatorType, str, EvalConfig],
-    eval_llm: BaseLanguageModel,
+    eval_llm: Optional[BaseLanguageModel],
     run_type: str,
     data_type: DataType,
     example_outputs: Optional[List[str]],
@@ -563,7 +562,6 @@ def _load_run_evaluators(
     Returns:
         A list of run evaluators.
     """
-    eval_llm = config.eval_llm or ChatOpenAI(model="gpt-4", temperature=0.0)
     run_evaluators = []
     input_key, prediction_key, reference_key = None, None, None
     if (
@@ -580,7 +578,7 @@ def _load_run_evaluators(
     for eval_config in config.evaluators:
         run_evaluator = _construct_run_evaluator(
             eval_config,
-            eval_llm,
+            config.eval_llm,
             run_type,
             data_type,
             example_outputs,