Return feedback with failed response if there's an error (#9223)

In Evals

libs/langchain/langchain/smith/evaluation/string_run_evaluator.py

@@ -313,17 +313,30 @@ class StringRunEvaluatorChain(Chain, RunEvaluator):
         self, run: Run, example: Optional[Example] = None
     ) -> EvaluationResult:
         """Evaluate an example."""
-        result = self({"run": run, "example": example}, include_run_info=True)
-        return self._prepare_evaluator_output(result)
+        try:
+            result = self({"run": run, "example": example}, include_run_info=True)
+            return self._prepare_evaluator_output(result)
+        except Exception as e:
+            return EvaluationResult(
+                key=self.string_evaluator.evaluation_name,
+                comment=f"Error evaluating run {run.id}: {e}",
+                # TODO: Add run ID once we can declare it via callbacks
+            )
 
     async def aevaluate_run(
         self, run: Run, example: Optional[Example] = None
     ) -> EvaluationResult:
         """Evaluate an example."""
-        result = await self.acall(
-            {"run": run, "example": example}, include_run_info=True
-        )
-        return self._prepare_evaluator_output(result)
+        try:
+            result = await self.acall(
+                {"run": run, "example": example}, include_run_info=True
+            )
+            return self._prepare_evaluator_output(result)
+        except Exception as e:
+            return EvaluationResult(
+                key=self.string_evaluator.evaluation_name,
+                comment=f"Error evaluating run {run.id}: {e}",
+            )
 
     @classmethod
     def from_run_and_data_type(

libs/langchain/tests/unit_tests/smith/evaluation/test_string_run_evaluator.py

@@ -0,0 +1,27 @@
+"""Tests for the string run evaluator."""
+
+from unittest.mock import MagicMock
+
+from langchain.evaluation import criteria
+from langchain.smith.evaluation.string_run_evaluator import (
+    ChainStringRunMapper,
+    StringRunEvaluatorChain,
+)
+from tests.unit_tests.llms import fake_llm
+
+
+def test_evaluate_run() -> None:
+    run_mapper = ChainStringRunMapper()
+    example_mapper = MagicMock()
+    string_evaluator = criteria.CriteriaEvalChain.from_llm(fake_llm.FakeLLM())
+    evaluator = StringRunEvaluatorChain(
+        run_mapper=run_mapper,
+        example_mapper=example_mapper,
+        name="test_evaluator",
+        string_evaluator=string_evaluator,
+    )
+    run = MagicMock()
+    example = MagicMock()
+    res = evaluator.evaluate_run(run, example)
+    assert res.comment.startswith("Error evaluating run ")
+    assert res.key == string_evaluator.evaluation_name