update eval type check

eval/run_eval.py

@@ -11,7 +11,6 @@ from langgraph.pregel.remote import RemoteGraph
 
 client = Client()
 
-TOLERANCE = 0.10  # should match within 10%
 NUMERIC_FIELDS = (
     "total_funding_mm_usd",
     "latest_round_amount_mm_usd",
@@ -41,8 +40,8 @@ EVALUATION_PROMPT = f"""You are an evaluator tasked with assessing the accuracy
 
 
 def evaluate_agent(outputs: dict, reference_outputs: dict):
-    if "info" not in outputs or not isinstance(outputs["info"], dict):
-        return 0.0
+    if "info" not in outputs:
+        raise ValueError("Agent output must contain 'info' key")
 
     class Score(BaseModel):
         """Evaluate the agent's output against the expected output."""