[issue-6025] fix: mapping callables not reading task output

Author: kriogenia

sdks/python/src/opik/evaluation/metrics/arguments_helpers.py

@@ -59,7 +59,7 @@ def create_scoring_inputs(
     else:
         for key, value in scoring_key_mapping.items():
             if callable(value):
-                mapped_inputs[key] = value(dataset_item)
+                mapped_inputs[key] = value(mapped_inputs)
             else:
                 if value not in mapped_inputs:
                     LOGGER.debug(

sdks/python/tests/unit/evaluation/test_create_scoring_inputs.py

@@ -37,14 +37,20 @@ def test_create_scoring_inputs_string_mapping():
 
 
 def test_create_scoring_inputs_callable_mapping():
-    """Test when scoring_key_mapping contains callable mappings for nested dictionaries"""
+    """Test when scoring_key_mapping contains callable mappings for nested dictionaries in dataset"""
     dataset_item = {
         "input": {"message": "hello"},
         "expected_output": {"message": "world"},
     }
-    task_output = {"result": "hello world"}
+    task_output = {
+        "result": "hello world",
+        "actual_output": {"message": "foo"},
+    }
 
-    mapping = {"reference": lambda x: x["expected_output"]["message"]}
+    mapping = {
+        "reference": lambda x: x["expected_output"]["message"],
+        "from_output": lambda x: x["actual_output"]["message"],
+    }
 
     result = create_scoring_inputs(
         dataset_item=dataset_item, task_output=task_output, scoring_key_mapping=mapping
@@ -55,6 +61,7 @@ def test_create_scoring_inputs_callable_mapping():
         "expected_output": {"message": "world"},
         "result": "hello world",
         "reference": "world",
+        "from_output": "foo",
     }
     assert_dicts_equal(result, expected)