dag

deepeval/metrics/__init__.py

@@ -4,6 +4,7 @@
     BaseMultimodalMetric,
 )
 
+from .dag.dag import DAGMetric
 from .bias.bias import BiasMetric
 from .toxicity.toxicity import ToxicityMetric
 from .hallucination.hallucination import HallucinationMetric

deepeval/metrics/dag/dag.py

@@ -57,11 +57,8 @@ def measure(
                     self.a_measure(test_case, _show_indicator=False)
                 )
             else:
-                dag = DeepAcyclicGraph(
-                    root_node=self.root_node, metric=self, test_case=test_case
-                )
-                self.score = dag._run()
-                self.success = self.score >= self.threshold
+                self.root_node._execute(metric=self, test_case=test_case)
+                self.success = self.is_successful()
                 return self.score
 
     async def a_measure(
@@ -77,11 +74,8 @@ async def a_measure(
         with metric_progress_indicator(
             self, async_mode=True, _show_indicator=_show_indicator
         ):
-            dag = DeepAcyclicGraph(
-                root_node=self.root_node, metric=self, test_case=test_case
-            )
-            self.score = await dag._a_run()
-            self.success = self.score >= self.threshold
+            await self.root_node._a_execute(metric=self, test_case=test_case)
+            self.success = self.is_successful()
             return self.score
 
     def is_successful(self) -> bool:

deepeval/metrics/dag/graph.py

@@ -1,6 +1,6 @@
 from typing import Set, List
 
-from deepeval.metrics.dag import BaseNode, VerdictNode
+from deepeval.metrics.dag import BaseNode
 from deepeval.test_case import LLMTestCase
 from deepeval.metrics import BaseMetric
 
@@ -16,42 +16,8 @@ def __init__(
         self.metric = metric
         self.test_case = test_case
 
-    def _topological_sort(
-        self, node: BaseNode, visited: Set[BaseNode], stack: List[BaseNode]
-    ) -> None:
-        if node in visited:
-            return
-        visited.add(node)
+    def execute(self) -> None:
+        self.root_node._execute(metric=self.metric, test_case=self.test_case)
 
-        if isinstance(node, VerdictNode):
-            if node.child is not None:
-                self._topological_sort(node.child, visited, stack)
-        elif hasattr(node, "children"):
-            for child in node.children:
-                self._topological_sort(child, visited, stack)
-
-        stack.append(node)
-
-    def evaluate(self) -> None:
-        visited: Set[BaseNode] = set()
-        stack: List[BaseNode] = []
-
-        self._topological_sort(self.root_node, visited, stack)
-
-        while stack:
-            node = stack.pop()
-            if isinstance(node, BaseNode):
-                node._execute(metric=self.metric, test_case=self.test_case)
-
-    async def a_evaluate(self) -> None:
-        visited: Set[BaseNode] = set()
-        stack: List[BaseNode] = []
-
-        self._topological_sort(self.root_node, visited, stack)
-
-        while stack:
-            node = stack.pop()
-            if isinstance(node, BaseNode):
-                await node._a_execute(
-                    metric=self.metric, test_case=self.test_case
-                )
+    async def a_execute(self) -> None:
+        self.root_node._a_execute(metric=self.metric, test_case=self.test_case)