pass flag instead of determining end-to-end utterance on the fly.

rasa/core/actions/action.py

@@ -6,6 +6,7 @@
 import aiohttp
 
 import rasa.core
+from rasa.core.policies.policy import PolicyPrediction
 
 from rasa.shared.core import events
 from rasa.core.constants import DEFAULT_REQUEST_TIMEOUT
@@ -78,14 +79,15 @@ def default_actions(action_endpoint: Optional[EndpointConfig] = None) -> List["A
     ]
 
 
-def action_for_index(
-    index: int, domain: Domain, action_endpoint: Optional[EndpointConfig]
+def action_for_prediction(
+    prediction: PolicyPrediction,
+    domain: Domain,
+    action_endpoint: Optional[EndpointConfig],
 ) -> "Action":
-    """Get an action based on its index in the list of available actions.
+    """Gets an instantiated `Action` based on the winning policy's prediction.
 
     Args:
-        index: The index of the action. This is usually used by `Policy`s as they
-            predict the action index instead of the name.
+        prediction: The prediction for the next action.
         domain: The `Domain` of the current model. The domain contains the actions
             provided by the user + the default actions.
         action_endpoint: Can be used to run `custom_actions`
@@ -95,17 +97,26 @@ def action_for_index(
         The instantiated `Action` or `None` if no `Action` was found for the given
         index.
     """
+    index = prediction.max_confidence_index
     if domain.num_actions <= index or index < 0:
         raise IndexError(
             f"Cannot access action at index {index}. "
             f"Domain has {domain.num_actions} actions."
         )
 
-    return action_for_name(domain.action_names[index], domain, action_endpoint)
+    return action_for_name(
+        domain.action_names[index],
+        domain,
+        action_endpoint,
+        prediction.is_end_to_end_prediction,
+    )
 
 
 def action_for_name(
-    action_name: Text, domain: Domain, action_endpoint: Optional[EndpointConfig]
+    action_name: Text,
+    domain: Domain,
+    action_endpoint: Optional[EndpointConfig],
+    is_end_to_end_prediction: bool = False,
 ) -> "Action":
     """Create an `Action` object based on the name of the `Action`.
 
@@ -115,6 +126,7 @@ def action_for_name(
             provided by the user + the default actions.
         action_endpoint: Can be used to run `custom_actions`
             (e.g. using the `rasa-sdk`).
+        is_end_to_end_prediction: `True` if it is an end-to-end prediction.
 
     Returns:
         The instantiated `Action` or `None` if no `Action` was found for the given
@@ -124,7 +136,9 @@ def action_for_name(
     if action_name not in domain.action_names:
         domain.raise_action_not_found_exception(action_name)
 
-    return action_from_name(action_name, domain, action_endpoint)
+    return action_from_name(
+        action_name, domain, action_endpoint, is_end_to_end_prediction
+    )
 
 
 def is_retrieval_action(action_name: Text, retrieval_intents: List[Text]) -> bool:
@@ -148,14 +162,18 @@ def is_retrieval_action(action_name: Text, retrieval_intents: List[Text]) -> boo
 
 
 def action_from_name(
-    name: Text, domain: Domain, action_endpoint: Optional[EndpointConfig]
+    name: Text,
+    domain: Domain,
+    action_endpoint: Optional[EndpointConfig],
+    is_end_to_end_prediction: bool = False,
 ) -> "Action":
     """Retrieves an action by its name.
 
     Args:
         name: The name of the action.
         domain: The current model domain.
         action_endpoint: The endpoint to execute custom actions.
+        is_end_to_end_prediction: `True` if it is an end-to-end prediction.
 
     Returns:
         The instantiated action.
@@ -170,7 +188,7 @@ def action_from_name(
     ):
         return ActionRetrieveResponse(name)
 
-    if name.startswith(UTTER_PREFIX) or name in domain.action_texts:
+    if name.startswith(UTTER_PREFIX) or is_end_to_end_prediction:
         return ActionUtterTemplate(name)
 
     is_form = name in domain.form_names

rasa/core/processor.py

@@ -354,8 +354,8 @@ def predict_next_action(
         """
         prediction = self._get_next_action_probabilities(tracker)
 
-        action = rasa.core.actions.action.action_for_index(
-            prediction.max_confidence_index, self.domain, self.action_endpoint
+        action = rasa.core.actions.action.action_for_prediction(
+            prediction, self.domain, self.action_endpoint
         )
 
         logger.debug(

tests/core/test_ensemble.py

@@ -177,8 +177,8 @@ def test_fallback_mapping_restart():
         tracker, domain, RegexInterpreter()
     )
     index_of_mapping_policy = 1
-    next_action = rasa.core.actions.action.action_for_index(
-        prediction.max_confidence_index, domain, None
+    next_action = rasa.core.actions.action.action_for_prediction(
+        prediction, domain, None
     )
 
     assert (
@@ -222,8 +222,8 @@ def test_mapping_wins_over_form(events: List[Event]):
         tracker, domain, RegexInterpreter()
     )
 
-    next_action = rasa.core.actions.action.action_for_index(
-        prediction.max_confidence_index, domain, None
+    next_action = rasa.core.actions.action.action_for_prediction(
+        prediction, domain, None
     )
 
     index_of_mapping_policy = 0
@@ -265,8 +265,8 @@ def test_form_wins_over_everything_else(ensemble: SimplePolicyEnsemble):
         tracker, domain, RegexInterpreter()
     )
 
-    next_action = rasa.core.actions.action.action_for_index(
-        prediction.max_confidence_index, domain, None
+    next_action = rasa.core.actions.action.action_for_prediction(
+        prediction, domain, None
     )
 
     index_of_form_policy = 0
@@ -291,8 +291,8 @@ def test_fallback_wins_over_mapping():
         tracker, domain, RegexInterpreter()
     )
     index_of_fallback_policy = 0
-    next_action = rasa.core.actions.action.action_for_index(
-        prediction.max_confidence_index, domain, None
+    next_action = rasa.core.actions.action.action_for_prediction(
+        prediction, domain, None
     )
 
     assert (

tests/core/test_policies.py

@@ -286,9 +286,7 @@ def test_missing_classes_filled_correctly(
             new_tracker = DialogueStateTracker(DEFAULT_SENDER_ID, default_domain.slots)
             for e in tr.applied_events():
                 if isinstance(e, ActionExecuted):
-                    new_action = rasa.core.actions.action.action_for_index(
-                        np.random.choice(classes), default_domain, action_endpoint=None
-                    ).name()
+                    new_action = default_domain.action_names[np.random.choice(classes)]
                     new_tracker.update(ActionExecuted(new_action))
                 else:
                     new_tracker.update(e)