Make end call tool can be cancelled

backend/services/pipecat/processors.py

@@ -230,9 +230,14 @@ def message_text(message: dict) -> str:
 class UserInputProcessor(FrameProcessor):
     """Validate app inputs and coordinate their interruption boundary."""
 
-    def __init__(self, should_ignore_input: Callable[[], bool] | None = None):
+    def __init__(
+        self,
+        should_ignore_input: Callable[[], bool] | None = None,
+        cancel_pending_end: Callable[[], bool] | None = None,
+    ):
         super().__init__()
         self._should_ignore_input = should_ignore_input or (lambda: False)
+        self._cancel_pending_end = cancel_pending_end or (lambda: False)
         self._register_event_handler("on_user_input")
         self._register_event_handler("on_client_ready")
 
@@ -256,6 +261,9 @@ class UserInputProcessor(FrameProcessor):
             await self.push_frame(frame, direction)
             return
 
+        if user_input.run_immediately and user_input.interrupt:
+            self._cancel_pending_end()
+
         if self._should_ignore_input():
             logger.debug("当前会话输入已暂停,忽略用户输入")
             await self._emit_result(
@@ -371,15 +379,25 @@ class SessionUpdateProcessor(FrameProcessor):
 
 
 class CallEndingUserMuteStrategy(BaseUserMuteStrategy):
-    """Keep user media muted after an end-call tool starts terminating a call."""
+    """Mute committed endings and cancel interruptible ones on user activity."""
 
-    def __init__(self, is_call_ending: Callable[[], bool]):
+    def __init__(
+        self,
+        should_mute: Callable[[], bool],
+        *,
+        is_user_interruption: Callable[[object], bool] | None = None,
+        cancel_pending_end: Callable[[], bool] | None = None,
+    ):
         super().__init__()
-        self._is_call_ending = is_call_ending
+        self._should_mute = should_mute
+        self._is_user_interruption = is_user_interruption or (lambda _frame: False)
+        self._cancel_pending_end = cancel_pending_end or (lambda: False)
 
     async def process_frame(self, frame) -> bool:
         await super().process_frame(frame)
-        return self._is_call_ending()
+        if self._is_user_interruption(frame):
+            self._cancel_pending_end()
+        return self._should_mute()
 
 
 @dataclass
@@ -570,9 +588,14 @@ class RealtimeInputAudioGateProcessor(FrameProcessor):
 class RealtimeUserInputProcessor(FrameProcessor):
     """Route text-only user-input messages to a realtime service."""
 
-    def __init__(self, should_ignore_input: Callable[[], bool] | None = None):
+    def __init__(
+        self,
+        should_ignore_input: Callable[[], bool] | None = None,
+        cancel_pending_end: Callable[[], bool] | None = None,
+    ):
         super().__init__()
         self._should_ignore_input = should_ignore_input or (lambda: False)
+        self._cancel_pending_end = cancel_pending_end or (lambda: False)
         self._register_event_handler("on_user_input")
 
     async def process_frame(self, frame, direction: FrameDirection):
@@ -590,6 +613,8 @@ class RealtimeUserInputProcessor(FrameProcessor):
         if user_input is None:
             await self.push_frame(frame, direction)
             return
+        if user_input.run_immediately and user_input.interrupt:
+            self._cancel_pending_end()
         if self._should_ignore_input():
             await self._emit_error(
                 user_input.input_id,