feat: add deterministic message interaction stages

backend/services/pipecat/call_lifecycle.py

@@ -2,6 +2,8 @@
 
 from __future__ import annotations
 
+import asyncio
+from collections import deque
 from collections.abc import Awaitable, Callable
 
 from loguru import logger
@@ -19,6 +21,7 @@ class CallEndCoordinator:
         self._speaking = False
         self._response_speech_started = False
         self._tracked_speeches = 0
+        self._tracked_speech_completions: deque[asyncio.Future[None]] = deque()
         self._finish_after_tracked_speech = False
         self._finished = False
         self._reason = "completed"
@@ -39,9 +42,12 @@ class CallEndCoordinator:
         """Wait for the next observed bot speech to finish."""
         self._armed = True
 
-    def track_speech(self) -> None:
-        """Register one fixed utterance before its TTSSpeakFrame is queued."""
+    def track_speech(self) -> Awaitable[None]:
+        """Register fixed speech and return its transport completion signal."""
+        completion = asyncio.get_running_loop().create_future()
+        self._tracked_speech_completions.append(completion)
         self._tracked_speeches += 1
+        return completion
 
     async def arm_after_tracked_speech(self) -> None:
         """Finish after every already queued fixed utterance has played."""
@@ -73,6 +79,9 @@ class CallEndCoordinator:
             self._speaking = False
             if self._tracked_speeches > 0:
                 self._tracked_speeches -= 1
+                completion = self._tracked_speech_completions.popleft()
+                if not completion.done():
+                    completion.set_result(None)
             if (
                 self._finish_after_tracked_speech
                 and self._tracked_speeches == 0