fix text input no context update

engine/text_stream.py

@@ -1,5 +1,9 @@
 from __future__ import annotations
 
+from dataclasses import dataclass
+
+from loguru import logger
+
 from pipecat.frames.frames import (
     Frame,
     InterruptionFrame,
@@ -9,7 +13,16 @@ from pipecat.frames.frames import (
     OutputTransportMessageUrgentFrame,
     TTSSpeakFrame,
 )
+from pipecat.processors.aggregators.llm_context import LLMContext
 from pipecat.processors.frame_processor import FrameDirection, FrameProcessor
+from pipecat.utils.time import time_now_iso8601
+
+
+@dataclass
+class ProductAssistantTurnStoppedMessage:
+    content: str
+    interrupted: bool
+    timestamp: str
 
 
 class ProductTextStreamProcessor(FrameProcessor):
@@ -29,10 +42,13 @@ class ProductTextStreamProcessor(FrameProcessor):
     started/delta/final sequence for its fixed text.
     """
 
-    def __init__(self) -> None:
+    def __init__(self, context: LLMContext | None = None) -> None:
         super().__init__()
+        self._context = context
         self._aggregation: list[str] = []
         self._turn_active = False
+        self._turn_start_timestamp = ""
+        self._register_event_handler("on_assistant_turn_stopped")
 
     async def process_frame(self, frame: Frame, direction: FrameDirection) -> None:
         await super().process_frame(frame, direction)
@@ -62,6 +78,7 @@ class ProductTextStreamProcessor(FrameProcessor):
             return
         self._turn_active = True
         self._aggregation = []
+        self._turn_start_timestamp = time_now_iso8601()
         await self._emit("response.text.started")
 
     async def _delta(self, text: str) -> None:
@@ -78,11 +95,26 @@ class ProductTextStreamProcessor(FrameProcessor):
         full_text = "".join(self._aggregation)
         self._turn_active = False
         self._aggregation = []
+        if self._context and full_text:
+            self._context.add_message({"role": "assistant", "content": full_text})
+            logger.info(
+                "Assistant committed to LLM context before TTS: "
+                f"{full_text[:120]}"
+            )
         await self._emit(
             "response.text.final",
             text=full_text,
             interrupted=interrupted,
         )
+        await self._call_event_handler(
+            "on_assistant_turn_stopped",
+            ProductAssistantTurnStoppedMessage(
+                content=full_text,
+                interrupted=interrupted,
+                timestamp=self._turn_start_timestamp or time_now_iso8601(),
+            ),
+        )
+        self._turn_start_timestamp = ""
 
     async def _emit(self, event_type: str, **payload: object) -> None:
         await self.push_frame(