TTSServices: for now just specify a single text aggregator

CHANGELOG.md

@@ -21,8 +21,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Added new `BaseTextAggregator`. Text aggregators are used by the TTS service
   to aggregate LLM tokens and decide when the aggregated text should be pushed
   to the TTS service. They also allow for the text to be manipulated while it's
-  being aggregated. Multiple text aggregators can be passed with
+  being aggregated. A text aggregator can be passed via `text_aggregator` to the
-  `text_aggregators` to the TTS service.
+  TTS service.
 
 - Added new `UltravoxSTTService`.
   (see https://github.com/fixie-ai/ultravox)

examples/foundational/35-pattern-pair-voice-switching.py

@@ -119,7 +119,7 @@ async def main():
         tts = CartesiaTTSService(
             api_key=os.getenv("CARTESIA_API_KEY"),
             voice_id=VOICE_IDS["narrator"],
-            text_aggregators=[pattern_aggregator],
+            text_aggregator=pattern_aggregator,
         )
 
         # Initialize LLM

src/pipecat/services/ai_services.py

@@ -239,7 +239,7 @@ class TTSService(AIService):
         # TTS output sample rate
         sample_rate: Optional[int] = None,
         # Text aggregator to aggregate incoming tokens and decide when to push to the TTS.
-        text_aggregators: Sequence[BaseTextAggregator] = [],
+        text_aggregator: Optional[BaseTextAggregator] = None,
         # Text filter executed after text has been aggregated.
         text_filters: Sequence[BaseTextFilter] = [],
         text_filter: Optional[BaseTextFilter] = None,
@@ -257,10 +257,7 @@ class TTSService(AIService):
         self._sample_rate = 0
         self._voice_id: str = ""
         self._settings: Dict[str, Any] = {}
-        # Ensure there's at least one text aggregator.
+        self._text_aggregator: BaseTextAggregator = text_aggregator or SimpleTextAggregator()
-        self._text_aggregators: Sequence[BaseTextAggregator] = text_aggregators or [
-            SimpleTextAggregator()
-        ]
         self._text_filters: Sequence[BaseTextFilter] = text_filters
         if text_filter:
             import warnings
@@ -358,8 +355,8 @@ class TTSService(AIService):
             # pause to avoid audio overlapping.
             await self._maybe_pause_frame_processing()
 
-            sentence = self._text_aggregators[-1].text
+            sentence = self._text_aggregator.text
-            self._reset_aggregators()
+            self._text_aggregator.reset()
             self._processing_text = False
             await self._push_tts_frames(sentence)
             if isinstance(frame, LLMFullResponseEndFrame):
@@ -405,8 +402,7 @@ class TTSService(AIService):
 
     async def _handle_interruption(self, frame: StartInterruptionFrame, direction: FrameDirection):
         self._processing_text = False
-        for aggregator in self._text_aggregators:
+        self._text_aggregator.handle_interruption()
-            aggregator.handle_interruption()
         for filter in self._text_filters:
             filter.handle_interruption()
 
@@ -418,25 +414,12 @@ class TTSService(AIService):
         if self._pause_frame_processing:
             await self.resume_processing_frames()
 
-    def _reset_aggregators(self):
-        for aggregator in self._text_aggregators:
-            aggregator.reset()
-
     async def _process_text_frame(self, frame: TextFrame):
         text: Optional[str] = None
         if not self._aggregate_sentences:
             text = frame.text
         else:
-            current_text = frame.text
+            text = self._text_aggregator.aggregate(frame.text)
-
-            # Process all aggregators except the last one.
-            for aggregator in self._text_aggregators[:-1]:
-                aggregator.aggregate(current_text)
-                current_text = aggregator.text
-
-            # The last aggregator decides whether we are sending text to the
-            # TTS or not.
-            text = self._text_aggregators[-1].aggregate(current_text)
 
         if text:
             await self._push_tts_frames(text)

src/pipecat/services/cartesia.py

@@ -7,7 +7,7 @@
 import base64
 import json
 import uuid
-from typing import AsyncGenerator, List, Optional, Sequence, Union
+from typing import AsyncGenerator, List, Optional, Union
 
 from loguru import logger
 from pydantic import BaseModel
@@ -91,7 +91,7 @@ class CartesiaTTSService(AudioContextWordTTSService):
         encoding: str = "pcm_s16le",
         container: str = "raw",
         params: InputParams = InputParams(),
-        text_aggregators: Sequence[BaseTextAggregator] = [],
+        text_aggregator: Optional[BaseTextAggregator] = None,
         **kwargs,
     ):
         # Aggregating sentences still gives cleaner-sounding results and fewer
@@ -109,7 +109,7 @@ class CartesiaTTSService(AudioContextWordTTSService):
             push_text_frames=False,
             pause_frame_processing=True,
             sample_rate=sample_rate,
-            text_aggregators=text_aggregators or [SkipTagsAggregator([("<spell>", "</spell>")])],
+            text_aggregator=text_aggregator or SkipTagsAggregator([("<spell>", "</spell>")]),
             **kwargs,
         )
 

src/pipecat/services/rime.py

@@ -7,7 +7,7 @@
 import base64
 import json
 import uuid
-from typing import AsyncGenerator, Optional, Sequence
+from typing import AsyncGenerator, Optional
 
 import aiohttp
 from loguru import logger
@@ -80,7 +80,7 @@ class RimeTTSService(AudioContextWordTTSService):
         model: str = "mistv2",
         sample_rate: Optional[int] = None,
         params: InputParams = InputParams(),
-        text_aggregators: Sequence[BaseTextAggregator] = [],
+        text_aggregator: Optional[BaseTextAggregator] = None,
         **kwargs,
     ):
         """Initialize Rime TTS service.
@@ -100,7 +100,7 @@ class RimeTTSService(AudioContextWordTTSService):
             push_stop_frames=True,
             pause_frame_processing=True,
             sample_rate=sample_rate,
-            text_aggregators=text_aggregators or [SkipTagsAggregator([("spell(", ")")])],
+            text_aggregator=text_aggregator or SkipTagsAggregator([("spell(", ")")]),
             **kwargs,
         )