Merge pull request #117 from daily-co/llm-use-aggregator-pass-through…

…-fix

aggregators: fix LLMUserResponseAggregator passs-through

examples/foundational/06-listen-and-respond.py

@@ -10,8 +10,8 @@
 from dailyai.services.open_ai_services import OpenAILLMService
 from dailyai.services.ai_services import FrameLogger
 from dailyai.pipeline.aggregators import (
-    LLMAssistantContextAggregator,
-    LLMUserContextAggregator,
+    LLMAssistantResponseAggregator,
+    LLMUserResponseAggregator,
 )
 from runner import configure
 
@@ -55,11 +55,9 @@ async def main(room_url: str, token):
             },
         ]
 
-        tma_in = LLMUserContextAggregator(
-            messages, transport._my_participant_id)
-        tma_out = LLMAssistantContextAggregator(
-            messages, transport._my_participant_id
-        )
+        tma_in = LLMUserResponseAggregator(messages)
+        tma_out = LLMAssistantResponseAggregator(messages)
+
         pipeline = Pipeline(
             processors=[
                 fl,
@@ -78,8 +76,6 @@ async def on_first_other_participant_joined(transport, participant):
                 {"role": "system", "content": "Please introduce yourself to the user."})
             await pipeline.queue_frames([LLMMessagesFrame(messages)])
 
-        transport.transcription_settings["extra"]["endpointing"] = True
-        transport.transcription_settings["extra"]["punctuate"] = True
         await transport.run(pipeline)
 
 

examples/foundational/06a-image-sync.py

@@ -47,13 +47,12 @@ async def main(room_url: str, token):
             token,
             "Respond bot",
             5,
+            camera_enabled=True,
+            camera_width=1024,
+            camera_height=1024,
+            mic_enabled=True,
+            mic_sample_rate=16000,
         )
-        transport._camera_enabled = True
-        transport._camera_width = 1024
-        transport._camera_height = 1024
-        transport._mic_enabled = True
-        transport._mic_sample_rate = 16000
-        transport.transcription_settings["extra"]["punctuate"] = True
 
         tts = ElevenLabsTTSService(
             aiohttp_session=session,

examples/foundational/07-interruptible.py

@@ -67,7 +67,6 @@ async def run_conversation():
                 pre_processor=LLMUserResponseAggregator(messages),
             )
 
-        transport.transcription_settings["extra"]["punctuate"] = False
         await asyncio.gather(transport.run(), run_conversation())
 
 

examples/foundational/10-wake-word.py

@@ -129,12 +129,6 @@ async def main(room_url: str, token):
             camera_width=720,
             camera_height=1280,
         )
-        transport._mic_enabled = True
-        transport._mic_sample_rate = 16000
-        transport._camera_enabled = True
-        transport._camera_width = 720
-        transport._camera_height = 1280
-        transport.transcription_settings["extra"]["punctuate"] = True
 
         llm = OpenAILLMService(
             api_key=os.getenv("OPENAI_API_KEY"),

examples/foundational/11-sound-effects.py

@@ -82,7 +82,6 @@ async def main(room_url: str, token):
             mic_sample_rate=16000,
             camera_enabled=False,
         )
-        transport.transcription_settings["extra"]["punctuate"] = True
 
         llm = OpenAILLMService(
             api_key=os.getenv("OPENAI_API_KEY"),

examples/image-gen.py

@@ -77,8 +77,6 @@ async def on_participant_joined(transport, participant):
             async for audio in audio_generator:
                 transport.output_queue.put(Frame(FrameType.AUDIO_FRAME, audio))
 
-    transport.transcription_settings["extra"]["punctuate"] = False
-    transport.transcription_settings["extra"]["endpointing"] = False
     await asyncio.gather(transport.run(), handle_transcriptions())
 
 

examples/internal/11a-dial-out.py

@@ -127,8 +127,6 @@ async def on_call_state_updated(transport, state):
                     transport.start_recording()
                     transport.dialout(phone)
 
-        transport.transcription_settings["extra"]["punctuate"] = True
-
         await asyncio.gather(transport.run(), handle_transcriptions())
 
 

examples/starter-apps/chatbot.py

@@ -139,8 +139,6 @@ async def run_conversation():
                 pre_processor=LLMUserResponseAggregator(messages),
             )
 
-        transport.transcription_settings["extra"]["endpointing"] = True
-        transport.transcription_settings["extra"]["punctuate"] = True
         await asyncio.gather(transport.run(), run_conversation())
 
 

examples/starter-apps/patient-intake.py

@@ -340,8 +340,6 @@ async def handle_intake():
                 pre_processor=OpenAIUserContextAggregator(context),
             )
 
-        transport.transcription_settings["extra"]["endpointing"] = True
-        transport.transcription_settings["extra"]["punctuate"] = True
         try:
             await asyncio.gather(transport.run(), handle_intake())
         except (asyncio.CancelledError, KeyboardInterrupt):

examples/starter-apps/storybot.py

@@ -278,8 +278,6 @@ async def storytime():
                 pipeline,
             )
 
-        transport.transcription_settings["extra"]["endpointing"] = True
-        transport.transcription_settings["extra"]["punctuate"] = True
         try:
             await asyncio.gather(transport.run(), storytime())
         except (asyncio.CancelledError, KeyboardInterrupt):

examples/starter-apps/translator.py

@@ -99,8 +99,6 @@ async def main(room_url: str, token):
         ts = TranslationSubtitles("spanish")
         pipeline = Pipeline([sa, tp, llm, lfra, ts, tts])
 
-        transport.transcription_settings["extra"]["endpointing"] = True
-        transport.transcription_settings["extra"]["punctuate"] = True
         await transport.run(pipeline)
 
 

src/dailyai/pipeline/aggregators.py

@@ -1,5 +1,6 @@
 import asyncio
 import re
+import time
 
 from dailyai.pipeline.frame_processor import FrameProcessor
 
@@ -8,6 +9,7 @@
     EndPipeFrame,
     Frame,
     ImageFrame,
+    InterimTranscriptionFrame,
     LLMMessagesFrame,
     LLMResponseEndFrame,
     LLMResponseStartFrame,
@@ -106,6 +108,7 @@ def __init__(
         start_frame,
         end_frame,
         accumulator_frame,
+        interim_accumulator_frame=None,
         pass_through=True,
     ):
         self.aggregation = ""
@@ -115,31 +118,75 @@ def __init__(
         self._start_frame = start_frame
         self._end_frame = end_frame
         self._accumulator_frame = accumulator_frame
+        self._interim_accumulator_frame = interim_accumulator_frame
         self._pass_through = pass_through
-
+        self._seen_start_frame = False
+        self._seen_end_frame = False
+        self._seen_interim_results = False
+
+    # Use cases implemented:
+    #
+    # S: Start, E: End, T: Transcription, I: Interim, X: Text
+    #
+    #        S E -> None
+    #      S T E -> X
+    #    S I T E -> X
+    #    S I E T -> X
+    #  S I E I T -> X
+    #
+    # The following case would not be supported:
+    #
+    #    S I E T1 I T2 -> X
+    #
+    # and T2 would be dropped.
     async def process_frame(self, frame: Frame) -> AsyncGenerator[Frame, None]:
         if not self.messages:
             return
 
+        send_aggregation = False
+
         if isinstance(frame, self._start_frame):
+            self._seen_start_frame = True
             self.aggregating = True
         elif isinstance(frame, self._end_frame):
-            self.aggregating = False
-            # Sometimes VAD triggers quickly on and off. If we don't get any transcription,
-            # it creates empty LLM message queue frames
-            if len(self.aggregation) > 0:
-                self.messages.append(
-                    {"role": self._role, "content": self.aggregation})
-                self.aggregation = ""
-                yield self._end_frame()
-                yield LLMMessagesFrame(self.messages)
-        elif isinstance(frame, self._accumulator_frame) and self.aggregating:
-            self.aggregation += f" {frame.text}"
+            self._seen_end_frame = True
+
+            # We might have received the end frame but we might still be
+            # aggregating (i.e. we have seen interim results but not the final
+            # text).
+            self.aggregating = self._seen_interim_results
+
+            # Send the aggregation if we are not aggregating anymore (i.e. no
+            # more interim results received).
+            send_aggregation = not self.aggregating
+        elif isinstance(frame, self._accumulator_frame):
+            if self.aggregating:
+                self.aggregation += f" {frame.text}"
+                # We have receied a complete sentence, so if we have seen the
+                # end frame and we were still aggregating, it means we should
+                # send the aggregation.
+                send_aggregation = self._seen_end_frame
+
             if self._pass_through:
                 yield frame
+
+            # We just got our final result, so let's reset interim results.
+            self._seen_interim_results = False
+        elif self._interim_accumulator_frame and isinstance(frame, self._interim_accumulator_frame):
+            self._seen_interim_results = True
         else:
             yield frame
 
+        if send_aggregation and len(self.aggregation) > 0:
+            self.messages.append({"role": self._role, "content": self.aggregation})
+            yield self._end_frame()
+            yield LLMMessagesFrame(self.messages)
+            # Reset
+            self.aggregation = ""
+            self._seen_start_frame = False
+            self._seen_end_frame = False
+            self._seen_interim_results = False
+
 
 class LLMAssistantResponseAggregator(LLMResponseAggregator):
     def __init__(self, messages: list[dict]):
@@ -160,6 +207,7 @@ def __init__(self, messages: list[dict]):
             start_frame=UserStartedSpeakingFrame,
             end_frame=UserStoppedSpeakingFrame,
             accumulator_frame=TranscriptionFrame,
+            interim_accumulator_frame=InterimTranscriptionFrame,
             pass_through=False,
         )
 

src/dailyai/pipeline/frames.py

@@ -164,6 +164,17 @@ def __str__(self):
         return f"{self.__class__.__name__}, text: '{self.text}' participantId: {self.participantId}, timestamp: {self.timestamp}"
 
 
+@dataclass()
+class InterimTranscriptionFrame(TextFrame):
+    """A text frame with interim transcription-specific data. Will be placed in
+    the transport's receive queue when a participant speaks."""
+    participantId: str
+    timestamp: str
+
+    def __str__(self):
+        return f"{self.__class__.__name__}, text: '{self.text}' participantId: {self.participantId}, timestamp: {self.timestamp}"
+
+
 class TTSStartFrame(ControlFrame):
     """Used to indicate the beginning of a TTS response. Following AudioFrames
     are part of the TTS response until an TTEndFrame. These frames can be used

src/dailyai/transports/daily_transport.py

@@ -10,6 +10,7 @@
 from typing import Any
 
 from dailyai.pipeline.frames import (
+    InterimTranscriptionFrame,
     ReceivedAppMessageFrame,
     TranscriptionFrame,
     UserImageFrame,
@@ -88,9 +89,11 @@ def __init__(
             "model": "2-conversationalai",
             "profanity_filter": True,
             "redact": False,
+            "endpointing": True,
+            "punctuate": True,
+            "includeRawResponse": True,
             "extra": {
-                "endpointing": True,
-                "punctuate": False,
+                "interim_results": True,
             },
         }
 
@@ -368,8 +371,12 @@ def on_transcription_message(self, message: dict):
             elif "session_id" in message:
                 participantId = message["session_id"]
             if self._my_participant_id and participantId != self._my_participant_id:
-                frame = TranscriptionFrame(
-                    message["text"], participantId, message["timestamp"])
+                is_final = message["rawResponse"]["is_final"]
+                if is_final:
+                    frame = TranscriptionFrame(message["text"], participantId, message["timestamp"])
+                else:
+                    frame = InterimTranscriptionFrame(
+                        message["text"], participantId, message["timestamp"])
                 asyncio.run_coroutine_threadsafe(
                     self.receive_queue.put(frame), self._loop)