refactor party tonight

src/samples/theoretical-to-real/04-utterance-and-speech.py

@@ -2,9 +2,11 @@ import argparse
 import asyncio
 import re
 
+from dailyai.services.ai_services import SentenceAggregator
 from dailyai.services.daily_transport_service import DailyTransportService
 from dailyai.services.azure_ai_services import AzureLLMService, AzureTTSService
 from dailyai.queue_frame import QueueFrame, FrameType
+from dailyai.services.elevenlabs_ai_service import ElevenLabsTTSService
 
 async def main(room_url:str):
     global transport
@@ -22,34 +24,46 @@ async def main(room_url:str):
     transport.camera_enabled = False
 
     llm = AzureLLMService()
-    tts = AzureTTSService()
+    azure_tts = AzureTTSService()
+    elevenlabs_tts = ElevenLabsTTSService(voice_id="ErXwobaYiN019PkySvjV")
+
+    messages = [{"role": "system", "content": "tell the user a joke about llamas"}]
+
+    # Start a task to run the LLM to create a joke, and convert the LLM output to audio frames. This task
+    # will run in parallel with generating and speaking the audio for static text, so there's no delay to
+    # speak the LLM response.
+    buffer_queue = asyncio.Queue()
+    llm_response_task = asyncio.create_task(
+        elevenlabs_tts.run_to_queue(
+            buffer_queue,
+            SentenceAggregator().run(
+                llm.run([QueueFrame(FrameType.LLM_MESSAGE, messages)])
+            ),
+            True,
+        )
+    )
 
     @transport.event_handler("on_participant_joined")
     async def on_joined(transport, participant):
         if participant["id"] == transport.my_participant_id:
             return
 
-        # queue two pieces of speech: one specified as a text literal,
-        # and one generated by an llm. We'll kick off the llm first, and let
-        # it generate a response while we're speaking the literal string.
-        #
-        # Note that in this case, we don't use `run_llm_async` because we're
-        # taking advantage of the time spent speaking the first phrase to generate
-        # the entire LLM response, and this happens asynchronously in a task.
-        llm_response_task = asyncio.create_task(llm.run_llm(
-            [{"role": "system", "content": "tell the user a joke about llamas"}]
-        ))
+        await azure_tts.run_to_queue(
+            transport.send_queue,
+            [QueueFrame(FrameType.SENTENCE, "My friend the LLM is now going to tell a joke about llamas.")]
+        )
 
-        async for audio_chunk in tts.run_tts("My friend the LLM is now going to tell a joke about llamas."):
-            transport.output_queue.put(QueueFrame(FrameType.AUDIO, audio_chunk))
+        async def buffer_to_send_queue():
+            while True:
+                frame = await buffer_queue.get()
+                await transport.send_queue.put(frame)
+                buffer_queue.task_done()
+                if frame.frame_type == FrameType.END_STREAM:
+                    break
 
-        llm_response = await llm_response_task
-        async for audio_chunk in tts.run_tts(llm_response):
-            transport.output_queue.put(QueueFrame(FrameType.AUDIO, audio_chunk))
+        await asyncio.gather(llm_response_task, buffer_to_send_queue())
 
-
-        # wait for the output queue to be empty, then leave the meeting
-        transport.output_queue.join()
+        transport.wait_for_send_queue_to_empty()
         transport.stop()
 
     await transport.run()
@@ -61,6 +75,6 @@ if __name__ == "__main__":
         "-u", "--url", type=str, required=True, help="URL of the Daily room to join"
     )
 
-    args: argparse.Namespace = parser.parse_args()
+    args, unknown = parser.parse_known_args()
 
     asyncio.run(main(args.url))