Merge pull request #109 from daily-co/only-use-fps

transport: only use fps to set maxFramerate

.github/workflows/publish_test.yaml

@@ -15,6 +15,8 @@ jobs:
         uses: actions/checkout@v4
         with:
           ref: ${{ github.event.inputs.gitref }}
+          fetch-tags: true
+          fetch-depth: 100
       - name: Set up Python
         id: setup_python
         uses: actions/setup-python@v4
@@ -35,6 +37,7 @@ jobs:
       - name: Upload wheels
         uses: actions/upload-artifact@v4
         with:
+          name: wheels
           path: ./dist
 
   publish-to-pypi:
@@ -42,7 +45,7 @@ jobs:
     runs-on: ubuntu-latest
     needs: [ build ]
     environment:
-      name: pypi
+      name: testpypi
       url: https://pypi.org/p/dailyai
     permissions:
       id-token: write
@@ -50,6 +53,7 @@ jobs:
       - name: Download wheels
         uses: actions/download-artifact@v4
         with:
+          name: wheels
           path: ./dist
       - name: Publish to PyPI
         uses: pypa/gh-action-pypi-publish@release/v1

.gitignore

@@ -3,6 +3,7 @@ env/
 __pycache__/
 *~
 venv
+.venv
 #*#
 
 # Distribution / packaging

README.md

@@ -58,7 +58,7 @@ By default, in order to minimize dependencies, only the basic framework function
 dependencies that you can install with:
 
 ```
-pip install dailyai[option,...]
+pip install "dailyai[option,...]"
 ```
 
 Your project may or may not need these, so they're made available as optional requirements. Here is a list:

dot-env.template

@@ -2,8 +2,16 @@
 ANTHROPIC_API_KEY=...
 
 # Azure
-SPEECH_KEY=...
-SPEECH_REGION=...
+AZURE_SPEECH_REGION=...
+AZURE_SPEECH_API_KEY=...
+
+AZURE_CHATGPT_API_KEY=...
+AZURE_CHATGPT_ENDPOINT=https://...
+AZURE_CHATGPT_MODEL=...
+
+AZURE_DALLE_API_KEY=...
+AZURE_DALLE_ENDPOINT=https://...
+AZURE_DALLE_MODEL=...
 
 # Daily
 DAILY_API_KEY=...

examples/foundational/05a-local-sync-speech-and-text.py

@@ -3,8 +3,9 @@ import asyncio
 import logging
 import tkinter as tk
 import os
+from dailyai.pipeline.aggregators import LLMFullResponseAggregator
 
-from dailyai.pipeline.frames import AudioFrame, ImageFrame
+from dailyai.pipeline.frames import AudioFrame, ImageFrame, LLMMessagesFrame, TextFrame
 from dailyai.services.open_ai_services import OpenAILLMService
 from dailyai.services.elevenlabs_ai_service import ElevenLabsTTSService
 from dailyai.services.fal_ai_services import FalImageGenService
@@ -22,7 +23,7 @@ async def main():
     async with aiohttp.ClientSession() as session:
         meeting_duration_minutes = 5
         tk_root = tk.Tk()
-        tk_root.title("Calendar")
+        tk_root.title("dailyai")
 
         transport = LocalTransport(
             mic_enabled=True,
@@ -43,7 +44,7 @@ async def main():
             api_key=os.getenv("OPENAI_API_KEY"),
             model="gpt-4-turbo-preview")
 
-        dalle = FalImageGenService(
+        imagegen = FalImageGenService(
             image_size="1024x1024",
             aiohttp_session=session,
             key_id=os.getenv("FAL_KEY_ID"),
@@ -60,18 +61,33 @@ async def main():
 
             return all_audio
 
+        async def get_month_description(aggregator, frame):
+            async for frame in aggregator.process_frame(frame):
+                if isinstance(frame, TextFrame):
+                    return frame.text
+
         async def get_month_data(month):
             messages = [{"role": "system", "content": f"Describe a nature photograph suitable for use in a calendar, for the month of {
                 month}. Include only the image description with no preamble. Limit the description to one sentence, please.", }]
 
-            image_description = await llm.run_llm(messages)
+            messages_frame = LLMMessagesFrame(messages)
+
+            llm_full_response_aggregator = LLMFullResponseAggregator()
+
+            image_description = None
+            async for frame in llm.process_frame(messages_frame):
+                result = await get_month_description(llm_full_response_aggregator, frame)
+                if result:
+                    image_description = result
+                    break
+
             if not image_description:
                 return
 
             to_speak = f"{month}: {image_description}"
             audio_task = asyncio.create_task(get_all_audio(to_speak))
             image_task = asyncio.create_task(
-                dalle.run_image_gen(image_description))
+                imagegen.run_image_gen(image_description))
             (audio, image_data) = await asyncio.gather(audio_task, image_task)
 
             return {
@@ -82,19 +98,14 @@ async def main():
                 "audio": audio,
             }
 
+        # We only specify 5 months as we create tasks all at once and we might
+        # get rate limited otherwise.
         months: list[str] = [
             "January",
             "February",
             "March",
             "April",
             "May",
-            "June",
-            "July",
-            "August",
-            "September",
-            "October",
-            "November",
-            "December",
         ]
 
         async def show_images():

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

@@ -5,7 +5,8 @@ from typing import AsyncGenerator
 import aiohttp
 from PIL import Image
 
-from dailyai.pipeline.frames import ImageFrame, Frame
+from dailyai.pipeline.frames import ImageFrame, Frame, TextFrame
+from dailyai.pipeline.pipeline import Pipeline
 from dailyai.transports.daily_transport import DailyTransport
 from dailyai.services.ai_services import AIService
 from dailyai.pipeline.aggregators import (
@@ -14,7 +15,6 @@ from dailyai.pipeline.aggregators import (
 )
 from dailyai.services.open_ai_services import OpenAILLMService
 from dailyai.services.elevenlabs_ai_service import ElevenLabsTTSService
-from dailyai.services.fal_ai_services import FalImageGenService
 
 from runner import configure
 
@@ -53,6 +53,7 @@ async def main(room_url: str, token):
         transport._camera_height = 1024
         transport._mic_enabled = True
         transport._mic_sample_rate = 16000
+        transport.transcription_settings["extra"]["punctuate"] = True
 
         tts = ElevenLabsTTSService(
             aiohttp_session=session,
@@ -64,57 +65,30 @@ async def main(room_url: str, token):
             api_key=os.getenv("OPENAI_API_KEY"),
             model="gpt-4-turbo-preview")
 
-        img = FalImageGenService(
-            image_size="1024x1024",
-            aiohttp_session=session,
-            key_id=os.getenv("FAL_KEY_ID"),
-            key_secret=os.getenv("FAL_KEY_SECRET"),
+        messages = [
+            {
+                "role": "system",
+                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be converted to audio so it should not include any special characters. Respond to what the user said in a creative and helpful way.",
+            },
+        ]
+
+        tma_in = LLMUserContextAggregator(
+            messages, transport._my_participant_id)
+        tma_out = LLMAssistantContextAggregator(
+            messages, transport._my_participant_id
+        )
+        image_sync_aggregator = ImageSyncAggregator(
+            os.path.join(os.path.dirname(__file__), "assets", "speaking.png"),
+            os.path.join(os.path.dirname(__file__), "assets", "waiting.png"),
         )
 
-        async def get_images():
-            get_speaking_task = asyncio.create_task(
-                img.run_image_gen("An image of a cat speaking")
-            )
-            get_waiting_task = asyncio.create_task(
-                img.run_image_gen("An image of a cat waiting")
-            )
-
-            (speaking_data, waiting_data) = await asyncio.gather(
-                get_speaking_task, get_waiting_task
-            )
-
-            return speaking_data, waiting_data
+        pipeline = Pipeline([image_sync_aggregator, tma_in, llm, tma_out, tts])
 
         @transport.event_handler("on_first_other_participant_joined")
         async def on_first_other_participant_joined(transport):
-            await tts.say("Hi, I'm listening!", transport.send_queue)
+            await pipeline.queue_frames([TextFrame("Hi, I'm listening!")])
 
-        async def handle_transcriptions():
-            messages = [
-                {
-                    "role": "system",
-                    "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be converted to audio. Respond to what the user said in a creative and helpful way.",
-                },
-            ]
-
-            tma_in = LLMUserContextAggregator(
-                messages, transport._my_participant_id)
-            tma_out = LLMAssistantContextAggregator(
-                messages, transport._my_participant_id
-            )
-            image_sync_aggregator = ImageSyncAggregator(
-                os.path.join(
-                    os.path.dirname(__file__), "assets", "speaking.png"), os.path.join(
-                    os.path.dirname(__file__), "assets", "waiting.png"), )
-            await tts.run_to_queue(
-                transport.send_queue,
-                image_sync_aggregator.run(
-                    tma_out.run(llm.run(tma_in.run(transport.get_receive_frames())))
-                ),
-            )
-
-        transport.transcription_settings["extra"]["punctuate"] = True
-        await asyncio.gather(transport.run(), handle_transcriptions())
+        await transport.run(pipeline)
 
 
 if __name__ == "__main__":

examples/foundational/10-wake-word.py

@@ -5,6 +5,7 @@ import os
 import random
 from typing import AsyncGenerator
 from PIL import Image
+from dailyai.pipeline.pipeline import Pipeline
 
 from dailyai.transports.daily_transport import DailyTransport
 from dailyai.services.open_ai_services import OpenAILLMService
@@ -133,6 +134,7 @@ async def main(room_url: str, token):
         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"),
@@ -145,45 +147,34 @@ async def main(room_url: str, token):
         )
         isa = ImageSyncAggregator()
 
+        messages = [
+            {
+                "role": "system",
+                "content": "You are Santa Cat, a cat that lives in Santa's workshop at the North Pole. You should be clever, and a bit sarcastic. You should also tell jokes every once in a while.  Your responses should only be a few sentences long.",
+            },
+        ]
+
+        tma_in = LLMUserContextAggregator(
+            messages, transport._my_participant_id)
+        tma_out = LLMAssistantContextAggregator(
+            messages, transport._my_participant_id
+        )
+        tf = TranscriptFilter(transport._my_participant_id)
+        ncf = NameCheckFilter(["Santa Cat", "Santa"])
+
+        pipeline = Pipeline([isa, tf, ncf, tma_in, llm, tma_out, tts])
+
         @transport.event_handler("on_first_other_participant_joined")
         async def on_first_other_participant_joined(transport):
-            await tts.say(
+            await transport.say(
                 "Hi! If you want to talk to me, just say 'hey Santa Cat'.",
-                transport.send_queue,
-            )
-
-        async def handle_transcriptions():
-            messages = [
-                {
-                    "role": "system",
-                    "content": "You are Santa Cat, a cat that lives in Santa's workshop at the North Pole. You should be clever, and a bit sarcastic. You should also tell jokes every once in a while.  Your responses should only be a few sentences long.",
-                },
-            ]
-
-            tma_in = LLMUserContextAggregator(
-                messages, transport._my_participant_id)
-            tma_out = LLMAssistantContextAggregator(
-                messages, transport._my_participant_id
-            )
-            tf = TranscriptFilter(transport._my_participant_id)
-            ncf = NameCheckFilter(["Santa Cat", "Santa"])
-            await tts.run_to_queue(
-                transport.send_queue,
-                isa.run(
-                    tma_out.run(
-                        llm.run(
-                            tma_in.run(
-                                ncf.run(tf.run(transport.get_receive_frames())))
-                        )
-                    )
-                ),
+                tts,
             )
 
         async def starting_image():
             await transport.send_queue.put(quiet_frame)
 
-        transport.transcription_settings["extra"]["punctuate"] = True
-        await asyncio.gather(transport.run(), handle_transcriptions(), starting_image())
+        await asyncio.gather(transport.run(pipeline), starting_image())
 
 
 if __name__ == "__main__":

examples/foundational/11-sound-effects.py

@@ -3,6 +3,7 @@ import asyncio
 import logging
 import os
 import wave
+from dailyai.pipeline.pipeline import Pipeline
 
 from dailyai.transports.daily_transport import DailyTransport
 from dailyai.services.open_ai_services import OpenAILLMService
@@ -81,6 +82,7 @@ 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"),
@@ -92,47 +94,31 @@ async def main(room_url: str, token):
             voice_id="ErXwobaYiN019PkySvjV",
         )
 
+        messages = [
+            {
+                "role": "system",
+                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be converted to audio. Respond to what the user said in a creative and helpful way.",
+            },
+        ]
+
+        tma_in = LLMUserContextAggregator(
+            messages, transport._my_participant_id)
+        tma_out = LLMAssistantContextAggregator(
+            messages, transport._my_participant_id
+        )
+        out_sound = OutboundSoundEffectWrapper()
+        in_sound = InboundSoundEffectWrapper()
+        fl = FrameLogger("LLM Out")
+        fl2 = FrameLogger("Transcription In")
+
+        pipeline = Pipeline([tma_in, in_sound, fl2, llm, tma_out, fl, tts, out_sound])
+
         @transport.event_handler("on_first_other_participant_joined")
         async def on_first_other_participant_joined(transport):
-            await tts.say("Hi, I'm listening!", transport.send_queue)
+            await transport.say("Hi, I'm listening!", tts)
             await transport.send_queue.put(AudioFrame(sounds["ding1.wav"]))
 
-        async def handle_transcriptions():
-            messages = [
-                {
-                    "role": "system",
-                    "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be converted to audio. Respond to what the user said in a creative and helpful way.",
-                },
-            ]
-
-            tma_in = LLMUserContextAggregator(
-                messages, transport._my_participant_id)
-            tma_out = LLMAssistantContextAggregator(
-                messages, transport._my_participant_id
-            )
-            out_sound = OutboundSoundEffectWrapper()
-            in_sound = InboundSoundEffectWrapper()
-            fl = FrameLogger("LLM Out")
-            fl2 = FrameLogger("Transcription In")
-            await out_sound.run_to_queue(
-                transport.send_queue,
-                tts.run(
-                    fl.run(
-                        tma_out.run(
-                            llm.run(
-                                fl2.run(
-                                    in_sound.run(
-                                        tma_in.run(transport.get_receive_frames())
-                                    )
-                                )
-                            )
-                        )
-                    )
-                ),
-            )
-
-        transport.transcription_settings["extra"]["punctuate"] = True
-        await asyncio.gather(transport.run(), handle_transcriptions())
+        await asyncio.gather(transport.run(pipeline))
 
 
 if __name__ == "__main__":

examples/foundational/13-whisper-transcription.py

@@ -1,12 +1,16 @@
 import asyncio
 import logging
 
+from dailyai.pipeline.frames import EndFrame, TranscriptionFrame
 from dailyai.transports.daily_transport import DailyTransport
 from dailyai.services.whisper_ai_services import WhisperSTTService
 from dailyai.pipeline.pipeline import Pipeline
 
 from runner import configure
 
+from dotenv import load_dotenv
+load_dotenv(override=True)
+
 logging.basicConfig(format=f"%(levelno)s %(asctime)s %(message)s")
 logger = logging.getLogger("dailyai")
 logger.setLevel(logging.DEBUG)
@@ -26,17 +30,27 @@ async def main(room_url: str):
     stt = WhisperSTTService()
 
     transcription_output_queue = asyncio.Queue()
+    transport_done = asyncio.Event()
 
-    pipeline = Pipeline([stt])
-    pipeline.set_sink(transcription_output_queue)
+    pipeline = Pipeline([stt], source=transport.receive_queue, sink=transcription_output_queue)
 
     async def handle_transcription():
         print("`````````TRANSCRIPTION`````````")
-        while True:
+        while not transport_done.is_set():
             item = await transcription_output_queue.get()
-            print(item.text)
+            print("got item from queue", item)
+            if isinstance(item, TranscriptionFrame):
+                print(item.text)
+            elif isinstance(item, EndFrame):
+                break
+        print("handle_transcription done")
 
-    await asyncio.gather(transport.run(pipeline), handle_transcription())
+    async def run_until_done():
+        await transport.run()
+        transport_done.set()
+        print("run_until_done done")
+
+    await asyncio.gather(run_until_done(), pipeline.run_pipeline(), handle_transcription())
 
 
 if __name__ == "__main__":

examples/foundational/13a-whisper-local.py

@@ -15,11 +15,10 @@ async def main():
     meeting_duration_minutes = 1
 
     transport = LocalTransport(
-        mic_enabled=False,
+        mic_enabled=True,
         camera_enabled=False,
         speaker_enabled=True,
         duration_minutes=meeting_duration_minutes,
-        start_transcription=False,
     )
 
     stt = WhisperSTTService()
@@ -27,8 +26,7 @@ async def main():
     transcription_output_queue = asyncio.Queue()
     transport_done = asyncio.Event()
 
-    pipeline = Pipeline([stt])
-    pipeline.set_sink(transcription_output_queue)
+    pipeline = Pipeline([stt], source=transport.receive_queue, sink=transcription_output_queue)
 
     async def handle_transcription():
         print("`````````TRANSCRIPTION`````````")
@@ -42,11 +40,11 @@ async def main():
         print("handle_transcription done")
 
     async def run_until_done():
-        await transport.run(pipeline)
+        await transport.run()
         transport_done.set()
         print("run_until_done done")
 
-    await asyncio.gather(run_until_done(), handle_transcription())
+    await asyncio.gather(run_until_done(), pipeline.run_pipeline(), handle_transcription())
 
 
 if __name__ == "__main__":

pyproject.toml

@@ -51,4 +51,4 @@ where = ["src"]
 pythonpath = ["src"]
 
 [tool.setuptools_scm]
-# Empty
+local_scheme = "no-local-version"

src/dailyai/services/azure_ai_services.py

@@ -19,7 +19,7 @@ try:
 except ModuleNotFoundError as e:
     print(f"Exception: {e}")
     print(
-        "In order to use Azure TTS, you need to `pip install dailyai[azure]`. Also, set `SPEECH_KEY` and `SPEECH_REGION` environment variables.")
+        "In order to use Azure TTS, you need to `pip install dailyai[azure]`. Also, set `AZURE_SPEECH_API_KEY` and `AZURE_SPEECH_REGION` environment variables.")
     raise Exception(f"Missing module: {e}")
 
 from dailyai.services.openai_api_llm_service import BaseOpenAILLMService

src/dailyai/transports/abstract_transport.py

@@ -21,9 +21,10 @@ class AbstractTransport:
         self._camera_enabled = kwargs.get("camera_enabled") or False
         self._camera_width = kwargs.get("camera_width") or 1024
         self._camera_height = kwargs.get("camera_height") or 768
+        self._camera_bitrate = kwargs.get("camera_bitrate") or 250000
+        self._camera_framerate = kwargs.get("camera_framerate") or 10
         self._speaker_enabled = kwargs.get("speaker_enabled") or False
         self._speaker_sample_rate = kwargs.get("speaker_sample_rate") or 16000
-        self._fps = kwargs.get("fps") or 8
 
         self._logger: logging.Logger = logging.getLogger("dailyai.transport")
 

src/dailyai/transports/daily_transport.py

@@ -222,9 +222,9 @@ class DailyTransport(ThreadedTransport, EventHandler):
                             "maxQuality": "low",
                             "encodings": {
                                 "low": {
-                                    "maxBitrate": 250000,
+                                    "maxBitrate": self._camera_bitrate,
                                     "scaleResolutionDownBy": 1.333,
-                                    "maxFramerate": 8,
+                                    "maxFramerate": self._camera_framerate,
                                 }
                             },
                         }

src/dailyai/transports/local_transport.py

@@ -19,6 +19,7 @@ class LocalTransport(ThreadedTransport):
         self._sample_width = kwargs.get("sample_width") or 2
         self._n_channels = kwargs.get("n_channels") or 1
         self._tk_root = kwargs.get("tk_root") or None
+        self._pyaudio = None
 
         if self._camera_enabled and not self._tk_root:
             raise ValueError(
@@ -51,7 +52,7 @@ class LocalTransport(ThreadedTransport):
         if self._mic_enabled:
             self._audio_stream.write(frame)
 
-    def read_frames(self, desired_frame_count):
+    def read_audio_frames(self, desired_frame_count):
         bytes = b""
         if self._speaker_enabled:
             bytes = self._speaker_stream.read(
@@ -62,7 +63,8 @@ class LocalTransport(ThreadedTransport):
 
     def _prerun(self):
         if self._mic_enabled:
-            self._pyaudio = pyaudio.PyAudio()
+            if not self._pyaudio:
+                self._pyaudio = pyaudio.PyAudio()
             self._audio_stream = self._pyaudio.open(
                 format=self._pyaudio.get_format_from_width(self._sample_width),
                 channels=self._n_channels,
@@ -84,6 +86,8 @@ class LocalTransport(ThreadedTransport):
             self._image_label.pack()
 
         if self._speaker_enabled:
+            if not self._pyaudio:
+                self._pyaudio = pyaudio.PyAudio()
             self._speaker_stream = self._pyaudio.open(
                 format=self._pyaudio.get_format_from_width(self._sample_width),
                 channels=self._n_channels,

src/dailyai/transports/threaded_transport.py

@@ -393,7 +393,7 @@ class ThreadedTransport(AbstractTransport):
                     this_frame = next(self._images)
                     self.write_frame_to_camera(this_frame)
 
-                time.sleep(1.0 / self._fps)
+                time.sleep(1.0 / self._camera_framerate)
         except Exception as e:
             self._logger.error(f"Exception {e} in camera thread.")
             raise e