working on theoretical API examples

src/khk-working/theoretical/01-say-one-thing.py

@@ -0,0 +1,47 @@
+
+from dailyai.services.transport.DailyTransport import DailyTransportService
+from dailyai.services.tts.AzureTTSService import AzureTTSService
+
+
+transport = None
+tts = None
+
+
+def main():
+    global transport
+    global tts
+
+    # create a transport service object using environment variables for
+    # the transport service's API key, room url, and any other configuration.
+    # services can all define and document the environment variables they use.
+    # services all also take an optional config object that is used instead of
+    # environment variables.
+    #
+    # the abstract transport service APIs presumably can map pretty closely
+    # to the daily-python basic API
+    transport = DailyTransportService()
+
+    # similarly, create a tts service
+    tts = AzureTTSService()
+
+    # ask the transport to create a local audio "device"/queue for
+    # chunks of audio to play sequentially. the "mic" object is a handle
+    # we can use to inspect and control the queue if we need to. in this
+    # case we will pipe into this queue from the tts service
+    mic = transport.audio_queue()
+    tts.set_output(mic)
+
+    transport.on("error", lambda e: print(e))
+    transport.on("joined-meeting", say_one_thing)
+    transport.start()
+
+
+def say_one_thing():
+    # say one thing, then leave
+    tts.run_tts("hello world")
+    transport.on("audio-queue-empty", shutdown)
+
+
+def shutdown():
+    transport.stop()
+    tts.close()

src/khk-working/theoretical/02-llm-say-one-thing.py

@@ -0,0 +1,47 @@
+
+from dailyai.services.transport.DailyTransport import DailyTransportService
+from dailyai.services.llm.AzureLLMService import AzureLLMService
+from dailyai.services.tts.AzureTTSService import AzureTTSService
+
+transport = None
+mic = None
+llm = None
+tts = None
+
+
+def main():
+    global transport
+    global mic
+    global llm
+    global tts
+
+    transport = DailyTransportService()
+    llm = AzureLLMService()
+    tts = AzureTTSService()
+    mic = transport.audio_queue()
+    tts.set_output(mic)
+
+    # similarly, we can tell the llm to pipe infeference output to our tts
+    # service. the design idea here is that any time we call llm.run_llm()
+    # we are creating a cancelable inference call, and somehow behind the
+    # scenes the full pipeline from the llm to the tts service to the
+    # transport's audio queue is managed in such a way as to be
+    # introspectible and cancelable. also, instead of piping the
+    # output to the tts service directly, we could pipe it through an
+    # adapter object that does chunking or processing or whatever.
+    llm.set_output(tts)
+
+    transport.on("error", lambda e: print(e))
+    transport.on("joined-meeting", make_one_inference_call)
+    transport.start()
+
+
+def make_one_inference_call():
+    # ask our llm to say one thing, then leave
+    llm.run_llm("tell me a joke about llamas")
+    transport.on("audio-queue-empty", shutdown)
+
+
+def shutdown():
+    transport.stop()
+    tts.close()

src/khk-working/theoretical/notes.txt

@@ -0,0 +1,14 @@
+
+-01 just say one thing and exit
+-02 llm say one thing and exit
+-03 send "still frame" video
+-04 manual intro utterance and then llm say one thing and exit
+-05 queue 10 spoken image prompts and synchronize the audio with the generated image frames
+-06 chat: llm speak and respond (ignoring transcription input while speaking)
+-07 chat: llm speak and respond (interruptible)
+-08 two llms arguing about a topic (in the same process)
+-09 two llms arguing about a topic (two separate bots)
+-10 listen for wake word before sending commands to llm
+-11 06 plus sound effects queued from sound file
+-12 06 plus background music played through a second "mic" device 
+