livekit_agent_deepgram_stt_audio_transcription.py

1import asyncio
2
3from livekit.agents import JobContext, WorkerOptions, cli
4from livekit.plugins import deepgram
5
6
7async def entrypoint(ctx: JobContext):
8    # Create a Deepgram STT instance
9    # It will use the DEEPGRAM_API_KEY environment variable by default
10    stt = deepgram.STT()
11
12    print(f"Connecting to room {ctx.room.name}")
13    await ctx.connect()
14
15    # This example simply initializes the STT plugin. 
16    # In a real agent, you would typically use this with a VoiceAssistant 
17    # or subscribe to audio tracks to transcribe them.
18    
19    @ctx.room.on("track_subscribed")
20    def on_track_subscribed(track, publication, participant):
21        if track.kind == "audio":
22            print(f"Transcribing track {track.sid} from participant {participant.identity}")
23            
24            async def transcribe():
25                audio_stream = livekit.rtc.AudioStream(track)
26                stt_stream = stt.stream()
27                
28                # Push audio to STT and handle results
29                async for audio_frame in audio_stream:
30                    stt_stream.push_frame(audio_frame)
31                    
32                async for event in stt_stream:
33                    if event.type == "transcript":
34                        print(f"Transcript: {event.transcript}")
35
36            asyncio.create_task(transcribe())
37
38if __name__ == "__main__":
39    cli.run_app(WorkerOptions(entrypoint_fnc=entrypoint))